Encyclopedia of Atmospheric Sciences - 1st Edition (Volume 1-6) (2003) PDF

ENCYCLOPEDIA
OF
ATMOSPHERIC
SCIENCES
Editor-in-Chief
JAMES R. HOLTON
Editors
JUDITH A. CURRY
JOHN A. PYLE
ACADEMIC PRESS
An imprint of Elsevier Science
Amsterdam Boston London New York Oxford Paris
San Diego San Francisco Singapore Sydney Tokyo
This book is printed on acid-free paper.
Copyright 2003 by Elsevier Science Ltd. All Rights Reserved
BOUNDARY LAYERS: Overview

UK Crown Copyright 2002
The following articles are US government works in the

public domain and not subject to copyright:
PLANETARY ATMOSPHERES: Mars, RADAR: Cloud Radar, RADIATIVE TRANSFER: Scattering
DYNAMIC METEOROLOGY: Balanced Flows; Potential Vorticity

Copyright Professor Michael McIntyre
The following article has a CSIRO author

and is not subject to copyright: BOUNDARY LAYERS: Complex Terrain
N o part of this publication may be reproduced or transmitted in any form or by any

means, electronic or mechanical, including photocopying, recording, or any
information storage and retrieval system, without permission in writing from the publisher.
Permissioils may be sought directly from Elsevier’s Science & Technology Rights
Department in Oxford, UK: phone: ( + 4 4 ) 1865 843830, fax: ( + 4 4 ) 1865 853333,
e-mail: perrnissions@elsevier.co.uk. You may also complete your request on-line via
the Elsevier Science homepage (http://www.elsevier.com), by selecting
‘Customer Support’ and then ‘Obtaining Permissions’.
Academic Press
An Imprint of Elsevier Science
84 Theobald’s Road, London W C l X 8RR, UK
http://www.academicpress.com
Academic Press
An Imprint of Elsevier Science
525 B Street, Suite 1900, San Diego, California 92101-4495, USA
http://www.academicpress.com
ISBN 0-12-227090-8
Library of Congress Control Number: 20021 14120
A catalogue record for this book is available from the British Library
Printed in Spain
EDITORS v
EDITOR-IN-CHIEF
James R. Holton
University of Washington
Department of Atmospheric Sciences
Seattle, WA 98195, USA
EDITORS
Judith A. Curry
Georgia Institute of Technology
School of Earth and Atmospheric Sciences
Atlanta, GA 30332, USA
John A. Pyle
University of Cambridge
Centre for Atmospheric Science
Department of Chemistry
Cambridge, CB2 1EW, UK
vi EDITORIALADVISORY BOARD
William Blumen' Stanley Q. Kidder

University of Colorado Boulder Colorado State University
Program in Atmospheric and Ocean Sciences Cooperative Institute for Research in the Atmosphere
Campus Box 31 1 Fort Collins, CO 80523, USA
Boulder, CO 80309, USA
Yutaka Kondo
The University of Tokyo
Lance F. Bosart
Research Center for Advanced Science and Technology
University at Albany
4-6-1 Komaba, Meguro
State University of New York
Tokyo 153-8904, Japan
Department of Earth and Atmospheric Sciences
1400 Washington Avenue
Albany, NY 12222, USA Dennis Lamb
Pennsylvania State University
Department of Meteorology
Frank Bradley University Park, PA 16802, USA
CSlRO Land and Water
PO BOX 1666
William K.-M. Lau
Canberra, ACT 2601, Australia
NASA Goddard Space Flight Center
Greenbelt, MD 20771, USA
Roy M. Harrison
University of Birmingham Conway Leovy
Institute of Environmental Health University of Washington
School of Chemistry Department of Atmospheric Sciences
Birmingham B15 2TT, UK PO Box 351640
Dennis Hartmann
University of Washington Peter S. Liss
Department of Atmospheric Sciences University of East Anglia
Box 351640 School of Environmental Sciences
Seattle WA 98195, USA Norwich NR4 7TJ, UK
Stefan L. Hastenrath Mankin Mak

University of Wisconsin Madison University of Illinois
Department of Atmospheric and Ocean Sciences Department of Atmospheric Sciences
1225 W. Drayton Street 105 S. Gregory Street
Madison, WI 53706, USA Urbana, IL 61801, USA
Julian C. R. Hunt, FRS Mitchell W. Moncrieff

University College London National Center for Atmospheric Research
Department of Space and Climate Physics Mesoscale and Microscale Meteorology Division
Gower Street PO Box 3000
London WC1 E 6BT, UK Boulder, CO 80307, USA
Richard H. Johnson Roger A. Pielke, Sr

Colorado State University Colorado State University
Department of Atmospheric Sciences Department of Atmospheric Science
Fort Collins, CO 80523, USA Fort Collins, CO 80523, USA
Michael Kelley William J. Randel

Cornell University National Center for Atmospheric Research
Department of Electrical Engineering Atmospheric Chemistry Division
318 Rhodes Hall PO Box 3000
Ithaca, NY 14850, USA Boulder, CO 80307, USA
EDITORIAL ADVISORY BOARD vii
A. R. Ravishankara John Thuburn

National Oceanic and Atmospheric Administration University of Reading
Aeronomy Laboratory Department of Meteorology
325 S. Broadway Earley Gate
Boulder, CO 80303, USA Reading RG6 6BB, UK
Richard P. Wayne
James M. Russell
Hampton University Oxford University
Physical and Theoretical Chemistry Laboratory
Center for Atmospheric Sciences
Hampton, VA 23668, USA South Parks Road
Oxford OX1 3Q2, UK
Hanwant B. Singh Peter J. Webster

NASA Ames Research Center
University of Colorado Boulder
Earth Science Division Program in Atmospheric and Ocean Sciences
Moffett Field, CA 94035, USA Campus Box 31 1
Akimasa Sumi
University of Tokyo Wilford F. Weeks
Center for Climate System Research University of Alaska Fairbanks
4-6-1 Komaba, Meguro-Ku Department of Geology and Geophysics
Tokyo 153, Japan Fairbanks, AK 99775, USA
Peter A. Taylor Morris Weisman

York University National Center for Atmospheric Research
Department of Earth and Atmospheric Science Mesoscale Dynamics Group
Toronto, Ontario M3J 1P3, Canada PO Box 3000
Alan J. Thorpe
University of Reading Paul 0. Wennberg
Department of Meteorology California Institute of Technology
Earley Gate Division of Geological and Planetary Sciences
Reading RG6 6BB, UK Pasadena, CA 91 125, USA
'deceased
viii FOREWORD
When I started research in atmospheric physics half a century ago, I was able to read almost everything published
in my field of cloud physics as well as keeping track of major developments in other branches of meteorology.
Today this is impossible because, in the meantime, the atmospheric and related sciences have expanded
enormously in scope, scale and complexity while the number of scientific journals has grown to such an extent
that it is difficult for an active research scientist to keep abreast of the literature even in his own subject.
I therefore greatly welcome this unique Encyclopedia of six volumes containing 330 contributions, each of
approximately 4000 words, on all major aspects of atmospheric science and cognate subjects such as
oceanography and hydrology, ranging from Acoustic Waves to the World Climate Research Programme. They
treat the physical, dynamical and chemical processes active in the atmosphere and the interactions of the
atmosphere with the land, the oceans, and the biosphere. These operate on space scales ranging from the atomic
to the global and on all time scales up to those governing major changes of climate. The major advances of the
last half century in our knowledge and understanding of the atmosphere, and in our ability to predict its changes,
have resulted largely from the development of novel techniques to observe and sense the global atmosphere, the
oceans and land surface from space, and from the astonishing growth in computing power. These have made
possible the simulation of weather and climate systems by the development of enormously complex physico-
mathematical models of the global atmosphere, the oceans, the cryosphere and the land surface, their complex
interactions and feedbacks, leading to major improvements in the accuracy and range of weather forecasts and
credible predictions of future changes in global and regional climates caused by man-made emissions of
greenhouse gases and aerosols.
Of only marginal interest 30 years ago, climate change and variability now occupy centre stage and are
represented by 30 articles in the Encyclopedia. Atmospheric chemistry, greatly stimulated by the discovery of the
Antarctic ozone hole and the serious problems of atmospheric pollution, has developed into a whole new subject
represented by no less than 50 articles.
Although opinions may differ on the relative importance attached to some topics, the list of titles and authors
encourages me to believe that they will be both authoritative and up-to-date. The six volumes promise to be the
most comprehensive and widely consulted publication in the atmospheric sciences for years to come. Every
scientist engaged in post-graduate teaching and research in the subject will need access to a copy.
The publishers, the editors, the advisory board and the authors are to be congratulated on providing a most
valuable service to their scientific colleagues worldwide.
h
--
Sir John Mason F.R.S.
Imperial College of Science, Technology and Medicine
London, UK
PREFACE ix
A half century ago the American Meteorological Society published the Compendium of Meteorology, which in a
single volume of 1334 pages summarized the state of understanding of the atmosphere at that time. A perusal of
the contents of that volume indicates that although a broad range of topics was covered, the vast bulk of the
volume was devoted to traditional meteorological topics such as atmospheric dynamics, cloud physics, and
weather forecasting. Barely 4 percent of the volume was devoted to articles related to atmospheric chemistry or
air pollution and, of course, none of the volume was devoted to techniques such as satellites and remote sensing.
As Sir John Mason aptly notes in his foreword to the present work, the atmospheric sciences have expanded in
scope enormously over the past 50 years. Topics such as atmospheric chemistry and global climate change, of
only marginal interest 50 years ago, are now central disciplines within the atmospheric sciences. Increasingly,
developing areas within the atmospheric sciences require students, teachers, and researchers to familiarize
themselves with areas far outside their own specialties. This work is intended to satisfy the need for a convenient
and accessible reference source covering all aspects of the atmospheric sciences. It is written at a level that allows
undergraduate science and engineering students to understand the material, while providing active researchers
with the latest information in the field.
More than 400 scientists, from academia, government, and industry have contributed to the 330 articles in
this work. We are very grateful to these authors for their success in providing concise and authoritative
summaries of complex subjects. As editors, we have benefited from the chance to learn from these articles, and
we believe that all students and active scientists who want to increase their knowledge of the atmosphere will
benefit enormously from access to this work.
We are also grateful to the 31 members of the Editorial Advisory Board who have guided us in our coverage of
the very broad range of topics represented in this encyclopedia. Their willingness to suggest topics and authors,
and to carefully review draft articles has contributed significantly to our success.
The production of this multivolume encyclopedia would not have been possible without the dedicated work of
the staff of the Major Reference Works group at Academic Press. We are especially grateful to the Major
Reference Work Development Manager, Colin McNeil, who has worked closely with us during the entire
process. Finally, we appreciate the liberal use of color figures in the printed encyclopedia.
James R Holton, Judith A Curry, and John A Pyle

x GUIDE TO USE OF THE ENCYCLOPEDIA
Introductory Points
In devising the vision and structure for the Encyclopedia, the Editors have striven to unite and interrelate all
current knowledge that can be designated ‘Atmospheric Sciences’. To aid users of the Encyclopedia, this new
reference work offers intuitive searching and extensive cross-linking of content. These features are explained in
more detail below.
Structure of the Encyclopedia

The material in the Encyclopedia is arranged as a series of entries in alphabetical order. Some entries comprise a
single article, whilst entries on more diverse subjects consist of several articles that deal with various aspects of
the topic. In the latter case the articles are arranged in a logical sequence within an entry.
To help you realize the full potential of the material in the Encyclopedia we have provided three features to
help you find the topic of your choice.
1. Contents Lists
Your first point of reference will probably be the contents list. The complete contents list appearing in each
volume will provide you with both the volume number and the page number of the entry. O n the opening page of
an entry a contents list is provided so that the full details of the articles within the entry are immediately
available.
Alternatively you may choose to browse through a volume using the alphabetical order of the entries as your
guide. To assist you in identifying your location within the Encyclopedia a running headline indicates the current
entry and the current article within that entry.
You will find ‘dummy entries’ where obvious synonyms exist for entries or where we have grouped together
related topics. Dummy entries appear in both the contents list and the body of the text. For example, a dummy
entry appears for Rainbows which directs you to Optics, Atmospheric: Optical Phenomena, where the material
is located.
Example
If you were attempting to locate material on Waves via the contents list.
WAVES See BUOYANCY AND BUOYANCY WAVES: Optical Observations; BUOYANCY AND
BUOYANCY WAVES: Theory; KELVIN WAVES; MIDDLE ATMOSPHERE: Gravity Waves; ROSSBY
WAVES; SOLITARY WAVES; STATIONARY WAVES (OROGRAPHIC AND THERMALLY FORCED)
At the appropriate location in the contents list, the page numbers for these articles are given.
If you were trying to locate the material by browsing through the text and you looked up Waves then tht
following information would be provided.
See BUOYANCY AND BUOYANCY WAVES: Optical Observations; BUOYANCY AND BUOYANCY WAVES:
Theory; KELVIN WAVES; MIDDLE ATMOSPHERE: Gravity Waves; ROSSBY WAVES; SOLITARY
WAVES; STATIONARY WAVES (OROGRAPHIC AND THERMALLY FORCED)
GUIDE TO USE OF THE ENCYCLOPEDIA xi
2. Cross References
All of the articles in the Encyclopedia have been extensively cross referenced. The cross references, which appear
at the end of each article, have been provided at three levels:
To indicate if a topic is discussed in greater detail elsewhere.
LIDAR: Backscatter
ii. To draw the reader’s attention to parallel discussions in other articles.
LIDAR: Backscatter
See also: Aerosols: Climatology of Tropospheric Aerosols; Observations and Measurements: Physics and Chemistry
of Aerosols; Role in Cloud Physics; Role in Radiative Transfer. Cloud Micro-physics. Lidar: Atmospheric Sounding
Introduction; Observations for Chemistry (Remote Sensing): Lidar. Optics,
Atmospheric: Airglow Instrumentation; Optical Remote Sensing Instruments. Radiative Transfer: Cloud-radiative
Processes.
I
iii. To indicate material that broadens the discussion.
LIDAR: Backscatter
Seea/so:Aerosols: Climatologyof TroposphericAerosols: Observations and Meaurements; Physics and Chemistry of
Aerosols; Role in Cloud Physics; Role in Radiative
Introduction; DIAL; Dop
Processes.
3. Index
The index will provide you with the page number where the material is to be located, and the index entries
differentiate between material that is a whole article, is part of an article or is data presented in a table. On the
opening page of the index detailed notes are provided.
4. Appendices
In addition to the articles that form the main body of the Encyclopedia, there are a number of appendices which
provide lists of physical constants used throughout the Encyclopedia, units and their SI equivalents, the periodic
table of the elements, the geological time scale and a list of abbreviations used in the encyclopedia.
The appendices are located in volume 6, before the index.
5. Contributors
A full list of contributors appears at the beginning of each volume.
xii CONTRIBUTORS
Alexander, M J Ayrault, F
Northwest Research Associates, Inc. Meteo-France
Colorado Research Associates Division Centre National de Recherches Meteorologiques
3380 Mitchell Lane 42 av. G. Coriolis
Boulder, CO 80301, USA F-31057 Toulouse cedex, France
Anderson, D
Baer, F
European Centre for Medium-RangeWeather Forecasts
University of Maryland
Shinfield Park
Reading RG2 9AX, UK
College Park, MD 20742, USA
Angevine, W M
University of Colorado Baines, P G
Cooperative Institute for Research in Environmental CSlRO Division of Atmospheric Research
Sciences PMB1
Boulder, CO 80305, USA Aspendale, VIC 3195, Australia
Antikainen, V Baker, M B
Vaisala Oyj University of Washington
PO Box 26 Department of Atmospheric Sciences
Helsinki FIN-00421, Finland Box 351650
Aplin, K L
The University of Reading
Bales, R C
Department of Meteorology University of Arizona
Earley Gate
Department of Hydrology and Water Resources
Reading RG6 6BB, UK
PO Box 21001 I
Tucson, AZ 85721, USA
Arakawa, A
University of California Los Angeles
Department of Atmospheric Sciences Balsley, B B
Los Angeles, CA 90095, USA University of Colorado
Cooperative Institute for Research in Environmental
Argall, P S Sciences
The University of Western Ontario Boulder, CO 80309, USA
Department of Physics and Astronomy
London, Ontario N6A 3K7, Canada Barnes, I
Bergische Universitat GesamhochschuleWuppertal
Ashmore, M R Fachbereich 9-Physikalische Chemie
University of Bradford Gauss Strasse 20
Department of EnvironmentalScience D-42097 Wuppertal, Germany
Bradford BD7 1DP, UK
Bartello, P
Asman, W A H
McGill University
Danish Institute of Agricultural Sciences
Departments of Atmospheric and Oceanic Sciences
Department of Crop Physiology and Soil Science
805 Sherbrooke Street West
Research Center Foulum
Montreal, Quebec H3A 2K6, Canada
DK-8830 Tjele, Denmark
Avallone, L M Bass, H E
University of Colorado at Boulder University of Mississippi
Laboratory of Atmospheric and Space Physics National Center for Physical Acoustics
Boulder, CO 80309, USA University, MS 38677, USA
CONTRIBUTORS xiii
Baum, B A Blier, W
University of Wisconsin Madison National Weather Service
Cooperative Institute for Meteorological Satellite Studies 21 Grace Hopper Avenue, Stop 5
1225 W. Drayton Street Monterey, CA 93943, USA
Madison, WI 53706, USA
Bluestein, H B
Baumgardner, D University of Oklahoma
Universidad Nacional Autonoma de Mexico School of Meteorology
Centro Ciencias de la Atmosfera 100 E. Boyd Street, Rm 1310
Circuit0 Exterior, Ciudad Universitaria Norman, OK 73019, USA
04510 Mexico City DF, Mexico
Bluestein, M
Becker, K H Indiana University - Purdue University Indianapolis
Bergische Universitat Gesamhochschule Wuppertal Department of Mechanical Engineering Technology
Fachbereich 9-Physikalische Chemie 799 West Michigan Street
Gauss Strasse 20 Indianapolis, IN 46202, USA
D-42097 Wuppertal, Germany
Blumen, Wt
Bedard Jr, A J University of Colorado at Boulder
National Oceanic and Atmospheric Administration Program in Atmospheric and Oceanic Sciences
Environmental Technology Laboratory Campus Box 31 1
Boulder, CO 80303, USA Boulder, CO 80309, USA
Boering, K A
Beljaars, A University of California Berkeley
European Centre for Medium-Range Weather Forecasts Department of Chemistry
Shinfield Park Berkeley, CA 94720, USA
Reading RG2 9AX, UK
Bogden, P
Benilov, A Gulf of Maine Ocean Observing System
Stevens Institute of Technology PO Box 4919
Davidson Laboratory Portland, ME 041 12, USA
Hoboken, NJ 07030, USA
Boyd, J P
Betts, A University of Michigan
Atmospheric Research Department of Atmospheric, Oceanic and Space Sciences
58 Hendee Lane 2455 Hayward Avenue
Pittsford, VT 05763, USA Ann Arbor, MI 481 09, USA
Biondi, F Bradley, E F
University of Nevada CSlRO Land and Water
Department of Geography PO Box 1666
Mail Stop 154 Canberra, ACT 2601, Australia
Reno, NV 89557, USA
Branscome, L E
Blake, D R Climatological Consulting Corporation
University of California lrvine Palm Beach Gardens, FL 33418, USA
Imine, CA 92697, USA Brenninkmeijer, C A M
Max Planck Institute for Chemistry
Blake, N J Air Chemistry Department
University of California lrvine PO Box 3060
Department of Chemistry D-55020 Mainz, Germany
Irvine, CA 92697, USA
Bretherton, C S
Blanken, P D University of Washington
University of Colorado at Boulder Department of Atmospheric Sciences
Department of Geography Box 351640
Boulder, CO 80309, USA Seattle, WA 98195, USA
'deceased
xiv CONTRIBUTORS
Brimblecombe, P Catling, D
University of East Anglia University of Washington
School of Environmental Sciences Department of Atmospheric Sciences
Norwich NR4 7TJ. UK Box 35 1640
Brooks, H E
National Severe Storms Laboratory Chang, P
1313 Halley Circle Texas A&M University
Norman, OK 73069, USA Department of Oceanography
College Station, TX 77843, USA
Browell, E V
NASA Langley Research Center Changnon, S A
Mail Stop 401A lllinos State Water Survey
Hampton, VA 23681, USA Midwestern Climate Center
Champaign, IL 61820, USA
Buizza, R
Chipperfield, M P
European Centre for Medium-Range Weather Forecasting
University of Leeds
Shinfield Park
School of the Environment
Reading RG2 9AX, UK
Leeds LS2 9JT, UK
Burn, C R Chylek, P
Carleton University Dalhousie University
Department of Geography and Environmental Studies Department of Physics and Atmospheric Science
1125 Colonel By Drive Halifax, Nova Scotia B3H 3J5, Canada
Ottawa, Ontario K1S 5B6, Canada
Clayson, C A
Burns, A G Purdue University
National Center for Atmospheric Research Earth and Atmospheric Science
High Altitude Observatory 1397 Civil Engineering Building
PO Box 3000 West Lafayette, IN 47907, USA
Clemitshaw, K C
Butler, J H Imperial College of Science, Technology, and Medicine
National Oceanic and Atmospheric Administration Department of Environmental Science and Technology
325 Broadway Silwood Park, Ascot SL5 7PY, UK
Coakley, J A
Oregon State University
Calanca, P
College of Oceanic and Atmospheric Sciences
Atmospheric and Climate Science ETH
104 Ocean Admin Building
Winterthurestr. 190
Corvallis, OR 97331, USA
CH-8057 Zurich, Switzerland
Coen, J
Cantrell, C A National Center for Atmospheric Research
National Center for Atmospheric Research Boulder, CO 80307, USA
1850 Table Mesa Drive
Boulder, CO 80305, USA Coffey, M T
National Center for Atmospheric Research
Cape, J N Boulder, CO 80307, USA
Centre for Ecology and Hydrology
Edinburgh Research Station Cole, H
Bush Estate National Center for Atmospheric Research
Penicuik EH26 OQB, UK Boulder, CO 80307, USA
Carslaw, K S Colle, B A
University of Leeds State University of New York at Stony Brook
School of the Environment Institute for Terrestrial and Planetary Atmospheres
Leeds LS2 9JT, UK Stony Brook, NY 11794, USA
CONTRIBUTORS xv
Collett, J L Deshler, T
Colorado State University University of Wyoming
Atmospheric Science Department Department of Atmospheric Science
Fort Collins, CO 80523, USA Laramie, WY 82071, USA
Collins, W D DeWeaver, E
National Center for Atmospheric Research University of Washington
Climate and Global Dynamics Division Joint Institute for the Study of the Atmosphere and Ocean
Boulder, CO 80307, USA Box 354 235
Colucci, S J
Dickinson, R E
Cornell University
Department of Earth and Atmospheric Sciences Georgia Institute of Technology
School of Earth and Atmospheric Sciences
Ithaca, NY 14853, USA
Atlanta, GA 30332, USA
Conde, M Dlugokencky, E
University of Alaska Fairbanks National Oceanic and Atmospheric Administration
Geophysical Institute Climate Monitoring and Diagnostics Laboratory
903 Koyukuk Drive, PO Box 757320
325 Broadway
Fairbanks, AK 99775, USA Boulder, CO 80303, USA
Cortinas Jr, J V Donaldson, D J

University of Oklahoma University of Toronto
Cooperative Institutefor Mesoscale MeteorologicalStudies Department of Chemistry
1313 Halley Circle 80 St. George Street
Norman, OK 73069, USA Toronto, Ontario M5S 3H6, Canada
Croft, P J Doswell 111, C A

University of Louisiana at Monroe University of Oklahoma
Department of Geosciences Cooperative lntitute for Mesoscale Meteorological Studies
700 University Avenue 100 East Boyd Street, Room 1110
Monroe, LA 71209, USA Norman, OK 73071, USA
Cunningham, P Doviak, R J
Florida State University National Severe Storms Laboratory
Department of Meteorology 1313 Halley Circle
Tallahassee, FL 32306, USA Norman, OK 73069, USA
Curry, J A Drarin, P G
Georgia Institute of Technology University of Bath
School of Earth and Atmospheric Sciences Department of Mathematical Sciences
Atlanta, GA 30332, USA Bath BA2 7AY, UK
Dabberdt, W F Du, S
Vaisala Inc. California Air Resources Board
8401 Baseline Road PO Box 2815
Boulder, CO 80303, USA Sacramento, CA 95812, USA
Dudhia, A
Dameris, M University of Oxford
lnstitut fur Physik der Atmosphare Department of Physics
DLR-Oberpfaffenhofen Parks Road
D-82234 Wessling, Germany Oxford OX1 3PU, UK
Davies, H C Dunkerton, T J
Institute for Atmospheric and Climate Science ETH Northwest Research Associates
Honggerberg, HPP PO Box 3027
CH-8093 Zurich, Switzerland Bellevue, WA 98009, USA
xvi CONTRIBUTORS
Durran, D R Finnigan, J J
University of Washington CSlRO Atmospheric Research
Department of Atmospheric Sciences FC Pye Laboratory
Box 351640 Black Mountain, ACT 2601, Australia
Fischer, H
Forschungszentrum Karlsruhe
Edwards, C A
lnstitut fur Meteorologie und Klimaforschung
University of California
Postfach 3640
Ocean Sciences Department
D-76021 Karlsruhe, Germany
Santa Cruz, CA 95064, USA
Flather, R A
Egger, J Proudman Oceanographic Laboratory
Universitat Munchen Bidston Observatory
Meteorologisches lnstitut Birkenhead CH43 7RA, UK
Theresienstrasse 37
D-80333 Munchen, Germany Forbes, J
University of Colorado
Ehernberger, L J Department of Aerospace Engineering Sciences
NASA Dryden Flight Research Center Campus Box 429
Edwards, CA 93523, USA Boulder, CO 80309, USA
Forster, P
Ehhalt, D H
University of Reading
lnstitut fur Atmospharische Chemie
Forschungszentrum Julich
Earley Gate
D-52425 Julich, Germany
Reading RG6 6BB, UK
Ellrod, G P Fowler, A C
National Environmental Satellite Data and Information University of Oxford
Service Mathematical Institute
601 World Weather Building, 5200 Auth Road 24-29 St Giles’
Camp Springs, MD 20746, USA Oxford OX1 3LB, UK
Emery, W J Fowler, D
University of Colorado Centre for Ecology and Hydrology
Program in Atmospheric and Oceanic Sciences Edinburgh Research Station
Boulder, CO 80309, USA Bush Estate
Penicuik EH26 OQB, UK
Epifanio, C C
Texas A&M University Fox, J L
Department of Atmospheric Sciences Wright State University
College Station, TX 77843, USA Department of Physics
3640 Colonel Glenn Highway
Fasullo, J Dayton, OH 45435, USA
University of Colorado Boulder
Program in Atmospheric and Oceanic Sciences Frazier Doviak, M E
Campus Box 31 11 University of Oklahoma
Boulder, CO 80309, USA 100 East Boyd Street
Norman, OK 73019, USA
Feely, R A
National Oceanic and Atmospheric Administration Frederick, J E
Pacific Marine Environmental Laboratory University of Chicago
7600 Sandpoint Way NE The Department of the Geophysical Sciences
Seattle, WA 981 15, USA 5734 South Ellis Avenue
Chicago, IL 60637, USA
Fernando, H J S
Arizona State University Friedl, R R
Department of Mechanical and Aerospace Engineering California Institute of Technology
Mail Code 9809 Jet Propulsion Laboratory
Tempe, A2 85287, USA Pasadena, CA 91 109, USA
CONTRIBUTORS xvii
Fritts, D C Ghil, M
North West Research Associates Inc. University of California Los Angeles
Colorado Research Associates Division Department of Atmospheric Sciences
3380 Mitchell Lane Los Angeles, CA 90095, USA
Gierasch, P J
Fu, Qiang Cornell University
University of Washington Astronomy Department
Department of Atmospheric Sciences Ithaca, NY 14850, USA
Box 351640
Gilbert, K E
Gaidos, E J University of Mississippi
California Institute of Technology National Center for Physical Acoustics
Division of Geology and Planetary Science 1 Coliseum Drive
Mail Stop 150-21 University, MS 38677, USA
Pasadena, CA 91 125, USA
Golden, J H
Gardner, C S National Oceanic and Atmospheric Administration
University of Illinois at Urbana-Champaign Forecast Systems Laboratory
Department of Electrical and Computer Engineering, CSRL 325 Broadway
1308 W. Main Street Boulder, CO 80305, USA
Urbana, IL 61801, USA
Golding, B W
Garratt, J R Met Office
CSlRO Atmospheric Research London Road
Aspendale, VIC 3195, Australia Bracknell RG12 2SZ, UK
Gates, W L
University of California Goody, R
Lawrence Livermore National Laboratory Harvard University
7000 East Avenue Division of Engineering and Applied Sciences
Livermore, CA 94550, USA Cambridge, MA 02138, USA
Gayet, J-F Grant, W B

Universite Blaise Pascal NASA Langley Research Center
Laboratoire de Meteorologie Physique Mail Stop 401A
34, avenue Carnot Hampton, VA 23681, USA
F-63006 Clermont - Ferrand Cedex, France
Gras, J L
Gedzelman, S D CSlRO Atmospheric Research
The City College of New York PMB1
Department of Earth and Atmospheric Sciences Aspendale, VIC 3195, Australia
Convent Avenue at 138th Street
New York, NY 10031, USA
Grasso, L D
Ge, S Colorado State University
University of Colorado Cooperative Institute for Research in the Atmosphere
Department of Geological Sciences Fort Collins, CO 80523, USA
220 Colorado Avenue
Boulder, CO 80309, USA Grotjahn, R
University of California
Geernaert, G L Department of Land, Air, and Water Resources
Los Alamos National Laboratory One Shields Avenue
Institute of Geophysics and Planetary Physics Davis, CA 95616, USA
Los Alamos, NM 87545, USA
Guenther, A
Gerber, H National Center for Atmospheric Research
Gerber Scientific Inc. Atmospheric Chemistry Division
1643 Bentana Way PO Box 3000
Reston, VA 20190, USA Boulder, CO 80307, USA
xviii CONTRIBUTORS
Gutrler, D S Harries, J E
University of New Mexico Imperial College of Science, Technology and Medicine
Department of Earth and Planetary Sciences Space and Atmospheric Physics Group
200 Yale Boulevard NE The Blackett Laboratory
Albuquerque, NM 87131, USA Prince Consort Road
London SW7 2BW, UK
Haberle, R M
NASA Ames Research Center Harris, N R P
Space Science Division European Ozone Research Coordinating Unit
Moffett Field, CA 94035, USA 14 Union Road
Cambridge CB2 1HE, UK
Hagan, M
National Center for Atmospheric Research Harrison, R G
High Altitude Observatory The University of Reading
PO Box 3000 Department of Meteorology
Boulder, CO 80307, USA Earley Gate
Reading RG6 6BB, UK
Hahn, C J
University of Arizona Hartmann, D L
Department of Atmospheric Sciences University of Washington
Tucson, A2 85721, USA Department of Atmospheric Sciences
Haigh, J D
Hase, F
Imperial College of Science, Technology and Medicine
Forschungszentrum Karlsruhe
Space and Atmospheric Physics Group
lnstitut fur Meteorologie und Klimaforschung
The Blackett Laboratory
Postfach 3640
Prince Consort Road
D-76021 Karlsruhe, Germany
London SW7 2BW. UK
Hasse, L
Hakim, G J Universitat Kiel
University of Washington lnstitut fur Meereskunde
Department of Atmospheric Sciences D-24105 Kiel, Germany
Box 351640
Hastenrath, S
University of Wisconsin
Hamill, T M Department of Atmospheric and Ocean Sciences
National Oceanic and Atmospheric Administration 1225 W Drayton Street
ClRES Climate Diagnostics Center Madison, WI 53706, USA
Haynes, P H
University of Cambridge
Hamilton, K
Department of Applied Mathematics and Theoretical
University of Hawaii at Manoa
Physics
International Pacific Research Center
Wilberforce Road
Honolulu, HI 96822, USA
Cambridge CB3 OWA, UK
Hanisco, T F Hays, P B
Harvard University University of Michigan
Department of Chemistry and Chemical Biology Space Physics Research Laboratory
Cambridge, MA 02138, USA 2455 Hayward St.
Ann Arbor, MI 48109, USA
Hardesty, R M
National Oceanic and Atmospheric Administration Healy, S B
Environmental Technology Laboratory The Met Office
325 Broadway London Road
Boulder, CO 80305, USA Bracknell RG 12 2SY, UK
CONTRIBUTORS xix
Heard, D E Holton, J R
University of Leeds University of Washington
Department of Chemistry Department of Atmospheric Sciences
Leeds LS2 9JT. UK Box 351640
Hedger, M M
Environment Agency Holtslag, A A M
Rio House, Waterside Drive Wageningen University
Almondsbury Meteorology and Air Quality Section
Bristol BS32 4UD, UK Duivendaal 2
6701 AP Wageningen, The Netherlands
Heintzenberg, J
Institute for Tropospheric Research Horhammer, J
Permoserstr. 15 Vaisala Oyj
D-04318 Leipzig, Germany PO Box 26
Helsinki FIN-00421, Finland
Hemsley, J M
Houghton, D D
National Data Buoy Center
University of Wisconsin-Madison
Stennis Space Center, MS 39529, USA
Department of Atmospheric and Oceanic Sciences
1225 West Dayton Street
Herckes, P Madison, WI 53706, USA
Colorado State University
Atmospheric Science Department
Houweling, S
Fort Collins, CO 80523, USA Universiteit Utrec ht
lnstitut voor Marien and Atmosterich Onderzoek
Herring, D D Princetonplein 5
Science Systems and Applications Inc. 3584 CC Utrecht, The Netherlands
Lanham, MD 20716, USA
Hu, H
Hess, G D Harvard University
Bureau of Meteorology Department of Chemistry and Chemical Biology
Research Centre Cambridge, MA 02138, USA
Melbourne, VIC 3001, Australia
Huaman, M M
Hess, P Cornell University
National Center for Atmospheric Research Department of Electrical and Computer Engineering
Atmospheric Chemistry Division 378 Rhodes Hall
PO Box 3000 Ithaca, NY 14853, USA
Hungate, B A
Hintsa, E Northern Arizona University
Woods Hole Oceanographic Institution Department of Biological Sciences
Department of Marine Chemistry and Geochemistry Flagstaff, AZ 8601 1, USA
Woods Hole, MA 02543, USA
Hunt, J C R
Hoerling, M University College London
Climate Diagnostics Center Department of Space and Climate Physics
National Oceanic and Atmospheric Administration 17 Gordon Street
325 Broadway London WC1 H OAH, UK
Hurrell, J W
Hofmann, D J National Center for Atmospheric Research
National Oceanic and Atmospheric Administration Climate Analysis Section
Climate Monitoring and Diagnostics Laboratory PO Box 3000
xx CONTRIBUTORS
Ismail, S Jones, P D
NASA Langley Research Center University of East Anglia
Mail Stop 401A Climatic Research Unit
Hampton, VA 23681, USA Norwich NR4 7TJ, UK
Jaffe, D A Juckes, M
University of Washington University of Oxford
Interdisciplinary Arts and Sciences Atmospheric, Oceanic and Planetary Physics
18115 Campus Way NE Clarendon Laboratory, Parks Road
Bothell, WA 9801 1, USA Oxford OX1 3PU, UK
Jakob, C Kantha, L
European Centre for Medium-Range Weather Forecasting SACLANT Undersea Research Center
Shinfield Park Viale S. Botolemeo 400
Reading RG2 9AX, UK 19138 La Spezia, Italy
James, I N
University of Reading Kaufman, Y J
Department of Meteorology NASA Goddard Space Flight Center
Earley Gate Mail Code 913
Reading RG6 6BB, UK Greenbelt, MD 20771, USA
Jenkins, G S Kelley, M C
Pennsylvania State University Cornell University
Department of Meteorology Department of Electrical and Computer Engineering
University Park, PA 16802, USA 318 Rhodes Hall
Ithaca, NY 14853, USA
Jennings, S G
National University of Ireland Key, R M
Department of Physics Princeton University
Galway, Ireland Department of Geosciences
Princeton, NJ 08540, USA
Jenniskens, P
NASA Ames Research Center
Keyser, D
SET1 Institute
University at Albany, State University of New York
Mail Stop 239-4
Department of Earth and Atmospheric Sciences
Moffett Field, CA 94035, USA
Albany, NY 12222, USA
Jensen, N O Kidder, S Q
Ris~National Laboratory Colorado State University
Wind Energy and Atmospheric Physics Department
Cooperative Institute for Research in the Atmosphere
Frederiksborgvej 399 Fort Collins, CO 80523, USA
DK-4000 Roskilde, Denmark
Jickells, T D Killeen, T L
University of East Anglia National Center for Atmospheric Research
School of Environmental Sciences High Altitude Observatory
Norwich NR4 7TJ, UK PO Box 3000
Johnson, R H
Colorado State University King, M D
Department of Atmospheric Science NASA Goddard Space Flight Center
Fort Collins, CO 80523, USA Mail Code 900
Joly, A
Meteo-France Klinger, B
Centre National de Recherches Meteorologiques Center for Ocean-Land-Atmosphere Studies
42 av. G. Coriolis 4041 Powder Mill Road, Suite 302
F-31057 Toulouse cedex, France Calverton, MD 20705, USA
CONTRIBUTORS xxi
Knight, C Lamarque, J F
National Center for Atmospheric Research National Center for Atmospheric Research
Mesoscale and Microscale Meteorology Division Atmospheric Chemistry Division
PO Box 3000 PO Box 3000
Knight, N Lamb, D
National Center for Atmospheric Research Pennsylvania State University
Mesoscale and Microscale Meteorology Division Department of Meteorology
PO Box 3000 519 Walker Building
Boulder, CO 80307, USA University Park, PA 16802, USA
Knox, J A Larsen, M F
University of Georgia Clemson University
Faculty of Engineering Department of Physics and Astronomy
Athens, GA 30602, USA Clemson, SC 29634, USA
Koch, O W Lau, K-M

Northern Arizona University NASA Goddard Space Flight Center
Department of Biological Sciences Climate and Radiation Branch
Flagstaff, AZ 86011, USA Greenbelt, MD 20771, USA
Kondo, Y Leary, C A
The University of Tokyo Texas Tech University
Research Center for Advanced Science and Technology Atmospheric Science Group
4-6-1 Komaba, Meguro Lubbock, TX 79409, USA
Tokyo 153-8904, Japan
Le Bras, G
Korolev, A Centre National de Recherches
Meteorological Service of Canada Laboratoire de Combustion et Systemes Reactifs
Downsview, Ontario M3H ST4, Canada F-45071 Orleans, France
Kreidenweis, S M Ledley, T S
Colorado State University TERC
Department of Atmospheric Science Center for Earth and Space Science Education
Fort Collins, CO 80523, USA 2067 Massachusetts Avenue
Cambridge, MA 02140, USA
Kumar, A
National Weather Service Headquarters LeMone, M A
Climate Prediction Center National Center for Atmospheric Research
World Weather Building, 5200 Auth Road PO Box 3000
Washington, DC 20746, USA Boulder, CO 30307, USA
Lacis, A Lenton, T
NASA Goddard Institute for Space Studies Centre for Ecology and Hydrology
2880 Broadway Edinburgh Research Station
New York, NY 10025, USA Bush Estate
Penicuik EH26 OQB, UK
Laing, A G
University of South Florida Lester, P F
Department of Geography 416 Guthrie Street
Tampa, FL 33620, USA Ashland, OR 97520, USA
Lal, D Lewis, A C
Scripps Institution of Oceanography University of Leeds
University of California San Diego School of the Environment
9500 Gilman Drive Woodhouse Lane
La Jolla, CA 92093, USA Leeds LS2 9JT, UK
xxii CONTRIBUTORS
Lippmann, M Malardel, S
New York University Meteo-France
School of Medicine Centre National de Recherches Meteorologiques
Nelson Institute of Environmental Medicine 42 av. G. Coriolis
57 Old Forge Road F-31057 Toulouse cedex, France
Tuxedo, NY 10987, USA
Mankin, W G
Liu, Guosheng National Center for Atmospheric Research
Florida State University Boulder, CO 80307, USA
Tallahassee, FL 32306, USA Mansfield, D
Met Office
Liu, W T London Road
California Institute of Technology Bracknell RG12 2SZ, UK
Jet Propulsion Laboratory
4800 Oak Grove Drive Marks, F D
Pasadena, CA 91 109, USA National Oceanic and Atmospheric Administration
Atlantic Oceanographic and Meteorological Laboratories
Lopez-Puertas, M Hurricane Research Division
lnstituto de Astrofisica de Andalucia 4301 Rickenbacker Causeway
Camino Baja de Huetor, 24 Miami, FL 33149, USA
Granada, CSIC, Spain
Martinson, D G
Lorenc, A C Columbia University
The Met Office Lamont-Doherty Earth Observatory
Forecasting Research Division PO Box 1000,61 Route 910,
London Road Palisades, NY 10964, USA
Bracknell RG12 2SZ, UK
Mason, P J
Loutre, M F Met Office
Universite Catholique de Louvain London Road
lnstitut d’Astronomie et de Geophysique Georges Bracknell RG12 2SZ, UK
Lemaitre
2, Chemin du Cyclotron McCulloch, A
B-1348 Louvain-la-Neuve,Belgium University of Bristol
School of Chemistry
LY, L Cantocks Close
Naval Postgraduate School Bristol BS8 lTS, UK
Department of Oceanography
833 Dyer Road Mclntyre, M E
Monterey, CA 93943, USA University of Cambridge
Department of Applied Mathematics and Theoretical
Lyons, W A Physics
FMA Research Inc. Wilberforce Road
Yucca Ridge Field Station Cambridge CB3 OWA, UK
46050 Weld County Road 13
Fort Collins, CO 80524, USA McKenzie, R
National Institute for Water and Atmospheric Research
Madden, R A Private Bag 50 061, Omaku
National Center for Atmospheric Research Lauder, Central Otago 9182, New Zealand
McMurry, P H
Madronich, S University of Minnesota
National Center for Atmospheric Research Department of Mechanical Engineering
Boulder, CO 80307-3000, USA 111 Church Street SE
Minneapolis, MN 55455, USA
Mahrt, L
Oregon State University McPeters, R D
College of Oceanic and Atmospheric Sciences NASA Goddard Space Flight Center
104 Ocean Admin Building Atmospheric Chemistry and Dynamics Branch
Corvallis, OR 97331, USA Greenbelt, MD 20771, USA
CONTRIBUTORS xxiii
Mechoso, C R Nicholls, N
University of California Los Angeles Bureau of Meteorology Research Centre
Department of Atmospheric Sciences GPO Box 1289K
Los Angeles, CA 90095, USA Melbourne, VIC 3001, Australia
Midgley, P M
M & D Consulting Nieuwstadt, F T M
Ludwigstrasse 49 Delft University of Technology
D-70771 Leinfelden Musberg, Germany Laboratory of Aero and Hydrodynamics
Leeghwaterstraat 21
Miller, M 2628 CA Delft, Netherlands
European Centre for Medium-Range Weather Forecasting
Shinfield Park
Reading RG2 9AX, UK Nigam, S
University of Maryland
Minnis, P Department of Meteorology
NASA Langley Research Center College Park, MD 20742, USA
21 Langley Boulevard
Hampton, VA 23681, USA
Nightingale, P D
Plymouth Marine Laboratory
Mishchenko, M
Prospect Place, West Hoe
NASA Goddard Institute for Space Studies
Plymouth PL1 3DH, UK
2880 Broadway
New York, NY 10025, USA
Nisbet, E
Royal Holloway, University of London
Mitchell, N J
Department of Geology
University of Bath
Egham TW20 OEX, UK
Department of Electronic and Electrical Engineering
Bath BA2 7AY, UK
Nkemdirim, L C
Moeng, C-H University of Calgary
National Center for Atmospheric Research Department of Geography
Mesoscale and Microscale Meteorology Division 2500 University Drive NW
Boulder, CO 80307, USA Calgary, Alberta T2N 1N4, Canada
Molteni, F
Abdus Salam International Centre for Theoretical Physics North, G R
Strada Costiera 11 Texas A&M University
Trieste 34100, Italy Climate System Research Program
Moncrieff, M W College Station, TX 77843, USA
National Center for Atmospheric Research
Mesoscale & Microscale Meteorology Division
PO Box 3000 Norton, W A
Boulder, CO 80307, USA University of Oxford
Atmospheric, Oceanic and Planetary Physics
Moore, D W Clarendon Laboratory
Pacific Marine Environmental Laboratory Parks Road
7600 Sand Point Way NE Oxford OX1 3PU, UK
Novelli, P C
National Oceanic and Atmospheric Administration
Munthe, J
Climate and Diagnostic Laboratory
IVL Swedish Environmental Research Institute
Box 47086
S-402 58 Goteborg, Sweden
NUSS, W A
Newman, P A Naval Postgraduate School
NASA Goddard Space Flight Center Department of Meteorology
Greenbelt, MD 20771, USA Monterey, CA 93943, USA
xxiv CONTRIBUTORS
O’Neill, A Platt, C M R
University of Reading Colorado State University
Department of Meteorology Department of Atmospheric Science
Earley Gate Colorado, CO 80523, USA
Reading RG6 6BB, UK
Politovich, M K
Osterkamp, T E National Center for Atmospheric Research
University of Alaska Boulder, CO 80307, USA
Geophysical Institute
Fairbanks, AK 99775, USA
Pommereau, J-P
Service d’Aeronomie, CNRS
Paillard, D
F-91371 Verrieres le Buisson Cedex, France
Commissariat a I’Energie Atomique
Laboratoiredes Sciences du Climat et de I’Environnement
F-91191 Gif-sur-Yvette, France Poulet, G
University of Orleans
Parish, T R Laboratoire de Physique et Chimie de I’Environnement
University of Wyoming F-45071 Orleans Cedex. France
Department of Atmospheric Science
Laramie, WY 82071, USA
Price, C
Parker, D J Tel Aviv University
University of Leeds Department of Geophysics and Planetary Sciences
School of the Environment Tel Aviv 69978, Israel
Leeds LS2 9JT, UK
Pyle, J A
Parks, G K University of Cambridge
University of California Berkeley
Centre for Atmospheric Science
Space Science Laboratory
Chemistry Department
Berkeley, CA 94720, USA
Cambridge CB2 1EW, UK
Paukkunen, A
Vaisala Oyj Quayle, R
PO Box 26 National Oceanic and Atmospheric Administration (Retd.)
Helsinki FIN-00421, Finland 1 Botany View Court
Asheville, NC 28805, USA
Pendleton Jr, W R
Utah State University
Physics Department Rajeevan, M
Logan, UT 84322, USA India Meteorological Department
University Road Shivajinagar
Pfeffer, R L Pune 41 1005, India
Florida State University
Ramamurthy, M K
Tallahassee, FL 32306, USA
University of Illinois Urbana-Champaign
PielkeSr, R A
105 S.Gregory Street
Colorado State University
Department of Atmospheric Science
Fort Collins, CO 80523, USA
Randel, W J
Pinnick, R National Center for Atmospheric Research
US Army Research Laboratory Atmospheric Chemistry Division
2800 Powder Mill Road Boulder, CO 80307, USA
Adelphi, MD 20783, USA
Plane, J M C Rangno, A L
University of East Anglia University of Washington
School of Environmental Sciences Department of Atmospheric Sciences
Norwich NR4 7TJ, UK Seattle, WA 98195, USA
CONTRIBUTORS xxv
Rauber, R M Roulstone, I
University of Illinois Urbana-Champaign University of Reading
Department of Atmospheric Sciences Joint Centre for Mesoscale Meteorology
105 S. Gregory Street Reading RG6 6BB, UK
Rudolph, J
Ravishankara, A R York University
NOAA Aeronomy Laboratory Centre for Atmospheric Chemistry
325 Broadway 4700 Keele Street
Boulder, CO 80305, USA Toronto, Ontario M3J 1P3, Canada
Renfrew, I A Sabine, C L
British Antarctic Survey University of Washington
Physical Sciences Division 7600 Sandpoint Way NE
High Cross, Madingley Road Seattle, WA 981 15, USA
Cambridge CB3 OET, UK
Salstein, D A
Reynolds, R Atmospheric and Environmental Research, Inc.
University of Reading 131 Hartwell Avenue
Department of Meteorology Lexington, MA 02421, USA
Early Gate
Reading RG6 6BB, UK Schar, C
Atmospheric and Climatic Science ETH
Rhines, P B Winterthurerstr. 190
University of Washington CH 8057 Zurich, Switzerland
School of Oceanography
Box 351640 Schmidt, U
Seattle, WA 98195, USA Johann Wolfgang Goethe- Universitat Frankfurt am Main
lnstutut fur Meteorologie und Geophysik
Richmond, A George-Voigt Strasse 14,
National Center for Atmospheric Research D-60325 Frankfurt am Main, Germany
High Altitude Observatory
PO Box 3000 Schoeberl, M R
Boulder, CO 80307, USA NASA Goddard Space Flight Center
Roble, R G
National Center for Atmospheric Research Schulz, J
High Altitude Observatory University of Bonn
PO Box 3000 Meteorological Institute
Boulder, CO 80307, USA Auf dem Hugel 20
D-53121 Bonn, Germany
Robock, A Seinfeld, J H
Rutgers University California Institute of Technology
Department of Environmental Sciences Pasadena, CA 91 125, USA
14 College Farm Road
New Brunswick, NJ 08901, USA
Senff, C J
University of Colorado
Rotach, M W Cooperative Institute for Research in Environmental
Swiss Federal Institute of Technology Sciences
Institute for Atmospheric and Climate Science 325 Broadway
Winterthurerstrasse 190 Boulder, CO 80305, USA
CH-8057 Zurich, Switzerland
Serreze, M C
Rotunno, R University of Colorado
National Center for Atmospheric Research Cooperative Institute for Research in Environmental
Mesoscale and Microscale Meteorology Division Sciences
xxvi CONTRIBUTORS
Shaw, G Smith, A K
University of Alaska National Center for Atmospheric Research
Geophysical Institute Mesa Laboratory
Fairbanks, AK 99775, USA PO Box 3000
Shellhorn, R
Vaisala Inc. Smith, L A
8401 Baseline Road London School of Economics
Boulder, CO 80303, USA Department of Statistics
Houghton Street
London WC2A 2AE, UK
Shepherd, G G
York University
Department of Earth and Atmospheric Science Smith, R B
4700 Keele Street Yale University
Toronto, Ontario M3J 1P3, Canada Department of Geology and Geophysics
PO Box 208109
New Haven, CT 06520, USA
Shepherd, T G
University of Toronto
Department of Physics Sokolik, I N
60 St. George Street University of Colorado at Boulder
Toronto, Ontario M5S 1A7, Canada Program in Atmospheric and Oceanic Sciences
Showman, A P
University of Arizona Soloviev, A
Lunar and Planetary Laboratory Nova Southeastern University
Tucson, AZ 85721, USA Oceanographic Center
8000 North Ocean Drive
Dania Beach, FL 33004, USA
Sica, R J
The University of Western Ontario
Sommar, J
Department of Physics and Astronomy
Goteborg University
London, Ontario N6A 3K7, Canada
S-412 96 Goteborg, Sweden
Singh, H B
NASA Ames Research Center Sousounis, P J
Earth Science Division Michigan State University
Moffett Field, CA 94035, USA Department of Atmospheric, Oceanic and Space Sciences
1541D Space Research Building
Skinner, W R Ann Arbor, MI 481 09, USA
University of Michigan
Space Physics Research Laboratory Stamnes, K
2455 Hayward Street Stevens Institute of Technology
Ann Arbor, MI 48109, USA Department of Physics and Engineering Physics
Castle Point on Hudson
Hoboken, NJ 07030, USA
Slingo, J
University of Reading
Department of Meteorology Steig, E J
Earley Gate University of Washington
Reading RG6 6BB, UK Department of Earth and Space Sciences
Smit, H G J
Research Centre Julich Stensrud, D J
Institute for Chemistry and Dynamics of the Geosphere National Severe Storms Laboratory
PO Box 1913 1313 Halley Circle
D-52425 Julich, Germany Norman, OK 73069, USA
CONTRIBUTORS xxvii
Stolarski, R S Thapliyal, V
NASA Goddard Space Flight Center India Meteorological Department
Atmospheric Chemistry and Dynamics Branch University Road Shivajinagar
Greenbelt, MD 20771, USA Pune 41 1005, India
Sturm, M Thomas, 0 E
US Army Cold Regions Research & Engineering University of Colorado
Laboratory-Alaska Laboratory for Atmospheric and Space Physics
PO Box 35170 Boulder, CO 80309, USA
Fort Wainwright, AK 99703, USA
Thomson, D J
Suess, S T Met Office
NASA Marshall Space Flight Center London Road
Huntsville, AL 35801, USA Bracknell RG12 2SZ. UK
Sullivan, P Thuburn, J
National Center for Atmospheric Research University of Reading
Mesoscale and Microscale Meteorology Division Department of Meteorology
Boulder, CO 80307, USA Earley Gate
Reading RG6 6BB, UK
Sulrer, M P
Arecibo Observatory Toggweiler, J R
HC 3 Box 53995 National Oceanic and Atmospheric Administration
Arecibo, PR 00612, USA Geophysical Fluid Dynamics Laboratory
PO Box 308
Princeton, NJ 08542, USA
Takle, E S
Iowa State University Tomas, R A
Department of Agronomy University of Colorado Boulder
Ames, IA 50011, USA Program in Atmospheric and Oceanic Sciences
Campus Box 31 1
Tanre, D Boulder, CO 80309, USA
Universite de Sciences et Techniques de Lille
Laboratoire d’Optique Atmospherique, CNRS Toohey, D
Villeneuve d’Ascq, France University of Colorado
Program in Atmospheric and Oceanic Sciences
Tao, W-K Boulder, CO 80309, USA
NASA Goddard Space Flight Center
Mesoscale Modeling and Dynamics Group Travis, L
Greenbelt, MD 20771, USA NASA Goddard Institute for Space Studies
2880 Broadway
Taylor, F W New York, NY 10025, USA
Oxford University
Atmospheric, Oceanic and Planetary Physics Trier, S 6
Clarendon Laboratory National Center for Atmospheric Research
Parks Road Mesoscale and Microscale Meteorology
Oxford OX1 3PU, UK Boulder, CO 80307, USA
Taylor, M J Tsurutani, B T
Utah State University Jet Propulsion Laboratory
Space Dynamics Laboratory 4800 Oak Grove Drive
Logan, UT 84322, USA Pasadena, CA 91 109, USA
Taylor, P K Turner, J
Southampton Oceanography Centre British Antarctic Survey
Empress Dock High Cross, Madingley Road
Southampton SO14 3ZH, UK Cambridge CB3 OET, UK
xxviii CONTRIBUTORS
Twohy, C Wang, B
Oregon State University University of Hawaii
College of Oceanic and Atmospheric Sciences Department of Meteorology
104 Ocean Admin Building 2525 Correa Road
Corvallis, OR 97331, USA Honolulu, HI 96822, USA
Uttal, T Wang, G
National Oceanic and Atmospheric Administration University of Connecticut
Environmental Technology Laboratory Department of Civil and Environmental Engineering
325 Broadway Storrs, CT 06269, USA
Warren, S G
Vaughan, G University of Washington
University of Wales Atmospheric Sciences and Geophysics
Physics Department Box 351640
Aberystwyth SY23 3BZ, UK Seattle, WA 98195, USA
Vaughan, W W Waters, J
University of Alabama in Huntsville California Institute of Technology
Atmospheric Science Department Jet Propulsion Laboratory
Huntsville, AL 35899, USA 4800 Oak Grove
Venkatram, A
University of California Riverside Wayne, R P
Department of Mechanical Engineering University of Oxford
Riverside, CA 92521, USA Physical and Theoretical Chemistry Laboratory
South Parks Road
Vincent, R K Oxford OX1 3Q2, UK
Bowling Green State University
Department of Geology Webster, P J
Bowling Green, OH 43403, USA University of Colorado Boulder
Program in Atmospheric and Oceanic Sciences
Vong, R J Campus Box 31 11
Oregon State University Boulder, CO 80309, USA
College of Oceanic and Atmospheric Sciences
104 Ocean Admin Building Weeks, W F
Corvallis, OR 97331, USA 6533 SW 34th Avenue,
Portland, OR 97201, USA
Wahner, A
lnstitut fur Atmospharische Chemie Weinheimer, A
Forschungszentrum Julich Scripps Institution of Oceanography
D-52425 Julich, Germany University of California San Diego
La Jolla, CA 92093, USA
Wakimoto, R M
University of California Weinstock, E
Department of Atmospheric Sciences Harvard University
Los Angeles, CA 90095, USA Department of Chemistry and Chemical Biology
12 Oxford Street
Walker, D E Cambridge, MA 02138, USA
State University of New York
Institute for Terrestrial and Planetary Atmospheres Weisman, M L
Marine Sciences Research Center National Center for Atmospheric Research
Stony Brook, NY 11794, USA Mesoscale Dynamics Group
Wallace, J M
University of Washington Weiss, S J
Department of Atmospheric Sciences National Severe Storms Laboratory
Box 351640 1313 Halley Circle
Seattle, WA 98195, USA Norman, OK 73069, USA
CONTRIBUTORS xxix
Weiss-Penzlas, P S Woodman, R F
University of Washington lnstituto Geofisico del Peru
Interdisciplinary Arts and Sciences Apartado 13-0207
18115 Campus Way NE Lima 13, Peru
Bothell, WA 9801 1, USA
Wuebbles, D J
Wells, N C University of Illinois
University of Southampton Department of Atmospheric Sciences
Southampton Oceanography Centre 105 S. Gregory Avenue
European Way Urbana, IL 61801, USA
Southampton SQ 14 3ZH, UK
Wernli, H Yang, P
Institute for Atmospheric and Climate Science ETH Texas A&M University
Honggerberg HPP Department of Atmospheric Sciences
CH-8093 Zurich. Switzerland College Station, TX 77843, USA
Westwater, E R Yang, S
University of Colorado National Oceanic and Atmospheric Administration
Cooperative Institute for Research in Environmental Climate Prediction Center
Sciences Camp Springs, MD 20746, USA
325 Broadway
Boulder, CO 80305, USA Yiou, P
Laboratoire des Sciences du Climat et de I’Environnement
Wetter, T CEA-CNRS, CE Saclay I’Orme des Merisiers
Johann Wolfgang Goethe- University Frankfurt am Main F-91191 Gif-sur-Yvette, France
lnstutut fur Meteorologie und Geophysik
George-Voigt Strasse 14 Young, J A
D-60325 Frankfurt am Main, Germany University of Wisconsin
Department of Atmospheric and Oceanic Sciences
Wheeler, M C Madison, WI 53706, USA
Bureau of Meteorology Research Centre
PO Box 1289K Yu, z
Melbourne, VIC 3001, Australia University of Nevada Las Vegas
Department of Geoscience
Las Vegas, NV 89154, USA
White, A A
Met Office
Yung, Y L
London Road
California Institute of Technology
Bracknell RG12 2SZ, UK
Division of Geology and Planetary Science
Whiteman, D
NASA Goddard Space Flight Center Yuter, S E
Laboratory for Atmospheres University of Washington
Greenbelt, MD 20771, USA Department of Atmospheric Sciences
Box 351640
Wilhite, D A Seattle, WA 98195, USA
University of Nebraska Lincoln
National Drought Mitigation Center Zahnle, K
Lincoln, NE 68583, USA NASA Ames Research Center
Space Science Division
Williams, E R Moffett Field, CA 94035, USA
Massachusetts Institute of Technology
Parsons Laboratory Zebiak, S E
77 Massachusetts Avenue International Research Institute for Climate Prediction
Cambridge, MA 02139, USA Palisades, NY 10964, USA
CONTENTS xxxi
Volume 1
A
ACOUSTIC WAVES KE Gilbert, HE Bass 1
AEOLIAN TRANSPORT See DUST
AEROSOLS
Climatology of Tropospheric Aerosols J L Crus 13
Observations and Measurements PH McMurry 20
Physics and Chemistry of Aerosols J Heintzenberg 34
Role in Cloud Physics SM Kreidenweis 40
Role in Radiative Transfer W D Collins 48
AGRICULTURAL METEOROLOGY AND CLIMATOLOGY ES Takle 54
AIRCRAFT EMISSIONS R R Fried1 60
AIRCRAFT ICING M K Politovich 68
AIR-SEA INTERACTION
Freshwater Flux J Schulz 75
Gas Exchange PD Nightingale 84
Momentum, Heat, and Vapor Fluxes PK Taylor 93
Sea Surface Temperature W j Emery 100
Storm Surges R A Flather 109
Surface Waves L Ly, A Benilov 118
ANGULAR MOMENTUM OF THE ATMOSPHERE D A Salstein 128
ANTARCTIC CLIMATE J Turner 135
ANTICYCLONES SJ Colucci 142
ARCTIC CLIMATE M C Serreze 146
ARCTIC HAZE G Shaw 155
ATMOSPHERIC TIDES M Hagan, J Forbes, A Richmond 158
AVIATION WEATHER HAZARDS AJ Bedard J r 166
B
BAROCLINIC INSTABILITY R Grotjahn 179
BEAUFORT WIND SCALE L Hasse 189
BIOGEOCHEMICAL CYCLES
Carbon Cycle E Nisbet 196
Heavy Metals T D Jickells 201
Nitrogen Cycle D A Jaffe, PS Weiss-Penzias 205
Sulfur Cycle P Brimblecombe 213
BOUNDARY LAYERS
Overview PJ Mason, DJ Thomson 22 1
Coherent Structures FTM Nieuwstadt, J C R Hunt 22 8
xxxii CONTENTS
Complex Terrain JJ Finnigan 234

Convective Boundary Layer M A LeMone 244
Modeling and Parameterization A A M Holtslag 253
Neutrally Stratified Boundary Layer J R Garratt, G D Hess 262
Observational Techniques-Remote W M Angevine, CJ Senff, E R Westwater 271
Observational Techniques In Situ EF Bradley 280
Ocean Mixed Layer L Kantha, C A Clayson 29 1
Stably Stratified Boundary Layer L Mahrt 298
Surface Layer G L Geernaert 305
BOW ECHOS AND DERECHO M Weisman 311
BUOYANCY AND BUOYANCY WAVES See also PARAMETERIZATION OF PHYSICAL
PROCESSES: Gravity Wave Fluxes
BUOYANCY AND BUOYANCY WAVES
Optical Observations MJ Taylor, W R Pendleton 323
Theory TJ Dunkerton 330
C
CARBON DIOXIDE C L Sabine, R A Feely 335
CHAOS See PREDICTABILITY AND CHAOS
CHEMISTRY OF THE ATMOSPHERE
Chemical Kinetics RP Wayne 343
Gas Phase Reactions G Le Bras 352
Ion Chemistry J L Fox 359
Laboratory Studies G Poulet 3 75
Principles of Chemical Change RP Wayne 3 84
CIRRUS CLOUDS See CLOUDS: Classification
CLEAR AIR TURBULENCE GP Ellrod, PF Lester, LJ Ehernberger 393
CLIMATE See also TROPICAL METEOROLOGY: Tropical Climates
CLIMATE
Overview D L Hartmann 403
CLIMATE PREDICTION (EMPIRICAL AND NUMERICAL) S Hastenruth 411
CLIMATE VARIABILITY
Decadal to Centennial Variability D G Martinson 418
Glacial, Interglacial Variations P Yiou, D Paillard 424
Nonlinear Aspects M Ghil 432
Volume 2
CLIMATE VARIABILITY
North Atlantic and Arctic Oscillation JW Hurrell 439
Seasonal and Interannual Variability DS Gutzler 445
CLOUD CHEMISTRY J L Collett, P Herckes 45 1
CLOUD MICROPHYSICS D Lamb 459
CLOUDS See also PARAMETERIZATION OF PHYSICAL PROCESSES: Clouds
CONTENTS xxxiii
CLOUDS
Classification A L Rangno 467
Climatology SG Warren, CJ Hahn 476
Cumuliform C A Leary 484
Measurement Techniques In Situ D Baumgardnec J-F Gayet, H Gerber, A Korolev, C Twohy 489
COASTAL METEOROLOGY WA Nuss 498
COLD AIR DAMMING B A Colle 5 04
CONTRAILS P Minnis 509
CONVECTION
Convection in the Ocean A Soloviev, B Klinger 52 1
Laboratory Models HIS Fernando 528
CONVECTIVE CLOUD SYSTEMS MODELING W-K Tao, M W Moncrieff 539
CONVECTIVE STORMS
Overview M L Weisman 548
Convective Initiation SB Trier 560
CORIOLIS FORCE D W Moore 570
COUPLED OCEAN-ATMOSPHERE MODELS D Anderson 5 74
CRITICAL LAYERS PH Haynes 5 82
CYCLOGENESIS GJ Hakim 589
CYCLONES See HURRICANES; POLAR LOWS; THERMAL LOW
CYCLONES, EXTRA TROPICAL A Joly, F Ayrault, S Malardel 5 94
DATA ANALYSIS See also WEATHER PREDICTION: Data Assimilation

DATA ANALYSIS
Empirical Orthogonal Functions and Singular Vectors CS Bretherton 617
Time Series Analysis G R North 62 1
DENSITY CURRENTS PG Baines 625
DEPOSITION RJ Vong 630
DERECHO See BOW ECHOS AND DERECHO
DESERTS AND DESERTIFICATION G Wang, GS Jenkins 633
DIURNAL CYCLE A Betts 640
DOWNSLOPE WINDS D R Durran 644
DROUGHT D A Wilhite 650
DRY LINE LD Grass0 658
DUST IN Sokolik 668
DYNAMIC METEOROLOGY See also BAROCLINIC INSTABILITY; HAMILTONIAN
DYNAMICS; INSTABILITY: Inertial Instability; INSTABILITY: Symmetric Stability;
INSTABILITY: Wave-CISK; KELVIN-HELMHOLTZ INSTABILITY;
LAGRANGIAN DYNAMICS; QUASI-GEOSTROPHIC THEORY; ROSSBY WAVES;
VORTICITY
xxxiv CONTENTS
DYNAMIC METEOROLOGY
Overview J R Holton 673
Balanced Flows M E McIntyre 680
Potential Vorticity M E McIntyre 685
Primitive Equations AA White 694
Waves J R Holton 703
EL NIRO AND THE SOUTHERN OSCILLATION

Observation N Nicholls 713
Theory P Chang, SE Zebiak 719
ELECTRICITY, ATMOSPHERIC
Global Electrical Circuit E R Williams 724
Ions in the Atmosphere R G Harrison, K L Aplin 733
Sprites WA Lyons 73 8
ENERGY BALANCE MODEL, SURFACE TS Ledley 747
EVOLUTION OF ATMOSPHERIC OXYGEN D Catling, K Zahnle 754
EVOLUTION OF EARTH’S ATMOSPHERE E ] Gaidos, Y L Yung 762
FLOODING C A Doswell I I I 769

FOG PJCroft 777
FORENSIC METEOROLOGY LE Branscome 793
FRONTOGENESIS W Blumen 799
FRONTS W Blumen 805
GAIA HYPOTHESIS T Lenton 815

GENERAL CIRCULATION See also ANGULAR MOMENTUM OF THE ATMOSPHERE;
HADLEY CIRCULATION; HURRICANES
GENERAL CIRCULATION
Overview J M Wallace 82 1
Energy Cycle R Grotjahn 829
Mean Characteristics R Grotjahn 841
Momentum Budget J Egger 855
Models C R Mechoso, A Arakawa 861
GLACIERS A C Fowler 869
GLOBAL CHANGE
Biospheric Impacts and Feedbacks B A Hungate, G W Koch 876
Human Impact of Climate Change M M Hedger 886
Ozone Trends N R P Harris 892
Surface Temperature Trends PD Jones 898
Upper Atmospheric Change R G Roble 910
CONTENTS xxxv
GRAVITY WAVES See BUOYANCY AND BUOYANCY WAVES: Optical Observations;

BUOYANCY AND BUOYANCY WAVES: Theory; MIDDLE ATMOSPHERE: Gravity Waves
GUST FRONTS R Rotunno 915
Volume 3
HADLEY CIRCULATION IN James 919

HAIL AND HAILSTORMS C Knight, N Knight 924
HALOES See OPTICS, ATMOSPHERIC: Optical Phenomena
HAMILTONIAN DYNAMICS T G Shepherd 929
HUMIDITY VARIABLES J Curry 939
HURRICANES FD Marks 942
HYDRAULIC FLOW RB Smith 966
HYDROLOGY
Overview RC Bales 968
Ground and Surface Water S Ge 973
Modeling and Prediction Z Yu 980
Soil Moisture A Robock 987
I
ICE AGES (MILANKOVITCH THEORY) MF Loutre 995
INSTABILITY See also BAROCLINIC INSTABILITY
INSTABILITY
Inertial Instability J A Knox 1004
Symmetric Stability HB Bluestein 1014
Wave-CISK CS Bretherton 1019
IONOSPHERE M C Kelley 1022
ISENTROPIC ANALYSIS J R Holton 1030
ISOTOPES, STABLE CAM Brenninkmeijer 1033
J
JET STREAKS P Cunningham, D Keyser 1043
KATABATIC WINDS T R Parish 1057

KELVIN WAVES B Wang 1062
KELVIN-HELMHOLTZ INSTABILITY PG Drazin 1068
KINEMATICS D D Houghton 1072
xxxvi CONTENTS
LABORATORY GEOPHYSICAL FLUID DYNAMICS R L Pfeffer 1081

LABORATORY KINETICS D] Donaldson 1090
LAGRANGIAN DYNAMICS I Roulstone 1098
LAKE-EFFECT STORMS P] Sousounis 1104
LAND-ATMOSPHERE INTERACTIONS See also GLOBAL CHANGE: Biospheric Impacts and
Feedbacks
LAND-ATMOSPHERE INTERACTIONS
Overview RE Dickinson 1116
Canopy Processes PD Blanken 1121
Trace Gas Exchange ]N Cape, D Fowler 1130
LAND-SEA BREEZE R A Pielke Sr 1136
LARGE EDDY SIMULATION C-H Moeng, P Sullivan 1140
LEE VORTICES CC Epifanio 1150
LEE WAVES AND MOUNTAIN WAVES D R Durran 1161
LIDAR
Atmospheric Sounding Introduction PS Argall, R] Sica 1169
Backscatter C M R Platt 1176
DIAL E V Browell, S Ismail, W B Grant 1183
Doppler RM Hardesty 1194
Raman D N Whiteman 1202
Resonance CS Gardner 1212
LIGHTNING
Overview M B Baker 1216
Production of Nitric Oxide C Price 1223
M
MAGNETOSPHERE GK Parks 1229
MESOSCALE METEOROLOGY See also BOW ECHOS AND DERECHOS; COLD AIR
DAMMING; DRY LINE; FRONTS; LAND-SEA BREEZE; LEE WAVES AND MOUNTAIN
WAVES
MESOSCALE METEOROLOGY
Overview D Parker 1237
Cloud and Precipitation Bands RM Rauber, MK Ramamurthy 1243
Mesoscale Convective Systems A G Laing 1251
Models R Pielke 1261
MESOSPHERE
Metal Layers J M C Plane 1265
Polar Summer Mesopause M M Huaman, M C Kelley 1271
METEORS P Jenniskens 1279
METHANE E Dlugokencky, S Houweling 1286
MICROBURSTS RM Wakimoto 1294
MICROCLIMATE M W Rotach, P Calanca 1301
CONTENTS xxxvii
MIDDLE ATMOSPHERE
Gravity Waves DC Fritts 1308
Planetary Waves AK Smith 1314
Polar Vortex M R Schoeberl, PA Newman 1321
Quasi-Biennial Oscillation TJ Dunkerton 1328
Semiannual Oscillation K Hamilton 1336
Stratospheric Sudden Warmings A O’Neill 1342
Transport Circulation W A Norton 1353
Zonal Mean Climatology WJ Randel 1358
MIRAGES See OPTICS, ATMOSPHERIC: Optical Phenomena
MONSOON
Overview J Slingo 1365
Dynamical Theory PJ Webster, J Fasullo 1370
ENSO-Monsoon Interactions KM Lau 1386
Prediction V Thapliyal, M Rajeeuan 1391
MOUNTAIN METEOROLOGY RB Smith 1400
MOUNTAIN WAVES See LEE WAVES AND MOUNTAIN WAVES
Volume 4
N
NOCTILUCENT CLOUDS GE Thomas 1407
NORTH ATLANTIC AND ARCTIC OSCILLATION See CLIMATE VARIABILITY: North Atlantic
and Arctic Oscillation
NUMERICAL MODELS
Chemistry Models MP Chipperfield 1414
Methods J Thuburn 1423
0
OBSERVATION PLATFORMS
Balloons J-P Pommereau 1429
Buoys JM Hemsley 1438
Kites BB Balsley 1443
Rockets MF Larsen 1449
OBSERVATIONS FOR CHEMISTRY ( I N SITU)
Chemiluminescent Techniques C A Cantrell 1454
Gas Chromatography AC Lewis 1460
Ozone Sondes HGJ Smit 1469
Particles T Deshler 1476
Resonance Fluorescence LM Auallone 1484
Water Vapor Sondes E Weinstock, E Hintsa 1490
OBSERVATIONS FOR CHEMISTRY (REMOTE SENSING)
INFIR H Fischer, F Hase 1499
Lidar G Vaughan 1509
Microwave J Waters 1516
xxxviii CONTENTS
OCEAN CIRCULATION
General Processes N C Wells 1528
Surface -Wind-Driven Circulation P Bogden, C A Edwards 1540
Thermohaline Circulation J R Toggweiler, R M Key 1549
Water Types and Water Masses W ] Emery 1556
OPERATIONAL METEOROLOGY JV Cortinas Jr, W Blier 1567
OPTICS, ATMOSPHERIC
Airglow Instrumentation M Conde 1576
Optical Phenomena SD Gedzelman 1583
Optical Remote Sensing Instruments GG Shepherd 1595
OROGRAPHIC EFFECTS
Lee Cyclogenesis C Schur 1602
Mountain Waves and Stratospheric Chemistry KS Carslaw 1614
OZONE See also SATELLITE REMOTE SENSING: TOMS Ozone
OZONE
Ozone as a UV Filter ] E Frederick 1621
Ozone Depletion ] A Pyle 1627
Ozone Depletion Potentials D] Wuebbles 1636
Photochemistry of Ozone A R Ravishankara 1642
Role in Climate P Forster 1649
Surface Ozone (Human Health) M Lippmann 1655
Surface Ozone Effects on Vegetation M R Ashmore 1663
P
PALEOCLIMATOLOGY
Ice Cores E ] Steig 1673
Varves A Weinheimer, F Biondi 1680
PALMER DROUGHT INDEX L C Nkemdirim 1685
PARAMETERIZATION OF PHYSICAL PROCESSES
Clouds CJakob, M Miller 1692
Gravity Wave Fluxes M]Alexander 1699
Turbulence and Mixing A Beljaars 1705
PERMAFROST T E Osterkamp, C R Burn 1717
PLANETARY ATMOSPHERES See also EVOLUTION OF ATMOSPHERIC OXYGEN;
EVOLUTION OF EARTH’S ATMOSPHERE
PLANETARY ATMOSPHERES
Jupiter and the Outer Planets A P Showman 1730
Mars R M Haberle 1745
Venus P] Gierasch, Y L Yung 1755
POLAR LOWS I A Renfrew 1761
POLAR THERMOSPHERE A G Burns, T L Killeen 1768
PREDICTABILITY AND CHAOS LA Smith 1777
Q
QUASI-GEOSTROPHIC THEORY H C Davies, H Wernli 1787
CONTENTS xxxix
R
RADAR
Cloud Radar T Uttal 1795
Doppler Radar RJ Doviak, ME Frazier Doviak 1802
Incoherent Scatter Radar MP Sulzer 1812
Meteor Radar NJ Mitchell 1819
Volume 5
RADAR
MST and ST Radars and Wind Profilers RF Woodman 1825
Precipitation Radar SE Yuter 1833
Synthetic Aperture Radar (Land Surface Applications) RK Vincent 1851
RADIATION (SOLAR) Q Fu 1859
RADIATIVE TRANSFER
Absorption and Thermal Emission R Goody, H H u 1863
Cloud-Radiative Processes Q Fu 1871
Non-Local Thermodynamic Equilibrium F W Taylor, M L6pez-Puertas 1874
Scattering M Mishchenko, L Travis, A Lacis 1882
RADIOACTIVITY
Cosmogenic Radionuclides D Lal 1891
RADIOSONDES WF Dabberdt, R Shellhorn, H Cole, A Paukkunen, J Horhammer, VAntikainen 1900
RAINBOWS See OPTICS, ATMOSPHERIC: Optical Phenomena
REFLECTANCE AND ALBEDO, SURFACE J A Coakley 1914
ROSSBY WAVES PB Rhines 1923
SATELLITE REMOTE SENSING

Aerosol Measurements YJ Kaufman, D Tanre' 1941
Cloud Properties P Yang, BA Baum 1956
GPS Meteorology S B Healy 1965
Precipitation Guosheng Liu 1972
Surface Wind WT Liu 1979
Temperature Soundings A Dudhia 1985
TOMS Ozone RS Stolarski, R D McPeters 1999
Water Vapor J E Harries 2005
Wind, Middle Atmosphere PB Hays, W R Skinner 2012
SATELLITES
Orbits SQ Kidder 2024
Research (Atmospheric Science) M D King, D D Herring 2038
SEA ICE WF Weeks 2047
SEVERE STORMS C A Doswell III 2054
SNOW (SURFACE) M Sturm 2061
SOLAR TERRESTRIAL INTERACTIONS J D Haigh 2072
SOLAR WINDS ST Suess, BT Tsurutani 2078
XI CONTENTS
SOLITARY WAVES ] P Boyd 2086

SOOT P Chylek, SGJennings, R Pinnick 2093
SPECTRAL MODELS F Baer 2099
STANDARD ATMOSPHERE WW Vaughan 2107
STATIC STABILITY ] A Young 2114
STATIONARY WAVES (OROGRAPHIC AND THERMALLY FORCED) S Nigam, E DeWeauer 2121
STRATOSPHERE-TROPOSPHERE EXCHANGE
Global Aspects J R Holton 2137
Local Processes I F Lamarque, P Hess 2143
STRATOSPHERIC CHEMISTRY AND COMPOSITION
Overview ] A Pyle 2150
Halogen Sources, Anthropogenic A McCulloch, PM Midgley 2154
Halogen Sources, Natural ] H Butler 2162
Halogens D Toohey 2167
HO, TF Hanisco 2174
Hydrogen Budget ] E Harries 2180
Hydroxyl Radical DE Heard 2185
Reactive Nitrogen (NO, and NO,) Y Kondo 2193
STRATOSPHERIC OZONE RECOVERY DJ Hofmann 2202
STRATOSPHERIC WATER VAPOR ] E Harries 2208
SURFACE LAYER MEASUREMENTS OF TURBULENCE N O Jensen 2215
SYNOPTIC METEOROLOGY
Forecasting D Mansfield 2220
Weather Maps R Reynolds 2230
Volume 6
T
TELECONNECTIONS S Nigam 2243
THERMAL LOW R H Johnson 2269
THERMODYNAMICS
Moist (Unsaturated) Air ] A Curry 2274
Saturated Adiabatic Processes ] A Curry 2278
THERMOSPHERE R G Roble 2282
TORNADOES H B Bluestein 2290
TRACERS K A Boering 2297
TROPICAL METEOROLOGY See also EL NINO AND THE SOUTHERN OSCILLATION:
Observation; EL NIfiO AND THE SOUTHERN OSCILLATION: Theory; HURRICANES;
MONSOON: Overview; MONSOON: Dynamical Theory; MONSOON: ENS0 - Monsoon
Interactions; MONSOON: Prediction
TROPICAL METEOROLOGY
Overview and Theory R A Tomas, P] Webster 2306
Equatorial Waves M C Wheeler 2313
CONTENTS xli
Inter Tropical Convergence Zones D E Waliser 2325

Intraseasonal Oscillation (Madden-Julian Oscillation) R A Madden 2334
Tropical Climates S Hastenrath 2338
TROPOPAUSE M Dameris 2345
TROPOSPHERIC CHEMISTRY AND COMPOSITION
AerosolsA'articles J H Seinfeld 2349
Aliphatic Hydrocarbons J Rudolph 2355
Ammonia and Ammonium W A H Asman 2365
Aromatic Hydrocarbons I Barnes, K H Becker 2376
Biogenic Hydrocarbons (inc. Isoprene) A Guenther 2385
Carbon Monoxide CAM Brenninkmeijer, PC Novelli, U Schmidt, T Wetter 2389
Hl U Schmidt, T Wetter 2397
Hydroxyl Radical KC Clemitshaw 2403
Mercury J Munthe, J Sommar 2411
Oxidizing Capacity D H Ehhalt, A Wahner 2415
PAN H B Singh 2424
Sulfur Chemistry, Organic I Barnes 2429
VOCs: Overview NJ Blake, D R Blake 2438
TURBULENCE AND MIXING P Haynes 2446
TURBULENCE, TWO DIMENSIONAL P Bartello 245 1
TURBULENT DIFFUSION A Venkatram, S D u 2455
ULTRAVIOLET RADIATION K Stamnes 2467

ULTRAVIOLET, SURFACE R McKenzie, S Madronich 2474
VALLEY WINDS J Egger 2481

VOLCANOES
Composition of Emissions M T Coffey, W G Mankin 2490
Role in Climate A Robock 2494
VORTICITY ] R Holton 2500
W
WALKER CIRCULATION K-M Lau, S Yang 2505
WATER VAPOR See SATELLITE REMOTE SENSING: Water Vapor
WATERSPOUTS J H Golden 2510
WAVE MEAN-FLOW INTERACTION MJuckes 2526
WAVES See BUOYANCY AND BUOYANCY WAVES: Optical Observations; BUOYANCY AND
BUOYANCY WAVES: Theory; KELVIN WAVES; MIDDLE ATMOSPHERE: Gravity Waves;
ROSSBY WAVES; SOLITARY WAVES; STATIONARY WAVES (OROGRAPHIC AND
THERMALLY FORCED )
xlii CONTENTS
WEATHER FORECASTING See SYNOPTIC METEOROLOGY: Forecasting SYNOPTIC

METEOROLOGY: Weather Maps
WEATHER MODIFICATION
Inadvertant SA Changnon 2533
WEATHER PREDICTION
Adaptive Observations T M Hamill 2537
Data Assimilation AC Lorenc 2542
Ensemble Prediction R Buizza 2546
Regional Prediction Models B W Golding 2557
Seasonal and Interannual Weather Prediction M Hoerling, A Kumar 2562
Severe Weather Forecasting DJ Stensrud, H E Brooks, SJ Weiss 2568
WEATHER REGIMES AND MULTIPLE EQUILIBRIA F Molteni 2577
WILDFIRE WEATHER J Coen 2586
WIND CHILL M Bluestein, R Quayle 2597
WIND PROFILERS See RADAR: MST and ST radars and Wind Profilers
WORLD CLIMATE RESEARCH PROGRAM WL Gates 2602
Appendices
1. Physical Constants 2607
2. Units and their SI Equivalents 2609
3. Periodic Table of the Elements 2610
4. The Geologic Time Scale 2611
5. Abbreviations 2612
Index 2625
ACOUSTIC WAVES 1
K E Gilbert and H E Bass, National Center for Physical and the minimum pressure regions by dashed lines.
Acoustics, University of Mississippi, MS, USA The horizontal line perpendicular to the wavefronts is
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. called an acoustic 'ray'. Acoustic rays are a concise
way to indicate the travel paths taken by acoustic
wavefronts as they propagate through space. In Figure
Introduction lC, the regions of compression and rarefaction, often
called the 'acoustic' pressure, are shown moving to the
This article is concerned with acoustic waves in the right with a speed c, which for dry air is 33 1.6 m s at
atmosphere, Owing to space constraints, the discus- 0°C. For a compact representation of the pressure
sion is limited to audible acoustic waves (soundwaves). wave, one could, for example, omit the wavefronts
Hence, two important topics - ultrasound (above and simply show an acoustic ray moving to the right
audible) and infrasound (below audible) - are not with a speed c.
discussed. Further, in order to provide a more in-depth For sinusoidal pressure variations, a planar acoustic
discussion of the effect of the atmospheric boundary wave can be represented mathematically as p =
layer on sound waves, some traditional topics such as p o cos(kx - wt + e), where p o is the acoustic
ground effects, nonlinear effects, and noise control are pressure amplitude, and the entire argument of the
omitted. The interested reader should refer to the cosine is called the 'phase' of the wave. The angular
resources cited under Further Reading for information frequency, o,is 2nf, where f is the frequency in cycles
on aspects of acoustic waves not covered here. per second or, more commonly, in hertz (Hz). The
At the atomic level, the Earth's atmosphere is a wavenumber k is 2nli,, where A is the wavelength
collection of gas molecules, mainly nitrogen and shown in Figure 1C. Since if = c, the wavenumber
oxygen, bound to the planet by gravity. The micro- can also be written as w/c. The quantity B is called the
scopic properties of the atmosphere are thus described
by the kinetic theory of gases and quantum mechanics.
In contrast, at the macroscopic level, the atmosphere C-
can be regarded as a fluid, and, in principle, can be
described by the equations of fluid dynamics. Both
points of view, molecular and fluid dynamical, are
needed to fully understand the generation, propaga-
tion, and absorption of the disturbances in the
atmosphere that are familiar to us as acoustic waves
or 'sound'.
Unlike wave motion on a stretched string or ripples
on the surface of water, acoustic waves in the atmos-
phere have no direct visual representation. Conse-
quently, one must in general rely heavily on a
mathematical description. It is useful, nevertheless,
to try to connect the mathematical description of
sound with an intuitive, physical picture, even if the
picture is an approximate representation of reality.
Hence, for purposes of visualization, one can sche-
matically represent a planar acoustic wave as shown
in Figure 1. In Figure lA, regions of compression
(positive pressure relative to the ambient background (C)
pressure) and regions of rarefaction (negative pressure Figure I Visualization of a planar acoustic wave moving to the
relative to the ambient background pressure) are right at speed c. (A) Schematic representation of regions of
indicated schematically by the density of points. compression (denser points) and regions of rarefaction (less dense
Closely spaced points represent a compression, and points). (B) Wavefronts (regions of constant pressure); maximum
less closely spaced points represent a rarefaction. The and minimum pressure regions are represented, respectively, by
solid and dashed vertical lines. An acoustic ray is drawn perpen-
vertical lines in Figure 1B indicate regions of constant dicular to the wavefronts. (C) Pressure variation in space at
pressure that are called 'wavefronts'. The maximum two instants of time for a sinusoidal plane wave of the form
pressure regions are indicated by solid vertical lines p = po cos(& - w f + e).
2 ACOUSTIC WAVES
‘phase angle’ and gives the phase of the wave at x = 0

and t = 0.
The compressions and rarefactions in an acoustic
wave cause variations in density and temperature that
also propagate with the wave. For all three quantities -
pressure, temperature, and density - the acoustic
amplitude is extremely small relative to the ambient PR (X-Ct) PL ( x + c t )
background. For example, po might be O.1Pa or Figure 2 Solutions to the one-dimensional wave equation. The
approximately one millionth of the nominal atmos- functionpR(x - ct) is a right-going solution and the function
pheric pressure. +
p ~ ( x ct) is a left-going solution. The complete solution is the
superposition of the left- and right-going solutions.
The Acoustic Wave Equation Here r = dx2+ + y2 z2. By comparing the form of
As noted above, acoustic waves in the atmosphere can eqn [3] with that of eqn [l], one can deduce that the
be viewed as small disturbances on an ambient general solution of eqn [3]is given by eqn [4], in which
background fluid, just as water waves are seen as Y,,f is an arbitrary reference distance, generally taken
disturbances on a calm surface. For the extremely to be 1m in the MKS system.
small pressure perturbations typical of sound, the
equations of fluid dynamics can be linearized to arrive
at the ‘acoustic wave equation’, which is the conven-
tional mathematical description of acoustic pressure The quantities, (r,,i/r)PouT(r - ct) and (Y,,~/Y)PIN
waves. In one dimension, the acoustic wave equation is (Y+ ct) are out-going and in-going spherical waves,
given by eqn [l], where p is the acoustic pressure, x is respectively. Note that the spherical wave solution has
distance, and t is time. the same mathematical form as the plane-wave solu-
tion except that the amplitude falls off as l / r .
For a source far away from boundaries, the acoustic
pressure is given by an out-going wave having the same
shape in the time domain as the source function.
The general solution to eqn [ l ] is of the form For example, for a time-harmonic source, the acoustic
p(x,t ) = ~ R ( X- ct) + ~ +
L ( X ct), where ~ R ( X - ct) field is a traveling sinusoidal wave of the form
is a right-going wave and p ~ ( x+ ct) is a left-going
p = ( Y , , f / Y ) p o sin(kr - o t ) , where po is the pressure
wave. The right-going wave, for example, could be a amplitude at the reference distance. Pictorially, an out-
transmitted pulse, and the left-going wave could be going spherical wave can be represented as shown in
an echo. Continuous waves as well as pulses satisfy Figure 3, where pressure maxima and minima of the
the wave equation. For example, since c = u / k , the
sinusoidal pressure wave discussed above satisfies the
one-dimensional wave equation. Moreover, as indi-
cated in Figure 2, any function of (x- ct) or (x+ ct)
satisfies eqn [l]. Further, the perturbations in density
f
and temperature associated with an acoustic pressure
wave satisfy the same wave equation as the acoustic
pressure except that, instead of pressure, the variable is
density or temperature, respectively.
The three-dimensional form of eqn [l]is eqn [ 2 ] ,
where (x,y, z ) are Cartesian coordinates.
For a symmetrical source, such as a small explosion

high above the ground, the three-dimensional wave
equation has spherical symmetry and can be written as
eqn ~31. Figure 3 Schematic representation of an out-going spherical
wave. The circles are the wavefronts and the straight lines are
a2 1 a2 acoustic rays. The solid and dashed circles denote, respectively,
-(Ye) = - -(@I 131 wavefronts
ar2 c2 at2 for maximum and minimum pressure.
ACOUSTIC WAVES 3
wavefronts are represented, respectively, by solid lines To express in decibels the variation of the rms
and dashed lines. The radial lines perpendicular to pressure, prms(r),with distance, it is conventional to
wavefront are acoustic rays. define the SPL at r = rrefas the ‘source level’ (SL) as in
eqn [6] and to define the decibel decrease in acoustic
pressure with distance as the ‘transmission loss’ (TL)
Sound Pressure Levels and Decibels as in eqn [7].
Acoustic pressure amplitudes encountered in practice
typically vary over several orders of magnitude.
Consequently, it has become conventional to use a
logarithmic scale to describe the amplitudes. For
continuous waves, the amplitude of interest is the
root-mean-square pressure amplitude, prms,and is 171
referenced to some standard reference pressure pref.
For pulses, some ‘peak’ pressure is often chosen. In In these equations, as noted previously, the reference
either case, a logarithmic amplitude measure called the distance, rref,is l m in the MKS system. (Note that
‘sound pressure level’ (SPL)is commonly used, defined since pressure usually decreases with distance, trans-
mission loss is usually positive.) Using the above
by eqn PI.
definitions for source level and transmission loss, one
can write the sound pressure level at a distance r from
the source as SPL = SL -TL. For example, in MKS
units, the rms pressure amplitude for a spherically
The pressure of interest is denoted here as the rms spreading wave can be written as erms(Y) = ~ I / Y ,
pressure, but could be any pressure, depending on the where p 1 is the rms pressure at 1 m . Thus, for a
application. Although SPL is actually a dimensionless spherically spreading wave, the source level is
quantity, one refers to the ‘units’ as decibels (dB), 20 loglo (pl/pref)and the transmission loss is simply
referenced to a reference pressure, pref.In atmospheric 20 log10 (Y).
acoustics, the reference level is usually chosen to be In general, the transmission loss is not a simple
2 x 10 - Pa or 20 pPa, which is the approximate function and must be computed numerically. With
threshold of hearing. Note that with the above numerical computations, it is often useful, for plotting
conventions, the SPL for 20 pPa is 0 dB. purposes, to subtract the transmission loss due to
The frequency range for audible sound for the spherical spreading, that is, to subtract 20 loglo (Y).
human ear is from approximately 20 Hz to approxi- Such a convention is equivalent to giving the sound
mately 20kHz. Typical sound pressure levels for pressure level relative to a spherically spreading wave,
sounds in the audible range are given in Table 1. and hence is given the name ‘relative sound pressure
In addition to being used as a measure for absolute level’. Thus, by definition, the relative sound pressure
pressure, decibels are also used to describe relative level for a spherically spreading wave is OdB. Ex-
changes in pressure. For example, if a pressure pressed as a relative sound pressure level, an SPL above
amplitude decreases with distance by a factor of 10, or below that for spherical spreading will be, respec-
it is conventional to say that, over the distance, the tively, greater than or less than zero.
acoustic pressure has decreased by 20 loglo( 10) =
20dB. If one pressure amplitude were 100 times The Speed of Sound in the Atmosphere
greater than another, one would say the first pressure
was 40 dB greater than the second pressure. To a good approximation, the atmosphere can be
treated as an ideal gas, and the acoustic pressure
variations in it can be treated as adiabatic; that is, there
Table 1 Representative list of audible sound pressure levels
is no heat flow from the higher pressure (hotter)
Sound SPL regions to the lower pressure (cooler) regions. For an
ideal gas and adiabatic compression (or rarefaction),
Threshold of hearing
Leaves rustling
0 dB
20 dB
the speed of sound is given by c = m, where y =
Quiet conversation 40 dB 1.40 is the ratio of the constant-volume specific heat
Normal conversation 60 dB for air, C V , to the constant-pressure specific heat, c p .
Average street traffic 80 dB The quantity, R = 286.69 J kg-’ K-l, is the gas con-
Diesel truck (at 10 m) 90 dB stant for dry air, and T i s the absolute temperature (K).
Jet take-off (at 10 m) 120dB
Threshold of pain
With an ideal gas model, the theoretical value for c at
140 dB
0°C (273.16K) is 331.lmsW1,which is in excellent
4 ACOUSTIC WAVES
agreement with the experimental value of 33 1.6 m s - the expanded regions are slightly cooler, a small
given earlier. For values of T not far from O'C, the amount of heat flows from the compressions to the
square root expression for the speed of sound can be rarefactions. The conduction of heat converts the
expanded linearly and written approximately as organized motion associated with the sound wave into
c = (331 + 0.6Tc) m sP1, where TCis the temperature random thermal motion of the gas molecules. Because
in degrees Celsius. Thus, for an increase in tempera- the heat flow lowers the temperature of the compres-
ture of one degree Celsius, the speed of sound increases sions and raises the temperature of the rarefactions,
by 0.6ms-I. both the pressure maxima and minima are reduced.
The reduction manifests itself as a decay of the
acoustic wave with distance. The component of
Absorption of Sound in the absorption due to thermal conduction is given by
Atmosphere eqn [9], where IC is the coefficient of thermal conduc-
In addition to the decrease in pressure amplitude of an tivity in J ( k g m o l ) - ' K - ' k g r n - ' ~ - ~ .
acoustic wave due to propagation effects such as
'geometrical' spreading (e. g., spherical spreading), the 191
amplitude is also reduced by atmospheric absorption.
A sound wave propagating through 'clean' air (no
solid particles) is attenuated owing to two basic In addition to energy loss due to classical mecha-
mechanisms: nisms (viscosity and heat conduction), energy can also
be lost via quantum-mechanical 'relaxation' processes
0 classical losses due to momentum transfer across a involving the internal degrees of freedom (rotation and
velocity gradient (viscosity) and heat flow across a vibration) of oxygen and nitrogen molecules. The
temperature gradient transfer of translational energy to internal degrees of
Quantum-mechanical losses due to relaxation pro- freedom and back takes place through an extended
cesses, mainly relaxation of rotational and vibra- sequence of molecular collisions, so there is a time
tional states in nitrogen and oxygen molecules. delay associated with the energy transfer. Because of
the time delay, relaxation processes cause energy to be
For both mechanisms, the effects of absorption can be
lost from the organized translational motion that
represented by an absorption coefficient, a, which has
constitutes the acoustic wave. As a result, just as with
units of m-'. The absorption coefficient enters via an
the classical mechanisms, the pressure amplitude of
exponential, so that the pressure is given by
the wave decreases as the wave propagates.
p = poe-cts, where po is the unattenuated pressure For any particular relaxation process, the associated
amplitude and s is the distance the wave has traveled.
absorption coefficient has the general form of eqn [lo].
To indicate the choice of the Napierian base, e,
the attenuation coefficient is, by convention, said
to have units of nepersm-'. The corresponding
attenuation coefficient, a, for decibels (base 10) is
a = 20a loglo (e) = 8 . 6 8 6 4 and, by convention, has
units of dB m - '. In eqn [lo],5, is the relaxation strength (in nepers), cis
Experimental and theoretical studies indicate that the sound speed, f is the frequency, and f r is called the
the total absorption coefficient can be represented as a 'relaxation' frequency. The relaxation frequency is the
sum of absorption coefficients, with each distinct frequency for maximum absorption and is roughly
physical process having its own unique coefficient. the reciprocal of the characteristic time delay for the
That is, the total absorption coefficient, ET, can be transfer between kinetic energy and internal energy of
written as ET = E, ai, where is the absorption the gas molecules.
coefficient associated with a particular mechanism. For air, there are three important relaxation pro-
For example, the component of sound absorption due cesses: (1)0 2 vibration, (2) N2 vibration, and ( 3 ) N2
to viscosity is given by eqn [8], in which o is the rotation. The relaxation frequency for N2 rotational
angular frequency, c is the speed of sound, po is the relaxation is very high, so that, below lOMHz, the
density of air, and p is the coefficient of viscosity of air. denominator in eqn [lo] is approximately unity. Thus,
N2 rotational relaxation varies as f 2 and can be
combined with the classical absorption coefficient. If
one denotes the classical-plus-rotational absorption
coefficient (i.e., the coefficient for viscosity, heat flow,
Since the compressed regions in an acoustic wave and N2 rotation) as ac, the absorption coefficient for
are slightly hotter than the ambient temperature, and 0 2 vibration as E O , and the absorption coefficient for
ACOUSTIC WAVES 5
10-0 point source in an atmosphere with a constant

temperature, and hence a constant sound speed.
lo-’ With a constant sound speed, acoustic ray trajectories
are straight lines. In reality, however, the atmospheric
10-2 temperature is never constant in space or time.
c
1 Consequently, the speed of sound is not constant but
E
2 10-3 varies spatially and temporally. In a typical daytime
a,
situation, the temporally averaged temperature is
- Q
E 10-4
independent of range but decreases with height (‘lapse’
.-0
a condition). Thus, on average, the sound speed de-
E 10-5 creases with height, and sound rays curve upward, as
8 shown in Figure 5 . The ray paths shown are for a sound
C
speed that decreases linearly with height. (For the
’= 10-6
0
simple case of linear variation, the ray paths are arcs of
$ circles.)
2 10-7 For a general sound speed variation in a stratified
atmosphere (i.e., no horizontal variation), ray paths
10-8 are governed mathematically by Snell’s law, which
states that the quantity c ( z ) / cos O(z) is invariant,
10-9 where at a height z, the quantities, c ( z ) and cos Q(z)
10’ 102 103 104 105 are, respectively, the sound speed and the cosine of the
Frequency (Hz) angle of a ray with respect to horizontal. Thus, Snell’s
law says that, if c ( z ) decreases with height, cos O(z)
Figure 4 Components and general behavior of the total absorp- must increase, so that the ray bends upward. In
tion coefficient for air. The contributions to the total absorption ( T )
are the classical plus N2 rotation (C), the 0 2 vibration (0)and the
general, acoustic rays bend toward regions of lower
N2 vibration (N). (Reproducedwith permission from Bass (1991), sound speed and away from regions of higher sound
VCH Publishers, Inc., Weinheim, Germany.) speed. The bending of acoustic rays is given the name
‘refraction’.
Nz vibration as EN, then the total absorption coeffi- The physical basis for refraction can be understood
cient, MT for air can be written as eqn [Ill. using the situation in Figure 5 . Consider a small
section of wavefront associated with a ray that leaves
+ +
?T = ?C EO ?N 1111 the source nearly parallel to the ground. For a small
enough section, the wavefront is nearly planar and
Figure 4 shows the total absorption coefficient, UT, nearly vertical. Since the sound speed decreases with
together with the components, M C , M O , and M N . Note height, the lower portion of the wavefront travels
that below about 10 000 Hz the absorption is domi- faster than the upper portion, causing the wavefront to
nated by vibrational relaxation. Further, note that, turn upward. In terms of rays, we would say that the
above about 1000 Hz, atmospheric absorption is ray is refracted upward due to the decrease in the
significant for propagation distances of a kilometer sound speed with height.
or more, which accounts for the lack of long-range Refraction of acoustic waves is caused by spatially
propagation of high-frequency sound. varying wind as well as by spatially varying temper-
ature. The effect of the wind on acoustic waves can be
Refraction of Sound in the Atmosphere accounted for approximately by defining an ‘effective’
sound speed, ce, which is the previously defined
The spherical wavefronts and associated rays shown in ‘adiabatic’ sound speed, c, = m, plus the compo-
Figure 3 represent acoustic waves radiating from a nent of the vector wind in the direction of propagation.
Figure 5 Acoustic rays and shadow zones for an upward-refracting daytime atmosphere. For clarity, rays that are reflected off the
ground are not shown. (Reproduced with permission from Bass (1 991).(VCH Publishers, Inc., Weinheim, Germany.)
6 ACOUSTIC WAVES
Figure 6 Acoustic rays and shadow zone for an atmosphere that is upward-refractingin the upwind direction and downward-refractingin
the downwind direction. For clarity, rays that are reflected off the ground are not shown. (Reproduced with permission from Bass (1991),
VCH Publishers, Inc., Weinheim, Germany.)
For example, let 1 be a unit vector pointing from the than the sound pressure level one would expect with
source to a receiver. Then, if the vector wind is denoted spherical spreading alone. Because of upward refrac-
as v ( u x ,u y ,u z ) ,the effective sound speed is given by tion, daytime sound pressure levels for a near-ground
c, = ca + i . v , where 1 -v is the component of the source fall off dramatically with horizontal distance as
vector wind in the direction of sound propagation. In one enters the shadow region, which, for strong
general, near the ground, the horizontal wind speed upward refraction, can be within 100-200m of the
increases with increasing height. For upwind propa- source.
gation of sound, therefore, the horizontal wind pro- At night, in contrast to the daytime situation,
gressively reduces the effective sound speed with ground-to-ground propagation is very good. Owing
increasing height. For downwind propagation, the to radiative cooling of the ground, both the near-
effect is reversed. That is, the horizontal wind ground air temperature and the sound speed are lower
progressively increases the effective sound speed than at higher altitudes (an 'inversion' condition). As a
with increasing height. In the daytime, for example, result, acoustic rays launched near to horizontal (less
where the temperature and adiabatic sound speed than about 10" with respect to horizontal) are bent
decrease with height, upwind propagation adds to the downward, causing sound to be trapped in a 'sound
upward refraction already present. Downwind, if the duct' near the ground. Rays launched at steeper angles
wind speed gradient is sufficiently large, the horizontal escape the duct and continue upward (see Figure 7).
wind can overcome the upward refraction due to the With strong trapping and small ground-bounce loss
daytime temperature profile and lead to downward (e.g., over water) the acoustic field in the near-surface
refraction. The ray paths for upwind and downwind sound duct undergoes approximately cylindrical
propagation are illustrated in Figure 6 . For propaga- spreading (l/fi)instead of spherical spreading (l/r).
tion directly across the wind, there is little effect due to As a consequence of daytime upward refraction and
the wind, but upward refraction persists because of the nighttime downward refraction, noise sources that are
decreasing temperature with height. not generally heard during the day can often be easily
As indicated in Figures 5 and 6 , for up- heard at long distances (e.g., several kilometers) at
ward refraction, there is a region, called an acoustic night. The long-range propagation of acoustic waves
'shadow', where no acoustic rays can penetrate. In the at night makes noise control much more difficult than
shadow region, the acoustic levels are much lower during the day.
Figure 7 Acoustic rays for a downward-refracting nighttime atmosphere. Rays launched at small angles with respect to the horizontal
(less than about 10") are trapped in a 'sound duct' near the ground and can propagate to ranges of several kilometers. For steeper launch
angles, the rays escape the duct and continue upward.
ACOUSTIC WAVES 7
Diffraction of Acoustic Waves wave equation for a 500Hz point source in an

in the Atmosphere upward-refracting atmosphere. The color plot in the
figure, which is for the relative sound pressure level as
As discussed above, one can approximately represent a function of range and height, shows the effects of
an acoustic field in terms of wavefronts whose both diffraction and refraction. The edge of the
propagation directions (i.e., rays) are governed by shadow boundary in Figure 8 would become more
refraction. Such a representation is useful visually and diffuse at lower frequencies (longer wavelengths),
can be valid computationally when the acoustic until finally, at very low frequencies, the shadow
wavelengths are much smaller than the smallest boundary would not be discernible at the ranges
sound-speed structure in the atmosphere. The main shown.
effect left out in the so-called ‘ray theory’ of sound When the atmosphere is downward-refracting (e.g.,
propagation is the wave phenomenon known as at night), the presence of diffraction again causes a
‘diffraction’. Diffraction is responsible for the well- ‘blurring’ of the features of the acoustic field, just as
known ability of sound to ‘bend’ around corners and with upward refraction. There are no sharp disconti-
obstacles. In outdoor sound propagation, diffraction nuities in the structure of the acoustic field. Rather,
fills in gaps in the acoustic field that would be present because of the finite wavelengths in the acoustic field,
in a purely ray-based representation. Full-wave solu- the changes in the field are continuous and smooth, as
tions to the wave equation (usually numerical) auto- shown in Figure 9, which is also for a 500Hz point
matically include both diffraction and refraction. source. In general, the longer the acoustic wavelengths
Because of diffraction, every acoustic field has an (i.e., the lower the frequency), the smoother the
intrinsic smallest possible scale length that is roughly a features of the acoustic field.
quarter of the smallest wavelength present in the field. In addition to limiting the sharpness of the acoustic
Owing to the scale limitation, there can be no sharp field, diffraction is responsible for the scattering of
edges in the acoustic field. For example, instead of the acoustic waves from the complex small-scale structure
sharp shadow boundary obtained with rays (Figures5 of the real atmosphere. In a realistic model of the
and 6), a smooth, diffuse boundary is obtained when atmosphere, the instantaneous temperature and vec-
diffraction is included. Such a situation is illustrated in tor wind fields are not smooth but are highly irregular,
Figure 8, which shows a numerical solution of the containing eddies of all sizes. The eddy sizes of most
Figure8 Numericalsolution for the sound field in an upward-refractingdaytime atmosphere without turbulence. Owing to diff raction, the
edge of the shadow boundary is diffuse. The color plot shows the relative sound pressure level as afunction of range and height. (Note that
the vertical scale is much less than the horizontal scale, so that the actual propagation angles with respect to the horizontal are much
smaller than shown.)
8 ACOUSTIC WAVES
Figure 9 Numerical solution for the sound field in a downward-refracting nighttime atmosphere without turbulence. Note that downward
refraction ‘traps’ sound in the near-ground duct. As in Figure 8, the features of the acoustic field are blurred owing to diffraction. The color
plot shows the relative sound pressure level as a function of range and height. (Note that the vertical scale is much less than the horizontal
scale, so that the actual propagation angles with respect to the horizontal are much smaller than shown.)
concern for audible sound are in the region called the inertial-subrange eddies. As a consequence of diffrac-
‘inertial subrange’, which typically begins at a few tens tion, new wavefronts emanate from every eddy, with
of meters and goes down to a few millimeters. In the the strongest scattering occurring in the near-forward
inertial subrange, the eddy structure is governed by direction. The diffracted acoustic waves that are
the well-known Kolmogorov spectrum. As a conse- scattered downward act to fill in the shadow region.
quence, at any instant of time, the small-scale spatial An example of this phenomenon is shown in Figure 11,
structure of the sound speed field, which depends on which was computed numerically using realistic rep-
temperature and vector wind, can also be described by resentations for .? and 6c. With a realistic model for the
a Kolmogorov spectrum. In the daytime turbulent sound speed, the predicted mean near-ground levels
boundary layer, for example, the sound speed can be ( -20 dB to - 30 dB relative to spherical spreading) in
approximated as a time-independent mean sound the shadow region (0.2-2 km) are in good agreement
speed, .?(z), that varies only with height, plus a with observation. The relative sound pressure level for
fluctuating part, 6 c ( x ,y, z, t ) , that varies with time, a longer-range interval is shown in Figure 12. It is
horizontal distance, and height. Hence the total sound apparent from Figure 12 that, even with scattering into
speed, c(x,y, z, t ) ,can be represented as in eqn [12]. the shadow region, daytime levels near the ground are
very low at ranges beyond a few kilometers.
It can be observed in Figures 11 and 12 that the
effects of turbulence are most apparent in the shadow
The quantity .?(z) approximates the slow, large-scale region, where the sound levels would be extremely
variations in the sound speed profile and 6c(x,y, z, t ) small in the absence of turbulence. Above the daytime
describes the rapid, smaller-scale fluctuations. shadow region, in the ‘insonified’region, the levels are
As noted earlier, the quantity & ( x , y , z , t ) follows much higher, so that the effect of scattering from
the same Kolmogorov statistics as do the temperature turbulence is less dramatic, though the effect increases
and wind fluctuations. At a particular instant of time, with increasing distance from the source. Similarly, for
an approximate ‘snapshot’ of the sound-speed fluctu- nighttime propagation, in the near-ground acoustic
ation field, & ( x , y, z, t ) ,can be synthesized by adding duct, where the mean levels are high, the effect of
together, with random phase, the wavenumber com- scattering from turbulence is not as dramatic as in the
ponents for a Kolmogorov spectrum. The result of shadow region above the duct (see Figure 13).Further,
such a synthesis is shown in two dimensions in Figure the nocturnal boundary layer, being more stable,
10. Because audible sound has wavelengths compa- intrinsically has weaker turbulence than the daytime
rable in size to small-scale atmospheric structure, it is boundary layer. As a consequence, mean sound levels
scattered in all directions as it propagates through at night near the ground are not affected by turbulence
ACOUSTIC WAVES 9
Figure 10 A two-dimensional 'snapshot' of small-scale turbulent fluctuations in the sound-speedfield synthesized using a Kolmogorov
spectrum and random Fourier components. The fluctuation magnitudes are typical of those created by turbulence in the daytime.
nearly as much as the near-ground daytime levels. It Acoustic Remote Sensing of the
should be noted, however, that large-scale nocturnal Atmosphere
phenomena such as flow down a slope and gravity
waves can have a significant effect on sound levels at As illustrated above, the sensitivity of acoustic waves
night. The effects of such large-scale dynamical to atmospheric wind and temperature variations
features on sound propagation are the subject of makes accurate prediction of ground-to-ground sound
current research in atmospheric acoustics. propagation a challenging problem. Conversely,
Figure 11 Same as Figure8 except that the small-scale turbulence shown in Figure 10 is included. Note that the effects of turbulence
are most apparent in the shadow region. Scattering of acoustic waves from turbulence 'fills in' the shadow region so the levels there are
much higher than the no-turbulence case is shown in Figure 8.
10 ACOUSTIC WAVES
Figure 12 Same as Figure 11 except that the range extends to 10 km. Note that, even with scattering into the shadow region, daytime
sound pressure levels near the ground are very low at ranges beyond a few kilometers.
however, that same sensitivity makes sound a remark- cated, is common and is known as a ‘monostatic’
ably valuable probe for remotely sensing the complex sodar. (A less common geometry has the receiver
features of the atmospheric boundary layer. The separated horizontally from the transmitter and is
most widely used acoustic tool for atmospheric called a ‘bistatic’ sodar.) When the upward-going pulse
sensing is a pulse-echo probe called a ‘sodar’, after encounters wind and temperature inhomogeneities
the more familiar ‘radar’, which is an electromagnetic produced by turbulence, faint scattered waves are
pulse-echo device. (Note: Sometimes the name ‘echo- created within the air itself. With monostatic sodar, the
sonde’ is also used, but that designation is less common part of the acoustic wave scattered back toward the
than ‘sodar’.) ground, the echo, is detected using the same trans-
The first sodars, which appeared in the early 1970s, ducers that produced the probe beam.
emitted an acoustic pulse in a single vertically pointing Early sodars were used primarily as instruments for
beam as shown in Figure 14. The sodar geometry detecting turbulence. The time delay between the
shown, with the acoustic source and receiver collo- emitted pulse and its echo determined the height of the
Figure 13 Same as Figure 9 except that small-scale turbulence typical of a nighttime atmosphere has been included. Since sound
levels are relatively high in the near-ground nocturnal sound duct, the effects of small-scale turbulence are less evident than during the
daytime.
ACOUSTIC WAVES 11
Figure 14 Geometry for a monostatic sodar. An array of

transducers projects a burst of acoustic waves vertically in a
beam. Turbulence-generated inhomogeneities in wind and tem-
perature scatter sound back toward the transmitting transducers,
which act as a directional receiver for the faint echoes received on
the ground. Figure 15 Record from a sodar taken over a diurnal cycle. The
vertical scale is 0-500 m, and time increases from left to right. The
vertical white lines are hour markers. (A) Unstable daytime
boundary layer with thermal plumes generated by solar heating
turbulence (one-half the time delay times the average of theground. (B) Decreasingdaytime boundary layer followed bya
speed of sound), while the strength of the echo was a growing stable nocturnal boundary layer showing evidence of initial
measure of the turbulence intensity. The evolving gravity wave activity. (C) Stable nocturnal boundary layer after
structure of the atmospheric boundary layer could be midnight with fully developed internal wave activity. (Reproduced
with permission from Atmospheric acoustics, Encyclopedia
‘mapped’ by plotting the delay time and echo strength of Applied Physics, vol. 2, VCH Publishers, Inc., Weinheim,
on a vertically moving strip of paper. For example, the Germany, 1991.)
horizontal distance on the strip could be proportional
to the time delay of the echo, and the darkness could be
proportional to the intensity of the echo. With many turbulent boundary layer descending in the late
repeated pulses, the evolution of the boundary layer afternoon and evening as solar heating of the ground
could be followed visually. With its debut in the 1970s, diminishes. The undulations in the latter part of the
the sodar immediately provided important insights record indicate the onset of internal gravity waves.
into the spatial structure and temporal evolution of the Panel (C)shows fully developed internal wave activity
atmospheric boundary layer. A typical sodar record is after midnight.
shown in Figure 15. The figure has time moving from In addition to visual displays of boundary layer
left to right and shows the evolution of boundary layer structure and dynamics, modern sodars can provide
structures over a typical diurnal period. The vertical quantitative measures of wind and temperature. The
scale in the figure is 0-500m, and the thin white so-called ‘Doppler sodar’, for example, which uses
vertical streaks are hour markers. Panel (A) shows a two slant beams in addition to the usual vertically
typical daytime record of thermal plumes carried pointing beam, can map vector wind versus height. A
through the vertical sodar beam. Panel (B) shows the typical geometry would have a vertical beam, together
12 ACOUSTIC WAVES
with slant beams pointing north and east, respectively, See also
at 60" above horizontal. Using the Doppler shift Boundary Layers: Convective Boundary Layer: Neutrally
in the echoes (up-shift for winds moving toward Stratified Boundary Layer; Stably Stratified Boundary Lay-
the receiver and down-shift for winds moving away er; Surface Layer. Buoyancy and Buoyancy Waves:
from the receiver), the three vector components Optical Observations; Theory. Diurnal Cycle. Parame-
(up-down, east-west, and north-south), can be meas- terization of Physical Processes: Turbulence and
ured as a function of height. Such advances as the Mixing.
Doppler sodar are due, in large part, to the vast
increase during the past 20 years in the computing
power available with small computers. In addition to Further Reading
providing greatly increased signal processing power, Bass HE (1991) Atmospheric acoustics. In: Trigg GL (ed.)
small, powerful computers have also made remote Encyclopedia of Applied Physics, vol. 2, pp. 145-179.
sensing instruments like the Doppler sodar sufficiently New York: Wiley-VCH.
'user friendly' that nonexperts can operate them Brekhovskikh LM (1980) Waves in Layered Media (trans-
successfully. lated by RT Beyer). New York: Academic Press.
A second important advance in acoustic remote Crocker M (1997) Introduction. In: Crocker M (ed.)
Handbook of Acoustics. New York: Wiley.
sensing is the 'Radio Acoustic Sounding System', or
Elmore WC and Heald MA (1969) Physics of Waves. New
RASS, which can provide accurate temperature pro- York: McGraw-Hill.
files as a function of height. A RASS uses a single Kinsler LE, Fry AR, Coppens AB and Sanders JV (2000)
vertically pointing sodar beam together with two Fundamentals of Acoustics. New York: Wiley.
radar beams that converge in the air column over the Morse PM (1981) Vibration and Sound. New York: Acous-
sodar. The radar is used in a bistatic geometry with the tical Society of America/American Institute of Physics.
transmitter on one side of the sodar and the receiver on Morse PM and Ingard KU (1968) Theoretical Acoustics.
the other side. Using coherent radar backscatter from New York: McGraw-Hill.
the upward-going acoustic beam, the RASS measures Neff WD and Coulter RL (1986) Acoustic remote sensing.
the speed of the acoustic beam as it propagates In: Lenschow DW (ed.)Probing the Atmospheric Bound-
upward. After making corrections for the vertical ary Layer. Boston, MA: American Meteorological
wind, one can estimate the adiabatic sound speed (a Society.
Ostashev VE (1977) Acoustics in Moving Inhomogeneous
function of temperature only) as a function of height, Media. London: E&FN Spon.
which then yields the temperature as a function of Pierce AD (1989)Acoustics: An Introduction to Its Physical
height. Extensive comparisons between RASS meas- Principles and Applications. Woodbury, NY: Acoustical
urements and in situ measurements have shown that a Society of America.
RASS provides reliable estimates of temperature at Piercy JE, Embleton TFW and Sutherland LC (1977)Review
heights from a few hundred meters to up to several of noise propagation in the atmosphere. Journal of the
kilometers. Acoustical Society of America 16: 1403-1418.
Salomons EM (2001) Computational Atmospheric Acous-
tics. Dordrecht: Kluwer Academic.
Acknowledgements Stull RB (1993) An Introduction to Boundary Layer
Meteorology. Boston, MA: Kluwer Academic.
The author would like to thank Mr RC Clark for the Sutherland LC and Daigle GA (1997) Atmospheric sound
acoustic propagation calculations and Ms EA Furr for propagation. In: Crocker M (ed.) Handbook of Acous-
expert technical editing assistance. tics. New York: Wiley.
AEOLIAN TRANSPORT
See DUST
ACOUSTIC WAVES 1
K E Gilbert and H E Bass, National Center for Physical and the minimum pressure regions by dashed lines.
Acoustics, University of Mississippi, MS, USA The horizontal line perpendicular to the wavefronts is
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. called an acoustic 'ray'. Acoustic rays are a concise
way to indicate the travel paths taken by acoustic
wavefronts as they propagate through space. In Figure
Introduction lC, the regions of compression and rarefaction, often
called the 'acoustic' pressure, are shown moving to the
This article is concerned with acoustic waves in the right with a speed c, which for dry air is 33 1.6 m s at
atmosphere, Owing to space constraints, the discus- 0°C. For a compact representation of the pressure
sion is limited to audible acoustic waves (soundwaves). wave, one could, for example, omit the wavefronts
Hence, two important topics - ultrasound (above and simply show an acoustic ray moving to the right
audible) and infrasound (below audible) - are not with a speed c.
discussed. Further, in order to provide a more in-depth For sinusoidal pressure variations, a planar acoustic
discussion of the effect of the atmospheric boundary wave can be represented mathematically as p =
layer on sound waves, some traditional topics such as p o cos(kx - wt + e), where p o is the acoustic
ground effects, nonlinear effects, and noise control are pressure amplitude, and the entire argument of the
omitted. The interested reader should refer to the cosine is called the 'phase' of the wave. The angular
resources cited under Further Reading for information frequency, o,is 2nf, where f is the frequency in cycles
on aspects of acoustic waves not covered here. per second or, more commonly, in hertz (Hz). The
At the atomic level, the Earth's atmosphere is a wavenumber k is 2nli,, where A is the wavelength
collection of gas molecules, mainly nitrogen and shown in Figure 1C. Since if = c, the wavenumber
oxygen, bound to the planet by gravity. The micro- can also be written as w/c. The quantity B is called the
scopic properties of the atmosphere are thus described
by the kinetic theory of gases and quantum mechanics.
In contrast, at the macroscopic level, the atmosphere C-
can be regarded as a fluid, and, in principle, can be
described by the equations of fluid dynamics. Both
points of view, molecular and fluid dynamical, are
needed to fully understand the generation, propaga-
tion, and absorption of the disturbances in the
atmosphere that are familiar to us as acoustic waves
or 'sound'.
Unlike wave motion on a stretched string or ripples
on the surface of water, acoustic waves in the atmos-
phere have no direct visual representation. Conse-
quently, one must in general rely heavily on a
mathematical description. It is useful, nevertheless,
to try to connect the mathematical description of
sound with an intuitive, physical picture, even if the
picture is an approximate representation of reality.
Hence, for purposes of visualization, one can sche-
matically represent a planar acoustic wave as shown
in Figure 1. In Figure lA, regions of compression
(positive pressure relative to the ambient background (C)
pressure) and regions of rarefaction (negative pressure Figure I Visualization of a planar acoustic wave moving to the
relative to the ambient background pressure) are right at speed c. (A) Schematic representation of regions of
indicated schematically by the density of points. compression (denser points) and regions of rarefaction (less dense
Closely spaced points represent a compression, and points). (B) Wavefronts (regions of constant pressure); maximum
less closely spaced points represent a rarefaction. The and minimum pressure regions are represented, respectively, by
solid and dashed vertical lines. An acoustic ray is drawn perpen-
vertical lines in Figure 1B indicate regions of constant dicular to the wavefronts. (C) Pressure variation in space at
pressure that are called 'wavefronts'. The maximum two instants of time for a sinusoidal plane wave of the form
pressure regions are indicated by solid vertical lines p = po cos(& - w f + e).
2 ACOUSTIC WAVES
‘phase angle’ and gives the phase of the wave at x = 0

and t = 0.
The compressions and rarefactions in an acoustic
wave cause variations in density and temperature that
also propagate with the wave. For all three quantities -
pressure, temperature, and density - the acoustic
amplitude is extremely small relative to the ambient PR (X-Ct) PL ( x + c t )
background. For example, po might be O.1Pa or Figure 2 Solutions to the one-dimensional wave equation. The
approximately one millionth of the nominal atmos- functionpR(x - ct) is a right-going solution and the function
pheric pressure. +
p ~ ( x ct) is a left-going solution. The complete solution is the
superposition of the left- and right-going solutions.
The Acoustic Wave Equation Here r = dx2+ + y2 z2. By comparing the form of
As noted above, acoustic waves in the atmosphere can eqn [3] with that of eqn [l], one can deduce that the
be viewed as small disturbances on an ambient general solution of eqn [3]is given by eqn [4], in which
background fluid, just as water waves are seen as Y,,f is an arbitrary reference distance, generally taken
disturbances on a calm surface. For the extremely to be 1m in the MKS system.
small pressure perturbations typical of sound, the
equations of fluid dynamics can be linearized to arrive
at the ‘acoustic wave equation’, which is the conven-
tional mathematical description of acoustic pressure The quantities, (r,,i/r)PouT(r - ct) and (Y,,~/Y)PIN
waves. In one dimension, the acoustic wave equation is (Y+ ct) are out-going and in-going spherical waves,
given by eqn [l], where p is the acoustic pressure, x is respectively. Note that the spherical wave solution has
distance, and t is time. the same mathematical form as the plane-wave solu-
tion except that the amplitude falls off as l / r .
For a source far away from boundaries, the acoustic
pressure is given by an out-going wave having the same
shape in the time domain as the source function.
The general solution to eqn [ l ] is of the form For example, for a time-harmonic source, the acoustic
p(x,t ) = ~ R ( X- ct) + ~ +
L ( X ct), where ~ R ( X - ct) field is a traveling sinusoidal wave of the form
is a right-going wave and p ~ ( x+ ct) is a left-going
p = ( Y , , f / Y ) p o sin(kr - o t ) , where po is the pressure
wave. The right-going wave, for example, could be a amplitude at the reference distance. Pictorially, an out-
transmitted pulse, and the left-going wave could be going spherical wave can be represented as shown in
an echo. Continuous waves as well as pulses satisfy Figure 3, where pressure maxima and minima of the
the wave equation. For example, since c = u / k , the
sinusoidal pressure wave discussed above satisfies the
one-dimensional wave equation. Moreover, as indi-
cated in Figure 2, any function of (x- ct) or (x+ ct)
satisfies eqn [l]. Further, the perturbations in density
f
and temperature associated with an acoustic pressure
wave satisfy the same wave equation as the acoustic
pressure except that, instead of pressure, the variable is
density or temperature, respectively.
The three-dimensional form of eqn [l]is eqn [ 2 ] ,
where (x,y, z ) are Cartesian coordinates.
For a symmetrical source, such as a small explosion

high above the ground, the three-dimensional wave
equation has spherical symmetry and can be written as
eqn ~31. Figure 3 Schematic representation of an out-going spherical
wave. The circles are the wavefronts and the straight lines are
a2 1 a2 acoustic rays. The solid and dashed circles denote, respectively,
-(Ye) = - -(@I 131 wavefronts
ar2 c2 at2 for maximum and minimum pressure.
ACOUSTIC WAVES 3
wavefronts are represented, respectively, by solid lines To express in decibels the variation of the rms
and dashed lines. The radial lines perpendicular to pressure, prms(r),with distance, it is conventional to
wavefront are acoustic rays. define the SPL at r = rrefas the ‘source level’ (SL) as in
eqn [6] and to define the decibel decrease in acoustic
pressure with distance as the ‘transmission loss’ (TL)
Sound Pressure Levels and Decibels as in eqn [7].
Acoustic pressure amplitudes encountered in practice
typically vary over several orders of magnitude.
Consequently, it has become conventional to use a
logarithmic scale to describe the amplitudes. For
continuous waves, the amplitude of interest is the
root-mean-square pressure amplitude, prms,and is 171
referenced to some standard reference pressure pref.
For pulses, some ‘peak’ pressure is often chosen. In In these equations, as noted previously, the reference
either case, a logarithmic amplitude measure called the distance, rref,is l m in the MKS system. (Note that
‘sound pressure level’ (SPL)is commonly used, defined since pressure usually decreases with distance, trans-
mission loss is usually positive.) Using the above
by eqn PI.
definitions for source level and transmission loss, one
can write the sound pressure level at a distance r from
the source as SPL = SL -TL. For example, in MKS
units, the rms pressure amplitude for a spherically
The pressure of interest is denoted here as the rms spreading wave can be written as erms(Y) = ~ I / Y ,
pressure, but could be any pressure, depending on the where p 1 is the rms pressure at 1 m . Thus, for a
application. Although SPL is actually a dimensionless spherically spreading wave, the source level is
quantity, one refers to the ‘units’ as decibels (dB), 20 loglo (pl/pref)and the transmission loss is simply
referenced to a reference pressure, pref.In atmospheric 20 log10 (Y).
acoustics, the reference level is usually chosen to be In general, the transmission loss is not a simple
2 x 10 - Pa or 20 pPa, which is the approximate function and must be computed numerically. With
threshold of hearing. Note that with the above numerical computations, it is often useful, for plotting
conventions, the SPL for 20 pPa is 0 dB. purposes, to subtract the transmission loss due to
The frequency range for audible sound for the spherical spreading, that is, to subtract 20 loglo (Y).
human ear is from approximately 20 Hz to approxi- Such a convention is equivalent to giving the sound
mately 20kHz. Typical sound pressure levels for pressure level relative to a spherically spreading wave,
sounds in the audible range are given in Table 1. and hence is given the name ‘relative sound pressure
In addition to being used as a measure for absolute level’. Thus, by definition, the relative sound pressure
pressure, decibels are also used to describe relative level for a spherically spreading wave is OdB. Ex-
changes in pressure. For example, if a pressure pressed as a relative sound pressure level, an SPL above
amplitude decreases with distance by a factor of 10, or below that for spherical spreading will be, respec-
it is conventional to say that, over the distance, the tively, greater than or less than zero.
acoustic pressure has decreased by 20 loglo( 10) =
20dB. If one pressure amplitude were 100 times The Speed of Sound in the Atmosphere
greater than another, one would say the first pressure
was 40 dB greater than the second pressure. To a good approximation, the atmosphere can be
treated as an ideal gas, and the acoustic pressure
variations in it can be treated as adiabatic; that is, there
Table 1 Representative list of audible sound pressure levels
is no heat flow from the higher pressure (hotter)
Sound SPL regions to the lower pressure (cooler) regions. For an
ideal gas and adiabatic compression (or rarefaction),
Threshold of hearing
Leaves rustling
0 dB
20 dB
the speed of sound is given by c = m, where y =
Quiet conversation 40 dB 1.40 is the ratio of the constant-volume specific heat
Normal conversation 60 dB for air, C V , to the constant-pressure specific heat, c p .
Average street traffic 80 dB The quantity, R = 286.69 J kg-’ K-l, is the gas con-
Diesel truck (at 10 m) 90 dB stant for dry air, and T i s the absolute temperature (K).
Jet take-off (at 10 m) 120dB
Threshold of pain
With an ideal gas model, the theoretical value for c at
140 dB
0°C (273.16K) is 331.lmsW1,which is in excellent
4 ACOUSTIC WAVES
agreement with the experimental value of 33 1.6 m s - the expanded regions are slightly cooler, a small
given earlier. For values of T not far from O'C, the amount of heat flows from the compressions to the
square root expression for the speed of sound can be rarefactions. The conduction of heat converts the
expanded linearly and written approximately as organized motion associated with the sound wave into
c = (331 + 0.6Tc) m sP1, where TCis the temperature random thermal motion of the gas molecules. Because
in degrees Celsius. Thus, for an increase in tempera- the heat flow lowers the temperature of the compres-
ture of one degree Celsius, the speed of sound increases sions and raises the temperature of the rarefactions,
by 0.6ms-I. both the pressure maxima and minima are reduced.
The reduction manifests itself as a decay of the
acoustic wave with distance. The component of
Absorption of Sound in the absorption due to thermal conduction is given by
Atmosphere eqn [9], where IC is the coefficient of thermal conduc-
In addition to the decrease in pressure amplitude of an tivity in J ( k g m o l ) - ' K - ' k g r n - ' ~ - ~ .
acoustic wave due to propagation effects such as
'geometrical' spreading (e. g., spherical spreading), the 191
amplitude is also reduced by atmospheric absorption.
A sound wave propagating through 'clean' air (no
solid particles) is attenuated owing to two basic In addition to energy loss due to classical mecha-
mechanisms: nisms (viscosity and heat conduction), energy can also
be lost via quantum-mechanical 'relaxation' processes
0 classical losses due to momentum transfer across a involving the internal degrees of freedom (rotation and
velocity gradient (viscosity) and heat flow across a vibration) of oxygen and nitrogen molecules. The
temperature gradient transfer of translational energy to internal degrees of
Quantum-mechanical losses due to relaxation pro- freedom and back takes place through an extended
cesses, mainly relaxation of rotational and vibra- sequence of molecular collisions, so there is a time
tional states in nitrogen and oxygen molecules. delay associated with the energy transfer. Because of
the time delay, relaxation processes cause energy to be
For both mechanisms, the effects of absorption can be
lost from the organized translational motion that
represented by an absorption coefficient, a, which has
constitutes the acoustic wave. As a result, just as with
units of m-'. The absorption coefficient enters via an
the classical mechanisms, the pressure amplitude of
exponential, so that the pressure is given by
the wave decreases as the wave propagates.
p = poe-cts, where po is the unattenuated pressure For any particular relaxation process, the associated
amplitude and s is the distance the wave has traveled.
absorption coefficient has the general form of eqn [lo].
To indicate the choice of the Napierian base, e,
the attenuation coefficient is, by convention, said
to have units of nepersm-'. The corresponding
attenuation coefficient, a, for decibels (base 10) is
a = 20a loglo (e) = 8 . 6 8 6 4 and, by convention, has
units of dB m - '. In eqn [lo],5, is the relaxation strength (in nepers), cis
Experimental and theoretical studies indicate that the sound speed, f is the frequency, and f r is called the
the total absorption coefficient can be represented as a 'relaxation' frequency. The relaxation frequency is the
sum of absorption coefficients, with each distinct frequency for maximum absorption and is roughly
physical process having its own unique coefficient. the reciprocal of the characteristic time delay for the
That is, the total absorption coefficient, ET, can be transfer between kinetic energy and internal energy of
written as ET = E, ai, where is the absorption the gas molecules.
coefficient associated with a particular mechanism. For air, there are three important relaxation pro-
For example, the component of sound absorption due cesses: (1)0 2 vibration, (2) N2 vibration, and ( 3 ) N2
to viscosity is given by eqn [8], in which o is the rotation. The relaxation frequency for N2 rotational
angular frequency, c is the speed of sound, po is the relaxation is very high, so that, below lOMHz, the
density of air, and p is the coefficient of viscosity of air. denominator in eqn [lo] is approximately unity. Thus,
N2 rotational relaxation varies as f 2 and can be
combined with the classical absorption coefficient. If
one denotes the classical-plus-rotational absorption
coefficient (i.e., the coefficient for viscosity, heat flow,
Since the compressed regions in an acoustic wave and N2 rotation) as ac, the absorption coefficient for
are slightly hotter than the ambient temperature, and 0 2 vibration as E O , and the absorption coefficient for
ACOUSTIC WAVES 5
10-0 point source in an atmosphere with a constant

temperature, and hence a constant sound speed.
lo-’ With a constant sound speed, acoustic ray trajectories
are straight lines. In reality, however, the atmospheric
10-2 temperature is never constant in space or time.
c
1 Consequently, the speed of sound is not constant but
E
2 10-3 varies spatially and temporally. In a typical daytime
a,
situation, the temporally averaged temperature is
- Q
E 10-4
independent of range but decreases with height (‘lapse’
.-0
a condition). Thus, on average, the sound speed de-
E 10-5 creases with height, and sound rays curve upward, as
8 shown in Figure 5 . The ray paths shown are for a sound
C
speed that decreases linearly with height. (For the
’= 10-6
0
simple case of linear variation, the ray paths are arcs of
$ circles.)
2 10-7 For a general sound speed variation in a stratified
atmosphere (i.e., no horizontal variation), ray paths
10-8 are governed mathematically by Snell’s law, which
states that the quantity c ( z ) / cos O(z) is invariant,
10-9 where at a height z, the quantities, c ( z ) and cos Q(z)
10’ 102 103 104 105 are, respectively, the sound speed and the cosine of the
Frequency (Hz) angle of a ray with respect to horizontal. Thus, Snell’s
law says that, if c ( z ) decreases with height, cos O(z)
Figure 4 Components and general behavior of the total absorp- must increase, so that the ray bends upward. In
tion coefficient for air. The contributions to the total absorption ( T )
are the classical plus N2 rotation (C), the 0 2 vibration (0)and the
general, acoustic rays bend toward regions of lower
N2 vibration (N). (Reproducedwith permission from Bass (1991), sound speed and away from regions of higher sound
VCH Publishers, Inc., Weinheim, Germany.) speed. The bending of acoustic rays is given the name
‘refraction’.
Nz vibration as EN, then the total absorption coeffi- The physical basis for refraction can be understood
cient, MT for air can be written as eqn [Ill. using the situation in Figure 5 . Consider a small
section of wavefront associated with a ray that leaves
+ +
?T = ?C EO ?N 1111 the source nearly parallel to the ground. For a small
enough section, the wavefront is nearly planar and
Figure 4 shows the total absorption coefficient, UT, nearly vertical. Since the sound speed decreases with
together with the components, M C , M O , and M N . Note height, the lower portion of the wavefront travels
that below about 10 000 Hz the absorption is domi- faster than the upper portion, causing the wavefront to
nated by vibrational relaxation. Further, note that, turn upward. In terms of rays, we would say that the
above about 1000 Hz, atmospheric absorption is ray is refracted upward due to the decrease in the
significant for propagation distances of a kilometer sound speed with height.
or more, which accounts for the lack of long-range Refraction of acoustic waves is caused by spatially
propagation of high-frequency sound. varying wind as well as by spatially varying temper-
ature. The effect of the wind on acoustic waves can be
Refraction of Sound in the Atmosphere accounted for approximately by defining an ‘effective’
sound speed, ce, which is the previously defined
The spherical wavefronts and associated rays shown in ‘adiabatic’ sound speed, c, = m, plus the compo-
Figure 3 represent acoustic waves radiating from a nent of the vector wind in the direction of propagation.
Figure 5 Acoustic rays and shadow zones for an upward-refracting daytime atmosphere. For clarity, rays that are reflected off the
ground are not shown. (Reproduced with permission from Bass (1 991).(VCH Publishers, Inc., Weinheim, Germany.)
6 ACOUSTIC WAVES
Figure 6 Acoustic rays and shadow zone for an atmosphere that is upward-refractingin the upwind direction and downward-refractingin
the downwind direction. For clarity, rays that are reflected off the ground are not shown. (Reproduced with permission from Bass (1991),
VCH Publishers, Inc., Weinheim, Germany.)
For example, let 1 be a unit vector pointing from the than the sound pressure level one would expect with
source to a receiver. Then, if the vector wind is denoted spherical spreading alone. Because of upward refrac-
as v ( u x ,u y ,u z ) ,the effective sound speed is given by tion, daytime sound pressure levels for a near-ground
c, = ca + i . v , where 1 -v is the component of the source fall off dramatically with horizontal distance as
vector wind in the direction of sound propagation. In one enters the shadow region, which, for strong
general, near the ground, the horizontal wind speed upward refraction, can be within 100-200m of the
increases with increasing height. For upwind propa- source.
gation of sound, therefore, the horizontal wind pro- At night, in contrast to the daytime situation,
gressively reduces the effective sound speed with ground-to-ground propagation is very good. Owing
increasing height. For downwind propagation, the to radiative cooling of the ground, both the near-
effect is reversed. That is, the horizontal wind ground air temperature and the sound speed are lower
progressively increases the effective sound speed than at higher altitudes (an 'inversion' condition). As a
with increasing height. In the daytime, for example, result, acoustic rays launched near to horizontal (less
where the temperature and adiabatic sound speed than about 10" with respect to horizontal) are bent
decrease with height, upwind propagation adds to the downward, causing sound to be trapped in a 'sound
upward refraction already present. Downwind, if the duct' near the ground. Rays launched at steeper angles
wind speed gradient is sufficiently large, the horizontal escape the duct and continue upward (see Figure 7).
wind can overcome the upward refraction due to the With strong trapping and small ground-bounce loss
daytime temperature profile and lead to downward (e.g., over water) the acoustic field in the near-surface
refraction. The ray paths for upwind and downwind sound duct undergoes approximately cylindrical
propagation are illustrated in Figure 6 . For propaga- spreading (l/fi)instead of spherical spreading (l/r).
tion directly across the wind, there is little effect due to As a consequence of daytime upward refraction and
the wind, but upward refraction persists because of the nighttime downward refraction, noise sources that are
decreasing temperature with height. not generally heard during the day can often be easily
As indicated in Figures 5 and 6 , for up- heard at long distances (e.g., several kilometers) at
ward refraction, there is a region, called an acoustic night. The long-range propagation of acoustic waves
'shadow', where no acoustic rays can penetrate. In the at night makes noise control much more difficult than
shadow region, the acoustic levels are much lower during the day.
Figure 7 Acoustic rays for a downward-refracting nighttime atmosphere. Rays launched at small angles with respect to the horizontal
(less than about 10") are trapped in a 'sound duct' near the ground and can propagate to ranges of several kilometers. For steeper launch
angles, the rays escape the duct and continue upward.
ACOUSTIC WAVES 7
Diffraction of Acoustic Waves wave equation for a 500Hz point source in an

in the Atmosphere upward-refracting atmosphere. The color plot in the
figure, which is for the relative sound pressure level as
As discussed above, one can approximately represent a function of range and height, shows the effects of
an acoustic field in terms of wavefronts whose both diffraction and refraction. The edge of the
propagation directions (i.e., rays) are governed by shadow boundary in Figure 8 would become more
refraction. Such a representation is useful visually and diffuse at lower frequencies (longer wavelengths),
can be valid computationally when the acoustic until finally, at very low frequencies, the shadow
wavelengths are much smaller than the smallest boundary would not be discernible at the ranges
sound-speed structure in the atmosphere. The main shown.
effect left out in the so-called ‘ray theory’ of sound When the atmosphere is downward-refracting (e.g.,
propagation is the wave phenomenon known as at night), the presence of diffraction again causes a
‘diffraction’. Diffraction is responsible for the well- ‘blurring’ of the features of the acoustic field, just as
known ability of sound to ‘bend’ around corners and with upward refraction. There are no sharp disconti-
obstacles. In outdoor sound propagation, diffraction nuities in the structure of the acoustic field. Rather,
fills in gaps in the acoustic field that would be present because of the finite wavelengths in the acoustic field,
in a purely ray-based representation. Full-wave solu- the changes in the field are continuous and smooth, as
tions to the wave equation (usually numerical) auto- shown in Figure 9, which is also for a 500Hz point
matically include both diffraction and refraction. source. In general, the longer the acoustic wavelengths
Because of diffraction, every acoustic field has an (i.e., the lower the frequency), the smoother the
intrinsic smallest possible scale length that is roughly a features of the acoustic field.
quarter of the smallest wavelength present in the field. In addition to limiting the sharpness of the acoustic
Owing to the scale limitation, there can be no sharp field, diffraction is responsible for the scattering of
edges in the acoustic field. For example, instead of the acoustic waves from the complex small-scale structure
sharp shadow boundary obtained with rays (Figures5 of the real atmosphere. In a realistic model of the
and 6), a smooth, diffuse boundary is obtained when atmosphere, the instantaneous temperature and vec-
diffraction is included. Such a situation is illustrated in tor wind fields are not smooth but are highly irregular,
Figure 8, which shows a numerical solution of the containing eddies of all sizes. The eddy sizes of most
Figure8 Numericalsolution for the sound field in an upward-refractingdaytime atmosphere without turbulence. Owing to diff raction, the
edge of the shadow boundary is diffuse. The color plot shows the relative sound pressure level as afunction of range and height. (Note that
the vertical scale is much less than the horizontal scale, so that the actual propagation angles with respect to the horizontal are much
smaller than shown.)
8 ACOUSTIC WAVES
Figure 9 Numerical solution for the sound field in a downward-refracting nighttime atmosphere without turbulence. Note that downward
refraction ‘traps’ sound in the near-ground duct. As in Figure 8, the features of the acoustic field are blurred owing to diffraction. The color
plot shows the relative sound pressure level as a function of range and height. (Note that the vertical scale is much less than the horizontal
scale, so that the actual propagation angles with respect to the horizontal are much smaller than shown.)
concern for audible sound are in the region called the inertial-subrange eddies. As a consequence of diffrac-
‘inertial subrange’, which typically begins at a few tens tion, new wavefronts emanate from every eddy, with
of meters and goes down to a few millimeters. In the the strongest scattering occurring in the near-forward
inertial subrange, the eddy structure is governed by direction. The diffracted acoustic waves that are
the well-known Kolmogorov spectrum. As a conse- scattered downward act to fill in the shadow region.
quence, at any instant of time, the small-scale spatial An example of this phenomenon is shown in Figure 11,
structure of the sound speed field, which depends on which was computed numerically using realistic rep-
temperature and vector wind, can also be described by resentations for .? and 6c. With a realistic model for the
a Kolmogorov spectrum. In the daytime turbulent sound speed, the predicted mean near-ground levels
boundary layer, for example, the sound speed can be ( -20 dB to - 30 dB relative to spherical spreading) in
approximated as a time-independent mean sound the shadow region (0.2-2 km) are in good agreement
speed, .?(z), that varies only with height, plus a with observation. The relative sound pressure level for
fluctuating part, 6 c ( x ,y, z, t ) , that varies with time, a longer-range interval is shown in Figure 12. It is
horizontal distance, and height. Hence the total sound apparent from Figure 12 that, even with scattering into
speed, c(x,y, z, t ) ,can be represented as in eqn [12]. the shadow region, daytime levels near the ground are
very low at ranges beyond a few kilometers.
It can be observed in Figures 11 and 12 that the
effects of turbulence are most apparent in the shadow
The quantity .?(z) approximates the slow, large-scale region, where the sound levels would be extremely
variations in the sound speed profile and 6c(x,y, z, t ) small in the absence of turbulence. Above the daytime
describes the rapid, smaller-scale fluctuations. shadow region, in the ‘insonified’region, the levels are
As noted earlier, the quantity & ( x , y , z , t ) follows much higher, so that the effect of scattering from
the same Kolmogorov statistics as do the temperature turbulence is less dramatic, though the effect increases
and wind fluctuations. At a particular instant of time, with increasing distance from the source. Similarly, for
an approximate ‘snapshot’ of the sound-speed fluctu- nighttime propagation, in the near-ground acoustic
ation field, & ( x , y, z, t ) ,can be synthesized by adding duct, where the mean levels are high, the effect of
together, with random phase, the wavenumber com- scattering from turbulence is not as dramatic as in the
ponents for a Kolmogorov spectrum. The result of shadow region above the duct (see Figure 13).Further,
such a synthesis is shown in two dimensions in Figure the nocturnal boundary layer, being more stable,
10. Because audible sound has wavelengths compa- intrinsically has weaker turbulence than the daytime
rable in size to small-scale atmospheric structure, it is boundary layer. As a consequence, mean sound levels
scattered in all directions as it propagates through at night near the ground are not affected by turbulence
ACOUSTIC WAVES 9
Figure 10 A two-dimensional 'snapshot' of small-scale turbulent fluctuations in the sound-speedfield synthesized using a Kolmogorov
spectrum and random Fourier components. The fluctuation magnitudes are typical of those created by turbulence in the daytime.
nearly as much as the near-ground daytime levels. It Acoustic Remote Sensing of the
should be noted, however, that large-scale nocturnal Atmosphere
phenomena such as flow down a slope and gravity
waves can have a significant effect on sound levels at As illustrated above, the sensitivity of acoustic waves
night. The effects of such large-scale dynamical to atmospheric wind and temperature variations
features on sound propagation are the subject of makes accurate prediction of ground-to-ground sound
current research in atmospheric acoustics. propagation a challenging problem. Conversely,
Figure 11 Same as Figure8 except that the small-scale turbulence shown in Figure 10 is included. Note that the effects of turbulence
are most apparent in the shadow region. Scattering of acoustic waves from turbulence 'fills in' the shadow region so the levels there are
much higher than the no-turbulence case is shown in Figure 8.
10 ACOUSTIC WAVES
Figure 12 Same as Figure 11 except that the range extends to 10 km. Note that, even with scattering into the shadow region, daytime
sound pressure levels near the ground are very low at ranges beyond a few kilometers.
however, that same sensitivity makes sound a remark- cated, is common and is known as a ‘monostatic’
ably valuable probe for remotely sensing the complex sodar. (A less common geometry has the receiver
features of the atmospheric boundary layer. The separated horizontally from the transmitter and is
most widely used acoustic tool for atmospheric called a ‘bistatic’ sodar.) When the upward-going pulse
sensing is a pulse-echo probe called a ‘sodar’, after encounters wind and temperature inhomogeneities
the more familiar ‘radar’, which is an electromagnetic produced by turbulence, faint scattered waves are
pulse-echo device. (Note: Sometimes the name ‘echo- created within the air itself. With monostatic sodar, the
sonde’ is also used, but that designation is less common part of the acoustic wave scattered back toward the
than ‘sodar’.) ground, the echo, is detected using the same trans-
The first sodars, which appeared in the early 1970s, ducers that produced the probe beam.
emitted an acoustic pulse in a single vertically pointing Early sodars were used primarily as instruments for
beam as shown in Figure 14. The sodar geometry detecting turbulence. The time delay between the
shown, with the acoustic source and receiver collo- emitted pulse and its echo determined the height of the
Figure 13 Same as Figure 9 except that small-scale turbulence typical of a nighttime atmosphere has been included. Since sound
levels are relatively high in the near-ground nocturnal sound duct, the effects of small-scale turbulence are less evident than during the
daytime.
ACOUSTIC WAVES 11
Figure 14 Geometry for a monostatic sodar. An array of

transducers projects a burst of acoustic waves vertically in a
beam. Turbulence-generated inhomogeneities in wind and tem-
perature scatter sound back toward the transmitting transducers,
which act as a directional receiver for the faint echoes received on
the ground. Figure 15 Record from a sodar taken over a diurnal cycle. The
vertical scale is 0-500 m, and time increases from left to right. The
vertical white lines are hour markers. (A) Unstable daytime
boundary layer with thermal plumes generated by solar heating
turbulence (one-half the time delay times the average of theground. (B) Decreasingdaytime boundary layer followed bya
speed of sound), while the strength of the echo was a growing stable nocturnal boundary layer showing evidence of initial
measure of the turbulence intensity. The evolving gravity wave activity. (C) Stable nocturnal boundary layer after
structure of the atmospheric boundary layer could be midnight with fully developed internal wave activity. (Reproduced
with permission from Atmospheric acoustics, Encyclopedia
‘mapped’ by plotting the delay time and echo strength of Applied Physics, vol. 2, VCH Publishers, Inc., Weinheim,
on a vertically moving strip of paper. For example, the Germany, 1991.)
horizontal distance on the strip could be proportional
to the time delay of the echo, and the darkness could be
proportional to the intensity of the echo. With many turbulent boundary layer descending in the late
repeated pulses, the evolution of the boundary layer afternoon and evening as solar heating of the ground
could be followed visually. With its debut in the 1970s, diminishes. The undulations in the latter part of the
the sodar immediately provided important insights record indicate the onset of internal gravity waves.
into the spatial structure and temporal evolution of the Panel (C)shows fully developed internal wave activity
atmospheric boundary layer. A typical sodar record is after midnight.
shown in Figure 15. The figure has time moving from In addition to visual displays of boundary layer
left to right and shows the evolution of boundary layer structure and dynamics, modern sodars can provide
structures over a typical diurnal period. The vertical quantitative measures of wind and temperature. The
scale in the figure is 0-500m, and the thin white so-called ‘Doppler sodar’, for example, which uses
vertical streaks are hour markers. Panel (A) shows a two slant beams in addition to the usual vertically
typical daytime record of thermal plumes carried pointing beam, can map vector wind versus height. A
through the vertical sodar beam. Panel (B) shows the typical geometry would have a vertical beam, together
12 ACOUSTIC WAVES
with slant beams pointing north and east, respectively, See also
at 60" above horizontal. Using the Doppler shift Boundary Layers: Convective Boundary Layer: Neutrally
in the echoes (up-shift for winds moving toward Stratified Boundary Layer; Stably Stratified Boundary Lay-
the receiver and down-shift for winds moving away er; Surface Layer. Buoyancy and Buoyancy Waves:
from the receiver), the three vector components Optical Observations; Theory. Diurnal Cycle. Parame-
(up-down, east-west, and north-south), can be meas- terization of Physical Processes: Turbulence and
ured as a function of height. Such advances as the Mixing.
Doppler sodar are due, in large part, to the vast
increase during the past 20 years in the computing
power available with small computers. In addition to Further Reading
providing greatly increased signal processing power, Bass HE (1991) Atmospheric acoustics. In: Trigg GL (ed.)
small, powerful computers have also made remote Encyclopedia of Applied Physics, vol. 2, pp. 145-179.
sensing instruments like the Doppler sodar sufficiently New York: Wiley-VCH.
'user friendly' that nonexperts can operate them Brekhovskikh LM (1980) Waves in Layered Media (trans-
successfully. lated by RT Beyer). New York: Academic Press.
A second important advance in acoustic remote Crocker M (1997) Introduction. In: Crocker M (ed.)
Handbook of Acoustics. New York: Wiley.
sensing is the 'Radio Acoustic Sounding System', or
Elmore WC and Heald MA (1969) Physics of Waves. New
RASS, which can provide accurate temperature pro- York: McGraw-Hill.
files as a function of height. A RASS uses a single Kinsler LE, Fry AR, Coppens AB and Sanders JV (2000)
vertically pointing sodar beam together with two Fundamentals of Acoustics. New York: Wiley.
radar beams that converge in the air column over the Morse PM (1981) Vibration and Sound. New York: Acous-
sodar. The radar is used in a bistatic geometry with the tical Society of America/American Institute of Physics.
transmitter on one side of the sodar and the receiver on Morse PM and Ingard KU (1968) Theoretical Acoustics.
the other side. Using coherent radar backscatter from New York: McGraw-Hill.
the upward-going acoustic beam, the RASS measures Neff WD and Coulter RL (1986) Acoustic remote sensing.
the speed of the acoustic beam as it propagates In: Lenschow DW (ed.)Probing the Atmospheric Bound-
upward. After making corrections for the vertical ary Layer. Boston, MA: American Meteorological
wind, one can estimate the adiabatic sound speed (a Society.
Ostashev VE (1977) Acoustics in Moving Inhomogeneous
function of temperature only) as a function of height, Media. London: E&FN Spon.
which then yields the temperature as a function of Pierce AD (1989)Acoustics: An Introduction to Its Physical
height. Extensive comparisons between RASS meas- Principles and Applications. Woodbury, NY: Acoustical
urements and in situ measurements have shown that a Society of America.
RASS provides reliable estimates of temperature at Piercy JE, Embleton TFW and Sutherland LC (1977)Review
heights from a few hundred meters to up to several of noise propagation in the atmosphere. Journal of the
kilometers. Acoustical Society of America 16: 1403-1418.
Salomons EM (2001) Computational Atmospheric Acous-
tics. Dordrecht: Kluwer Academic.
Acknowledgements Stull RB (1993) An Introduction to Boundary Layer
Meteorology. Boston, MA: Kluwer Academic.
The author would like to thank Mr RC Clark for the Sutherland LC and Daigle GA (1997) Atmospheric sound
acoustic propagation calculations and Ms EA Furr for propagation. In: Crocker M (ed.) Handbook of Acous-
expert technical editing assistance. tics. New York: Wiley.
AEOLIAN TRANSPORT
See DUST
AEROSOLS / Climatology of Tropospheric Aerosols 13
Contents
Climatology of Tropospheric Aerosols

Observations and Measurements
Physics and Chemistry of Aerosols
Role in Cloud Physics
Role in Radiative Transfer
plant debris. Most particles include some water,

Climatology of Tropospheric although cloud elements are usually excluded from
the description. Aerosol particles have both natural
Aerosols sources (e.g., wind-blown sea spray, dusts, and
volcanic debris) and anthropogenic sources (such as
J L Gras, CSIRO Atmospheric Research, Aspendale, smokes, fumes, and exhausts). In some cases, such as
Victoria, Australia
wind-blown dusts from poorly managed agricultural
Copyright 2003 Elsevier Science Ltd. All Rights Resewed. areas, this distinction may be unclear. In many regions,
particularly Northern Hemisphere mid-latitudes,
anthropogenic components now frequently dominate
Introduction
particle number and mass. In general, an aerosol
Although the idea of a climatology of tropospheric comprises particles of mixed composition both intern-
aerosol appears relatively straightforward, practical ally (i.e., within one particle) and externally (between
implementation of a comprehensive global-scale particles); this makes aerosol unlike a gas, which is the
climatology turns out to be very difficult; arguably, it same wherever it is measured. Particle lifetimes are
is far from well established. Most problems stem from dependent on size and location in the atmosphere.
the inhomogeneous nature and spatial distribution of Typically, lifetimes range from days near the surface to
the aerosol, combined with sparse and relatively months in the upper troposphere. There are many
uncoordinated measurements. Despite this, much is sources of particles, ranging from point sources
known about the distribution of aerosol throughout including volcanoes, fires, and industrial plants,
the troposphere and how it varies seasonally. New through to cities, clouds, entire regions, oceans, and
remote-sensing technologies and analysis methodolo- even the atmosphere itself. In general, the spatial and
gies allow detailed observations and global mapping temporal distribution of aerosol is notably inhomo-
of some aerosol parameters. Determination of the geneous. Substantial changes in aerosol properties can
global distribution of aerosol intrinsic properties, such occur over distances of only a few kilometers
as chemical composition, and the long-term trends of horizontally and much less vertically. The atmospheric
all aerosol properties remains a major challenge. aerosol is extremely dynamic, evolving and changing
An aerosol is a stable suspension of a solid or liquid properties from the point of production until its
in a gas, in this case air. In practical terms, for the ultimate removal by sedimentation or wet processes.
ambient atmosphere this includes particles from During its lifetime, a particle may amalgamate with
nanometers to tens and sometimes, hundreds of other particles, exchange material through gas-phase
micrometers in diameter (around lop9 to 10-4m, or reactions, and most likely pass through a number of
around five decades in size, and spanning an even cloud cycles where it is incorporated into cloud
greater range of concentrations). Constituent particles droplets, reappearing in a modified form when the
in the aerosol may be primary, that is, emitted as a droplets evaporate.
particle, or secondary, being produced from gases in The troposphere is that part of the atmosphere lying
the atmosphere (by condensation or chemical reaction between the surface and the stratosphere. The name
of certain species such as dimethyl sulfide (DMS), derives from the Greek tropos for ‘turning’, and this
sulfur dioxide, and volatile organic compounds). They part of the atmosphere is characterized by air motion
may include inorganic, organic, and biological entities and mixing. Near the poles the troposphere reaches to
such as spores, pollens, viruses, bacteria, waxes, and about 8 km altitude and in the tropics to around 18 km
14 AEROSOLS I Climatology of Tropospheric Aerosols
altitude. The concept of a well-mixed troposphere is the aerosol properties or elements. For tropospheric
somewhat misleading though, since many of the aerosol on the global scale, no truly comprehensive
important tropospheric aerosol features are associated ‘observational’ climatology exists for any intrinsic
with discrete layers in the atmosphere and the aerosol property and it may be unachievable in the
transport of aerosol in layers, often in large quantities. foreseeable future. None exists even for properties as
Tropospheric aerosols impact on global climate, the apparently fundamental as total mass or number.
atmospheric environment, and even human health. Sufficient reliable data exist for certain aerosol pro-
Aerosols are a significant component in the global perties at selected sites or in certain regions for
radiation balance, scattering and absorbing solar the development of limited ‘observational’ aerosol
radiation and changing the properties of clouds. Their climatologies. Very few records of any aerosol para-
climatic effect is of similar magnitude to the green- meter are multidecadal, and establishing these
house effect of carbon dioxide, but the net effect of records remains an important task for international
particles is a cooling. Aerosols are important in the science.
precipitation process: all cloud drops form on aerosol Contrasting with this is a ‘fully modeled’ aerosol
particles, known as cloud condensation nuclei, and ice climatology. A comprehensive aerosol model should
nuclei play a similar role in ice clouds. Aerosols be able to distribute precursor material between new
interact with reactive trace gases (heterogenous particles and existing particles and generate a full
processes) and are implicated in biogeochemical description of the distribution of mass and intrinsic
cycling in the atmosphere, including transport of trace aerosol properties, as a function of size, from around
nutrients such as iron to the world’s oceans and of lop9m to lop4m diameter, giving the three-dimen-
wind-generated sea-salt to the land, and playing roles sional spatial distribution of these properties over
in the sulfur, nitrogen, and carbon cycles. At the urban time. Aerosol dynamical models and chemical trans-
and rural levels they are the dominant cause of port models exist but none currently has all these
reduction of visibility and add to acidification and capabilities.
transport significant quantities of chemicals, such as An intermediate approach, in effect generating a
minerals, sulfate, nitrate, and carbonaceous material. ‘hybrid’ aerosol climatology, has also been used. It
Some aerosol components are toxic, others are comprises a model, based on an external mixture of
carcinogenic. Epidemiological studies have shown aerosols from a set of generalized sources (e.g.,
significant correlation between various aerosol prop- biomass burning emissions, mineral dust, and sulfate),
erties, including certain mass fractions, and human each with a given emissions inventory. Each compo-
mortality and morbidity. Bioaerosol (molds, spores, nent contributes its own generic properties, such as
pollens, dust mite feces, bacteria and viruses, and descriptions of the size distribution and optical
possibly fragments) are implicated in diseases includ- properties, derived from a variety of in situ or remotely
ing asthma. sensed measurements.
How do aerosol properties vary across the globe and Data quality is central to all climatologies. Mea-
as a function of time, location, and altitude in the surements with unknown accuracy or poorly specified
atmosphere? These factors are usually studied through conditions of measurement are ultimately of little or
development of a ‘climatology’. Strictly speaking, a no value.
climatology is a statistical description of a defined In addition to the spatial or temporal variation of
system, composed of various elements. These elements aerosol properties, other useful descriptions include
exhibit spatial and temporal variations to which the the amount of material or mass flux emitted by various
climate concept applies. For the global climate, for sources. Major aerosol flux contributors include
example, this would include temperature and rainfall. mineral aerosol, sea-salt, smoke from biomass burn-
The most fundamental elements for the tropospheric ing, anthropogenic emissions (as sulfate and carbo-
aerosol system are the size-dependent concentration naceous material), and secondary aerosol sources
and aerosol intrinsic properties; the latter properties (including natural precursor gases). Table 1 gives
are independent of concentration and include chemi- typical estimates of the annual flux of various sources
cal composition, refractive index, and shape. Other from the 1970s and 1990s, although the relative
elements or means of descriptions of the aerosol are importance for both environmental and climate
also possible, such as light extinction coefficient and impacts is not simply related to these mass fluxes.
integral mass. Particles in the range 50 nm to 1pm, for example,
Unfortunately, no single definition of an aerosol interact strongly with solar radiation and have the
climatology is universally accepted. The definition biggest effect on cloud properties. For mineral and sea-
above, which describes an ‘observational’ aerosol salt aerosol, massive particles with very short lifetimes
climatology, requires comprehensive measurements of complicate the definition of the flux. Flux estimates
AEROSOLS / Climatology of TroposphericAerosols 15
Table 1 Estimated strength of tropospheric aerosol sources Satellite-based remote sensing methodologies are
Early 1970s 1990s
very effective for geographically mapping various
(TdY - I ) (TdY-') measures of aerosol 'amount', but the determination
of intrinsic size-dependent properties such as compo-
Natural sition and particle shape require in situ determination.
Mineral 10-500 1000-3000a
Consequently, there are far fewer data available for
Sea-salt 200-1 000 1000-3000'
Volcanic 3-1 50 15-90 these properties, and the intercomparability of sam-
Biogenic 50 pling protocols becomes very important. Geographic
Gas to particle 300-2000 200-1 300 distributions of some species, such as sulfate, have
Extraterrestrial 0.1-50 10 received considerable attention and, while no really
Anthropogenic comprehensive global climatology of intrinsic proper-
Industry, fossil fuel, 100-400 300
carbon, sulfate, ties currently exists, there are a growing number of
nitrate, organics national networks addressing the issue of aerosol
Biomass burning 3-1 50 100-450 composition, including sulfate, nitrate, and carbona-
'Depends strongly on upper size limit.
ceous components.
In the following sections the troposphere is con-
sidered in two altitudes ranges, greater than and less
have been evolving since the early 1970s as informa- than 5 km, with emphasis placed on aerosol mass and
tion on aerosol emission rates and the extent of sources number concentration.
improve.
Lower Troposphere
Spatial and Temporal Distribution of
The altitude range from the surface to around 5 km
Particle Properties includes the most intensively studied region of the
Convection and advection control the transport of troposphere through surface networks, mountain top
atmospheric particles. Most particle sources are near observatories, and aircraft measurements. This part of
the surface and concentrations generally decrease by the atmosphere is complex, involving the diurnally
about two-thirds for each 1km altitude to about 5 km varying planetary boundary layer, nocturnal inver-
over the continents and 2 km over oceans. Above this, sions, the marine boundary layer, clouds, and
particle mixing ratios are relatively constant with precipitation. Surface measurements are not always
altitude to near the tropopause. Concentrations of good indicators of the free troposphere or even the
larger particles increase again in the lower strato- boundary layer. Measurement networks are also not
sphere and total particle concentrations decline. uniformly distributed spatially and have widely
Tropospheric air enters the stratosphere at low different measurement parameters and methodolo-
latitudes, returning to the troposphere, via tropopause gies. Remote sensing of column-integrated aerosol
perturbations (folding), polewards of about 30" N a n d properties from satellites using spectrally resolved,
30" S. This return air carries some stratospheric scattered solar radiation also primarily senses lower-
particles into the upper troposphere and sedimenta- tropospheric aerosol properties, with roughly equal
tion provides a flux of larger particles into the upper contributions from the boundary layer and free
troposphere at all latitudes. Stratospheric particles can troposphere.
be identified to about 1-2 km below the tropopause. From late in the 19th century, systematic recording
Extraterrestrial particle sources are relatively minor. of meteorological observations by ships' crews
Very small particles have an altitude profile more included aerosol phenomena (particularly haze).
indicative of a distributed source in the upper free Compiled in the 1930s, these represent one of the
troposphere, and freshly nucleated particles have been most extensive aerosol climatologies of the period and
observed in the upper troposphere when the integrated several decades following. The geographic distribu-
aerosol surface area becomes small. Both features are tion and seasonal variation from more than 50 years'
consistent with a relatively homogeneous tropo- observations show most features evident in today's
spheric aerosol away from major sources, the so- satellite observations. These include Saharan dust
called tropospheric background. Adding to this are plumes over the Atlantic ocean, the eastern Atlantic
remnant aerosol from various sources and the many European plume, a North American summer plume in
major enhancements in tropospheric aerosol concen- the North Atlantic, spring dust over the north-western
tration associated with layer transport, particularly of Pacific, a summer plume over the Arabian sea, and dry-
mineral dusts, smoke plumes, and anthropogenic season plumes in north-western Australia and Indo-
material. nesia. Currently, satellite-based remote measurements
16 AEROSOLS / Climatologyof Tropospheric Aerosols
afford the most extensive geographic mapping of those

components of the aerosol that interact appreciably
with visible or near-visible wavelengths, although
surface-based measurements of column-integrated
optical properties afford higher precision and the
stability needed to assess subtle long-term changes.
Examples of dust and smoke distributions are
shown in Figure 1 in a monthly average ‘Wabsorp-
tion aerosol index’ from the Total Ozone Mapping
Spectrometer (TOMS), for October 1997. This
demonstrates the extensive smoke plumes from fires
in Indonesia and Brazil as well as some dust over
Africa. Figure 2 shows extensive mineral aerosol from
Africa extending over the Atlantic Ocean during April
2000, as well as other dust including some over eastern
China. Individual daily mapping, such as for the 1997 Figure 2 TOMS UV-absorbing aerosol index for April 2000.
Indonesian fires, is also possible, as shown in Figure 3. (Image courtesy NASA.)
High-resolution spectrally resolved measurements of
scattered solar radiation, for example, from NOAA’s
AVHRR (Advanced Very High Resolution Radio- cities in developing countries. Over ocean regions,
meter) satellites, also allow detailed mapping of concentrations vary with wind speed and proximity to
equivalent column-integrated aerosol light extinction continental sources, but values down to 10 pg rnp3 are
(optical depth) over ocean regions. The major aerosol observed. Total aerosol mass is measured less exten-
plumes can be seen clearly in maps of the distribution sively than mass in aerosol size ranges associated with
of optical depth, such as Figure 4. health or climate effects. In some jurisdictions the mass
loading of particles with aerodynamic diameter
Particle Mass D , < 10 pm (PM10) is regulated. Concentrations
of PMlO in many cities in developing regions exceed
As indicated by remote sensing and in situ measure- 100 pgmP3 (24-hour average), but in developed
ments, aerosol mass concentration varies strongly countries are typically less than 50 pg m-3. In more
with location and time, depending on the proximity to remote continental regions, such as around the
sources and the effectiveness of dispersal and removal American Rocky Mountains, PMlO concentrations
mechanisms. Particle total mass concentrations range are usually less than 10 pg mF3.A seasonal variation in
from around 100 ng m-3 in the upper troposphere to mass loading is common and is frequently dominated
greater than several hundred pgmP3 in many large by local factors, including the seasonal pattern of
Figure 1 TOMS UV-absorbing aerosol index for October 1997. Figure 3 TOMS UV-absorbing aerosol index for 26 October
(image courtesy NASA.) 1997, showing Indonesian fire plumes. (Image courtesy NASA.)
AEROSOLS / Climatology of Tropospheric Aerosols 17
Figure 4 NOAA AVHRR equivalent aerosol optical thickness, annual composite 1989-1991. (Image courtesy NOAA.)
emissions, for example, anthropogenic sources, bio- observational studies have addressed the long-term
mass burning, mineral dusts, changes in convection, change in sea-salt aerosol.
dispersion (wind patterns and thermal structure of the Smoke aerosol loading is also highly variable,
atmosphere), and removal (precipitation). Other size reflecting local seasonal patterns of biomass burning.
fractions such as PM2.5 (mass of particles with Prescribed and uncontrolled fires occur in all vegetated
D , <2.5 pm) and particles of submicrometer diameter ecosystems around the globe, although statistics on
are increasingly studied for health and climate burning are relatively poor. Major fire regions include
research, but observations currently have much more Africa (10"-20" S) (August-October maximum); Bra-
limited spatial and temporal cover. Historical long- zil (July-September maximum); sporadically through
term mass data are sparse, requiring the use of proxy Indonesia (dry season, typically June-October) and
records. Observations of visual haze over the eastern northern Australia (May-October maximum); and
United States from the 1950s to the 1980s, for northern boreal forests, e.g., SiberiaKhina (spring
example, show a summertime increase in light- maximum, also autumn). Concentrations depend on
extinction coefficient but a reduction during winter the proximity to sources, and effectiveness of dis-
over the 1970s and 1980s. Primarily these reflect mass persal, with mass loading exceeding 1mgm-3 near
changes in the 0.1 to 1pm diameter range. large fires. Residual black carbon mass loading less
Although it may be feasible to assemble a mean- than 5ngmP3 (comprising roughly 10% of smoke
ingful global climatology from the diverse urban and mass) is frequently observed in well-scavenged marine
regional network measurements of aerosol mass, none air. Smoke aerosol transported over long distances is
exists at present. Different regions and subregions often present in well-defined layers up to around 5 km.
have individual patterns and many developing net- Few long-term measurements of biomass burning
works will in time allow assessment of trends. aerosol exist and historic records are sparse. Emissions
Contributions to the mass from different sources have been estimated to have increased 2-fold to 3-fold
also vary uniquely. Oceanic sea-salt aerosol produc- over the past century.
tion, for example, depends strongly on wind speed. Mineral aerosol is also highly regional, and season-
Within the marine boundary layer, sea-salt typically ally dependent. Satellite remote sensing affords daily
dominates aerosol mass, although mineral, smoke, measurement and mapping of the spread of aerosol
and anthropogenic aerosol can be regionally impor- from individual events. Longer-term records are now
tant, as shown in Figures 1-4. Representative sea- compiled from satellite analyses (e.g., AVHRR,
salt aerosol mass concentrations lie in the range TOMS) and some historic context can be inferred
1-50 pg m-3. Concentrations can change rapidly from ships' observations from the late 19th century.
and strongly with synoptic events, and seasonal Quantitative measurement of aerosol from in situ
variation is typical with a local winter concentration surface sampling has been the primary source of
maximum. Seasonal amplitude ranges vary from less information on composition, mass loading, lifetime,
than about 1 . 5 1 (Southern Hemisphere mid-lati- and variability. Studies show substantial variability
tudes) to greater than 8:l (North Atlantic). Surface even during a single day, with concentrations changing
observations are the primary source of information on up to several orders of magnitude. The spatial
sea-salt aerosol intrinsic properties. Few, if any, distribution of mineral aerosol is governed primarily
18 AEROSOLS / Climatology of Tropospheric Aerosols
~ ~ ~ ~
by the extent of source regions and climatological a similar range is found on the Antarctic coast, around
wind patterns. Thus, major dust areas include the 50-600 cmP3at Mawson, for example, but the range is
Atlantic Ocean downwind of the Sahara, the western much smaller at the South Pole with a seasonal cycle
Pacific downwind of China, and the Arabian Sea. around 10-150 crnp3. A smaller range is also found in
Northern Hemisphere mineral aerosol fluxes are an the Tropics; for example, at Samoa the monthly mean
order of magnitude greater than those in the Southern particle number ranges from about 200 to 300 cmP3.
Hemisphere owing to the location of active dust source Photoxidation of DMS, emitted by marine phyto-
regions. plankton, is considered the major natural source of the
As with smoke aerosol, the long-range transport of climatically important submicrometer aerosol par-
mineral aerosol frequently occurs in well-defined ticles over the global oceans.
layers in the lower 5 km of the atmosphere. Mineral In north-eastern Atlantic air (e.g., Mace Head,
aerosol concentrations are strongly dependent on Ireland) the number concentration in clean air ranges
season and distance from the source regions. Aerosol from about 100 to 700 cmP3 with no seasonal cycle,
from the dust areas in East Asia is observed most although a summer peak is found in methyl sulfonate,
frequently in spring (March-April-May ) . Lowest an oxidation product of DMS, the principal natural
concentrations occur in summer and winter when precursor. Long-range transport from the United
dust production and transport mechanisms are unfa- States may inhibit a winter number concentration
vorable. In spring, boundary-layer concentration of minimum. Closer to major continental source re-
mineral aerosol ranges from several hundred pg m-3 gions, for example, over the north-western Atlantic
near the Chinese coast to values up to about 9 pg m-3 and north Pacific, concentrations are more typically
at Mauna Loa (on Hawaii). The central Pacific has a 4000-6000 ~ m - ~ .
relatively small mineral aerosol loading, whereas The most widely quoted particle concentrations for
moderate concentrations of dust from Australian continental sites date to the first half of the 20th
deserts are observed over the south-western Pacific in century and may be geographically biased; concentra-
the Austral spring and summer. Saharan and sub- tion values around l o 4 cm-3 are typical of areas away
Saharan mineral aerosol are transported across the from population centers. In truly clean continental
Atlantic, with a maximum occurrence in June-August, areas (for example, across Australia), particle con-
and observed in both North and South America. Daily centrations are more typically around 700 cmP3 in the
average concentrations of mineral aerosol in excess of boundary layer and around 2 0 0 ~ m - above ~ the
100 pg mP3 have been recorded in Miami and monthly boundary layer.
mean concentrations in the dust season regularly Arctic aerosol and its associated haze have been
exceed 20 pg m P 3 in Miami and at Barbados. Con- studied intensively. Typically a phenomenon of the
centrations in the dust plumes over the Atlantic may be lowest 3 km, occasionally up to 5 km, haze events
as high as 400 pg mP3. result from the long-range transport of anthropogenic
aerosol from Eurasia and North America. Particle
numbers (monthly mean) at the surface are typically
50-400 ~ m - ~occasionally
, up to 800 cmP3, with
Particle Number
spring and summer maxima, whereas light scattering
The integral number concentration was one of the first (and absorption) peak in spring and winter. Recent
aerosol properties explored in the late 19th century, analyses for Barrow, Alaska, indicate a significant
with the development by John Aitken of a portable positive trend in particle number (around 8 % per year
‘dust counter’. Although essentially the same type of since 1976) and a decreasing trend in spring light-
measurement is used today, systematic long-term scattering coefficients. Natural sulfur sources are
records are a relatively recent phenomenon. Total important in summer.
particle number concentrations vary from less than Above the boundary layer, number concentrations
10cmP3 over the Antarctic Plateau in winter, and are lower. For example, in Hawaii (Mauna Loa,
infrequently at mid-latitudes in polar outbreaks, to 3.4 km) monthly mean concentration is typically
around lo5 cmP3 in large cities. Concentrations 260 cmP3; there is a weak seasonal cycle (concentra-
decrease with altitude, to around 100-200 cmP3 in tion range around 1.5:l) with a September maximum.
the upper troposphere. Over clean southern ocean A trend of +1.7% per year was seen in the 1970s
regions, concentrations are typically several hundred and 1980s. Similarly, in free tropospheric air over
~ m - varying
~, seasonally, with a summer maximum. the European Alps, (JungFrauJoch, 3.6 km), typical
The amplitude of the seasonal cycle is latitude summer concentrations are around 800 cmP3 and
dependent. At southern mid-latitudes, monthly med- about half of this concentration in winter. Across
ian concentrations of 100-600 cm-3 are observed, and eastern Europe, average particle concentrations of
AEROSOLS / Climatology of TroposphericAerosols 19
1 0 0 - 5 0 0 ~ m - ~are typical at 3 km (mid 1900s). Over Limb-scanning satellite data (from 1984 to 1991)
the continental United States during the.l98Os, at alti- give an extensive depiction of the upper-tropospheric
tudes of 2.5-5 km, number concentrations were typi- aerosol via the light extinction coefficient. The data
cally 1800 cmP3 in summer, and 700 cmP3 in winter, have also been used to infer aerosol mass concentra-
with an overall negative trend around 5% per year. tion and effective size in the submicrometer size range.
Longer-term measurements usually show interann- These extinction data and in situ determinations show
ual variations in concentration. In Southern Hemi- a relatively uniform aerosol between about 5 km
sphere oceanic regions there is evidence of varia- altitude to near the tropopause. In both hemispheres
tions with decadal time scales, although records (20"-40") a seasonal cycle is seen in light extinction,
are relatively short (Figure 5 ) . At Cape Grim (41"s with a maximum in the spring-summer period when
145" E) annual-average particle concentration particle size is also maximum. In the Northern
increased by an average 1.4+0.3% per year from Hemisphere the spring period is characterized by a
1976 to 1998. Historic records comprising full maximum in mineral aerosol production and trans-
seasonal data at regionally representative sites do not port. At mid-latitudes in Australia, balloon-borne
exist. Proxy data, specifically small-ion conductivity, measurements show marked increases in particle
suggest that number concentrations in the northern concentrations in the entire troposphere during spring,
Atlantic may have approximately doubled in the 60 probably linked to long-range transport of biomass
years to 1970. Similar (sparse)records from the south burning emissions. Enhanced concentrations are also
Pacific suggest little or no change in this period. found over the south-western Pacific during this
season. At southern mid-latitudes in spring, mixing
ratios of submicrometer aerosol are comparable to
those at northern mid-latitudes in summer, represent-
Upper Troposphere (>5 km) ing an enhancement of more than an order of
Compared to the lower Skm, surprisingly few magnitude over other periods.
systematic long-term studies have addressed the In general, significantly more light extinction and
upper-tropospheric aerosol. Methodologies that mass are observed in the Northern Hemisphere,
access this region include satellite-based light-extinc- particularly at mid-latitudes and polewards, and the
tion measurement, in situ balloon-borne samplers, seasonal cycle in extinction is about five times greater
high-altitude aircraft, and various lidar (laser radar) in the Northern Hemisphere. AVHRR data indicate
systems. The widest geographic cover is from satellites around 50% more column light extinction in the
such as SAGE-2 (Stratospheric Aerosol and Gas Northern Hemisphere than the Southern Hemisphere.
Experiment), looking at long-path light extinction Multidecadal measurements of particle size and
(limb-scanning) during sunrise and sunset. These number at northern mid-latitudes, made from bal-
provide information on extinction coefficients at loon-borne samplers, also show a spring-summer
1Fm and submicrometer wavelengths and indirect peak in an annual cycle of optically active particles,
data on particle sizes. Balloon-borne measurements where concentrations vary by close to an order of
over more than two decades have produced consistent magnitude. For total particle number, the seasonal
in situ particle information in both hemispheres. variation is smaller, around a factor of 3, with a
summer peak. All of these observations include
quiescent and enhanced conditions; in situ aerosol
1000 L I I backscatter measurements at mid-latitudes from only
quiescent periods indicate relatively little interhemi-
7
n
r 800- spheric difference.
E
0
v
.-+ 600-
E
See also
3
C
400; Aerosols: Observations and Measurements: Physics and
c Chemistry of Aerosols; Role in Cloud Physics; Role in
8 Radiative Transfer. Biogeochemical Cycles: Sulfur
200 -
Cycle. Deposition. Dust. Global Change: Biospheric
I
Impacts and Feedbacks. Lidar: Backscatter. Observa-
0 ~ " " " " " " ~ tion Platforms: Balloons. Observationsfor Chemistry
76 78 80 82 84 86 88 90 92 94 96 98 00
(In Situ): Particles. Satellite Remote Sensing: Aerosol
Year, 1976-2001
Measurements. Soot. Tropospheric Chemistry and
Figure 5 Monthly median concentration of particles in clean Composition: Aerosols/Particles. Volcanoes: Role in
Southern Ocean air at Cape Grim Tasmania, 1976-2001. Climate.
20 AEROSOLS I Observations and Measurements
Further Reading Hobbs PV (ed.) (1993) Aerosol-Cloud-Climate Interac-

tions. San Diego: Academic Press.
Knott CG (ed.) (1923) Collected Scientific Papers of john Charlson RJ and Heintzenberg J (1995) Aerosol Forcing of
Aitken L.L.D, E R.S. Cambridge: Cambridge University Climate. Chichester: Wiley.
Press. Henderson-Sellers A (ed.) (1995) World Survey of Climatol-
McDonald WF (1938) Atlas of Climatic Charts of the ogy, vol. 16, Future Climates of the World: A Modelling
Oceans. Washington, DC: US Department of Agricul- Perspective. Amsterdam: Elsevier.
ture, W.B. no. 1247. Houghton JT, Meira Filho LG, Callandev BA, et al. (eds)
Junge CE (1963) Air Chemistry and Radioactivity. New (1996) Climate Change 1995, The Science of Climate
York: Academic Press. Change. Cambridge: Cambridge University Press, for the
D’Almeida G, Koepke P and Shettle EP (1991)Atmospheric International Panel on Climate Change.
Aerosols Global Climatology and Radiative Character- Pruppacher HR and Klett JD (1997)Microphysics of Clouds
istics. Hampton, VA: A Deepak Publishing. and Precipitation. Dordrecht: Kluwer Academic.
Gotz G, MCsziros E and Vali G (1991) Atmo- Koepke P, Schult I and Shettle EP (1997) Global Aerosol
spheric Particles and Nuclei. Budapest: AkadCmiai Data Set. Hamburg: Max-Plank-Institut fur Meteorolo-
Kiad6. gie Report No. 243.
Observations and Measurements

P H McMurry, University of Minnesota, Minneapolis, ably emitted directly into the atmosphere as sea spray,
MN, USA while the sulfate crystals attached to it may have been
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. produced in the atmosphere by chemical transforma-
tions. The rutile in Figure 2C probably originated as
wind-blown dust and may have accumulated the
introduction sulfate coating as a result of atmospheric processing.
Atmospheric aerosols consist of a suspension of solid The variabilities in shape and composition illustrated
or liquid particles in air. These particles range from by the photographs in Figure 2A-C introduce ambig-
molecular clusters less than one nanometer in diam- uities in aerosol measurements since the responses of
eter to pollens, wind-blown soil dust and sea salt of particle measuring instruments depend on such prop-
lOpm or larger. Figure 1 shows an idealized atmos- erties. In addition, particles often contain volatile or
pheric aerosol size distribution with the characteristic
modes in which particles tend to be found. Homoge-
neous nucleation leads to the clustering of molecules
that grow to particles several nanometers in diameter
at the small end of the spectrum. Aitken or nuclei mode
particles in the 10 to 100nm diameter range often
contain ‘primary’ particles that are directly emitted
into the atmosphere by combustion processes. Accu-
mulation mode particles contain most of the submi- Accumulation
cron mass. Most of this mass is typically ‘secondary’ in
origin (i.e., formed in the atmosphere by chemical
transformations). Sulfates, nitrates, and organics typ-
ically comprise most of the accumulation mode mass,
and this mass can accumulate in different submodes,
depending on the chemical mechanisms by which they
are formed. Coarse-mode particles tend to be pro-
duced by mechanical processes.
Figures 2A-C show photographs of atmospheric 0.001 0.01 0.1 1 10 102
particles obtained by electron microscopy. The chain- Particle diameter (km)
agglomerate soot particle in Figure 2A is typical of
Figure 1 Idealized atmospheric aerosol size distribution. Parti-
particles emitted by diesel engines or other combus- cles tend to be found in modes that reflect the different processes
tion sources. Figures 2B and C illustrate particles of by which they are formed in the atmosphere by chemical transfor-
mixed composition. The salt in Figure 2B was prob- mations or are emitted directly into the atmosphere from sources.
AEROSOLS / Observations and Measurements 21
Figure 2 Photographs of atmospheric particles obtained by transmission electron microscopy. (A) Chain agglomerate soot particle.
(B) Sulfate attachedtoa sodium chloride particle. (C) Mineral dust particlecoated with sulfate. (Reproducedwith permission.Aand Cfrom
Buseck PR, Jacob DJ, Posfai M, Li J, and Anderson JR (2000) Minerals in the air: an environmental perspective. lnternationalGeology
Review42,577-593, Q V.H. Winston & Son, Inc. 360 South Ocean Boulevard, Palm Beach, FL 33480, USA. All rights reserved.; B from
Buseck PR and Posfai M (1999) Airborne minerals and related aerosol particles: effects on climate and the environment. Proceedings of
the National Academy of Sciences of the United States of America 96: 3372-3379. Q 1999, National Academy of Sciences, USA.)
semivolatile species such as water, nitric acid, ammo- densed and gas phases, and the gadparticle
nia, and various organic compounds that are also distribution can be perturbed during sampling.
present in the gas phase. This adds another challenge This article summarizes the most important meth-
to obtaining accurate measurements, since such spe- ods for measuring atmospheric aerosols. It first
cies undergo continuous exchange between the con- discusses aerosol sampling and chemical analysis of
22 AEROSOLS / Observations and Measurements
aerosol, then concludes with methods used to measure exceeds the true ambient concentration (Figure 3A),
physical properties of particles such as particle size and while the reverse is true when the wind speed is less
number concentration. Throughout, discussion focus- than the sampling speed (superisokinetic sampling;
es on in situ particle measurements; a great deal of Figure 3B). An ideal isokinetic sampler is aligned with
valuable information on global distributions of aero- the wind and draws in air at a speed equal to the wind
sols has been obtained with satellites, but such speed. In practice, wind speeds and directions vary, so
measurements are not discussed here (see Satellite achieving this ideal with samplers at a fixed location is
Remote Sensing: Aerosol Measurements). impractical.
It might appear that an aircraft would provide an
ideal platform for sampling coarse particles. Isokinetic
Sampling sampling would be achieved automatically if the
sampler pointed straight ahead and sampled at a
Accurate aerosol measurements require the collection velocity equal to the aircraft velocity. Unfortunately,
of a representative sample. Obtaining representative this is not the case. Even if the probe sampling velocity
samples of particles smaller than 1pm is usually not equals the aircraft speed, variations in the aircraft
too difficult. Under typical sampling conditions par- orientation lead to misalignment between the probe
ticles in this size range travel with the gas. Difficulties and the sampled airstream, and the high sampling
may be encountered with particles smaller than about velocities required lead to turbulent deposition of
lOnm in diameter since, because of their high particles within the sampling probe. Thus, it can be
diffusivities, they can be lost on the walls of sampling difficult to transport particles through the probe to the
tubes during transport to the detector. It is usually measuring instrumentation. The design of aircraft
possible to design sampling systems that minimize this sampling inlets that deliver a known fraction of larger
effect. particles is a current area of research.
Supermicron particles are difficult to sample. Such Omnidirectional sampling inlets are typically used
particles have so much inertia that they tend not to for fixed-point sampling. Such inlets draw aerosol
follow the gas flow as it enters the sampler. Figure 3 from all directions through a horizontal annular
shows trajectories of large particles as they are drawn opening. Sampling efficiencies for omnidirectional
into an aerosol sampler. When the wind speed exceeds inlets are independent of wind speed, and can be nearly
the velocity of the air flow into the sampling tube 100% for particles up to 10 pm.
(subisokinetic sampling), the sampled concentration In summary, obtaining representative aerosol sam-
ples for particles smaller than a few micrometers in
diameter is usually straightforward. Obtaining repre-
Subisokinetic sampling sentative samples for particles larger than this is
difficult since, owing to their inertia, large particles
+ =
............. tend not to follow air as it flows into a sampler.
-+- ............. Sample Collection and Analysis:

Filters and Impactors
Superisokinetic sampling Filters are the device most commonly used for
collecting particles for analysis. Samples are collected
by removing particles from a known volume of air, and
are analyzed for particulate mass or species concen-
~r trations. Electron micrographs of several types of filter
-+- A
media commonly used to sample atmospheric parti-
cles are shown in Figure 4A-C. All of these media
contain passages through which the gas flows and
(B) solid substrates on which particles are collected.
Figure 3 Trajectories of large particles as they are drawn into an Filters collect particles by three different mecha-
aerosol sampling probe. (A) Subisokinetic sampling occurs when nisms, as illustrated in Figure 5. Very small
the velocity of the sampled aerosol at the probe entrance is less particles (diameter less than about 0.1 pm) are col-
than the velocity of the wind relative to the probe. In this case the
concentration of ‘large’ particles in the sampled aerosol is higher
lected by diffusion, and because diffusivities increase
than the ambient concentration. (B) For superisokineticsampling, with decreasing size, particle collection efficien-
the sampled concentration is less than the true concentration. cies increase with decreasing size. Large particles
AEROSOLS I Observations and Measurements 23
Figure 4 Electron micrographs of several types of filter media commonly used to sample atmospheric particles. The filter media
illustrated in these figures include (A) a 3.0 pm Teflon membrane filter, (B) a glass fiber filter, and (C) a 1.O pm polycarbonate membrane
nuclepore filter. (Courtesyof Professor Benjamin Y. H. Liu, Departmentof Mechanical Engineering,University of Minnesota, Minneapolis,
MN.)
(diameter >-1 pm) are collected when, due to their the filter are typically much larger than the sizes of the
inertia, they are unable to follow the gas as it flows collected particles. All filters have a ‘most penetrating
around a solid surface (e.g., a fiber) and impact on the size’ that falls typically in the 0.05 to 0.3 pm diameter
fiber. Because inertia increases with increasing particle range, where particle collection by diffusion, inter-
size, filter collection efficiencies increase with size in ception, and impaction reaches a minimum. For filters
this range. Intermediate-sized particles are collected typically used for air filtration, however, collection
by interception when the gas flow carries them to efficiencies even at this size usually exceed 99%.
within one particle radius of the filter medium. Note Therefore, it is generally reasonable to assume that all
that ‘sieving’ is not a significant particle collection particles delivered to the filter are collected. Filters are
mechanism for air particle filters; as can be seen in typically operated downstream of an inlet that re-
Figure 4, the dimensions of the air passages through moves all particles larger than a specified size.
24 AEROSOLS I Observations and Measurements
Interception work is underway to extend this methodology to

\ Diffusion
various organic compounds. Sampling such com-
pounds presents a challenge because they can volatil-
-b
ize from the filter during sampling, thereby leading to
-b an underestimate of the particle-borne concentration.
-b
-b A schematic of a diffusion denuder sampler is shown in
-b
-
Figure 6. The aerosol is first drawn through an annular
/ - denuder, the walls of which are coated with a chemical
Irnpaction that reacts readily with the gas to be removed. Basic
Figure 5 Filters collect particles by three different mechanisms. compounds such as sodium carbonate are often used
Diffusion is primarily responsible for the collection of particles to remove acidic gases such as sulfur dioxide and nitric
smaller than about 0.1 pm. Inertial impaction, which causes acid, and acidic compounds such as citric acid are used
particles to cross air streamlines and deposit on the filter substrate, for ammonia. Owing to their high diffusivities, gases
is responsible for the collection of particles larger than about 1 pm.
Interception,which occurs when the air streamline brings a particle diffuse to the coated walls and are removed with
to within one particle radius of a filter substrate, is important for nearly 100% efficiency. Particle diffusivities are orders
particles in the intermediate size range. of magnitude lower than those for gases, so particles
pass through the denuders to a filter where they are
collected with nearly 100% efficiency. Volatilization
Substantial effort has been made to develop diffu- from the collected particles occurs during sampling,
sion denuder filter samplers for measuring the gas and and a suitable collector is located downstream of the
particle phase concentrations of semivolatile species. filter to collect these volatilized compounds. For
The design of such samplers for inorganic species example, nylon filters are known to collect nitric
including ammonium nitrate is well developed, and acid vapor with nearly 100% efficiency, and porous
H+, NH,: so:-, NO;
Citric acid
NH,
coating
Annular
denuders Na2C0, NO2, HNO,, NHO,, SO2
for basic and coating
acidic gases
NaCl HNO,, SO2

coating
Impactor to remove
particles 72.5pm
I I\ I
YYt
Aerosol inlet
Figure 6 Schematic of a diffusion denuder sampler for collecting the gas and particle phases of semivolatile compounds.
Next Page
AEROSOLS / Observations and Measurements 25
polyurethane foam is used to collect certain organic Aerosol inlet

vapors. The particle phase concentration of the
semivolatile compound is found by adding the load-
ings collected by the filter and the downstream
adsorber. Several different approaches are used to
measure the gas phase concentration. In some cases the
collected gas can be extracted from the diffusion
denuder and analyzed directly. Alternatively, two
samplers can be operated in parallel, one with a
diffusion denuder and one without. The sampler
without the diffusion denuder collects the total gas-
plus-particle concentration. The gas phase is then
determined by difference.
Cascade impactors are used to collect size-segregat-
ed samples for subsequent analysis. A schematic
diagram of a cascade impactor is shown in Figure 7.
The aerosol is first accelerated through a large nozzle
towards a collection substrate that deflects the air and
collects the largest particles by inertia. The aerosol is
then transported to a stage with a slightly smaller
nozzle that collects smaller particles due to the higher
speed of the gas through the nozzle. Impactors can
provide a very sharp size separation down to sizes as
small as 10 nm. Measurements of solid particles can be
adversely affected by bounce, which can be reduced by
greasing the collection substrates. A limitation of
impactors is that no strategy has been developed to
measure evaporative losses of semivolatile compounds
from the individual impactor stages.
Virtual impactors replace the collection substrate
with a receiving tube that draws a small fraction of the
flow and into which all particles larger than the cut size
are projected. Virtual impactors avoid bounce and use
conventional filter substrates, but have the disadvan- Figure 7 Schematic diagram of a cascade impactor. These
instruments separate particles into several different size fractions
tages that they cannot be designed with cut points that for subsequent measurement of size-resolved mass or species
are as sharp or as small as can be achieved with concentrations.
conventional impactors. Cascade impactors segregate
particles according to aerodynamic diameter, which
depends on shape, physical size, and particle density. particles can be collected directly in liquid water and
Samples collected with filters or impactors are typi- analyzed by ion chromatography to determine concen-
cally analyzed to determine particulate mass or species trations of major aqueous ions, including sulfate,
concentrations. Table 1 provides a summary of typical nitrate, ammonium, chloride, etc. When employing
minimum detection limits that can be achieved by denuders these methods can also be used to measure
commonly used analytical methods. gas phase concentrations of semivolatile compounds.
Other approaches involve collecting particles by im-
paction on a substrate from which they can be volati-
Real-Time, In-Situ Analysis of Aerosol lized by rapid heating and analyzed by a conventional
gas analyzer. These semicontinuous samplers provide
Chemical Properties species concentrations with a time resolution down to
Filtration with subsequent analysis in the laboratory is 3 min and show that aerosol concentrations often vary
by far the most commonly used approach for deter- significantly over time scales required for filter samples.
mining the composition of atmospheric aerosols. A tremendously exciting new development in aero-
Considerable progress has been made in the last few sol measurement involves the use of mass spectrome-
years, however, on techniques that provide real-time try to measure the composition of individual particles
information on aerosol composition. For example, in real time. Particles are drawn through an inlet into a
54 AGRICULTURAL METEOROLOGY AND CLIMATOLOGY
E S Takle, Iowa State University, Ames, IA, USA Agriculture is arguably the most weather-sensitive
Copyright 2003 Elsevier Science Ltd. All Rights Reserved sector of society. Forty percent of the land surface of
the Earth is classed as arid, semi-arid, and dry
subhumid but is home to millions of people, particu-
larly in developing countries. For some of these areas,
Introduction
frequent crop failure due to adverse weather must be a
Agricultural meteorology is an interdisciplinary part of interannual planning by farmers and regional
science concerned with discovering, defining, and and state policy makers. International agricultural
applying knowledge of the interactions between me- meteorologists, therefore, have significant concern
teorological and hydrological factors and biological for famine and food security because of their close
systems to practical use in agriculture. An ultimate link to interannual variability of weather and climate.
goal of agricultural meteorology is to extend and fully Even in areas having what are considered ideal
deploy knowledge of atmospheric and related pro- climates for crops, water management is a major
cesses to optimize agricultural production, and hence concern, particularly in regions where competing uses
to increase profitability, decrease risk, and feed an of fresh water put increased pressure on agricultural
expanding global population. A second goal that is uses of water.
taking on increased importance is to help conserve The historical focus of agriculture to produce food
natural resources and protect our soil, plant, and and fiber for an expanding global population has been
water resources. supplemented by a new thrust at the beginning of the
Environmental interactions of a wide range of twenty-first century. Agriculturists now also have roles
agriculturally related organisms are of interest to in managing soil and landscapes to regulate flows of
agricultural meteorologists. Although most attention carbon, nutrients, soil amendments, and pesticides.
has been focused on agricultural and horticul- Atmospheric transport of pesticides, spores, and
tural crops and forests, this segment of atmospheric pollens (particularly those originating from genetical-
science also includes environmental interactions ly modified plant materials) must be quantified with
with animals grown to provide food and fiber, insects, increased accuracy. Although uncertainty remains
plant and animal pathogens, and aquaculture species. large, consensus estimates of sources and sinks of
Agricultural meteorology, like the entire field of greenhouse gases such as carbon dioxide, methane,
meteorology, has its roots in the study of temperate and nitrous oxide reveal that agriculture has a signif-
(mid-latitude) regions of the Northern Hemisphere. icant role. For instance, agricultural sources of meth-
In parallel with its parent discipline, agricul- ane from ruminant animals, rice production, and
tural meteorology has more recently intensified its biomass burning are comparable to, or may exceed,
focus on tropical agriculture, with some of the natural emissions on a global scale. Agriculture may
same difficulties of paucity of data faced by tropical play a significant role in moving society from its fossil-
meteorology. fuel base for energy and materials to one that relies
The vagaries of weather always have been a leading more heavily on bio-energy and bio-based materials.
cause of variability in agricultural production, Tillage practices on natural prairie lands have reduced
but the technological era has increased this vulnera- soil carbon by up to 50% in the US Midwest.
bility even as it has provided some means of insulating Opportunities for agricultural recapture of soil carbon
agriculture from adverse conditions. So, for instance, by use of high-yield plant varieties, reduced tillage,
disease-resistant crops, wide availability of soil and improved management of crop residues, fertiliza-
amendments and chemicals for pest control, and tion, and irrigation are under consideration.
efficient tillage, planting, and harvesting equipment Most biological and chemical processes in the
have reduced agricultural vulnerability and increased biosphere are highly temperature- and moisture-
yields; however, larger fields and wide use of mono- dependent, and meteorology is the study of underlying
cultures have exposed crops to vector-borne diseases physics and chemistry that governs these processes.
and insects, and exposed soil to erosion by wind and Emerging recognition of the importance of bio-com-
water. The use of chemicals, new varieties, and plexity and ecosystem services and the need for
genetically modified organisms has brought new sustainable methods of agriculture and economic
weather dependencies. development are creating new roles for agricultural
AGRICULTURAL METEOROLOGY AND CLIMATOLOGY 55
meteorology. Agricultural meteorologists, therefore, matter per megajoule. Chlorophyll in leaves makes
can be expected to play an increasingly larger role in plants much less reflective in the PAR region than in
working with scientists from many disciplines to meet the near-infrared (by a factor of 4 for corn and
the challenges of these new environmental concerns. soybeans), a fact that allows remote assessment of
photosynthesis by use of the normalized differ-
ence vegetation index (NDVI) derived from satellite
Fundamental Principles observations.
Radiation
Heat Balance
Agricultural meteorology is built on a foundation of
Thermodynamic principles provide the basis for rela-
fundamental physical laws with applications to the
tions among atmospheric pressure, temperature, and
plant, animal, and soil environments. The principles of
density (ideal gas law), as well as the transfer and
radiation describe how radiant energy received direct-
conversion of energy (first law of thermodynamics). A
ly from the sun and in diffuse form from the atmos-
phere is made available to plants for photosynthesis primary focus of agricultural meteorology is the
balance of energy (conservation of energy) for the
and converted by solid and liquid surfaces into other
system being studied, such as a metabolizing organism
forms of energy. Approximately 40% of the radiation
or a plant-covered or soil surface. For an organism we
emitted by the Sun is in the visible wavelength band
can describe the steady state heat balance by eqn [2].
from 0.4 to 0.7 pm, sometimes referred to as the short-
wave band or, for biological applications, as the band
of photosynthetically active radiation (PAR). Radia-
R, + M = C + LE + G [21
tion with wavelengths just shorter than those in the In eqn [2], R, is the net gain of heat from radiation; M
PAR zone is called ultraviolet radiation (UV-A from is the net gain of heat from metabolism; C is the loss
0.32 to 0.40 pm, and UV-B from 0.28 t o 0.32 pm). of sensible heat to air by convection; LE is the loss of
Ultraviolet radiation is not used by plants for photo- latent heat by evaporation of water; G is the loss of
synthesis but it can damage living tissue of plants and heat to ground and vegetation by conduction. E is the
animals, particularly simple organisms. evaporation rate (flux of water vapor per unit area),
Visible radiation may arrive at a leaf surface either and L is the latent heat of vaporization. All quantities
directly from the Sun or indirectly by reflection from are considered to be averaged values per unit area. For
atmospheric molecules, clouds, or solid or liquid applications to animal agriculture, M is likely to be a
surfaces (including other plant leaves). Leaf orienta- significant factor, but for a soil or vegetated surface,
tions and solar zenith angle cause the amount of the metabolism contribution is negligible. The radiant
radiation received to vary over the course of the day. component of energy consists of absorbed incoming
The fraction of plan area of leaves to ground area short-wave energy less net emitted long-wave energy.
covered by the plant (including only one side of the A plant canopy uses a portion of the short-wave
leaf) is called the leaf area index (LAI) and is used to component of this net incoming radiant energy for
describe the area of the plant available for photosyn- photosynthesis. In a balanced condition, the plant uses
thesis. its evapotranspiration capacity to regulate its temper-
Radiation of wavelengths just larger than visible ature by converting excess sensible heat to latent heat.
light is called long-wave or infrared radiation, with the Most agricultural animals, like humans, also have the
band from about 0.7 to 1.5 pm being referred to as the capacity to rid themselves of excess heat by means of
near-infrared region, which accounts for about 40% evaporation.
of the solar spectrum. Wien’s displacement law (eqn
[I]) relates the wavelength of radiated energy to Surface Aerodynamics
temperature. The aerodynamics of plant interactions with the
I 2897 atmosphere provides a basis for understanding how
A=- plants exchange moisture, trace gases, and heat energy
T
with and extract momentum from the free atmosphere
In eqn [l], 3, is the wavelength (in pm) and T is the through turbulent processes. Descriptions of the
temperature in K. Wien’s law can be used to show that movement of pollen, spores, insects, and chemical
most terrestrial surfaces emit radiation of 8-12 pm, sprays also require information about mean and
with most growing plants radiating at about 10 pm. turbulent flow processes on scales of centimeters to
The amount of dry matter produced by a plant per unit hundreds of kilometers.
absorbed PAR is a measure of its light-use efficiency Simple representations of atmosphere-surface in-
(LUE), with typical values being from 1.5 t o 4.5 g dry teractions are given by drag coefficient formulations of
56 AGRICULTURAL METEOROLOGY AND CLIMATOLOGY
vertical fluxes of quantity S from a surface as given in coefficient for moisture, p = u*/CD, RH, is the
eqn 131. relative humidity at the plant or soil surface,
RH, is the ambient relative humidity, u* is the
F, = - U t ( & - Ssurf 1 131 friction velocity, and CD is the drag coefficient; and
Ut is the transport velocity for the interface and the y = cp/L, where cp is the specific heat capacity of
values of S are taken at height t and at the surface. The air at constant pressure. In some implementations, A
transport velocity at the Earth’s surface is usually and F, are replaced by factors that include canopy
parametrized by use of a drag coefficient ( CD,)for the and atmospheric resistances to flow of heat and
quantity S and the mean wind speed at some level momentum.
} [4]).
(usually taken to be lorn), i.e. { V ~ O(eqn Both amount and timeliness of precipitation and
evaporation are of critical importance to agriculture.
ut = cDs{ VIO} 141 Irrigation scheduling requires reliable climate infor-
mation as well as good weather forecasts, particularly
Drag coefficients depend on atmospheric stability with increased competition for fresh water due to
but are typically in the range 1x1OP3 to S X ~ O - increased
~ population and expanded uses of water.
(dimensionless). Food security exacerbates the vulnerability of many
Concepts of gradient or Fickian diffusion have been precipitation-deficient developing countries to inter-
used to describe fluxes by measuring vertical gradients annual variability of precipitation and raises the
and using assumptions or additional measurements to urgency of improved seasonal to interannual forecasts
estimate transfer coefficients. Under this approach, the of weather and climate.
turbulent flux of a quantity is proportional to the
vertical gradient of its mean quantities above the
surface, eqn [5].
Instrumentation, Measurements,
as and Networks
F, = -K,- 151
az Agricultural climatology relies on records of basic
where K, is the turbulent diffusion transfer coefficient meteorological measurements having been taken over
for variable s, usually estimated to be ku*z (kbeing von extensive areas and significantly long periods of time.
Karman’s constant (0.4) and u* being the friction These records form the basis for understanding
velocity) with an additional stability correction factor climate variability and change and also for extracting
and constant for each variable s. Equation [5] with an statistically significant relationships between meteor-
assumed form of K , is used to derive vertical profiles of ological variables and soil and plant processes, plant,
temperature and horizontal windspeed over homoge- animal, and pest development, and seasonal yield. In
neous surfaces. Profiles inside crop canopies are more addition to standard atmospheric measurements,
complicated and are usually specified by empirical agriculturists need measurements of soil temperature
relations. and soil moisture. These measurements are less widely
recorded although they (especially soil moisture) are
Evaporation and Precipitation being recognized for their role in climate memory and
Agriculture is practiced over large regions of the Earth hence seasonal forecasting. More such measurements
where water excess or water deficit is a major and networks for measurements are needed, particu-
limitation for successful crops. Therefore, a major larly in developing countries where use of technology
focus of agricultural meteorology and climatology is to reduce vulnerability to climate variability is severely
the study of precipitation and evaporation. The heat limited.
balance equation can be used to provide an estimate of The central role of the surface energy balance in
the evaporation rate for a surface from knowledge agricultural meteorology calls for accurate methods of
of other components of the heat budget by a type of evaluating fluxes of heat, momentum, moisture, and
Penman-Monteith equation (eqn [6]). trace gases from crop, soil, and forest surfaces.
Unfortunately, this is not an easy task for heterogene-
LE = A(% - H G ) + F w
ous surfaces typically encountered in agricultural
A+Y
16
1 applications. Estimates of surface fluxes can be
made: by drag coefficient formulations and gradient
In eqn [6] A = RH, aqs/aT, where RH, is the diffusion estimates, or by eddy correlation methods.
saturation relative humidity and q, is the saturation The most direct measurement of vertical fluxes is
specific humidity; HG is the soil heat flux; F, = accomplished by using eddy correlation methods,
C&(RH, - RH,), where CE is the bulk transfer which have seen increased use due to wider availability
AGRICULTURAL METEOROLOGYAND CLIMATOLOGY 57
of improvements in sensors and recording and in data appropriate averaging time must be selected to be long
archiving equipment and methods. Eddy correlation enough to ensure a sufficiently large sample but not so
methods are based on the principle that turbulent flow long as to mix turbulent processes with phenomena of
near the Earth’s surface can lead to vertical fluxes longer time scales. Perhaps a more serious problem is
of heat, moisture, momentum or trace gases in the the ‘representativeness’ issue: Fluxes at a point over an
absence of a mean vertical flux of dry air. We express agricultural field are not completely vertical, particu-
the vertical flux of quantity s as F, = c,pw(t)s(t),with larly where inhomogeneities exist in the field. It can be
a time-averaged value given by {F,} = csp{w(t)s(t)}, difficult to specify the ‘surface footprint’ from which
where c, is a constant for the particular quantity being the surface flux emerged for situations having chang-
transported, p is the dry air density, and w is the ing wind directions, terrain irregularities, changing
vertical wind speed. We can express w as a sum of a levels of atmospheric stability, or inhomogeneities of
time-independent mean and a time-dependent turbu- surface vegetation, soil moisture, or soil type. Year-
lent component, w ( t ) = wo + w ’ ( t ) ,and similarly for long measurements of COZ flux over a mixed-species
s, s ( t ) = so + s’(t). We can then write eqn [ 7 ] . forest in irregular terrain, for instance, would require
considerably more care in interpretation than daytime
measurements over a flat field of corn. Despite the
additional expense and care needed in conducting
measurements and additional effort for analysis, eddy
After time averaging, this becomes eqn [81. correlation measurements are increasingly being used
for evaluating surface fluxes of COZ and other trace
+
{F,} = c,p{wo~o} c,p{wos’(t)} gases and moisture.
Measurement networks have been established by
some local, state, federal, and international agencies to
The first two terms on the right-hand side of eqn [8] are provide both an expanding climate database and a
zero because the mean vertical wind speed is zero. The basis for near-term and seasonal agricultural decision
third term vanishes because, by definition, the mean making. There is an urgent need to expand these
fluctuation of the vertical wind is zero. The last term networks to meet the increasing food needs, particu-
can be nonzero, however, if the fluctuation of the larly in developing countries. Remote sensing by
vertical wind has correlation other than zero with the satellite is finding expanded use in providing large-
fluctuating part of s. The time-averaged turbulent flux scale data of relevance to agriculture, but its use for
of s then reduces to {F,} given by eqn [9], which can be individual farmers is limited.
computed by combining measured w’ and s’ taken
from simultaneous recordings of fast response meas-
urements of w ( t ) and s ( t ) . Modeling and Theory
{FSI = csp{w’(t)s’(t)) [91 Modeling of plant interactions with the atmosphere
has emerged from at least two directions: global
The Bowen ratio is defined as the ratio of heat flux to climate modelers seeking more accurate representa-
moisture flux near the surface (eqn [lo]). tion of energy, momentum, and moisture budgets at
the Earth’s surface, and crop modelers seeking ways of
understanding plant responses to climate and of
projecting yields of agricultural crops. Climate scien-
From eqn [2], ignoring metabolic contributions, we tists use the so-called soil-vegetation-atmosphere
can express the sensible heat flux and latent heat flux, transfer (SVAT)models as ‘surface packages’ to which
respectively, from the surface as in eqns [ll]and [12]. they supply meteorological data at each surface grid
point at each model time step (a few minutes to hours).
The SVAT model then calculates the response of the
soil and plants (e.g., evaporation or transpiration,
temperature change, soil moisture content, moisture
R, - G uptake by roots, rain or dew held on leaves, precip-
LE=- itation runoff, momentum extracted) and returns to
l+B
the climate model the surface fluxes of heat, moisture,
Flux measurements by the eddy correlation method and momentum consistent with these soil- and plant-
present challenges that can lead to uncertainty of based changes. Computational constraints limit the
5-30%. For a particular situation being sampled, an detail to which plant processes can be described, but,
58 AGRICULTURAL METEOROLOGYAND CLIMATOLOGY
simplistic as they are, the models provide a conceptual through use of irrigation, glass-houses, shelterbelts
framework for eventual coupling of more detailed and windbreaks, snow fences, wind machines, surface
crop, forest, and ecosystem models. mulches, certain tillage practices, alley-cropping, and
Crop models may be physiologically based or agroforestry. The design and operation of such mod-
statistically based. Crop growth models are built on ifications require considerable information on the
plant biophysical processes of agricultural crops and mean, extremes, and interannual variability of climate
their relationship to environmental factors. They at the specific location where the practice is imple-
predict growth, development, and yield based on mented.
complex interactions between weather, soil character- Horticulture crops, which typically have a higher
istics, nutrients, and plants. A practical application of value per unit area than grain crops, are sensitive to
crop growth models is t o estimate agricultural pro- small changes in microclimate. Also in contrast to
duction as a function of weather and soil conditions grains, horticulture crops are more sensitive to weath-
under alternative management conditions. Basic me- er-induced reduction in product quality or market
teorological information needed t o drive these models value. For instance, the desirable red coloration on
includes air temperature, precipitation, and solar some fruits is sensitive to optimal amounts of solar
radiation (or sunshine hours). More advanced models radiation at a critical stage. Manipulation of micro-
might additionally use dew-point temperature, wind climates for horticultural crops is more cost-effective
speed, and soil temperature. Statistically-based crop than for cereals because of both their high value and
models provide large-area yield predictions based on their sensitivity of quality to microclimate. Weather
correlations of past yields with regional average extremes may have multiyear impacts on agricultural
weather conditions. These models tend t o be much crops grown as perennials (e.g., fruits, nuts, grapes),
less computationally intensive, but also more location- which raises the cost-effectiveness of microclimate
specific and hence less transferable to other regions. modifications for reducing such extremes.
The fate of fugitive agricultural chemicals and
movement of insects and pollen are addressed by
models of atmospheric flow on scales of turbulent Agriculture Meteorology Forecasts
eddies t o mesoscale meteorology. Large-eddy simula-
Agriculturists can use weather forecasts with valid
tion (LES) models and models developed for use in air
times of a few minutes to several months. Weather
pollution regulation are sometimes adapted for sim-
forecasts are used for planning tillage and planting
ulating transport of agriculturally related materials.
operations, seed purchase, chemical application, frost
Recent advances in numerical simulation of turbulent
suppression, grain harvesting, transport and storage,
flow through vegetation have been used to understand
pest and disease management, and marketing, as well
the aerodynamic functioning of agricultural shelter-
as crop growth calculations and long-range planning.
belts. Extensions of these models t o simulate the
Major improvements over the past 10 years in our
complete microclimate provide opportunities for
understanding of the El NifiolSouthern Oscillation
exploring, by use of first principles, complex phy-
(ENSO)-related phenomena have enhanced prospects
sical relationships in heterogeneous ecosystems and
for seasonal to interannual forecasts of agriculture-
landscapes.
sensitive climate information. Such information now
Concern for national and international food security
is being used in early warning systems for planning,
has prompted the need for models of seasonal yield of
management, and operations in some tropical areas. In
various food crops. Private organizations as well as
regions where the climate correlation with ENSO is
governmental agencies have developed yield models
strong, projected ENSO factors have been used to
based on long-range weather conditions. The Food and
create projections of stress indices. Other than in the
Agriculture Organization of the United Nations has
Tropics and a few extratropical locations, the ENSO
developed agrometeorological models that forecast
signal in climate is muted or absent. However, current
yield on the basis of cumulative weekly or ten-day crop
research on this and related areas may offer future
water balances for providing early warning of poten-
progress in seasonal to interannual forecasts.
tial food security problems in developing countries.
Climate Data
Manipulating Microclimates to
Agricultural climatologists use long-term records of
Enhance Productivity and Reduce Risk standard meteorological data to compute derived
Agriculturists have a long history of enhancing crop agriculturally related variables such as growing degree
growth by manipulating soil and plant microclimates days, heat stress units, frost-free days, Palmer drought
AGRICULTURAL METEOROLOGY AND CLIMATOLOGY 59
index, and temperature-humidity index. These varia- dioxide will have a beneficial effect on crops both
bles have been developed from physiological con- through direct fertilization and through increased
cepts to correlate with crop development stage, crop water use efficiency. The mean yield increase for C4
yield, daily milk production, or other agriculturally crops (e.g., maize, sugar cane, millet, and sorghum)
important parameters. under a doubling of atmospheric COZ is minimal,
whereas increases for C3 crops (most other plants) are
approximately 30%, other factors being equal. Loss of
Education and Training soil organic matter, leaching of soil nutrients, and
salinization and erosion of soils will occur in some
Basic education for agricultural meteorologists is
climatic zones, which will call for more effective
usually acquired by supplementing a conventional
agricultural land use practices. Crop yields and pro-
meteorology, physics, or environmental science cur-
ductivity will vary considerably across different cli-
riculum with courses on plant, soil, or animal science,
mate zones under global warming, with low-latitude
forestry, or horticulture. Only a few US and European
and low-income countries being most negatively
universities offer undergraduate or graduate degrees
affected and some high-latitude countries experienc-
specifically in agricultural meteorology, and only one
ing more favorable crop growing conditions than
university on the whole African continent offers such a
under the current climate.
degree at the batchelor’s level. Most courses of study
Advances in our knowledge in the traditional areas
with emphasis in agricultural meteorology are con-
of agricultural meteorology - surface fluxes of energy,
nected with more traditional programs in agriculture
moisture, and trace gases, and the study of precipita-
such as agronomy. India has taken a more coordinated
tion and evaporation processes - will be urgently
approach than almost any other country to the
needed for coping with interannual variability and
university education of agricultural meteorologists.
long-term change of future climates. And practical
The practical applications of agricultural meteorol-
applications of this new knowledge require timely,
ogy have created needs for training programs aimed at
efficient, and worldwide distribution networks.
changing the knowledge, skills, and behavior of
Better understanding of basic agricultural micro-
personnel to achieve the objectives of the organiza-
meteorology and associated plant and soil processes
tions they work for. The World Meteorological
will allow for continuing advances in applied agricul-
Organization provides in-service training through
tural meteorology as well as at larger scales of
regional meteorological training centers that offer
meteorology, Measurements of heterogeneities in soil
specialized courses in basic agricultural meteorology,
and microclimate across a field are increasingly being
database management, agricultural meteorology
used for site-specific management of plant environ-
modeling, and hydrometeorology. These short courses
ments and yield improvement. At larger scales, the
tend to be task-oriented, focusing on improving and
subtle processes regulating the exchange of moisture
standardizing the practice of agricultural meteor-
and energy of plants and soil with the atmosphere are
ology, particularly in relation to observations and
key to improvements in numerical models of meso-
data management.
scale meteorology and global climate.
Rapid advances in the field of meteorology have
increased the need for more lifelong learning oppor-
tunities both in basic education and in training.
Increased interest in global observing networks for See also
monitoring a wider range of environmental variables Boundary Layers: Surface Layer. Energy Balance
exacerbates this need. The Internet offers a potential Model, Surface. General Circulation: Energy Cycle.
means of delivering standardized and authoritative Hydrology: Soil Moisture. Large Eddy Simulation.
educational and training materials to larger fractions Synoptic Meteorology: Forecasting. Ultraviolet Radi-
of the global agricultural meteorology community. ation. Weather Prediction: Adaptive Observations; En-
semble Prediction: Seasonal and Interannual Weather
Prediction: Severe Weather Forecasting.
Future Issues
The potential impact of global climate change on Further Reading
agriculture has been a subject of intense study in recent Doraiswamy PC, Pasteris PA, Jones KC, Motha RP and
years. Uncertainties in projections of future climates at Nejedlik P (2000)Techniques for methods of collection,
regional scales limit the accuracy with which agricul- database management and distribution of agrometeoro-
tural impacts can be estimated. However, there is high logical data. Agricultural and Forest Meteorology 103:
confidence that increases in atmospheric carbon 83-97.
60 AIRCRAFT EMISSIONS
Hanks J and Ritchie JT (eds)(1991)Modeling Plant and Soil Salinger MJ and Stigter CJ and Das HP (2000) Agromete-
Systems. American Society of Agronomy. orological adaptation strategies to increasing climate
Hoogenboom G (2000)Contribution of agrometeorologyto variability and climate change. Agricultural and Forest
the simulation of crop production and its applications. Meteorology 103: 167-184.
Agricultural and Forest Meteorology 103: 137-157. Sivakumar MVK, Gommes R and Baier W (2000)Agrome-
International Rice Research Institute (1989) Climate and teorology and sustainable agriculture. Agricultural and
Food Security, International Symposium on Climate Forest Meteorology 103: 11-26.
Variability and Food Security in Developing Countries. Stigter CJ, Sivakumar MVK and Rijks DA (2000)
Manila, Philippines: International Rice Research Insti- Agrometeorology in the 21st century: workshop summa-
tute. ry and recommendations on needs and perspec-
Lomas J, Milford JR and Mukhala E (2000)Education and tives. Agricultural and Forest Meteorology 103:
training in agricultural meteorology: current status and 209-227.
future needs. Agricultural and Forest Meteorology 103: Strand JF (2000) Some agrometeorological aspects of pest
197-208. and disease management for the 21st century. Agricul-
Maracchi G, Pkrarnaud V and Kleschenko AD (2000) tural and Forest Meteorology 103: 73-82.
Applications of geographical information systems and Stull R (1988)An Introduction to Boundary Layer Meteor-
remote sensing in agrometeorology. Agricultural and ology. Dordrecht: Kluwer Academic.
Forest Meteorology 103: 119-136. Wang H, Takle ES and Shen J (2001) Shelterbelts and
Monteith JL (2000) Agricultural meteorology: evolution windbreaks: mathematical modeling and computer sim-
and application. Agricultural and Forest Meteorology ulations of turbulent flows. Annual Reviews of Fluid
103: 5-9. Mechanics 33: 549-586.
Monteith JL and Unsworth MH (1990) Principles of Watson RT, Zinyowera MC and Moss RH (eds) (1996)
Environmental Physics, 2nd edn. London: Edward Arn- Climate Change 1995 - Impacts, Adaptations and
old. Mitigation of Climate Change: Scientific-Technical
Ogallo LA, Boulahya MS and Keane T (2000)Applications Analyses. Cambridge: Intergovernmental Panel on Cli-
of seasonal to interannual climate prediction in agricul- mate Change, Cambridge University Press.
tural planning and operations. Agricultural and Forest Whitmore JS (2000) Drought Management on Farmland.
Meteorology 103: 159-166. Dordrecht: Kluwer Academic.
R R Friedl, California Institute of Technology, Jet aircraft burning hydrocarbon-based fossil fuels
Pasadena, CA, USA transport the bulk of air passengers and freight.
Copyright 2003 Elsevier Science Ltd. All Rights Reserved Currently there are over 15 000 aircraft serving nearly
10 000 airports worldwide and burning nearly 140 Tg
of fuel per year. By the year 2015, fuel burn by aviation
Introduction
is forecast to increase to approximately 300 Tg per
Human society is becoming increasingly dependent on year. As with other fossil fuel transportation technol-
aircraft for long-distance travel and shipping. Among ogies, jet aircraft operation results in gaseous and
transportation modes, aviation is the fastest-growing; particle combustion byproducts. Aircraft engines emit
the current passenger growth rate is approximately principally carbon dioxide (C02) and water (H20)
4% per year and the average growth rate since 1960 with minor contributions from nitrogen oxides (NO,),
has been nearly 9 % per year. The fraction of transport sulfur oxides (SO,), unburned hydrocarbons (HC),
fuel use by aviation has risen steadily to about 13% and soot. All of these exhaust species are atmospheric
currently. Because of the robust growth rate, concern pollutants. C 0 2 and H20 are greenhouse gases that
has been expressed over possible environmental affect the Earth’s climate directly. NO, and H C are
impacts of future aircraft operation. Vigorous science reactive gases that affect atmospheric ozone and
and technology programs have been pursued over the methane levels. Soot, SO,, HC, and H20 are aerosol
last decade to define potential atmospheric impacts and cloud precursors that affect ozone and climate.
and identify technological strategies t o reduce specific A major difference between aviation and other
exhaust emissions. Environmental compatibility transportation modes is in the atmospheric placement
issues have also been central t o efforts t o develop of the combustion exhaust products. Unlike autos and
future aircraft technologies such as high-speed (i.e., trucks, by far the greater part ( > 8 5 % ) of aircraft
supersonic) civil transport. exhaust is released above the planetary boundary layer
Hanks J and Ritchie JT (eds)(1991)Modeling Plant and Soil Salinger MJ and Stigter CJ and Das HP (2000) Agromete-
Systems. American Society of Agronomy. orological adaptation strategies to increasing climate
Hoogenboom G (2000)Contribution of agrometeorologyto variability and climate change. Agricultural and Forest
the simulation of crop production and its applications. Meteorology 103: 167-184.
Agricultural and Forest Meteorology 103: 137-157. Sivakumar MVK, Gommes R and Baier W (2000)Agrome-
International Rice Research Institute (1989) Climate and teorology and sustainable agriculture. Agricultural and
Food Security, International Symposium on Climate Forest Meteorology 103: 11-26.
Variability and Food Security in Developing Countries. Stigter CJ, Sivakumar MVK and Rijks DA (2000)
Manila, Philippines: International Rice Research Insti- Agrometeorology in the 21st century: workshop summa-
tute. ry and recommendations on needs and perspec-
Lomas J, Milford JR and Mukhala E (2000)Education and tives. Agricultural and Forest Meteorology 103:
training in agricultural meteorology: current status and 209-227.
future needs. Agricultural and Forest Meteorology 103: Strand JF (2000) Some agrometeorological aspects of pest
197-208. and disease management for the 21st century. Agricul-
Maracchi G, Pkrarnaud V and Kleschenko AD (2000) tural and Forest Meteorology 103: 73-82.
Applications of geographical information systems and Stull R (1988)An Introduction to Boundary Layer Meteor-
remote sensing in agrometeorology. Agricultural and ology. Dordrecht: Kluwer Academic.
Forest Meteorology 103: 119-136. Wang H, Takle ES and Shen J (2001) Shelterbelts and
Monteith JL (2000) Agricultural meteorology: evolution windbreaks: mathematical modeling and computer sim-
and application. Agricultural and Forest Meteorology ulations of turbulent flows. Annual Reviews of Fluid
103: 5-9. Mechanics 33: 549-586.
Monteith JL and Unsworth MH (1990) Principles of Watson RT, Zinyowera MC and Moss RH (eds) (1996)
Environmental Physics, 2nd edn. London: Edward Arn- Climate Change 1995 - Impacts, Adaptations and
old. Mitigation of Climate Change: Scientific-Technical
Ogallo LA, Boulahya MS and Keane T (2000)Applications Analyses. Cambridge: Intergovernmental Panel on Cli-
of seasonal to interannual climate prediction in agricul- mate Change, Cambridge University Press.
tural planning and operations. Agricultural and Forest Whitmore JS (2000) Drought Management on Farmland.
Meteorology 103: 159-166. Dordrecht: Kluwer Academic.
R R Friedl, California Institute of Technology, Jet aircraft burning hydrocarbon-based fossil fuels
Pasadena, CA, USA transport the bulk of air passengers and freight.
Copyright 2003 Elsevier Science Ltd. All Rights Reserved Currently there are over 15 000 aircraft serving nearly
10 000 airports worldwide and burning nearly 140 Tg
of fuel per year. By the year 2015, fuel burn by aviation
Introduction
is forecast to increase to approximately 300 Tg per
Human society is becoming increasingly dependent on year. As with other fossil fuel transportation technol-
aircraft for long-distance travel and shipping. Among ogies, jet aircraft operation results in gaseous and
transportation modes, aviation is the fastest-growing; particle combustion byproducts. Aircraft engines emit
the current passenger growth rate is approximately principally carbon dioxide (C02) and water (H20)
4% per year and the average growth rate since 1960 with minor contributions from nitrogen oxides (NO,),
has been nearly 9 % per year. The fraction of transport sulfur oxides (SO,), unburned hydrocarbons (HC),
fuel use by aviation has risen steadily to about 13% and soot. All of these exhaust species are atmospheric
currently. Because of the robust growth rate, concern pollutants. C 0 2 and H20 are greenhouse gases that
has been expressed over possible environmental affect the Earth’s climate directly. NO, and H C are
impacts of future aircraft operation. Vigorous science reactive gases that affect atmospheric ozone and
and technology programs have been pursued over the methane levels. Soot, SO,, HC, and H20 are aerosol
last decade to define potential atmospheric impacts and cloud precursors that affect ozone and climate.
and identify technological strategies t o reduce specific A major difference between aviation and other
exhaust emissions. Environmental compatibility transportation modes is in the atmospheric placement
issues have also been central t o efforts t o develop of the combustion exhaust products. Unlike autos and
future aircraft technologies such as high-speed (i.e., trucks, by far the greater part ( > 8 5 % ) of aircraft
supersonic) civil transport. exhaust is released above the planetary boundary layer
AIRCRAFT EMISSIONS 61
-
( > 2 km) and a large fraction ( 70%) of it is released
in the upper troposphere (UT) and lower stratosphere
species in the fuel is sulfur, which can represent up to
0.3% (by weight) of the fuel content.
(LS) between 9 and 13 km. Consequently, the major Combustion of the fuel hydrocarbons to produce
polluting effects of aircraft are expected to occur in the C 0 2 and H20 is nearly complete (>99.5%) in
UT/LS region of the atmosphere. commercial aircraft engines. In addition, the fuel
The dynamics of the UT/LS region differ from those sulfur is converted to sulfur dioxide and sulfuric acid,
of the boundary layer in that there is less vertical although the precise mechanism for this process
mixing and less diurnal variation in wind direction. remains uncertain. The small fraction of incompletely
Because of these differences, pollutants emitted into combusted fuel hydrocarbons give rise to C O and
the UT/LS reside there longer and can spread over various smaller gaseous hydrocarbons (HC) such as
considerable longitudinal and, in some cases, latitudi- ethene, ethine, and formaldehyde. Under fuel-rich
nal distances. Although aircraft exhaust is released in combustor conditions, breakdown of the fuel hydro-
geographically narrow flight routes and corridors, its carbons leads to formation of soot particulates com-
injection into the UT/LS means that the polluting posed primarily of carbonaceous material. The rate-
effects of aircraft will be felt on regional and, perhaps, limiting process in soot formation appears to involve
global scales. The longer residence times also enable the oxidation of C2 species such as acetylene (C2H2).
some pollutants, such as NO,, to spend extended Decomposition of ambient nitrogen and oxygen also
times cycling through catalytic chemical reaction sets occurs in the high-temperature portions of the com-
that create or destroy ozone. Because of such enhanced bustor, giving rise to the important atmospheric
catalytic chemical cycling in the UT/LS, the impact of a pollutants nitric oxide (NO) and nitrogen dioxide
given amount of aircraft emissions on atmospheric (NO2) (Le., NO,) (Table 1).
ozone and climate may be much greater than the same
amount of emissions from ground transportation Aircraft Technology Considerations
sources.
There are also important dynamical and chemical Aircraft engine and airframe technologies have under-
differences between the UT and LS regions that com- gone dramatic improvements over the last 30 years.
plicate any analysis of aircraft effects. For example, One result of these improvements has been a 70%
the lifetime of ozone and the chemical mechanisms reduction in fuel burned per passenger seat from early
controlling its concentration are sensitive functions of to current jets. Gains in fuel efficiency are of benefit
altitude in the vicinity of the tropopause. Because of both economically and in environmental terms by
this altitude dependence, the sign of the ozone reducing fuel costs and uniformly lowering C02,
response to injections of NO, shifts from positive H20, and SO, emissions. These gains have derived
(net ozone formation) to negative (net ozone destruc- primarily from increasing gas temperatures and pres-
tion) at altitudes slightly above the tropopause (i.e., sures inside the engines. Without concomitant changes
the transition between the stratosphere and tropo- in engine design, increasing engine temperature leads
sphere). The partitioning of aircraft exhaust between to increasing NO, emissions.
the UT and LS is difficult to define (estimates differ by Concern over urban pollution has led to increas-
factors of two) because of the high variability and ingly stringent standards being adopted by the Inter-
latitudinal dependence of the tropopause height. national Civil Aviation Organization (ICAO)
regarding emissions of smoke, CO, HC, and NO,.
Aircraft smoke refers to visible particulates in the
aircraft plume and presumably includes the large
Aircraft Exhaust Products diameter ( >1pm) part of the soot population. The
Jet engines on modern aircraft are composed of three
essential elements: compressors that increase the
Table 1 Approximate emission index levels for cruise level
pressure and temperature of the entering air, combus-
operation of current commercial jet aircraft
tors that mix and burn fuel with incoming air, and
turbines that convert the hot gas energy, through Species Emission index (g kg- ’)
compressor activity, to bypass airflow that propels the COP 31 60
aircraft. The fuel-to-air ratio in modern combustors is H20 1240
approximately 1:9, hence large quantities of ambient co 2
air are processed in aircraft engines. Jet fuel is HC 1
composed predominately of high-weight ( CI2-Cl5) NO, (as NOz) 12
SO, (as SO2) 0.8
alkanes, with substantially smaller quantities of soot 0.04
alkenes and aromatics present. An important trace
ICAO standards have both reflected and motivated perturbations are expected to be localized in regions
improvements in engine design and manufacture. How- around the flight corridors. A great deal of work has
ever, because the service lifetime of an individual aircraft been done to compile accurate inventories of aircraft
is between 25 and 40 years, the current fleet consists of a emissions. These efforts have involved development of
combination of older and newer technologies. aircraft movement databases based on simplifying
Measurement of aircraft cruise emissions is an assumptions about the airframe-engine combinations
important facet of assessing impacts and documenting used and the paths flown between various city pairs.
technological advances. These difficult measurements Combining these movement databases with informa-
are made either in altitude simulation test cells or by in- tion on individual aircraft emission rates enables
flight measurements utilizing target and chase aircraft. construction of global emissions inventories. For
atmospheric modeling purposes, the aircraft emission
Aircraft Operational Considerations databases are divided into spatial bins that are 1"
Airline traffic patterns are highly inhomogeneous, longitude x 1"latitude x 1km altitude.
with the bulk of current traffic located inside well-
defined 'flight corridors' in the Northern Hemisphere
(Figure 1).The chemical lifetimes of aircraft exhaust
Impacts on Carbon Dioxide and Water
products such as NO,, soot, and sulfate injected in the Although they are the most prevalent exhaust prod-
UT/LS are comparable to atmospheric mixing times. ucts, emissions of COZ and H20 from aircraft repre-
Consequently, a number of the aircraft chemical sent relatively small sources of these species compared
Figure 1 Calculated NO, emissionsfor all aircraft traffic in May 1992 as afunction of altitude and latitude, summed over longitude (A),
and as a function of latitude and longitude summed over altitude (B). Values greater than the range maximum are plotted as black. (From
NASA reference publication 1400.)
with the many other large natural and anthropogenic (where M represents a gaseous third body such as N2
sources. Given past and current emission rates, aircraft or 0 2 ) . An analogous mechanism, which includes the
are responsible for increasing atmospheric C 0 2 levels reaction between N O and CH302, exists for CH4
by approximately 1ppmv or 2 % over the last 50 years. oxidation.
Because C 0 2 is very long-lived in the atmosphere and The overall rate of ozone production from CO and
is well mixed, it is impossible to distinguish the CH4 oxidation decreases generally with height be-
C02 emitted from aircraft from any other source. cause of decreasing ambient concentrations of CO and
Perturbations due to aircraft H20 emissions are far CH4. However, as can be seen from the above reaction
less than 1%globally. These small perturbations are sequence, the production rate depends also on the
impossible to detect on the large scale because water ambient concentrations of N O and OH. For example,
vapor has a short (days to weeks) tropospheric increasing O H and N O will increase the rates of
residence time and its ambient concentrations are reactions [I] and [111], respectively, thereby increasing
highly variable. At very small spatial scales, H20 the ozone production rate. At high enough concentra-
perturbations from aircraft are substantial and can tions of NO, ( > 500 pptv) the ozone production rate
lead to contrail and cirrus cloud formation. These begins to decrease owing to the increasing importance
effects have important climate consequences that will of NO2 reactions that remove HO, species, i.e.,
be discussed below.
OH+ NO2 + M + H O N 0 2 + M [VI]
Impacts on Ozone and Methane HO2 + NO2 + M + HO2N02 + M [VII]

Ozone chemistry throughout the stratosphere and In the LS region, the primary influence of NO, is
troposphere is driven by solar-initiated free radical on destruction of ozone by the following radical-
reactions. Aircraft emit a number of species (i.e., NO,, catalyzed processes:
SO,, H20, CO, and soot) that participate in ozone-
controlling reactions of free radicals and free radical 0 + xo -+ x+0 2 [VIII]
precursors. The relationship between aircraft exhaust
products and ozone is complex and depends on the X+03+XO+02 [IXI
balance between a number of ozone-forming and
-depleting chemical processes. These processes are Net: 0 +0 3 + 202
summarized in the next two sections along with

(where X =NO, C1 and OH). Increases in ambient
observational evidence that addresses the magnitude
NO, due to aircraft emissions will enhance ozone
of the aircraft effect on ozone.
destruction for the case where X = N O but will
Atmospheric Chemistry decrease ozone destruction for the case X = C1 by
removing C10 through reaction [XI:
Nitrogen oxides in the UT/LS participate in both
ozone-forming and ozone-depleting reaction cycles. C10 + NO2 + M -+ ClON02 +M [XI
The balance between these processes, and their
response to changes in ambient NO, levels, are The exact balance between these contrary effects
sensitive functions of altitude. In the UT region, the depends on the background concentrations of NO,
primary influence of NO, is on the production of and C10,. Throughout most of the year, the net effect
ozone from CO and CH4 oxidation. The CO cycle of increasing LS NO, levels through aircraft emissions
involves the following reactions: will be to increase ozone concentrations. An exception
is at high latitudes in summer, when NO, levels are
O H + C O -+ H + C02 [I1 high. In that case, addition of NO, will decrease
ozone.
H +0 2 +M -+ HOl +M Increasing NO, levels due to aircraft exert an
PI indirect effect on atmospheric CH4 concentrations.
This effect is initiated by the formation of O H radicals
in reaction [III]. The increased levels of O H in the air
traffic corridors lead to decreases in carbon monoxide
NO2 + Sunlight -+ NO +0 [IVI (CO) through reaction [I]. Because CO lifetimes are
longer than NO, lifetimes, the region of decreased CO
0 +0 2 + M -+ 03 +M [VI concentrations spreads out from air traffic corridors to
a much greater extent than do the aircraft NO,
Net: C O + 202 -+ C 0 2 + O3 emissions. The CO perturbation spreads all the way to
tropical and subtropical regions where much of the in the total ozone column. In terms of climate effects,
global oxidation of CH4 takes place through its the radiative forcing changes due to increased ozone
reaction with OH: appear to be largely offset by the predicted decreases in
methane. Considerable uncertainty is attached to
O H + CH4 -+CH3 + H20 [XI1 these calculations, however (see Figure 2).
As CO levels are lowered in the tropics, O H levels are Observing Ozone Impacts
raised correspondingly. The higher levels of O H serve
to lower CH4 concentrations, which, in turn, lead to a Dense air traffic in Northern Hemisphere flight corri-
further increase in OH. As a result of the complex dors will give rise to distinct geographical perturba-
interplay (Le., atmospheric feedback cycle) between tions of NO,, aerosols and ozone under two condi-
NO,, OH, CO, and CH4, an increase in NO, will lead tions. First, large-scale dispersion of the exhaust must
to an amplified decrease in CH4. The amplification be slower than the chemistry that removes and/or links
factor is approximately 1.5. these emissions to ozone. Second, the strength of the
Aircraft emissions of SO,, H20 and soot also effect aircraft emissions must be significant relative to other
atmospheric ozone concentrations by serving as natural and anthropogenic sources of NO, and
aerosol precursors. In the UT, sulfate- and water-ice- aerosols. The total NO, emission from current global
containing aerosols promote ozone decreases by aviation is approximately 0.5Tg per year, of which
acting as surfaces for heterogeneous removal of the roughly 60% is released into the upper troposphere
ozone precursors NO, and HO,. A major identified and 15% is released into the lower stratosphere.
heterogeneous reaction involves conversion of the The major source of NO, in the lower stratosphere
temporary NO, reservoir species nitrogen pentoxide is chemical oxidation in situ of nitrous oxide ( N 2 0 ) :
(N205) into the longer-term reservoir nitric acid
(HN03). 0 3 + sunlight + O(lD) 0 2 + [XIVI
N205 + H2S04/H20 -+ 2 HNO3 [XII]
O('D) + N20 -+ 2N0 [XVI
HN03, along with a number of other nitrogen and
hydrogen acids and peroxides (e.g., H N 0 4 and H202) The global production rate of NO from N20 (- 12 Tg
are absorbed onto sulfate and water-ice. The absorbed per year) far exceeds that from current subsonic
species can be removed from the UT by sedimentation. aircraft emissions in the lower stratosphere. Hence
In the LS, sulfate- and water-ice-containing aerosol there is no expectation, nor observational evidence,
particles not only remove HO, and NO, species but that current aircraft are significantly perturbing
also liberate ozone-destroying C10, by heterogeneous stratospheric NO, levels.
reactions such as In the upper troposphere, the major non-aircraft
sources of NO, include fossil fuel combustion
ClON02 + H20 -+ HOC1 +HN03 [XIII]
The net effect of the heterogeneous processes is to
decrease ozone in the LS and UT. However, the effect Aircraft NO,
of the aircraft-derived aerosols on LSAJT ozone offsets
only partially the effect of the NO, emissions.
Much less is known about the effect of soot
particulates on ozone. Ozone is observed to react
directly on laboratory soot surfaces, but the reaction
slows as the surface is modified. Heterogeneous
reactions of NO, and nitrogen reservoir species also
occur on soot surfaces - in some cases the reactions
lead to more reactive species, in others to less reactive
ones. Consequently, the effect of aircraft soot on
Aircraft Aircraft Aircraft SO,
atmospheric ozone concentrations is poorly deter- soot H2O and H 2 0
mined at present.
According to the current scientific understanding, Figure 2 Influence of aircraft emissions on chemical balance in
the overall effect of aircraft emissions in the UT/LS is the UTlLS region. Atmospheric chemical reactions couple together
0 3 , CO, NO, and CH4. Among aircraft emissions, NO, is
to increase ozone levels. Model calculations indicate calculated to have the greatest effect on the coupled species,
that aircraft have increased ozone by about 6% in acting to increase ambient NO, and O3levels and decrease CO
heavy traffic areas, with an associated 0.4% increase and CH4.
Table 2 Present-day sourcesof NO, in the troposphere and their ing air is very dry and/or warm, contrails may be short-
approximate strengths lived or may not form at all. In either case, the emitted
Source Emission rate Emission rate soot and sulfate nuclei will remain in the atmosphere
(Ts yr- (TsYr- for days and weeks and possibly promote natural ice
Total 9-13km altitude band (cirrus) cloud formation in locations far from the
Aviation 0.5 0.3
initial aircraft plume. These same nuclei, upon contact
Fossil fuel combustion 22 0.7 with cirrus clouds, may change properties of the cloud
Biomass burning 8 0.2 particles such as size distribution, number density, and
Soil emissions 7 0.2 chemical composition.
Lightning 5 1 Like other naturally occurring clouds, contrails and
N20 oxidation 12 0.6
aircraft-induced (or modified) clouds impact the
Earth’s climate by affecting the radiation balance.
For typical particle properties, cirrus clouds trap
surface outgoing long-wave radiation more effectively
(autos, trucks, etc.), biomass burning, soil emissions, than they reflect solar incoming short-wave radiation.
lightning, and N20 oxidation. Of these, only lightning As a result, cirrus clouds tend to warm the climate.
deposits NO, directly into the UT. The fractions of However, the magnitude and even the sign of a cloud’s
NO, transported into the UT from sources at the radiative effect on climate is a sensitive function of
Earth’s surface or in the stratosphere are small, cloud particle size and shape as well as altitude and
occurring only during convective events, such as geographical location.
frontal activity or thunderstorms or during strato-
sphere-troposphere exchange events triggered by
Cloud Formation Processes
meteorological features such as extratropical cy-
clones. Source strength estimates for the various Clouds or contrails can form when air moisture
NO, sources are listed in Table 2. As shown in the becomes supersaturated with respect to ice. The ice
table, aircraft emissions into the UT are of comparable formation process takes place by one of several
strength to other sources and contribute a significant mechanisms. At higher supersaturations, low-volatil-
fraction of UT NO,. ity gas phase species will cluster together to form
Chemical sampling of the UT in and around traffic liquid particles. These liquids can subsequently freeze
corridors has revealed each individual aircraft per- in a process known as homogeneous freezing if
turbs ambient NO, levels substantially for distances of the air is cooled by upward dynamical motion.
several kilometers behind it. At larger spatial scales, At lower supersaturations, the freezing process
aircraft signatures have not been discerned, owing to may be aided by the presence of a solid particle surface
the high variability of background NO,. Likewise, in a process termed heterogeneous freezing. The liquid
there have been no identifiable spatial patterns in and solid particles that readily promote freezing and
ozone concentrations that unambiguously point to ice crystal growth are typically in the 0.05-1 pm
production by aircraft NO,. Long-term ozone trend diameter size range; they are referred to as cloud
observations at specific measuring stations (e.g., condensation nuclei (CCN) and ice nuclei (IN),
Hohenpeissenburg, Germany, and Wallops Island, respectively. Aircraft emissions may enhance the
USA), do not correlate with the growth rate of air frequency of these freezing events by increasing the
traffic from 1970 to the present, indicating that abundances of CCN and IN.
aircraft emissions are not a major factor in the Aircraft soot emissions have attracted attention
observed upper-tropospheric trends. as a possible source of IN in the UT. The median
size of a fresh aircraft soot particulate is approxi-
mately 0.02ym. In order for a soot particulate
to become an IN it must be activated (i.e., become
Impacts on Clouds more hydrophilic) by reaction with suitable species.
Trails of ice particles - contrails - are the most readily Lab studies have shown that acids such as H2S04
identifiable exhaust signatures of aircraft (see Con- induce this activation but that others such as H N 0 3 do
trails). Contrails often form, even under clear-sky not. Hydrated samples of soot have been obtained
conditions, because aircraft H20 emissions raise the from non-sulfur-containing flames, indicating the
relative humidity of the air near the exhaust plume presence of other, as yet unidentified, activating
above 100%. Water vapor in the supersaturated air species.
subsequently condenses on aircraft-derived soot and Aircraft emissions of condensable gases such as
sulfate nuclei and freezes to form ice. If the surround- sulfur oxides and oxygenated hydrocarbons can
contribute to CCN formation. In an aircraft plume, clouds, aircraft-derived CCN and IN may influence
large numbers (- l o T 6particles per kilogram of fuel) of the properties of the cloud particles in one of
small particles (<0.01 pm radius) are formed from several ways. If the aircraft particles are larger and
nucleation of sulfuric acid and water. The formation function as more active growth nuclei than ambient
and subsequent growth of these particles may be particles, they may compete effectively for the avail-
accelerated by chemi-ions that are emitted into the able water vapor and induce growth of larger ice
plume following their production in high-temperature crystals at the expense of crystal number density. If, on
reactions occurring in the combustor. As the plume the other hand, the aircraft particles increase the
expands and is diluted by entrainment of ambient air, number of CCN and IN, but do not change the overall
the small plume particles may continue to grow by rate of crystal growth, then increases in the crystal
uptake of additional gaseous species or they may be number density are expected, with concomitant
scavenged by larger ambient particles. The competi- decreases in average crystal size. The radiative prop-
tion between these two processes depends on a number erties of the resulting clouds will be altered, but the
of environmental variables such as air temperature, magnitudes and characteristics of these modifications
relative humidity, and background aerosol concentra- are uncertain.
tion. Under low background aerosol conditions, such
as exist during wintertime, a significant number of
Observing Cloud Impacts
plume particles are expected to survive long enough to
grow to CCN size (see Figure 3). Cirrus cloud coverage, as documented by surface and
Addition of aircraft-derived CCN and IN to the UT satellite observations, has been increasing over a
will increase cirrus cloud occurrence in areas where number of regions in the last two decades, with the
the air is supersaturated with respect to ice but crystal largest increases observed over regions of heavy air
growth is limited by a lack of sufficient numbers of traffic in the United States and the North Atlantic.
nuclei. Relative humidity measurements taken in the Growth of cloud cover in air corridor regions has been
UT reveal that ice supersaturation occurs in more than approximately 1-2% per decade greater than in other
10% of the clear-sky cases examined. Cloud growth in areas; attributing this growth rate to aircraft impacts
these regions should be particularly susceptible to implies that there has been an overall 5% increase in
aircraft IN and CCN. In areas of developing cirrus traffic route cloud cover during the last 30 years of air
m
Y
g 10'6
100 101 102

Plume age (h)
Figure 3 Calculated time evolution of the 'apparent' emission indices of aircraft-generated particles for various size thresholds. Solid
and dashed lines are for low and high ambient aerosol conditions, respectively. Appreciable concentrations of CCN size particles
(>50nm) are predicted only for low ambient aerosol conditions. (Adapted with permission from Yu F and Turco RP (1999) Geophysical
Research Letters 26: 1703-1706. Washington, DC: American Geophysical Union.)
travel. Contrails are clearly a significant part of this Table 3 Ozone and climate impacts of present day aviation.
increase. For instance, analysis of satellite images has Radiative forcing is used as a measure of the climate impact
indicated that contrail coverage over Europe is on the Impact Ozone column Radiative forcing
order of 1%. ("h change) (wrn-')
Indirect effects of aircraft-derived aerosols on cloud
COP +0.016
formation may be responsible for the rest of the
NO, +0.5 +0.024
observed increase. However, a number of other CH4 - 0.015
natural and anthropogenic causes such as changes Hz0 +0.002
in UT temperature and humidity, greenhouse gas Sulfate aerosol -0.1 - 0.003
concentrations, and upper-atmosphere dynamics Soot particulate - 0.01 +0.003

Contrails +0.021
may be contributing to the increased cloudiness.
Cirrus clouds 5 +0.04
Currently, there is little observational information
from which to determine if aircraft are significant Numbers are taken from IPCC (1999).
sources of CCN and IN. In a few cases, sampled
contrail and cirrus cloud particles have contained
significant amounts of soot and metals, suggestive of
an aircraft influence. But the few direct measurements tion of the aircraft-specific contribution to climate and
of CCN in aircraft wakes have yielded concentrations ozone change is not possible at present. Future growth
ranging from very low to very high values. Similar of aircraft emissions is related to economic and
types of measurement for IN have suggested population growth. Under some plausible high-
that aircraft wakes do not contain large numbers of growth scenarios, aircraft emissions may increase
these nuclei. fivefold over the next 50 years and contribute larger
fractions of the anthropogenic climate forcing and
ozone change. In addition, the possible introduction
Impact Summary: Present and Future
of a supersonic aircraft fleet may contribute to
Scientific progress on aviation impacts has progressed significant changes in stratospheric water vapor and
to a point where it is providing important guidance ozone. The potential impacts of supersonic aircraft
to technology and policy decision-makers. Aircraft emissions have been studied in the 1970s and again
C 0 2 emissions are known to be small but signi- in the 1990s in concert with industry interest in
ficant contributors to the observed rise in ambient building a fleet of such aircraft. From these studies,
levels, accounting for approximately 2 % of the the higher altitude (i.e. 16-20km) release of
current total anthropogenic C 0 2 emissions. In supersonic NO, and H20 emissions is predicted
addition, it is now clear that the overall radiative to generally enhance ozone-depleting cycles
forcing by aircraft is substantially larger, by a factor (reactions VI11 and IX) relative to ozone-forming
of 2 to 5, than the forcing by aircraft C 0 2 alone. ones (reactions IV), and lead to a net decrease of
In terms of possible mitigation strategies, this ozone. The water vapor emissions from supersonic
finding underscores the need to reduce many of aircraft, occurring predominately in the relatively dry
the aircraft emissions simultaneously or consider lower stratosphere, are also predicted to contribute
tradeoffs in aircraft designs and operational measures. to climate warming.
Examples of tradeoffs include the development Even if these future subsonic and supersonic
of more efficient (i.e., emitting lower COZ) engines scenarios come to pass, detection of specific aircraft
at the expense of higher NO, emissions and/or climate and ozone impacts will continue to be difficult.
promotion of contrails, and increases in average Consequently, technology and policy decision-making
flight altitudes to reduce contrail formation at the will have to rely on observations of overall global
expense of increased 0 3 sensitivity to NO,. Consid- change along with a robust scientific understanding of
eration of these tradeoffs is tempered by the fact that aviation effects.
the impacts from aircraft C 0 2 are far better charac-
terized than the impacts from NO,, contrails, and
aerosols (Table 3 ) .
Acknowledgements
The calculated impacts of the current aircraft fleet I dedicate this article to my wife Myrna; her
on climate and ozone are small relative to the natural long battle with illness ended sadly during its prepa-
atmospheric variability and have not been observed. ration. I thank Drs S. Baughcum and E Yu for use
Like climate and ozone impacts of other individual of their data in Figures 1 and 3. Preparation of the
sectors, aircraft emissions represent a small fraction of article was carried out at the Jet Propulsion Labo-
the total anthropogenic impact. Consequently, detec- ratory, California Institute of Technology, under
68 AIRCRAFT ICING
contract with the National Aeronautics and Space Further Reading

Administration. Brasseur G, Cox RA, Hauglustaine D, etal. (1998)European
Scientific Assessment of the Atmospheric Effects of
Glossary Aircraft Emissions. In: Brasseur G, Amanatidis GT,
and Angeletti G. (eds) Atmospheric Environment 32:
CCN cloud condensation nuclei 2327-2418.
emission index measure of aircraft emission in units Brasseur GP, Orlando JJ, and Tyndall GS (eds) (1999)
grams of exhaust per kilogram of fuel burned Atmospheric Chemistry and Global Change. New York:
IN ice nuclei Oxford University Press.
ICAO International Civil Aviation Organization Fried1 RR, Anderson BE, Baughcum SL, et al. (1997)
LS lower stratosphere Atmospheric Effects of Subsonic Aircraft: Interim
NO, reactive oxides of nitrogen (NO, = NO+ N02) Assessment Report of the Advanced Subsonic Technol-
ppmv parts per million by volume ogy Program. NASA Reference Publication 1400. Green-
pptv parts per trillion by volume belt, MD: NASA Goddard Space Flight Center.
SO, reactive oxides of sulfur (SO, = SO2 so3) +
Intergovernmental Panel on Climate Change (1999) Avia-
tion and the Global Environment, eds. Penner JE, Lister
UT upper troposphere
DH, Griggs DJ, Dokken DJ, etal. Cambridge: Cambridge
University Press.
Turco RP (1997)Earth Under Siege, From Air Pollution to
See also Global Change. New York: Oxford University Press.
Chemistry of the Atmosphere: Chemical Kinetics; Gas SONEX/POLINAT Special Section (2000) Geophysical
Phase Reactions; Principles of Chemical Change. Cli- Research Letters 26: 3053-3084. Washington, DC:
mate: Overview. Contrails. Global Change: Ozone American Geophysical Union.
Trends. Soot. Stratospheric Chemistry and Compo- SONEX/POLINAT Special Section (2000)Journal of Geo-
sition: HO,; Reactive Nitrogen (NO, and NO,). Strato- physical Research 105: 3595-3892. Washington, DC:
spheric Water Vapor. Tropospheric Chemistry and American Geophysical Union.
Composition: Aerosols/Particles;Carbon Monoxide; Sul- Wayne RP (1991) Chemistry of Atmospheres, 2nd edn.
fur Chemistry, Organic. Oxford: Clarendon Press.
M K Politovich, National Center for Atmospheric adhering to the wings of an airplane prior to takeoff,
Research, Boulder, CO, USA and carburetor icing, will not be covered.
Copyright 2003 Elsevier Science Ltd. Ail Rights Reserved.
Effects on an Aircraft
Introduction Although the basic concept of in-flight icing is a simple
Aircraft icing is the accretion of supercooled liquid onto one, the processes contributing to icing, and the results
an airplane during flight. Accreted ice adversely affects of icing, are at once quite complex and fascinating.
flight; thus, it is an important component of an aviation Meteorologists, aerospace engineers, and pilots need
weather forecast. Meteorology associated with in-flight and want information about icing because it can
icing begins with the microscale, addressing growth of adversely affect the flight characteristics of an aircraft.
supercooled droplets and their collision with and Icing can increase drag, decrease lift, and cause control
adhesion to airframes. Cloud-scale and mesoscale problems. The added weight of the accreted ice is
processes control the amount and distribution of generally a factor only for light aircraft.
supercooled liquid water. Synoptic weather patterns Aircraft can fly in icing conditions, and to do so
govern the movement and overall location of icing legally they must first be certified. For certification of a
environments. Any discussion of aircraft icing must particular type of airplane, it must be flown in a range
also include the development and use of numerical of natural icing conditions and demonstrate that these
weather prediction models as well as in situ and remote conditions result in no significant effect on the
sensors for icing detection, diagnosis, and forecasting. airplane’s performance. The range of conditions was
There are isolated cases of snow and frost adhesion first developed from measurements obtained in the
during flight, but since these rarely occur they will not 1940s and is illustrated in Figure 1, which was
be discussed here. Similarly, precipitation or frost designed to envelop 99.9% of icing conditions found
68 AIRCRAFT ICING
contract with the National Aeronautics and Space Further Reading

Administration. Brasseur G, Cox RA, Hauglustaine D, etal. (1998)European
Scientific Assessment of the Atmospheric Effects of
Glossary Aircraft Emissions. In: Brasseur G, Amanatidis GT,
and Angeletti G. (eds) Atmospheric Environment 32:
CCN cloud condensation nuclei 2327-2418.
emission index measure of aircraft emission in units Brasseur GP, Orlando JJ, and Tyndall GS (eds) (1999)
grams of exhaust per kilogram of fuel burned Atmospheric Chemistry and Global Change. New York:
IN ice nuclei Oxford University Press.
ICAO International Civil Aviation Organization Fried1 RR, Anderson BE, Baughcum SL, et al. (1997)
LS lower stratosphere Atmospheric Effects of Subsonic Aircraft: Interim
NO, reactive oxides of nitrogen (NO, = NO+ N02) Assessment Report of the Advanced Subsonic Technol-
ppmv parts per million by volume ogy Program. NASA Reference Publication 1400. Green-
pptv parts per trillion by volume belt, MD: NASA Goddard Space Flight Center.
SO, reactive oxides of sulfur (SO, = SO2 so3) +
Intergovernmental Panel on Climate Change (1999) Avia-
tion and the Global Environment, eds. Penner JE, Lister
UT upper troposphere
DH, Griggs DJ, Dokken DJ, etal. Cambridge: Cambridge
University Press.
Turco RP (1997)Earth Under Siege, From Air Pollution to
See also Global Change. New York: Oxford University Press.
Chemistry of the Atmosphere: Chemical Kinetics; Gas SONEX/POLINAT Special Section (2000) Geophysical
Phase Reactions; Principles of Chemical Change. Cli- Research Letters 26: 3053-3084. Washington, DC:
mate: Overview. Contrails. Global Change: Ozone American Geophysical Union.
Trends. Soot. Stratospheric Chemistry and Compo- SONEX/POLINAT Special Section (2000)Journal of Geo-
sition: HO,; Reactive Nitrogen (NO, and NO,). Strato- physical Research 105: 3595-3892. Washington, DC:
spheric Water Vapor. Tropospheric Chemistry and American Geophysical Union.
Composition: Aerosols/Particles;Carbon Monoxide; Sul- Wayne RP (1991) Chemistry of Atmospheres, 2nd edn.
fur Chemistry, Organic. Oxford: Clarendon Press.
M K Politovich, National Center for Atmospheric adhering to the wings of an airplane prior to takeoff,
Research, Boulder, CO, USA and carburetor icing, will not be covered.
Copyright 2003 Elsevier Science Ltd. Ail Rights Reserved.
Effects on an Aircraft
Introduction Although the basic concept of in-flight icing is a simple
Aircraft icing is the accretion of supercooled liquid onto one, the processes contributing to icing, and the results
an airplane during flight. Accreted ice adversely affects of icing, are at once quite complex and fascinating.
flight; thus, it is an important component of an aviation Meteorologists, aerospace engineers, and pilots need
weather forecast. Meteorology associated with in-flight and want information about icing because it can
icing begins with the microscale, addressing growth of adversely affect the flight characteristics of an aircraft.
supercooled droplets and their collision with and Icing can increase drag, decrease lift, and cause control
adhesion to airframes. Cloud-scale and mesoscale problems. The added weight of the accreted ice is
processes control the amount and distribution of generally a factor only for light aircraft.
supercooled liquid water. Synoptic weather patterns Aircraft can fly in icing conditions, and to do so
govern the movement and overall location of icing legally they must first be certified. For certification of a
environments. Any discussion of aircraft icing must particular type of airplane, it must be flown in a range
also include the development and use of numerical of natural icing conditions and demonstrate that these
weather prediction models as well as in situ and remote conditions result in no significant effect on the
sensors for icing detection, diagnosis, and forecasting. airplane’s performance. The range of conditions was
There are isolated cases of snow and frost adhesion first developed from measurements obtained in the
during flight, but since these rarely occur they will not 1940s and is illustrated in Figure 1, which was
be discussed here. Similarly, precipitation or frost designed to envelop 99.9% of icing conditions found
AIRCRAFT ICING 69
Icing tends to affect general aviation less than

commuter or air carrier operations; there are several
reasons for this. The smaller aircraft included in the
general aviation category tend to fly at lower altitudes
where icing is more prevalent. Those aircraft may have
less de-icing capability and reserve power in case of
encountering icing conditions, and their pilots may
have less experience of operating under icing condi-
tions. Air carriers tend to quickly penetrate icing-
bearing clouds on ascent and descent from airports
and cruise at altitudes far above those at which icing
occurs. Commuter aircraft are caught in the middle, in
terms of both their ability to handle ice and the
altitudes at which they fly. With the burgeoning
Figure 1 Icing envelopes defined by liquid water content, droplet business in this area, they find themselves susceptible
size, and temperature. The shaded areas denote the limits of these to icing and need accurate forecasts.
environmental parameters in which aircraft must be able to fly
safely to be certified for flight into icing conditions. (From FAA
Federal Aviation Regulations Part 25, Appendix C.)
Severity and Intensity of Icing
in stratiform clouds. More recent studies have con- Icing is currently classified into four severity catego-
firmed that the indicated range provides reasonable ries: trace, light, moderate, and severe. Severity is a
limits for certification, although, as will be presented combination of the state of the icing environment, the
in a later section, it does not address the problem of aircraft’s response, and the pilot’s assessment of the
large supercooled drops. response.Table 1shows descriptions for icing severity
Certified aircraft are commonly equipped with that are being adopted for official use by the FAA.
devices that either serve to prevent ice from adhering Even with these severity descriptions, there is as yet
to the airframe or remove it once it has adhered. Such no official quantification of environmental parame-
anti-icing or de-icing equipment may be deployed ters. The most important parameters are the liquid
manually or through an automatic system triggered by water content, outside air temperature, and droplet
an icing detection probe. Equipment includes pneu- size. The more liquid water there is, the more is
matic ‘boots’, heat, and liquid. All three can be applied available to accrete on the airframe, and thus higher
to the leading edges of the wings and tail, and liquid water contents are associated with more severe
occasionally to propellers. conditions. Temperature controls what happens to
Tailplane icing is a subset of icing and refers to icing that liquid once it impacts the airframe - either it
that accretes on the vertical and horizontal stabilizers. freezes in place or it runs back along the surface to
It is not necessarily caused by unique atmospheric possibly unprotected areas. Droplet size controls the
conditions but is usually considered separately be- collection efficiency of those droplets onto the air-
cause it results in vastly different response of the frame. Overall, droplet size is not as important as
airplane from that produced by icing on the wings. liquid water content or temperature in determining
Table 1 Icing severity descriptions

Category Description
~ ~ ~ ~ ~ ~~
Trace Ice becomes perceptible. Rate of accumulation is slightly greaterthan rate of sublimation. It is not hazardous even
though de-icing/anti-icingequipment is not utilized, unless encountered for an extended period of time (over 1
hour).
Light The rate of accretion may create a problem if flight is prolonged in the environment (over 1 hour). Occasional use
of de-icinglanti-icingequipment removes/preventsaccretion. Itdoes not present a problem if the de-icing/anti-
icing equipment is used.
Moderate The rate of accretion is such that short encounters become potentially hazardous and use of de-icing/anti-icing
equipment, or diversion, is necessary.
Severe The rate of accretion is such that de-icing/anti-icing equipment fails to reduce or control the hazard. Immediate
diversion is necessary.
70 AIRCRAFT ICING
severity until droplets reach drizzle sizes, with diam- It is formed when the droplets deform andlor flow
eters exceeding -50 ym. Thus, environments with along the surface prior to freezing. Glaze icing can be
high amounts of liquid water with large droplet more serious to the aircraft than rime since it tends to
diameter at warm (but supercooled) temperatures run back along the airframe, covering more surface
would represent the most hazardous conditions for area than rime icing - perhaps flowing onto and
aircraft icing. Research is being conducted to deter- adhering to unprotected areas. Glaze icing can be hard
mine appropriate limiting values for these parameters to see from inside the aircraft, so that the pilot may be
to define each severity category. To be useful, the unaware of ice buildup. Mixed icing often occurs in
definitions must relate atmospheric conditions to layers, similar to wet and dry hailstone growth, as a
observable information as well as effect on flight. transition from rime to clear conditions is encoun-
tered. These icing types are illustrated in Figure 2.
The type of icing is related to the air temperature,
Types of Icing the liquid water content, and the size of the droplets.
There are two main physical types of icing: glaze and Glaze is generally associated with higher tempera-
rime. Mixed icing is a combination of the two. Rime tures, higher supercooled liquid water (SLW)contents
ice is brittle and opaque and tends to grow into the and larger droplets. Rime is usually created at lower
airstream. It is formed as the droplets freeze immedi- temperatures, low SLW contents, and small droplet
ately upon impact. Glaze icing, sometimes referred to size. There are also effects dependent on the airplane
as clear icing, can be nearly transparent and has a itself, including wing shape, airspeed, and type of de-
smoother surface, sometimes with a waxy appearance. icinglanti-icing equipment.
(C) (D)
Figure 2 Post-flight photographs of ice encountered by the NASA Glenn Research Center’s instrumented Twin Otter aircraft. The
leading edge of the left wing is shown in each photograph: (A) light rime ice, (B) severe glaze ice, (C) moderate mixed ice, (D) supercooled
large droplet ice. Note how much farther aft the ice in (D) hasaccreted compared to the othertypes. (Photographs courtesy of NASA Glenn
Research Center.)
AIRCRAFT ICING 71
Location and Frequency of Icing Light icing is the most frequent severity category
Conditions reported by pilots, accounting for -60-70% of all
reports. Severe icing, which indicates a condition in
Since icing occurs in clouds or precipitation at which flight cannot be sustained, is reported in only a
temperatures below O"C, any icing climatology must few percent of cases. Rime icing is reported much
be associated with cold, cloudy conditions. Figure 3 more frequently than glaze or mixed, comprising
shows that icing frequency is most strongly related to
latitude in the contiguous United States, with some
-70-75% of reports. For both icing type and severity,
the largest joint frequency is for light rime icing, which
preference for the northeastern part of the country. covers nearly half of all reports.
Icing-related fatal aircraft accidents average approx-
imately 30 per year in the United States, with the
highest incidence in the winter months. Alaska has by
far the highest accident rate, followed by the north-
Relation of Icing to Weather Features
west mountains, Great Lakes, western Pacific states Icing takes place in clouds or precipitation and is thus
and the central states. closely related to weather features providing such
The average altitude of icing environments is conditions. Most PIREPs of icing occur near fronts.
around 3000 m above mean sea level (msl), with few Regions ahead of or near surface warm fronts are
encounters above 6000 m. Cumuliform clouds, with favorable icing regions if temperatures are in the right
their greater depth and transport of significant liquid range (- 0 to - ZOOC), since they provide widespread
amounts to higher altitudes, have on average higher lifting of generally moist air. Cold fronts also provide
altitude coverage than stratiform clouds. Frequency opportunities for icing, with narrower regions of more
of icing 'PIREPs' (pilot reports) by time of day is a intense lifting near the surface front. Moist, maritime
direct reflection of the frequency of flights, with few air masses are associated with higher frequencies of
reports overnight. The weekly pattern also follows air icing PIREPs, whereas continental air masses, espe-
traffic trends, with most reports on Tuesday through cially those well behind arctic fronts, have fewer
Thursday. reports. Topography also influences icing, providing
Figure 3 Frequency of icing expected over the contiguous 48 states of the United States in February. Frequency is expressed as
percentage of total time available. These data are for February 1992 through 1997, for altitudes <3000m above msl. The plots were
constructed by correlating icing pilot reports to surface weather measurements, then extrapolating the icing pilot report frequency to the
entire month, to account for overreporting in high air traffic areas, and underreporting during nighttime and in low or no air traffic areas.
(Courtesy of Barbara Brown, National Center for Atmospheric Research.)
72 AIRCRAFT ICING
local sources of uplift. For example, cold fronts are 30-50 pm, depending on the data set. Cumuliform
progressing southward through the central United clouds tend to have larger droplets than stratiform
States often provide widespread icing conditions along clouds, and clouds in continental areas have smaller
the front range of the Rocky Mountains from Wyo- droplets than those in maritime areas. Liquid water
ming through New Mexico. Cold, moist air is forced content and droplet size generally increase with
up the gentle slope leading to the steep mountain altitude in single cloud layers but the behavior is less
range. Lake-effect storms tend to be efficient in predictable in multilayered clouds. These are general
forming snow, but near the leading edges of lake- guidelines - individual clouds can and do vary
effect clouds, where ice has not yet begun to form, considerably from one another, and variations occur
significant supercooled liquid water concentrations within clouds.
can be found. There is only limited anecdotal evidence that ice
One might assume that icing will not be present aloft crystals will adhere to an airplane in flight. High
where there is significant precipitation at the surface concentrations of ice crystals alone, as in cirrus
due to scavenging of cloud liquid water. However, thunderstorm anvils, have been related to engine
examination of PIREPs has not borne this out; the problems due to ice ingestion, but they are not an
chances are about even that a PIREP will be associated airframe icing problem. A surprising number of icing
with surface snow or rain as opposed to no precipi- encounters are in mixed-phase conditions, that is,
tation. PIREPs of moderate or greater severity com- when ice crystals are present along with the super-
bined with clear or mixed icing type are significantly cooled liquid water. This appears to be the norm rather
more likely to be reported near locations of surface than the exception, and has implications for remote
observations of freezing precipitation than those detection and forecasting. Mixed-phase conditions are
locations with rain, snow, or no precipitation. This usually thought to represent a transitory state as ice
feature can be used by the forecaster as a clue for crystals will tend to grow at the expense of the liquid
potentially hazardous icing and will be discussed more droplets. However, in a case with sufficient moisture
thoroughly in a later section on large droplet icing. supply and updraft speed, enough condensate can be
produced for both deposition on ice and condensation
on droplets to occur.
Microphysical Characteristics
Measurements of the microphysical characteristics of
icing environments have been obtained from the 1940s
Supercooled Large Droplet Icing
to the present. Temperatures range from 0°C to below It has been recognized for some time that supercooled
-25"C, with a mean around -10°C. Few icing large droplets (SLDs), which are those with diameters
encounters occur at temperatures below - 20°C. At exceeding 50 pm, pose an especially serious threat to
temperatures above about - S'C, adiabatic compres- flight. Their larger size means they are not as likely as
sion of air may increase the actual temperature along small droplets to be carried around the airframe with
the leading edges of the airframe to above freezing; the airstream but more readily impact on the airframe.
typical dynamic heating corrections are 1-2°C for SLDs can impact farther aft than small droplets, which
small, slow aircraft, to as much as 6 4 ° C or higher for means that they may land on and freeze on areas not
large, faster-flying air carriers. usually protected by de-icing or anti-icing devices. The
Liquid water content (LWC) of icing environments larger droplets may also flow along the aircraft before
tends to be, on average, fairly low. In convective freezing, sending them to these unprotected locations.
clouds, 90% of the values are < 0.5-0.7 g m - 3 and for Roughness resulting from this type of ice accretion can
stratiform cloud < 0.3-0.5 g m - 3 . Maximum values create a greater aerodynamic penalty than that at the
for LWC are typically 1.2-1.3 g mP3, but can reach
N very forward edge, near the 'stagnation point' where
higher values in deep convective clouds. The distance the airflow splits to go under or over the wing. Cases of
over which certain LWC values continuously exceed increased degradation of performance due to flight in
certain limits decreases with increasing amount of SLD conditions are well documented for several
liquid; continuous LWC > 0 . 5 g m P 3 was limited to research aircraft.
distances < 13 km in one study, whereas in the same There are two general situations for formation of
study one encounter had LWC > 0.1 g m P 3 persisting SLDs. The first is the classic freezing rain process, by
along 83 km of flight path. which snow forms aloft, falls into an intruding warm
Droplets are typically small, with average mean (T>O"C) layer, melts, continues to fall into lower
diameter or median volume diameter usually between cold air (T<O"C), and becomes supercooled, ready
10 and 20ym. Maximum values for mean diameter to adhere to an airplane. This is a relatively easy
AIRCRAFT ICING 73
forecasting problem since it requires a specific ther- this time there is no one instrument that will remotely
modynamic profile. The other general case is forma- and accurately determine where supercooled liquid
tion of SLDs by coalescence of liquid drops, and is not resides in the atmosphere. Research into the pro-
so easily recognized using operationally available data per interpretation of remote sensor data for icing
sets. Wind shear (differences in wind speed and/or applications is very active at this time. TDWR
direction) at cloud top in stratiform clouds may (Terminal Doppler Weather Radar: C-band, 5 cm) and
encourage the formation of SLDs, but specific mech- NEXRAD (Next Generation Weather Radar: S-band,
anisms for SLD formation in these cases have not yet 10 cm) operational radars were not specifically de-
been identified. There is some evidence that minimum signed for icing detection, but they may yield infor-
thresholds of liquid water content must be exceeded mation that, when combined with that from other
for drizzle formation to occur: 0.2-0.25gm-3 in sources such as numerical weather prediction models,
continental clouds and around 0.1 g m - in maritime satellite imagery, or surface observations, provides
clouds. clues to the location and intensity of icing. Altering
The observation of freezing precipitation - freezing scan strategies to sample the lower atmosphere more
drizzle, freezing rain or ice pellets - at the ground can effectively, or retrofitting these radars with polariza-
provide an important clue for SLD conditions aloft. tion capability, may aid in their utility as icing
This makes physical sense since all three are super- diagnosis tools. Short-wavelength radars (such as
cooled (or already frozen) large drops: if they are K-band, 0.86cm or W-band, 3 m m ) have shown
present at the surface, they must be present for some potential for detecting icing conditions, especially in
depth above the surface. The more difficult part of nonprecipitating clouds. Dual- or triple-wavelength
using this to diagnose SLD conditions aloft is to systems, using combinations of W-, K-, X-, and longer
determine how far aloft the SLDs will be present. wavelengths, also have shown promise in quantifying
Knowledge of the moisture and thermal structure of liquid along the radar beam. These systems take
the atmosphere is needed to infer this depth. advantage of the differences in attenuation of micro-
wave radiation by atmospheric liquid at the different
wavelengths.
Since icing often occurs in mixed-phase conditions,
Detection of Icing Conditions some means must be available to discriminate between
In situ detection of icing is done visually or by the use ice and liquid in the radar data. Use of reflectivity
of instrumentation. Pilots generally have a poor view alone, even with accompanying temperature informa-
of the wings of their aircraft, so they may use the icing tion, is not sufficient. Polarization is one method by
accreting on windshields, wipers, or pitot tubes near which this may be done through determination of the
the nose of the aircraft to assess the presence and shape of the hydrometeor.
amount of ice. The pilot can also notice changes in Multichannel microwave radiometers, which pas-
aircraft performance due to icing as described in the sively detect radiation emitted from atmospheric
table of severity indicators. Icing detectors warn the constituents, have also been shown to be useful in
pilot when ice is accreting on the aircraft. In some cases identifying icing aloft. Their drawback is that they
these instruments are sensitive enough to provide an currently do not identify the altitudes at which icing
early warning before the ice becomes noticeable to the exists (although this is an active area of research), and
pilot. These airframe-mounted detectors are a fairly whether the detected liquid is supercooled. As with
mature technology, although new systems are still radar data, combining radiometer-based information
being developed. Examples of detector types are those with that from other instruments can help the
that can be flush-mounted on the wing and detect forecaster gain insight into the nature of the icing
differences in capacitance on the surface, or have a environment.
vibrating rod protruding into the airstream that Multispectral Geostationary Operational Environ-
detects the difference in resonant frequency as ice mental Satellites (GOES) show great promise as icing
accretes. Special camera-like systems that use infrared diagnosis tools. Several algorithms have been devel-
or other wavelengths of light are also being developed. oped that use combinations of visible and long- and
The advantage of in situ systems is that they provide a short-wavelength infrared channels to determine lo-
definite detection of icing conditions. However, they cations of supercooled liquid cloud tops. These algo-
have the drawback that the aircraft must necessarily be rithms will not diagnose where all icing conditions
within the icing environment, and in many cases that is exist or are absent; for example, ice-bearing cirrus may
not a desirable place to be. overlie a supercooled liquid cloud and prevent its
The use of remote sensors for detecting icing is in a detection, or a supercooled liquid layer may be present
relatively young stage of research and development. At above a cloud consisting predominantly of ice crystals.
74 AIRCRAFT ICING
~ ~ ~ ~ ~ ~~~
Nevertheless, satellite-based techniques provide the forecast using temperature and relative humidity
forecaster with an additional clue as to where icing predictions from numerical weather models. The
resides and should be used in combination with other methods tend to produce a maximum probability of
detection of icing (that is, where there was a PIREP of
data. Methods have also been devised to use micro-
wave information from satellites t o quantify the total
integrated amount of atmospheric liquid water con-
-
icing there was also a forecast of icing) of 75%. The
actual threshold values will depend on model specifics
tent over oceans. Progress in adapting these to such as horizontal and vertical resolution, the treat-
overland use is slow, owing t o the wide variations in ment of moisture in the boundary layer, and so on.
background radiative emission from the Earth’s land Adjustments can be made to mimic situations where
surfaces compared to the relatively constant values icing probabilities can be enhanced. For example,
over water. These data could be combined with other relative humidity thresholds might be lowered in cases
observations or model outputs to provide the needed where deep convection is expected, since rising turrets
information on the location and nature of icing may penetrate altitudes where model humidity is low.
conditions in clouds. Low-level stratiform clouds with relatively warm tops
and no overlying cloud layers are especially favorable
for supercooled liquid water, and algorithms can be
Forecasting Icing Conditions developed to recognize these situations from model-
Forecasting in-flight icing is the same as predicting the generated temperature and humidity profiles.
presence of supercooled liquid water in clouds-not With the advent of improved numerical weather
exactly on the list for undergraduate weather fore- prediction models with prognostic cloud liquid water
casting laboratories. Following a ‘forecast funnel’ content, the potential exists to determine where icing
process, the forecaster seeks really exists (as opposed to inferring it from smoothed
temperature and humidity fields), and to quantify the
Clouds or precipitation hazard in terms of icing severity and type. Generally,
cloud microphysical parametrizations are first devel-
Favorable temperature regime ( <O’C, > -20°C)
oped and tested on research models such as the Penn
Lift to create liquid State/NCAR Mesoscale Model (MM5), or the Colo-
Lack of significant ice to encourage glaciation rado State University RAMS, run in a post-analysis
‘diagnostic’ mode. Concepts are then coded and tested
Various guidelines linking weather features observ- for use in the operational models deployed by the US
able on the synoptic or mesoscale with the occurrence National Centers for Environmental Prediction
of icing conditions have been used with some success. (NCEP). Currently, the meso-Eta model includes
However, a better route is to apply knowledge of the liquid water prediction, and preliminary verification
atmosphere to the available information. Knowing for icing looks promising. The Rapid Update Cycle
what creates and depletes liquid, in combination with Model includes a microphysical parametrization with
information about where clouds are expected and prognostic equations for additional hydrometeor
the temperature structure of the atmosphere, should types.
provide a much more reliable and geographically One of the roadblocks to inclusion of cloud liquid
robust prediction. These concepts can also be incor- water fields in the operational models is the demand
porated into automated systems that provide the for extra and computer power and storage capacity.
forecaster with initial guidance or the nonmeteorolo- However, inclusion of an explicit liquid water field is
gist with a reasonable ‘final answer’ of where to expect really the only means toward the goal of automated
icing. prediction of severity and type of icing.
Current icing forecasting methods vary but they No one observational tool or weather prediction
share some common characteristics. They tend t o be model provides us with all we need to know about
human-intensive, time-consuming, and somewhat where icing is located, or any of its attributes such as
subjective, and they have output formats that can be type or severity. Forecasters combine the information
difficult to interpret. Until recently, operational from various sources to get the complete story on
weather prediction models did not include an icing icing; it makes sense to develop automated algorithms
product or clouds or liquid water fields. Thus, the to accomplish the same goal. Automated versions of
forecast problem was t o determine, using the available this human technique are being developed and to date
output, where supercooled cloud was likely to be have proved quite successful in diagnosing where icing
located. conditions reside. Predictive capabilities, by which
In the absence of explicit information on the sensor-based icing features are identified and extrap-
locations of clouds, it is possible to make a reasonable olated forward, represent the means by which this
AIRSEA INTERACTION/ Freshwater Flux 75
combined approach may be taken into the future. The Mountain Meteorology. Parameterization of Physi-
key to successful icing forecasting lies in understand- cal Processes: Clouds. Radar: Precipitation Radar.
ing the physical processes resulting in supercooled Satellite Remote Sensing: Precipitation; Temperature
liquid water production, how these processes relate to Soundings; Water Vapor. Thermodynamics: Moist (Un-
observable phenomena, and how to combine infor- saturated) Air. Weather Prediction: Regional Prediction
Models.
mation from as many sources as possible to gain the
most complete picture of the icing situation.
Further Reading
Glossary Cooper MA, Sand WR, Politovich MK and Veal DL (1984)
Effects of icing on performance of a research airplane.
Supercooled Refers to liquid water cooled to below
]ournu1 ofAircraft 21: 708-715.
0°C without becoming a solid (ice). Hansman RJ Jr (1985) Droplet size distribution effects on
Glaze ice Ice with a translucent, glossy appearance. aircraft ice accretion. Journal of Aircraft 22: 503-508.
May be smooth or have embedded lumps. Some- Lankford TT (2000) Aircraft Icing: A Pilot’s Guide (Prac-
times referred to as ‘clear’ ice. tical Flying Series). New York: McGraw-Hill.
Rime ice Opaque, brittle ice that tends to form Matrosov SY, Reinking RF, Kropfli RA and Bartram BW
‘feathers’ into the airstream. (1996) Estimation of ice hydrometeor types and shapes
Mixed ice A combination of rime and glaze ice, from radar polarization measurements. Journal of
caused by variations in atmospheric parameters Atmospheric and Oceanic Technology 13: 85-96.
resulting in either type of ice. Politovich MK (1995) Response of a research aircraft to
Severity Refers to the combination of environmental icing and evaluation of severity indices. Journal of
Aircraft 33: 291-297.
icing intensity, aircraft response, and pilot interpre-
Schultz P and Politovich MK (1991) Toward the improve-
tation of that response. ment of aircraft icing forecasting for the continental
United States. Weather and Forecasting 7: 491-500.
Thompson G, Bruintjes RT, Brown BG and Hage F (1997)
See also Intercomparison of in-flight icing algorithms. Part I:
Cloud Microphysics. Cyclones, Extra Tropical. Hu- WISP94 Real-time Icing Prediction and Evaluation Pro-
midity Variables. Mesoscale Meteorology: Overview. gram. Weather and Forecasting 12: 878-889.
Contents
Freshwater Flux
Gas Exchange
Momentum, Heat and Vapor Fluxes
Sea Surface Temperature
Storm Surges
Surface Waves
Introduction
Freshwater Flux The world ocean is a key element of the physical
climate system. The ocean contains 97% of the world’s
J Schulr, Meteorological Institute, University of Bonn, water and covers an area of 71% of the globe. As a
Bonn, Germany reservoir, the ocean supplies water vapor to the
atmosphere that brings rain and snow over land
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. surfaces. About one-third of the precipitation over
combined approach may be taken into the future. The Mountain Meteorology. Parameterization of Physi-
key to successful icing forecasting lies in understand- cal Processes: Clouds. Radar: Precipitation Radar.
ing the physical processes resulting in supercooled Satellite Remote Sensing: Precipitation; Temperature
liquid water production, how these processes relate to Soundings; Water Vapor. Thermodynamics: Moist (Un-
observable phenomena, and how to combine infor- saturated) Air. Weather Prediction: Regional Prediction
Models.
mation from as many sources as possible to gain the
most complete picture of the icing situation.
Further Reading
Glossary Cooper MA, Sand WR, Politovich MK and Veal DL (1984)
Effects of icing on performance of a research airplane.
Supercooled Refers to liquid water cooled to below
]ournu1 ofAircraft 21: 708-715.
0°C without becoming a solid (ice). Hansman RJ Jr (1985) Droplet size distribution effects on
Glaze ice Ice with a translucent, glossy appearance. aircraft ice accretion. Journal of Aircraft 22: 503-508.
May be smooth or have embedded lumps. Some- Lankford TT (2000) Aircraft Icing: A Pilot’s Guide (Prac-
times referred to as ‘clear’ ice. tical Flying Series). New York: McGraw-Hill.
Rime ice Opaque, brittle ice that tends to form Matrosov SY, Reinking RF, Kropfli RA and Bartram BW
‘feathers’ into the airstream. (1996) Estimation of ice hydrometeor types and shapes
Mixed ice A combination of rime and glaze ice, from radar polarization measurements. Journal of
caused by variations in atmospheric parameters Atmospheric and Oceanic Technology 13: 85-96.
resulting in either type of ice. Politovich MK (1995) Response of a research aircraft to
Severity Refers to the combination of environmental icing and evaluation of severity indices. Journal of
Aircraft 33: 291-297.
icing intensity, aircraft response, and pilot interpre-
Schultz P and Politovich MK (1991) Toward the improve-
tation of that response. ment of aircraft icing forecasting for the continental
United States. Weather and Forecasting 7: 491-500.
Thompson G, Bruintjes RT, Brown BG and Hage F (1997)
See also Intercomparison of in-flight icing algorithms. Part I:
Cloud Microphysics. Cyclones, Extra Tropical. Hu- WISP94 Real-time Icing Prediction and Evaluation Pro-
midity Variables. Mesoscale Meteorology: Overview. gram. Weather and Forecasting 12: 878-889.
Contents
Freshwater Flux
Gas Exchange
Momentum, Heat and Vapor Fluxes
Sea Surface Temperature
Storm Surges
Surface Waves
Introduction
Freshwater Flux The world ocean is a key element of the physical
climate system. The ocean contains 97% of the world’s
J Schulr, Meteorological Institute, University of Bonn, water and covers an area of 71% of the globe. As a
Bonn, Germany reservoir, the ocean supplies water vapor to the
atmosphere that brings rain and snow over land
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. surfaces. About one-third of the precipitation over
76 AIR-SEA INTERACTIONI Freshwater Flux
land originates from water evaporated from the ocean. written as eqn [ 11.
The water vapor in the atmosphere is the most
important gaseous absorber for solar and terrestrial P=E+D 111
radiation and accounts for about half of the atmos-
phere’s natural greenhouse effect. The process of In eqn [l],Pis precipitation, E is evaporation, and D is
evaporation accounts for approximately half of the discharge or river runoff. This balance states that
surface cooling balancing the heating by absorption of water added to the surface by precipitation is parti-
solar radiation. tioned between E and D . Although globally precipi-
Because of the asymmetric insolation of the Earth’s tation and evaporation are balanced, the large
surface by solar radiation, the oceans act as a large differences in the components P , E , and D over land
energy and heat transport system from the Equator to and over ocean produce the world’s water cycle,
the poles. The deep-ocean circulation that is critical shown schematically in Figure 1. The units used
for this transport is mostly driven by variations in the throughout the chapter for E, P , and E - P are
density of sea water. Ocean salinity is an important mm d - In general, total evaporation exceeds pre-
contributor to these variations and varies with latitude cipitation over oceans, which is compensated for by
in the upper layers of the oceans. The surface salinity the runoff of rivers from the continents, where
depends on the fresh water flux at the ocean surface precipitation exceeds evaporation.
and is relatively high in the subtropics where evapo- In the following sections the focus will be on the
ration exceeds precipitation, whereas it is relatively different methods used to determine the freshwater
low in the tropics and middle and high latitudes where flux at the oceans surface: E , P , E - P will be
precipitation dominates. considered in detail. The heat transport within the
The most comprehensive publication about the oceans is beyond the scope of this article (see Air-Sea
world water balance was written by Baumgartner and Interaction: Momentum, Heat and Vapor Fluxes).The
Reichel in 1975. They assembled different estimates of second section gives an overview of different
the water balance over continents, river basins, and techniques that can be used to determine the fresh
oceans to calculate a global water balance. This water flux, followed by a description of the fresh water
pioneering work resulted in global maps of evapora- flux climatology as derived from satellite data. This is
tion and precipitation thar even today are widely used followed by a short consideration of the role of river
by meteorologists, oceanographers, climatologists, runoff, and finally some conclusions are presented.
and hydrologists. The definition of the water balance
and its components used here follows the work of Methods for Determining
Baumgartner and Reichel.
Assuming that the amount of water on Earth is not
E , P , and E - P
changing with time, the long-term average of the water Basically, there exist three different methods for
balance for a unit area of the Earth’s surface can be determining the components E and P as well as
Figure 1 Schematic representation of the world water cycle E and P denote evaporation and precipitation over oceans and land
(denoted by subscript 0 and L, respectively). D is the discharge or river runoff of water from the continents to the oceans and MT is the
moisture transport in the atmosphere.
AIR-SEA INTERACTION/ Freshwater Flux 77
E - P. Traditional estimates of E are based on irt situ during the last few years, the global heat balance has
measurements of surface wind speed U, specific not been closed, mostly because of the low observation
humidity of air Qa, and sea surface temperature Ts. density and deficiencies in the bulk aerodynamic
These are used within the so-called bulk aerodynamic formula. The determination of precipitation is even
formula to parametrize the evaporation (seeTurbulent more difficult. It is largely based on the observed actual
Diffusion). P is estimated by analyzing actual weather weather and parametrizations that convert the weath-
reports using an empirical parametrization. The er code into rainfall amount. The conversion schemes
resulting estimates are interpolated and extrapolated were developed from data that are not representative
to construct global maps of E and P. of the global oceans, so it became necessary to correct
Recently, satellite data have been used to derive the under-estimated precipitation in the Tropics by em-
same basic state variables near and at the ocean surface pirical temperature-dependent corrections. Incorpo-
using empirical and physical retrieval schemes. Again ration of P measurements from islands in data-sparse
these quantities are used to parametrize E using the regions is also very difficult, because of the influence of
bulk aerodynamic formula. Many algorithms for island terrain on the rainfall. Comparisons of these
estimating rainfall using almost the whole electro- precipitation fields with satellite-derived fields exhibit
magnetic spectrum have been developed during the large differences even at the climatological scale.
last 20 years (see Satellite Remote Sensing: Precipita-
tion). These satellite algorithms are used alone and in
conjunction with irt situ data and model results to give
Remote Sensing of E -P
best estimates. Remote sensing of evaporation is based mostly upon
The third method is the so-called moisture budget the derivation of basic state variables, wind speed, sea
method, which make use of global-scale analyzed surface temperature, and near-surface atmospheric
water vapor fields or measurements of atmospheric specific humidity, and the parametrization of the
water vapor by rawinsondes in the form of four- evaporation using the bulk aerodynamic formula.
dimensional data assimilation (see Data Analysis: Wind speed can be obtained from either passive or
Time Series Analysis). The global distribution of E - P active microwave systems. The active system relates
is then computed from the residuals of water vapor the backscattered energy to the wind speed at a
transport in the atmosphere using large-scale numer- reference level over sea surface (e.g., 10 m) and is also
ical models. able to deliver wind direction information. The
passive systems rely on the surface emission change
due to wind-induced sea surface roughness and partial
Traditional Estimates from in situ Measurements
foam coverage. The root-mean-square (rms)errors on
Most of our present knowledge of fresh water fluxes is an instantaneous time scale for both systems are on the
derived from weather observations on special weather '.
order i1.3 to k 2 m s - Remote sensing methods for
ships, buoy data, and also data from merchant ships the near-surface specific humidity make use of a
participating in the Voluntary Observing Ship system. vertically integrated water vapor content (obtained
Many of these data have been organized into the from a passive microwave instrument such as the
Comprehensive Ocean Atmosphere Data Set which Special Sensor Microwave/Imager (SSWI)) as a pre-
has been used to derive climatologies of the energy dictor. Several techniques ranging from linear regres-
fluxes and the fresh water flux at the sea surface. The sion analysis to neural networks have been used to
major disadvantage of ship-based estimates of E and P
is that the observation base is not very good for either
parameter. The coverage is mostly obtained along
-
deduce the near-surface humidity with rms errors of
k 0 . 7 g k g - I on the monthly time scale. Estimates
of sea surface temperature are deduced from passive
shipping lanes, which may be sufficient in the North infrared sensors like the Advanced Very High Resolu-
Atlantic, North Pacific, and the Mediterranean but is tion Radiometer (AVHRR).The largest problem with
not sufficient in the Tropics and all southern oceans. this method is the cloud clearance of the satellite scene,
Additionally, the concentration along shipping lanes because otherwise the determined Ts would be nega-
can introduce a fair weather bias, since ships try to
avoid bad weather.
Whereas the measurement quality for the basic state
-
tively biased. Accounting for the surface skin effect,
rms errors for the best satellite methods are f0.2 K.
Estimates of sea surface temperature with not much
variables used for the parametrization of E is relatively less accuracy are also possible employing passive
good on ocean weather ships and research quality
buoys, it is less good on the voluntary observing ships.
Although much effort has been put into correcting
-
microwave measurements at frequencies between
5-10 GHz that are available from TRMM's TMI
and will shortly be available from the new Advance
errors on the basis of individual ship measurements Multifrequency Scanning Radiometer onboard the
78 AIR-SEA INTERACTION/ Freshwater Flux
Aqua and ADEOS-I1 satellites. The big advantage of classify the quality of the satellite algorithms. A
those estimates is the much better coverage because prominent data set using combinations of geostation-
clouds are almost transparent at those frequencies ary satellite data, passive microwave data, and rain
allowing an undisturbed view of the ocean surface. gauge data is that produced by the Global Precipita-
However, infrared estimates of sea surface tempera- tion Climatology Project (GPCP).
ture remain of high importance for the computation of With the launch of the satellite of the Tropical
evaporation climatology since estimates of sea surface Rainfall Measuring Mission (TRMM)in 1997, for the
temperature from SSWI measurements were not first time a space-borne radar can be used to derive the
possible with sufficient accuracy. Recently, some new three-dimensional structure of rainfall and the surface
methods have been developed that circumvent the rainfall. The combination of instruments onboard the
bulk formula in the retrieval process by relating the TRMM satellite can be considered as a reference for
satellite data directly t o an existing flux data set, e.g., methods applied to other instruments in space. A
re-analyses of fields derived from in situ data that are calibration of rainfall estimates from other satellites
assumed to be true. The accuracy of all methods like the SSM/I then delivers an optimal combination of
is comparable at a level of f 3 0 W m P 2 and accuracy and temporal/spatial sampling and subse-
i l 5 W m P 2 at weekly and monthly time scales, quently consistent rainfall distributions.
respectively.
As mentioned in the previous section, there are not Moisture Budget Methodology
sufficient conventional and surface-based radar rain- The moisture budget methodology tries to compute
fall estimates over the oceans for the derivation of E - P as a residual from the large-scale atmospheric
rainfall fields. A reasonable alternative is the use of transports of water vapor using global analyses and re-
satellite remote sensing. Remote sensing of rainfall analysis data sets produced with four-dimensional
from satellites started with the statistical analysis of data assimilation schemes. This technique has a long
the reflectivity and emissivity of the upper cloud layers history, although it usually makes use of rawinsonde
at visible and infrared wavelengths, respectively. data directly. E - P is computed from eqn [2],which is
Because of the small physical correlation between the vertically integrated (from the top of the atmos-
the signal and the rainfall at the surface, this technique phere to the surface) equation for the conservation of
leads to acceptable results only if the derived rainfall is water vapor.
integrated over space and time. Owing to the strong
variability of rainfall, rainfall climatologies derived
using this technique with data from geostationary
satellites with their high repetitive cycle build are still
the backbone of today’s rainfall analyses. W is the total precipitable water, q is the specific
Over water surfaces, passive microwave radiome- humidity, p is pressure (with p s being the surface
ters deliver a much better information base. The signal pressure), v is the velocity vector, and g is the standard
at frequencies below 30 GHz is mostly determined by gravity. Many comparison studies between precipita-
the emission from rain water, which leads t o a strong tion fields produced routinely and the GPCP data set
increase of the brightness temperature over the cold found discontinuities in the analyses due to changes in
background of the sea surface. For higher frequencies, the data assimilation system. Another problem was
the brightness temperature decreases owing to scat- that rainfall maxima in the analyses were often in the
tering by ice particles. This information can also be wrong place and too strong. In general, the assimila-
used to estimate the rainfall rate at the surface. Since tion systems have been much improved during the last
the launch of the first SSM/I onboard the satellites of ten years and are more or less consistent with satellite-
the Defense Meteorological Satellite Program (DMSP) derived data sets. However, use of this method with
in 1987, a continuous time-series of data exists from at different analysis and re-analysis products from the
least one satellite. Many algorithms have been devel- European Center for Medium-range Weather Fore-
oped to analyze rainfall using these measurements. On casts (ECMWF), the National Centers for Environ-
the basis of numerous algorithm intercomparison mental Prediction (NCEP), and the National
projects, it has proved almost impossible t o find a so- Aeronautics and Space Administration (NASA) as
called standard algorithm that performs best for most input and comparisons to pure model-computed
of the situations investigated. The variability of the E - P exhibited large differences. The most critical
cases analyzed showed the quality of some algorithms part of this method is the dependence of the moisture
under certain conditions, but not of one prevailing budget on the divergence of the velocity field. This
algorithm. In many cases it was found that the is of special importance in the Tropics, where the
accuracy of the validation data was not sufficient to divergence field is not very well known.
River Inflow Maximum values of evaporation up to 9 mm d - '

are observed over the Kuroshio and Gulf Stream
The inflow of fresh water from rivers is not included in
regions in winter (Figure 2A) and minimum values
most ocean surface fresh water data sets derived from
below 1mm d - l are seen in the eastern equatorial
satellites and is also neglected in the residual approach,
Pacific and Atlantic during all months. Large areas
but it is significant to the global fresh water balance of
with high evaporation rates of 5-6 mm d - are found
the ocean. Baumgartner and Reichel estimated from a
in the main Trade Wind belts between about 10" and
global hydrological balance calculation that the con-
40" latitude in both the Northern and Southern
tribution of river runoff to the balance is as high as
hemispheres. These high evaporation regions are the
10% of the contribution of precipitation. It might be
major sources of atmospheric water for the global
thought that the impact of the fresh water inflow from
hydrological cycle. Whereas during Northern Hemi-
the rivers on the buoyancy would be local in compar-
sphere winter and spring the maximum extent and the
ison to the size of the ocean basins. However, the
highest values are found north of the Equator (Figures
impact of major rivers like the Amazon or Congo is
2A and 2B), maximum evaporation is observed in the
observable several hundred kilometers away from the
southern Indian, Atlantic, and Pacific Oceans during
mouths of the rivers. Neglecting this contribution
Northern Hemisphere summer and fall (Figures 2C
would increase the average salinity of the upper
and 2D).
Atlantic ocean (the first 5 0 m ) by 1.5 psu after 10
The global precipitation pattern is dominated by a
years of integration of a numerical ocean model.
strong band of precipitation circling the globe just
Two major runoff data sets are the Global River
north of the Equator. This is the region where the
Discharge Catalogue published by the International
northern and southern Hadley circulation cells meet,
Hydrography Program (IHP) and that issued by the
forming a region of strong surface convergence known
Global Runoff Data Center, and these form the
as the Intertropical Convergence Zone (ITCZ).Where
backbone of information on river runoff. IHP data
the maximum precipitation on an annual scale exceeds
consists of a selection of monthly discharges at 949
stations over six continents (Africa, Asia, North
'.
6 mm d - Another convergence zone in the western
tropical Pacific, known as the South Pacific Conver-
America, South America, Europe and Australia/
gence Zone (SPCZ), is somewhat broader, with
Oceania), but only 219 stations are listed as corre-
precipitation values similar to those in the ITCZ. It
sponding to a direct discharge into an ocean basin. The
extends from the region of Indonesia and the Philip-
length of the individual data sets is 19.3 years on
pines south-east across the southern Pacific. With the
average, but varies from 1 year to 100 years. Addi-
onset of the summer monsoon, the ITCZ, which was
tionally, a great disparity exists between the different
earlier located in the southern Indian Ocean, shifts to
continents, with Europe and North America present-
its northernmost location and merges with the Mon-
ing the longest records. Not included in either data set
soon trough giving rise to copious rainfall over the
is the runoff f r q n Arctic and Antarctic regions and the
Indian subcontinent and the adjacent seas, namely the
inflow of fresh water from ground water sources.
Arabian Sea and Bay of Bengal (Figure 3C).
Analysis of the monthly mean climatology of the
On an annual scale, the region off the Indonesian
direct contribution of rivers to the fresh water flow
islands receives a maximum rainfall of more than
into the ocean, considering only rivers for which time
-
records longer than two years exist, sums to 0.57 Sv
(1Sv = lo6 m3 s - ').This estimate is much lower than
10 mm d - Outside the two convergence zones,
precipitation rates are significantly lower, with the
exception of two regions. Precipitation rates are quite
that of Baumgartner and Reichel, who found 0.73 Sv,
high over the Gulf and Kuroshio Streams, with values
which also includes contributions from regions be-
as high as in the ITCZ during the period November to
yond the polar circles, which they estimated from
March (Figure 3A). This feature has not been recog-
other sources.
nized in rainfall climatologies derived from routine
weather observations.
Climatology of E - P Derived from At tropical and subtropical latitudes, between about
1 5 " N and 40"N and between 5 " s and 30"S, the
Satellite Data eastern parts of the Pacific and the Atlantic ocean are
Figures 2-4 show the seasonally averaged global maps regions where precipitation is below 1mm d-'. A
of evaporation, precipitation, and E - P derived from similar region is found in the Indian Ocean along
AVHRR and SSWI data, on a grid with 1"x 1" the east coast of Africa and Saudi Arabia during all
resolution. The data set was constructed by averaging months, with a maximum extent over the
instantaneous estimates of the components and the whole Arabian Sea during winter (Figure 3A). In the
flux over 11 years (1987-1998). Southern Hemisphere, west of Australian coast,
80 AIR-SEA INTERACTION/ Freshwater Flux
-’as derived from satellite data: (A) December, January, February; (B) March, April, May; (C) June, July, August; (D)
r
t
h
h
U
m
I
Figure 2 Climatological seasonal averages of evaporation in mmd

E
E
E
a
i
m
id
September, October, November.

-’asderivedfrom satellite data: (A) December,January, February;(6)March,April, May; (C) June, July, August; (D)September,
r
U
C
E
E
Figure3 Climatologicalseasonalaveragesof rainfallin mm d

October, November.
82 AIR-SEA INTERACTION I Freshwater Flux
’
n
n
n
7
u-
Lu
e
-P
a
.-C
al
i
Figure 4 Climatological seasonal averages of E - P in mm d - as derived from satellite data: (A) December,January, February;(B)March, April, May; (C) June, July, August; (D) September,
(u
P
E
E
al
i
m
4
i.
8cn
0
C
P
0
20
eal
E
E
I
October, November.
another minimum is observed, which has its maximum Table 1 Global climatological averages for E , P, and E - P
extension during the Southern Hemisphere spring (in mmd-') over oceans, from different studies
(Figure 3D). Source E P E-P
The patterns of precipitation and evaporation
exhibit quite different spatial distributions. Precipita- Baumgartner and Reichel 1177 1066 111
(1975)
tion maxima occur in the global convergence regions, Chahine (1992) 1202 1088 114
while evaporation maxima occur in regions of high ECHAM4a(Todini and 1246 1147 99
surface humidity gradient and wind speed. On a Dumenil, 1999)
monthly time scale, values of E exhibit much less ECMWF~+GPCP~ rain 1194 1083 111
spatial structure than fields of P. From this it is clear (Oki, 1999)
HOAPSd(Grassl eta/.,2000) 1086 908 178
that the E - P monthly temporal variability is dom-
inated by variations in location and intensity of aECHAM4: Cimate model of the Max-Planck Institute for Meteor-
rainfall and the spatial structure in E - P is also ology, Hamburg.
dominated by the P field. bECMWF: European Centre for Medium-range Weather Fore-
casts.
However, on the climatological time scale, fields of 'GPCP: Global Precipitation Climatology Project.
E - P consist of signatures of both evaporation and dHOAPS: Hamburg Ocean Atmosphere Parameters and Fluxes
precipitation fields. The ITCZ and SPCZ appear from Satellite Data.
prominently as regions of fresh water supply to the
ocean. In these regions the fresh water flux from
atmosphere to ocean is generally larger than
Discussion and Conclusion
4 mm d - I. With the exception of the SPCZ, precipi-
tation decreases rapidly with latitude to the north and Table 1 shows how existing estimates of the fresh
to the south of the ITCZ, while evaporation remains water flux and its components differ from some
strong or even increases, causing positive values of the examples of estimates from different sources found
fresh water flux. in the literature. Whereas the older estimates from
The strongest gradients in the fresh water flux fields observations are comparable to the results from
occur in the boundary regions between the negative General Circulation Models, the satellite estimates
values of 4 m m d - ' within the ITCZ and the differ considerably from all of them. However, today
strong positive flux regions to the north and south, there is no agreed true value for E - P. Current results
with values up to Gmmd-I. Poleward from the of global analyses seem not to be very reliable, but
evaporation regions, the fresh water flux is relatively centers like ECMWF are improving the assimilation of
small. The evaporation fields generally decrease rainfall estimates from satellite data, and these will be
toward the poles primarily as a result of the decrease operational in a few years. Satellite data sets have great
in the humidity difference, except during wintry arctic potential to be improved in the future by using
cold air outbreaks which often lead to very high sophisticated methods of intercalibration between
evaporation rates and therefore to positive fresh water different satellites. In the case of basic state variables
fluxes. Although evaporation is quite high, large U , Qa, and Ts, improvements are expected from
negative values of E - P can be found in the Gulf intercomparison of the satellite estimates with high-
Stream and Kuroshio regions, below - 4 mm d - * quality surface-based measurements.
during the winter owing to high precipitation, while
in all other months evaporation almost balances
precipitation. See also
An analysis of the fresh water flux on a seasonal
Aerosols: Climatology of Tropospheric Aerosols. Air-Sea
scale (Figures 4A-D) reveals that the eastern parts of
Interaction: Momentum, Heat and Vapor Fluxes; Sea
the Arabian Sea, the Bay of Bengal, and the South Surface Temperature. Clouds: Climatology. Data Anal-
China Sea all have negative values of fresh water flux ysis: Time Series Analysis. Humidity Variables. Satel-
during summer (Figure 4C) and autumn (Figure 4D). lite Remote Sensing: Precipitation, Turbulent
Further, on an annual scale, it can be seen that the Diffusion.Weather Prediction: Data Assimilation.
eastern equatorial Indian Ocean, the Bay of Bengal,
and the Kuroshio and the Gulf Stream regions all
exhibit negative fluxes. The regions of positive flux are Further Reading
over the north-west Arabian Sea and the southern Baumgartner A and Reichel E (1975) The World Water
Indian Ocean south of 20" S. Also, the North Atlantic Balance. Muenchen Wien: R. Oldenbourg Verlag.
and the South Atlantic exhibit positive fluxes on both Chanine M T (1992)The hydrological cycle and its influence
sides of the ITCZ. on climate. Nature 359: 373-380.
84 AIRSEA INTERACTIONI Gas Exchange
Grass1 H, Jost V, Schulz J, et al. (2000) A Climatological Taylor PK (ed.) (2000) lntercomparison and validation of
Atlas of Satellite-derived Air-Sea Interaction Parameters ocean-atmosphere energy flux fields. Final report of the
over the Worlds Ocean, Max-Planck Report No. 312. Joint WCRP/SCOR Working Group on Air-Sea fluxes,
Hamburg: Max-Planck Institute for Meteorology. URL: WCRP-112, WMO/TD NO. 1036. URL: http://
http:Nwww.mpimet.mpg.de/Depts/Physik/HOAPS www.soc.soton.ac.uk!JRD/METlWGASF
Hartmann DL (1994) Global Physical Climatology. San Todini E and Dumenil L (1999) Estimating large-scale
Diego: Academic Press. runoff. In: Browning KA and Gurney RJ (eds) Global
Josey S, Kent EC and Taylor PK (1999)New insights into the Energy and Water Cycles, pp. 265-277. Cambridge:
ocean heat budget closure problem from analysis of the Cambridge University Press.
SOC air-sea flux climatology. Journal of Climate 12: Trenberth KE and Guillemot CJ (1999) Estimating evapo-
2 856-2 8 80. ration-minus-precipitation as a residual of the atmos-
Oki T (1999) The global water cycle. In: Browning KA and pheric water budget. In: Browning KA and Gurney RJ
Gurney RJ (eds) Global Energy and Water Cycles, pp. (eds) Global Energy and Water Cycles, pp. 236-246.
10-27. Cambridge: Cambridge University Press. Cambridge: Cambridge University Press.
Gas Exchange
P D Nightingale, Plymouth Marine Laboratory, nisms, with the aim of predicting concentration fields
Plymouth, UK by using oceanic/atmospheric models.
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. Rather less progress has been made in understand-
inn- the basic mechanisms behind air-sea gas- transfer,
although the two most important variables are
Introduction molecular diffusivity and the degree of turbulence
The air-sea interface acts as the conduit for the close t o the air-water interface. One of the main
transfer of gases such as oxygen, dimethyl sulfide, parameters linked to turbulence is wind speed, al-
carbon monoxide, carbon dioxide, and methyl iodide though observations from laboratory studies suggest
between the oceans and atmosphere. The air-sea that other variables such as surfactants, breaking
interface is therefore of fundamental importance in waves, and bubbles also influence k. Although models
studies of marine productivity, biogeochemical cycles, predict that k should rise with increasing wind speed
atmospheric chemistry, climate, and human health. (or more correctly friction velocity), the dominant
For example, about 30% of the world’s population is mechanism is not agreed upon. A major problem is a
at risk of iodine deficiency disorders that impair dearth of knowledge concerning the sea surface
mental development. The main supply of iodine to microlayer (usually defined as the top l m m of the
land is the transfer of volatile iodine compounds sea) and its properties.
produced in the oceans to the atmosphere via the air- Given the lack of a comprehensive understanding of
sea interface. A further example is that the oceans take the processes controlling gas exchange, various ingen-
up about 30% of the anthropogenic C 0 2 annually ious techniques have been developed t o try to measure
emitted to the atmosphere and represent the long-term k in situ in order to develop and test parameterizations
sink for most of the predicted future emissions of based on easily measurable and readily available
anthropogenic COS. environmental variables such as wind speed. An
It has proved t o be extremely difficult t o measure improvement on wind speed-based parameterizations
air-sea gas fluxes in situ. As a result, they have been is dependent on a better understanding of the proc-
calculated from the product of the concentration esses controlling air-sea exchange and on new tech-
difference between the two phases and a kinetic (or niques to measure k over short time scales.
rate) term known as the gas transfer coefficient (k).
The gas transfer coefficient is also known as the
transfer velocity as it has dimensions of length per unit
The Two-Film Model of Gas Exchange
time. As most gases of interest are produced and/or The simplest model of air-sea gas transfer is the two-
destroyed within the ocean or atmosphere, there is film model illustrated in Figure 1. Although physically
considerable spatial and temporal variability in con- unrealistic, it is useful in visualizing how gas transfer
centration fields. Large international research pro- between the two interfaces may occur. The model
grams have attempted t o quantify this variability and assumes that the main bodies of air and water are well
to understand gas production and removal mecha- mixed and that transfer through the two thin films is by
AIR-SEA INTERACTION /Gas Exchange 85
Concentration sparingly soluble (high H ) gases (e.g., oxygen, carbon
(3
dioxide, methane, methyl bromide) the rate-limiting
step is transfer through the aqueous thin film because
Perfect Eddy diffusion molecular diffusion through water is considerably
slower than in air. The term Ilk, then dominates and,
mixing ignoring the second equality, eqn [ 3 ] simplifies to the
ca more familiar expression for estimating air-sea gas
fluxes, eqn [ 5 ] .
Air-sea Molecular diffusion only F = k,($-C,)
interface
Molecular diffusion only

For some gases that either react with water or are
highly soluble, the term 1 / H k , dominates. These gases
include hydrogen chloride, sulfur dioxide, water, and
probably ammonia. There are relatively few gases
for which eqn [ 3 ] does not simplify, although recent
laboratory experiments suggest that transfer through
both films may need to be considered for flux estimates
of dimethyl sulfide at low temperatures or moderate
Figure 1 The two-film model of gas exchange. The various wind speeds and of some organic gases (e.g., PCBs and
parameters are defined in the text. (Adapted with permission from pesticide vapors). As there is much information on the
Nature ( L i s PS, Slater PG (1974) Flux of gas across the air-sea term ka from research into air-sea fluxes of water
interface 247: 181-1 84) Copyright (1974) Macmillan Magazines vapor (see Air-Sea Interaction: Momentum, Heat and
Limited.)
Vapor Fluxes) the remainder of this article will
concentrate on the determination of k, and air-sea
fluxes of sparingly soluble gases.
molecular diffusion only. The net flux of gas through If k, is known for any particular gas, then it can be
either film is given by the product of the concentration derived for another gas using eqn [ 6 ] .
difference across the film and a kinetic term known as
the gas transfer coefficient ( k ) (eqn [ l ] ) .
In eqn [ 6 ] ,Sc (the Schmidt number) is temperature

If the gas obeys Henry’s law and assuming the system is dependent and is defined as the ratio of the kinematic
at steady state, then the relationship between C,i and viscosity of sea water and the molecular diffusivity of
C,i is given by eqn [2] where H is the Henry’s Law the gas of interest. Note that the thin-film model
coefficient. implies that the value of n (the Schmidt number
dependence) is unity as k , would be proportional to
molecular diffusivity.
121
It can be shown that eqn [ 3 ] holds, where K, and Ka Chemical Enhancement

are given by eqn [4].
The transfer rate of CO2 may be enhanced by reaction
with O H - in the aqueous thin film. This effect is most
important at low wind speeds and high water temper-
atures (e.g., in equatorial regions) and may cause an
1 1 1 1 H 1 enhancement of up to 8 % in k, in these regions. N o
-
K w k w Hka
+-, -- -_
Ka kw k a
+- 141 experiment has yet shown that this effect operates
at sea.
The flux of gas between the air-water interface is
therefore given by the concentration difference be-
tween the bulk air and bulk sea water after correcting Laboratory Experiments with
for solubility (i.e., the degree of disequilibrium be- WindMave Tanks
tween the two phases) and an overall transfer coeffi-
cient that is itself dependent on the individual transfer The vast majority of experiments designed to under-
coefficients in the air and water. In practice, for most stand the mechanisms behind the process of air-sea
86 AIRSEA INTERACTION/ Gas Exchange
gas exchange and to identify variables that influence

k , have been conducted in wind tunnels. These vary Atmosphere
greatly in scale and sophistication and may have the
ability to mechanically generate waves and bubbles Surfacing
and possess a return flow for the air and water phases -
Le., a closed system.
These experiments have shown that the thin-film
model is too simple. At low wind speeds k , was found exchange
to vary with SC-'.~'. This behavior is in agreement
with boundary layer models. However, once wind-
Ocean
induced waves were observed on the surface of the
water, k , varied with SC-'.~. This is in agreement with
surface renewal models, where water in the aqueous
Figure 2 A schematic of bubble evolution after entrainment into
thin film is periodically replaced by fluid from the bulk the upper ocean. (Reproduced with permission from Woolf DK
water. The exact wind speed (or friction velocity) at (1997) Bubbles and their role in gas exchange. In: Liss PS,Duce
which this regime changes was found to vary with the RA (eds) The Sea Surface and Global Change, pp. 173-205.
facility being used. At high wind speeds when wave Cambridge University Press.)
breaking and bubbles are present, an enhancement in
k , was observed that was dependent on the solubility
of the gas. This solubility effect is due to the limited bubble then either dissolves completely or rises to the
volume of air contained within a bubble. surface, where it again disrupts the microlayer causing
Artificial surfactants have long been known to surface renewal. It is not yet clear how much bubbles
reduce k , in windlwave tanks. Laboratory experi- and breaking waves contribute to gas exchange at sea,
ments with sea water have shown that the decrease in although models of bubble-mediated gas transfer
k , correlates inversely with bulk-water chlorophyll predict that k , for insoluble gases will be significantly
and with colored dissolved organic matter. Except for enhanced.
low wind speeds, the reduction in k , is not caused by To summarize, it seems reasonable to assume that
insoluble surfactant films (slicks) but is due to the there are several mechanisms, some of them compet-
presence of soluble surfactants that alter the hydro- ing, that influence gas exchange rates in the seas and
dynamic properties of the sea surface. Soluble surfact- oceans. Although wind and wave tank experiments
ants can reduce gas exchange even in the presence of suggest that there are three clear regimes - smooth,
breaking waves. Laboratory studies have found, ripples, and breaking waves - in which gas exchange
however, that k, showed a reasonable correlation rates can be predicted from wind speed, such predic-
with the total mean square wave slope (particularlyfor tion is much more difficult in a system as complex and
the shorter wind waves), suggesting that this param- variable as the ocean, where wind is not intimately or
eter, although difficult to measure at sea, might be a immediately linked to wave field.
useful predictor of gas exchange rates.
The good correlation with the mean square wave
slope of shorter wind waves has led investigators to In Situ Measurements of Gas
assume that a wave-related mechanism controls k , at Transfer Rates
low and intermediate wind speeds. This mechanism is Large-Scale Measurements
thought to be increased turbulence due to microscale
wave breaking that leads to surface renewal. However, Radiocarbon Radiocarbon ( 14C) is produced natu-
these microscale breakers might be suppressed at rally in the atmosphere and transferred across the air-
higher wind speeds owing to the presence of large sea interface into the oceans. If the system is in
breaking waves. equilibrium, the flux of 14C from the atmosphere is
Large breaking waves are also known to be associ- balanced by radioactive decay in the water column. An
ated with high levels of near-surface turbulence and estimate of the global annual value for kCOz of
with bubble formation. Laboratory experiments have 21 f5 cm h - can therefore be produced. Similarly,
shown that bubbles are extremely efficient in increas- by measuring increased 14C present in the oceans as a
ing k, for poorly soluble gases (see Figure 2). The result of atmospheric nuclear bomb tests, an inde-
initial injection of the bubble enhances gas transfer by pendent estimate of k c o n of 22 f3 cm h - can be
disrupting the microlayer; the air inside the bubble derived, in good agreement with the natural 14Cvalue.
then exchanges with the bulk sea water (air-water gas Although the 14C technique gives good estimates of
transfer that bypasses the air-sea interface) and the globally averaged k c o n , it yields little information on
Next Page
AIRSEA INTERACTIONI Gas Exchange 87
how k varies in time and space, or indeed how to processes within the water column. Additionally,
calculate k for other gases. Parameterizations of k , advection and dispersion at the sampling site are
have therefore been proposed by fitting quadratic and major problems. The use of deliberate tracer tech-
cubic curves such that, when averaged over the range niques may help to overcome this by enabling meas-
of global wind speeds, they are in agreement with the urements to be made within a Lagrangian framework.
global mean k , determined above (Figure 3). The
assumption is also made that the value for the Schmidt
Radon Radioactive decay of radium-226 (226Ra)to
number dependence ( a )in eqn [6] is 0.5. the gas radon-222 ("'Rn) occurs within the water
column and radon is therefore transferred from the
Oxygednitrogen ratios A more recent technique for surface mixed layer to the atmosphere. A mass budget
deriving regional estimates of the annual mean k , is can be made of the 'missing' radon by assuming steady
based on high-precision measurements of atmospheric state with deeper waters and a value for kRn can be
oxygednitrogen ( 0 2 / N 2 ) ratios from baseline sites derived. The mean value for kco2 obtained using this
situated around the globe. The technique depends on technique is about 1 4 c m h - 1 (corrected from kRn by
the use of CO2 data to correct for the effects of land/ assuming a = 0.5). The radon data show a large
atmosphere fluxes on 0 2 / N 2 ratios and on an atmos- amount of scatter with wind speed and the technique
pheric transport model to simulate oceanic fluxes. has shortcomings in that the condition of steady state
Annual values for kcOz of 24 f6 and 29 f1 2 cm h - is rarely fulfilled.
were calculated for ocean areas north of 30"N and
south of 30" S respectively. Although higher than
estimates derived from 14C, these values agree with Deliberate tracers Sulfur hexafluoride (SF6) is an
them within measurement uncertainties. inert, volatile tracer and has been deliberately
added to water bodies to determine k , Experiments
Local Measurements in enclosed lakes showed a good correlation between
k, and wind speed and a parameterization of k ,
Mass balance This technique involves time-series developed from these studies predicts a similar global
measurements of a gas (typically CO2 or 0 2 ) that is out mean ( 1 2 c m h - I ) to that obtained from the radon
of equilibrium with the atmosphere to try to obtain the data. Co-release of helium-3 (3He) has allowed a
flux of the gas across the air-sea interface by use of a value of 0.5 for the Schmidt number dependence
water-side mass budget. However, it is difficult to to be measured, in agreement with surface renewal
accurately quantify all the production and/or removal models.
90
80
70
60
20
10
0
0 2 4 6 8 10 12 14 16
Wind speed (m s-l)
Figure3 Three parameterizationsof kwwith wind speed are shown. The dashed line is a cubic fit to the global mean kwderived from 14C,
whereas the dotted line is a quadratic fit to the same 14Cvalue. The solid line is based on data from an SF6release in a lake and from wind
tunnel experiments. The dual tracer data ( 0 ) are from various experiments in coastal regions and the Atlantic and Pacific Oceans.
(Reproducedwith permissionfrom Nightingale PD, Liss PS, Schlosser P (2000) Geophysical Research Letters 27: 21 17-2120, copyright
(2000) by the American Geophysical Union.)
128 ANGULAR MOMENTUM OF THE ATMOSPHERE
D A Salstein, Atmospheric and Environmental by means of transient eddies. The transport of angular
Research, Inc., Lexington, MA, USA momentum is also accomplished vertically, carrying
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. angular momentum as part of the Hadley and other
mean meridional circulations.
The atmosphere, however, is far from a closed
Introduction
system in this respect, and a streamfunction analysis
Angular momentum is a property of mass in motion of zonal mean angular momentum has source
about a given axis, which in a closed domain is and sinks at the atmosphere’s lower boundary.
conserved. In the context of the atmosphere, angular Indeed, there are exports and imports of angular
momentum is a useful parameter for studying dynam- momentum across its lower interface by means of
ics on different temporal and spatial scales. When the torques. But the whole Earth, including its fluid
reference axis is identified with that of the Earth’s components, functions mostly as a closed system
figure, which we may call the principal axis, the with respect to the angular momentum budget (but for
resulting globally integrated axial angular momentum the influence of certain well-known tides, principally
value may be treated as a fundamental index of with the Moon).
atmospheric circulation. As such, this parameter If we consider atmospheric angular momentum
mirrors many aspects of the signature of climate and about the fundamental axis, the relative angular
weather. Furthermore, how angular momentum is momentum is largely dependent on the westerly
exchanged across its lower boundary, by means of the component of the wind, with the component related
interactive torques with the oceans and solid Earth to mass changes rotating with the Earth very small.
below, is important to quantify so that one can Because of the variations in this axial angular mo-
understand how the Earth acts as a system. Small but mentum quantity, the angular momentum in the other
measurable changes in the Earth’s rotation rate components of the Earth system must change in
are a consequence of the exchanges of angular compensation. Indeed, observations using space-
momentum between the solid Earth and its fluid geodetic techniques have demonstrated that the
envelope; this aspect of the variability is of importance Earth’s rotation rate changes perceptibly on many
to the study of Earth physics and to the monitoring of time scales. Such a change is most conveniently
reference frames for satellite orbits and navigation. expressed in terms of variations in the length of day,
The relevance of atmospheric angular momentum which are very nearly proportional to those in atmos-
changes to geodesy and geophysics has been recog- pheric angular momentum.
nized by the formal organization of the Special Bureau Besides the principal axis, angular momentum may
for the Atmosphere of the International Earth Rota- also be calculated about pairs of other axes in the plane
tion Service to supply such atmospheric data to perpendicular to the principal axis, in the equatorial
geoscientists. plane. In the components in the equatorial axes, the
The angular momentum of a parcel of air in the term related to the mass of the atmosphere dominates.
perpendicular plane about an axis is given as its mass These changes of atmospheric angular momentum
multiplied by the length of the radius arm to the lead to motions of the Earth’s pole about the mean
reference axis, multiplied by the component of the rotation axis - the wobble of the Earth. Such polar
velocity of the parcel in that plane, normal to the motions have also been measured by several space-
radius arm. The angular momentum of the global geodetic techniques.
atmosphere about such an axis is the sum of the Changes in the angular momentum of a body must be
angular momentum of all its air parcels, which may be produced by an imposed torque. In the case of the
calculated by integration over the volume of the atmosphere, such interactions occur across its lower
atmosphere. Because the atmosphere is a fluid, vari- interface, with the solid Earth and the ocean below.
ations in its angular momentum relate to changes in These torques are related principally to two mechanisms.
both motion terms (relative to the Earth), as well as to In one, winds at the surface transfer angular momentum
changes in its mass distribution. by tangential stresses across the surface, yielding a so-
As a conservative property, angular momentum in a called friction torque on the Earth. A second mechanism
closed system is constant in total but can be redistrib- comes as a result of the existence of surface pressure
uted within that system. For example, the atmosphere variability near areas of high topography. Such ‘moun-
transfers angular momentum northward principally tain’ torques result from the variability of the normal
ANGULAR MOMENTUM OF THE ATMOSPHERE 129
pressure gradient forces that push harder on one side of

the mountain than they do on other.
In the following, we will examine the distribution of
-5
C
- - - __ __ All
-_-_ DJF
months
angular momentum in the atmosphere, principally the

axial component, and then we discuss its variability on a
number of time scales. We also expand upon the
relationship to the corresponding motions of the Earth.
Axial Angular Momentum in Regions

The dominant relative angular momentum about
Earth's axis depends on the strength of zonal winds,
which tend to be persistent features of the atmospheric Figure 2 Long-term mean angular momentum distribution in a
set of 46 equal area belts spanning a 30-year time period. Shown
circulation. In Figure 1we present the latitude-pressure
are values for all months, for December/January/February (DJF)
distribution of the long-term zonal mean zonal wind and June/July/August (JJA) months.
based on the so-called reanalysis dataset from the US
National Centers for Environmental Prediction and
National Center for Atmospheric Research (NCEP-
NCAR); we used the most recent 30 years to form a difference between the December/January/February
climatology of the mean wind in the figure. A similar periods and the June/July/August periods. It is clear
signature exists in the two hemispheres. Mean easterly that middle latitude belts have their largest values
winds are found in the tropical regions, with a broader during their winter, in both the Southern and Northern
latitudinal extent at the surface than higher in the Hemispheres, but the annual cycle is larger in the
atmosphere. The winds are westerly over most of the northern than in the southern hemispheric belts. Most
extratropics. Increases in the strength of these winds of the atmosphere has westerly relative angular
with height in the atmosphere lead to very strong momentum, indicating that in these regions the
westerlies in the upper troposphere, at levels near atmosphere superrotates with respect to the underly-
200hPa; above this jet level, they tend to decrease ing planet.
again. Regional maps of such jets at this level reveal that
the strongest values are located over the eastern North
American and Asian continents. The strong winds at Global Atmospheric Angular
these regions contribute heavily to the relative compo- Momentum
nent of the axial angular momentum of the atmosphere.
Angular momentum can be computed in zonal belts From series of the four-times daily zonal winds given,
from values of the zonal wind, so that a profile of based on the NCEP-NCAR reanalyses, global values
angular momentum (Figure 2) reveals the general of the relative atmospheric angular momentum are
distribution with latitude. The seasonality of the calculated by integration over the volume of the
angular momentum can be noted as the substantial atmosphere. In the resulting series of atmospheric
angular momentum values, shown in Figure 3, a host
of interesting signals on a number of important time
scales emerges. For example, it is clear that a strong
annual cycle exists, whose phase yields a peak around
January, during the period of the strongest Northern
Hemisphere jets in the boreal winter. The angular
momentum signal in each hemisphere peaks in its
respective winter months, but the annual signal in the
Northern Hemisphere has a stronger amplitude than
that of the Southern Hemisphere, due to the greater
continentality of the Northern Hemisphere. As
a result, the phase of the global signal is that of
the Northern Hemisphere, though the amplitude is
reduced.
Figure 1 Latitude-pressure cross-section of zonal mean zonal
winds, from which relative atmospheric angular momentum is In the figure is also evidence of a superimposed
derived. Based on 30 years (1970-1999) of the NCEP-NCAR semiannual signal, which can be noted as a combina-
analysis system. Units are m s - ' . tion of a dip during the middle of the northern winter,
130 ANGULAR MOMENTUMOF THE ATMOSPHERE
5 ~ 1 1 1 1 1 1 1 1 1 1, 1 1 1 1 1 1 1 , 1 1 1 , 1 1 1 1 1 1
70 72 74 76 78 80 82 84 86 88 90 92 94 96 98
Year
Figure 3 Series, since 1970, of globally integrated atmospheric angular momentum between the 1000 and 10 hPa levels, based on the
NCEP-NC A R reanalyses.
and a sharp plunge in the middle of the southern different latitudes that peak six months out of phase.
winter. This semiannual signal arises largely from the These varying patterns can be noted in the time-
corresponding wind signal in the stratosphere. Such an latitude diagram of angular momentum in the strato-
overall signature is derived from annual patterns at sphere in Figure 4.
Figure 4 Time-latitude diagram of angular momentum of the stratosphere in 46 equal area belts, between the 100 and 10 hPa pressure
levels. The resolution of data here is monthly. Units are loz4kg rn‘s-’.
On interannual time scales, we find two prominent between 100 and 10hPa in Figure 4 in the strato-
signals in the global signature in Figure 3, one on scales spheric pressure levels in the belts surrounding the
slightly longer than 2 years, and a second on time equator. Because angular momentum is calculated
scales closer to 4 years. The shorter of the two relates with weights related to the distance to the rotation
to the so-called ‘quasi-biennial oscillation’, a result of axis, the contributions from the zonal winds at the
the reversal of the zonal winds in the tropical strato- lowest latitudes, farthest from the rotation axis, are
sphere. The second of the two has a signature that most important here. The vertical curve on the right-
attains a maximum around the time of peaks in the hand side of Figure 4, the sum over all the belt values in
El Niiio Southern Oscillation (ENSO) over the trop- the stratosphere, clearly reflects an alternation every
ical Pacific Ocean. other year in the global time series of angular
momentum in the stratosphere.
Quasi-Biennial Oscillation in
Atmospheric Angular Momentum and ENSO Influence on Angular Momentum
Stratospheric Winds The influence of the El Nifio Southern Oscillation
The distribution of winds in the stratosphere is such produces a clear signature in the evolution of angular
that westerly winds predominate in middle latitudes momentum. The origin of the strong peaks in the
and easterly winds are found in the tropics. In alternate global relative atmospheric angular momentum may be
years, approximately, however, the tropical easterly noted in the time-latitude diagram, such as Figure 5,
winds tend to diminish substantially or even reverse which, to emphasize the ENSO time scales, filters out
their direction to become westerly. Such an alternating signals longer than 4 years and is confined to the region
signal, though noted first over the western Pacific, has below 100hPa. During periods of El Nifio, the
been observed at other longitudes, and it is very well tropospheric zonal winds have westerly anomalies,
captured by a zonal average. It can be observed equivalent to weakened easterlies or to westerlies in
Figure 5 Time-latitude diagram of atmospheric angular momentum, based on monthly mean anomalies from the average of the
calendar month, bandpass filtered to emphasize the time scale associated with the ENSO signal. Units are ~ ’.
kg m2s
132 ANGULAR MOMENTUM OF THE ATMOSPHERE
part of the tropics initially, and then anomalously weather events. Thus, important changes in global
strong westerlies more poleward on the order of six angular momentum have been tied to individual weath-
months to one year later. During the peak of the er patterns across the Rocky, Andes, and Himalayan
westerly anomaly period, especially, the globally Mountains. Indeed, a considerable percentage of the
integrated atmospheric angular momentum is notably rapid fluctuations in the Northern Hemisphere winter
strong. Two such strong values recently were during can be tied simply to surface pressure differences
the 1982-83, and the 1997-98 El Nifio events. During between two stations on the opposite side of the Rockies.
these episodes, the global signal in relative atmos- At lower frequencies, the mountain and friction
pheric angular momentum was exceptionally high. torques have approximately the same amount of
However, the record value, in January 1983 came power. However, the Madden-Julian oscillations
about as a result of the superposition of the El Nifio may be dominated by the friction torques over the
signal with that of the normally strong seasonal signal Pacific Ocean. Determining the mechanism for the
during northern winter. However, the 1997-98 had an seasonal and interannual angular momentum varia-
exceptionally extended period in which there were tions, such as in the generation of El NiAo conditions in
strong positive atmospheric angular momentum atmospheric angular momentum is somewhat more
anomalies. difficult; such lower frequency variations likely result
With the cooling of the waters in the Pacific, the La in a combination of effects.
Niiia ushers in a different circulation from that of the Other mechanisms for exchange of angular mo-
El Nifio, and anomalously easterly winds create a mentum have been theorized. That due to gravity wave
negative anomaly in atmospheric angular momentum. drag, which exchanges momentum in internal waves,
The transition can be quite abrupt, as occurred during typically over uneven topography, is similar to the
May 1998, a month that featured a reversal of the sign friction mechanism, but on larger spatial scales. Lastly,
of angular momentum anomaly across a very wide gravitational torque, involving the attraction of the
meridional band from the middle latitude of the planet with the varying atmospheric mass, is a
Southern Hemisphere to the middle latitudes of the relatively small contributor for the axial component
Northern Hemisphere. of angular momentum.
Torques Across Atmosphere’s Concurrent Changes in Atmospheric

Lower Boundary Angular Momentum and Length of Day
The angular momentum of the atmosphere may Because of the exchange of angular momentum
fluctuate quite rapidly, and so it is apparent that between Earth and atmosphere, Earth’s rotation rate
means must exist to accomplish this change at the fluctuates in very close connection to the changes in the
atmosphere’s lower boundary. Two principal torque global atmospheric angular momentum. This conser-
mechanisms to effect the angular momentum transfer vation would imply a strict proportion between
have been identified. In one, the atmosphere sets up a variations in atmospheric angular momentum and
pressure gradient force on opposite sides of moun- those in length of day (1.o.d.). The relationship
tainous topography, and when considered at a dis- between l.o.d., determined using observations by
tance from the axis, the normal force near topography means of space-geodetic systems, like very-long base-
creates a so-called mountain torque on the atmosphere line interferometry, and satellite-laser ranging, and
and Earth. Thus, a relatively low pressure on the west those of atmospheric angular momentum, obtained
of a mountain range and relatively high pressure on the from the NCEP-NCAR reanalyses, integrated
east will tend to decelerate the Earth and thus through 99% of the atmosphere, to 10 hPa, is shown
accelerate the atmosphere. A second major torque over a 2-year period in Figure 6 . The very good
results from the tangential forces of the winds against agreement between the two series is remarkable
the ocean or land below. This force on the atmosphere because of the extremely different data types from
will be counter to the direction of the zonal winds; thus which they are derived. The seasonal and intrasea-
westerlies will tend to diminish by the action of sonal fluctuations occur quite closely in both series.
friction, and the Earth will gain the angular momen- Indeed, coherence between the two is very strong on
tum transferred from the atmosphere. scales down to 7 days, and has some significance on
The time scales on which the mountain and friction scales as short as daily. Interestingly, the intrasea-
torques operate are quite different. Mountain torques sonal fluctuations in 1.o.d. were discovered indepen-
have primary responsibility for the atmospheric angular dently from those of the atmospheric Madden-Julian
momentum fluctuations on the synoptic scales of oscillation, related to fluctuations across the Pacific
5 coupled atmosphere-ocean model would help resolve

6 % the changes that would be needed for the prediction of
4 $7 angular momentum trends.
E b Also, because of the close relationship on time scales
-%NE from days to a few years, between 1.o.d and atmospheric
0 $2 angular momentum, earlier records of 1.o.d. may be used
a m
L "0
-2 .o as a proxy for the global variations in the atmosphere. A
"
-42
n
z record of 1.o.d. back to the dawn of the telescope era in
-1 .o]
I"
-6 I: the seventeenth century has been examined, though it is
E of insufficient accuracy for atmospheric purposes until
'LSblNd
'J ' + M i M J JILSGND'J ' F M i M J ' J the end of the nineteenth or the beginning of the
1998 1999
twentieth century. Signals relating to changes in varia-
Figure6 Global integral of atmospheric angular momentum and bility of the atmosphere during certain decades (like the
departures of the length of day, for a recent 2-year period. Mean 1920s and the 1940s, which had high and low interan-
terms have been removed. nual variability, respectively, in l.o.d.), and dominant
interannual time scales (3.4 and 2.1 years) have been
determined from such a proxy record.
Ocean and observed in both the tropics and extra-
tropics. Differences in angular momentum between
the Earth and atmosphere point to either errors Atmospheric Angular Momentum in
in the data sets, or to the role in the exchange of
angular momentum of a third component, such
the Equatorial Plane and Polar Motion
as the oceans. Besides its rotation about the principal axis discussed
for most of this article, the other two components of
the atmospheric angular momentum vector, namely
Models and Historical Series of those in the equatorial plane, can be determined.
Though not of clear fundamental interest to atmos-
Atmospheric Angular Momentum pheric studies, this component of angular momentum
The atmosphere has been simulated by a large number is related importantly to certain motions of the Earth
of models that are driven solely by the temperature of known as Earth wobble, or polar motion. Related
the underlying ocean. Based on these models, atmos- fluctuations of angular momentum in these compo-
pheric angular momentum has been calculated and nents are stronger in the so-called matter (surface
used, moreover, as a parameter for model validation to pressure) term than in the motion (wind) term. Thus,
determine the success of model simulations. Aside pressure variability over certain regions like the
from observations of angular momentum, independ- northern Pacific and Atlantic (Aleutian and Icelandic
ent measurements of 1.o.d. have been used to examine lows, respectively), the southern oceans, and over
the results of models. Lengthy runs of models are Eurasia have been determined to be important
possible because sea surface temperatures are avail- to fluctuations of equatorial angular momentum
able for most of the twentieth century; these models (Figure 7).
are unlike atmospheric analyses, whose dependence When atmospheric pressure fluctuations over the
on upper air winds, are confined to the second half of oceans are observed carefully, it can be noted that
the century. Such runs indicate that an increase in such those on time scales of several days and longer
values over the last half century appear to have influence the distribution of the ocean mass below.
occurred, possibly related to the relative increase in This effect, the so-called inverted barometer, acts so
El Nifio activity. Increases in the short-term variability that a high atmospheric pressure will depress the
of atmospheric angular momentum moreover appear surface below, moving ocean mass away from that
to have taken place. region; the opposite action occurs with a relatively low
Models can be run in a prognostic mode as well, to atmospheric pressure. Such an inverted barometer
determine, for example, the effect of an increase in relationship has the effect of dramatically reducing the
greenhouse gases on the angular momentum of the mass component of the effective angular momentum
Earth. Such effects may include changes in the annual signal of the atmosphere over the oceans.
signature, and a possible decrease in angular momen- For the continental regions remaining, the mass
tum could be related to the warming of the higher fluctuations over Eurasia, predominantly, and North
latitudes, which could induce a reduction in temper- America, secondarily, appear to be the biggest regional
ature gradient and the strength of zonal winds. Use of a atmospheric influentes exciting polar motions on
134 ANGULAR MOMENTUMOF THE ATMOSPHERE
Figure7 Variability of atmosphericexcitation of polar motion, in nondimensional units x l 0-l6based upon surface pressureanalysesfor
a recent 30-year period.
subseasonal a n d other time scales which have been Grotjahn R (1993j Global Atmospheric Circulations, Obser-
monitored using space-geodetic methods. On longer vations and Theories. New York: Oxford University Press.
time scales, the ocean plays a role of similar impor- Hide R and Dickey JO (1991) Earth’s variable rotation.
tance to that of the atmosphere. The more dominant Science 2.53: 629-637.
such ones for polar motion are the annual term and Lambeck K (1980) The Earth’s Variable Rotation: Geo-
that of the natural modal response of the Earth’s polar physical Causes and Consequences. Cambridge: Cam-
bridge University Press.
motion, with period near 430 days, known as the
Lorenz EN (1967) The Nature and Theory of the General
Chandler wobble. At the other end of the frequency Circulation of the Atmosphere. Geneva: World Meteor-
spectrum, a motion of the Earth on near daily time ological Organization.
scales, known as nutation, is also driven partially by Newton CW (1971j Global angular momentum balance:
atmospheric angular momentum forcing. Earth torques and atmospheric fluxes. Journal of the
Atmospheric Sciences 28: 1329-1341.
Oort AH (1989) Earth’s angular momentum cycle in
the atmosphere-earth-solid earth system. Bulletin of
See also the American Meteorological Society 70: 1231-1242.
Peixoto JP and Oort AH (1992) Physics of Climate, ch. 11.
Air-Sea Interaction: Momentum, Heat and Vapor Flux- New York: American Institute of Physics.
es. ClimateVariability:Decadalto CentennialVariability; Rosen RD (1993)The axial momentum balance of Earth and
Seasonal to lnterannual Variability. El Niiio and the its fluid envelope. Surveys in Geophysics 14: 1-29.
Southern Oscillation: Observation. General Circula- Salstein DA, Kolaczek B and Gambis D (edsj (1999) The
tion: Momentum Budget. Land-Atmosphere Inter- impact of El Nifio and other low-frequency signals on Earth
actions: Overview. Middle Atmosphere: Quasi- rotation and global Earth system parameters, IERS Tech-
Biennial Oscillation; Semiannual Oscillation. Mountain nical Note 26. Paris, France: Central Bureau of the IERS.
Meteorology. Salstein DA, Kann DM, Miller AJ and Rosen RD (1993)The
Sub-bureau for atmospheric angular momentum of the
International Earth Rotation Service: A meteorological
data center with geodetic applications. Bulletin of the
Further Reading American Meteorological Society 74: 67-8 1.
Barnes RTH, Hide R, White AA and Wilson CA (1983) Weickmann KM, Kiladis GN and Sardeshmukh PD (1997)
Atmospheric angular momentum fluctuations, length-of- The dynamics of intraseasonal atmospheric angular
day changes and polar motion. Proceedings of the Royal momentum oscillations. Journal of the Atmospheric
Society of London A 387: 31-73. Sciences 54: 1445-1461.
ANTARCTIC CLIMATE 135
J Turner, British Antarctic Survey, Cambridge, UK conditions across the Southern Hemisphere. This
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. article presents a brief overview of the climate of
Antarctica, which is taken to be the area south of 60"S
(a map of the Antarctic indicating topographic
Introduction features and the locations of many of the research
Antarctica is the coldest, windiest, and driest conti- stations is shown in Figure 1).Statistics on mean and
nent on Earth, with a remote location far from the extreme atmospheric conditions are provided, along
major centers of population. Yet as one of the two heat with selected mean meteorological fields. Particular
sinks in the global climate system it plays a crucial role attention is paid to the factors that maintain the
in the general circulation of the atmosphere and has a climate of the Antarctic and the interactions with
profound effect on the atmospheric and oceanic lower latitudes.
90" E
Figure 1 A map of the Antarctic showing topographic features and the locations of a selection of research stations.
136 ANTARCTIC CLIMATE
The Antarctic Continent from numerical weather prediction systems have

shown that the Antarctic coastal region has the highest
The climate of the Antarctic is influenced heavily by
incidence of cyclogenesis anywhere in the Southern
the orography of the continent, the highest part of
Hemisphere. The zone of 60-70" S is therefore char-
which is located close to but slightly offset from the
acterized by many depressions, both declining and
South Pole. The Antarctic represents about 10% of the
active, which results in a belt of low surface pressure
land surface of the Earth and elevations increase very
called the circumpolar trough. This can be seen in the
rapidly inland of the coast, with much of the continent
maps of mean sea-level pressure (MSLP) for the four
being above 2 k m in height and small areas above
seasons shown in Figure 2. The circumpolar trough is
4 km. The high interior is therefore isolated from the
present throughout the year, and in the mean fields has
warm air masses of midlatitudes and is characterized
an approximate wavenumber 3 pattern with low-
by very cold, dry, relatively cloud-free conditions
pressure centers close to 30" E, 90" E, and 150"W. This
(Table 1).
pattern affects a number of aspects of the Antarctic
The other landmasses of the Southern Hemisphere
climate, such as the northward extension of sea ice
are well north of the Antarctic, so that the oceanic and
close to the Greenwich meridian, as a result of the
atmospheric flow is much more zonal than in the
climatological southerly flow at this longitude and the
Northern Hemisphere. This further isolates the Ant-
relatively mild temperatures experienced on the west-
arctic from the influences of lower latitudes and is one
ern side of the Antarctic Peninsula because of the
of the reasons the Antarctic is significantly colder than
predominance of north to north-westerly airstreams
the Arctic.
affecting the area. From Figure 2 it can be seen that
MSLP values within the circumpolar trough are lowest
during the spring and autumn, and higher during the
Weather Systems in the Antarctic
summer and winter. This semiannual oscillation can
The large difference in temperature between the be detected in the MSLP observations from the coastal
Antarctic and the tropics is responsible for the many stations and also in the number of reports of precip-
active depressions found over the Southern Ocean. itation. The oscillation is a result of changes in the
These carry warm (cold) air southwards (northwards) position and intensity of the circumpolar trough over
on their eastern (western) flanks and attempt to the year, with it being furthest south (north) and most
remove the radiatively induced pole to tropics tem- pronounced (weak) in the intermediate (summer and
perature difference. Because of the distribution of the winter) seasons. The cycle is present because of the
landmasses in the Southern Hemisphere the atmos- phase difference between the seasonal cycle of surface
pheric long waves have a smaller amplitude than their pressure values over the contiqent and the sub-
counterparts in the north, so that the depressions play Antarctic latitudes.
a greater role in the poleward transport of heat than in One aspect of the climate of the Antarctic that was
the Northern Hemisphere. apparent only once satellite imagery became generally
The depressions over the Southern Ocean generally available is the degree to which mesoscale low-
move from west to east, with those that developed at pressure systems (mesocyclones) are a feature of the
more northerly latitudes spiraling in towards the Southern Ocean and Antarctic coastal regions. These
Antarctic coastal region. The meeting of cold Antarc- systems have a horizontal length scale of less than
tic and midlatitude air masses in the Antarctic coastal about 1000 km and a lifetime of less than about 1 day
region results in a moderate to strong horizontal and are therefore difficult to analyse and forecast.
temperature gradient (baroclinicity), so that this area However, they have a major impact on the weather
is also one of frequent cyclogenesis (development of experienced at coastal sites and so are important in the
depressions). In fact, recent studies using the analyses forecasting processes. Although mesocyclones are rare
Table I Mean temperature data for selected Antarctic stations

Station Latitude Longitude Nevation Period Mean annual Mean January Mean July
(m) temperature temperature temperature
Vostok 78.5"s 106.9"E 3488 1958-2001 - 55.4% - 32.2"C -67.1%

South Pole 90.0's 2800 1958-2001 - 49.5"C - 28.2"C - 59.7"C
Mawson 67.6" S 62.9" E 16 1954-2001 - 11.2"C +O.l"C - 18.0"C
Faraday Vernadsky 65.4" S 64.4" W 11 1951-97 - 3.9% +0.7"C - 9.2"C
Bellingshausen 62.2" S 58.9" W 16 1968-2001 - 2.4"C + 1.5% - 6.6'C
0"
0081 008 1
0" 0"
Figure 2 Average mean sea-level pressure fields (hPa) for the four seasons.
over the high interior, they are a common feature on imagery does show that some frontal bands associated
the ice shelves. Here there is low-level convergence of with lows in the circumpolar trough can be seen on the
air that has descended from the plateau, which aids the plateau, although automatic weather station data
spin-up of vortices, coupled with the presence of mild, suggest that the pressure signals across these features
oceanic air masses that provide moisture for the are small. The conditions that favor depressions
formation of cloud. However, during the winter having an impact in the interior are amplified long
season the automatic weather stations on the Ross waves and strong steering flow aloft. Under such
Ice Shelf have indicated that low-level mesoscale conditions mild air masses over the plateau can give
vortices can be present that do not have any cloud relatively large falls of precipitation, resulting in a
associated with them, but which have a clear surface significant fraction of the year's accumulation falling
circulation. in a day or two. When the longwaves are strongly
Because of the rapid increase in elevation inland of amplified, maritime air masses can affect the South
the coast, few major weather systems penetrate far Pole and even Vostok Station on the high plateau of
into the interior of the continent. However, satellite East Antarctica, but such conditions are rather rare.
The Role of Sea Ice the high plateau is that temperatures increase with
height over the lowest few hundred meters of the
Another major influence on the climate of the Antarc-
atmosphere. This temperature inversion is a result of
tic is the seasonally varying belt of sea ice that rings the
the intense radiative cooling of the surface and lower
continent. Unlike in the Arctic, most of the sea ice
atmosphere and is therefore strongest during the
melts during the summer so that by February there is
winter season, although in other seasons it is still
only about 3.5x106km2 of ice, most of which is
significant. The mean strength of the winter inversion,
located over the western Weddell Sea. From February
i.e., the temperature difference between the surface
and throughout the winter and early spring the sea ice
and the maximum temperature in the lower tropo-
advances in a divergent fashion around the whole
sphere, varies from about 5°C in the coastal region to
continent, reaching a maximum in September, when
more than 25°C over the highest parts of East
the ice covers around 19x106km'.
Antarctica.
The Antarctic sea ice is generally about 1m thick
Across the Antarctic there is a very large range of
and provides an effective cap on the upper layers of the annual surface air temperatures, although it is only in
ocean, limiting the fluxes of heat and moisture into the
the northernmost part of the Antarctic Peninsula that
lower layers of the atmosphere. However, the effects of
mean summer temperatures rise above freezing. Over
the many weather systems over the Southern Ocean on
the Antarctic Peninsula and along the coast of East
the sea ice is to open up linear cracks (leads) or larger
Antarctica the annual cycle of temperatures is similar
areas of open water (polynyas), which can provide
to those found in midlatitudes with a broad summer
local sources of heat or moisture, resulting in cloud.
maximum and a minimum in July o t August. How-
This can be important for the climate at the coastal
ever, at more southerly latitudes the cycle is different,
stations during the winter months when the opening
with a sharp summer maximum and a 'coreless'
up of coastal leads and polynyas can significantly
winter, during which temperatures vary by only a
increase the temperature and humidity, sometimes
small amount. This form of the annual cycle comes
leading to fog formation.
about for a number of reasons, including the abrupt
The other major effect of sea ice is to increase the
change in solar radiation at the start and end of the
surface albedo. This is not a significant factor during
period of Austral winter darkness, the effects of the
the period of winter darkness, but can be important in
semiannual oscillation of the annual cycle of advection
the spring when the sun returns to the high latitude
of warm air into the Antarctic, and the heat reservoir
areas slowing the surface heating.
effect of the Antarctic snow pack.
At higher levels in the troposphere the atmosphere is
strongly stratified, much more so than in the midlat-
The Temperature Field itude areas of the Southern Hemisphere. This is the
Because of the low solar elevations in summer and the case in all seasons, with the stability being strongest
long period of winter darkness, the bulk of the below about 4 km during the winter. In radiosonde
Antarctic receives little incoming solar radiation and ascents a tropopause in usually evident in the summer,
is always very cold. In addition, the snow-covered but it can become very indistinct during the winter
surface reflects much of the incoming solar radiation when the stratosphere cools rapidly.
back to space, so giving rise to a positive feedback that
helps maintain the frigid conditions.
The high plateau of East Antarctica experiences the The Wind Field
coldest temperatures on Earth and at Vostok Station
(78.5" S 106.9"E, 3488 m elevation) they have The strong, persistent winds recorded at a number of
an annual mean temperature of - 55.4"C and have sites around the Antarctic are one of the most
recorded the lowest temperature on Earth of remarkable features of the continent's climate. It is
- 89.6"C, measured on 2 1 July 1983. The very cold now known that these winds are katabatic in origin
temperatures are experienced not just because of the and occur because of the drainage of cold, dense air at
lack of midlatitude air masses reaching the area, but low levels from the interior plateau down towards the
because of the total lack of solar heating during the coast. The katabatic winds are most pronounced
winter months, the high elevation of the Antarctic during winter, when there is no incoming solar
plateau and the very limited amounts of cloud and heating, and a large pool of cold air over the interior
water vapor in the atmosphere, which allow much of is formed to feed the katabatic flow.
the emitted long-wave radiation from the surface to be Surface winds over the interior show a high direc-
lost to space. Although the near-surface layers on the tional constancy, indicating that they are dictated by
plateau are very cold, a characteristic of the climate of the local orography through katabatic forcing rather
than by the highly variable synoptic-scale weather These suggest that the greatest cloud cover is found
systems. The wind speeds are closely related to the over the ocean area north of the edge of the continent,
slope of the orography, with the strongest winds being with about 85% cloud cover throughout the year near
measured at stations on the coastal escarpment and the 60"s. In the coastal region near 70"S, the surface
weakest on the flattest areas of the plateau. Along the observations indicate that the total cloud cover is
coast of Adtlie Land the orography channels the about 45-50%, with little seasonal variability and
katabatic flow onto a small stretch of coast, resulting only a small decrease during the winter months. Inland
in very strong and persistent winds with a very high of the coast, the amounts of thick cloud decrease
directional constancy. It was in this area that Maw- rapidly, since few synoptic-scale weather systems are
son's 1912-13 expedition recorded the world record found here. However, the interior is characterized by
annual mean wind speed of 19.4ms-1 and experi- extensive, very thin cirrus cloud which gives a semi-
enced gale-force winds on all but one of 203 consec- permanent veil of ice crystals. This type of cloud causes
utive winter days. problems for observers, who have to decide whether to
As the katabatic winds descend from the plateau report no cloud or 10/10 cloud cover. The mean annual
they turn to the left because of the Coriolis force cloud cover at South Pole station is 45%, but anyone
and merge with the coastal easterlies that are present using such statistics has to be aware of the nature of the
because of the circumpolar trough north of the cloud that occurs there and the problem facing
coast. The near-surface flow therefore appears as an observers of how to report the thin cirrus.
anticyclonic vortex, with cold air outflow from the The amount of precipitation across the Antarctic
continent. In some parts of the coastal region, such as generally follows the distribution of thick cloud. In
south of the Weddell Sea, the coastal easterly other words, the greatest precipitation totals are found
comes up against high orography and the cold, in the coastal region, with a rapid decrease inland.
stably stratified air at low levels does not have the Figure 3 shows the mean annual precipitation across
kinetic energy to cross the barrier. The air is then the continent as estimated from ice cores. These
dammed up against the barrier until a pressure glaciological measurements of accumulation are very
gradient develops that results in the air moving north similar to precipitation, since there is little evaporation
as a 'barrier wind'. With the strong static stability in the interior. However, they are not identical,
encountered at low levels in the Antarctic, barrier because of the effects of blowing snow and summer
winds are relatively common in the coastal areas of melt in some areas. But with so few in situ measure-
the continent. ments of precipitation they have been used extensively
as a proxy for precipitation. In Figure 3 it will be noted
that no data are presented for the northern part of the
Antarctic Peninsula, since precipitation varies so
Clouds and Precipitation rapidly in this area. It can be seen that the area of
Clouds are very important in the climate system as greatest precipitation is along the coast of the southern
they can reflect a significant proportion of the incom- Bellingshausen Sea, where there is over 1m water
ing solar radiation back to space. However, since the equivalent per year. This peak is found because of the
surface of the Antarctic already has a high albedo by high frequency of northerly airstreams bringing mild,
virtue of the year-round snow cover, clouds over the moist air onto the coast. Other areas of high precip-
continent tend to have less of an effect on the incoming itation are found where there is frequent cyclonic
solar radiation because the surface and cloud have activity, such as north of Enderby Land and along the
similar albedos. Nevertheless, clouds play a very coast of East Antarctica. The lowest precipitation in
important role in controlling surface temperatures the coastal region is found on the low-lying Ross and
through their effect on the long-wave radiation budget. Ronne Ice Shelves. Inland of the coast the amounts of
In cloud-free conditions, the dry atmosphere allows precipitation drop very rapidly, so that over the vast
most of the emitted terrestrial radiation to escape to majority of East Antarctic there is less that 50 mm of
space, resulting in very low temperatures. However, precipitation a year.
when thick cloud cover is present surface temperatures A number of estimates have been made of the mean
are much higher because of the downward long-wave and total snow accumulation across the whole of the
radiation emitted from the cloud. Antarctic ice sheet using glaciological data gathered
Since most of the research stations are located in the in situ. These estimates have improved as additional
coastal region, it is difficult to get an accurate picture surveys have been carried out, and the latest studies
of the distribution of cloud across the continent. suggest a figure of around l'60mm water equivalent
However, using in situ data and satellite imagery, per year. This equates to a total input of approximately
climatologies of cloud cover have been prepared. 2205 Gt year -
Figure 3 Precipitation over the Antarctic estimated from ice core data. Lines are accumulation isopleths in 100 kgm-*year-’ (or,
equivalently, 100 mm year-’). (Reproduced with permission from Bromwich DH (1988) Snowfall in high southern latitudes. Reviews of
Geophysics 26: 152. 0American Geophysical Union.)
The nature of the mechanisms behind precipitation ing of climate variability and change is limited in the
is different across the Antarctic, with most precipita- Antarctic because of the shortness of the records and
tion in the coastal area coming from synoptic-scale the fact that most research stations are on the coast,
weather systems, while in the interior most falls in the with only Vostok and South Pole stations having long
form of clear-sky precipitation, or ‘diamond dust’ as it records from the interior.
is sometimes known. This is an almost continuous The time series of annual mean surface air temper-
fallout of ice crystals from a thin veil of cirrus covering atures at a number of stations are shown in Figure 4.
the sky. Clear-sky precipitation has not been investi- These stations are located in different climatic re-
gated to any great degree, but is thought to result from gimes, at the South Pole (Amundsen-Scott), on the
the cooling of air over the plateau and the formation of high interior plateau (Vostok),on the coast (Mirny and
ice crystals as it descends into the cold near-surface Halley) and at Orcadas in the South Orkney Islands. It
layer. Just inland of the coast there is a zone where both can be seen that all the stations show a high degree of
synoptic-scale weather systems and clear-sky precip- interannual variability, but that this is largest on the
itation both play a role. Over Dronning Maud Land western side of the Antarctic Peninsula at Faraday/
studies have shown that clear sky precipitation falls on Vernadsky Station. This station is located close to the
most days, but that a few major weather systems can northern limit of the sea ice in the Bellingshausen Sea,
give a significant fraction of the year’s accumulation in and small variations in the ice extent are amplified into
a few days. much larger surface temperature variations, depend-
ing on whether the ocean west of the station is ice-
covered or ice-free in a particular winter. Most of the
Climate Variability and Change stations show a small warming trend, the exception
Both the Arctic and Antarctic exhibit a greater degree being Amundsen-Scott Station, where there has been a
of interannual and interdecadal climate variability slight cooling since the late 1950s. The warming on the
than locations at lower latitudes. This is thought to be western side of the Antarctic Peninsula is larger than
a result of the complex interactions between the elsewhere in the Antarctic, and even though the record
atmospheric circulation and the cryosphere, including is not long by the standards of stations outside the
a number of positive-feedback mechanisms that am- Antarctic, the warming trend is statistically significant
plify the climate variability. However, our understand- at the 99% level. Antarctic-wide temperature trends
-47 I -53,
E -48 -
Amundsen-Scott
v
Vostok
f
+
-49 - g
c
-55
$ -50 - a,
a
-56 -
2 -51 - E -57 -
c
5
-2- ' -52
1955 1965 1975 1985 1995 1955 1965 1975 1985 1995
-16, -9 I
1955 1965 1975 1985 1995 '1955 1965 1975 1985 1995
-18 -26
h
E -19 E -27
f -20 f -28 .._.._..-*--
*
I
9
a -21
c
5 -29
F -22 E -30
F
-23 -31
1955 1965 1975 1985 1995 1955 1965 1975 1985 1995
0 0
-a -2
h
y
v
-2
f
c
-4
I
f -4
2
a, -6 $a -6
E
F
-8
j -8
-10 -10
1945 1955 1965 1975 1985 1995 1900 1920 1940 1960 1980
Figure 4 Time series of annual mean temperature at a number of Antarctic coastal stations.
have been considered by various investigations within ated with the increased anthropogenic emission of
which seasonal and annual average temperatures for greenhouse gases.
Antarctica have been calculated by computing areally General circulation models are being used to
weighted means of all available station data. examine how the climate of the Earth will change
The annual mean Antarctic temperature showed a over the coming decades, and it is important to
warming trend of 0.029"C per year for the period consider how the climate of the Antarctic may change.
1957-82, which was significant at the 95% level. Although the various models all differ to some degree,
However, the greatest contribution from this temper- most models predict relatively modest temperature
ature rise came from the stations on the western side of rises around Antarctica over the next 50 years. As
the Antarctic Peninsula. temperatures rise so it is expected that there will be
In parallel with the warming trend there has increased snowfall over the continent, especially in the
also been a statistically significant increase in the coastal region, which should more than compensate
number of precipitation reports at the stations on for increased melting of Antarctic ice, and will thus
the western side of the Peninsula. We do not know partially offset the rise in sea level resulting from
yet whether the climatic changes observed in this area thermal expansion of the oceans and melting of
are the result of local factors or due to broader- icecaps and glaciers elsewhere in the world. However,
scale circulation changes across the Pacific region. many processes occurring in the polar regions are not
However, there is no evidence that the warming is well represented in climate models at present, and
taking place because of any 'global warming' associ- further research is needed to improve our confidence in
142 ANTICYCLONES
these predictions. This is particularly true for predictions Jones PD and Limbert DW (1987)A Data Bank ofAntarctic
beyond 50 years, when Antarctica may start to warm Surface Temperature and Pressure Data. Washington,
enough to have a significant impact on the ice sheets. DC: US Department of Energy.
King JC and Turner J (1997) Antarctic Meteorology and
Further Reading Climatology. Cambridge: Cambridge University Press.
Schwerdtfeger W (1970) The climate of the Antarctic. In:
Bromwich DH (1988) Snowfall in high southern latitudes. Orvig S (ed.) World Survey of Climatology, vol. 14,
Reviews of Geophysics 26: 149-168. Climates of the Polar Regions, 253-355. New York:
Carleton AM (1992)Synoptic interactions between Antarc- Elsevier.
tica and lower latitudes. Australian Meteorological Schwerdtfeger W (1984) Weather and Climate of the
Magazine 40: 129-141. Antarctic. Amsterdam: Elsevier.
S J Colucci, Cornell University, Ithaca, NY, USA temperatures. The convectively stable air of anticy-
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. clones may allow air pollutants to concentrate near the
Earth’s surface. Finally, the blocking action of anti-
cyclones aloft may cause persistently anomalous
Introduction
weather conditions at the Earth’s surface.
Anticyclones are regions of relatively high pressure on
horizontal surfaces, or high geopotential height on
isobaric surfaces, around which air circulates clock-
wise in the Northern Hemisphere and counterclock-
Structure
wise in the Southern Hemisphere. Anticyclones are Anticyclones may either be cold-core or warm-core
therefore characterized by negative relative vorticity systems. An example of each type is presented in
and low but positive absolute vorticity in the Northern Figure 1. Cold-core anticyclones are typically found
Hemisphere, while in the Southern Hemisphere they on the poleward side of the midlatitude baroclinic
are distinguished by positive relative vorticity and zone. These are shallow systems with an anticyclonic
low but negative absolute vorticity. On sea-level circulation confined to the lower troposphere.
pressure or geopotential height analyses, they may be The geostrophic relative vorticity (curvature in the
subjectively identified by closed isobars or height isobars or geopotential height contours) is anti-
contours, whereas in vorticity analyses they may cyclonic near the Earth’s surface but becomes cyclonic
objectively be identified by relative vorticity minima (or less anticyclonic) by the middle troposphere. The
in the Northern Hemisphere and maxima in the region of high sea-level pressure over south-western
Southern Hemisphere. Canada in Figure 1 is a cold-core anticyclone; note its
At sea level, anticyclones typically originate as proximity to a local minimum in the 1000-500mb
cold, shallow circulations that migrate Equatorward thickness which, hydrostatically, is proportional to the
and evolve into warm, subtropical high-pressure vertically averaged temperature in the 1000-500 mb
systems penetrating well into the troposphere. layer. This sea-level anticyclone is located between
Aloft, anticyclones may appear at middle and high cyclonic and anticyclonic features at the 500 mb level
latitudes on isobaric surfaces. From hydrostatic con- (Figure 2 ) .
siderations, these are relatively warm systems. Anti- Warm-core anticyclones are found Equatorward of
cyclones aloft are often stationary or westward baroclinic zones and are characterized by circulations
drifting and thus may block the eastward progress of that remain or may become increasingly anticyclonic
other weather systems. Anticyclonic circulations at from sea level to the middle troposphere. The region of
high latitudes may penetrate into the stratosphere high sea-level pressure over the Arctic in Figure 1 is a
where they may be associated with sudden strato- warm-core anticyclone. Even though this system is at a
spheric warmings. higher latitude than the cold-core anticyclone over
Although not as actively researched as cyclones, south-western Canada, it is characterized by higher
anticyclones are important because the clear, dry 1000-500 mb thickness (vertically averaged tempera-
conditions usually associated with them may allow ture) and it is located near a local maximum in
strong nighttime radiative cooling and cold surface the thickness field. Note in Figure 2 that the sea-level
142 ANTICYCLONES
these predictions. This is particularly true for predictions Jones PD and Limbert DW (1987)A Data Bank ofAntarctic
beyond 50 years, when Antarctica may start to warm Surface Temperature and Pressure Data. Washington,
enough to have a significant impact on the ice sheets. DC: US Department of Energy.
King JC and Turner J (1997) Antarctic Meteorology and
Further Reading Climatology. Cambridge: Cambridge University Press.
Schwerdtfeger W (1970) The climate of the Antarctic. In:
Bromwich DH (1988) Snowfall in high southern latitudes. Orvig S (ed.) World Survey of Climatology, vol. 14,
Reviews of Geophysics 26: 149-168. Climates of the Polar Regions, 253-355. New York:
Carleton AM (1992)Synoptic interactions between Antarc- Elsevier.
tica and lower latitudes. Australian Meteorological Schwerdtfeger W (1984) Weather and Climate of the
Magazine 40: 129-141. Antarctic. Amsterdam: Elsevier.
S J Colucci, Cornell University, Ithaca, NY, USA temperatures. The convectively stable air of anticy-
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. clones may allow air pollutants to concentrate near the
Earth’s surface. Finally, the blocking action of anti-
cyclones aloft may cause persistently anomalous
Introduction
weather conditions at the Earth’s surface.
Anticyclones are regions of relatively high pressure on
horizontal surfaces, or high geopotential height on
isobaric surfaces, around which air circulates clock-
wise in the Northern Hemisphere and counterclock-
Structure
wise in the Southern Hemisphere. Anticyclones are Anticyclones may either be cold-core or warm-core
therefore characterized by negative relative vorticity systems. An example of each type is presented in
and low but positive absolute vorticity in the Northern Figure 1. Cold-core anticyclones are typically found
Hemisphere, while in the Southern Hemisphere they on the poleward side of the midlatitude baroclinic
are distinguished by positive relative vorticity and zone. These are shallow systems with an anticyclonic
low but negative absolute vorticity. On sea-level circulation confined to the lower troposphere.
pressure or geopotential height analyses, they may be The geostrophic relative vorticity (curvature in the
subjectively identified by closed isobars or height isobars or geopotential height contours) is anti-
contours, whereas in vorticity analyses they may cyclonic near the Earth’s surface but becomes cyclonic
objectively be identified by relative vorticity minima (or less anticyclonic) by the middle troposphere. The
in the Northern Hemisphere and maxima in the region of high sea-level pressure over south-western
Southern Hemisphere. Canada in Figure 1 is a cold-core anticyclone; note its
At sea level, anticyclones typically originate as proximity to a local minimum in the 1000-500mb
cold, shallow circulations that migrate Equatorward thickness which, hydrostatically, is proportional to the
and evolve into warm, subtropical high-pressure vertically averaged temperature in the 1000-500 mb
systems penetrating well into the troposphere. layer. This sea-level anticyclone is located between
Aloft, anticyclones may appear at middle and high cyclonic and anticyclonic features at the 500 mb level
latitudes on isobaric surfaces. From hydrostatic con- (Figure 2 ) .
siderations, these are relatively warm systems. Anti- Warm-core anticyclones are found Equatorward of
cyclones aloft are often stationary or westward baroclinic zones and are characterized by circulations
drifting and thus may block the eastward progress of that remain or may become increasingly anticyclonic
other weather systems. Anticyclonic circulations at from sea level to the middle troposphere. The region of
high latitudes may penetrate into the stratosphere high sea-level pressure over the Arctic in Figure 1 is a
where they may be associated with sudden strato- warm-core anticyclone. Even though this system is at a
spheric warmings. higher latitude than the cold-core anticyclone over
Although not as actively researched as cyclones, south-western Canada, it is characterized by higher
anticyclones are important because the clear, dry 1000-500 mb thickness (vertically averaged tempera-
conditions usually associated with them may allow ture) and it is located near a local maximum in
strong nighttime radiative cooling and cold surface the thickness field. Note in Figure 2 that the sea-level
ANTICYCLONES 143
Figure 1 Sea-level pressure (solid contours in millibars with leading 9 or 10 omitted, at 4mb intervals) and 1000-500mb thickness
(dashed contours in dekameters, at 6 dm intervals) at 0000 UTC, 16 December 2000.
anticyclone is located beneath a well-defined anticy- mass convergence occurs over a lower-tropospheric
clone (closed contours around relatively high geopo- cold-air pool which, hydrostatically, would be associ-
tential height) at the 500-mb level. ated with relatively high sea-level pressure. Warm
Both warm-core and cold-core anticyclones are anticyclogenesis may occur if mass convergence
characterized by gently subsiding vertical motion in occurs over relatively high sea-level pressure at lower
the troposphere. This subsidence favors clear skies latitudes.
promoting strong nighttime radiative cooling of the More commonly, cold anticyclones evolve into
Earth’s surface near the centers of these anticyclones. warm anticyclones as follows. The circulation
The adiabatic warming of the sinking air coupled with around cold anticyclones draws cold air Equatorward,
radiative cooling at the surface often produces an forcing the sea-level pressure to rise locally. The
inversion in the vertical temperature profile; this anticyclone relocates towards rising sea-level pressure.
inversion may be eroded or destroyed by daytime Thus, cold anticyclones usually drift Equatorward
radiative heating and vertical mixing in the boundary with time.
layer. Regardless, anticyclones are distinguished by The mass convergence over the cold anticyclone
strong static stability. forces air to sink through the troposphere and to
adiabatically warm. The anticyclone thus becomes
warmer over time and may eventually be located
Dynamics Equatorward of the midlatitude baroclinic zone. Fric-
Convergence of mass in the upper troposphere is the tionally induced mass divergence at the Earth’s surface
primary mechanism responsible for the relatively high forces the sea-level pressure to fall at the anticyclone
sea-level pressure at anticyclone centers. From center, which then weakens. The anticyclone may
considerations of gradient wind balance, this reintensify as a warm system if mass convergence aloft
mass convergence occurs downwind of anticyclonic exceeds the lower tropospheric mass divergence near
circulations, or near regions of anticyclonic vorticity the anticyclone center in magnitude.
advection. The formation of new, cold-core anticy- Other mechanisms may contribute to anticyclone
clones (or anticyclogenesis) is favored when this formation and intensification. While there is usually
144 ANTICYCLONES
Figure 2 The 500 mb geopotentialheight (solidcontours in dekameters,at 6 dm intervals)and temperature(dashedcontours in Celsius,

at 5°C intervals) at 0000 UTC, 16 December2000.
very little temperature advection in the lower tropo- investigations reveal that, over the Northern Hemi-
sphere over anticyclone centers, the advection of cold sphere (Figure 3), sea-level anticyclones are most
air in the upper troposphere over a sea-level anticy- frequently observed in a band over the midlatitude
clone center may contribute to its intensification. Pacific Ocean, in a broadband near North America and
Clear conditions near cold anticyclone centers may centered on the Great Lakes, and near Mongolia. Over
result in the formation of ice fogs, and radiative heat the Southern Hemisphere, similarly defined sea-level
loss from these fogs may contribute to sea-level anticyclones are concentrated within a midlatitude
pressure rises and anticyclone intensification. How- (25-45" S) band with little longitudinal variation
ever, this effect is believed to be small. (Figure 4).
In the middle troposphere over the Northern Hem-
isphere (Figure 5 ) , anticyclones identified objectively
from maxima in geopotential height fields are most
Climatology frequently observed during the summer at low lati-
Early climatologies of weather systems were construct- tudes, especially over continental regions. During the
ed through manual inspection of sea-level pressure winter, these systems are rare but observed occasion-
charts and subjective identification of centers of closed ally over high latitude oceans. These latter systems
isobars. Contemporary investigations employ auto- may be associated with blocking, as discussed below.
mated procedures to identify objectively these systems No comparable results exist for the Southern Hemi-
from maxima in sea-level pressure analyses. Such sphere to date.
ANTICYCLONES 145
eastward motion of other weather systems. In order

for the blocking system to be nearly stationary and
persistent, the anticyclonic circulation aloft must be
continually maintained on its upstream flank by fluxes
of anticyclonic potential vorticity (lower tropospheric
warm air and middle tropospheric anticyclonic vorti-
city). These fluxes are provided by smaller-scale waves
approaching the blocking system. Additionally, the
strong static stability of anticyclones helps maintain
them against dissolution by convective mixing. Sink-
ing motion in the statically stable, anticyclonic environ-
ment promotes adiabatic warming and maintenance
of a deep, warm anticyclonic system.
Blocking anticyclones may be observed anywhere
and at any time, but are favored during the cool season
over the oceans, particularly the eastern oceans of the
Northern Hemisphere and the eastern Pacific Ocean in
the Southern Hemisphere. These locations are also
downstream of the principal storm tracks and are also
Figure 3 Sea-level anticyclone track density, in number of locations of climatologically preferred diffluent flow
centers per month, in the 1980-86 analyses of the European fields in the middle troposphere. The repeated inter-
Center for Medium-Range Weather Forecasts. (Reproduced with action of small-scale systems with the diffluent flow
permission from Sinclair MR and Watterson IG (1999) Journal of enhances the diffluence (by making its poleward
Climate 12: 3467-3485. Boston: American Meteorological branch more anticyclonic) until an anticyclonic circu-
Society.)
lation is established aloft. Quasi-geostrophically, this
diffluence is enhanced by the local deposition of
Blocking anticyclonic potential vorticity. Equatorward of the
blocking anticyclone, the normal westerly flow may
Warm anticyclones that penetrate deep into the reverse to easterlies over a considerable longitudinal
troposphere from the Earth’s surface in midlatitudes
distance and for periods of a week or more. The
may, if stationary and persistent, block the normal
blocking anticyclone may even penetrate into the
stratosphere, causing a reversal of the flow there from
westerly to easterly and sudden stratospheric warm-
ings. Thus, blocking anticyclones are of considerable
scientific and practical importance.
Impact
Because they are distinguished by clear skies and
subsiding air, anticyclones are typically associated
with fair weather. A stationary and persistent anticy-
clone may produce prolonged fair and dry weather
conditions, depleting soil moisture and stressing crops
and water supplies. The strong stability of anticy-
clones may stagnate air near the Earth’s surface,
leading to enhanced concentrations of pollutants.
Clear skies near anticyclone centers favor strong
nocturnal cooling near the Earth’s surface; these
conditions during the growing season may damage
crops. The Equatorward circulation of cold air around
anticyclones may cause sudden cold-air outbreaks
over midlatitudes. The cold air associated with anti-
Figure 4 As in Figure 3 but for the Southern Hemisphere.
(Reproduced with permission from Sinclair MR and Watterson IG
cyclones may become wedged or dammed against
(1999) Journal of Climate 12: 3467-3485. Boston: American mountain ranges, leading to freezing rain or ice if
Meteorological Society.) warm moist air is circulated over the dammed, cold air.
146 ARCTIC CLIMATE
Winter
Summer
Figure5 Number of twice-daily 500 mb geopotential height analyses during 1963-77 with anticyclones in 2" latitude by 5" grid boxes.
(Reproduced with permission from Bell GD and Bosari LF (1989) Monthly Weather Review 117: 2142-2163. Boston: American
Meteorological Society.)
Thus, while perhaps not as dramatic as cyclones, Further Reading

have their Own unique and interesting Bluestein HB (1992) Synoptic-Dynamic Meteorology in
features and impacts. Midlatitudes. New York: Oxford University Press.
Holton JR (1992) A n Introduction to Dynamic Meteorol-
ogy. New York: Academic Press.
See also Palmen E and Newton CW (1969)Atmospheric Circulation
Cold Air Damming. Dynamic Meteorology: Potential Systems: Their Stvucture and Physical Interpretation.
Vorticity. Quasi-geostrophicTheory. Static Stability. New York: Academic Press.
M C Serrere, University of Colorado, Boulder, CO, USA of the atmosphere. Key features of the Arctic, such as
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. its as low mean annual air temperatures, stable
boundary layer, sea ice cover, permafrost and seasonal
snow cover, in large part result ultimately from limited
Introduction solar radiation receipts. The high albedo of the snow
Annual receipts of solar radiation at the top of the and ice cover helps to maintain the Arctic in a low
Arctic atmosphere are much weaker than in equatorial energy state. Regional features of the atmospheric and
regions. Hence, the Arctic functions as the Northern ocean circulation, surface cover and topography
Hemisphere heat sink that establishes latitudinal modify primary latitudinal controls to result in a
pressure gradients which drive the general circulation wide variety of climate conditions. Owing to radiation
146 ARCTIC CLIMATE
Winter
Summer
Figure5 Number of twice-daily 500 mb geopotential height analyses during 1963-77 with anticyclones in 2" latitude by 5" grid boxes.
(Reproduced with permission from Bell GD and Bosari LF (1989) Monthly Weather Review 117: 2142-2163. Boston: American
Thus, while perhaps not as dramatic as cyclones, Further Reading

have their Own unique and interesting Bluestein HB (1992) Synoptic-Dynamic Meteorology in
features and impacts. Midlatitudes. New York: Oxford University Press.
Holton JR (1992) A n Introduction to Dynamic Meteorol-
ogy. New York: Academic Press.
See also Palmen E and Newton CW (1969)Atmospheric Circulation
Cold Air Damming. Dynamic Meteorology: Potential Systems: Their Stvucture and Physical Interpretation.
Vorticity. Quasi-geostrophicTheory. Static Stability. New York: Academic Press.
M C Serrere, University of Colorado, Boulder, CO, USA of the atmosphere. Key features of the Arctic, such as
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. its as low mean annual air temperatures, stable
boundary layer, sea ice cover, permafrost and seasonal
snow cover, in large part result ultimately from limited
Introduction solar radiation receipts. The high albedo of the snow
Annual receipts of solar radiation at the top of the and ice cover helps to maintain the Arctic in a low
Arctic atmosphere are much weaker than in equatorial energy state. Regional features of the atmospheric and
regions. Hence, the Arctic functions as the Northern ocean circulation, surface cover and topography
Hemisphere heat sink that establishes latitudinal modify primary latitudinal controls to result in a
pressure gradients which drive the general circulation wide variety of climate conditions. Owing to radiation
ARCTIC CLIMATE 147
feedbacks and couplings between the Arctic and (Figure 1).The Arctic Ocean is hence often referred
global ocean, the Arctic has gained a prominent role to as a Mediterranean-type sea. The dominant feature
in the climate change debate. of the ocean surface is its sea ice cover, which ranges
in areal extent from about 14.8 x 106 km2 in March
to about 7.8 x 106km2 in September, but with
Key Physical Features large seasonal and interannual variability. The ice
Most of the area north of 70"N is occupied by the is typically 1-5m thick but also highly variable
Arctic Ocean. Except for the sector between about (Barry et al. 1993). Most of the land surface is
20"E and 2OoW, the ocean is surrounded by land snow-covered from October through May, with the
150'
30"
180 0"
30"
150
.,.- - - Absolute max.
Average max.
Seasonal ice
Average min.
- Absolute min.
\7
120" 90" w 60"
Figure 1 Geography of the Arctic region and the average and extreme limits of sea ice (Reproduced with permission from Barry RG
(1983) Arctic Ocean ice and climate: perspectives on a century of polar research. Annals, Association of American Geographers 73:
485-501 .)
148 ARCTIC CLIMATE
duration of snow cover increasing with latitude. Over

the central Arctic Ocean, snow cover is generally
present for 10 months of the year. However, precip-
itation is generally scant in the Arctic. Much of the
land is classified as polar desert, often with less than
5% plant cover. In the low Arctic, the tundra
commonly includes shrub vegetation of birch and
willow. Permanent land ice is restricted primarily to
Greenland (the Greenland ice sheet) and the ice caps
and glaciers of north-eastern Canada. However, most
of the land is underlain by perennially frozen ground
(permafrost), overlain by an active layer exhibiting
seasonal thaw. Permafrost acts to keep moisture near
the surface. Many areas are covered by shallow thaw
lakes.
Atmospheric Circulation
Large-Scale Features
The primary feature of the northern high-latitude,
mid-tropospheric circulation is the polar vortex.
The vortex is strongly asymmetric during winter
(Figure 2A) with major troughs over eastern North
America and eastern Asia and a weaker trough
over western Eurasia (the Urals trough). A strong
ridge is located over western North America.
The lowest winter pressure heights are located over
northern Canada. These features are related to
orography, land-ocean distribution and radiative
forcing. The polar vortex weakens during summer
and is more symmetric than its winter counterpart
(Figure 2B).
The dominant sea level features of the mean winter
circulation (Figure 3A) are the Icelandic Low off the
southeast coast of Greenland, the Aleutian Low in the
north Pacific basin, and the Siberian high over central
Figure 2 Mean 500 hPa height fields (m) for (A) January and
Eurasia. The Icelandic and Aleutian lows are main- (8)July based on NCEP/NCAR reanalysis data for the period
tained by low-level thermal effects of the compara- 1960-99.
tively warm underlying ocean, position downstream
of the major mid-tropospheric stationary troughs
where eddy activity is favored and regional cyclone Cyclones in this area take a northerly to easterly track,
development processes. The Siberian High is a cold, and collectively represent part of the North Atlantic
shallow feature, driven largely by long-wave radiative cyclone track. Activity peaks in the vicinity of the
cooling. The Icelandic and Aleutian lows are much Icelandic Low. This is a region of frequent cyclogenesis
weaker during summer (Figure 3B). Summer also sees and strong cyclone deepening rates. Development is
replacement of the Siberian High by mean low enhanced because of contrasts between the warm,
pressure. A weak high pressure cell is found in the northward-flowing North Atlantic drift current and
Beaufort Sea. An area of mean low pressure is also the cold, southward-flowing East Greenland current,
found near the pole. proximity to the sea ice margin, and vorticity produc-
tion on the lee of the Greenland ice sheet. The North
Extratropical Cyclone Activity and Polar Lows
Atlantic track is weaker in summer, but cyclone acti-
In accord with the mid-tropospheric steering currents vity increases over land (Figure 4B). Summer cyclo-
(Figure 2A), winter cyclone activity is most prominent genesis occurs in preferred regions over east central
over the Atlantic side of the Arctic (Figure 4A). Eurasia and over Alaska and extending south-east.
ARCTIC CLIMATE 149
Figure 3 Mean sea-level pressure fields (hPa) for (A) January Figure 4 Average seasonal number of extratropical cyclone
and (B) July based on NCEP/NCAR reanalysis data for the period centers for (A) winter and (B) summer. Results are based on an
1960-99. Note the different contour interval for January (4hPa) automated cyclone identification algorithm applied to 6-hourly sea
compared with July (2 hPa). level pressure fields from the NCEP/NCAR reanalysis for the
period 1970-99. Dotted contours are used to highlight areas with
more than 3.5 systems per season.
A summer cyclone maximum is also found over the
central Arctic Ocean. This feature arises largely from
the migration and subsequent occlusion of systems Frontal Activity
generated over Eurasia and along the weakened
North Atlantic track. Serreze (1995) provides further Early Canadian analysis schemes adopted a three-
reading. front model of the westerlies, with the northernmost
Polar lows are mesoscale systems that form within representing Arctic fronts, hence separating Arctic
or at the leading edge of polar airstreams. They are from polar air masses. More recent studies based on
commonly found in the Arctic peripheral seas during aircraft data collected during the winter season
the winter season. Polar lows are typically less than present clear evidence of separate Arctic jet streams
500 km in diameter. They may intensify rapidly and with well-defined tropopause folds (as diagnosed from
surface winds speeds can reach hurricane force potential vorticity) between the lower (approximately
(Businger and Reed 1989). 5 km) Arctic tropopause to the north and the higher
150 ARCTIC CLIMATE
( 7 k m ) polar tropopause to the south (Shapiro et al.

1987). 340 I
The concept of preferred geographical regions of 300
frontal activity in northern high latitudes emerging as
250
distinct from frontal activity in middle latitudes
(termed the ‘Arctic Frontal Zone’) has a long history. 200
A maximum in frontal frequencies is found during
summer along northern Eurasia from about 60” to 150 i--
70”N, best expressed over the eastern half of the
100
continent. A similar relative maximum is found over N
-
Alaska, which although best expressed in summer is E

present year-round. These features are clearly separate 3
Y 50
X
from the polar frontal zone in the middle latitudes of 2 0
the Pacific basin. While some separation between C
.-0
high- and middle-latitude frontal activity is observed Z -50
2
in every season, the summer is distinguished by the a: -100
development of a mean baroclinic zone aligned along
the Arctic Ocean coastline and associated wind
maxima in the upper troposphere. While it has been
postulated that the frontal zone arises from contrasts
-150 1
in energy balance between the tundra and boreal
forest, it appears that coastal baroclinicity and focus-
ing of the baroclinicity by orography play stronger
roles. Regions of maximum summer frontal frequency
correspond to preferred areas of summer cyclogenesis
over Eurasia and Alaska (Serreze et al. 2001).
-3401 , i , l i l i
J F M A M J J A S O N
Months
Figure 5 Monthly radiation balance components (W m-‘) for the

Surface Energy Budget central Arctic Ocean. Notation is as follows: 6,
incoming longwave
radiation Fr,incoming solar radiation; Fr(l- u s ) , solar radiation
Figure 5 shows typical monthly values of radiative flux absorbed at the surface (us is surface albedo): 6 - m T 4 , net
components for the central Arctic Ocean. The outgo- longwave radiation; mT4,outgoing longwave radiation, and Rn,net
ing long-wave flux from the surface decreases from radiation. (Reproduced with permission from Barry eta/.,1993).
about 320 W mP2 in summer (when the sea ice surface
is melting) to about 200 W m P 2 in winter. The incom-
ing long-wave flux varies between 160 W m P 2 in emanating from wide (>10km) open water areas
winter to 3 0 0 W r 1 - ~in July. For all months, the net (leads and polynyas) that extend to 4 k m in the
long-wave flux is directed away from the surface. The atmosphere have been observed (Barry et al. 1993).
downwelling short-wave flux is zero during winter, The fundamental difference between the surface
rising to about 300 W mP2 in June. Because of the high energy budgets of the Arctic Ocean, glaciers and
surface albedo (exceeding 0.80 when covered with tundra is the portion energy used to melt snow and ice.
fresh snow), comparatively little of the solar flux is Once the snow is melted from the tundra, energy can
absorbed by the surface. A fraction of the incoming be used in sensible heating and to evaporate water. The
solar radiation (typically 1 5 % ) penetrates into the consumption of heat through melt on the ocean and
snow and ice. Net radiation is directed away from the glaciers is about 4-6 times larger than on the tundra.
surface from October through March and peaks in Consequently, sensible heat is transferred from the
June at about 80Wrnp2. During winter there is a atmosphere to the surface of the oceans and glaciers,
conductive heat flux through the ice to the surface. On while it is carried from the surface to the atmosphere in
an annual basis, the sensible and latent heat fluxes the tundra. Evaporation is, on average, the most
together account for 20-50% of the net radiation. significant heat sink on tundra, and is considerably
During summer, most of the net radiation is used to larger than on the ocean and glaciers (Ohmura 1984).
melt snow and ice. Locally, over areas of open water or A key control on Arctic surface energy budgets is
thin ice where strong temperature gradients are cloud cover. Winter cloud fractions range from 40 to
formed in the boundary layer, winter sensible heat 70%, greatest over the Atlantic side. Total cloud
fluxes may reach 600 W mP2. Condensate plumes fractions rise to 70-90% in summer. There is a rapid
ARCTIC CLIMATE 151
increase between April and May, characterized by the in instrument types and reporting practices, and the
development of extensive low-level stratus over the sparse precipitation monitoring network. Figure 7
ocean. For most of the year, the cloud radiative forcing shows the distribution of annual precipitation based
is positive, meaning that clouds have a warming effect on a gridded climatology that compiles data from
at the surface. The sign and magnitude of the cloud several sources. The highest totals are found off the
radiative forcing depends on the solar flux above the south-east coast of Greenland (locally > 2400 mm),
clouds, cloud albedo, optical thickness and tempera- with amounts decreasing north-east to about 400 mm
ture, surface albedo, and multiple reflections between in the Kara Sea. This pattern manifests the pattern of
the surface and cloud base. Curry et al. (1996)reviews cyclone activity shown in Figure 2A. High totals are
Arctic cloud characteristics, radiative forcing, and also found over southern Alaska. The lowest annual
feedback processes. totals ( < 200 mm) are found over the Beaufort Sea and
northern Canada. The winter pattern is qualitatively
Air Temperature and Boundary Layer similar to that seen in the annual mean. For example,
January precipitation ranges from over 200 mm in the
The winter surface air temperature field is strongly northern North Atlantic to less than 1 0 m m over
controlled by downwelling long-wave radiation. northern Canada and east-central Eurasia. The Atlan-
However, as seen in Figure 6A, winter surface air tic side maximum largely disappears in summer.
temperatures decrease sharply from the northern Precipitation is more uniform across the Arctic, with
North Atlantic to the central Arctic Ocean. The higher markedly higher totals as compared with winter over
temperatures over the Atlantic sector arise from ocean land areas. This is consistent with seasonal changes in
heat fluxes (which keep the region free of sea ice), synoptic activity (Figure 2B). Convective precipitation
extensive cloud cover, and horizontal atmospheric is not unknown over Arctic land areas during summer.
heat transports associated with the North Atlantic Winter precipitation is largely stored in the snow-
cyclone track. The lowest winter air temperatures are pack. Maximum spring snow depths are highly
found over east-central Eurasia in association with the variable due to differences in precipitation, tempera-
Siberian high. Heat fluxes through areas of open water ture, topographic setting and redistribution by wind.
and thin ice result in comparatively higher tempera- Values of 20-50cm over the Arctic Ocean and
tures over the central Arctic Ocean. Low temperatures 40-70 cm over the subarctic can be considered typical.
over Greenland reflect elevation. Summer air temper- Mean hydrographs for Arctic rivers exhibit a late
atures exhibit a much more zonal distribution (Figure spring to early summer peak in discharge due to melt of
6B). Because of the melting sea ice cover, summer the snow pack.
temperatures over the Arctic Ocean are close to zero. Direct estimates of evaporation are very scanty.
Higher temperatures over land reflect lower latitude However, large-scale estimates of precipitation minus
and sensible heating of the snow-free surface. Note the evaporation ( P - E ) (net precipitation) can be ob-
strong temperature gradients along the coastline. tained through evaluation of the atmospheric vapor
Winter radiation deficits give rise to strong surface- flux convergence (Cullather et al. 2000). Estimated
based temperature inversions (Kahll990). Away from mean annual P - E (Figure 8) is typically 150-300 mm
the moderating effects of the Atlantic sector, winter over land, 200 mm over the central Arctic Ocean and
inversions are typically 1000 m deep, with a temper- over 1OOOmm in the vicinity of the Icelandic Low.
ature difference across the inversion layer of 10-12°C. Although precipitation over much of the land area
Inversion depth and strength vary widely, however, in peaks in summer, P - E for this season (not shown)
response to local topographic conditions, winds, and tends to be small or even negative, pointing to
cloud cover. Inversions over the central Arctic ocean significant regional recycling of water vapor.
tend to be weaker than over land owing to heat fluxes
through areas of open water and thin ice. Inversions Freshwater Budget and Circulation of the
are also common in summer, although they are weaker Arctic Ocean
than their winter counterparts and are typically The Arctic Ocean is unique in receiving discharge from
separated from the surface by a mixed layer. four of the world’s major rivers (the Ob, Yenisei, Lena,
and Mackenzie). River discharge contributes about
Hydrologic Budget 360 mm of fresh water to the Arctic Ocean annually.
Along with the inflow of ocean water through the
Precipitation and Precipitation Minus Evaporation
Bering Strait and P - E over the Arctic Ocean itself,
(P - €1 river discharge helps to maintain a relatively fresh
Precipitation in the Arctic is difficult to measure be- surface layer. This layer extends down to about 200 m,
cause of gauge undercatch of blowing snow, changes and is often well mixed down to about 50m.
152 ARCTIC CLIMATE
(B)
Figure 6 Mean surface air temperature ("C) for (A) January and (B) July, based on the University of Washington International Arctic
Buoy Programme/Polar Exchange at the Sea Surface (IABPIPOLES) data set. (Reproduced with permission from Rigor IG, Colony RL
and Martin S (2000) Variations in surface air temperature observations in the Arctic, 1979-1997. Journal of Climate 13: 896-914.)
ARCTIC CLIMATE 153
mixing with the Atlantic layer, allowing sea ice to

form readily in winter and inhibiting melt during
summer.
The large-scale mean annual drift of the sea ice cover
is characterized by the clockwise Beaufort Gyre,
centered in the Canada Basin, and a mean drift of ice
from the Siberian coast, across the pole and through
Fram Strait, known as the Transpolar Drift Stream.
This pattern reflects roughly equal contributions by
winds and surface currents, the latter ultimately wind
driven to a large extent. The mean annual sea ice
circulation hence bears a strong resemblance to the
mean annual sea level circulation of the atmosphere
(Figure 9). Fresh water exported out of the Arctic
Ocean, largely via Fram Strait in the form of low-
salinity sea ice and liquid water, is believed to impact
on the overturning cell of the global ocean through
influencing convection in the subarctic gyres which in
Figure 7 Mean annual precipitation (mm) with estimated adjust- turn feed the North Atlantic (Lewis 2000).
ments for wind-induced gauge undercatch, changes in instrument
types and differences in observing methods (compiled from data
provided by P. Groisman, D. Yang, J. Eischeid and C. Willmott).
Variability and Change in Arctic
Dotted contours are used to highlight areas with precipitation Climate
exceeding 800 mm.
Arctic climate exhibits pronounced variability on
interannual to decadal scales. A major source of
Relatively warm waters of Atlantic origin are found at variability is associated with the North Atlantic
depths between 200 and 900m, which if brought to Oscillation (NAO), which describes mutual strength-
the surface would quickly melt the sea ice cover. ening and weakening of the Azores High and the
However, at low water temperatures of the Arctic Icelandic Low. The NAO has concentrations of
Ocean, the density structure is determined by salinity. power at 24, 8, and 2.1 years and also has a
Hence the fresh surface layer suppresses vertical multidecadal signal. Under the positive mode of the
NAO (a deep Icelandic Low), positive temperature
anomalies are found over the western Eurasian Arctic
with negative anomalies over north-eastern Canada
and the northern North Atlantic. In turn, the North
Atlantic cyclone track extends deeper into the Arctic
Ocean. Roughly opposing anomalies are associated
with negative NAO states. More recently, attention
has been paid to the 'Arctic Oscillation' or AO, which
represents the leading empirical orthogonal function
of monthly sea-level pressure anomalies poleward of
20" N (Thompson and Wallace 1998). Pressure vari-
ability associated with the A 0 is characterized by a
primary center of action over the Arctic Ocean and
opposing anomalies in midlatitudes of the Pacific and
Atlantic basins. The A 0 signal represents a strength-
ening and weakening of the polar vortex. Time series
of the A 0 and NAO are highly correlated, and
the exact relationship between these teleconnections
is being debated. Arctic climate variability is also
linked to the El Nifio Southern Oscillation (ENSO),
Figure 8 Mean annual precipitation minus evaporation (P- E )
particularly as it relates to variability in the strength
(mm) based on calculations of the vapor flux convergence using
NCEPlNCAR reanalysis data for the period 1070-1 999 (data and location of the Aleutian Low, as well as
providedby B. Brornwich).Areas with negative P - E are indicated other teleconnections, such as the North Pacific
with dotted contours. Oscillation (NPO).
154 ARCTIC CLIMATE
Figure9 Mean annual sea ice drift (cm s- I ) , based on data from drifting buoys, manned and unmannedcamps and mean annual sea-
level pressure (hPa). (Reproducedwith permission from Barry RG, Serreze MC, Maslanik JA and Preller RH (1993) The Arctic sea-ice
climate system: obsetvations and modelling. Reviews of Geophysics 31: 397-422.)
There has been pronounced change in the northern eastern Canada (Serreze et al. 2000). Paleoclimate
high-latitude environment from the 1970s through the evidence suggests that Arctic temperatures of the late
1990s, in part linked to generally positive phases of the twentieth century are the highest of the past 400 years
A 0 and NAO. This includes winter and spring (Overpeck et al. 1997).
warming over both continents (partly compensated General circulation models predict that the effects
by cooling over the northern North Atlantic). Warm- of anthropogenic greenhouse warming will be ampli-
ing is also evident over the central Arctic Ocean. There fied in the Arctic owing to feedbacks in which
has been a downward tendency in sea ice extent and variations in snow and sea ice extent, the stability of
thickness and increased areal extent of the Arctic the lower troposphere, and thawing of permafrost
Ocean’s Atlantic layer. Negative snow cover anoma- play key roles. However, regional patterns of Arctic
lies have dominated over both continents since the late warming differ greatly among simulations. Projected
1980s and terrestrial precipitation has increased in warming is greatest for late autumn and winter, largely
some areas. Small Arctic glaciers have exhibited because of the delayed onset of sea ice and snow cover.
generally negative mass balances. While permafrost Retreat of snow cover and sea ice is accompanied by
has warmed in Alaska and Russia, it has cooled in increased winter precipitation.
ARCTIC HAZE 155
Further Reading Ohmura A (1984) Comparative energy balance study for

Arctic tundra, sea surface, glaciers and boreal forests.
Barry RG (1983)Arctic Ocean ice and climate: perspectives Geojournal8: 221-228.
on a century of polar research. Annals, Association of Overpeck J, Hughen K, Hardy D, et al. (1997) Arctic
American Geographers 73: 485-501. environmental change of the last four centuries. Science
Barry RG, Serreze MC, Maslanik JA and Preller RH 278: 1251-1256.
(1993) The Arctic sea-ice climate system: observat- Rigor IG, Colony RL and Martin S (2000) Variations in
ions and modeling. Reviews of Geophysics 31: surface air temperature observationsin the Arctic, 1979-
397-422. 1997. Journal of Climate 13: 896-914.
Businger S and Reed RJ (1989) Cyclogenesis in cold air Serreze MC (1995)Climatological aspects of cyclone devel-
masses. Weather and Forecasting 4: 133-156. opment and decay in the Arctic. Atmosphere-Ocean 33:
Cullather RI, Bromwich DH and Serreze MC (2000) 1-23.
The atmospheric hydrologic cycle over the Arctic Serreze MC, Lynch AH and Clark MP (2001) The Arctic
basin from reanalyses. Part I: Comparison with observa- frontal zone as seen in the NCEP/NCAR reanalysis.
tions and previous studies. Journal of Climate 13: Journal of Climate (in press).
923-937. Serreze MC, Walsh JE, Chapin FS 111, et al. (2000) Obser-
Curry JA, Rossow WB, Randall D and Schramm JL (1996) vational evidence of recent change in the northern high
Overview of Arctic cloud and radiation characteristics. latitude environment. Climatic Change 46: 159-207.
Journal of Climate 9: 1731-1764. Shapiro MA, Hampel T and Krueger AJ (1987)The Arctic
Kahl JD (1990)Characteristics of the low-level temperature tropopause fold. Monthly Weather Review 115(2):
inversion along the Alaskan Arctic coastline. Inter- 444454.
national Journal of Climatology 10: 537-548. Thompson DWJ and Wallace JM (1998)The Arctic Oscil-
Lewis EL (ed.) (2000)The Freshwater Budget of the Arctic lation signature in the wintertime geopotentialheight and
Ocean. NATO Science Series. Series 2. Environmental temperature fields. Geophysical Research Letters 25:
Security; vol. 70. Dordrecht: Kluwer. 1297-1 300.
G Shaw, University of Alaska, Fairbanks, AK, USA sance missions in the Arctic. Mitchell recognized that
Copyright 2003 Elsevier Science Ltd. All Rights Resewed. the constituents were probably of the same order or
possibly smaller than the wavelength of visible light.
He speculated that the origin of the unknown haze
Introduction particles might be quite distant since small sub-
Arctic haze is composed of aerosol and gas pollution micrometer particles have long lifetimes in the
that builds up in the lower atmosphere across the atmosphere. The note that Mitchell published received
arctic regions in late winter and spring. Arctic haze is little recognition at the time.
derived from industrial by-products from continental Investigations of chemical composition and optical
regions surrounding the arctic basin. The gas and properties of arctic haze were conducted in the early
aerosol material constituting arctic haze congregates 1970s. Planning strategies that resulted from an
mainly within the meteorological boundaries of the informal symposium on arctic aerosols held in Nor-
arctic front, a system surrounding the North Pole, way in 1977 initiated investigations into arctic air
reaching its southernmost extent in late winter. During pollution and led to the recognition of an arctic-wide
late winter the system of arctic haze is roughly as large pollution phenomenon. The general picture of arctic
in areal extent (the region encircled in Figure 1) as the haze being pervasive, extensive and caused by
African continent. During its late winter maximum circumpolar industrial emission unfolded during
intensity, the ‘haze’ is visible to the eye, especially research carried out in the 1970s and 1980s.
when viewed edgewise from an aircraft. The haze If one considers the geographic location of pollution
reduces direct solar radiation at the surface, whitens sources and the nature of atmospheric circulation at
the sky and causes slight warming of the Earth- high latitudes (see the next section), it is quite apparent
atmosphere system. The haze is pollution transported (Figure 1)that northern Eurasian pollution sources are
to the Arctic over spatial scales of several thousands of likely to contribute larger amounts of pollution to the
kilometers. Arctic than from North America.
The haze was first noted by Murray Mitchell Jr, in The lack of sunlight, coupled with the snow-covered
1956, when flying on ‘Ptarmigan’ weather reconnais- surface that reflects what little sunlight there is back to
ARCTIC HAZE 155
Further Reading Ohmura A (1984) Comparative energy balance study for

Arctic tundra, sea surface, glaciers and boreal forests.
Barry RG (1983)Arctic Ocean ice and climate: perspectives Geojournal8: 221-228.
on a century of polar research. Annals, Association of Overpeck J, Hughen K, Hardy D, et al. (1997) Arctic
American Geographers 73: 485-501. environmental change of the last four centuries. Science
Barry RG, Serreze MC, Maslanik JA and Preller RH 278: 1251-1256.
(1993) The Arctic sea-ice climate system: observat- Rigor IG, Colony RL and Martin S (2000) Variations in
ions and modeling. Reviews of Geophysics 31: surface air temperature observationsin the Arctic, 1979-
397-422. 1997. Journal of Climate 13: 896-914.
Businger S and Reed RJ (1989) Cyclogenesis in cold air Serreze MC (1995)Climatological aspects of cyclone devel-
masses. Weather and Forecasting 4: 133-156. opment and decay in the Arctic. Atmosphere-Ocean 33:
Cullather RI, Bromwich DH and Serreze MC (2000) 1-23.
The atmospheric hydrologic cycle over the Arctic Serreze MC, Lynch AH and Clark MP (2001) The Arctic
basin from reanalyses. Part I: Comparison with observa- frontal zone as seen in the NCEP/NCAR reanalysis.
tions and previous studies. Journal of Climate 13: Journal of Climate (in press).
923-937. Serreze MC, Walsh JE, Chapin FS 111, et al. (2000) Obser-
Curry JA, Rossow WB, Randall D and Schramm JL (1996) vational evidence of recent change in the northern high
Overview of Arctic cloud and radiation characteristics. latitude environment. Climatic Change 46: 159-207.
Journal of Climate 9: 1731-1764. Shapiro MA, Hampel T and Krueger AJ (1987)The Arctic
Kahl JD (1990)Characteristics of the low-level temperature tropopause fold. Monthly Weather Review 115(2):
inversion along the Alaskan Arctic coastline. Inter- 444454.
national Journal of Climatology 10: 537-548. Thompson DWJ and Wallace JM (1998)The Arctic Oscil-
Lewis EL (ed.) (2000)The Freshwater Budget of the Arctic lation signature in the wintertime geopotentialheight and
Ocean. NATO Science Series. Series 2. Environmental temperature fields. Geophysical Research Letters 25:
Security; vol. 70. Dordrecht: Kluwer. 1297-1 300.
G Shaw, University of Alaska, Fairbanks, AK, USA sance missions in the Arctic. Mitchell recognized that
Copyright 2003 Elsevier Science Ltd. All Rights Resewed. the constituents were probably of the same order or
possibly smaller than the wavelength of visible light.
He speculated that the origin of the unknown haze
Introduction particles might be quite distant since small sub-
Arctic haze is composed of aerosol and gas pollution micrometer particles have long lifetimes in the
that builds up in the lower atmosphere across the atmosphere. The note that Mitchell published received
arctic regions in late winter and spring. Arctic haze is little recognition at the time.
derived from industrial by-products from continental Investigations of chemical composition and optical
regions surrounding the arctic basin. The gas and properties of arctic haze were conducted in the early
aerosol material constituting arctic haze congregates 1970s. Planning strategies that resulted from an
mainly within the meteorological boundaries of the informal symposium on arctic aerosols held in Nor-
arctic front, a system surrounding the North Pole, way in 1977 initiated investigations into arctic air
reaching its southernmost extent in late winter. During pollution and led to the recognition of an arctic-wide
late winter the system of arctic haze is roughly as large pollution phenomenon. The general picture of arctic
in areal extent (the region encircled in Figure 1) as the haze being pervasive, extensive and caused by
African continent. During its late winter maximum circumpolar industrial emission unfolded during
intensity, the ‘haze’ is visible to the eye, especially research carried out in the 1970s and 1980s.
when viewed edgewise from an aircraft. The haze If one considers the geographic location of pollution
reduces direct solar radiation at the surface, whitens sources and the nature of atmospheric circulation at
the sky and causes slight warming of the Earth- high latitudes (see the next section), it is quite apparent
atmosphere system. The haze is pollution transported (Figure 1)that northern Eurasian pollution sources are
to the Arctic over spatial scales of several thousands of likely to contribute larger amounts of pollution to the
kilometers. Arctic than from North America.
The haze was first noted by Murray Mitchell Jr, in The lack of sunlight, coupled with the snow-covered
1956, when flying on ‘Ptarmigan’ weather reconnais- surface that reflects what little sunlight there is back to
156 ARCTIC HAZE
Figure 2 Schematic diagram showing why arctic haze builds up

in the high latitude regions. Removal for scrubbing out the
atmosphere operates more slowly in the polar regions.
Figure 1 Annual emissionsof sulfur dioxide (in units of millions of
metric tons) in the high latitude Northern Hemispherethat influence
the Arctic. (Map courtesy of Leonard Barrie.)
planet. The turnover and almost constant drizzle is
instrumental in cleansing the arctic atmosphere,
space, results in especially strong cooling of the surface reducing concentrations of arctic haze to insignificant
layers and buildup of surface-based temperature levels. As a result, the concentration of arctic haze
inversions. Mixing is greatly inhibited under such undergoes a strong seasonal variation with maximum
circumstances. In much of the arctic basin, most in winter and minimum in summer. This ‘arctic air
especially around the Siberian High, cloudiness is thin mass’ picture provides a rough understanding of the
and sparse. These factors enhance the residence time of more persistent and larger features of arctic air
pollutants in the Arctic. It is mostly because of the pollution.
increase in residence time, rather than simply large By correlating synoptic meteorological patterns
masses of injected pollutants, that the arctic atmo- with episodes of arctic haze, it was concluded that
sphere becomes generally contaminated (Figure 2). there is often a close connection between outbreaks of
arctic haze and transport along or within anticyclonic
pathways in the atmosphere. North American sources
Seasonal and Geographic Variations of evidently contribute only in a minor way to the arctic
the Arctic Haze and Meteorological pollution burden. Polluted air masses originating from
the east coastal region of North America usually travel
Transport east across the Atlantic Ocean and experience rapid
In winter, the arctic air mass extends throughout the removal by storms within the Icelandic cyclone
high Arctic and extends down over Eurasia and North system.
America. Arctic haze has been described as having a Air trajectories associated with flow across the
‘dome’ 7-8km deep over the pole with shallow arctic basin are frequently associated with pollution
tongues of air 0-5 km deep spilling southward over episodes. During such instances, transport is often
the land masses. This air mass is statically stable controlled by the mid-northern Eurasian high, extend-
because of the strong temperature inversions and has ing over Kamchatka and another cell over the Alaska
relatively small liquid cloud water or ice content. area, or steered by cyclonic systems in the Barents Sea.
During summer, cloud cover is extensive throughout Studies of the meteorological patterns indicate that
the arctic basin. The uniform cloud deck forms when transport of arctic haze is associated with a quasi-
warm air flows over cold icepack; it is often associated stationary ‘blocking’ pattern in the atmosphere. This
with drizzle. The summer arctic stratus is one of the has been proposed to be a fundamental mechanism
most pervasive and persistent cloud systems on the that provides conditions for poleward transport of
ARCTIC HAZE 157
midlatitudinal air pollution, particularly in the Eur- 4000

Aerosol sulfate at alert, NWT
opean sector. The seasonal variation of such blocking 3500

seems to be an important factor, along with scavenging
and other removal processes, for understanding the
annual cycle of the arctic air pollution.
‘k 3000
2500
2. 2000
The isentropic dome of cold air centered roughly II
over the north pole helps explain the many elevated G 1500
1000
layers seen in arctic haze. The large static stability 500
inhibits vertical mixing of the material throughout the 0
troposphere. Also, the ‘dome’ causes midlatitude 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991
Year
ground-based pollutants to ascend adiabatically as
winds carry them poleward into the arctic basin. Figure 3 Concentration of sulfate (in nanograms of sulfate per
The map of annual emissions of sulfur dioxide cubic meter) in the Canadian Arctic measured at Alert, in the
with superimposed arctic air mass (Figure 1) helps Northwest Territories. (Courtesy of Leonard Barrie.)
identify the major sources and currents of pollution-
derived material affecting the Arctic. Thus, on the
basis of the relatively strong source region in the strongly over Eurasia, suggests major contribution to
central and western Eurasia sector, the occurrence of a arctic haze from Eurasian sources.
deep lobe of the arctic air mass over much of this The origin of the arctic haze was determined to be
source, the occurrence of a poleward flowing circula- Eurasian using a chemical fingerprint based on the
tion over this source area, and the absence of ratio of vanadium to manganese concentrations in the
precipitation, clouds and turbulence along the path- collected filtered samples of the arctic haze. This
way, one can conclude that Eurasia is of greater simple tracer system immediately suggested that the
importance than North America as a source region for greatest fraction of arctic haze aerosol derived from
the Arctic haze. Eurasian industrial sources, especially in the eastern
In the past decade or so, the occurrence of arctic sectors. The reason is that the former Soviet Union and
haze has become weaker and less frequent, perhaps eastern European nations are coal burning societies,
due to breakup of the Soviet Union and associated while the western nations are heavy users of petroleum
decrease in heavy industrial emissions, though sys- products, laced with vanadium used as catalyst in the
tematic changes in meteorological patterns may also be cracking process. The arctic haze value of V/Mn was
involved. very low, consistent with a coal burning source.
Concentrations of black carbon were also elevated,
consistent with the dirtier combustion that takes place
Chemistry of the Arctic Haze
in inefficient coal power plants of the type used in the
Samples of aerosol were collected in various arctic air former Soviet Union. It is quite interesting to see that
masses in the 1970s and analyzed for their chemical the essential picture about the sources and transport
composition in the hope that the origin of the hazy mechanisms for the arctic haze was recognized quite
layers might become clear. Although sulfate dominates clearly during the 1970s. During later aircraft missions
aerosol chemistry, the haze contains trace amounts of conducted by NOAA (the AGASP missions, Arctic
heavy metals, some of which suggest industrial Gas and Aerosol Sampling Program), the arctic haze
pollution sources. was sampled directly and the concentrations and
The chemical sampling also uncovered strong compositions of a number of minor constituents were
seasonal variation (discussed in the last section), determined. Again, sulfur in the form of sulfate was
with maximum occurrence of the arctic haze in the the predominant chemical compound in the arctic
late winter and early spring. Figure 3 shows the strong haze, even at the higher altitudes.
seasonal variation of sulfate aerosol concentration as More elaborate multivariate analysis of heavy metal
sampled in the Canadian Arctic. composition of the arctic haze aerosol confirmed that
The largest fraction of the arctic haze is sulfate. Note the greatest fraction of the arctic haze indeed originates
from Figure 1 that the Eurasian sulfur dioxide in Eurasia. These deductions were in large part made
emissions in areas liable to influence the Arctic are from wide and generous sharing of data among the
about a factor of 2-4 times larger than for North circumpolar nations in an informal ‘network’. By the
America. Note also that the major Eurasian sulfur mid-l980s, chemical fingerprinting methods had
dioxide emission sources are 5-1 0” higher in latitude determined that Eurasia and North America were the
in comparison to those in North America. This, along first and second primary source regions for the general
with the fact that the arctic air mass lobes down pollution phenomenon in the Arctic.
158 ARCTIC HAZE
More recent investigations combined back wind Optical Transparency and Climatic
meteorological trajectory models with geographic Effects
information and inventories of chemical emissions.
With the use of such models one can deduce In the early 1970s, unexpectedly high values for
partitioning of the source regions for the differing atmospheric turbidity were reported at the McCall
chemical species, although it has to be kept in mind Glacier in the Brooks Range in Alaska. In trying to
that the accuracy suffers since the network of synoptic understand the physical cause of the high turbidity,
stations in the Arctic is sparse. Such investigations additional measurements were made of the wave-
have indicated that a number of geographic locations length dependence of optical extinction caused by the
in Eurasia have high potential as major emission haze and the angular distribution of sky brightness.
sources. It was found that a Ni-Cu smelting complex These measurements confirmed Mitchell’s earlier
at Norislk is probably one of the major contributors to suspicions that the winter arctic haze consisted mainly
the haze. of small aerosols. During the early 1970s, arctic haze
Arctic haze in mid-winter mainly consists of over Alaska was found to be layered by flying a sun
sulfur dioxide. At that time the polar atmosphere photometer aboard a light (Cessna) aircraft. The
is a large chemical reactor. The gaseous form AGASP experiments of the 1980s demonstrated that
converts to particulate sulfates when the sun rises in the layers consist of aerosol particles that can have
spring. climatic effects by interacting with the solar radiation
More than sulfur-containing particles and heavy field all across the Arctic. It was later determined that
metals reach the Arctic. Organic compounds such as in addition to the submicrometer haze, there is often
polychlorinated aromatics and pesticides have been another source of haziness in the Arctic from
measured in arctic air pollution. precipitations of ice crystals, which sometimes form
in clear air. Such ‘diamond dust’, so called because of
its sparkling appearance, can also reduce visibility and
Residence Time for Arctic Air Pollution interact with the radiation field.
During the early investigations of the arctic haze Scattering and absorption of sunlight by arctic haze
phenomenon, in the 1970s, scientists were hampered was shown to have a slight warming effect on the
by the erroneous, but pervasive, belief that aerosol Earth-atmosphere column, of magnitude about 0.1”C
pollution is confined to regional scales up to a few per day, because carbonaceous material lowers
hundred kilometers at most. It was then clear that albedo. There is estimated to be a slight cooling at
under conditions of great atmospheric stability and the surface caused by scattering of light. In the winter
low water content, the residence time for aerosols and months there may be a slight warming of the arctic
the corresponding air parcel travel distance increase atmosphere caused by the interaction of the aerosol
dramatically. The majority of air pollution studies in with outgoing infrared radiation.
the 1960s and 1970s had concluded that the lifetime of However, in addition to slight climate effects caused
atmospheric aerosols is short, being only a few days. by absorption and scattering of light by the arctic haze
This assumption breaks down in the polar regions aerosols, there may be subtle and so far not well
since particles may remain airborne for weeks in the evaluated influences on climate from indirect radiative
stable arctic air. effects. These may result from the modification of
Two notable exceptions to the rule of ‘a few days cloud parameters, since the arctic haze introduces new
residence time’ for aerosols had, however, been sources of cloud condensation nuclei and, possibly,
recognized for a long time. One was the phenomenon new sources of freezing nuclei into the arctic basin.
of stratospheric haze from explosive volcanic erup-
tions, which is known to have lifetimes of several
years. The other was the long-range transport of
Ecological Implications
windblown dust from deserts. We now believe arctic Compounds such as pesticides, PCBs, persistent
haze to have an average lifetime shorter than strato- organics, as well as trace metals are detected through-
spheric dust veils, but comparable to long-distance out the arctic basin in the atmosphere and also in the
transport of desert dust outbreaks, which also occur in surface land and sea domains and scattered through-
dry air masses. out the region’s biome. Most of the atmospheric
The connection between arctic haze and anti- pollution and some of the surface and biological
cyclonic conditions, which was mentioned earlier, pollution is undoubtedly caused from the arctic haze
suggests that one reason why haze lasts so long is phenomena. But it should also be recognized that in
partly due to lowered exposure to rain and snow places in the Arctic the surface concentrations of
cleansing mechanisms. pollutants can be extremely small, even during times
ATMOSPHERIC TIDES 159
when the air is quite contaminated. This of course is Norwegian Sea and possibly in the Bering Sea. These
due to the long residence time of the haze. are important world fisheries and the consequences of
In addition to simple transport of material through dumping industrial pollutants that had congregated in
the atmosphere, another depositional mechanism was the arctic atmosphere are unknown.
suggested that involves a fractional distillation. The
surprisingly high concentration of organic material is Further Reading
determined in part by the temperature-dependent
partitioning of the low volatility compounds. A Barrie L (1986)In: Stonehouse B (ed.) Arctic Air Pollution.
process of global fractionation may be occurring, Cambridge: Cambridge University Press.
during which time organic compounds become Curry JA (1995) Interactions among aerosols, clouds and
climate of the Arctic Ocean. Science of the Total
latitudinally fractionated, condensing at different
Environment 160: 777-791.
ambient temperatures depending on their volatility. Rahn KA and Lowenthal DH (1984) Elemental tracers of
Substances with low vapor pressure preferentially distant regional pollution aerosols. Science 223: 132-139.
accumulate in the polar regions, much like deposition Schnell RC (1984) Arctic Haze and the Arctic Gas and
of vapor products on cold regions in vacuum systems. Aerosol Sampling Program (AGASP). Geophysical
Aerosols and gases are scrubbed out of the atmo- Research Letters 11: 361-364.
sphere in precipitating air masses. Many of these are Shaw GE (1995)The Arctic Haze phenomenon. Bulletin of
over oceanic regions in the North Atlantic, the the American Meteorological Society 76: 2403-2413.
M Hagan, National Center for Atmospheric Research, 24 h and 1 2 h periods, respectively. The lunar diurnal
Boulder, CO, USA tidal period is about 24.8 h, while the lunar semidiur-
J Forbes, University of Colorado, Boulder, CO, USA nal period is 12.4 h. Scientists often use a shorthand
notation to represent solar and lunar tides. SI and S2
A Richmond, National Center for Atmospheric refer respectively to the solar diurnal and semidiurnal
Research, Boulder, CO, USA
tides. Their lunar counterparts are M1 and M2.
Copyright 2003 Elsevier Science Ltd. All Rights Reserved Atmospheric tides have been studied for many
years, since they are evident in both surface pressure
and magnetic observations that date back to the early
Introduction
part of the twentieth century. Figure 1 illustrates a time
Atmospheric tides are ubiquitous features of the series of surface pressure measurements made at
Earth’s atmosphere. They are the persistent global Batavia (now known as Jakarta, Indonesia) during
oscillations that are observed in all types of atmos- the first 5 days of January in 1925. The dominant
pheric fields, including wind, temperature, pressure, feature of this time series provides evidence of the solar
density, and geopotential height. Tidal oscillations semidiurnal atmospheric tide. Specifically, there is a
have periods that are some integer fraction of a solar or 1-2 hPa deviation from the average pressure of about
lunar day. The solar diurnal and semidiurnal tides have 1011 hPa that occurs regularly at 12 h intervals. This
1017
h
a
n
E.
2
2 1013
v)
2
Q
.-0
i 1009
QE
3
1005 I , , , l a I I I I I I I 1 I I
1 2 3 4 5
Day
Figure 1 Surface pressure (hPa) at Batavia (Jakarta, Indonesia) against time during the first 5 days of January 1925.
when the air is quite contaminated. This of course is Norwegian Sea and possibly in the Bering Sea. These
due to the long residence time of the haze. are important world fisheries and the consequences of
In addition to simple transport of material through dumping industrial pollutants that had congregated in
the atmosphere, another depositional mechanism was the arctic atmosphere are unknown.
suggested that involves a fractional distillation. The
surprisingly high concentration of organic material is Further Reading
determined in part by the temperature-dependent
partitioning of the low volatility compounds. A Barrie L (1986)In: Stonehouse B (ed.) Arctic Air Pollution.
process of global fractionation may be occurring, Cambridge: Cambridge University Press.
during which time organic compounds become Curry JA (1995) Interactions among aerosols, clouds and
climate of the Arctic Ocean. Science of the Total
latitudinally fractionated, condensing at different
Environment 160: 777-791.
ambient temperatures depending on their volatility. Rahn KA and Lowenthal DH (1984) Elemental tracers of
Substances with low vapor pressure preferentially distant regional pollution aerosols. Science 223: 132-139.
accumulate in the polar regions, much like deposition Schnell RC (1984) Arctic Haze and the Arctic Gas and
of vapor products on cold regions in vacuum systems. Aerosol Sampling Program (AGASP). Geophysical
Aerosols and gases are scrubbed out of the atmo- Research Letters 11: 361-364.
sphere in precipitating air masses. Many of these are Shaw GE (1995)The Arctic Haze phenomenon. Bulletin of
over oceanic regions in the North Atlantic, the the American Meteorological Society 76: 2403-2413.
M Hagan, National Center for Atmospheric Research, 24 h and 1 2 h periods, respectively. The lunar diurnal
Boulder, CO, USA tidal period is about 24.8 h, while the lunar semidiur-
J Forbes, University of Colorado, Boulder, CO, USA nal period is 12.4 h. Scientists often use a shorthand
notation to represent solar and lunar tides. SI and S2
A Richmond, National Center for Atmospheric refer respectively to the solar diurnal and semidiurnal
Research, Boulder, CO, USA
tides. Their lunar counterparts are M1 and M2.
Copyright 2003 Elsevier Science Ltd. All Rights Reserved Atmospheric tides have been studied for many
years, since they are evident in both surface pressure
and magnetic observations that date back to the early
Introduction
part of the twentieth century. Figure 1 illustrates a time
Atmospheric tides are ubiquitous features of the series of surface pressure measurements made at
Earth’s atmosphere. They are the persistent global Batavia (now known as Jakarta, Indonesia) during
oscillations that are observed in all types of atmos- the first 5 days of January in 1925. The dominant
pheric fields, including wind, temperature, pressure, feature of this time series provides evidence of the solar
density, and geopotential height. Tidal oscillations semidiurnal atmospheric tide. Specifically, there is a
have periods that are some integer fraction of a solar or 1-2 hPa deviation from the average pressure of about
lunar day. The solar diurnal and semidiurnal tides have 1011 hPa that occurs regularly at 12 h intervals. This
1017
h
a
n
E.
2
2 1013
v)
2
Q
.-0
i 1009
QE
3
1005 I , , , l a I I I I I I I 1 I I
1 2 3 4 5
Day
Figure 1 Surface pressure (hPa) at Batavia (Jakarta, Indonesia) against time during the first 5 days of January 1925.
160 ATMOSPHERIC TIDES
semidiurnal variation is modulated by other varia- tide is diurnal and moves or migrates westward in
tions, but the former is such a persistent oscillation longitude with the apparent motion of the Sun from
that the semidiurnal tide is also the dominant oscilla- the perspective of a ground-based observer. Further,
tion in monthly, yearly, and even multiyear averages of Cph = -(2z/24) h-l. Similarly, if n = 2 and s = -2,
daily surface pressure measurements made at Batavia. then the wave is a migrating semidiurnal tide. The
Atmospheric tides are further characterized by their remaining set of global scale waves with tidal periods
sources. The Moon’s gravity forces the lunar atmos- that are not Sun-synchronous are known as non-
pheric tide, while solar atmospheric tides can be migrating tides. Nonmigrating tides may be viewed as
excited in several ways, including the absorption of waves that propagate to the west more rapidly or
solar radiation, large-scale latent heat release associ- slowly than the Sun, or that propagate eastward, or
ated with deep convective clouds in the troposphere, that are standing.
and the gravitational pull of the Sun. The restoring All tides contain components that propagate in the
force that acts on atmospheric tides is gravity, so tides vertical direction. The effects of upward-propagating
are a special class of buoyancy or gravity waves. tidal components are particularly important because
Unlike other gravity waves, tides are affected by the these waves grow in amplitude as they conserve energy
Earth’s rotation and sphericity because of their com- in an atmosphere whose density decreases with
paratively large periodicities and horizontal scales. increasing altitude. Thus, tides with insignificant
The general mathematical expression for a tidal amplitudes in their lower atmospheric regions of
oscillation is given by eqn [ 11,where A is the magnitude excitation often affect the upper atmosphere pro-
of the variation in some atmospheric field, o is its foundly because they introduce large atmospheric
frequency, t is universal time; A is longitude, and s is the variations with local time and because they may
zonal wavenumber. The magnitude of s represents the dissipate and deposit their energy and momentum
number of wave crests that occur along a latitude therein.
circle, while the sign of s is indicative of the zonal
direction of propagation. 4 is the so-called tidal phase.
Migrating Solar Tides
A crest of the wave occurs when 4 satisfies eqn [2].
The absorption of radiation by a longitudinally
A cos(ot - SI. - 4) [11 invariant atmosphere is the primary source of migrat-
ing solar tides. Owing to the rotation of the Earth, this
4 = ot - s2 PI absorption is periodic in time from the perspective of
the ground-based observer. The resultant heating gives
The horizontal phase speed of the tide, Cph, is defined
rise to migrating tidal oscillations. Solar radiation is
by differentiating eqn [2] and holding 4 constant (eqns
absorbed throughout the Earth’s atmosphere, thereby
[31 and 141). exciting migrating solar tides at almost all altitudes.
odt - SdA = 0 [31 Atomic oxygen, which is the most abundant atmos-
pheric constituent at altitudes about 150 km above the
d2 - c Earth’s surface, absorbs the shortest-wavelength solar
Cph = dt - s 141 radiation, known as the extreme ultraviolet. Increas-
ingly longer wavelengths are absorbed as the solar
s > 0 corresponds to eastward tidal propagation, and
radiation approaches the Earth’s surface. Molecular
s < 0 corresponds to westward tidal propagation.
oxygen, ( 0 2 ) absorbs the far-ultraviolet radiation
For solar tides, the nth harmonic frequency is
(100-200 nm) at altitudes near about 100-150 km,
o, = nol, where n is a positive integer and 01 =
and ozone ( 0 3 ) absorbs the 200-300 nm solar ultra-
(27c/24) h-l . Rewriting the mathematical expression
violet radiation at middle atmospheric altitudes be-
for a tide in terms of local solar time (hours),
tween about 30 and 70 km. Solar infrared radiation
t L = t + 2/01, results in a mathematical expression
may be absorbed by water vapor, (H2O)in the lowest
of the form of eqn [5].
part of the atmosphere.
+
A cos(o,tL - (S #)A - 4) 151 Even though there is little, if any, tidal forcing due to
solar heating in the upper mesosphere (- 80-100 km),
For the subset of atmospheric tides known as migra- measurements of winds and temperatures exhibit
ting solar tides, s = -n and eqn [5] reduces to eqn [6]. strong tidal signatures in this region. Figure 2 illus-
trates an example of the magnitude of the mean winds
A cos(o,tL - 4) [GI and the tidal oscillations over Adelaide, Australia at
Thus, migrating solar tides have the same local time these altitudes. The data points represent the eastward
variation at all longitudes. If n = 1 and s = -1, the winds that were measured with the Buckland Park
semidiurnal forcing in the middle atmosphere. Thus,

the diurnal tidal growth occurs over a deeper altitude
region than the semidiurnal growth and it is reason-
able to anticipate a diurnal amplitude that is larger
than the semidiurnal amplitude in the upper meso-
sphere. The aggregate characteristics of the mean
winds and tides that are illustrated in Figure 2 support
the claim that upward-propagating migrating tides
p
W 251 govern the large-scale dynamics of the upper meso-
sphere.
Migrating tides exhibit somewhat complicated
0 6 12 18 0 6 12 18 0 behavior in that the latitudinal structure of the
Local solar hours horizontal wind oscillations is dramatically different
from the temperature, pressure, or vertical velocity
Figure 2 Eastward winds (ms-') over Adelaide, Australia
against local time (h) on 8-9 August 1994 at 98km (top), 94 km structure. For example, the upward-propagating mi-
(middle),and 90 km (bottom). Data pointsare radar measurements grating diurnal tide is characterized by a primary
and dashed curves are model predictionsthat includethe migrating temperature amplitude maximum over the Equator
diurnal and semidiurnal tidal components. (Professor RA Vincent with secondary maxima near f30". The horizontal
provided the Adelaide radar data.) wind amplitudes exhibit minima over the Equator and
nearly symmetric amplitude peaks at low to middle
latitudes ( f(20-30')). While ground-based observa-
radar during two days in August 1994. The dashed tions provide an important perspective on the local
curves illustrate global-scale wave model (GSWM) behavior of waves with tidal frequencies, it is impos-
tidal predictions for August at the location of Adel- sible to decipher global structures from local struc-
aide. While the GSWM differs from the measurements tures without conducting correlative analysis of
in detail, the model captures the salient features of the measurements made at multiple locations over a
observed winds, particularly on 8 August. Differences broad range of latitudes. Further, in order to distin-
may be attributable to small-scale waves that are not guish migrating from nonmigrating tidal components,
included in GSWM or to sources of day-to-day tidal it is necessary to have a longitudinal distribution of
variability that are also omitted. measurements.
The GSWM predictions include mean winds
(17-18 m s - and both migrating diurnal and semi-
Nonmigrating Tides
diurnal components. Notably, the migrating diurnal
amplitudes (20-30 m s-') are larger than the mean Near the surface of the Earth the strong longitudinal
winds. The GSWM diurnal tidal amplitudes are at differences in topography, land-sea contrast, and
least a factor of 2 larger that the semidiurnal ampli- surface interactions produce zonal variations in the
tudes (8-15ms-l) and the phases of both compo- local time behavior of the atmosphere and thus excite
nents shift to earlier times with increasing altitude nonmigrating tides. Figure 3 provides a good example.
between 90 and 98 km. This behavior, which is known It illustrates the latitude-longitude distributions of the
as downward phase progression, is indicative of annual-average semidiurnal ( a )and diurnal (b)surface
upward-propagating wave energy. There are clear pressure amplitudes derived from a global array of
signatures of downward phase progression in the measurements. Both components are maximum at low
GSWM predictions in Figure 2. That is, the wind latitudes where the absorption of solar radiation is
predictions are most westward near 15.00 hours at greatest. The semidiurnal surface pressure oscillation,
98 km and there are similar features at progressively S2(p)is more uniform with respect to longitude, while
later times and progressively lower altitudes. The the diurnal amplitudes, Sl(p),are primarily restricted
vertical wavelength of the migrating diurnal tide over to land masses. The S2(p)behavior is dominated by the
Adelaide is much shorter than that of the migrating migrating (s = -2) response, while the S l ( p )behavior
semidiurnal tide, so the phase of the former progresses reflects the importance of sensible heating from the
far more rapidly than the latter phase and their ground. (Note that if S2(p) were solely due to a
combined effects result in a pattern of wave maxima migrating semidiurnal tide, Figure 3 would consist of a
and minima that evolves between altitudes. series of horizontal lines.) The S2(p) and S l ( p )
The migrating diurnal tide originates primarily in amplitudes are comparable over land and vary be-
the troposphere. Although tropospheric semidiurnal tween about 80 and 140Pa ( 8 and 14dPa). S2(p)
forcing is nonnegligible, there is comparatively more dominates over the oceans at latitudes lower than 30".
162 ATMOSPHERICTIDES
90" N
60" N
30" N
a
s.E 0"
m
1
30" S
60" S
90" s
180" 120" W 60" W 0" 60"E 120"E 180"
(A) Longitude
90" N
60" N
30" N
a
-
D
a
.-
m
-1
0"
30" S
60" S - w- . - -
90" s I I I I I I
180" 120" W 60" W 0" 60"E 120"E 180"
(B) Longitude
Figure 3 Contours of annual-average semidiurnal (A) and diurnal (B) surface pressure amplitudes (dPa) against longitude and latitude.
(Dr. A. Dai provided these figures, which are adapted with permission from Dai and Wang (1999).)
A significant fraction of S2(p) is due to a migrating based remote-sensing measurements made at these
semidiurnal component that is excited by absorption altitudes. During the 1990s remote sensing from the
of ultraviolet radiation by ozone at higher altitudes Upper Atmosphere Research Satellite (UARS)consid-
-
( 30-70 km) and propagates down to the surface.
The corresponding migrating diurnal component is
erably ameliorated this problem. Figure 4 illustrates
the latitude-longitude distribution of nonmigrating
excited less efficiently aloft because of its shorter diurnal northward wind amplitudes at 95 km for the
vertical wavelength, which explains the significant month of October, averaged over three years. The data
amplitude of S2(p) relative to Sl(p) and its relative utilized to construct this figure consist of wind
longitudinal uniformity. measurements made by the High Resolution Doppler
Through Fourier analysis the longitude variations
depicted in Figure 3 may be decomposed into a series
of wave components with different zonal wavenum-
bers s. Each of the resultant nonmigrating tidal
components possesses different vertical propagation
characteristics that depend, for example, on its sensi-
tivity to the prevailing zonal winds and its vertical
wavelength. It is difficult to track vertical tidal
propagation into the upper atmosphere because the
distribution of ground-based observations is spatially
limited. This is particularly true for altitudes between
about 30 and 180 km above the Earth's surface, which
encompass the region where the upward-propagating
waves attain large amplitudes and subsequently dissi- Figure 4 Contours of UARS nonmigrating diurnal northward
pate. It is generally impossible to distinguish migrating wind amplitudes (rn s -') at 95 km during October against longitude
from nonmigrating tides in the analysis of ground- and latitude.
Interferometer (HRDI) and the Wind Imaging Inter- structure is also distorted by the added presence of
ferometer (WINDII) instruments on UARS. In this the s = 0 and s = 3 oscillations. Latent heat release
depiction the migrating tidal component (amplitudes associated with raindrop formation in deep tropo-
as large as 50 m s - has been removed. The ampli- spheric clouds is a plausible source of these nonmi-
tudes in Figure 4 are residuals of order 10-20 m s - I, grating diurnal waves.
which add a substantial longitude dependence to the
diurnal tide. Spectral analysis of these particular data
reveals that the longitudinal structure is dominated by Lunar Tides
the westward-propagating s = -2 and the standing Lunar atmospheric tides are only about 5-10% as
s = 0 components, with some contribution from the
large as solar tides, but they have clearly detectable
eastward-propagating s = 3 diurnal tidal component. effects. The lunar tidal pressure at the ground maxi-
Complementary numerical modeling studies suggest
mizes at low latitudes, with an average amplitude of
that these waves are probably generated in the lower about 7 Pa. The corresponding wind amplitude at the
levels of the atmosphere.
Equator is about 0.03 m s-'. The wind amplitude
Spectral analyses of the Japanese Kyushu University increases with altitude up to about 110 km, where it
general circulation model (KUGCM) results reveal reaches an amplitude on the order 10 m s -
that diurnal and semidiurnal nonmigrating tides are Unlike the solar tides, the lunar atmospheric tides
primarily generated in the lower atmosphere and are entirely driven by gravitational forces, as illustrat-
propagate to higher levels. The KUGCM is one of the ed schematically in Figure 6. Because the lunar
few models of the atmosphere that realistically simu- gravitational acceleration decreases as the inverse
lates the physics and dynamics of the lower, middle, square of the distance from the center of the Moon,
and upper atmosphere. It includes realistic topography
this acceleration is not exactly uniform near the Earth,
and latent heating associated with tropospheric con-
so that atmospheric air parcels at various locations
vection, and the model domain extends into the
around the Earth experience slightly different lunar
dissipative thermosphere above 100 km. The accelerations from those of the Earth as a whole. Air
KUGCM meridional wind component associated
parcels in the hemisphere most distant from the Moon
with the westward-propagating s = -2 diurnal tide are accelerated toward the Moon less strongly than is
is illustrated in Figure 5. The wave amplitude exhibits
the Earth, in effect creating a relative acceleration
double maxima on either side of the Equator near
away from the Moon for these air parcels, in the
about f20". Double maxima can also be discerned
Earth's reference frame. Conversely, air parcels in the
near 120-180" and 300-360" longitude in the satellite
moonward hemisphere of the Earth experience a
measurements that are shown in Figure 4. These
relative acceleration toward the Moon. In each
maxima are nearly 180" apart, which is consistent
with a wavenumber 2 oscillation. However, this
Total lunar gravitational acceleration
140
120
-L
-
h
Y
E
E 100
.-a
m Acceleration relative to Earth
I
80
60
90"N 60"N 30"N 0"
Latitude
30"s 60"s
Figure 5 Latitude-height (km) contours of northward wind

90"s
7-3
amplitude (ms-') for the s = -2 westward-propagating non- Figure 6 Schematics of lunar tidal forcing in the reference
migrating diurnal tide from the Kyushu University general circula- frames of the Moon (top) and the Earth (bottom). The Earth is
tion model results. (Professor S.Miyahara and Dr. Y . Miyoshi viewed from above the North Pole and the Moon (not shown) is to
provided the model results.) the right.
164 ATMOSPHERIC TIDES
hemisphere, parcels to the west of a line passing position but also of geographical location, and gener-
through the centers of the Earth and Moon experience ate nonmigrating tides in addition to the primary
an eastward component of acceleration, while those to migrating lunar tides.
the east of this line experience a westward accelera- It is possible to develop a model of the M2 lunar
tion. As the Earth rotates, during a lunar day semidiurnal atmospheric tide that produces results
(24.8412 h on the average) an air parcel at the Equator that agree very satisfactorily with the observed tide in
successively passes twice through regions of westward the surface pressure. Such a model must account for all
and eastward acceleration, comprising two lunar direct and indirect lunar forcing effects, and include
semidiurnal cycles of period 12.4206h. When the realistic atmospheric wind and temperature struc-
Moon is north or south of the Earth’s Equator, an tures. Figure 7 illustrates prototypical M2 model
additional diurnal lunar cycle (period 24.8412 h) of results and shows how the lunar tidal amplitude varies
acceleration exists at nonequatorial latitudes. There is with latitude and longitude over the Earth for atmos-
also a monthly periodicity to the forcing as the Moon pheric conditions representing the month of Decem-
cycles between the Northern and Southern Hemi- ber. The largest amplitudes in this month are at low
spheres of the Earth. Both this cycling and the latitudes, but they vary from less than 4 Pa (40 dPa) on
ellipticity of the Moon’s orbit create amplitude and the east coast of South America to more than 12Pa
frequency modulation of the lunar semidiurnal and over the mid-Pacific. A secondary maximum appears
diurnal forcings that can be expressed as combinations over the northern Atlantic.
of multiple closely spaced periods. The dominant On the average, the amplitude of the M2 tide at low
lunar period, representing the average lunar semidi- latitudes is larger around the solstices than at the
urnal tide, is referred to as the M2 tide, with a equinoxes, by roughly 50%. Some of the largest
12.4206 hour period. geophysical effects of atmospheric lunar tides appear
In addition to the direct forcing of lunar gravity on in the low-latitude ionosphere. Like solar tidal winds,
the atmosphere, lunar atmospheric tides are indirectly lunar tidal winds generate a dynamo effect in the
forced by lunar gravity through deformation of the ionosphere by moving the electrically conducting
Earth’s surface due to ocean and Earth tides. The medium through the Earth’s magnetic field, creating
vertical velocity associated with this deformation an electromotive force with ensuing electric currents
significantly affects the atmospheric tide, and the and polarization electric fields. During the daytime,
modulation of terrestrial gravity by the deformation of when the conductivity is large owing to ionizing solar
the Earth’s mass distribution also has an effect. These radiation, the electric currents produce perturbations in
indirect forcing effects can be determined accurately the Earth’s magnetic field that are readily measured at
from measurements of ocean and Earth tides, so that the ground. Figure 8 shows the northward component
the total lunar tidal forcing is well known. This is of the magnetic perturbation at Huancayo, Peru, for
beneficial for testing theoretical models of tidal prop- two phases of the lunar tide: 5 and 12 days following
agation and dissipation in the atmosphere. The indi- new moon. These represent average conditions in
rect forcing effects, because they depend on land-sea 1957-58 for the months of November to February,
differences, are a function not only of apparent lunar when the lunar tide in the ionosphere is generally
Figure 7 Contours of M2 lunar semidiurnal surface pressure amplitude (dPa) against longitude and latitude for the month of December.
(Dr. F. Vial collaborated with Professor J. M.Forbes to produce the model results.)
2501 I I I I I I I I I I I I I I I I I I I I I I I 1 phere: Gravity Waves. Parameterization of Physical

Processes: Gravity Wave Fluxes. Radar: Meteor Radar.
Satellite Remote Sensing: Wind, Middle Atmosphere.
200 -
Further Reading
Chapman S and Bartels J (1940) Geomagnetism. Oxford:
Clarendon Press.
s- 150 -
h
Chapman S and Lindzen RS (1970) Atmospheric Tides.
Dordrecht: Reidel.
-
C
.-0
m
-E
Dai A and Wang J (1999) Diurnal and semidiurnal tides in
global surface pressure fields. Journal of the Atmospheric
r
a,
Sciences 6: 3874-3891.
a Hagan ME, Forbes JM and Vial F (1995) On modeling
.-0
100 - migrating solar tides. Geophysical Research Letters 22:
m 893-8 9 6.
z
E
Hays PB, Abreu VJ, Dobbs ME, et al. (1993) The High-
Resolution Doppler Imager on the Upper Atmosphere
c
r
Research Satellite. Journal of Geophysical Research 98:
- 10713-10723.
2 50
Haurwitz B and Cowley AD (1973) The diurnal and
semidiurnal oscillations, global distribution and
annual variation. Pure and Applied Geophysics 102:
193-222.
0 Matsushita S (1967) Solar quiet and lunar daily variation
fields. In: Matsushita S, Campbell WH (eds) Physics of
Geomagnetic Phenomena, pp. 3 0 1 4 2 7 . New York:
Academic Press.
Matsushita S (1967) Lunar tides in the ionosphere. In:
-50 1 1 1 1 1 " 1 1 1 1 1 1 1 1 1 1 ' 1 1 ' 1 1 Handbuch der Physik, pp. 547-602. Berlin: Springer-
0 6 12 18 24 Verlag.
Local time (h) Miyahara S, Miyoshi Y and Yamashita K (1999)Variations
of migrating and non-migrating tides simulated by the
Figure8 The northward component of magnetic perturbation (in middle atmosphere circulation model at Kyushu Univer-
nanoteslas)at Huancayo, Peru against local time (h) on the 5th and
sity. Advances in Space Research 24: 1549-1558.
12th days following new moon reveals evidence of lunar tidal
effects.
Reber CA, Trevathan CE, McNeal RJ and Luther MR
(1993) The Upper Atmosphere Research Satellite
(UARS) mission. Journal of Geophysical Research 98:
largest. The larger solar diurnal and semi-diurnal 10643-10647.
tides produce a northward perturbation that maxi- Shepherd GG, Thuillier G, Gault WA, et al. (1993)WINDII:
mizes daily a t around 11.00 local solar time (LT).On The Wind Imaging Interferometer on the Upper Atmos-
day 5 , the lunar tide enhances the magnetic perturba- phere Research Satellite. Journal of Geophysical Re-
tion a t 11.00 LT, but reduces it in the late afternoon. On search 98: 10725-10750.
Vial F and Forbes JM (1994) Monthly simulations of the
day 12 the phase of the lunar tide and its effects are
lunar semi-diurnal tide. Journal of Atmospheric and
reversed from those on day 5 . Clear lunar effects in Terrestrial Physics 56: 1591-1607.
the low-latitude ionospheric electron density are also Vincent RA, Kovalam S, Fritts DC and Isler JR (1998)Long-
found. term MF radar observations of solar tides in the low-
latitude mesosphere: Interannual variability and com-
See also parisons with the GSWM. Journal of Geophysical
Research 103: 8667-8683.
Buoyancy and Buoyancy Waves: Optical Observations; Volland H (1988)Atmosbheric Tidal and Planetary Waves.
Theory. Diurnal Cycle. Ionosphere. Middle Atmos- Dordrecht: Kluwer Academic.
166 AVIATION WEATHER HAZARDS
A J Bedard Jr, National Oceanic and Atmospheric

Administration, Boulder, CO, U S A
Copyright 2003 Elsevier Science Ltd. All Rights Reserved.
Introduction
A broad range of aviation weather hazards affect
operations from takeoff and landing to in-route flight
at high altitudes. A goal here is to provide an overview
of key meteorological processes disrupting flight,
reducing lift, increasing drag, influencing instrument
readings, or reducing visibility. These atmospheric
factors include gravity currents including thunder-
storm gust fronts and sea breeze fronts. Because the
Figure 1 Summary of the altitude impact ranges of meteorolog-
motions and decay of aircraft wake vortices are ical hazards and typical wind speed strengths involved.
controlled by local winds, turbulence, and stability,
these dangerous wake effects are also discussed.
Hazard types covered appear in Table 1, together and strengths are summarized. Also, discussions of
with brief descriptions of their potential impacts on efforts at hazard prediction, detection, and warning
flight. Figure 1 summarizes some of these meteoro- illustrate the progress that has been made in mitigating
logical disturbances, indicating typical flow strengths atmospheric impacts on the aviation system.
and altitudes affected. Encounters with turbulence
aloft can disrupt flight paths and cause injuries to
crews and passengers. Such strong encounters can
Atmospheric Gravity Currents
result from organized instabilities of limited duration One important type of atmospheric gravity current is
or extent, such as breaking gravityhhear waves. On the outflow from a distant thunderstorm downdraft.
the other hand, more random turbulence aloft, when For these systems, because the downdraft region is
of long duration and covering extended areas, can usually quite large in diameter (typically 10 km or
contribute to structural fatigue and reduce aircraft more), the wind shears near the downdraft can be
operating lifetimes. relatively weak. However, the wind speed and wind
For selected hazards, atmospheric causal processes direction change that accompany the leading edge can
are reviewed and key properties such as dimensions cause significant relative air speed changes for aircraft.
Table 1 A summary of aviation weather hazards and their areas of impact upon flight operations
~ ~~
Hazard Areas of impact
Atmospheric gravity current wind shears (e.g., thunderstorm Relative air speed changes
gust front and sea breeze front) Require runway changes
Microburst wind shears Flows can exceed performance capabilities of modern
aircraft
Vertical wind shear Deviations from glide slopes
Gravitykhear waves Flight disruption and structural fatigue
Icing Increased drag and reduced lift
Reduced stall angle
Flight disruption
Terrain-induced disturbances (e.g., lee waves, rotors, bora) Deviations from assigned flight altitudes
Structural damage
Flows can exceed performance capabilities of modern
aircraft
Vicinity of thunderstorms (e.g., hail, funnels, obstacle flows) Deviations from flight altitudes
Structural damage
Aircraft wake vortices transported to unexpected locations Roll moments disrupting flight of following aircraft
Altimeter errors Deviations from assigned or expected flight altitudes
AVIATION WEATHER HAZARDS 167
Sudden aircraft performance changes caused by the Table 2 Gust front statistics derived from surface meteorological
atmosphere (whether increasing or decreasing per- stations in the Denver, Colorado region during the summer of 1982.
formance) are problematical. Figure 2 is a conceptual Number of events = 99
view of a gust front from a distant thunderstorm Parameter Average Minimum Maximum
crossing an airport. Although at times these bounda- value value
ries are clearly visible because of entrained dust, more
often the boundary will occur invisibly in clear air at
Wind speed change 9.4 m s- ’ 3.0 17.0
Wind vector change 13.7ms-’ 5.0 37.0
large distances from the originating thunderstorm. Temperature change - 1.9”C -5.1 +3.1a
The speed of motion, c, of a gravity current with no Pressure change 0.67 hPa - 0.4a 3.3
ambient wind can be estimated from the density Rain rate 24 mm h 0 3.0
current equation [11. These anomalous readings of temperature increase and pressure
decrease resulted from the passage of a gust front on top of and
eroding a ground-based inversion.
Here, Fr is the Froude number (-1), AT is the vertical forcing of ambient air occurs. This is especially
temperature change in the gust front air relative to the true where boundaries collide. Microbursts also have a
environmental air, T i s the mean temperature, g is local tendency to occur near outflow boundary regions.
gravity, and h is the height of the outflow boundary. Fortunately, Doppler radar can detect these gust front
Corrections for the ambient wind can be made. Table 2 boundaries effectively and provide an invaluable, all-
summarizes statistics for gust fronts measured in the weather resource to guide airport operations when
Denver, Colorado, region during an intensive field gravity currents are approaching.
program during the summer of 1982. Data from this
and other experiments demonstrate that the density
current equation applies quite well. The leading edge is
usually accompanied by a temperature drop and
Microbursts
pressure rise, unless complicated by the existence of Between 1964 and 1985, over 30 commercial aircraft
a ground based inversion layer. Figure 3 shows an crashes resulted from microbursts. A microburst is
acoustic sounder display of a gust front propagating defined as a downdraft region with a scale size less
on top of a ground-based inversion. than 4 km. The resulting strong outflows usually do
Another important aspect of thunderstorm outflow not travel radially outward for long distances
boundaries is that a preferred region for the initiation (>10km), and the durations are short (often less
of new convection is near the leading edges where than 10 min). However, in the vicinity of a microburst,
Figure 2 Conceptual view of a gust front from a distant thunderstorm crossing an airport.
Figure 3 Acoustic sounder detection of a gust front propagatingon a ground-basedinversion,
strong winds (>50ms-') and rapid wind direction said that a secret service officer jumped on top of the
changes of 180" can occur. When microbursts descend president to protect him as the winds buffeted the
near or on runways, they constitute an extreme flight aircraft.
hazard. Microburst flows are analogous to those There are two extreme types of microbursts: 'dry'
produced when squirting a water hose on a flat and 'wet'. Dry microbursts are especially hazardous
surface. The downflow jet interacts strongly with the because the visible virga (raindrops or a snow plume
surface, producing strong radially directed flows. The descending from cloud base as in the photograph in
large spatially concentrated horizontal wind vector Figure 4) related to the microburst initiation process
changes and the downdraft can produce increasing evaporates and becomes invisible as it approaches the
performance/decreasing performance couplets that surface. Since the downdraft descent time takes about
are difficult to predict and handle. For example, an 5 min, it can be difficult to relate an observation of virga
aircraft flying through a microburst that has impacted to a resulting microburst. Conversely, a wet microburst
the approach end of a runway will first encounter a has a strong rain shaft and is easily seen if not obscured
head wind, increasing performance and causing ex- by rain from a surrounding storm (Figure 5 ) .
cursions above the glide slope. As the pilot corrects for Doppler radar can detect microbursts once the
this, the aircraft enters the downdraft region, followed radial outflow is established by the intense downdraft
by an outflow region, rapidly degrading performance. penetrating to the surface. Also, the concentrated rain
Depending upon the timing and relative positions of shafts for wet microbursts can be detected. The
the aircraft and microburst to the runway, this
scenario can be catastrophic. The timing is so critical
that even landing differences of several minutes can be
important. The statistics of microbursts measured in Table 3 Impacting microburst statistics derived from surface
the vicinity of Denver, Colorado, are presented in meteorologicalstations in the Denver, Colorado region during the
summer of 1982.Number of events = 33
Table 3. Microbursts were identified by the winds
clearly radiating outward from a center, as distinct Parameter Average Minimum Maximum
from the essentially linear gust front winds. To date, value value
the largest wind speed documented for a microburst Wind speed change 13.5ms-' 2.5 27.5
was that related to a 'near miss' of Air Force One with Wind vectorchange 20.7ms-' 10.0 37.5
President Reagan on board when it was on the ground Temperature change - 1.5"C - 9.0 +5'
near Washington, DC, on 1August 1983 (a wind speed Pressurechange 0.66 hPa - 1.5' 2.0
surge over 60m s- I). The microburst occurred five Dew point change - 7°C + 7°C
Rain rate 16.4mm h - ' 0 2.75
minutes after the plane landed. Newspaper accounts
Figure 4 Photograph showing virga descending from cloud base.
uses of arrays of airport wind sensors, Doppler radars, wind speed, vertical changes in wind speed and
and improved controller/pilot training (both to recog- direction can also present a hazard, especially for
nize visual clues and to respond in the best possible way lower-level flight operations. A ground-based inver-
if a microburst is encountered) have helped to reduce sion is often accompanied by calm winds near the
microburst-related accidents. Also, the fact that the surface and strong winds just above the cooler, stable
lapse rate between 500 and 700 hPa is correlated with near-surface air. Aircraft descending or ascending
microburst probability provides forecasting potential through such layers can encounter strong wind
for dry microburst likelihood. Dry microbursts are shear-produced performance changes and turbulence,
more probable when the lapse rate is > 8"C/km. and rapid fluctuations associated with gravityhhear
waves. These waves have scale sizes from tens to
Vertical Wind Shear and GravityKhear hundreds of meters, resulting in aircraft interaction
times of seconds or less. Figure 6 is a conceptual view
Waves of such a situation in the vicinity of mountains. Figure
Whereas the thunderstorm gust front and microburst 7 is a Doppler lidar display of gravityhhear waves.
hazards result primarily from horizontal changes in Vertical wind shear conditions can be especially
Figure 5 Photograph of a 'wet' downdraft

.-
+-
3
2
7
I jet
/wind
Height of maximum
speed
-
Turbulent air
/-
motion
s
7-
-
At what rates do such stable pools of air get
removed?
What are some examples of situations where these
processes are important for flight operations?
The Froude number, Fr, is a measure of the relative

importance of inertial and gravity forces. For the
Wind speed
* situation of flow above a stable layer, Fr is given by
eqn [2], where U is the flow speed above the inversion,
Figure 6 Conceptual view of gravitykhear waves in the vicinity g is local gravity, h is the height of the inversion, dp is
of mountains. the density difference between the two layers, and p is
the mean density.
U
important for general aviation airports if relative Fr =
airspeed is suddenly reduced on a low-level approach
or during takeoff in the vicinity of terrain. At airports
(dp f) 'I2
where wind shear above stable air is a frequent
problem, boundary layer wind profilers or acoustic In field and laboratory experiments, the start of
sounders can provide valuable real-time monitoring disturbances and waves takes place when the Froude
capabilities. number exceeds about 0.6. Thus, if the height of the
The presence of gravityhhear wave activity often inversion and a temperature profile are available, the
complicates flight through layers of vertical wind threshold speed U can be estimated. Once the erosion
shear. A pure shear layer in a neutrally buoyant process starts, it can often continue at a slow and
atmosphere may be modeled as a vortex sheet, highly approximately constant rate. Values of vertical ero-
unstable to disturbances. If wind shear occurs in sion rates near 10 cm s-' have been measured near
conjunction with a stable layer, gravity provides a complex terrain. O n the positive side, the time scales
restoring force, and such a system will support wave are of the order of hours for changes, in contrast with
motion. Hence, the term gravity shear wave. Several minutes for microbursts. Thus, vertical wind profilers
questions naturally follow from this situation. for monitoring and knowledge of local climatologies
can be quite valuable, particularly for mountain
Under what conditions will the upper-level wind valleys and the lee sides of complex terrain.
shear start to erode the stable air and turbulence Another important dimensionless number is the
grow? Richardson number, Ri,which is an important index
Figure 7 Doppler lidar display of gravitykhear waves in the vicinity of mountains, showing the accompanying wind speed changes. The
numbers below the color bar are the radial wind speeds in meters per second. The numbers above the color bar are the distance in
kilometers from the lidar. (Courtesy L. Darby, NOAA.)
for turbulence. This number depends upon the gradi- about 5 min. Such discontinuities can represent sig-
ents of both temperature and wind speed (eqn [ 3 ] ) . nificant hazards, since they are unexpected weather
Ri = $/ (=)
dU
[31
features producing wind vector changes on scales that
can affect flight, especially during takeoff and landing.
In eqn [ 3 ] ,g is local gravity, 0 is the mean potential Icing

temperature, d0/dZ is the change in potential temper- Icing potential depends upon the probability of drops
ature with height, and dU/dZ is the change of wind of supercooled liquid impacting aircraft surfaces.
speed with height. The criterion for turbulence, Larger droplets are more likely to strike an airfoil,
Ri<1/4, has been shown to be a valuable index for since they do not easily follow the flow streamlines and
aircraft turbulence when temperature and wind speed pass around an obstacle as do smaller droplets. The
profiles are available to make the comparison between icing hazard can be insidious because of two factors:
predicted turbulent altitudes and actual turbulence
reports from pilots. A challenge is either to predict the 1. Only a small amount of ice deposition can have
temperature and wind speed gradients or to measure large, deleterious effects upon lift and drag, thus
them with sufficient accuracy to produce reasonable reducing aircraft performance.
estimates of Ri values. Wind profiling radars used in 2. Icing and the degradation of performance can
conjunction with the Radio Acoustic Sounding System increase slowly and imperceptibly, until an emer-
(RASS)to obtain temperature profiles will be valuable gency exists.
for such applications. Pilot reports as well as the use of
visual clues are also an invaluable component for Two situations account for most reported in-flight
avoiding regions of turbulence aloft. Figure 8 depicts icing encounters. Convective activity involves rela-
gravityhhear waves that are revealed by cloud forma- tively large amounts of supercooled liquid water, and
tions. the hazard can extend to higher altitudes with the
Large-amplitude, long-lived gravity waves repre- potential of significant ice accumulation in a short
sent another aspect of the hazard. Such waves have time.
been observed to propagate rapidly (at 35-40 rn s- ') The other frequent situation involves flight through
away from the region of a cyclone where they were layered cloud decks. Although the supercooled liquid
generated. The waves in this case traveled through content can be lower than in convective situations,
eastern New England in the United States. The high- aircraft typically spend more time in these layered
speed waves were accompanied by precipitation and clouds, accumulating significant ice on fuselage and
wind surges. For example, at Boston the wind control surfaces. Figure 9 depicts situations contrib-
increased from less than 10 knots to 5 7 knots over uting to this form of icing, which is especially
Figure 8 Photograph showing gravitykhear waves visualized by cloud formations

Figure 9 A summary of situations contributing to aircraft icing events. SLW stands for supercooled liquid water.
important for commuter and general aviation aircraft freezing rainstorms. This problem is treated at major
operating at lower altitudes. The wind shears and airports by application of de-icing fluids before takeoff
turbulence often accompanying such systems can be (see Aircraft Icing).
regions where larger drops are concentrated by mix-
ing. Sometimes, by making small altitude changes, a
plane can avoid these regions of enhanced icing
potential.
Terrain-induced Turbulence
Combinations of remote sensors have value for the Terrain effects can occur at all flight levels, with some
monitoring and forecasting of icing situations. Polari- disturbances affecting the stratosphere. The flow
metric radars, dual-wavelength radars, wind profilers, situations can range from lee waves, bora flows
RASS, and passive radiometric measurements can, in (a form of density current), and rotors, to mechani-
concert, indicate the locations of supercooled liquid. cally induced turbulence. Figure 10 is a Doppler lidar
Aircraft can also accumulate icing on the ground in display showing the roll-up of a vortex sheet in the
Figure 10 Doppler lidar display showing the rollup of avortex sheet in the lee of a mountain range. The numbers below the color bar are
the radial wind speeds in meters per second. The numbers above the color bar are the distance in kilometers from the lidar. (Courtesy L.
Darby, NOAA.)
Colorado Springs area. At times, organized instabil- 150% higher for a selected group of mountain airports
ities can occur in the forms of vertical or horizontal relative to a group of nonmountain airports.
axis vortices. These obstacle-involved situations can Table 4 indicates a pattern of sporadic encounters of
be exceptionally complex when the terrain flows aircraft with severe or extreme turbulence in the
interact with other meteorological factors (such as vicinity of mountains. In many cases, aircraft preced-
lee-side inversions). Since 1964, there have been 15 ing or following the aircraft involved in the event
major accidents and incidents in the vicinity of encountered some turbulence, but not the extreme
complex terrain (Table 4). turbulence of the encounters (which often exceeded
One study indicated that the general aviation structural limits). Thus, the regions of severe or
accident rate was 40% higher for US mountain states extreme turbulence may, at times, be spatially con-
than for all other continental states, and the rate was centrated and short-lived. This makes predicting the
Table 4 Turbulence-relatedaccidents and incidents occurring in the vicinity of mountains
Event Date Location Comment
Accident 31 Mar. 1993 Anchorage, Alaska Turbulence, 747 lost engine

Accident 22 Dec. 1992 West of Denver, Colorado Loss of wing section and tail assembly, 2-engine cargo
plane, lee waves present
Accident 9 Dec. 1992 West of Denver, Colorado DC-8 cargo plane, loss of engine, lee waves present
Unknown cause accident 3 Mar. 1991 Colorado Springs, Colorado 737 crash
Accident 12 Apr. 1990 Vacroy Island, Norway DC-6 Crash
Severe turbulence 24 Mar. 1988 Cimarron, New Mexico 767, 1.7G, Mountain wave
Severe turbulence 22 Jan. 1985 Over Greenland 747, +2.7G
Severe turbulence 24 Jan. 1984 West of Boulder, Colorado Saberliner, +0.4G to - 0.4G
Severe turbulence 16 Jul. 1982 Norton, Wyoming DC-10, + 1.6G to - 0.6G
Severe turbulence 3 Nov. 1975 Calgary, Canada DC-10, +1.6G
Accident 2 Dec. 1968 Pedro Bay, AK Fairchild F27B, wind rotor suspected
Accident 6 Aug. 1966 Falls City, Nebraska BAC 111, wind rotor suspected
Accident 5 Mar. 1966 Near Mt. Fuji, Japan BOAC 707, wind rotor suspected
Accident 1 Mar. 1964 Near Lake Tahoe, Utah Paradise Air Constellation, strong lee wave
Accident 10 Jan. 1964 East of Sangre de Cristo B52, wind rotor suspected
mountains in New Mexico
-
45000
40000 -
::
6
1813-1820 (GMT)
150
200
35000 -
4i:
Horizontal speed (m s-’)
._._.-._, ._._. ._.-._._._._.-._._ -

W-E
:+w
1828-1 834 (GMT)
0 30000 - 300 3
a
a a
cv f.
g, 25000 - 1730-1737 (GMT)
p!
.-
a
400 2
v)
I
20000 - 401 p!
23 --- 1849-1857 (GMT) 500
15000 miles 600
I n 0 10 20 30
0 10 20 30 4050 1700
1
km
5000 Boulder BAO 850
Figure 11 Horizontal wind speed changes encountered by a research aircraft in a lee wave situation on 25 January 1984. Data from
flight legs at four altitudes are shown.
time and location of these events more difficult. There guided the execution of field programs studying
is a great need to define the properties of mountain- mountain flows, which require the commitment of
related hazards, improve short-term forecasting of considerable scientific and measurement resources.
these events, improve pilot training resources, and Figure 11 shows an example of the changes in flows
develop detection methods. encountered by a research aircraft traversing a lee
Major field experiments have addressed this prob- wave in the Rocky Mountain region of the US. Similar
lem, which requires three-dimensional sampling of strong changes were also encountered in temperature,
large volumes of the atmosphere as a function of time, pressure, and vertical wind speed, with great differ-
documenting both surface and near-surface effects, as ences as a function of horizontal distance and flight
well as upper tropospheric and stratospheric effects. level. Doppler lidars have the ability to detect these
Both physical scale models and numerical models have hazards near the surface in clear air. Figure 12
Figure 12 Dopplerlidar display showingrotor circulationassociatedwith lee waves. The numbersbelowthe color bar are the radialwind
speeds in meters per second. The numbers above the color bar are the distance in kilometersfrom the lidar. (Courtesy L. Darby, N O M . )
is a Doppler lidar display illustrating the com- winds or turbulence), these vortices move downward
plexity possible for such upper-level mountain and than outward, diverging in a predictable way as
flows. they approach the surface of the Earth.
However, with light wind speeds that are close to the
transport speeds of the vortices, complex vortex
Thunderstorms trajectories can result. Figure 13, from numerical
simulations, shows how vortex transport can be
The most extensive study of thunderstorms was the
modified by ambient winds and for some situations
Thunderstorm Project. This study included surface
can remain stationary over runways. Such motions
meteorological stations, soundings, radar, and numer-
have been documented in field experiments. In addi-
ous aircraft flights. This study not only treated the
tion, atmospheric stability also affects transport,
larger-scale thunderstorm structure, circulation, and
perhaps causing the vortex pair to remain at a higher
surface effects, but also properties important for flight.
altitude than expected. The level of atmospheric
These thunderstorm effects influencing aircraft in-
turbulence is an important factor in controlling vortex
cluded turbulence, hydrometeors, downdrafts, and
decay times. Although generated by aircraft rather
updrafts, providing statistics for these thunderstorm
than being a true atmospheric meteorological phe-
parameters. The report also discussed the thunder-
nomenon, it is the variability of the atmosphere that
storm structure as detected by radar and geographical/
magnifies the hazard potential.
topographical effects on thunderstorm development.
Concern about the wake vortex hazard to following
This report by Byers and Braham is an important
aircraft is a limiting factor for the most efficient use of
resource addressing all aspects of thunderstorms (see
airport operations. This is because safety is currently
Further Reading).
ensured through the use of increased spacing between
Radar guidance permits aircraft to navigate
heavy aircraft and following aircraft. In spite of this,
gaps between growing thunderstorm cells, and it is
between 1983 and April 2002 there were nine docu-
presumed here that pilots will avoid penetrating
mented accidents with wake vortices as the cause.
such violent severe weather directly. On the other
Although most of these accidents involved general
hand, the clear air around cells or near benign-looking
aviation aircraft encountering the wakes of heavy
flattened anvil clouds often appears safe for flight,
commercial aircraft during approach or departure,
but these regions should be approached with caution
some involved in-route wake upsets between two
if they cannot be avoided. Table 5 outlines these
heavy aircraft. For example, on 2 September 1999 a
regions near thunderstorms and some of the potential
Boeing 737 out of Santa Barbara, California flew
hazards.
through severe turbulence from the wake of an
MD-11, with 15 injuries. This occurred in spite of a
separation of 11 miles. In the future, the use of re-
Aircraft Wake Vortices mote sensing technologies to track the positions
The strong vortex pairs behind heavy aircraft flying at of vortices could permit more effective use of
low speeds result from the lift being generated. The runways. Also, boundary layer wind profilers and
tangential winds near the cores can be greater than RASS could provide detailed real-time wind and
(measured at over 130ms-' for a Boeing temperature data as inputs to vortex transport predic-
757).In an ideal atmosphere (neutrally stable, with no tion models.
Table 5 Atmospheric hazards near thunderstorms
Potentially Hazardous Region Comment
1. Clear air regions to the lee of thunderstorms acting as obstacles Aircraft upsets have been caused by lee waves around
to ambient winds thunderstorms
2. Funnels may occur in the vicinities of anvil regions There have been funnel-related aircraft accidents
3. Near or in the tops of weak-looking convective cells when there These situations can strengthen regions of concentrated
is wind shear present vorticity
4. Lightning strikes in and near thunderstorms Frequent occurrence; damage can affect flight performance
and cockpit instruments. Composite material airframe
structures are at greater risk
5. The clear air above thunderstorm tops is a location for sprites, The hazard potential is not known.This may become a factor for
elves, blue jets and other electromagnetic phenomena aircraft manufactured from composite materials
9 0 0 ' -A0 ' -A0 ' -io ' -40 ' ; ' io ' i o ' $0 ' i o ' 1bo
Distance from center of flight path (m)
Figure 13 Computations of aircraft wake vortex transport affected by atmospheric winds
Altimeter Errors Also, the surfaces of constant pressure could be highly

perturbed (e.g., by strong lee wave conditions or in the
There are several critical ways in which altimeter
vicinities of frontal boundaries). Atmospheric winds
errors can affect the aviation system. Pilots may be at
around buildings housing altimeters can cause signif-
any of three distinct altitudes:
icant differences between true static atmospheric
0 Where they are assigned pressure and a measurement contaminated by dynam-
0 Where they think they are ic pressure effects. For example, winds between 10 and
0 Where they are 25 m s- can induce dynamic pressure errors as large
as about 2 hPa at an airport barometer (equivalent to
Several factors can contribute to these three places not altitude errors between 3 and 20 m). The development
being the same. There could be errors in calibrating of new static pressure probe designs and improved
aircraft altimeters. There could be malfunctions of calibration procedures is helping with these problems,
aircraft instruments (e.g., clogged pressure ports). which can be critical for mountain flying situations.
45000 - 150
1813-1820 (GMT)
40000 - 200
Pressure altitude (feet) W-E
35000
32 787 1828-1 834 (GMT)
0 30000
- 300 iij.
0 - 28 450 a
m G
c
v
-- .. 27 888 1730-1 737 (GMT) 2
.-
25000
- 27 325 - 400 1
I" 20000
t
15000 116188l
n 0
miles
10 20 30
4 600
10000 0 1020 30 4050
5000 850
Figure 14 Altitude changes encountered by a research aircraft traversing a lee wave on 25 January 1984.Data from flight legs at four
altitudes are shown. These measurements are for the same flight shown in Figure 11.
Table 6 Remaining areas of special need Acknowledgments

Improved knowledge of terrain-induced disturbances and the I am grateful to L. Darby of the Optical Remote
development of mitigation techniques Sensing Division of the Environmental Technology
Improved warnings for ‘free air’ turbulence, especially better
Laboratory of the National Oceanic and Atmospheric
knowledge of risks in the vicinity of thunderstorms
Earlier detection techniques for dry microbursts and increased Administration for making available the Doppler lidar
understanding concerning their initiation processes images used in Figures 7,lO and 12. I am also grateful
Improved warnings of icing situations for the valuable suggestions of an anonymous review-
Better predictions of the transport and decay of aircraft wake er. G. Salottolo of the US National Transportation
vortices
Broader application of site-specific remote sensing technol-
Board provided information about aircraft accidents
ogies for hazard prediction and monitoring applications attributed to wake vortices.
In the future, deviations from assigned flight alti-

tudes in the vicinity of pressure surface perturbations See also
will become more of a concern. As air traffic continues
Aircraft Icing. Gust Fronts.Lee Waves and Mountain
to increase, so will the motivation to allow smaller
Waves. Lidar: Doppler. Microbursts. Middle Atmos-
vertical separations of flight altitudes. There is a need phere: Gravity Waves. Radar: Doppler Radar. Turbu-
to document the statistics of these excursions as well as lence and Mixing.
the conditions under which they occur. The significant
altitude changes involved with an aircraft traversing a
lee wave on 25 January 1984 provide an example of the
importance of such encounters (Figure 14). Further Reading
Byers HR and Braham RR Jr (1949) The Thunderstorm:
Report on the Thunderstorm Project. Washington, DC:
Concluding Remarks United State Printing Office.
The early literature of aeronautical meteorology Bosart LF, Bracken WE and Seimon A (1998) A study of
contains vivid accounts of the frightening experiences cyclone mesoscale structure with emphasis on a large-
amplitude inertia-gravity wave. Monthly Weather Re-
of balloonists within thunderstorms. Fortunately, the
view 126: 1497-1527.
evolution of modern flight has made modern aircraft Fujita TTand Caracera F (1977)An analysis of three weather
largely immune to such experiences, but they are related aircraft accidents. Bulletin of the American
susceptible to other hazards, such as microbursts and Meteorological Society 58: 1164-1181.
wake vortices (in part because of our technological Fujita TT (1985) The Downburst. Chicago: The University
advances). On the other hand, modern lighter-than-air of Chicago Press.
vehicles, both tethered and untethered, can find Gregg WR (1930) Aeronautical Meteorology. New York:
themselves similarly challenged because of limited The Ronald Press.
agility. The future uses of lighter-than-air craft may be Horne TA (1999)Flying America’s Weather.Newcastle,WA:
expanded as our ability to make reliable long-range Aviation Supplies and Academics.
forecasts improves. Table 6 outlines challenges we still Lenschow DH (ed.) (1986) Probing the Atmospheric
Boundary Layer. Boston, MA: American Meteorological
face in continuing to mitigate aviation weather haz-
Society.
ards. The application of remote sensing technologies, Scorer RS (1978)Clouds ofthe World. Melbourne: Lothian
better knowledge of atmospheric dynamics, improved Publishing Company.
numerical models, and accurate longer-term forecasts Scorer RS (1978)Environmental Aerodynamics. Chichester:
will be critical to success. Ellis Horword.
BAROCLINIC INSTABILITY 179
R Grotjahn, University of California, Davis, CA, USA ature gradients at rigid bottom ( z = 0) and top
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. ( z = Z T ) boundaries. In the Q G system, PV can be
written as:
Introduction 4 = 02$+fo
v
+ Py
RV
Baroclinic instability refers to a process by which
perturbations draw energy from the mean flow
potential energy. The conversions of energy are
proportional to perturbation heat fluxes in both
horizontal and vertical directions. The atmosphere 1v
- BPV
requires heat fluxes to maintain the observed pattern
of net radiation (positive in the tropics, negative where
poleward of 38"N or S on an annual average). A zonal
mean meridional circulation, such as a tropical Hadley
cell, can generate these heat fluxes. However, in middle
latitudes, various factors cause eddies to accomplish
and
the bulk of the heat transport. Baroclinic instability
provides a mechanism to explain how these eddies
form and evolve while incorporating the necessary
heat fluxes. Theoretical models of baroclinic instabil-
ity can simulate various observed properties of where t jis the horizontal velocity streamfunction, p is
midlatitude eddies, including the dominant length density, g is the acceleration due to gravity, and IC is the
scales, propagation speed, vertical structure, and static stability from O s , the horizontal mean potential
energetics. temperature. The coordinates are x eastward, y
Baroclinic instability can be viewed as a shear northward, and z upward. Nondimensional length
instability. From thermal wind balance, the vertical scales are L in the horizontal and D in the vertical.
shear of the zonal wind is proportional to a meridional fo is the constant part while is the meridional
temperature gradient. The meridional temperature derivative (approximated as a constant) of the Coriolis
gradient is proportional t o the available potential parameter.
energy (APE) that the baroclinic instability mecha- An inherent horizontal length scale is the Rossby
nism taps. Another view of baroclinic instability radius of deformation (LR = N H f t l ) where N is the
emphasizes interacting potential vorticity (PV) anom- Brunt Vaisala frequency ( N 2 = glcD-l) and H =
alies. Baroclinic instability is usually studied by RTg-' is the scale height (an inherent vertical length
linearizing the dynamics equations and using scale). Thus, E = ( L H ) 2 ( L ~ D ) -relates
2 the assumed
eigenvalue or initial value techniques. These alterna- scales L and D to LR and H .
tive views and analysis procedures generally provide Quasi-geostrophic PV includes three distinct parts:
complementary means to understand better baroclinic absolute vorticity, which includes relative vorticity
instability. (RV); 'thermal' vorticity (TV); and boundary PV
(BPV). Positive PV is associated with an interior
trough (in geopotential) and/or a warm surface tem-
An Illustrative Model perature anomaly.
When the vorticity and potential temperature
An illustrative model provides mathematical rela-
conservation equations are combined, one obtains
tions and archetype solutions for the concepts that
a time-dependent equation for QGPV conser-
follow.
vation:
Mathematical Formulation
The model uses quasi-geostrophic (QG) approxima-
tions and nondimensional scaling appropriate for
midlatitude frontal cyclones. Potential vorticity
has contributions from the interior and from temper-
180 BAROCLINIC INSTABILITY
with boundary conditions at the bottom and top
-0
at z = O,ZT [2bI
‘Basic state’ variables are specified: U (independent of
x ) is zonal wind, and Q is the interior part of the
QGPV; meridional and vertical velocities are zero.
One can solve eqn [2] as an initial value problem by
specifying an initial streamfunction or potential
vorticity.
An eigenvalue problem can also be formulated from
eqn [2]. A common approach assumes time and space The volume integrals are over a closed domain and
dependence in the form: mass weighted by ps, the horizontal mean density. In
the QG system, a+/& is proportional to potential
+ ( x , Y,z , t) = Re {@(Y, z ) ~ X [ik(x
P - ct)]} [31
temperature 8, making the first term on the right-hand
for the ‘perturbation’ streamfunction being sought. side of eqn [sa] proportional to a meridional heat flux,
This solution has zonal wavenumber k and complex while the second term is proportional to a vertical heat
phase speed c. The growth rate is given by k Im { c}. If flux. The specified vertical shear, aU/az, is propor-
U has no meridional variation, then one can assume a tional to the available potential energy, A, of the basic
wavelike y dependence too: exp(i2y). When wave- state and is the energy upon which the baroclinic
number 1 = k, the solution is a ‘square wave’. Pertur- instability mechanism feeds. The first term on the
bation velocities are defined as u = -a+/ay and right-hand side of eqn [Sb] is a barotropic energy
v = -a+/ax. conversion. The barotropic conversion is proportional
Additional simplifying approximations are often to the divergence of eddy momentum flux and draws
made. A particularly simple form, commonly labelled energy from the mean flow kinetic energy K,. The
the ‘Eady model’, was described by E. T. Eady second term on the right-hand side of eqns [sa] and
in 1949. The Eady model assumes wavelike meridio- [5b] is the same, but with opposite sign indicating a
nal structure, a Q / a y = 0, U = z, incompressibility conversion between A, and K,.
( p = constant), and E = 1. Then eqn [2a] is reduced
simply to solving q’ = 0 in the interior where the prime Example Solutions
denotes the ‘perturbation’ sought. The Eady eigenval-
This QG eigenmodel of baroclinic instability is applic-
ue problem can be solved analytically to yield a pair of
able to the midlatitudes. In these regions zonal flow
normal modes, one growing, and one decaying, for
increases with height reaching a maximum near the
scaled wavenumber a< N 2.4. The scaled wave-
tropopause. Figure 1A is a representative nondimen-
number:
sional profile of U where the tropopause is at nondi-
CI = { (k2 + 22)E-’}1’2 [41 mensional z = 1.0. The growth rate and phase speed
spectra, along with the (growing normal mode)
is proportional to absolute wavenumber and static eigenfunction structures for different k are also shown
stability. in Figure 1. The growth rate has maximum value at a
Equations for perturbation kinetic energy, K , and specific value of a. The vertical structure tends to have
available potential energy, A, are: relative maxima at the surface and near the tropo-
pause, but it becomes progressively more bottom-
trapped for shorter waves. The phase varies such that
unstable modes tilt upstream with height, i.e., against
the mean flow shear. Other solutions to eqn [2],
labeled continuum modes, are relevant to nonmodal
growth.
For shorter waves the lower maximum tends to
dominate (when compressibility is included) and the
solution decays rapidly away from the boundary. For
longer waves the tropopause level maximum tends to
dominate (Figure 1D). Eady model normal modes
BAROCLlNlC INSTABILITY 181
~l,/-J,
-0 1.o
, , , , . ,
2.0
, , , , , , , , , , , , , , ,
3.0
I i
O,:
0 1 .o 2.0 3.0
(6) 1 (C) 1
1= 3.0
2.0 2.0 L I I
1.6
1.2
ry
0.8
0.4
0.0
-2.0 -1.0 0 1.0 2.0 -2.0 -1.0 0 1.0 2.0 -2.0 -1.0 0 1.0 2.0
(D) (E) (F)
Figure 1 Quasigeostrophic eigenanalysis. (A) Specified zonal wind U , and meridional gradient of interior potential vorticity Qou versus
scaled height. z = 1 is 10 km. (B) Growth rateand(C) phasespeedversusabsolutewavenumberCI. (D)-(F)AmplitudeA, andphaseP,for
the growing normal mode for a = 0.8, a = 2.0, and CI = 3.0, respectively. All three modes tilt westward (upstream) with increasing height.
Dimensionalwavelengths depend upon scaling assumptions, but reasonable choices imply that CI = 0.8, a = 2.0, and CI = 3.0 correspond
to 11.O, 4.4, and 2 . 9 lo3
~ km wavelengths, respectively. (Zonal and meridional scales are set equal.) The same scaling implies phase
-
speed of 9 m s - ’ and doubling time of 1.2 days for CI = 2.0. (Adapted with permission from Grotjahn R (1980) Linearized tropopause
dynamics and cyclone development. Journal of Atmospheric Sciences 37: 2396-2406.)
have interior q’ = 0; from eqn [I]: the Laplacian Classical View

increases as k 2 , requiring a rapid change with height
for short waves to make the thermal term comparable Baroclinic instability draws upon the APE of the
(this leads to boundary trapping of the solutions). For environment in which an eddy sits. Since APE is related
longer waves, the Laplacian becomes small and the to a horizontal temperature gradient, which is in turn
vertical structure is more evenly spread in the vertical, related to the vertical shear, it can be viewed as a type
hence these modes are ‘deeper’. of shear instability. One advantage of doing so is to
Typical geopotential patterns observed prior to make comparisons with barotropic instability, which
frontal cyclone development have separate surface draws energy from the horizontal shear. This view
and upper troposphere troughs, each equivalent provides a link to the eddy fluxes that are observed and
barotropic (vertical trough axis), with the upper- necessary for each conversion.
level trough more prominent. A crude simulation As demonstrated in eqn [5],heat fluxes are neces-
of that initial state is used to generate solutions sary to have a baroclinic energy conversion. Horizon-
shown in Figure 2. Time-series of the growth rates tal heat fluxes imply that the temperature and mass
of several quantities are tracked over several days. (here $) fields are offset. The offset implies that the
The time-series include potential enstrophy ( H = trough and ridge axes tilt upstream with elevation.
+
(q’)2)and total energy (TE = A, K,) integrated The QG formulation above is adiabatic, so individ-
over the whole domain. The solutions approach ual parcels conserve their potential temperature (e)
asymptotically to the most unstable normal-mode over time. For unstable modes, the horizontal and
growth rate as that eigenmode emerges to dominate vertical eddy heat fluxes must distort the 8 field over
the solution. The growth rate has transient peak time, as suggested schematically in Figure 3. An
values that can exceed the asymptotic (normal-mode) isentropic (e) surface is drawn in three-dimensional
value. perspective; it curves up and over colder areas and dips
182 BAROCLlNlC INSTABILITY
2'o r 3.5I
'
.......
3.0 -I\
~ ...........
............
. .
. . . . . .
; ,' ;' ,,,.' , .
-1
I
1.o j ; ( / !"
.
,
. ... .,.,...'..... . .
I
..........
),
. ..........
.....
2.5 - \
-1 \
b
2.0
1.5 I \
'\
1.0 I I 1.o
0.5
"." I I I
0.0 2.5 5.0 7.5 10.0 0.0 2.5 5.0 7.5 10.0
(C) Time (B) Time
Figure 2 Initial value calculation. (A) Zonal cross-section of initial streamfunction, dashed contours used for negative values. (B) Time-
series of growth rates, u, for domain average potential enstrophy (solid line) and its components: RV2(short dashed line), TV'(dot-dashed
line), and BPV2 (long dashed line). Growth rates approach asymptotically the most unstable normal mode rate for this wavenumber
( x = 2.0). (C) Similar to (B) except for total energy (solid line), kinetic energy (short dashed line), and available potential energy (dot-
dashed line). For reasonable scale choices, z is scaled by 10 km x by 700 km, and time by 1.2 days.
down over warmer areas. Prior to eddy development,

the isentropic surface did not vary in the x direction
and had a shape like its intersection with the wall at
x = 0. The isentropic surface is distorted by flow
around the high- and low-pressure centers and repre-
0 sentative cold (C) and warm (W)trajectories are also
drawn. When these trajectories are projected onto the
x = 0 wall, they appear to cross the initial zonal mean
isentrope and have a slope that is typically half the
slope of the mean isentrope. In fact, they are changing
the zonal mean of the isentrope to become more
Y-+ horizontal, thereby reducing the horizontal tempera-
Figure 3 Schematic diagram showing distortion of an isentropic ture gradient and thus reducing A,. In this classical
surface by a baroclinically amplifying frontal cyclone. Dotted lines view, A, is reduced while A, is increased by increasing
are used for objects underneath the three-dimensional isentropic the zonal undulations of the isentropic surface.
surface. Surface high, H, and low, L, are marked together with two Another aspect is that colder air is sinking while
representative contours of surface pressure. Trajectories of
representative parcels are shown in warm air, W, and cold air, C.
warmer air is rising, a process that lowers the center of
Subscript s denotes projection onto the bottom surface, while z mass and thus converts A, into K,. To lower the center
denotes projection onto the meridional plane (where x = 0). The of mass, the parcel paths must have the vertical
trajectories do not cross the isentropic surface but distort it. Initially component indicated, but they must also be less than
the isentropicsurface had negligiblevariationwith xand looked like the slope of the mean isentropes for instability to
the current pattern at x = 0. The projections W, and C, seem to
cross the initial isentropic surface but in fact are flattening it (which
occur.
reduces A,). Rising air is warm while sinking air is cold, which The classical view can incorporate latent heat
lowers the center of mass, converting A, into Ke. release as follows. The bulk of the precipitation in a
developing cyclone forms in the warm-air sector of the constructed for a PVanomaly associated with a surface
storm. The release of latent heat further depresses the temperature gradient.
isentropic surfaces where there is poleward motion, Potential vorticity anomalies are created by flow
implying additional conversion of A, into A, and K,. across PV contours. Figure 5 illustrates how two
sinusoidal PV anomalies can amplify each other. The
PV gradient is reversed between the two levels,
Potential Vorticity View increasing with y at upper levels and decreasing with
The potential vorticity view of instability tracks how y at the surface. This pattern is consistent with upper
two or more PVanomalies interact in a way that causes tropospheric W a n d the surface temperature gradient,
growth of the PV anomalies. Potential vorticity is a respectively. (Recall that q is positive for lower
fundamental conserved quantity for adiabatic mo- geopotential height or warmer surface temperature. j
tions. The illustrative model is designed around QGPV The associated winds cause propagation by distorting
conservation. the PV pattern. However, the meridional wind asso-
A PV pattern has an associated streamfunction and ciated with a PV anomaly is in quadrature with that
horizontal wind field. In general, eqn [ 11 implies that anomaly so the PV cannot amplify itself. Growth is
PV emphasizes smaller-scale variations than the described simply as advection at the PV extrema that
streamfunction field. Inverting eqn [ 13 obtains broad further amplifies the PV pattern. Since the associated
patterns of I)associated with isolated packets of q. An winds extend beyond the elevation of the PV anomaly,
illustrative example is shown in Figure 4 for PV there can be interaction with a second PV anomaly at
anomalies in the upper troposphere. The associated another level. When the second PV anomaly is offset
winds are displaced from the PV anomaly center by $ from the first, as in Figure 5A, the associated winds
wavelength ( - 1000 kmj. A similar diagram can be amplify the first anomaly.
100
-
.......
400 .-
h
a"
c
v
P
700 -
-
1000 -
(A)
100
400
P
i i
I I
700
0.004
1000
0 15 30 45 60 90 0 15 30 45 60
(C) x (") (Longitude) (D) x (") (Longitude)
Figure 4 Quasi-geostrophic potential vorticity inversion on a sphere. Zonal cross-sections are shown over 90' longitude. Potential
vorticity anomalies are specified and centered at 45' N; (A) and (B) use dotted lines to indicate PVcontours 0.5 and - 0.5 nondirnensional
unitsforwavenumbers4and8, respectively.Contoursarestreamfunction,$, derivedfrorn invertingthe PV relationof eqn [l].Lows ($ <O)
correspond with PV > 0 maxima, and vice versa. (C) and (D) show the meridional winds associated with $ in (A) and (B), respectively.
Dashedcontours denote negativevalues. The contour interval in (A) and (6) is 0.005; in (C) and (D) it is 0.002. Longer PV anomalies have
deeper and stronger strearnfunctions.
\ \ \
(A) (B)
Figure 5 Baroclinic instability from interacting PV anomalies at two levels. A representative PV contour (dot-dashed line) is drawn at
each level. Theoffset is (A)awavelength and (B)Jwavelength. Atypical wavelength might be4 x lo3km. Each anomaly has an associated
wind component parallel to the PV gradient; dashed arrows are winds from the lower PV anomaly, while solid arrows are from the upper
anomaly. The winds from each anomaly advect the associated anomaly. In (A) each PVanomaly has a wind component that amplifies the
each PV anomaly has a wind component that augments the propagation in
undulation in the other anomaly, thereby causing growth. In (6)
the manner indicated by the broad arrows; this causes the anomalies to migrate to a phase difference similar to diagram (A).
This mechanism also explains how developing A necessary condition for instability is that a
cyclones maintain a preferred tilt (i.e., become ‘phase steering level, where U = Re{c}, lies within the
locked’). The lower anomaly is shifted horizontally domain. A supportive kinematic argument is that air
to the right in Figure 5B so that upper and lower parcels remain with the system (rather than blow
anomalies are 180 degrees out of phase. The two through it or be left behind) and are more easily mixed
PV anomalies no longer amplify each other’s PV laterally. For really long waves, strong retrogressive
anomalies (shutting off the instability mechanism). motion caused by the p term leads to a different class
Furthermore, the two anomalies reinforce the of unstable eigenmodes for CI<W 1.1 (note cusp in
velocities midway between their positive and Figure 1B) than for larger CL
negative extremes, thereby enhancing the propa-
gation at each level; but the propagation is in
opposite directions at each level, thereby reducing Normal Modes
the phase shift to reestablish the pattern in Normal modes are physically meaningful eigenfunc-
Figure 5A. As with the classical view, normal tions. As in the illustrative model, the equations
modes are a special case where this phase locking are linearized about a specified basic state and
is optimized. perturbation solutions are sought. Most commonly,
The PV view provides theoretical weight to a classic the time and one or more space dependencies are
description of how cyclones develop: an upper-level assumed. By assuming a form like that in eqn [ 3 ] ,
trough (PV anomaly) approaches a low-level baro- unstable solutions grow exponentially. Models that
clinic zone (another PV anomaly), then growth com- are simple enough may be solved analytically.
mences. This paradigm is commonly labeled ‘type B’ More commonly, the eigenvalue problem is solved
cyclogenesis. numerically.
Observations show independent troughs at the Normal modes are consistent with many observed
upper and lower troposphere prior to development features:
with the upper approaching the lower. Neither trough
has upstream tilt initially; such tilt emerges only after 1. Unstable modes tend to be lined up along the jet
the two become favorably aligned and growth has axis (if present) in the mean flow.
commenced. 2. The most unstable wavelength is similar to the
A necessary condition for instability is that the observed median size. The normal-mode scale can
across-flow mean gradient of PV changes sign within be manipulated by varying the choices made for
the domain. In the illustrative model above, p > 0 nondimensional parameters, but is on the order of
meaning Q y > 0 in the interior, and the surface 4500 km.
temperature gradient (dO/dy < 0) implies Q y< 0 at 3 . Solutions tend to develop similar zonal and merid-
the ground. In the Eady model Qr = 0 everywhere in ional lengths, the latter responding to the width of
the interior, so the normal-mode instability comes the jet that provides one natural scale in the model.
from BPV having opposite sign at top and bottom Other properties (such as static stability) also
boundaries. influence the length scales.
4. The vertical structure of the most unstable modes cult. The vorticity equation also illustrates instability
tends to have relative maxima at the surface and by which the divergence term has positive vorticity
upper troposphere. tendency at a trough where vorticity is a maximum,
5. In growing normal modes the temperature lags the thus amplifying the peak vorticity (and vice versa for
mass field (typically by 20-50 degrees of phase for ridges).
the most unstable mode). Two consequences are: In addition to the normal modes, the eigenfunctions
0 First, the mass field must tilt upstream with include a class of solutions called ‘continuum’ modes.
height. There is typically to wavelength For an adiabatic model continuum modes have
(1-2 x l o 3 km) between the trough location at equivalent barotropic structure (no tilt), making
the surface and at tropopause level. them neutral. In the Eady model, continuum modes
0 Second, the lag allows across-flow heat fluxes have zero PV at all levels except at the critical level,
down the temperature gradient, as expected from where their amplitude has a ‘kink’. Continuum modes
eqn [Sa]. In the Eady model the heat flux is play a role in nonmodal growth.
uniform with height. Model improvements, most
notably compressiblity, can emphasize the eddy
heat flux in the lower troposphere (where obser
vations find it most prominent).
Nonmodal Growth
-
6 . The rate of propagation is 10-2Oms-’: slower
than jet stream level winds, but faster than (zonal
Nonmodal growth is seen when solving initial value
problems such as eqn [2]. The formulation can be
average) surface winds. The steering level is defined
linear, as in eqn [2], or nonlinear. This approach is
as the level where the propagation speed of the more general than eigenanalysis, since the time
storm is equal to the wind component along the dependence is not assumed.
storm’s track. The steering level for the most The solution at any time can be decomposed into a
unstable normal modes is typically between 700 combination of eigenfunctions. For an arbitrary initial
and 500 hPa, depending on the assumptions made. state, continuum and normal modes are present. These
For shorter waves, the steering level is closer to the
modes move at differing speeds. In a linear formula-
surface, and these modes move more slowly. tion the modes operate independently; as modes
Longer waves respond to competing effects: they disperse, positive and negative reinforcement varies.
have greater upper-level amplitude (where U is The interference between modes decays algebraically
faster) but greater sensitivity to (which enhances asymptotically. However, for some initial conditions it
retrograde motion). is possible to have sizable growth over a limited time
7. The rate of growth is similar to but slower than that
period.
of observed cyclones. Observed doubling times are For the Eady model, analytic solutions can be found
typically 1-2 days at upper levels.
-
that illustrate the process. Using an initial condition
with upstream tilt ( 4 exp (imz) in eqn [ 3 ] , where
8. Instability is inversely proportional to static stabil-
m>O) yields solutions with normal mode and
ity. For example, the peak growth rate depends on c(
( =2.0 in Figure 1). From eqns [lb] and [4], CI isalgebraic parts. The algebraic part has time depen-
dence proportional to { ( m - k t ) 2 + ci2}-’
proportional to static stability K. Hence, smaller K and
places the most unstable peak at larger k, making exp {i(m - k t ) z } . The amplitude increases as the tilt
the growth rate (k Im {c}) larger. Kinematically, becomes more vertical until t = mk-’.After that, the
vertical motion needed in eqn [5] becomes easier wave tilts downstream and decays.
for smaller IC. Initial upstream tilt becoming more vertical with
time has led to an expectation that RV increases at the
The fact that normal modes have fixed tilt is not expense of TV while interior PV remains conserved.
necessarily unrealistic. Observations of the vorticity However, exceptions can be found where large non-
equation terms support an approximately fixed struc- modal growth occurs (in H ) as upstream tilt ‘develops’
ture for a developing low because the divergence term from an initial state with no tilt. The explanation lies in
opposes the horizontal advection at upper levels but a rough cancellation between RV and TV, leaving the
reinforces the horizontal advection at low levels. The BPV evolution to dictate rapid growth in H .
normal modes (Figure 1)are special structures where A robust interpretation of nonmodal growth is
the net advection is exactly uniform throughout the progressively more favorable superposition of constit-
depth of the fluid. Tracking observed troughs over uent modes. Continuum modes, having mainly upper-
time shows some evidence for a period of fixed tilt level amplitude, tend to move quickly, while modes
during growth, though the 1 2 h interval between with mainly lower-level amplitude move slowly. De-
upper air observations makes the measurement diffi- composition into eigenmodes of an initial state with
upstream tilt finds faster continuum modes located between the two instabilities. The most unstable
upstream of slower continuum modes. Over time, the baroclinic eigenmode has optimal structure for a
modes become more favorably lined up; the tilt flow having only vertical shear, but when horizontal
becomes more vertical and the total amplitude in- shear is added to that flow a different structure is
creases. Figure 6 illustrates the process. needed or the eddy will be sheared apart. The
Nonmodal growth can be quite strong in simple subsequent structure is unlikely to be as optimal for
models like Eady’s. However, most improvements to baroclinic energy conversion. Hence, the baroclinic
the model such as adding compressibility, variable conversion will usually be reduced, though the baro-
Coriolis, and realistic vertical shear of U reduce tropic growth mechanism may compensate. Figure 7C
nonmodal growth. Using more realistic initial states illustrates such a calculation; in this case adding a
also tends to reduce nonmodal growth (e.g., using a purely barotropic flow reduced the growth rate,
wave packet instead of a wavetrain; using separate even though the barotropic growth mechanism was
untilted upper and lower features instead of connect- activated.
ing them with a tilt). Baroclinically unstable frontal cyclones prefer to
develop in certain regions. The preference may arise
from local conditions, such as lower static stability or
locally greater vertical shear. The illustrative model
Other Issues above assumes a wavetrain solution; when more
Baroclinic instability has links with barotropic insta- localized development is considered, a variety of issues
bility. First, each instability draws energy from mean are raised.
flow shear. Second, barotropic instability has a similar For example, if one uses a single low as the initial
stability criterion (absolute vorticity gradient chang- condition, the solution typically evolves into a chain of
ing sign in the domain). Third, there is interference waves as the modal constituents of the initial state
Time=O Tirne=l Tirne=n
0,251
0.00
I l Il l 1 1 0.00
0.25 u
0.50
0.25
0.00
0.0 1.6 3.1 4.7 6.2 0.0 1.6 3.1 4.7 6.2 0.0 1.6 3.1 4.7 6.2
1.oo 1.oo
0.75 0.75
0.50 0.50
0.25 0.25
0.00 0.00
0.0 1.6 3.1 4.7 6.2 0.0 1.6 3.1 4.7 6.2
1 .oo
0.75 -
0.50 0.50
::A
0.25
0.00 0.00 -.--
0.0 1.6 3.1 4.7 6.2 0.0 1.6 3.1 4.7 6.2 0.0 1.6 3.1 4.7 6.2
1.oo 1.oo
0.75 0.75
0.50 0.50
0.25 0.25
n nn
_.”_ n on
0.0 1.6 3.1 4.7 6.2 0.0 1.6 3.1 4.7 6.2
x-+
(C)
Figure 6 Nonmodal growth as a superposition process. Four initial value linear calculations are shown. The top three rows show three
individual neutral continuum modes at three times. The bottom row shows the sum of the three modes at the given time. (A) Initial condition;
(B)time when energy growth is a maximum in the sum; (C) time when growth rate is zero in the sum. Reasonable scales are 700 km in x
and 10 km in z. (Adapted with permission from Grotjahn R, Pedersen R and Tribbia J (1995) Linear instability with Ekman and interior
friction. Part II: Initial value analysis. Journal of Atmospheric Sciences 36, 764-777).
1.o
ru ru
0.0 0.0
-1.1 0 1.1 -1.1 0 1.1
(A) Y
1.o
ru
0.0
-1.1 0 1.1 -1.1 0 1.1
(D) Y
Figure 7 baroclinic energy conversion (A, - Ae)for four models. (A) Lowest-order, square wave solution for an Eady-type model but
includingcompressibility,increasingvertical shear in U , /3 = 1. (B) Solutionwhen asurface frontal zone, centered at Y = 0, is added to the
lowest-order mean flow Uo and leading ageostrophic advective effects are included (using geostrophic coordinates). The frontal zone
adds wind field: 0.2(2z - z‘) Ul where UI = bl(1 - tanh‘ ( a Y )- 9 - 3b3 Y‘ to UOwhere a = 1.6. bl = 3.2. 9 = 1.1, and b3 = 0.03.
Y is scaled by 1100 km. The geostrophic coordinate transform causes the asymmetry. (C) Correction to the conversion shown in (A)
+
when bsrotropically unstable horizontal shear U1 is added to Uo.If the total wind is U = UO p U l , then the total conversion is ( a )+ p(c).
The barotropic shear reduces the growth rate. (D) Modification due to all leading order ageostrophic corrections. If those corrections
are order p , then the total conversion is ( a ) + p ( d ) . Ageostrophic conversions reduce the conversion and introduce asymmetry.
(Adapted with permission from Grotjahn R (1979) Cyclone development along weak thermal fronts. Journal of Atmospheric Sciences
36,2049-2074.)
disperse. Alternatively, a wave packet initial condition Studies of regional development spawned subcate-
might be used consisting of a ‘carrier wave’ multiplied gories of baroclinic instability. ‘Absolute’ instability
by an amplitude envelope. The packet evolution occurs when the wave packet expands faster than it
depends upon the mean flow properties and assump- propagates; the amplitude at a point keeps growing.
tions made in the model. However, for reasonable ‘Convective’ (in the advection sense) instability occurs
choices of parameters, one might find a packet that when the packet moves fast enough so that growth
spreads while propagating downwind. The leading then decay occurs as the packet moves past a point.
edge of the packet has mainly faster, wider, and deeper ‘Global’ instability (like the eigensolutions shown
modes. The trailing edge has slower, shorter, and here) has growth that is invariant to a Galilean
shallower waves. It is possible to construct a localized transform. Such is not the case for ‘locally’ unstable
structure that resists this dispersion by making a modes. Normal modes for zonally varying basic states
judicious combination of eigenmodes having similar look like carrier waves modulated by a spatially fixed
phase speed, but different zonal wavenumber. Figure 8 amplitude envelope; the envelope locally modifies the
illustrates such an example using neutral continuum growth rate (sometimes called ‘temporal’ instability).
modes. When this model is solved as an initial value ‘Spatial’ instability allows wavenumber to be complex
problem the packet maintains a localized shape for a while phase speed remains real.
long time. Almost no growth occurs because the Nonlinear calculations raise other issues related to
normal modes were filtered out, and there is very slow baroclinic instability. One issue concerns equilibra-
phase shifting of the constituent modes. However, tion. The growing wave modifies the mean flow while
when nonlinear advection is allowed, modes interact drawing energy from it. This places a limit upon the
and soon amplitude is injected into all the eigenmodes, cyclone development. In PV theory, this may be where
including the growing normal modes, which grow the distortion shown in Figure 5 becomes comparable
rapidly in this example. to the cyclone width. Waves longer than the most
2.0 1.o
1.8 0.9 -
1.6 0.8 -
1.4 0.7 -
0.6 -
/.-
1.2 ‘E /’
lu 1.o 0.5 -
0.8 0.4 ;
0.6 0.3 -
0.4 0.2 - /’
0.2 -..- ./
...........
0.0 .......“... ...... ..............
L
0 1.6 3.1 4.7 6.3 7.9 9.4 0.0 4.7 9.4

0
’ 2.5 5.0 7.5 10.0 12.5 15.0
X X
(C) Time
Figure 8 Initial value calculations for a linearly localized initial condition. (A) Zonal cross-section showing contours of streamfunction
initially. Values < - 1.Oare shaded. (B)Horizontal pattern of streamfunction at tropopause level ( z = 1.O) initially. The initial condition is
constructed from neutral modes having similar phase speed. Growing or decaying normal modes are excluded. (C) Time-series of energy
growth rate, ‘E, for three integrations. The linear model (dotted line) shows little growth since the nonmodal mechanism is weak and
growing normal modes cannot develop. Also shown are nonlinear calculations for two amplitudes of the initial condition, where the solid
line uses three times the initial amplitude of the dot-dashed line. Growing normal modes are activated by nonlinear interaction. Some
evidence of nonlinear saturation is seen. Reasonable scales are 700 km in x and y , 10 km in z, and 1.2 days for time.
unstable wave tend to reach a larger amplitude than wave. Ageostrophy also causes contours t o be more
the linearly most unstable mode. One reason for this is closely spaced around a low and more widely spaced
that they are deeper and so can potentially tap more around a high.
APE in the mean flow. Another reason may be that the
larger scale in both horizontal dimensions provides a
longer time for PV contour distortion. Another pos- See also
sibility concerns the inversion of a PVanomaly: Figure 4 Cyclogenesis. Cyclones, Extra Tropical. Dynamic
uses the same magnitude of PV anomaly, but the Meteorology: Balanced Flows; Overview; Waves.
streamfunction amplitude is larger for the broader PV Fronts. Quasi-geostrophic Theory. Vorticity.
anomaly.
‘Life cycle’ studies model cyclones from birth to
peak amplitude to decay. These studies typically find Further Reading
baroclinic growth followed by barotropic decay. This
cycle fits the observed facts that eddies have a net heat Gill A (1982) Atmosphere-Ocean Dynamics. New York:
Academic Press.
flux and a net momentum convergence. These studies
Grotjahn R (1979) Cyclone development along weak
also reveal a characteristic evolution of the eddy thermal fronts. Journal of Atmospheric Sciences 36:
structure: upper-level amplification compared t o the 2049-2074.
linear eigenmodes. An explanation is that saturation is Grotjahn R (1993) Global Atmospheric Circulations: Ob-
reached sooner at the critical level and at the surface servations and Theories. New York: Oxford University
while upper levels continue to grow. When averaged Press.
over the life cycle, the vertical distribution of the zonal Holton J (1992)A n Introduction t o Dynamic Meteorology,
mean eddy heat and momentum fluxes becomes more 3rd edn. San Diego, CA: Academic Press.
realistic. Hoskins B, McIntyre M and Robertson A (1985)On the use
Finally, the atmosphere has higher-order processes and significance of isentropic potential vorticity maps.
Quarterly Journal of the Royal Meteorological Society
than the QG system. The biggest impact of ageostro-
111: 877-946.
phy is t o break symmetries in the solutions. Figure 7D James I (1994) Introduction t o Circulating Atmospheres.
shows the leading order ageostrophic effects for a Cambridge: Cambridge University Press.
linear model. Ageostrophy causes enhanced eddy Pedlosky J (1987) Geophysical Fluid Dynamics, 2nd edn.
development on the poleward side (mainly by negative New York: Springer-Verlag.
baroclinic conversion on the equatorward side), builds Pierrehumbert R and Swanson K (1995)Baroclinic instabil-
mean flow meridional shear, and slows down the ity. Annual Review of Fluid Mechanics 27: 419467.
BEAUFORT WIND SCALE 189
L Hasse, Universitat Kiel, Kiel, Germany the sea, wind effects on rigging, whistling of the wind,
Copyright 2003 Elsevier Science Ltd. All Rights Resewed
and other phenomena may have helped to pass on a
tradition from experienced observers to younger
colleagues, though not coded in words. Ship officers
Introduction in the transition time likely had sufficient training on
sail ships to estimate the wind force even on board
For ships at sea it has been widespread practice to steamers.
include weather information into ship’s logbooks to In 1927 a description of the Beaufort scale in terms
document the situation during its operations. For sail of sea state was formulated by Petersen as a result of
ships, wind information is most important. Beaufort many years of experience (Table 1). His description
adopted a scale to estimate the ‘force of the wind’ in was added later to the wind code by the International
1805 when he was commanding officer of a man of Meteorological Organization (IMO),the predecessor
war. The scale was formulated in terms of the effect of of the World Meteorological Organization (WMO).
wind on sail ships of a certain type, but was Obviously, IMO did not view reference to sea state as a
subsequently used for other sail ships and steamers redefinition, but rather as a written account of an
too. existing practice. It is common habit to call the
Beaufort’s scale of wind force was devised for use in redefined scale the Beaufort scale of wind force too.
the marine environment. However, since 1874 the The use of Beaufort numbers for coding purposes
Beaufort scale has been used in the international had initiated early investigations into wind speed
telegram code to transmit wind information from both equivalents to Beaufort numbers, often called a
sea and land. While Beaufort had used the behavior of ‘Beaufort equivalent scale’. Measured wind speeds
sail ships in a given wind to define for what conditions presumably could ensure consistent use of Beaufort
common language terms like ‘gentle breeze’, ‘moder- numbers over land and would help to alleviate the
ate gale’, ‘whole gale’, or ‘storm’ should be applied, a difficulties resulting from change of ship types with
different definition was required for land surfaces. time. Also, estimated wind force numbers do not really
Observers used certain indicators; e.g., behavior of fit into the concepts of theoretical meteorology, where
flags, trees or drag plates and feeling of wind in the wind velocity is used in the basic Navier-Stokes
face. equations.
With today’s knowledge of boundary layer meteor- From these introductory remarks, we see that there
ology, attempts to estimate the wind force over land are at least two tasks:
appear questionable. One would need to use indica-
tors of a known drag coefficient at a prescribed height 1. selection of a suitable relation between the force of
in an open, level area without obstructions to the flow.
wind and the wind velocity, and
Even then correction for the roughness of the under-
2 . correction for inhomogeneities in the time series of
lying terrain and for stability (e.g., dayhight) would be
winds at sea.
required in order to make estimates comparable.
Obviously, anemometer operation on land is a more
direct method to determine wind speed instead of Relation between Beaufort Force and
estimation. Use of the Beaufort scale to determine Wind Velocity: A Problem of Physics
wind speed on land is not recommended and will not
and Regression Interpretation
be discussed in the following.
Anemometer measurements on ships are difficult A sail ship is not a well-defined tool to measure wind
due to flow distortion. Also, only relative wind speed speed. The same is true for the appearance of the sea
and direction are measured. True wind needs to be surface that depends to some degree on the history
calculated by vector addition of the independently both of the wind field and the wave field and even on
measured course and speed of a ship, a source of turbulence in the atmospheric boundary layer over the
additional error variance. Estimates of Beaufort force sea. In retrospect it appears wise that a coarse scale
have therefore remained a tool for an extended period only was devised for a measure of wind force. As we
of time. can expect only a statistical relation, we need to define
With improved types of rigging and transition from requirements on such a scale.
sail ships to steamers, the original definition of Following Lindau, an optimal Beaufort equivalent
Beaufort forces was endangered. The appearance of scale is required to convert Beaufort estimates into
190 BEAUFORTWIND SCALE
Table 1 Definition of Beaufort force in terms of sea state
Beaufort Common name Definition

number
0 Calm Sea like a mirror

1 Light air Ripples with the appearance of scales are formed, but without foam crests
2 Light breeze Small wavelets, still short but more pronounced; crests have a glassy appearance and do not break
3 Gentle breeze Large wavelet; crests begin to break; foam of glossy appearance; perhaps scattered white horses
4 Moderate breeze Small waves, becoming longer; fairly frequent white horses
5 Fresh breeze Moderate waves, taking a more pronounced long form; many white horses are formed (chance of
some spray)
6 Strong breeze Large waves begin to form; the white foam crests are more extensive everywhere (probably some
spray)
7 Near gale Sea heaps up and white foam from breaking waves begins to be blown in streaks along the direction
of the wind
8 Gale Moderately high waves of greater length; edges of crests begin to break into the spin drift; the foam
is blown in well-marked streaks along the direction of the wind
9 Strong gale High waves; dense streaks of foam along the direction of the wind; crests of waves begin to topple,
tumble, and roll over; spray may affect visibility
10 Storm Very high waves with long overhanging crests; the resulting foam, in great patches, is blown in
dense white steaks along the direction of the wind; on the whole, the surface of the sea takes a
white appearance; the tumbling of the sea becomes heavy and shock-like; visibility affected
11 Violent storm Exceptionally high waves (small- and medium-sized ships might be fora time lost toview behind the
waves); the sea is completelycovered with long white patches of foam lying along the direction of
the wind; everywhere the edges of the wave crests are blown into froth; visibility affected
12 Hurricane The air is filled with foam and spray; sea completely white with driving spray; visibility very seriously
affected
Reproduced with permission from World Meteorological Organization (1970) The Beaufort Scale of Wind Force. Reports on Marine
Science Affairs No. 3. Geneva: World Meteorological Organization.
wind speeds such that derived climatological quanti- where typically means of the dependent variable are
ties like means and variances agree with respective calculated for intervals of the independent variable.
quantities from unbiased wind measurements. The physical problem of one-sided regressions is
The attempts to derive a Beaufort equivalent scale seen in the following. If we determine an equivalent
differ in the meteorological setup of the experiment scale from data including error variances in both
and in the subsequent statistical interpretation. Diffi- variables, the spread of the independent variable is
culties exist in both parts. It appears that misunder- increased by its random errors and the resulting
standings in the interpretation of statistical methods regression line has too small a slope. Applying this
have hampered the development of and agreement on regression to climatological variables (e.g., monthly
a Beaufort equivalent scale more than questions of mean wind speed and variance), the derived variances
measurements and exposure. will deviate systematically from respective quantities
determined from unbiased measured winds.
In the development of Beaufort equivalent scales,
One-sided Regressions
two approaches have been used:
Derivations of Beaufort equivalent scales typically use
regression techniques. For a pair of variables (x, y), 1. We take the Beaufort number as well determined
two regression curves can be derived, depending on and calculate the mean velocities for each interval
which of the variables are considered ‘independent’. A of Beaufort number; i.e., we use a nonlinear
common notion of regression sounds like: ‘The regression of wind velocity on Beaufort number.
regression of y on x gives the best estimate (by the 2. We argue that the average wind velocity is well
method of least squares) of y for a given independent x, measured and the random variations stem from
and, similarly, the regression x on y gives the best imperfect estimation of Beaufort numbers. In this
estimate of x as a function of independent y’. The one- case the regression of Beaufort force on wind
sided regression implies the ‘independent’ variable as velocity is the given choice.
nonrandom and attributes all random variation to the
other, as can be seen if one follows the derivation, e.g., The pitfall is seen in the following: in one-sided
for a linear fit. The same is true in nonlinear regression, regression the independent variables are treated as
BEAUFORTWIND SCALE 191
nonrandom and all variability from both axes is

ascribed to the dependent variable.
The first approach was applied in early investiga-
tions listing wind speeds averaged for each Beaufort
number, although difficulties both of Beaufort obser-
vations and of the exposure and calibration of wind
instruments were discussed.
However, there are good arguments to use the
second approach, the regression of estimates to
measured speeds. We know that Beaufort numbers
are estimates only, which are influenced by different
external conditions and by uncertainties in the ob-
server’s judgment. Also the quantization error is larger
for estimates due to the larger intervals of the Beaufort
scale compared to speeds in meters per second. It is not
unreasonable to infer the error of Beaufort number
estimation to be much larger than the uncertainty of 51 I I I I I I I I I
anemometer measurements. Koppen in 1888 already 2 3 4 5 6 7
advocated the averaging of Beaufort estimates for Beaufort
intervals of measured speeds in order to establish an Figure 1 Relation between Beaufort force and wind speed (kn),
equivalent scale. based on observations from OWS K and ships within 500 km from
IMO adopted a scale for official use, corresponding OWS K, period 1960-71. One-sided and ‘orthogonal’ linear
to today’s W M O code 1100, that was established in regression (thin lines) are compared with monthly means. The
1906 by Simpson, Meteorological Office, London, linear fit of monthly means (full line) deviates from the best linear
regression calling for explicit consideration of error variances.
using regression of Beaufort estimates t o measured (Reproduced with permission from Lindau R (1994) €he neue
speeds as recommended by Koppen. Beaufort-Aquiva/entska/a.Kiel: Berichte lnstitut fur Meereskunde
Kiel, No. 249.)
Two-sided Regressions
It is evident that both Beaufort force estimates and that the uncertainty of Beaufort estimates is a larger
velocity measurements contain random errors, and a fraction of the total variance than that of anemometer
two-way regression would be the appropriate tool. measurement.
The best relation between two random variables lies The orthogonal regression yields a better Beaufort
between the two one-sided regressions. For linear equivalent scale than any of the one-sided regressions -
regressions, in the absence of information on the in the sense to better reproduce the means of wind
respective error variances the best choice would be the speed from Beaufort estimates. The agreement is even
bisector of the angle between the two one-sided better than one would expect from error estimates of
regression lines. It minimizes the orthogonal distances wind speed measurements and Beaufort estimates.
of observed points from the line. This is equivalent to The reason for this is seen in the natural variability
assuming the error variances to be the same fraction of between measurements co-located in space and time
total variance for each of the variables. This so-called that enters as ‘unexplained’ variance in the regression
orthogonal regression is certainly better than a one- too.
sided regression, when both variables are subject to
random errors. However, in a given case, the fraction
An Optimized Beaufort EquivalentScale
of error in the total variance of each variable need not
be the same and an improved technique is required. Wind speed measurements at Ocean Weather Stations
For illustration consider the observations at the (OWSs) and Beaufort estimates from passing volun-
Ocean Weather Station (OWS)K in Figure 1. The two tary observing ships (VOSs) were used in several
one-sided regression lines and the orthogonal regres- attempts to derive an improved equivalent scale. Some
sion line are plotted together with the monthly means averaging is needed to reduce errors of measurements
of anemometer measurements at OWS K and Beaufort in respective observations and also to account for the
estimates from ships in the vicinity. The fit through the natural variability that enters because VOSs pass
monthly means deviates from the orthogonal regres- OWSs at some distance.
sion in the direction towards the one-sided regression Lindau developed a sophisticated method to
of Beaufort numbers on measured speeds. This implies determine effective variances. From simultaneous
192 BEAUFORT WIND SCALE
observations of pairs of VOSs, differences were 35 I I I I I I I I I I
obtained and variances calculated as a function of + + + + Code1100 I
distance between the ships. Extrapolating to fictitious 30 -OOOOOOO CMM-IV

zero distance, the error variances of VOSs were
obtained. Using the same technique on pairs of
OWSs and VOSs, the error variances of OWSs were h 25
also determined. 2
;20
Knowing the variances, influence of errors and
Q
natural variability could be reduced by averaging in n
time for OWSs and in space for VOSs. A suitable area 15
around an OWS was selected to contain the same i5
variance in space from VOS than at the OWS in time 10
for 1 day. This radius and the appropriate number of
VOS observations were determined separately for
5
each OWSandeachseason.
With the error variance at OWS and VOS reduced
by averaging the appropriate number of observations, 0
0 1 2 3 4 5 6 7 8 9 1 0 1 1 1 2
and the natural variability being assured to be the Beaufort
same, the ‘orthogonal’ regression yields the correct
relation. Instead of straight lines, Lindau used the Figure 2 Different Beaufort equivalent scales: code 1100 (10 m
method of cumulative frequencies in order to admit reference height), crosses; CMM-IV (25 m height), circles; Kaufeld
nonlinearities. Anemometer heights at weather ships (25 m height), dashes; Lindau (25 m height), dots. Code 1100
are near 25m. Wind measurements at OWSs were determined by regression of Beaufort on wind speed, CMM-IV by
reduced to 10 m height using a diabatic wind profile. wind speed on Beaufort, Kaufeld and Lindau from cumulative
frequencies.
The resulting equivalent scale is thus applicable to
1 0 m height. The results are given in Table 2 and
depicted in Figure 2.
Code 1100 in use from 1926 to 1946 was believed to
Discussion of Heightsfor Equivalent Wind Velocities
refer to 6 m height. Since the required standard height
Beaufort estimates per se have no natural height above of anemometers had changed from 6 to 10 m, a slightly
the sea surface. However, equivalent scales give wind changed code 1100was introduced in 1946 by WMO,
speeds. Because of the approximately logarithmic supposedly applicable to 10 m height. Both versions of
wind profile in the atmospheric boundary layer, for code 1100 were based on data sets into which
applications of the Beaufort equivalents the corre- observations from the Scilly Islands entered in a
sponding height needs to be known. relatively large number. However, wind speeds were
Table 2 Selection of Beaufort equivalent scales, given in m s ~ ’. ‘

Often knots (1 kn = 0.5144 m s- ’; 1 m s- = 1.944 kn) are used as the
unit. The last line gives the reduction factor to compare equivalent scalesfor reference heights of 25 and 10 m, derived for the ensemble of
measurements at North Atlantic OWSs
Beaufod 0 1 2 3 4 5 6 7 8 9 10 11 12 Source
a
IMO(1926)6m 0.0 1.1 2.5 4.3 6.3 8.6 11.1 13.8 16.7 19.9 23.3 27.1
Code1100 10m - 0.8 2.4 4.3 6.7 9.4 12.3 15.5 18.9 22.6 26.4 30.5 -34.8 a
Lindau 10 m 0.0 1.2 2.7 4.6 7.2 9.7 12.1 14.6 17.3 20.2 23.4 27.1 31.4
CMM-IV 18 m 0.8 2.0 3.6 5.6 7.8 10.2 12.6 15.1 17.8 20.8 24.2 28.0 -32.2 a
Lindau 25 m 0.1 1.2 2.8 4.9 7.7 10.5 13.1 15.9 18.9 22.2 26.0 30.3 35.4
Kaufeld 25 m 0.4 1.9 4.1 6.4 8.7 11.0 13.4 15.9 18.7 21.8 25.1 28.6 32.4
Reduction factor 0.96 0.94 0.94 0.92 0.92 0.92 0.92 0.91 0.90
25to10m
aReproducedfrom World MeteorologicalOrganization (1970) The Beaufort Scale of WindForce. Reportson Marine Science Affairs No. 3.
Geneva: World Meteorological Organization.
%eproduced from Lindau R (2001) Climate Atlas of the Atlantic Ocean. Berlin, Heidelberg: Springer Verlag.
‘Reproduced from Lindau R (1995) A new Beaufort equivalent scale. Proceedings of the lnternational COADS Winds Workshop. Kiel:
Berichte aus dem lnstitut fur Meereskunde Kiel, No. 265 (available from lnstitut fur Meereskunde, 24105 Kiel, Germany or NOAA,
Environmental Research Laboratories-CDC, Boulder, CO 80303, USA).
dReproduced from Kaufeld L (1981) The development of a new Beaufort equivalent scale. Meteorologische Rundschau34: 17-23.
taken at the small island of St Mary’s that features reversed to what Koppen advocated and used: i.e.,
several heights reaching 30-50 m above mean sea regression of Beaufort on wind speed.)
level. It is uncertain to what height in undisturbed flow The regression technique of cumulated frequencies
over water the measurements would correspond. applied by Lindau had been used by Kaufeld before
Lindau’s improved equivalent scale settled the him. It has the advantage to account for non-uniform
problem of reference height by using anemometer distribution of observations at the tails of the frequen-
measurements of known heights and reducing winds cy distribution. This technique can be seen as the
to heights of 1 0 m individually with the aid of the nonlinear equivalent to the ‘orthogonal’ two-sided
diabatic wind profile. He could also show that for the regression. Though the different error variances of the
WMO code 1100 of 1946 a reference height of 10 m is two variables were not accounted for, Kaufeld’s scale is
reasonable. The anemometer measurements at OWSs certainly preferable to the so-called scientific scales
are not corrected for flow distortion. One can hope derived by one-sided regression of wind speed on
that the exposure of instruments at OWSs and the Beaufort force.
mode of ship operation at the station will make this an Kent and Taylor reviewed a selection of Beaufort
acceptable error. equivalent scales and the techniques used in their
Often a reduction of velocity measurements from derivation, and concluded that the Beaufort equiva-
anemometer height to reference height is necessary. lent scale of Lindau is to be preferred when creating a
Typically, a constant reduction factor, derived from homogeneous monthly mean wind data set from
the neutral wind profile, is used. However, slightly anemometer and visual winds in COADS.
unstable conditions prevail at most parts of the oceans,
approaching near-neutral conditions at higher wind
speeds. For the mix of stabilities at the North Atlantic
- -
OWS, a reduction factor of 25-10m decreases with
wind speeds from 0.94 to 0.90; see Table 2.
Time Dependence of Beaufort
Estimates
Increasing interest in long time series of climate data
Discussionof Other Scales
inevitably leads to the question of whether the
The matter of Beaufort equivalent scales found reported winds from the oceans are homogeneous in
renewed interest in the second half of the twentieth time. We mentioned the slow change of observing
century. Especially within the wave modeling com- practices. There have been changes in coding practices
munity, still in 1990 the opinion prevailed that the too. Originally, Beaufort forces were used in trans-
WMO Beaufort equivalent scale (code 1100) is in mitting data. Effective 1948, W M O changed from
error. For example, the WMO Commission of Marine Beaufort forces to transmit wind speeds in knots
Meteorology produced a scale - known as CMM-IV (1knot equals 1nautical mile per hour or 0.51 m s -I).
scale - using regression of anemometer measurements For a short time of transition, the erroneous use of
on Beaufort estimates, similar to the scale of Cardone. codes may have influenced reported winds, but no
Regrettably, in the new derivations the variable with long-term trend is expected.
small error variances was regressed on a variable with There is an increasing number of ships carrying
obviously much larger error variances, leading to anemometers. Determination of true wind from ane-
biased scales. mometer measurements requires vector subtraction of
Fortunately, the governing bodies of W M O adhered course and speed of the ship, certainly an additional
to code 1100, though admitted application of CMM- source for errors and inaccuracies. Also the code
IV and similar scales for scientific purposes, so-called indicating the data as either measured (anemometer)
scientific equivalent scales. In retrospect, the scientific or estimated (Beaufort scale) is known to be less
scales are the wrong choice. They give biased climato- reliable. Peterson found frequency distributions of
logical means. The use of ‘scientific’ scales in wave estimated winds to show significant secular changes
modeling is even more questionable since this practice and ships carrying anemometers to report higher
is at variance to the use of code 1100 in operational estimated winds than ships without anemometers. The
weather analysis and forecasts. Seen in the light of the latter is most embarrassing, since no single reason and
correct derivation by Lindau, it turns out that the simple cure can be given.
insight of Koppen and Simpson around 1900 resulted On the other hand, a trend towards higher wind
in a scale, WMO code 1100, that is less biased than speeds could well be an indication of climate change.
some scientific scales of 70 years later. (It might be Growing amounts of greenhouse gases change the
noted that the description of Koppen’s method in the radiation balance. At the oceans surplus energy can
WMO Report on marine science affairs of 1970 is fuel atmospheric circulations; e.g., midlatitude and
194 BEAUFORT WIND SCALE
tropical storms. In fact, plotting the recorded wind

-
speeds at face value shows significant trends.
- 40N-50N January 1960 - 1971 -
Ramage showed that reported ocean wind speeds
exhibit secular changes. He studied monthly mean
wind speeds at the Singapore-South China route, 10
taking the along-track pressure differences for refer-
ence. He attributed the decrease of wind speed prior to r
h
World War I1 to transition from sail to steam and the b

increase thereafter to a growing portion of ships . E o
9
carrying anemometers; a true secular trend could not
be isolated.
Taking the evidence together suggests deriving time-
-1 0
dependent corrections for the Beaufort equivalent
scale. The obvious tool is to use independently
measured pressure gradients as reference, presuming
the relation between surface wind and pressure field to -20 1 ''' I ' ' ' ' ' ' I ' ' I ' 1 ' ' ' 1 ' I' ' 1 ' I' ' 1 ' ' ' II I I ' 1 ' 1 ' 1 1 6 1
be invariant. Unfortunately, ships tend to follow 0 90 1ao 270 360

certain lines providing pressure differences rather Relative wind direction, (")
than full horizontal pressure gradients. Lindau devel- Figure 3 Determination of geostrophic wind V, as reference for
oped a statistical method to derive wind velocities time-dependent calibration of Beaufort estimates, based on pres-
relative to pressure gradients. He used simultaneous sure differences from pairs of ships. Pressure differences are
wind vectors and pressure differences from pairs of shown as a function of wind direction relative to the direction of
paired observations. The dashed line is obtained considering the
ships within reasonable distance. Pressure differences
error variance of wind direction. Based on data from the North
were sorted according to the relative wind direction Atlantic between 40" N and 50" N, January, period 1960-71.
and the magnitude of pressure gradient obtained by fit (Reproduced with permission from Lindau R (1995) Time depend-
(Figure 3). Errors in wind direction estimated at VOSs ent calibration of marine Beaufort estimates using individual
are equivalent to a smoothing of the pressure differ- pressure differences. Proceedings of the lnternafional COADS
Winds Workshop. Kiel: Berichte lnstitut fur Meereskunde Kiel, No.
ences and have been accounted for.
265 (available from lnstitut fur Meereskunde, 24105 Kiel, Germany
Taking 1960-71 as the base period, time-dependent or NOAA, Environmental Research Laboratories-CDC, Boulder
corrections for estimated wind speeds have been co 80303, USA).)
derived for the North Atlantic between 20" N
and 60"N . From the uncorrected data one would
identify long-term trends of different sign before
and after World War 11, while corrected for drift of velocity for Beaufort 12, under the premise of
scale the apparent increase of wind speed after 1950 stationarity of the time series.
disappears (Figure 4). Considering the full time Beaufort equivalent scales have been determined
range from 1890 to 1990, there appears to be no mainly from observations at the North Atlantic but
physically significant trend over the North accepted by WMO for international use. If sorted by
Atlantic Ocean. ship call sign (corresponding roughly to national
origin of data) there are differences in mean results.
Noting differences in time as well as in national use, it
would be desirable to check the equivalence scale for
Unresolved Issues consistency at other oceans too. The indication that
The low and high velocity ends of the Beaufort scale anemometer-carrying ships report estimated winds
are less well determined. This is not important for the higher than ships without an anemometer calls for
low wind speeds. For the high wind speeds, obviously further study too.
too few observations are available to establish Beau- Collecting marine surface observations has been a
fort equivalents. In typical listings, speed ranges for the habit of generations. In the beginning, national serv-
Beaufort numbers are given as well as averages. ices collected logbooks. Later data were transferred
Beaufort number 12 is usually taken as unlimited, to punched cards and collected sets (referred to as
i.e., with no equivalent speed assigned. Bortkovskii 'decks') binationally exchanged. Multinational col-
fitted a Maxwell distribution to high wind speeds from lections finally resulted in the Comprehensive Ocean
winter Atlantic OWS observations for the description Atmosphere Data Set (COADS).Wind observations at
of storminess in air-sea interaction studies. This sea were collected more systematically since about
method could be used to establish a mean equivalent 1860 -welcome information for climate studies. It is
", Atlantic 20N-60N Uncorrected

E
v
-h
E 0.0
ea
'0
'C3-1.5
1890 1940 1990
[ , [ [ , [ [ I , I / , , I I I l l /
v
L
Atlantic 20N-60N , Corrected
-A
E 0.0
em
-
since 1946
-0.11 cm/s/a
u
C * 0,20 t 0.23
5 -1.5 I l l 1 I l l l l l l l I I I I I I I
1890 1940 1990
Figure 4 Mean wind speed anomalies for the North Atlantic Ocean between 20" N and 60" N, converted from Beaufort estimates using
the Lindau scale for 10 m height. Upper panel uncorrected, lower panel corrected with referenceto pressure gradients. (Reproducedwith
permission from Lindau R (2001) Climate Atlas of the Atlantic Ocean. Berlin, Heidelberg: Springer Verlag.)
hoped that the methods available now will be adopted Lindau R (1994) Time dependent calibration of
by the international community t o provide for the marine Beaufort estimates using individual pressure
production of a homogeneous, unbiased time series of differences. In: Diaz HF and Isemer H-J (eds) Pro-
ocean wind fields t o be used in future research. ceedings International COADS Workshop, Kid,
Germany. Berichte Institut fur Meereskunde,
No. 265.
Further Reading Lindau R (2001)Climate Atlas of the Atlantic Ocean. Berlin,
Heidelberg: Springer Verlag.
Cardone V, Greenwood J and Cane M (1990) On trends in Lindau R (2002) Rapport on Beaufort Equivalent Scales.
historical marine wind data. ]ournu1 of Climate 3: Advances in the Applications on Marine Climatology -
113-127. The Dynamic Part of the WMO Guide to the Applica-
Cheng C-L and Van Ness JW(1999) Statistical Regression tions of Marine Climatology. JCOMM Technical Report
with Measurement Error. London: Arnold. No. 13, WMO/TD-No. 1081. Geneva: World Meteoro-
Dobson FW (1980) Review of Reference Height for and logical Organization.
Averaging Time of Surface Wind Measurements at Sea. Peterson EW and Hasse L (1987) Did the Beaufort-scale or
Marine Meteorology and Related Oceanographic Activ- the wind climate change? Journal of Physical Oceano-
ities, Report No. 3. Geneva: World Meteorological graphy 17: 1071-1074.
Organisation. Petersen P (1927) Zur Bestimmung der Windstarke auf See.
Isemer H-J and Hasse L (1991) The scientific Beaufort Annalen der Hydrographie 55: 69-72.
equivalent scale: effects on wind statistics and climato- Woodruff SD, Diaz HF, Elms JD and Morley SJ (1998)
logical air-flux estimates in the North Atlantic Ocean. COADS release 2 data and metadata enhancements for
Journal of Climate 4: 819-836. improvements of marine surface flux fields. Physics and
Kent E and Taylor P (1997)Choice of a Beaufort equivalent Chemistry of the Earth 23: 517-526.
scale. Journal of Atmospheric and Oceanic Technology World Meteorological Organization (1970) The Beau-
14: 228-242. fort Scale of Wind Force. Reports on Marine Science
Kinsman B (1969) Historical notes on the original Beaufort Affairs No. 3. Geneva: World Meteorological Organ-
scale. Marine Observer 39: 116-124. ization.
196 BIOGEOCHEMICAL CYCLES / Carbon Cycle
Contents
Carbon Cycle
Heavy Metals
Nitrogen Cycle
Sulfur Cycle
out the Archean Eon (roughly 4.0-2.5 Ga ago) the air

Carbon Cycle was probably mainly carbon dioxide and nitrogen,
over deep oceans, possibly with methane-rich epi-
E G Nisbet, Royal Holloway University of London, sodes. During the Archean, life appeared, and has
Egham, Surrey, UK dramatically modified the composition of the atmos-
phere, such that the air has probably been a biological
Copyright 2003 Elsevier Science Ltd. All Rights Reserved
construction for most of the past 3.5 Ga.
The first possible evidence for carbon-managing life
Geological Sources of Carbon on Earth is in >3.7-Ga-old rocks in Greenland. By
around 3.5 Ga ago, certainly by 3.0 Ga ago, oxygenic
Historical Background
photosynthesis was the dominant control on the
Carbon has probably been present on the Earth’s carbon gas content of the atmosphere. Since then, 0 2
surface from the earliest accretion of the planet. and C02 have been obverse and reverse sides of the
During the Hadean Eon (4.56 to about 4.0Ga ago, same coin. The chief mediators of the interexchange of
where 1Ga is lo9 years), the atmosphere was probably carbon and oxygen between organic matter and air
rich in carbon gases. Frequent major impacts would were first the cyanobacteria and purple bacteria.
have occurred, the largest of which would have Today, descended from them, are the chloroplasts
removed any existing atmosphere, to be replaced by (from cyanobacteria) and mitochondria (from purple
fresh infall and degassing from the interior. Volcanoes bacteria) in plants and mitochondria in animals. The
in the early ocean would have hosted vigorous enzyme rubisco, which is of great antiquity, is the
hydrothermal systems, and basalt ejecta after each principal means of capture of carbon from the air, but
impact would have reacted with sea water, with rubisco in effect works in reverse if concentrations
precipitation of carbonate. At the close of the Hadean, change. In the Proterozoic Eon (2.5-0.56 Ga ago) the
the atmosphere was probably mainly carbon gases and air changed greatly. Most opinion is that initially in the
nitrogen, over a deep ocean. Archean the air may have been dominantly carbon
On Venus, the loss of hydrogen atoms from the top dioxide and nitrogen, but possibly around 2.3-2.0 Ga
of the atmosphere removed nearly all hydrogen from ago, there may have been a major event when
the water that once covered the surface of the planet, oxygen increased markedly in the air, and the atmos-
dehydrating it. The atmosphere became dominantly pheric burden of its reverse, carbon dioxide, dropped.
carbon dioxide, and is now in effective equilibrium Since then, the atmosphere has been made of nitro-
with the surface rocks, such that carbon dioxide and gen and oxygen with other minor gases, and water
rock silicate react to give carbonate and quartz, at vapor. Only the argon in the air is not biologically
roughly 500°C and 90 MPa. There is now very little managed.
hydrogen left in the planet, except in sulfuric acid in In the Phanerozoic Eon (the last 0.56Ga of Earth
clouds. The interior is probably consequently oxidized history), the development of animal life with hard
by residual oxygen. parts, such as shells, precipitated carbonate widely
On Earth, planetary evolution proceeded differently (although it should be noted that microbial precipita-
and the surface has probably sustained liquid water for tion had occurred extensively in the Precambrian).
over 4 Ga, with occasional excursions into somewhat The colonization of the land, and the development
more extreme states, including possible snowball of dense forests from the Carboniferous onwards, has
states (but always with some liquid water). Through- allowed the development of a subtle and complex
BIOGEOCHEMICALCYCLES / Carbon Cycle 197
atmospheric management regime, with short-term and North America, formed in ancient equatorial
regulation of the carbon dioxide/oxygen balance by rainforest settings.
processes such as fire and seasonal growth/decay Carbonate sedimentary rocks are very widespread,
cycles. Although Precambrian oil is known, the particularly in Mesozoic sequences. Today, carbonate
deposition of oil and coal has particularly occurred reefs still occur (e.g., the Australian Great Barrier
in the Phanerozoic, especially from the start of the Reef), but some Mesozoic reefs may have been
Carboniferous. This has had substantial impact on the immense. The Earth’s crust also contains a wide
global carbon budget. assemblage of crustal metamorphic rocks that include
Nevertheless, the basis of the biogeochemical reg- carbon (e.g., diamond) that has at one time been cycled
ulation of the air remains the same - chloroplasts and through the atmosphere.
mitochondria, descended from Archean bacteria, and
the inorganic chemical constraints of weathering and Geological Cycling
precipitation.
Geological processes cycle carbon continually from
the interior to the atmosphere and then back to the
Geological Stores of Carbon interior. Carbon is degassed from volcanoes as carbon
dioxide, carbon monoxide, and methane. Some vol-
Over time, large geological stores of carbon have
canoes also emit carbonate lavas (e.g., in the East
developed (Figure 1).Though there are estimates of
African rift region), and rare kimberlite eruptions
the size of these stores, they are very poorly con-
bring up carbon directly as diamonds.
strained. Isotopic evidence suggests that for every four
The volcanic carbon gases are either (75-80Y0)
carbon atoms degassed from the interior of the Earth,
rained out and, via bicarbonate, eventually deposited
three have been sequestered into carbonate, and one
as carbonate, when combined with calcium, magne-
captured by rubisco and incorporated into organic
sium, strontium, etc., weathered from rocks, or (20-
matter.
25 YO)captured by organic matter and eventually
The highest-level geological store of carbon is
deposited as such (eg., kerogen).
organic matter in soft sediment and in piles of
Carbon is returned through the interior of the planet
sediment only affected by burial metamorphism.
in several ways. Some sediment, especially carbonate,
This matter includes buried organic material that is
coal, and oil, is laid down directly on continents and
slowly being processed by microbial recyclers and by
eventually metamorphosed or deformed in mountain
the inorganic effects on buried organic matter of the
building events (with loss of COZ back to the air).
burial pressure and temperature. In addition, much
Other sediment is subducted and carbon dioxide is
charcoal is stored in sediment, deposited from erosion
either driven off into the overlying lithosphere at a few
after forest fires. Methanogens in sediment, and also
tens of kilometers depth, or driven off at deeper level,
inorganic metamorphism of deeper carbon-rich sedi-
eventually to ascend and return to the air via volca-
ment, produce methane. In the Arctic at a few hundred
noes. A small amount of carbon is not returned to the
meters depth, and elsewhere in thick piles of sediment
air but continues down into the deeper mantle and,
on continental margins worldwide under 500-1000 m
in turn, a small amount is emitted from deep mantle
water, very large quantities of methane are stored as
via deep-sourced mantle-plume volcanoes such as
methane hydrate, an icelike methane/water com-
Hawaii.
pound. Estimates of the total methane store are
around 1019g, an order of magnitude more carbon
than there is in the biosphere. Reservoirs and Fluxes of Carbon in the
Sedimentary rocks include stored carbon in oil, Ocean-Atmosphere-Biosphere
coal, and carbonate. Oil, gas, and coal are metamor-
phic products. Low-grade thermal maturation of
System
organic matter produces oil and gas; coal comes Figure 1 summarizes the main reservoirs of carbon in
especially from buried plant matter such as Carbon- the atmosphere-ocean system, and the main fluxes
iferous trees and bog deposits. Particular geological between them. Biological stores are small but the
periods have, for various reasons, produced remark- annual fluxes are large. Geological stores are vast,
able amounts of carbon-rich deposits. Cretaceous many orders of magnitude larger than the biological
rocks host a disproportionate proportion of the stores (though too poorly quantified for numbers to be
world’s oil, possibly reflecting the organic blooms given here), but annual fluxes are relatively small.
after huge carbon gas emissions in the Cretaceous Geological stores include carbon in carbonate (about
from mantle-plume volcanism. Carboniferous rocks, three-quarters of Earth’s near-surface store of carbon)
as their name implies, are rich in coal across Europe and also once-organic carbon (about one-quarter),
198 BIOGEOCHEMICAL CYCLES / Carbon Cycle
Figure 1 Main diagram: components of the global carbon cycle. Boxes show reservoirs in PgC, during the 1980s. Arrows show fluxes
per year in PgC. Numerical values from IPCC. Inset circles show details of specific parts of the cycle. Left: land biosphere; center: oceanic
exchanges; right: longer-term geological cycling.
such as coal, oil, and gas. Hydrate reserves of carbon returned to the water or ground, although it may be
(mainly methane in offshore sediments) are probably processed in the meanwhile. The first cycle is the
around 1019g. diurnal photosynthesdrespiration cycle, within the
The atmosphere contains about 730 Pg carbon, and seasonal cycle of growth and decay. Around this is
the ocean about 38 000 Pg where 1petagram is 10’’ g, wrapped the longer-term cycling of carbon between
or a billion tons, using the US version of billion). Soil air and the Earth’s surface (mostly ocean) over
carbon is of the order of 1500 Pg, and the total plant millennia (see exchanges in Figure 1).
carbon is around 500 Pg. Estimates of reservoirs and On an even longer time scale the geological cycles
fluxes given here are from the Intergovernmental Panel operate. Carbon dioxide is taken up as carbonate in
on Climate Change (IPCC), the best recent overall sediments and altered seafloor rock. The carbonate is
assessment. either carried down subduction zones by the plate
system, and there driven off to emerge as carbon
dioxide at volcanoes, or is accreted into continents,
where carbon dioxide is driven off by metamorphic
Global Carbon Cycles processes. Calcium, strontium, etc., erode from sili-
The global carbon system can be seen as a set of nested cate rock, then combine with the carbon dioxide to
cycles (Figure 1). The inmost cycles are the atmos- precipitate carbonate and return the cycle. In addition,
pheric cycles. Carbon dioxide has an effective lifetime some carbon is accreted to the bases of continents as
in the air of several centuries. Once emitted, an atom of diamond. On a longer time scale yet, carbon is
carbon in a molecule of carbon dioxide is likely to returned from the deep mantle to the midocean ridges,
survive in the air over this period before the carbon is and thence back to the surface.
BIOGEOCHEMICALCYCLES / Carbon Cycle 199
Geological sources, in addition to the volcanic The near-surface waters of the oceans host plankton
carbon gases already discussed, include metamorphic photosynthesizers, including both abundant cyano-
degassing, through carbonated water in hot springs bacterial picoplankton and a wide variety of euka-
and other emissions. The total flux from geological ryote plankton. Both cyanobacteria and the genetical-
sources is minuscule compared to annual fluxes within ly related chloroplasts of eukaryote phytoplankton use
the biosphere, but over hundreds of millions of years rubisco. Feeding on these are zooplankton and the
the geological cycling exchanges surface carbon complex chain of marine life. The GPP of the oceans is
effectively with the interior. On land and in the oceans, about 103 PgC year-', of which 58 PgC year-' are
the principal intermediary between atmospheric car- recycled by autotrophic respiration, leaving an NPP of
bon and captured carbon is now, as in the past, the 45 PgC year-' (compare with the 60 PgC year-' land
enzyme rubisco. This preferentially selects isotopically production - see below). In total for land and sea, the
light 12C, and consequentially inorganic carbonate annual NPP is over IOOPgC, and gross annual
made from the residual carbon is preferentially drawdown from the air is -200 PgC, contrasted to
enriched in the heavier 13C. By means of rubisco, the atmospheric reservoir of 730 PgC. Much of this is
plants capture carbon from the air. A substantial rapidly recycled back to the air.
amount of the carbon is respired back by the plants, However, some material remains abstracted from
and most of the rest by animals, bacteria, and archaea. the air. The detritus from particulate organic matter
(POC)in the oceans (everything from dead bacteria to
dead whales), plus the sinking of DOC, exports carbon
Carbon Cycling on Land
into the deeper water below the thermocline that
Most carbon capture on land is in the growth of plants. marks the lower boundary of the biologically produc-
All residents of temperate latitudes are familiar with tive upper waters. Physical exchange transports
the seasonal cycle of growth and decay. About 120 33PgCyear-' downwards as DIC. In this deeper
petagrams of C per year is captured by photosynthesis water below the thermocline, heterotrophic respira-
in plants (gross primary production: GPP), of which tion by marine animals occurs, and 42 PgC year-' is
half (60 PgC year-') is very rapidly respired. Thus net recycled upwards as DIC. Some material descends to
primary production (NPP) of the land is about the sea bottom. Also, at the seafloor, both deposition
60PgCyear-' - contrast this with the total carbon (in shallow water) and dissolution of carbonate occur
burden of the atmosphere of 730 Pg. Most of this NPP (in the deepest water).
is fairly rapidly returned to the air. About 4 PgC year-'
is returned via fires, mostly as C 0 2 but with significant
Geological Controls
CO and CH4. Fires produce significant charcoal,
which is very resistant and can survive for hundreds of Volcanism emits variable quantities of carbon gases to
millions of years abstracted from the air; thus, to this the air. In many years, few eruptions may occur; then
extent, fire is a long-term remover of carbon from the many in one year; rarely, cataclysmic eruptions occur.
air. Animals and microbially mediated decay return Metamorphism also causes emissions of carbon gases,
55 PgC year-' via heterotrophic respiration. Much of principally carbon dioxide via hydrothermal waters
this occurs via detritus (e.g., leaf fall) - some of the (e.g., in hot springs). Total geological emissions are
detritus decays quickly, while a part enters the soil
carbon reservoir and returns to the air over a much
- 0.02-0.05 PgC year-', but highly variable.
Chemical weathering is the long-term remover of
longer time period. The flux of dissolved carbon carbon from the ocean-atmosphere system. Rain-
captured via plants to inert soil carbon is about water and carbon dioxide form carbonic acid. This
0.4 PgC year-', eventually carried by rivers as dis- then reacts with rock silicate to extract the calcium and
solved organic carbon (DOC). Added to the precipitate with it as carbonate:
0.4 PgC year-' flux of dissolved inorganic carbon
(DIC) derived from CaCo3; this gives a total river
flux of about 0.8 PgC year-'.
The carbonate and quartz are eventually returned

Carbon Cycling in the Oceans
either to the mantle by subduction or to the deep crust
Air exchanges gas with water, for example in the by tectonic processes, and cycled back again to the
bubbles of wave crests under the winds. Carbon surface either via magma or as metamorphic volatiles.
dioxide dissolved in the ocean (DIC) occurs as C 0 2 , Dissolution of carbonate rock also occurs. Carbonic
HCOT, and COi-. The DIC input to the oceans is acid in soils attacks limestones, and the net result is
about 90 PgC year-', and return about 88 PgC year-'. transport of carbonate to the seafloor and return of
200 BIOGEOCHEMICAL CYCLES I Carbon Cycle
carbon dioxide to the air. As the carbonic acid is Overriding the year-on-year cycles, the steady
derived from the atmosphere, this dissolution/precip- anthropogenic growth is well illustrated in the dia-
itation cycle has no net impact on the carbon dioxide gram, especially in the South Pole record.
burden of the air, but is important in the fluxing of
dissolved inorganic carbon.
Methane and Carbon Monoxide
CaC03 (limestones) + H2CO3
Methane
+ CaC03 (shells, etc.) + C 0 2 + H2O The major sources of methane are methanogenic
archaea and combustion, both natural and human-
Global Distribution of Carbon Dioxide induced. Somewhat over 500Tg (1Tg is 1 0 l 2 g or a
The NOAA-CMDL ‘flying carpet’, diagram (more million tons) of methane are emitted each year, of
properly the zonally-averaged surface of mole frac- which roughly 300Tg are in some way linked to
tion), which incorporates measurements from a human activity and 200 Tg are from broadly natural
worldwide net of monitoring time series, illustrates sources. The major natural sources are wetlands,
carbon dioxide distribution in the world atmosphere including both tropical wetlands such as African
(Figure 2). The figure is built from long-term time papyrus swamps and South American swamps, and
series of many stations, using intercalibrated data sets. boreal (northern)wetlands and peat bogs (especially in
Several features of the diagram are notable. First, let Siberia and Canada). Natural grass and forest fires, set
us consider the interhemispheric gradient. There is far by lightning, are also major sources. Human-induced
more land in the Northern Hemisphere, and the emissions are from leaks from the gas industry, from
dominance of the north reflects the high biological coal mines, from cows and other domestic ruminants,
productivity of the northern land biomes. The marked from rice fields, and from human-lit fires, including
seasonality is caused by the seasonal variation - in tropical wildfires. Note that in some cases defining the
northern spring, massive uptake of carbon dioxide naturallhuman division is not simple: human-planted
occurs as leaf growth takes carbon dioxide out of the rice fields emit, but may replace natural wetlands that
air. In return, the rise in northern autumn, leaf fall, and also emitted methane.
microbial oxidation return the carbon dioxide to the Methane has a lifetime in the air of about a decade in
air (note that the word ‘autumn’ is used here to avoid the modern atmosphere. The principal sink of meth-
commenting on the ‘fall rise’). The diagram is shown ane is OH, from water vapor in the atmosphere. O H is
on a multiyear scale, but local continuous records most abundant in the tropical troposphere, and OH
show further diurnal detail, as carbon dioxide is taken abundance follows the sun in its progress north and
up in the warm afternoon, and respired back in the south. Thus this is the main locus of methane removal.
cooler evening. The second, much more minor sink of methane is soil
The Southern Hemisphere shows reverse seasona- microbial oxidation by methanotrophic bacteria.
lity, which records both local southern effects, and also The methane flying carpet diagram (Figure 3)
the windblown transport from the north. appears similar to the COZ diagram, but is subtly
very different. The major sources are seasonal -
wetland emissions vary logarithmically with temper-
ature and are thus much greater in hot mid-to-late
summer. In contrast, fossil fuel emissions are larger in
winter, especially gas leaks and emissions from coal
pulverization. In any place, the chief sink, OH, is
sharply seasonal, moving across the (mainly tropical)
latitudes with the sun.
The diagram illustrates the dominance of the
Northern Hemisphere, especially in human activity.
The marked seasonality in the north records both
production and the southward transport of methane
as low-methane summer winds enter the cool north
and displace the high-methane northern air south-
wards. In the Southern Hemisphere, fires contribute,
Figure 2 Carbon dioxide time series record by latitude. (>http:// as does the arrival of high-methane northern air.
www.cmdl.noaa.gov/ccgg/gallerylindex.php?pageType=folder The South Pole record illustrates the overall growth
&currDir=./Data-Figures) (Courtesy of NOAA-CMDL with thanks.) well.
BIOGEOCHEMICALCYCLES I Heavy Metals 201
etc., and anthropogenic gases including the chloro-

fluorocarbons (CFCs).
Further Reading
General Summary
Holman K (2000)The global carbon cycle. In: Jacobson MC,
Charlson RJ, Rodhe H and Orians GH (eds)Earth System
Science. San Diego: Academic Press, Chapter 11, pp.
282-321.
Intergovernmental Panel on Climate Change (2001)Climate
Change 2002. Cambridge: Cambridge University Press.
Figure 3 Methane time series record by latitude. (>http:// Special Topics

www.cmdl.noaa.gov/ccgg/galler/index.php?pageType=folder
&currDir=./Data-Figures) (Courtesy of NOAA-CMDL with thanks.) Harvey LDD and Huang Z (199.5) Evaluation of the
potential impact of methane clathrate destabilization on
future global warming. Journal of Geophysical Research
Carbon Monoxide and Minor Carbon Gases 100: 2905-2926.
Kvenvolden K (1988)Methane hydrate: A major reservoir of
Carbon monoxide is produced by incomplete com- carbon in the shallow geosphere? Chemical Geology 29:
bustion, and also as a stage in the atmospheric 1.59-162.
oxidation of methane. The chief natural sources of Lorirner GH (1981)The carboxylation and oxygenation of
carbon monoxide are wildfires. These include boreal ribulose 1,5-bisphosphate:The primary events in photo-
(northern) forest fires, that occur infrequently in any synthesis and photorespiration. Annual Review of Plant
one site, but sometimes spread over very large areas. Physiology 32: 349-383.
Nisbet EG and Sleep NH (2001)The habitat and nature of
Tropical savanna grassland fires, in contrast, are often
early life. Nature 409: 1083-1091.
annual, and emit very large amounts of carbon Pierson BK (1994)The emergence, diversification, and role
monoxide. Recently, equatorial tropical rainforest, of photosynthetic eubacteria. In: Bengtson S (ed.) Early
which rarely burns naturally, has been set on fire and Life on Earth, pp. 161-180. New York: Columbia
has become a source. University Press.
The other major source of carbon monoxide is Rosing MT (1999) I3C-depleted carbon in ,3700 Ma
combustion in industrial, heating, and transport seafloor sedimentary rocks from West Greenland. Science
activities. Partial combustion in inefficient power 283: 674-676.
plants and in inefficient heating systems (e.g., in Schidlowski M and Aharon P (1992) Carbon cycle and
east Asia) produces large amounts. Cars until carbon isotopic record: Geochemical impact of life over
recently were major emitters, but this source has 3.8 Ga of Earth history. In: Schidlowski M et al. (eds.)
Early Organic Evolution: Implications for Mineral and
been greatly reduced since the advent of catalytic
Energy Resources, pp. 147-175. Berlin: Springer-Verlag.
converters. Sundqvist ET and Broecker WS (eds) (1985) The carbon
There are a wide array of natural and human-made cycle and atmospheric COz: natural variations Archean
minor gases containing carbon, such as methyl com- to present. Geophysical Monograph 32, American Geo-
pounds (chloride, bromide, dimethylsulfide),terpenes, physical Union.
Heavy Metals
T D Jickells, University of East Anglia, Norwich, UK the periodic table except those in groups 1 and 2. This
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. definition will include metals such as aluminum,
which are not heavy in terms of their atomic weight
but which are emitted, transported, transformed and
Introduction deposited via the atmosphere in a similar way to other
‘Heavy metals’ is a widely used but poorly defined true heavy metals. Mercury is the only metal to behave
term. This article will consider all metallic elements in very differently since it exists in the atmosphere in the
202 BIOGEOCHEMICAL CYCLES / Heavy Metals
gas phase rather than on aerosols. In the context of The biggest increase from the 1950s was predomi-
global biogeochemical cycling, atmospheric transport nantly due to the use of lead additives in automobile
of heavy metals is a major transport route and one that fuels. This source has declined very rapidly from the
has for some metals been significantly perturbed by 1980s onward as a result of the removal of such
human activity. The following sections will briefly additives. Thus Table 1 serves only to estimate the
describe the sources of some heavy metals to the wide range of specific natural sources that are impor-
atmosphere, their transport and cycling through the tant for at least some elements, and the scale of
atmosphere, and their subsequent deposition, and will perturbations of the total fluxes arising from human
also consider the biogeochemical significance of the activity. The elements listed in Table 1 have been
atmospheric transport rate. Because of its fundamen- selected to represent those for which we know that
tally different behavior, a separate section is devoted to perturbation are large. For many other metals, such as
mercury. iron, aluminum and magnesium, perturbations appear
to be much smaller. This is illustrated in Table 2. Here
the data are based on direct measurements of the
Sources metals in the atmosphere, which are expressed as
Metals are emitted t o the atmosphere from a wide enrichments of the observed concentrations over those
range of sources. Some of these emissions are entirely predicted if soil dust were the only source. As is evident
natural, such as from volcanoes or from biological from Table 1, this assumption is simplistic, but large
emissions from land and the oceans. Others arise from enrichments such as seen in Table 2 do demand that
natural processes that may have been increased by there are large sources in addition to soil dust. In the
human activity, such as the formation of wind blown case of a few metals such as selenium, it is likely that
dust or biomass burning. Similarly, sea salt formation other natural sources, such as biological emissions of
is a natural process, but its significance as a source of alkylated selenium compounds from sea water, are
metals to the atmosphere may have been modified by responsible for the enrichments. However, in most
human perturbation of the concentrations of metals in cases it is generally accepted that the large enrichments
sea water. Other sources of metals to the atmosphere reflect the fact that anthropogenic emissions now
are essentially entirely anthropogenic, arising from overwhelm natural emissions for some metals as
combustion processes and industrial activity. illustrated in Table 1.
Table 1 lists best estimates of fluxes to the atmos-
phere from the known sources for a group of metals
whose fluxes are thought to have been significantly
Transport and Transformation
modified by human activity. The table is not intended Metals emitted to the atmosphere with soil dust retain
to represent a comprehensive list of sources. Indeed, this association and hence, like soil dust itself, are
for many metals we do not know sources well. associated with aerosols of an equivalent aerodynamic
Furthermore, fluxes from human activity can change mass median radius of 0.5-2pm during long-range
dramatically. Emissions of lead worldwide, for in- atmospheric transport. Larger-sized particles will be
stance, have increased more than 25-fold. Even in found close to sources or occasionally during highly
remote areas of Europe, it is possible to show energetic long-range transport. Metals derived from
increasing lead concentrations in sediments dating bursting of bubbles in sea water are associated with sea
back 2000 years associated with the Romans. Con- salt particles that typically have radii of 1-5 pm.
centrations increased further as populations and By contrast to these metals associated with aerosol
industrialization developed over the last 1000 years. particles formed directly by physical processes, metals
Table 1 Worldwide emissions of trace metals (109gy - ’)
Source Arsenic Cadmium Copper Mercury Lead Selenium Zinc
Wind-borne soil particles 2.6 0.2 8.0 0.05 3.9 0.2 19

Sea spray 1.7 0.1 3.6 0.02 1.4 0.5 0.4
Volcanoes 3.8 0.8 9.4 1.o 3.3 0.9 9.6
Forest fires 0.2 0.1 3.8 0.02 1.9 0.3 7.6
Continental biogenic emissions 1.6 0.2 2.9 0.6 1.5 3.7 5.1
Marine biogenic emissions 2.3 0.05 0.4 0.8 0.2 4.7 3.0
Total natural emissions 12 1.4 28 2.5 12 10.3 45
Anthropogenic emissions 19 7.6 35 3.6 332 6.3 132
Based on Nriagu (1989).

BIOGEOCHEMICAL CYCLES / Heavy Metals 203
Table 2 Average enrichments of metals in aerosols collected Table 3 Mass median radius of some aerosol trace metals at
from many sites around the world over the value expected if soil Enewetak Atoll in the Pacific Ocean
dust is the dominant source
Metal Mass median radius (pm)
Metal Enrichment
Sodium 4.3
AI uminurn 1.o Aluminum 1.o
Arsenic 190 Iron 1.1
Cadmium 1100 Manganese 1.3
Cobalt 1.9 Copper 0.3-0.4
Copper 25 Lead 0.3-0.6
Iron 1.3 Zinc 0.5-1 .O
Lead 320
Manganese 1.5 From Arimoto R, Duce RA, Roy BJ and Unni CK (1985) Journal of
Selenium 3500 Geophysical Research 90: 2391-2408.
Silicon 0.8
Zinc 50
Based on Wiersma and Davidson (1986).

powerful oxidizing agents. These reactions can also
Note: Because soil and rock metal concentrations and ratios vary modify metal solubility.
from place to place, small enrichments (<5) are probably not
significantly different from 1.
+ H20 +
light
[Fe(II)(H20),J2++ O H [I1
released as gases (either of biogenic origin or arising
from hot combustion processes) condense onto exist- A further process that may significantly modify the
ing particles on the basis of surface area. Hence, such behavior of trace metals during atmospheric transport
metals will be associated with finer modes of aerosol. and transformation is organic complexation. This
During aerosol transport, cloud cycling can induce process has been known for many years in soil, fresh
changes in size distribution, as different aerosol waters, and marine waters. It is now becoming clear
particles are incorporated into cloud droplets that that organic material represents a substantial compo-
subsequently evaporate to produce a mixed aerosol nent of many aerosols. This organic material is poorly
particle. Despite this complication, it is remarkable characterized but it is clear that some of the organic
that the aerosol size distribution appears to be very compounds, including relatively simple organic mol-
effectively retained over very long transport distances, ecules such as oxalic acid and also probably some of
as illustrated by data from Enewetak Atoll in the the more complex organic material, can complex
central Pacific Ocean shown in Table 3. metals and significantly modify the metals' solubility,
During atmospheric transport, aerosols containing bioavailability, and photochemistry.
metals will become hydrated and the metals may
become solubilized. This solubilization process is
highly pH-dependent, with all metals being more
Deposition
soluble at low pH. Crustal aerosol itself is often Trace metal removal processes from the atmosphere
alkaline and hence high dust loadings may suppress are similar to those of other aerosol species and involve
solubility. In contrast, anthropogenic emissions of wet and dry deposition processes. These are discussed
gases such as SO2 and NONO2 can acidify atmos- elsewhere and will not be described in detail here.
pheric aqueous solutions and thus promote solubility. However, a few important points arise that are of
Indeed, some trace metals are known to be able to specific relevance to metals. First, wet and dry depo-
catalyze oxidation of S 0 2 , and hence a positive sition processes are dependent on aerosol particle size.
feedback enhancing solubility is possible. During Metals associated with coarser material will therefore
long-range transport, aerosols will be cycled through be removed more rapidly from the atmosphere than
clouds a number of times, thereby encountering a metals associated with finer aerosol particles. The
range of p H values that may include highly acidic component of the metals in aerosol arising from gas-
waters. There is evidence to suggest that such pH to-particle conversion, predominantly the anthro-
cycling is not completely reversible and that repeated pogenic component, is therefore less efficiently re-
cycles enhance the solubility of metals at a particular moved than those components associated with soil
final p H relative to the initial aerosol solubility at that dust or sea spray. Thus anthropogenic emissions in
same pH. general not only increase emissions of metals to the
In the case of iron, photochemical reactions can lead atmosphere but also promote their long-range trans-
to the formation of OH radicals (eqn [I]), which are port. This is probably one reason why the enrichments
204 BIOGEOCHEMICAL CYCLES / Heavy Metals
Table 4 A comparison of atmospheric and riverine fluxes

(1O9 g y ') of some trace metals to the oceans
Fluvial flux Atmospheric flux
Metal Dissolved Particulate Dissolved Particulate
Iron 1100 110000 3200 29 000

Copper 10 1500 14-45 2-7
Lead 2 1600 80 10
Zinc 6 3900 33-170 11-60
From Duce, etal. (1991).

3 4 5 6 7 Note: Riverine particulate material is deposited primarily in estu-
aries and hence the best measure of comparison is probably of
PH
atmospheric flux with riverine dissolved flux.
Figure 1 Effect of pH on the solubility (masshass) of copper,
lead, and zinc in rain water. (Based on Jickells (1997).)
industrialization is preserved in the ice caps, particu-

seen in Table 2 are greater than predicted from the larly of the Northern Hemisphere, and in corals in
ratio between natural and anthropogenic emissions. central ocean regions remote from riverine sources
The effect of p H on metal solubility was noted (Figure 2).Thus it is clear that atmospheric transport
earlier. This results in marked changes in solubilities of has resulted in contamination of remote environments
metals over rather narrow p H ranges that vary from by trace metals, though the evidence for deleterious
metal to metal (Figure 1). This is important because effects arising from their contamination, and hence for
the impact of the metals on the environment varies pollution, is equivocal.
markedly depending on whether the metal is soluble or Atmospheric transport of some trace metals has
insoluble. However, this solubility will also depend on increased due to societal emissions, while that of
the p H of the receiving media, which will generally be
less acidic than rain water.
-
-
Mercury 0.75 -
v
Mercury is unique among metals in existing predom- C
inantly in the gas phase in the atmosphere as Hg'. In

this form it has long residence time of about 1 year
-
.-0
C
0
0
compared to only a few days for aerosol-bound 3 0.25
metals. It has substantial marine and terrestrial bio- D
logical sources, though on a global scale anthro-
pogenic emissions may now predominate (Table 1). 00
Atmospheric oxidation reactions slowly convert Hgo
to Hg(II), in which form it is essentially involatile and
is removed rapidly to aerosols and is deposited.
Biogeochemical Significance of
Atmospheric Transport of Trace
Metals
As noted earlier, the emissions of many metals to the vc I I I I I I
atmosphere have been estimated to have been in- 1850 1870 1890 1910 1930 1950 1970 1990
(B) Year
creased markedly by human activity and are now a
major route for global transport (Table 4).Atmos- Figure 2 (A) Concentration of lead in Greenland snow. (Repro-
pheric emissions, particularly on fine-mode aerosols, duced from Wolff E (1995) In: Wolff E, Bales RC (eds) Chemical
leads t o very effective long-range transport, as is Exchange between the Atmosphere and Polar Snow, pp. 1-18.
Berlin: Springer-Verlag,) (B) Concentration of cadmium in coral
evident from the enrichment of several trace metals in from Bermuda in the Sargasso Sea central Atlantic. (Reproduced
even the remote atmosphere (Table 2). The record of from Shen GT, Boyle EA and Lea DW (1987) Nature 328:
increasing atmospheric concentrations arising from 794-796.)
BIOGEOCHEMICAL CYCLESI Nitrogen Cycle 205
others such as iron has been little changed. Iron fluxes Ice Cores. Tropospheric Chemistry and Composi-
have changed on longer time scales, being higher tion: Mercury.
during the last glaciation owing to increased aridity
and stronger winds. Over the last few years the role of
iron has attracted increasing interest as it has become
Further Reading
Duce RA, Lino PS, Merrill JT, et al. (1991) The atmospheric
clear that iron is a key nutrient for phytoplankton in
the oceans and that in some areas remote from desert input of trace species to the world ocean. Global
regions (the dominant sources of dust), phytoplankton Biogeochemical Cycles 5: 193-259.
growth may be limited by iron availability. Higher Jickells TD (1985) Atmospheric inputs of metals and
dust loadings during the last glaciation may have nutrients to the oceans: their magnitude and effects.
Marine Chemistry 48: 199-201.
promoted high marine productivity and contributed to
Jickells TD (1997) Atmospheric inputs of some chemical
the lower atmospheric C 0 2 levels and hence to a species to the North Sea. German Journal of Hydro-
cooler climate. Furthermore, higher productivity graphy49: 111-118.
may increase emissions of dimethyl sulfide (DMS) Nriagu JO (1989)A global assessment of natural sources of
from the oceans, which would increase the acidity of atmospheric trace metals. Nature 338: 47-49.
clouds and promote increasing bioavailability of iron
Renberg I, Brannvall ML, Bindler R and Emteryd 0 (2000)
as well as increase global albedo and hence influence Atmospheric lead pollution during four millennia (2000
climate. BC to 2000 AD) in Sweden. Ambio 29: 150-156.
Spokes LJ and Jickells TD (1995)Speciation of metals in the
It has become clear over recent years that atmos-
pheric cycling of trace metals is a very important atmosphere. In: Ure AM, Davidson CM (eds.) Chemical
component of the global biogeochemical cycle. Speciation in the Environment. Edinburgh: Blackie.
Turner DR and Hunter KA (eds) (2001) The Biogeochem-
istry of Iron in Seawater. Chichester: Wiley.
Wiersma GB and Davidson CI (1986) Trace metals in the
atmosphere of rural and remote areas. In: Nriagu JO,
See also Davidson CI (eds.j Toxic Metals in the Atmosphere. New
Aerosols: Physics and Chemistry of Aerosols. Air- York: Wiley.
Sea Interaction: Gas Exchange. Biogeochemical Zhuang G, Yi Z, Duce RA and Brown PR (1992) Link
Cycles: Carbon Cycle. Deposition. Deserts and between iron and sulphur suggested by detection of Fe(I1)
Desertification. Gaia Hypothesis.Paleoclimatology: in remote marine aerosols. Nature 355: 537-539.
Nitrogen Cycle
D A Jaffe and P S Weiss-Penzias, University of the primary natural input of nitrogen to terrestrial and
Washington, Bothell, WA, USA marine ecosystems. Since the industrial revolution,
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. however, humans have learned how to fix nitrogen
without the aid of microbes and have used this ability,
driven by fossil fuel energy, to substantially increase
Introduction global food production. This fertilizer production has
Nitrogen is an essential element for all living organ- certainly been an important reason why global agri-
isms. Its global biogeochemical cycle links atmospher- culture can now feed 6 billion people. However, these
ic, terrestrial, marine, and biological processes to activities have caused the concentrations of many
sustain all life on Earth. However, the nitrogen cycle trace nitrogen species to increase substantially, alter-
has been substantially impacted by human activities, ing the global nitrogen cycle, and leading to numerous
primarily global agriculture, fossil fuel combustion, environmental problems such as acid rain, urban
and biomass burning. photochemical smog, and watershed eutrophication.
While atmospheric N2 is the most abundant form of Regarding atmospheric chemistry, N2 itself plays a
nitrogen (78.1% by volume in the atmosphere), only a very minor role in the lower atmosphere because it
small number of organisms are able to utilize this form is essentially inert. In contrast, other nitrogen species
directly. These organisms, principally some bacteria such as NO, NO2, NH3, and N20, which are
and algae, ‘fix’ atmospheric N2 into NH3 and organic present in trace amounts (parts per billion or less)
N compounds. Since all life needs nitrogen, other are involved in key atmospheric processes such as
organisms must get their nitrogen from food. This is regulating the oxidative power of the atmosphere
206 BIOGEOCHEMICAL CYCLES / Nitrogen Cycle
through tropospheric ozone formation, the acid-base low reactivity and low water solubility, N 2 0 has a
balance of precipitation, and stratospheric ozone lifetime of about 100 years in the atmosphere and is
chemistry. well mixed throughout the global troposphere. Hav-
ing no significant sinks in the troposphere, N 2 0
eventually gets transported to the stratosphere where
Key Atmospheric Nitrogen it can react or photolyze and is believed to be the major
Compounds contributor to stratospheric NO,.
There are dozens of nitrogen compounds present in the Because N 2 0 is created during denitrification, and
Earth’s atmosphere, oceans, soils, and biota, which are denitrification is often increased following the appli-
connected through the global nitrogen cycle. This cation of nitrogenous fertilizers, emissions of N2O
section will focus on five major categories of nitrogen have risen in the last century. This, along with
compounds, giving an overview of their chemistry, increases in biomass burning which also generates
sources and sinks. N20, has caused an accumulation of N 2 0 in the
The five categories are: atmosphere. N20 is a greenhouse gas with roughly
200 times the warming potential of C02. N 2 0 also
1. N2 and N 2 0 - stable nitrogen. plays a significant role in stratospheric ozone deple-
2. The NO,/NO, family (NO, NO2, H N 0 3 , and tion, supplying the stratosphere with nitrogen in a
others) - reactive nitrogen oxides. relatively reactive form. Thus, understanding the
3 . NH 3 and NH4f - ammonia and ammonium com- contributions from agricultural releases and biomass
pounds. burning of N20 and, if possible, reducing these fluxes
4. Organic N - amines, amides, alkyl nitrates, and is an ongoing area of investigation.
others.
5. Aerosol N - liquid and solid phase compounds.
NO, and NO,, Compounds
N2 and N20
This group of compounds is characterized by relatively
Nitrogen gas, N2, is a colorless gas at room temper- short atmospheric lifetimes (hours to days). They are
ature. In nitrogen fixation, a variety of bacteria and best thought of as a group in that N O is the primary
algae, both symbiotic and free-living, are able to source, and H N 0 3 is the primary sink. These com-
convert this relatively inert molecule into NH3 and pounds are responsible for the formation of photo-
organic nitrogen compounds via enzyme-catalyzed chemical smog and also substantially contribute to
electron transport chains. ‘Industrial nitrogen fixa- acid precipitation and watershed eutrophication.
tion’ refers to the industrial production of NH3 and Two subgroups within this broader class have been
nitrates from N2, mainly for fertilizers. This process is defined. NO, is defined as N O +NO,, and NO, refers
extremely energy-intensive, but has greatly increased to the sum of all reactive nitrogen oxides. NO,
yields from agricultural lands in the last century. In specifically excludes less reactive compounds such as
addition, since nitrogen-fixing organisms are found on N 20. NO, or nitric oxide, is a colorless gas at room
the roots of many legumes (clover, soybeans, chick- temperature. A small amount of N O is produced by
peas, etc.), these have been intentionally planted microbial action in natural soils, but much larger
as ‘green manures’ as a means of replenishing soil amounts are produced during fossil fuel combustion.
nitrogen. In combustion, N O is produced by the oxidation of
Natural nitrogen fixation from the atmosphere to nitrogen in the fuel, but also through the high
oceans and terrestrial ecosystems is balanced by a temperature reaction of N2 with 0 2 . The kinetics
nearly equal source of N2 t o the atmosphere via and thermodynamics of this reaction are strongly
another biological process called denitrification. In temperature-dependent, such that only at tempera-
denitrification, 80-100% of the nitrogen released is in tures exceeding about 1000 “C do significant quanti-
the form of N2, with the remainder being N20. Under ties of N O begin to form. Thus N O production is
certain environmental conditions, however, nitrous largely associated with auto emissions and power
oxide can become a major end product. plants.
Nitrous oxide, or N20, is a colorless, fairly unre- NO2, or nitrogen dioxide, is a brown-yellow gas at
active gas at room temperature. N20 has a geometric room temperature due to its light absorption in the
structure similar t o that of carbon dioxide, which blue spectral region. It has a very irritating odor and is
contributes to its inertness. It is the second most fairly toxic. Nitrogen dioxide is produced by the
abundant nitrogen compound in the atmosphere with oxidation of N O by several oxidants including O3and
a current (year 2000) average global mixing ratio of R02. NO2 also photolyzes back to NO, so that a
316 parts per billion by volume (ppbv).Because of its steady state between the N O and NO2 is quickly
Next Page
BIOGEOCHEMICAL CYCLES / Nitrogen Cycle 207
established which depends on the amount of oxidants gory includes the important compound peroxyacetyl
and available light. This explains why N O and NO2 nitrate or PAN, which is a lung and eye irritant formed
are often grouped together as NO,. Because of the in photochemical smog. Some of these compounds are
short atmospheric lifetime of NO, and the largely directly emitted from industrial emissions, including,
urban sources, NO, mixing ratios in the atmosphere for example, emissions of amines from sewage treat-
are highly nonuniform, ranging from parts per trillion ment or waste incineration, and emissions of nitro-
in remote regions to hundreds of parts per billion in arenes from fossil fuel combustion. Natural sources of
urban environments. A more detailed discussion of amines are also found, for example from decaying
photochemical ozone formation and urban smog organisms.
is given below, under ‘Photochemistry and Smog Secondary products, from atmospheric photochem-
Formation’. ical reactions in situ, include the alkyl nitrates,
Gaseous nitric acid, or H N 0 3 , is highly water- peroxyacyl nitrate, and other compounds. Although
soluble and a strong acid. It is principally formed from the contribution of these species is usually fairly
the reaction of NO2 with the O H radical. Its sinks are modest (a few percent) compared with the inorganic
wet deposition (contributing to acid rain) and dry species already mentioned, some of these compounds
deposition. Nitric acid is also produced industrially in have well-documented health effects (e.g., PAN,
large amounts as a feedstock to manufacture fertilizers nitrosamines, nitroarenes).
and explosives.
There are a number of other NO, compounds
Aerosol Nitrogen Compounds
involved in nitrogen oxide cycling, including N 0 3 ,
N205, HONO, and HN04. For a description of the Quantitatively, a relatively small fraction of total
sources, sinks and significance of these compounds, atmospheric nitrogen is found in aerosol particles.
the reader should consult one of the more specialized This is because most inorganic nitrogen compounds
references given in the bibliography. are relatively volatile and are therefore preferen-
tially found in the gas phase. Many of the higher-
NHJ NH; molecular-weight or polar organic nitrogen com-
pounds are commonly found as aerosols. Aerosol
Ammonia is a colorless gas with a very pungent odor; it
NOS can form either by uptake of gaseous nitric
is also toxic at high concentrations. NH3 is a weak
acid on a preexisting aerosol particle or by reaction of
base, forms strong hydrogen bonds, is highly soluble in
a precursor which forms NO; directly on the surface
water, and is fairly reactive. Because of this reactivity
of the particle.
and water solubility, atmospheric ammonia is short-
Observations show that N O T is found in greater
lived and, like NO,, has a very nonuniform distribu-
amounts on larger aerosol particles, i.e., those with
tion in the atmosphere: high concentrations near
diameters greater then about 2 pm. These larger
sources and extremely low concentrations in remote
particles are often composed mainly of alkaline,
regions. In the atmosphere, ammonia is one of the few
crustal material, which will retain a volatile but
‘reduced’ compounds and also one of a small number
acidic gas such as HN03. Smaller aerosol particles
of gaseous bases. For this reason, it will react with
(e.g., those with diameters less then 2 pm) are usually
acids in either the gas or the aqueous phase to produce
acidic, and so tend to retain less HN03. In urban areas,
salts, with ammonium sulfate and ammonium nitrate
high concentrations of aerosol NOT can be found
being common. These acid-base reactions reduce the
on both large and fine aerosol particles, presumably
level of acidity, or increase the pH, in rainwater. Under
due to the high levels of precursor NO,, N205,
most conditions, NH; ion, whether as solid, liquid, or
and HN03. In both urban and remote areas,
aerosol, will be present in greater amounts than
gaseous H N 0 3 concentrations are generally greater
gaseous NH3. Most ammonia gets removed from the
than aerosol NO; concentrations by a factor of three
atmosphere as NH;, principally in wet deposition.
or more.
Finally as mentioned above, ammonia is industrially
Other aerosol nitrogen compounds include high-
manufactured by the Haber process in large amounts
molecular-weight polycyclic organic compounds.
for use in fertilizers and explosives.
Some of these, such as I-nitropyrene, are known to
be animal carcinogens and possibly also human
Organic Nitrogen Compounds
carcinogens. Many of these polycyclic compounds
A wide variety of organic nitrogen compounds have are emitted in significant quantities in fuel com-
been identified in the atmosphere. They include, for bustion, especially diesel exhaust, and are typically
example, amines, amides, alkyl nitrates, nitrosamines, found associated with graphitic carbon particles
nitroarenes, and peroxyacyl nitrates. This later cate- or soot.
BOUNDARY LAYERSI Overview 221
Contents
Overview
Coherent Structures
Complex Terrain
Convective Boundary Layer
Modeling and Parameterization
Neutrally Stratified Boundary Layer
Observational Techniques-remote
Observational Techniques In Situ
Ocean Mixed Layer
Stably Stratified Boundary Layer
Surface Layer
much thinner than the atmospheric boundary layer, is

Overview called the benthic boundary layer. Here fluxes of heat
and buoyancy influences are weak and the boundary
P J Mason and D J Thomson, Met Office, Bracknell, layer is usually shear-driven. The surface of the oceans,
UK in contrast, can have substantial fluxes of buoyancy as
well as significant shear arising from the influence of
Copyright 2003 Crown Copyright.
the wind at the surface.
The term boundary layer is also sometimes used in
Introduction association with flows and currents that are concen-
Away from the surface the atmosphere is mainly trated against the sides of basins or orography. These
statically stable with little small-scale turbulent mix- features, such as the Gulf Stream in the North Atlantic
ing occurring outside frontal regions, areas with moist Ocean, arise from the flow dynamics without mixing
convective updrafts, and regions with breaking gravity playing a critical role. They are not discussed further
waves generated by hills and mountains. In contrast, here.
adjacent to the surface the flow is nearly always
turbulent, with the turbulence generated by the action Role in the Overall Atmosphere
of wind shear and/or buoyant convection. This layer The boundary layer has an important influence on the
adjacent t o the surface in which vertical mixing is behavior of the atmosphere as a whole, and activities
especially important is termed the boundary layer. In involving the representation of the atmosphere such as
fact, turbulent boundary layers involving shear or climate modeling and numerical weather prediction
buoyancy effects are a key feature of all bounded fluid cannot succeed without the boundary layer being
flows at high Reynolds number. However, the atmos- represented in some detail. The main influences on the
pheric boundary layer involves a variety of features atmosphere as a whole are as follow:
and processes in addition to shear and buoyancy. For
example, there is often a distinct interface at the height Over level terrain the boundary layer determines the
where the mixing reaches the base of the stable free drag between the atmosphere and the surface and
atmosphere above. Other dynamical and thermody- this drag is the main mechanism by which the energy
namical processes that affect the atmospheric bound- in the large-scale motion is dissipated. With small-
ary layer are Coriolis forces produced by the planetary scale orography the drag also occurs through pres-
rotation, and factors such as the formation of clouds sure forces whose magnitude is influenced by the
and radiative heat transfer. boundary layer turbulence.
The ocean is also mainly statically stable and has It provides a buffer between the surface and
boundary layers at its top and bottom. The boundary the atmosphere, thus influencing the transfer of
layer at the bottom of the ocean, which is generally heat and moisture between the surface and the
222 BOUNDARY LAYERS I Overview
atmosphere and, in particular, the way surface solar of heat H . This heat flux H is equal to pcp (w’T’),where
heating is partitioned into sensible and latent heat p and cp are the density and specific heat capacity of air
fluxes. and (w’T’)is the covariance of w’ and T’, the turbulent
It is critical in determining the properties of air fluctuations in vertical velocity and in temperature
entering the base of clouds that form the roots for about their means. Similarly, the covariance between
moist convection extending into the atmosphere the horizontal and vertical velocities gives rise to a
above the boundary layer. downward vertical flux of horizontal momentum t
It plays a central role in determining the occurrence which is given by -p(u’w’), where u’ is the fluctuation
of low-level cloud within the boundary layer and the in the horizontal component of velocity. Such a
consequential effects on radiation budgets. momentum flux is often referred to as a turbulent
It tends t o retain aerosols and pollutants from the stress.
surface, with the transfer of such polluted air t o the These turbulent fluxes have a major effect on
free troposphere being limited mainly to moist the flow. For example, if we consider the average
convection and frontal motions which, through flow in a horizontally homogeneous boundary layer
washout, leave the main atmospheric air freer of over an area of the earth surface, the governing
such material. equation of the horizontal momentum balance is given
by eqn [ll.
Role in the Local Atmospheric Environment
The boundary layer is of particular significance to = - - g r a1d p - f k x U + - -
-dU 1 dz
111
human activities and natural processes occurring on dt P P dz
the Earth’s surface. Here prediction and understand-
ing of the local environment requires an understanding Here p is the mean pressure, f is the Coriolis
of the boundary layer. In particular, the boundary layer parameter, k is a unit vertical vector, U is the
is important for predicting a range of parameters such (horizontal) mean velocity vector, and z the vertical
as coordinate, with the adjective ‘mean’ being used to
indicate quantities where the turbulent fluctuations
0 the near surface wind and turbulence; have been averaged out. Above the boundary layer
0 daily maximum and minimum temperatures; where z is zero, the flow will tend to adjust to a steady
0 visibility and fog; state with a balance between the pressure and Coriolis
0 the dispersion of pollutants and other material. forces, leading to a geostrophic flow at right angles to
the pressure gradient. Within the boundary layer,
Role in the Oceans however, there is a rotation of the flow away from the
As in the atmosphere, boundary layers in the oceans geostrophic direction, with the wind having a compo-
play an important part in the overall and local ocean nent directed down the pressure gradient. Similar
circulations: equations can be written for other variables such as
temperature or humidity. These equations will contain
0 In the deep ocean, the bottom benthic boundary a balance between the rate of change and the flux
layer has only a weak dissipating influence on the gradient, together with any source terms such as - in
ocean circulation, although it plays a more impor- the case of the temperature equation - the radiative
tant role on the continental shelf and in coastal transfer divergence.
regions. Field experiments have yielded many useful data on
0 The surface boundary layer plays a key role in flow covariances such as (u’w’). In a theoretical
influencing the rate of exchange of heat and mo- description the key issue is to estimate the covariance
mentum between the atmosphere and the ocean and from the flow properties. This is an impossible
is consequently critical to the ocean circulation. problem to resolve fully, because the flow eddies
determining the flux evolve through complex,
nonlinear and turbulent interactions, making it in-
Theoretical Framework tractable to obtain exact solutions for the fluxes in
In turbulent flow, the fluxes of flow variables are terms of the mean flow variables. In many cases,
caused by the differing properties of air ‘parcels’ however, progress is possible through simple closure
moving in different directions relative to the mean flow approximations (which relate the turbulence statistics
(the mean flow usually being defined, from a practical to the local mean flow) and/or through consideration
perspective, as an average over a period of order an of bulk models.
hour). For example, if the rising air parcels are warmer The simplest closure method is the eddy-viscosity
than those descending, then there will be a vertical flux approach in which the flux is assumed proportional
BOUNDARY LAYERS/ Overview 223
to local mean gradients through a turbulent ‘eddy magnitude of the mean wind. Integrating this equation
viscosity’ or ‘eddy diffusivity’, in the same way as leads to the log law (eqn [4]).
fluxes due to molecular motions are related to gradi-
ents via molecular viscosities and diffusivities. In the u*
u= - log -z
141
case of the velocity covariance in the momentum k zo
balance equation, this leads to the approximation of Here zo, the constant of integration, is related to the
eqn PI. height of the roughness elements and is called the
roughness length. It is typically about one-tenth of
dU the actual height of the roughness elements.
-(u’w’) = K - 121
dz The eddy viscosity approach can provide an exact
K is the eddy viscosity, which varies in space and time. match to these similarity relations if appropriate
K itself is usually determined by a combination of empirical choices for the coefficients and the various
dimensional considerations and empirical measure- scales are used. As a simple illustration, consider the
ments. It can usefully be thought of as a turbulent situation in the surface layer for the case of neutral
velocity scale multiplied by a length scale, by analogy stability. Here we can write eqn [5], where lo and uo are
with the way molecular mixing is related to the size of the length and velocity scale determining the eddy
molecular velocities and the length of the mean free viscosity K.
path. Close to the surface, dimensional considerations dU dU
imply a dependency on the distance from the surface (u’w’)= K - = l o ~ - o
dz dz
(physically this is because eddy sizes are restricted by
the presence of the ground and so tend to be smaller To match the similarity relation [3], we can choose
close to the ground) while toward the middle of the uo = u* with lo equal to kz. If we also note that (u’w’)
boundary layer some length scale related to the varies only slowly within the surface layer (the surface
boundary layer depth will prevail. A separate estimate layer is often as a result referred to as the constant-
of velocity scale is needed. This may be obtained from stress layer) and that it can be approximated by the
local or bulk considerations. A local determination surface stress u:, then we again obtain the log law. The
can be made by considering a mixing length model (see choice of u* as the correct velocity scale can be difficult
below) or by considering the turbulence energy equa- to understand (both here and in the pure similarity
tion with consideration of both shear and buoyancy approach above) and one way to motivate this is to
effects. consider a mixing length model. We assume air is
Although eddy viscosity approaches and other mixed in the vertical over a length scale lo (the mixing
closure models are only an approximate representa- length) and assume that the air starts off with the local
tion of the turbulence properties, under appropriate mean horizontal velocity and retains this velocity as it
conditions model-independent results can be derived travels. At the end of its journey it has a horizontal
-
through dimensional considerations combined with velocity that differs from the local mean by
an empirical determination of dimensionless constants u’ lo(dU/dz). Taking this as our velocity scale uo
or functions. Such similarity descriptions have exten- we obtain eqn [6].
sive application in a region near the ground where the
flow depends on only a few variables. This region,
called the surface layer, occupies the part of the flow
where the height z is much greater than the roughness
These surface relations can be refined to incorporate
elements but much less than the depth of the boundary
buoyancy effects. From dimensional considerations,
layer. The structure of the flow in this region is thought
the velocity scale wf at a height scale z due to a heat flux
to be characterized by a few parameters, namely, the
H is given by eqn [7].
friction velocity u* (defined as J ( z / p ) where z is the
magnitude of the surface stress), the surface flux of
[71
sensible heat H , and the height z above the surface. In
neutral stability conditions, with H = 0, dimensional
This order of magnitude estimate of the local buoy-
analysis then leads to eqn [3].
ancy-driven velocity scale could have been more
dU physically derived by considering the energy equation
_- U*
[31 or buoyancy accelerations subject to the key recogni-
dz -kz
tion of distance from the surface as the relevant
k is the von Karman constant (which has an empiri- length scale. From similar dynamical considera-
cally determined value of about 0.4) and U is the tions, we can derive an important length scale, the
224 BOUNDARY LAYERSI Overview
Monin-Obukhov length L , which is the height scale at horizontally homogeneous and steady conditions, and
which the shear and buoyancy velocity scales are these are best classified by stability.
equal. It is defined by eqn [8]. In the neutral boundary layer, where the surface heat
flux is negligible, the wind increases steadily with
height (although at a decreasing rate) and the wind
direction varies throughout the boundary layer. Such
neutral conditions, where shear flow turbulence gen-
This is a key parameter in the surface layer, and z / L is a eration dominates, occur as often because the wind
measure of the relative role of shear and buoyancy in speed is large because the heat flux is actually close to
the production of turbulence. The flow in the surface zero. The relevant criterion is that the boundary layer
layer is then determined by eqn [9]. depth is smaller than the magnitude of the Monin-
d U U+ Obukhov length L. Turbulence levels scale on the
- = - vM(;) 191 friction velocity u* and decrease steadily with height.
dz k z
With a geostrophic wind speed of 1 0 m s - typical
Here qIMis an empirical function which has been values of u* and of fluctuations in wind speed are
' '
estimated experimentally. Similar relations hold for about 0.3 m s - and 1m s - respectively, depending
other variables, and by integration stability-dependent of course on the surface roughness. If there are no
bulk formulas can also be obtained. q, is greater than thermal effects at all, then the boundary layer depth is
unity in stably stratified conditions ( L > 0 ) where of order 0.3u,/f. However, the term neutral boundary
turbulence transfer is weaker (i.e. for fixed dU/dz the layer usually refers only to the absence of a heat flux at
momentum flux is reduced) and less than unity in the surface, and stratification and subsidence above
convective conditions ( L< 0). At heights much less the boundary layer almost invariably play an impor-
than ILI, buoyancy effects are unimportant and tant role in restricting the actual boundary layer depth.
qM 1, with U approaching the log law appropriate
N In convective boundary layers the surface heating is
for neutral conditions. sufficient to make the Monin-Obukhov length very
In a model of the boundary layer, the net impacts can much less than the boundary layer depth. The vertical
be described by bulk drag and transfer coefficients that mixing is stronger and this tends to produce nearly
relate the surface stress to the wind U at some finite uniform profiles of velocity and temperature, with
height via eqn [ 101. variations largely confined to near the surface and the
boundary layer top. With heat being continually
t = cDIu/u POI added at the surface, a completely steady situation is
impossible and the boundary layer usually grows
It is trivial to relate this bulk formula to the similarity slowly as the stable stratification above the boundary
result - for example, in neutral conditions CDis given layer is eroded. The nature of this erosion often acts to
by C D = (k/log(z/zO))2* sharpen gradients at the boundary layer top and
Although the classical surface layer theory outlined frequently there is a strong temperature inversion
above has proved to be one of the cornerstones of extending over only a few tens of meters. Vertical
boundary layer theory, it is important to realize that it turbulent velocities tend to peak some way above the
may not be exactly correct. For example, the presence ground (at about a height of d / 3 where d is the
of large eddies that fill the boundary layer will cause boundary layer depth), with lower values near the
horizontal fluctuations in wind near the ground whose ground and boundary layer top. Horizontal turbulent
properties depend on the boundary layer depth. This is velocities, however, are more uniform as these are not
particularly true in convective conditions, but may blocked by the presence of the ground. The typical
also play a role in other types of boundary layer. velocity scale is linked to the boundary layer depth and
is of order the convective velocity scale w* which is
given by eqn [ 111.
Types of Boundary Layer
In both the atmosphere and the oceans, boundary
layers occur in forms that can be complex, involving a
diverse mixture of processes and space and time
dependencies. In spite of this general complexity, A typical value for the daytime convective boundary
boundary layer properties can often be understood in '.
layer in mid-latitudes is about 1or 2 m s - The largest
terms of the properties of a number of idealized cases, eddies extend throughout the depth of the boundary
and these cases provide an overview of the range of layer. In summer in mid-latitudes with clear skies,
possible behavior. The simplest cases are those with convective boundary layers will reach a depth of order
BOUNDARY LAYERS/ Overview 225
2 km by late afternoon. Pollutants emitted into con- ences that have a negligible effect on neutral and
vective boundary layers are dispersed rapidly. This convective boundary layers can become important -
leads to small ground level concentrations from near- examples are slight slopes, variations in the thermal
surface emissions. However, convective conditions properties of the ground, and spatial variations in
can lead to high near-surface concentrations from cloud cover. Pollutants emitted into stable boundary
releases from elevated stacks because the rapid mixing layers are dispersed slowly. This can lead to high
can bring material down to the ground quickly before ground level concentrations for near-surface sources.
it has been much diluted. In these conditions plumes However, ground level concentrations from elevated
are often seen to ‘loop’ up and down as they are sources can be small because the plume is mixed down
distorted by boundary layer eddies that are much to the ground only very slowly, and may be released at
larger than the plume. such a height or with sufficient buoyancy that it is
In stable boundary layers that generally occur at carried completely above the boundary layer. Typical
night or over other cool surfaces, mixing is greatly profiles of wind and temperature for neutral, convec-
reduced and temperature and wind vary across the tive, and stable boundary layers are shown in Figure 1.
whole depth of the boundary layer at a more constant In the real atmosphere, conditions always vary with
rate than in neutral conditions. Three-dimensional time, but usually the time variation is slow enough for
turbulent eddies tend to be smaller than in the neutral the main characteristics of the ideal boundary layers to
or convective cases and turbulence levels tend to be prevail. A primary cause of variation over land is the
low (although they scale with u* as in the neutral case, diurnal cycle. The boundary layer starts at night with a
u* itself tending to be smaller in stable cases). In shallow, stable layer. As the sun rises and heats the
addition to the three-dimensional turbulence, slow ground, convective turbulence is generated and a
lateral meandering motions are often present. Stable convective boundary layer is formed that gradually
boundary layers tend to be much shallower than erodes the stable layer above the boundary layer to
convective or neutral boundary layers, with a typical reach its maximum depth by late afternoon. Then the
depth of order 100-200m. Often, however, wind sun sets and the ground starts to cool, with thermal
shear is present above the true stable boundary layer effects now tending to suppress the turbulence. As a
due to the legacy from the daytime boundary layer result, the turbulence decays rapidly except in a
and/or inertial oscillations. In very stable cases it seems shallow layer near the ground, which forms a new
that the turbulence can be completely suppressed at stable boundary layer. Although this is a useful
times and becomes intermittent. This regime is very idealized description, it should be noted that there
hard to understand and predict because small influ- are many situations in which it does not apply. For
2t Neutral Convective Stable
E
s.
E
.g 1 -/- I
I
0 5 10 19 20 21 0 5 10 18 19 20 21 22 0 5 10 20 21 22
Wind components Potential Wind components Potential Wind components Potential
(m s-l) temperature (ms-l) temperature (m s-l) temperature
(“C) (“C) (“C)
Figure 1 Illustrationof typical profiles of mean wind ( U : V ) and potentialtemperature (6’) in neutral, convective, and stable atmospheric
’
boundary layers, for a geostrophic wind of 10 m s - and a surface temperature of 20°C. The dashed lines indicatethe boundary layer top,
which is generally well defined only in the convective case. The profiles are based on the results of large-eddy simulations for idealized
conditions but, for reasons of illustration, have been adjusted to indicate the typical relative depths of the three types of boundary layer.
Real profiles will show significant differences due to a range of influences and to the presence of turbulent fluctuations that will not be
completely removed even if the profiles are averaged over a period of an hour.
226 BOUNDARY LAYERSI Overview
example, in strong winds, the ‘mechanical’ effects may area average properties involve consideration of
dominate the thermal effects in the boundary layer boundary layer dynamics. Orography on a scale of
energy balance, leading to an effectively neutral several kilometers or less is another important surface
boundary layer. Also, at high latitudes the sun may feature. The consequent pressure forces on the flow
not be strong enough to make H positive. Over the sea give a large surface drag with high levels of turbulence
the diurnal cycle is almost absent owing to the greater and mixing.
heat capacity of the sea and hence the smaller changes Over the sea it is usual to use either roughness
in surface temperature. lengths or the equivalent bulk transfer relations. The
In this short introduction it is especially difficult to values are not constant but are mainly functions of
do justice to the full range of complexities that can wind speed. The momentum transfer occurs mainly
have some influence upon the boundary layer, but we through pressure forces on the waves and varies with
can note some key ones. Clouds and fog within the wind speed as a result of its influence on the wave
boundary layer do not release enough rain or drizzle to height. The values of roughness length are generally
be significant net sources of latent heating but rather smaller than over land surfaces. The scalar transfer
have influence through acting to redistribute heat and coefficients increase only weakly with wind speed and
moisture within their circulations and through the spray generation may be one cause of this. For waves in
effect of radiative cooling on the cloud tops. Such equilibrium with the wind there are reasonable
cooling can be the main source of turbulence energy empirical descriptions of the transfer coefficients.
production in situations with fog and stratocumulus- More refined approaches seek to use wave models to
capped boundary layers. Entrainment is a key process allow for the waves being out of equilibrium with the
in the heat and moisture budgets of such boundary wind. There also remains uncertainty over transfer in
layers, and the development of cloudy boundary layers very strong wind conditions when severe spray makes
is sensitive to this and the surface fluxes. The entrain- measurements impossible. These surface fluxes match
ment can be driven by the action of the boundary layer into the oceanic surface boundary layer, which also
eddies and is also influenced by wind shear and the receives the influence of any radiative transfer and
potential for dry entrained air to give evaporative input of fresh water from precipitation. The oceanic
cooling of the cloud. surface boundary layer is subject to dynamics compa-
The matching of the description of the turbulent rable with those of the basic neutral, stable, and
flow to the surface properties is a key issue because the convective atmospheric boundary layers. However,
surface fluxes are critical to the overall boundary layer the time scales for the variations in the mean structure
properties. At the surface a great range of complex of the temperature and salinity profiles differ and
issues prevail as the flow interacts with the obstacles derive from a combination of the seasonal heating and
and vegetation that comprise the surface. The avail- cooling modulated by the daily weather cycles. The
ability of empirical data is usually essential to quan- main areas of the ocean have a stable thermocline
titative prediction. Over homogeneous surfaces with somewhat analogous to the stable cap of most
level terrain we have noted that the momentum atmospheric boundary layers. Deep penetrating con-
transfer properties of the surface can be represented vective motion tends to occur in more limited areas
by the roughness length Z O . Corresponding roughness where strong cooling occurs. In the atmosphere,
lengths can also be used for heat and moisture and they vertical movement of air through a significant fraction
are generally much smaller because scalar fluxes do of the depth of the troposphere is similarly restricted in
not involve pressure forces. They are also less accurate its area, but the areas of occurrence are linked to moist
in use owing to variations with plant properties as they convection in deep cumulus clouds or frontal zones
interact with varying soil moisture, humidity, temper- and not directly to persistent heating.
ature, and radiative fluxes. It is common in models to
refine the treatment of vegetated surfaces with plant
Practical Boundary Layer Models PBL
canopy models that seek to represent key processes
such as those of the plant stomata, and even flow Models of a variety of types are used to represent and
within the plant canopy. predict boundary layers. Even the simplest work well
The task of representing the surface is made more in some cases, while even the most complex is not able
difficult by the presence of heterogeneity in the surface. to give reliable results in all circumstances. Fortunate-
The boundary layer tends to respond to an area ly we have a fair understanding of the reasons for
average of the surface properties. This is somewhat the success and failure and can at least anticipate these.
inverse of the way in which a pollutant from a point The simplest of these models are ‘bulk’ models that
source would occupy a large area at heights some estimate, for example, the drag and the boundary layer
distance above the surface. The methods of forming depth directly from the external parameters such as the
BOUNDARY LAYERS I Overview 227
wind above the boundary layer and the roughness argued that in the interior of the boundary layer little
length. More sophisticated models attempt to estimate more is needed from the subgrid parametrization than
the various fluxes that occur in the equations for the to allow dissipation to occur explicitly at the smallest
mean boundary layer properties. This is done either resolved scales. The only particular difficulty with LES
directly and locally within the flow in terms of the is that it depends on achieving a correct parameteriza-
mean quantities (mixing length models are the main tion near the surface. At the surface the flow eddies
example here) or by deriving equations for the fluxes become too small for the LES to describe them
and then attempting to estimate the unclosed terms in explicitly. The LES solutions depend upon the closure
those equations. The full equations for the fluxes can in this region and in the transition from this region to
be simply derived from the equations of motion. Just as the flow interior where the resolved eddies dominate
in the averaged equations of motion, where a closure the flow.
model is required for the fluxes, so in these new Models for predicting the dispersion of material in
equations a closure is needed for the higher-order the boundary layer also have a similar range of degrees
terms that appear. The terms requiring closure involve of sophistication. The simplest models estimate the
triple moments and correlations of velocity and width of any plume spread directly from a few
pressure. In a so-called second-order closure, a closure parameters, e.g., wind speed, time of day and year
assumption is made for these terms. Generally this (to estimate the solar elevation), and cloud cover
closure is, as in mixing length models, in terms of local (which affects the surface radiation budget). Some
flow properties. Such models usually only address the models involve mixing-length type assumptions or
very difficult issue of nonlocal influences through a higher-order closure models. For a dispersing plume,
more questionable derivation of the local turbulence the horizontal variations are critical as well as the
length scale and through diffusion of the second-order vertical ones and so, unlike boundary layer models
quantities. Typically, in situations where rnixing- that are often one-dimensional, a fully three-dimen-
length models are reasonable, higher-order closure sional description is required. A more sophisticated
models will do better, and are quite successful in approach is to simulate the motion of many elements
capturing the extra flow details. Neither model can of the pollutant in a ‘stochastic Lagrangian particle’
usually deal at all well with the details of flows such as model. Here the statistical properties of the flow (e.g.
convective ones with strong nonlocal influences. mean flow and velocity variances and covariances) are
Higher-order closures such as third-order ones have assumed known and a stochastic model for the
shown success in some cases but they seem unreliable random dispersion of pollutant ‘particles’ is construc-
when applied over a wide range of flows. ted to be consistent with these flow statistics. Finally,
More sophisticated again are large-eddy simulation the LES technique described above can be applied to
(LES) models. These models attempt to simulate calculate dispersion as well as the mean flow and
the detailed evolution of the flow eddies within the turbulence, either by tracking particles within the LES
boundary layer, although of course, because the flow or by solving an equation for the evolution of the
smallest structures have a size of order 10 - m, it is concentration field.
impossible to describe the entire flow in detail. Instead,
the largest eddies are simulated, with the effects of the Summary and Challenges
smallest scales parametrized (e.g., with a mixing
length model). The hope and evidence are that results Our understanding of the boundary layer has pro-
will be insensitive to the treatment of these small scales gressed greatly over recent decades. Much still remains
and so results will not be compromised by errors in the to be done, however, and progress will depend
parametrization. In using the LES technique it is not critically on the development of theory, numerical
expected that the instantaneous details of the flow will computations, and the continuing refinement of em-
correspond to a particular real case. Instead, the model pirical data. As theory and numerical computations
results are used to derive mean quantities which, by gain in potential it will remain essential to ensure that
averaging out the unpredictable turbulent variability, the underpinning empirical factors that are so critical
should be comparable with reality. In this sense the in turbulence continue to receive good attention.
model is used more in the way climate modelers use
large-scale meteorological models than in the way
weather forecasters use them. The results from LES See also
show great promise for the future and with adequate Boundary Layers: Modeling and Parameterization;
numerical resolution they offer a basis both to make Ocean Mixed Layer: Stably Stratified Boundary Layer;
direct predictions and to provide flow details with Surface Layer. Parameterization of Physical Pro-
which to develop other closure methods. It can be cesses: Turbulence and Mixing.
228 BOUNDARY LAYERSI Coherent Structures
Further Reading Monin AS and Yaglom AM (1971) Statistical

Fluid Mechanics, vol. 1. Cambridge, MA: MIT
Arya SPS (1988) Introduction to Micrometeorology. San Press.
Diego, CA: Academic Press. Nieuwstadt FTM and van Dop H (eds) (1982)Atmospheric
Garratt JR (1992) The Atmospheric Boundary Layer. Turbulence and Air Pollution Modelling. Dordrecht:
Cambridge: Cambridge University Press. Reidel.
Kaimal JC and Finnigan JJ (1994) Atmospheric Boundary Oke TR (1987)Boundary Layer Climates, 2nd edn. London:
Layer Flows - Their Structure and Measurement. New Routledge.
York: Oxford University Press. Panofsky HA and Dutton JA (1984) Atmospheric Turbu-
Kraus EB (ed.)(1977)Modelling and Prediction ofthe Upper lence - Models and Methods for Engineering Applica-
Layers of the Ocean. New York: Pergamon Press. tions. New York: Wiley.
Lumley JL and Panofsky HA (1964) The Structure o f Pasquill F and Smith FB (1983)Atmospheric Diffusion, 3rd
Atmospheric Turbulence. New York: Wiley. edn. Chichester: Ellis Horwood.
Coherent Structures
FT M Nieuwstadt, Delft Universityof Technology, Delft, the statistical approach (as Prandtl indeed first pointed
The Netherlands out in 1925), while the statistical analysis also
J C R Hunt, University College London, London, UK provides a quantitative model for the coherent struc-
tures (e.g. Holmes et af. 1996).
When the only systematic measurements were
obtained from time series recorded by isolated instru-
ments, it was natural that the study of atmospheric
Introduction
motions was essentially statistical, analyzing correla-
From the earliest times writers and artists have given tions and spectra (Panofsky and Dutton 1984) and
us their verbal and pictorial images of the character- then using the governing equations of fluid flow
istic and repeating patterns of the irregular eddying to calculate the mean dynamics and energetics,
motions in the lower part of the atmosphere. One can following Reynold’s analysis. However, the new
see their form in long streaks of snow (called by one technology and capability of multipoint tower
writer ‘rivers of wind’), vortices spurting from the and radar measurements (e.g. Kaimal and Finnigan
desert surfaces in sandstorms, billowing fog-banks 1994) together with the computational capacity
and towering cumulus clouds reaching 10 km upwards to perform numerical simulations (at varying degrees
to the tropopause (e.g. Scorer 1978). The scientific of accuracy) have now provided nearly complete
study of these coherent structures began in 1802-03 descriptions of the instantaneous and time develop-
when Lamarck in Paris and Luke Howard in London ment of coherent structures in different atmospheric
noted that there are only a few types of clouds which conditions. These broadly confirm earlier concepts
can therefore be usefully classified. Howard, whose based on observations, clouds, bird flight and dust
nomenclature (cumulus, stratus, cirrus) has only been patterns.
slightly changed by later research, had the insight to Structures tend to be particularly well defined where
realize that those structures define the essential one type of structure dominates the boundary layer,
dynamics governing the flow. The importance of his such as thermal plumes in unstable boundary layers or
insight was recognized by Osborne Reynolds in 1895 Kelvin-Helmholtz billows (which e.g. occur in the
when he was establishing the fundamental statistical stable layer), the undulating top of the mixed layer and
laws of the ‘new’ subject of turbulence. sea breeze fronts. When there are more types of
It is now generally accepted that the comprehensive structures or where these are evolving from one form
study of any type of turbulent flow, especially atmos- to another, there are no dominant structures and their
pheric flows, requires combining the statistical ap- descriptions are less precise. Quantitative evaluation
proach with the description and analysis of coherent of their velocity and pressure fields have shown that
eddy structures, i.e. motions driven by the unstable structures contribute significantly to the overall dy-
inertial and buoyancy forces in the flow that have a namics of the flow. That is why they tend to determine
characteristic, repeatable and persistent form yet are the form of the velocity spectra, e.g. of the vertical and
internally unpredictable and occur randomly in space horizontal components, especially in the distorted
and time. Development of the dual approach has eddy structure near the ground. Just as important in
improved the understanding of the approximations in practice, structures also affect the statistics of extreme
BOUNDARY LAYERS I Coherent Structures 229
events, such as very large gusts or high downbursts of The Unstable or Convective Boundary Layer
pollution. (h/LMO <0 )
This survey article reviews the various coherent An unstable or convective atmospheric boundary layer
structures in atmospheric turbulence. First the hori- usually occurs when the boundary layer is heated at
zontally homogeneous boundary layer is considered, the surface, which is being warmed by solar radiation.
then the structure of the boundary layer over inhomo- Convective layers can also form when the top of the
geneous and rough terrain and equally importantly boundary layer consists of a closed cloud deck, e.g.
over the oceans. stratocumulus clouds. In that case the top of the
boundary layer cools owing to long-wave radiation
The Stationary, Horizontally from the clouds. The cool air descends which sets up
convective motions in the boundary layer. Here,
Homogeneous Atmospheric Boundary
however, the discussion is restricted to the convective
Layer Over Flat Terrain boundary layer developing above a surface.
In Figure 1 are classified the horizontally homogene- The principle of convection applies to a layer of fluid
ous atmospheric boundary layer in terms of the ratio with a thickness h subjected to a vertical temperature
h/LMo between the boundary-layer height h and the difference AT with the high temperature at the bottom
Monin-Obukhov length LMOand the ratio zlh, where and the cold temperature at the top. The layer may
z is the height above the surface. The Monin-Obukhov become unstable when the buoyancy forces exceed the
length characterizes the height above the ground damping force of viscosity and the tendency of heat to
where buoyancy starts to dominate shear. Since the diffuse away from convecting elements. This criterion
forms of eddy structures depend on the balance of is usually expressed in terms of the Rayleigh number,
forces within them and on the proximity to bounda- defined as
ries, the type of coherent structure is characterized by /?ATh3
the two parameters h/LMo and zlh. Near the surface, Ra = -
KV
when h/ILMOI 5 1, the main statistical features of the
boundary layer tend to be determined locally in terms where p is the buoyancy parameter, K the heat
of z and local fluxes of momentum and buoyancy. conduction, and v the kinematic viscosity. In
However, in the upper part of the boundary layer the air, /? = g/To, with g the acceleration of gravity
large-scale eddying motion are also affected by inter- and To the absolute mean temperature of the fluid
actions between the troposphere and the free atmos- layer.
phere above the boundary layer if these are significant. The critical value of Ra when fluid motions first
This means that the typical frequency of the large appear lies in the neighborhood of Ra RZ 2000. The
eddies in the boundary layer, e.g. u,/h, where ug is the first instability mode just beyond this critical value is a
friction velocity, becomes comparable with the buoy- flow pattern of two-dimensional rolls. When the
ancy frequency of the stable troposphere NT or the Rayleigh number becomes larger, other modes come
Coriolis parameter f. into play and the flow assumes a pattern of hexagonal
\able wave motion

_Trapped
_ _ - _ - -K-H
- - - -waves
____---
I I
-
,c
/ I
Intermittent / /
K-H motions//
I
I Convective I Convective plumes
I
/
I
I -0.5 - roll eddies I puffs
I
I
1
/---
__-- I 1 Blocked convective eddiedsmall plumes
I Shearing + surface effects Sheared eddies near surface
I
I I
J
10 5 0 5 10 50 100
Stable hJLM0 -h/LMO Unstable
Figure 1 Diagram of the various coherent structures in the atmospheric boundary layer in term of the dimensionless height z / h and the
dimensionless stability parameter h/LMo (Adapted with permission from Holtslag AAM and Nieuwstadt FTM (1986) Scaling the
atmospheric boundary layer. Boundary-Layer Meteorology 36: 201-209.)
230 BOUNDARY LAYERSI Coherent Structures
cells with a width about equal to the layer of the fluid. plumes organize themselves into larger structures.
These are also known as Rayleigh-Bknard cells. When - h / L ~ o> 10 these take the approximate form
In the atmosphere the Rayleigh number is typically of hexagonal spoke patterns. The lateral extent of
Ra which is so high that the convective flow is large-scale structures in the surface layer is determined
highly turbulent with several types of flow structures. mainly by the blocking of the relatively wide down-
Hexagonal cloud patterns can sometimes be observed, drafts as they impact on the ground. This mechanism
e.g. when air is transported from a cold sea surface produces a horizontal wind shear close to the surface
over a land surface at higher temperature which sets up and generates sheared eddies with a typical velocity u*,
strong convection. However, the horizontal scale of as also occurs in unstratified and stable boundary
these cloud patterns is always much larger than the layers.
boundary layer depth, say from 50 to 100 km, which When 1< - h / L M O < 10 the convective boundary
makes them quite different from the Rayleigh-Bknard layer occurs in the combination with a mean wind.
cells. Generally, the structure of convective eddies, as In that case the resulting shear is able to organize
well as the profiles of the statistical properties of the the plumes into rolls with a horizontal separation of
- by the ratio of the surface
turbulence, are determined the order of magnitude close to the boundary layer
buoyancy flux FB = pw’6” to the rate of transfer of height. These can sometimes be observed as cloud
mechanical energy ( u ; / h ) caused by turbulent shear streets.
stress. Alternatively, this ratio may be expressed as
Stable Boundary Layer
h/LMo, where LMO= -ui/FB. Note that this ratio
also indicates the relative strength of the convective A stable boundary layer occurs when the vertical
turbulent eddies, with a typical velocity we, and the (potential) temperature gradient in the boundary layer
shear-dominated eddies, with a typical velocity u+, is positive. This is usually associated with the heat flux
since h/lLMoI = ( w * / u * )3 . and the associated buoyancy flux FBbeing negative at
Let us now concentrate on the flow patterns found in the surface, i.e. h/LMo > 0. However, stable boundary
the convective atmospheric boundary layer, which in layers may also occur in spatially developing flows,
Figure 1 is the region with -h/LMo > l o . In the such as in cold fronts when the boundary layer is
middle and upper part of the boundary layer, i.e. significantly cooler than the air above. In all cases the
z / h >0.1, the flow organizes itself in large-scale Richardson number Ri > 0 with Ri defined as
plume- or puff-like structures in which there is a
strong flow upwards carrying the warm air from the
surface to the top of the boundary layer. Plumes tend to
form when there is a constant heat flux at the surface,
for example by strong thermal radiation. The upward where 0 is the mean potential temperature and 77the
motion is compensated by a weak downward motion mean velocity. The coherent structures are not as
in the area outside the plumes. The consequence of this energetic as in convective and neutral flows (see the
velocity structure is that the area-averaged vertical following subsection on the shear-driven or neutral
velocity fluctuations have a positive skewness. boundary layer) because buoyancy forces suppress the
Near the top of the boundary layer, i.e. z / h >0.8, vertical displacements of fluid elements. By decorre-
where there is usually a temperature inversion, the lating vertical and horizontal motions when Ri is
plume structures impinge on this stable layer. As a a,
greater than its critical value of about the buoyancy
result the strong vertical motions are converted into forces effectively suppress the input of energy from the
horizontal velocity fluctuations which generate mean flow to turbulence. But in such situations wave
Kelvin-Helmholtz type instabilities in the velocity motions in the upper part of the boundary layer can
profile near the boundary layer top. Owing to these induce turbulence with significant energy. Turbulence
instabilities, air from above the inversion is mixed with can survive only if sufficient turbulent energy is
air of the convective layer. This is called entrainment, produced locally by the breakdown of such waves or
and it causes the boundary layer to grow during the by the mean shear. The influence of the stable
day. As a result of impinging convective eddies, stratification forces reduces the vertical dimensions
wave motions are set up which are partly trapped in of the eddy structures in relation to their horizontal
the inversion layer and partly propagate into the free dimensions to such an extent that, as clouds and
atmosphere. chimney plumes reveal, the eddies take the shape of
Near the surface, in the so-called surface layer, i.e. pancakes or ‘blink’. This is why the turbulent struc-
z / h < 0.1 heating produces structures with a size that tures in the middle of the stable boundary layer, i.e.
increases with distance z above the ground. Owing to z / h > 0.1, are not closely coupled to motions at the
their mutual entrainment or coalescence, these small surface. As with the convective boundary layer, small
BOUNDARY LAYERS I Coherent Structures 231
sheared eddy structures, e.g. longitudinal vortices with surface. This means that the structure is (weakly)
diameters of a few centimeters, are observed near the dependent on the ratio u,/(fio) and the boundary
surface. layer depth is proportional to u * / f . Owing to this
Experiments show that the turbulent pancake change in wind direction the velocity profile in the
structures persist even when the stable density gradient ‘Ekman’ layer (named after its discoverer) becomes
is so great that Ri exceeds unity. In this situation the three-dimensional and contains an inflection point
turbulence decays, but periodically it tends to be where the velocity gradient is maximum. Because of
reenergized by overturning internal waves of local both the instability of this profile and the anisotropy of
shear-driven instabilities as layers move over each Reynolds stresses, eddy structures develop in the form
other. In the quiescent periods between these events the of rolls, approximately directed along the wind
turbulence tends to collapse into layered chaotic direction. These longitudinal rolls enhance the vertical
motions in horizontal planes that have some resem- transport of momentum to such an extent that, except
blance to two-dimensional turbulence, as observed in in ideal neutral conditions, the change in mean wind
the ocean and the laboratory, and as produced by direction with height is usually less than predicted
numerical simulations. by Ekman’s theory and quite often in the opposite
direction.
The Shear-Driven or Neutral Boundary Layer
As the heat flux to and from the surface decreases and

the wind speed increases, i.e. h / l L ~ o I<< 1, progres- Eddy Structure in Boundary Layers
sively a greater proportion of the eddy motion in Over Very Rough and Inhomogeneous
the boundary layer becomes determined by
wind shear, and buoyancy forces become negligible.
Terrain
Then the boundary layer is neutrally stable and is Over snowfields, deserts and prairies, the land surface
similar to those found in engineering flows where is effectively made up from small, regular ‘roughness
the governing similarity parameter in the Reynolds elements’ such as snow and sand particles or grass and
number Re = ul/v,where u and e are the character- bushes. But over more irregular terrain and inhabited
istic velocity and length scale of the flow. However, areas there are larger obstacles such as mountains,
there are also differences because the Reynolds buildings and trees that affect the mean flow and the
number in an atmospheric boundary layer eddy structure significantly. Both types of surface
( R e FZ l o 6 ) is in general much larger than the roughness slow down the flow and affect the mass and
Reynolds number in an engineering flow ( R e 5 lo5). heat transfer to the surface. Note that even if these
In the atmosphere, as observation of ‘cat’s-paws’ obstacles are smoothly shaped, such as hills and
moving over the sea surface show very clearly, the 4)
valleys with low slopes (say less than or water waves
large-scale eddies interact with the surface by in case of air flow over the sea, they have a substantial
impinging downward and by ‘scraping’ along the effect on the mean structure of the atmospheric
surface. This top-down production of turbulence boundary layer, especially in stably stratified condi-
differs from the bottom-up instability or ‘bursting’ tions. But when the obstacle slopes are large (such as
processes which occur for low- and moderate- i)
mountains with slopes more than or when they are
Reynolds-number turbulence. The consequence is shaped with sharp angles (such as buildings) they are
that the eddy motion at different heights of the described as bluff and they generate a quite distinct
boundary layer are well correlated and the variance eddy structure.
of horizontal velocity fluctuations decreases only For the former case, of smoothly shaped obstacles in
slowly with height in the atmospheric surface layer, neutral stratification, it is found that in the ‘inner-
in contrast with the quite rapid variation as a function layer’ (which over a hill or wave is typically less than
of height in engineering boundary layers. Another about & of the length of the hill) the turbulence and
effect is that the length scale of the horizontal velocity eddy structures are similar to those over a flat surface.
fluctuation is much larger in the high-Reynolds- However in the outer region the eddy structures are
number atmospheric case. significantly distorted, leading for example to a
Above the surface layer, i.e. z / h > 0.1, if the amplification of vertical turbulence at the top of the
troposphere above the boundary layer is very weakly hill and a slight reduction of the horizontal fluctua-
stratified then the wind velocity profile and eddy tions. When several hills lie perpendicular to the wind,
structure is influenced by Coriolis acceleration f due to the distortion of the mean shear by the undula-
the earth’s rotation, and this results in a change of ting motion leads to large-scale rolls of streamwise
direction of the velocity with distance from the vortices.
232 BOUNDARY LAYERS I Coherent Structures
There are special characteristics of eddies generated when examined closely, even at the smallest scales,
by the flow over bluff obstacles, depending on whether turbulence contains structures with a distinct geomet-
their height 6 is much smaller than the boundary layer rical form that is elongated and far from isotropic.
depth h, such as in the case of buildings, or when 6 is These two findings are nevertheless consistent, be-
comparable with h, as in the case of mountains. cause these anisotropic structures have no particular
Roller eddies (or Kelvin-Helmholtz billows) are orientation, so that their statistical distribution is
produced in the intense shear layer separating isotropic. These regions consist of long, elongated
the fast-moving stream of air over the top of the threads of high vorticity (parallel to the threads). They
obstacles from the slow-moving wakes on the down- are sometimes called ‘worms’. These relatively long-
wind side. These roller eddies tend to transform lived structures are the residual motions resulting from
downwind into longitudinal eddy structures over a the rolling up of short-lived vortex sheets. The
distance of a few obstacle heights. There tends to be a dynamics of small-scale isotropic turbulence is deter-
distinct boundary between the eddies produced by mined by the vorticity production and dissipation in
the obstacle shear layers and those of the oncoming these vortex sheets, which are formed by strong
boundary layer which is visible when smoke is straining or deforming motions. The duration of this
released into the wake regions. Notable features of process is typically short, say of the order of the
isolated bluff bodies are the horseshoe vortices which Kolmogorov time scale (seconds in the atmosphere),
are wrapped around the obstacle and then extend and is therefore intermittent.
downwind. These are visible by an indented region on Despite their small size these structures play a
a sandy or snow-covered surface. They interact crucial role in environmental processes. The straining
unsteadily with the vortical eddy structures shed affects mixing and chemical reactions, while the
from the obstacle. vortices affect processes that involve small
By contrast convective turbulence over hills and particles, such as aerosols or cloud droplets. The
other rough terrain is broadly similar to that over level swirling motion in these vortices or worms cause
terrain, although the eddy structures have some weak particles heavier than air to be spun out of them.
correlation with the surface elevation, surface albedo, In other words the worms act like centrifuges, and
and temperature. When the slopes become quite large, this results in flow regions with few particles and
the plumes tend to develop preferentially on the slopes regions with many particles, also called preferential
and may converge into a single large plume at the top concentrations.
of the mountain. These phenomena are well known to The effect of preferential concentration may have
glider pilots, who make use of ascending air currents to two consequences in the process of warm-cloud rain
stay aloft. formation. The first is related to the fact that in regions
In the stably stratified flow over sloping terrain and where there is a small concentration of water droplets
smoothly shaped obstacles, such as valley flow at almost no condensation of water vapor on droplets
night, the mean flow pattern is determined by a can take place. As a result the water vapor density may
complex interaction between the geostrophic flow and grow and become supersaturated again, which implies
the buoyancy-driven flow down the slope. Since the that new cloud droplets can be formed, resulting in a
flow direction tends to vary with height, the eddy broad distribution of cloud droplets.
structures are of small scale and highly sheared. They The second effect is related to the process of
are also quite intermittent, because the rate of cooling collision and coalescence of cloud droplets. For
and the slope-driven flow varies from place to place. cloud droplets in the range of 20-50pm, the
Such slope flows tend to have a billowing eddy strong vortices may influence the trajectory of the
structure characteristic of gravity currents with a droplets such that they preferentially fall along the
strong local vertical mixing at the head of the current side of the vortex with the downward-moving
and weaker mixing downwind. velocity. As a result the settling speed is increased
and may reach a factor 80% more than the
Coherent Structures at the Small settling speed of a cloud droplet in quiescent air.
Together with the increase in concentration this may
Scales result in a larger collision probability and thus
Many measurements of spectra and correlations in droplet growth due to coalescence. When the
confirm hypotheses of G. I. Taylor and A. N. droplet becomes larger that about 50 pm the vortices
Kolmogorov that at the smallest scales of motion the can no longer deflect the droplet from its vertical fall
basic statistical structure of turbulence is isotropic. trajectory. In that case the vortices will decrease the fall
However, even qualitative observations, for example velocity somewhat with respect to the fall velocity in
of particle motion and smoke patterns, show that quiescent air.
BOUNDARY LAYERS 1 Coherent Structures 233
Significance of Coherent Structures broken up and the spectra change. The rapid change
for Practical Problems with height of the eddy motion over hills has an effect
on the performance of wind turbines, and this should
Coherent structures in the atmosphere need to be be taken into account in an assessment of the viability
understood and described in order to deal more of wind energy projects.
effectively with engineering and environmental prob-
lems. Examples are wind energy, wind loading on
structures, aircraft operation, blowing of dust and
snow, propagation of electromagnetic waves, wind See also
shelter design, and dispersion of air pollution. Study of
coherent structures is also helpful in interpreting Boundary Layers: Complex Terrain; Modeling and Pa-
statistical data (such as spectra, correlations, and rameterization; Overview. Clear Air Turbulence.
Clouds: Classification. Convection: Laboratory Models
probability distributions) and can also be used for
of. Kelvin-Helmholtz Instability.Turbulent Diffusion.
interpolation when data are not available or for Wave Mean-Flow Interaction.
extrapolation t o more complex situations. A few
examples are now given.
In the convective boundary layer, the structure of the
thin updrafts and the broad downdrafts causes the Further Reading
position of the maximum surface concentration of an
elevated source of pollution to be much closer to the Cantwell B (1990)Future directions in turbulence research
and the role of organized motions. In: Lumley JL (ed.)
source than in the neutral boundary layer. This effect
Whither Turbulence?Turbulence at the Crossroads, pp.
was not included in early atmospheric dispersion 97-131. New York: Springer-Verlag.
models, in which the turbulence was assumed to have a Etling D and Brown RA (1993)Roll vortices in the planetary
Gaussian distribution and which therefore made boundary layer: a review. Boundary Layer Meteorology
substantially incorrect predictions. The interaction 65: 215-248.
between the large, buoyancy-dominated eddy struc- Holmes P, Lumley JL and Berkooz G (1996) Turbulence,
ture in the upper part of the layer and the shear CoherentStructures,Dynamical Systems, and Symmetry.
structure near the surface also has to be understood New York: Cambridge University Press.
when estimating the variation of the dispersion from Holtslag AAM and Nieuwstadt FTM (1986) Scaling the
low-level sources. atmospheric boundary layer. Boundary-Layer Meteoro-
The eddy structure in the stable boundary layer is logy 36: 201-209.
also of importance in the dispersion of air pollutants, Hunt JCR, Sandham ND, Vassilicos JC, et al. (2001)
Developments in turbulence research: a review based
especially at night and in cold winter conditions. The on the 1999 Programme of the Isaac Newton
way the turbulence is unsteady and very sensitive to a Institute, Cambridge. lournal of Fluid Mechanics 436:
slope (even as low as lop2)helps explain anomalous 353-392.
effects and the limitations of simple predictive models. Kaimal JC and Finnigan JJ (1994) Atmospheric
For agriculture in valleys where frost pockets form, Boundary Layer Flows. New York: Oxford University
these intermittent mixing events need to be predicted Press.
and if possible avoided by artificial mixing (e.g. by fans Panofsky HA and Dutton JA (1984) Atmospheric Turbu-
or burners). lence. New York: Wiley.
When there are very strong winds, the boundary Koschmieder EL (1993) Be'nard Cells and Taylor Vortices.
layer is neutrally stratified. The calculation of wind Cambridge: Cambridge University Press.
energy and wind loads on structures needs informa- Robinson SK (1991) Coherent motions in the turbulent
boundary layer. Annual Review of Fluid Mechanics 23:
tion about the turbulent spectra. Spectra show that 601-639.
eddies near the ground are larger in the flow direction Scorer RS (1978)Environmental Aerodynamics. Chichester:
than those higher in the boundary layer. This surpris- Ellis Horwood.
ing result, widely used in engineering calculations, can Plate EJ, Federovich EE, Viegas DX and Wyngaard JC (eds)
be understood in terms of elongated eddies very close (1997)Buoyant convection in geophysical flows. NATO
to the ground. But near buildings, these eddies are AS1 series. Dordrecht: Kluwer.
234 BOUNDARY LAYERS I Complex Terrain
J J Finnigan,CSlRO Atmospheric Research, ant boundary layer flow patterns. Those over a very
Black Mountain, ACT, Australia large hill, whose pressure field is largely determined by
the displacement of the stratified synoptic flow above
the boundary layer, are quite different from those over
Introduction a smaller hill, where flow displacement is confined
The boundary layer is the layer of the atmosphere that within the neutral or unstable boundary layer. Here we
is influenced directly by the roughness and energy will confine our attention to smaller hills.
balance of the surface. Much of the character of the We will be concerned with the boundary layers that
boundary layer, therefore, is impressed upon it by the develop over terrain with these two kinds of complex-
particular nature of the underlying surface. Hence we ity, due to surface cover and to topography, and will
consider the kinds of boundary layers that develop concentrate especially on two aspects of their descrip-
over surfaces that are inhomogeneous either because tion: the windfields that we observe within them and
the surface cover is changing or because the surface is the surface stresses beneath them. Our introduction of
not flat. Although, more often than not, natural spatially averaged equations below suggests one mo-
surfaces exhibit both topography and changing sur- tivation for this. Mathematical model used for climate
face cover, it is more instructive to deal with these two or weather prediction have horizontal resolutions
elements of complexity separately. 50 x 50 km2 so that the windspeed averaged over grid
cells 500 x 500 km2 in area has to be related to some
Changing Surface Cover average of the surface properties within the cell. At the
same time we want to know how to relate measure-
Here we are concerned with horizontal inhomo-
ments of windspeed and other variables at points in an
geneity, ranging from simple changes between one
evolving boundary layer to the surrounding landscape.
surface type and another to continual changes such
as might be seen in farms with fields planted with
different crops. Atmospheric flow over such terrain Notation
is characterized by the appearance of internal
boundary layers over each new surface. If the new We use a right-handed rectangular Cartesian coordi-
surface continues sufficiently far downstream without nate system, xi(x,y , z ) with x ~ ( x )aligned with the
further change, the new internal boundary layer mean velocity at the surface and x ~ ( z normal
) to the
replaces the old boundary layer and eventually a new ground surface. Velocity components are denoted
geostrophic balance is struck between the surface and by ui(u,u, w ) with u ~ ( uthe
) streamwise component
the synoptic flow above the boundary layer. If the and u3 ( w ) the vertical component. Time averages
surface character changes continually, however, the are denoted by an overbar (e.g., M ) and departures
impact of each internal layer only extends up to some from the time mean by a prime (e.g., u’).Area averages
blending height, above which the total boundary layer over the x-y plane are represented by angle brackets
behaves as if it were flowing over a surface with (e.g. (.)I.
properties that are some average of the different
patches. Changing Surface Cover
Topography We will look first at simple changes of surface
roughness such as those between bare soil and an
Hills and valleys affect boundary layer flow because
irrigated crop. Once we have established the nature of
the pressure field that develops as the atmosphere
simple transitions from one type of surface to another
flows over them accelerates and decelerates the near-
we will be in a position to describe the boundary layer
surface flow. In a relatively thin layer near the surface,
over patchy surfaces.
analogous to an internal boundary layer, changes in
turbulent stresses strongly affect the mean flow, but at
Local Advection: The Wind Field
higher levels the changes in mean wind speed are
essentially inviscid. The pressure field that develops Local advection refers to situations where the effects of
about any given hill is strongly dependent upon the surface changes do not propagate above ~ A S L ,the
stratification of the atmosphere flowing over it, which depth of the atmospheric surface layer. Imagine an
can be characterized by a Froude number. Hence, the equilibrium surface layer flow, characterized by a
scale of the topography profoundly affects the result- logarithmic profile with roughness length zo1 and
BOUNDARY LAYERSI Complex Terrain 235
displacement height d ~eqn

, [ 11. origin of the z coordinate is adjusted appropriately to
include d.
The strength of the roughness change can be
PI characterized by the ratio of the roughness lengths,
M*, or its logarithm, M (eqn [2]).
u*1 = fi is the friction velocity, defined as the
square root of the kinematic surface stress TO, and IC is
von Karman’s constant. This flow encounters a new
surface with roughness length 202 and displacement
height d2. We will assume that the boundary is Within the internal boundary layer the flow displays
perpendicular to the surface wind vector. As the air characteristics of the downstream surface. Outside it,
flows over the new surface it either slows down apart from the small perturbation caused by the
because of increased surface friction (smooth-rough pressure pulse at the transition, the flow field is
change, 202 > z01) or speeds up because the surface identical to that upwind (see Figure 1B). The internal
friction falls (rough-smooth: 202 <zol).The effect of boundary layer depth & ( x ) is usually defined, there-
this acceleration or deceleration, which is initially fore, as the height at which the downwind velocity
confined to the air layers in contact with the new &(z) attains a fixed fraction (e.g., 99%) of its upwind
surface, is diffused vertically by turbulence, and the value at the same height.
effect of the change is felt through a steadily growing The growth of the internal boundary layer is caused
internal boundary layer of depth & ( x ) (see Figure 1A). by turbulent diffusion and, if we take the characteristic
The effects of the change are also transmitted by diffusion velocity as u*2 = 6, the downstream
pressure forces that are associated with any change in friction velocity, then we can write eqn [3].
streamline height that follows if dl is not equal to d2 d6i - B u * ~
and this pressure perturbation is not confined to the 131
dx u ~ ( z )
internal boundary layer. Its effect is negligible, how-
ever, except very close to the transition, even when the To integrate eqn [3] we need an expression for Zi,(z);
change in displacement height is significant and for for 6i < J,A.sL,the simplest assumption is that shown in
the rest of this section we will ignore it. From now on eqn 141.
we will also avoid writing z - d, assuming that the
141
Then stipulating that & ( x ) = 0 at x = 0, and locating

the origin of coordinates at the roughness change,
from eqns [3] and [4] we obtain eqn [5].
---------f Equation [5] provides a qualitative description of

the growth of the internal boundary layer and, with
the experimentally determined constant B 21 1.25,
provides a good quantitative measure of 6i(x) for
smooth-rough changes and also for rough-smooth
transitions if M is less than 2. When the rough-smooth
ru
0) change is larger (M >2), eqn [5] tends to underesti-
-0 mate the growth in & ( x ) because then diffusion
downstream of the roughness change is controlled
for some distance by the slowly decaying upstream
turbulence.
(B: To obtain eqn [5] we assumed that the velocity
profile within the internal boundary layer was loga-
Figure 1 (A) Schematic diagram of internal boundary layer
rithmic all the way up to 6 i ( x ) . This is a gross
growth. The inner equilibrium region is marked by the dashed
curve. This region is not expected to begin until some distance after oversimplification, however. In Figure 1A we have
the roughness change. (B) Logarithmic velocity profiles after a identified an inner equilibrium layer, & ( x ) at the
roughness change. The upwind profile is denoted by a dashed line. bottom of the internal boundary layer. Only in this
236 BOUNDARY LAYERSI Complex Terrain
layer has the flow attained local equilibrium with the rough-smooth (M = +4.8) changes are illustrated in
new surface with the shearing stress T ~ ( z )approxi- Figures 2A and B. In each case we see the internal
mately constant with height and the velocity profile boundary layer deepening with downstream distance
ziz(z) obeying eqn [4]. An estimate for & ( x ) can be and the velocity profile slowing in the smooth-rough
obtained by first writing an approximate equation for case and accelerating in the rough-smooth case. In
the streamwise momentum balance that ignores any both cases, the lower part of the internal boundary
pressure perturbation at the toughness change and layer is occupied by a logarithmic profile in equilib-
also assumes that the changes in the flow field are small rium with the new surface, although the true depth
of the equilibrium region is exaggerated by the
aAu aAT
logarithmic height scale. Measured in terms of
U1(z)- N -
ax az physical distance, the equilibrium region appears
where Au = t i 2 ( z ) - i i l ( z ) and AT = T ~ ( z )- z l ( z ) . If to be established more slowly in the rough-
we now insist that for local equilibrium to obtain smooth case, but in terms of dimensionless distance
below 6,(x), the integral from z = 0 to 6, of the x/zo there is little difference between the two transi-
advection term on the left-hand side of this equation tions in the rate at which equilibrium is reached.
must be negligible compared to the perturbation in
surface stress, 502 - 501 = u$2 - u i l , we obtain Local Advection: Surface Stress
2 In Figures 3A and B are plotted measurements of
-6, (2x )( 3
surface shearing stress from the experiment that
X
furnished the velocity profiles of Figure 2A,B. These
whence results are typical of those from experiments at a range
of scales. Two features are noteworthy: the overshoot
in stress at the transition and the rapid attainment of a
new equilibrium. The overshoot phenomenon is easily
explained. In the case of a smooth-rough transition,
For the kinds of roughness changes often studied in the air stream, traveling relatively rapidly over the
micrometeorology, the slope 6i/x x 1/10 while smooth surface, generates a high stress on first
6,/x M 1/100. Hence 6, corresponds to the height- encountering the increased roughness. As the region
to-fetch requirements traditionally adopted as a rule of of decelerated flow thickens into an internal boundary
thumb by researchers who wish to apply one-dimen- layer, the velocity of the air in contact with the surface
sional formulas downwind of a change in surface slows and the surface stress falls. In a rough-smooth
cover. transition, we see a stress undershoot with a relatively
For 6i > z > 6, we have a blending region, where slow airstream generating lower stress when the
the velocity profile changes smoothly between a ( z ) = surface roughness falls but the stress then rising as
u*~/[K ln(z/zo2)] and a(z) = .*I/[. ln(z/zol)]. In this the flow accelerates.
region and downwind of the immediate vicinity of Although sophisticated models of the magnitude of
the transition, the velocity and shear stress perturbat- the stress change have been developed, a simple
ions are self-preserving, that is, they can be written as expression can be derived by assuming that the
functions of a velocity scale uo and a dimensionless velocity profile obeys eqn [4] with u* = u*2 and zo =
height y(x) = z/6i(x):eqns [7] and [8]. 202 for the full depth of the inner region and then obeys
eqn [l] with u* = u*1 and zo = z01 above a sharp
171 discontinuity at z = 6i. Matching the two layers leads
directly to eqn [9].
A good choice for the velocity scale is uo = lua2 - u*1 I [91

and the functions g(y) and h(y) can be found by
substituting eqns [7] and [8] into the equations of The result of eqn [9] is plotted on top of the data points
motion and making a closure assumption to relate in Figures 3A and B and it is clear that it performs quite
T ~ ( z ) to ziz(z). Several theories have been developed well in the smooth-rough case but underestimates the
in this way and we will encounter one of them when stress change for the rough-smooth transition. Equa-
we consider continually changing surfaces. tion [9] relies on an accurate expression for 6i and we
Typical examples of the velocity profiles that have already noted that eqn [5], which is used to
develop following smooth-rough (M = -4.8) and generate the curves in Figure 3A,B, underestimates the
BOUNDARY LAYERSI Complex Terrain 237
I I I- I1 '' 1
I ' 1
0 12.20
6.42 t 0 6.10
o 2.32
1.18
A 0.32
Smooth-roug h Rough-smooth
I I I l l I
Figure 2 The development of logarithmic velocity profiles after a roughness change (data taken from Bradley (1968) see Kaimal and
Finnigan (1994) for details). (A) Smooth-rough change: zol = 0.02 mm, zo2= 2.5 mm, M = -4.8. (B) Rough-smooth change:
zol = 2.5 mm, zo2 = 0.02 mm, M = +4.8.
growth rate of 6i(x) in the rough-smooth case because Advection on Larger Scales
it discounts the influence of the energetic upstream
turbulence on the diffusion of the new internal The formulas we have derived above and the reasoning v
boundary layer. behind them strictly apply to internal boundary layers
.
I ' I
Smooth-rough
.6
N
c
.
0.41
-w
OF
Rough-smooth
I I I I 1 I 1 I 1
4 a 12 16 20
(6) x (m)
Figure3 Surface shear stress development after roughness changes (data taken from Bradley (1968), see Kaimal and Finnigan (1994)
for details). (A) Smooth-rough change: zol = 0.02 mm, 202 = 2.5 mm, M = -4.8. (B) Rough-smooth change; zot = 2.5 mm,
202 = 0.02 mm, M = +4.8. The solid line represents eqn [9] with & ( x ) calculated using eqn [5].
238 BOUNDARY LAYERS / Complex Terrain
that are no deeper than ~ A S L , the depth of the smooth-rough case illustrated in Figure 4,the early
atmospheric surface layer, because we have assumed equilibrium value of T O ~ / Z ON~ 3.5 will fall to
that the mean velocity iil ( z ) may be described by the 202/201 N 2.0 as the new boundary layer attains
logarithmic law. Above BASL, both the characteristic geostrophic balance. For the neutral case, this occurs
velocity and length scales of the turbulence change. between the point at which the new boundary layer
The length scale becomes O(z;),the depth of the whole replaces the old at x/zo l o 6 and x/zo N 10'.
N
boundary layer, while the velocity scale depends upon Attaining a new balance between the surface drag
whether the boundary layer is neutrally or unstably and the geostrophic wind will also change the
stratified. In a neutral boundary layer, the turbulent geostrophic departure, the angle between the surface
velocity scale is u* and, at higher levels, ii(z) changes and geostrophic wind direction. This angle will
more slowly with height than in the logarithmic increase in a smooth-rough change and decrease in a
surface layer. More usually, the surface layer is capped rough-smooth change.
by a convective mixed layer, where the turbulent
velocity scale is w* = [g/T0(w'B/)0~i]1/3 with (W'el), Patchwork Surfaces
the surface heat flux. In the mixed layer the mean Natural surfaces rarely consist of simple changes
velocity a ( z ) = U M is approximately constant with between two types; rather the surface cover changes
height. continuously. To describe flow over these surfaces
Inserting constant values for the turbulent velocity we generalize the concept of the internal boundary
scale (u* or w * ) and advection velocity ( i i ( z ) or U M ) layer to define the blending height, hg. Figure 4
into eqn [3], we see that we can expect 6i(x) to grow illustrates a hypothetical surface consisting of a
linearly above the surface layer with a slope between set of N patches of different surface cover, each
Bu,/ii(z) and Bw*/UM as the boundary layer varies occupying a plan area ai with streamwise extent Lj
between neutral stratification and convective mixing. and having roughness lengths and displacement
There are relatively few measurements in this regime, heights zoi and di, respectively. Over each surface an
but those that exist suggest that the surface layer value internal boundary layer grows and reaches a depth
B M 1.25 remains applicable. 6i(Lj) by the end of the patch. From the definition
The early attainment of a near equilibrium value of of the internal boundary layer we know that above
surface stress that is shown in Figure 3A,B masks the 6iMAX, the height of the deepest internal boundary
continual slow adjustment of this quantity as the layer, the velocity profile U ( z ) no longer varies
internal boundary layer grows out of the surface layer. horizontally but attains a spatially averaged value, so
The new internal boundary layer replaces the old we can identify the blending height with &MAX (eqn
boundary layer when & ( x ) equals the old boundary [101).
layer depth. This occurs at downstream distances of
order x/zo2 = lo6 in neutral conditions, but possibly hB = &MAX
much less in a convective boundary layer with a weak
mean wind. Current understanding of the magnitude If B ~ M A X is smaller than the depth of the surface layer
of the geostrophic drag coefficient u * / G , where G is ~ A S L then
, for SASL> z > hg the velocity profile will be
the geostrophic wind speed, suggests that in the logarithmic with the form given in eqn [ l l ] , where ( )
Figure 4 Schematic drawing of the flow structure over a series of surface patches with different roughness lengths zoj and streamwise
extents L i .
BOUNDARY LAYERS/ Complex Terrain 239
here denotes an average over the x-y plane. and hi = 6i(Li).
A central problem over natural surfaces is to find

an expression for the effective roughness length
zff in terms of patch level roughness lengths
zoi and other accessible parameters such as the Values for h g , hi and h,i are readily obtained using eqns
windspeed above &MAX so that the area-averaged [lo], [51 and 161.
surface momentum flux (TO) can be inferred from Equation [15] provides a much better estimate of zff
windspeed measurements or parameterized in than zy but it also starts to underestimate the
models that are unable to resolve the individual momentum-absorbing capacity of a heterogeneous
patches. surface when the streamwise length scale of the
One approach to finding zff is to assume once again patches, L, becomes small. This is because in deriving
that the flow within each internal boundary layer, eqn [15] it is assumed that the equilibrium value of
rather than being self-preserving (eqns [7] and [8]), can stress, zoi, applies over an entire patch ai and the
be represented by logarithmic profiles with local overshoots and undershoots in stress at the roughness
roughness lengths and then to average the profiles transitions that we saw in Figure 4 have been ignored.
across the x-y plane (eqn [12]). These strong perturbations in stress just following the
change are asymmetrical, the stress increases follow-
ing a smooth-rough change being generally greater
than the decrease after a rough-smooth transition. By
ignoring this asymmetry we can underestimate the
Whence we obtain average stress when the Li become very small.
Zeff = (u*iln(zoi)) and u*j = fi 1131 Larger-scale Surface Variability

0
(z0i)li2 When the scale of individual surface patches Li
Equation [13] is not a very useful formula because we becomes much larger than a kilometer, the blending
do not, in general, know the stress TO^ on each patch. height will be greater than the depth of the surface
The simplest recourse is to ignore the correlation layer. At much larger scales (Lj > > l okm), the new
between stress and roughness length and to write internal boundary layer will replace the entire plane-
eqn [14]. tary boundary layer and the regional surface stress can
be calculated by averaging the contributions of essen-
tially independent patches. At the intermediate scale,
where 10 km >Li > 1km, the blending height will be
Because, as we have seen, zof and zof are positively above the surface layer and formulas based on
correlated, z y will always be an underestimate of zff, assumptions of logarithmic velocity profiles are inap-
but it forms a useful reference value and provides a first propriate. Currently there are no simple descriptions
estimate of zff when the variation in roughness length of this scale of heterogeneity. Numerical models that
between patches is small. can accommodate the diabatic influences that are
A more accurate formula for zff has been derived by usually important above the surface layer have been
exploiting the fact that the flow in the internal used in particular cases and the average surface stress
boundary layers making up the blending region can be considered to be bounded by the values
between 6,, and hB is self-preserving and assuming appropriate to small- and large-scale heterogeneity.
that in this region a simple mixing-length expression is
adequate to express the relationship between shear
stress ~ ( z and
) velocity shear &/az These two Topography
assumptions allow the shear stress and velocity at the
Air Flow over Isolated Hills
blending height to be related to the local values within
the thin equilibrium layer over each patch so that the As we did in considering changing surfaces, we will
value of zof required to weight the local roughness first describe the flow over an isolated hill and then
length zof in eqn [13] can be inferred. The result is a go on to consider how the boundary layer adjusts
formula for the effective roughness length that is most to continuously hilly terrain. We will confine our
simply expressed as in eqn [15], where he, = Sez(Lf) attention to hills sufficiently small that the flow
Next Page
240 BOUNDARY LAYERS / Complex Terrain
Figure 5 Schematic drawing of the flow over a 2D ridge showing the formation of a downstream separation region when the ridge is
steep enough. On an axisymmetric hill, the upwind deceleration region is replaced by a region of lateral flow divergence.
perturbations they cause are confined within the z l l - 0.3. In the wake we see a substantial velocity
boundary layer. In practice this means that the hill deficit extending to at least z = H .
height H and the hill horizontal lengthscale L satisfy Much of the understanding we now have about the
H << zi and L << h", where h*, the 'relaxation length' dynamics of flow over hills derives from linear theory,
of the boundary layer, is defined as h" = ziUO/u, or which assumes that the mean flow perturbations
z i U ~ / w according
, to whether the flow is neutrally caused by the hill are small in comparison to the
stratified or convectively unstable. The velocity scale
UOis defined below. The horizontal length scale L is I I
defined as the distance from the hill crest to the half-

height point. In continuously hilly terrain it can be
more appropriate to use a characteristic wavelength i
as the horizontal length scale. For sinusoidal terrain,
L = i/4. 10
In Figure 5 we have sketched the main features of the
velocity field about an isolated hill. The figure could
represent flow approaching an axisymmetric hill or a
2D ridge at right angles. Close to the surface, the flow
decelerates slightly at the foot of the ridge before
accelerating to the summit. In the axisymmetric case
the deceleration is replaced by a region of lateral flow
divergence at the foot of the hill. The wind reaches its
maximum speed above the hill top and then deceler-
5
1
ates on the lee side. If the hill is steep enough
downwind, a separation bubble forms in which the
mean flow reverses direction. Whether the flow
separates or not, a wake region forms behind the hill
with a marked velocity deficit extending for at least
/
/
10 H downwind. .'
/
I / ,/ Prnr,
"IC'JL
The same information is made more concrete in ,
Figure 6, where velocity profiles well upwind, over the II / ,'
hill top and in the wake are plotted. The vertical 0.1 ~ I I
coordinate z measures height above the local surface. 0 1
In Figure 6 it is made dimensionless with the inner UIU,
layer height 1, defined below. Upwind we have a
Figure 6 Profiles of mean velocity observed upwind, on the crest,
standard logarithmic profile, but On the top the and in the wake region of a hill, The vertical Scale is made
profile is accelerated with the r f ~ x i m u mrelative dimensionless with the inner layer depth, 1. Note the position of the
speed-up occurring quite close to the surface at maximum speed-up on the crest at z //3. -
BOW ECHOS AND DERECHO 31 1
Kitaigorodskii SA (1973) Physics of Air-Sea Interaction, Panofsky HA and Dutton JA (1984) Atmospheric Turbu-
trans. from Russian by Baruch A. Jerusalem: Israel lence. New York: Wiley.
Program for Scientific Translations. Stull RB (1989) An Introduction to Boundary Layer
Lumley JL and Panofsky HA (1964) The Structure of Meteorology. Boston, MA: Kluwer.
Atmospheric Turbulence.New York: Wiley.
M Weisman, National Center for Atmospheric is envisioned to be composed of a sequence of

Research, Boulder, CO, USA relatively independent convective cells that contribute
Copyright 2003 Elsevier Science Ltd. All Rights Reserved collectively to a larger system-scale structure. The
individual convective cells can be ordinary cells,
multicells, or supercells, as described elsewhere in
Introduction the encyclopedia (see Convective Storms: Overview).
Windstorms produced by complexes of convective In the following, we emphasize the system-scale
storms (thunderstorms) pose a significant hazard to attributes that have led to these particular systems
life and property in many places of the world, being identified for their unique form of meso-
especially during the spring and summer months. convective organization.
The largest and most long-lived of these events have
been given a generic name of ‘derecho’, a term that
originated in the late 1800s to refer to convective
Bow Echoes
systems producing wide and long swaths of straight- Bow echoes, originally referred to as a line echo wave
line wind damage. Detailed studies of convective wind pattern (LEWP), are most readily identified by a
events, however, have shown that a vast majority are persistent bow-shape on a radar screen, and have
associated with a particular type of organized convec- become especially associated with the production of
tive system, more popularily referred to as a ‘bow long, narrow swaths of damaging surface winds.
echo’. This chapter describes the basic structures and Much of what we know observationally concerning
environments associated with bow echoes and der- bow echoes originated with Dr T. T. Fujita, who spent
echoes, and further highlights some of the recent much of his career trying to characterize and under-
research that clarifies the mechanisms critical to their stand the types of convective systems most apt to
development and maintenance. produce severe weather such as downbursts and
Bow echoes and derechoes form a subset under the microbursts.
more general heading of mesoscale convective sys- A typical evolution and morphology of radar echoes
tems, which include squall lines, mesoscale convective associated with a severe bow echo is presented in
complexes, and the like. In all of these cases, the system Figure 1. The system usually begins as a strong
Figure 1 A typical morphology of radar echoes associated with bow echoes that produce strong and extensive downbursts, labeled DB
on the figure. (Reproduced with permission from Fujita, 1978.)
31 2 BOW ECHOS AND DERECHO
convective cell or a short line of convective cells that pools. Such rear-inflow jets can reach magnitudes of
may be either isolated or embedded within a more 25 m s-' or greater above the ambient mid-level flow.
extensive squall line. A5 the strong surface winds The cold pool and rear-inflow jet represents the
develop, the initial cell evolves into a bow-shaped line primary source for the strong surface winds in such
segment of cells, with the strongest winds occurring systems.
near the apex of the bow. Cyclonic and anticyclonic Bow echoes are observed over a wide range of scales,
motion of the radar cells are often noted on the from tens of kilometers to over 200 km (along-line
northern and southern ends of the bow segment, length) in extreme cases. Lifetimes can range from a
respectively. During the declining stage, the system couple of hours to over 10h. Although a range of
often evolves into a comma-shaped echo with pre- scales is observed, the most intense bow echoes tend to
dominantly cyclonic motion of the radar echoes be 40-120 km in length and have lifetimes of 4-6 h.
evident on the northern end of the system (Figure 1E). Widespread surface winds of 25 m sC1 or greater are
At the core of most bow echoes is a strong surface commonly observed with severe bow echoes, with
cold pool and associated surface mesohigh, which is extreme cases producing swaths of damaging winds of
produced via the evaporation of rain and transport of greater than 50 ms-l, producing widespread falls of
potentially colder air to the surface from the mid- trees, toppling power poles, damaging buildings, and
troposphere [e.g., 3-5 km above ground level (AGL). blowing vehicles off highways.
Cold pools within severe bow echoes can be as much as An example of bow echo that passed through
10-15°C colder than the surrounding airmass, with Springfield, Illinois, on 6 August 1977 is presented in
surface pressure excesses reaching 5-8 mbar. In asso- Figure 2. This system evolved from a relatively isolated
ciation with this cold pool is also often an intense rear- cell to a comma-shaped echo over a 5 h period while
inflow jet that extends rearwards from the leading producing a continuous swath of damaging surface
edge of the active convection and may extend in depth winds over a 200 km path. A detailed damage survey
from 3-5km AGL down to the surface. This rear- taken during a portion of this event is presented in
inflow jet helps transport drier, mid-level air into the Figure 3, and demonstrates that the broad swath of
precipitating region behind the leading edge of the outflow winds is often made up of a series of individual
convection, increasing the potential for strong, evap- downbursts and microbursts (see also Microbursts). It
orationally produced downdrafts and resulting cold also demonstrates the tendency of bow echoes to
Figure 2 Evolution of radar e c h o e s associated with the Springfield downbursts and tornadoes of 6 August 1977. (Reproduced with
permission from Fujita, 1978.)
BOW ECHOS AND DERECHO
Aerial survey and Mapping by FORBES and WAKIMOTO
Figure 3 Eighteen tornadoes, 10 downbursts, and 17 microbursts are depicted in this map. One tornado (No. 11) was anticyclonic. Apparently, eight tornadoes formed on the left side of
microbursts. No trace of downbursts were found in the vicinity of other tornadoes. (Reproduced with permission from Fujita, 1978.)
313
314 BOW ECHOS AND DERECHO
generate tornadoes, especially along and to the north severe cells, sometimes supercellular, may be con-
of the apex of the bow (Figure 2). In the present case, tained within the larger-scale structure (e.g., note the
18 tornadoes were generated just to the north of the ‘intense convective cell’ with the tight reflectivity
bow echo apex as the northernmost cell evolved into a gradient and hook-like appendage at the southern end
cyclonically rotating head. The relationship between of the bow echo in Figure 4).
bow echoes and tornadoes has still not been adequate- Another example of a mature bow echo is shown in
ly explained. Figure 5 from 5 May 1996 near Paducah, Kentucky. In
A wide range of radar-echo configurations can this example, a large bow-shaped convective system
be associated with developing severe bow echoes has two smaller-scale bows embedded with the larger
(Figure 4).One of the common characteristics is the circulation. The Doppler winds clearly depict a large
development of a strong low-level reflectivity gradient rear-inflow jet behind the core of the system (dark
near the leading edge of the concave-shaped echo. blue), with weak anticyclonic shear to the south of the
Another significant feature is the presence of a weak bow and stronger cyclonic shear evident on the
echo channel or ‘rear inflow notch’ (RIN)on the back northern end of the bow. Additionally, the smaller
side of the bow, which often signifies the location of the embedded bows each have their own localized rear-
intense rear-inflow jet and the likelihood of downburst inflow jets with associated rotational features on
winds and possible downburst-induced tornadoes. the ends. This event emphasizes that a range of
Also, while a bow echo is generally organized on a bow-echo scales can exist, sometimes side by side, in
scale larger than a single convective cell, individual the same basic environment. Avertical cross-section of
Figure4 Radar analysis of the central Minnesota derecho between 2047 and 21 12 UTC from Minneapolis-St Paul, Minnesota (MSP).
Reflectivity contours are 18, 30, 41, and 46 dBZ. Shaded region represents reflectivity values greater than 50 dBZ. (Reproduced with
permission from Przybylinski, 1995.)
BOW ECHOS AND DERECHO 315
(A) Base reflectivity 5 May 1996 18.48 (6) Relative velocity 5 May 1996 18.48
Figure5 (A) Base reflectivity and (6) relative velocity from the PaducahWSR-88D radar at 18.48 GMT for 5 May 1996. Velocities are presentedrelative toa storm motionof 33 knots (17 m s ') ~
from 280". On the scales in the lower left of each figure, ND indicates no data and RF indicates range folding. For (B)blue colors represent flow towards the radar while red colors represent flow
away from the radar (RW Przybylinski,personal communication.)
Figure6 Vertical cross-sections of (A) reflectivity and (B) storm-relative at 1855 UTC for the Paducah, Kentucky bow echo. The vertical
cross-sections are taken at a 277" heading from KPAH. Velocities are presented relative to a storm motion of 20 knots (10 m s - ' ) from
277". For (B), green colors represent flow towards the radar while red colors represent flow away from the radar. (RW Przybylinski,
personal communication.)
reflectivity taken through the core of the bow Derechoes

(Figure 6A) depicts strong, upright convective cells at
the leading edge, with a weaker stratiform region The term 'derecho' is used to describe convective
extending rearward. A storm-relative velocity cross- systems that produce straight-line convective wind
section (Figure 6B) depicts front-to-rear ascending gusts greater than 26 m s- within a concentrated area
flow through the convective cells and extending aloft with a major axis length of at least 400 km. The gusts
within the anvil, with a strong rear-inflow jet beneath must also show a systematic pattern of progression,
the front-to-rear flow at mid-levels. with no more than 3 h elapsing between successive
Figure 7 Area affected by the convective windstorm of 5 July 1980 (dashed line). Three-hourly squall line positionsare indicated in UTC
',
(from 0300 to 2100 on 5 July). Officially measured convective gusts are indicated by wind barbs (full barb signifies 5 m s- flag signifies
25 m s- '). Personalinjuries (67) are indicatedby dots, and each death (6) is shown by an (Reproducedwith permissionfrom Johns and
'XI.
Hirt, 1987.)
wind damage events. Such systems have been observed bility and low-level moisture. Surface dew points are
to have lifetimes of as long as 18 h, producing a swath commonly greater than 20°C with lifted indices
of damaging winds hundreds of kilometers wide and averaging about - 9°C. Convective available poten-
1000 km long. An example of the extent and longevity tial energy (CAPE) is generally greater than
of a derecho event is presented in Figure 7. In this case a 2000 J kg-', with many cases exhibiting CAPEs
squall line produced damaging wind over a swath greater than 4000 J kg- Such large CAPEs support
hundreds of kilometers wide and 1000 km long over both the development of strong convective updrafts as
18 h. well as strong convective downdrafts and cold pools,
Two basic patterns of radar cells are often associ- the latter being especially critical for the development
ated with a derecho (Figure 8 ) . The first pattern of strong surface outflow. The development of strong
(referred to as a progressive derecho) consists of a cold pools can also be supported by dry mid-tropo-
single bowed segment of convective cells that often spheric conditions, although bow echoes and derech-
develops just on the cool side of a weak stationary oes are observed for both moist and dry mid-level
front. The bowed feature moves parallel to the front. conditions as long as sufficient CAPE is available.
The second pattern (referred to as a serial derecho) Environmental vertical wind shear magnitudes tend to
consists of a longer squall line that has evolved into a be in the moderate range for severe convective events,
series of bow echoes or LEWPs that propagate along with about 1 5 m s - ' of shear evident between the
the squall line. While bow echoes represent one of the surface and 700 mbar (roughly 0-3 km AGL), and
primary convective structures associated with derecho about 2 0 m s P 1 between the surface and 500mbar
events, severe bow echoes occur much more frequently (roughly 0-6 km AGL). Such shear magnitudes are
than derechoes, producing damaging surface winds generally weaker than are associated with supercell
over more limited regions and for shorter time periods storms, although bow echoes and derechoes are
than included in the strict definition of a derecho. observed in these more strongly sheared environments
as well.
Bow echoes and derechoes can occur in environ-
Bow Echo and Derecho Environments ments with strong synoptic-scale forcing, as with
severe, prefrontal squall lines, but also occur quite
and Climatology often in association with more benign synoptic
Bow echo and derecho environments are generally patterns. As presented in Figure 9, many events begin
characterized by large amounts of convective insta- along or to the north of a weak east-to-west oriented
318 BOW ECHOSAND DERECHO
Mean wind
direction
---h
-120 km
Figure 9 Idealized sketch of a midlatitude warm season synop-

tic-scale situation especially favorable for development of long-
lived progressive bow echo complexes producing extensive
swaths of damaging winds. The line B-M-E represents the track
of the bow echo complex. Thin lines denote sea level isobars in the
vicinity of a quasi-stationary frontal boundary. Broad arrows
represent low-level jet stream (LJ) and polar jet (PJ) in the upper
Mean wind troposphere. (Reproduced with permission from Johns and
Doswell, 1992.)
to help increase the thermodynamic instability, fills out

the ingredient list that promotes the development of
bow echoes and derechoes for this type of weather
-120 km
pattern.
Figure 10 shows the paths of 67 well-defined
derecho events over the USA during the period of
1983-93. A primary corridor of derecho activity is
evident over the upper mid-western states, with
secondary corridors along an axis from Kansas
through Oklahoma and Texas, and also in the
Figure 8 Schematic representation of features associated with south east. As presented in Figure 11, derechoes are
(A) progressive and (B) serial derechoes near the midpoint of their most frequent during the spring and summer
lifetimes. The total area affected by these derechoes are indi- months over the USA, but can occur at almost
cated by the hatching. The frontal and squall line symbols are
any time of the year. A similar climatology is believed
conventional. (Reproduced with permission from Johns and Hirt,
1987.) to exist for the even more frequent smaller-scale,
shorter-lived bow echoes, although a specific study
documenting bow echo occurrence has yet to be
quasi-stationary frontal boundary, in the vicinity of a undertaken.
mid-tropospheric ridge, and then move along the
boundary (as in the progressive bow echo in Figure 8).
The existence of a low-level jet impinging from the
Numerical and Dynamical Studies
south and flowing along the frontal region, along The tendency for a convective cell to evolve into a
with the polar jet oriented parallel to the front bow-shaped system of cells for certain environments is
farther to the north, leads to moderate magnitudes readily reproduced in numerical cloud modeling
of vertical wind shear in the lower and mid-tropo- studies. Fundamentally, an updraft produces rain
sphere. The addition of enhanced low-level conver- that falls and evaporates, thereby producing a pool
gence along the zone, to help initiate the convection, of cold air that spreads along the ground. This
along with the associated deepening of the moist layer, spreading cold pool produces convergence and lifting
Figure 10 Graphical plot of derecho path and centroid, along with the number and location of proximity soundings, as generated for a
climatological study of well-organized derecho events during the period of 1983-93 over the central and eastern USA. (Adapted from
Evans and Doswell, 2000.)
along its leading edge that can then trigger new cells. convective cells along a line 150 km in length in a
However, rather than a cold pool producing a com- horizontally homogeneous environment with moder-
plete circle of new cells around the initial storm, cells ate CAPE (2200J kg- and strong low-level vertical
are favored along a bow-shaped arc oriented perpen- wind shear (20 m s-' over the lowest 2.5 km AGL;
dicular t o the vertical wind shear vector. The ability to winds are kept constant above 2.5 km). In both cases, a
trigger new cells along this arc increases dramatically line of strong convective cells has become established
as the amount of vertical wind shear increases, and by 3 h. For the non-Coriolis case (Figure 12A),this line
also if the wind shear is confined to the lowest 2-5 km becomes significantly bow-shaped between 3 and 6 h,
AGL. with strong mirror image cyclonic and anticyclonic
Figure 1 2 demonstrates these results for numerical vortices developing at mid-levels behind the northern
model simulations of bow echoes, with and without and southern ends of the system, respectively. With
the effects of the Earth's rotation (Coriolis forcing) Coriolis forcing added (Figure 12B), the northern
included. These simulations are initiated with five cyclonic line-end vortex strengthens over time, while
10
9
8
7
6
5
a 5
5
2 4
3
2
1
0
Jan Feb Mar Apr May June July Aug Sept Oct Nov Dec
Month
I n S F m W F BHybridI
Figure 11 Monthly distribution of derecho events over the central and eastern USA for the years 1983-93, for the events shown in
Figure 10. Events are also subclassified based on the relative strength of the synoptic scale forcing, as either strong forcing (SF), weak
forcing (WF), or hybrid events. (Adapted from Evans and Doswell, 2000.)
the acceleration due to gravity. Within this

the southern anticyclonic vortex weakens, leading t o a
highly asymmetric system configuration by 6 h. The framework, the only source of horizontal vorticity
strengthening of the northern cyclonic vortex is other than the ambient vertical wind shear [e.g.,
y = aU(z)/az, where U ( z ) represents the ambient
directly related to the mid-level convergence of plan-
etary rotation. Figure 12B also demonstrates that a wind profile] is horizontal gradients of buoyancy.
range of bow-echo scales can be produced within the Thus, the analysis of the development of circula-
same convective system, very similar to the observation is simplified to understanding the evolution
tions of the Paducah bow echo case from 5 May 1996 of the buoyancy field and its interaction with
(Figure 5 ) . The tendency t o develop such sub-system-
the ambient shear. The two primary sources of
scale vortices within such simulations increases withbuoyancy within a convective system are the warm
increasing magnitudes and depth of the ambient convective updraft and the cold convective down-
vertical wind shear. draft and cold pool. The net circulation of the
system thus depends on the relative strengths of
Much of the strength and structural characteristics
of bow echo-type systems can be understood the ambient shear, the convective updrafts, and the
by considering the development of a two-dimensional cold pool.
circulation along a vertical cross-section through From the perspective of eqn El], a convective
the core of the convective line. This is most system initially leans in the direction of the ambient
easily accomplished via the two-dimensional hori- vertical wind shear vector (e.g., downshear) as the
zontal vorticity equation for inviscid, Boussinesq warm convective updraft feels the influence of the
flow (eqn [l]),where y = a u / a z - a w / a x , and sheared flow (Figure 13A). However, as the surface
where, B represents the buoyancy, defined by cold pool develops and strengthens over time, the
eqn PI. opposing circulation associated with the cold pool
forces the system to achieve a more upright
(Figure 13B) and then upshear-tilted (Figure 13C)
PI
configuration. Once the system begins to tilt up-
shear, a rear-inflow jet is generated in response to
the buoyant front-to-rear ascending current aloft
and rearward spreading cold pool at the surface.
Here 8 represents the potential temperature, qv, qc, For most convective systems, this rear-inflow jet
and qr represent the mixing ratio of water vapor, descends and spreads along the surface well behind
cloud water, and rainwater, respectively, and g is the leading edge of the convection, enhancing the
Figure 12 Horizontal cross-sections of system-relative flow, rainwater mixing ratio and vertical velocity at 2 km AGL for the
Us = 20 m s- 2.5 km shear (A) non-Coriolisand (B) Coriolis simulations at 3,4.5and 6 h, respectively.Vectors are presented every four
grid points (8 km), with a vector length of 8 km equal to a wind magnitude of 20 m s The rainwater is contoured for magnitudes greater
than 1 g kg-' (lightly shaded) and magnitudesgreaterthan 3 g kg-l (darkly shaded). Thevertical velocity iscontouredat 5 m s-' intervals,
with thezerocontoursommitted.Adomainspeedof urn= 18.5 m s-' has beensubtractedfromtheflowfield.Tickmarksarespaced20 km
apart. (Adapted from Weisman and Davis, 1998.)
surface outflow but generally weakening the convec- pair (e.g., Figure 12A),which focuses and strengthens
tive system. For the stronger-shear, large GAPE bow- the mid-level rear in-flow jet, thereby enhancing
echoes, however, this rear-inflow jet may remain the resultant convective downdrafts and surface
elevated, enhancing the lifting at the leading edge of outflow. The source of this line-end vorticity is
the system, and promoting an even stronger, and more both the downward tilting of the ambient vertical
long-lived convective system (Figure 13D). This con- wind shear layer as well as upward tilting of the
figuration of the vertical circulation and elevated rear- system-generated vertical wind shear associated
inflow jet is quite similar to the 5 May 1996 case, as with the ascending updraft current and descending
presented in Figure 6 . rear-inflow jet.
System strength and severity is enhanced even The processes described above contribute to the
further through the development of the line-end vortex evolution of all convective systems, but produce severe
weather for a relatively restricted range of environ-

mental conditions. Generally, long-lived, severe wind-
producing convective systems, such as bow echoes
and derechoes, are produced in idealized simula-
tions for environments with at least 2000Jkg-' of
CAPE and at least lOms-' of vertical wind shear
over the lowest 2-5km AGL. However, the more
coherent systems with significant book-end vortices
and strong, elevated rear-inflow jets, as presented in
Figure 12, are restricted to environments with at
least 15-20ms-' of vertical wind shear over the
lowest 2-5 km AGL.
See also
Convective Storms: Convective Initiation; Overview.
Density Currents. Gust Fronts. Mesoscale Meteor-
ology: Mesoscale Convective Systems. Microbursts.
Tornados.Weather Prediction: Severe Weather Fore-
casting.
Further Reading
Evans JS and Doswell CA I11 (2000)Examination of derecho
environments using proximity soundings. Weather and
Forecasting 16: 329-342.
Fujita TT (1978) Manual of Downburst Identification for
Project Nimrod. Satellite and Mesometeorology Re-
search Paper No. 156, Department of Geophysical
Sciences, University of Chicago.
-v warm 7
Johns RH (1993)Meteorological conditions associated with
bow echo development in convective storms. Weather
and Forecasting 8: 294-299.
Johns RH and Hirt WD (1987) Derechos: widespread
convectively induced windstorms. Weather and Fore-
casting 2: 32-49.
Johns R H and Doswell CA I11 (1992) Severe local
storm forecasting. Weather and Forecasting 7:
588-612.
Przybylinski RW (1995) The bow echo: Observations,
numerical simulations, and severe weather detection
Figure 13 Four stages in the evolution of an idealized bow echo methods. Weather and Forecasting 10: 203-218.
developing in a strongly sheared, large-CAPE environment. The Weisman ML (1993) The genesis of severe, long-lived
updraft current is denoted by the thick, double-lined flow vector, bow-echoes. Journal of Atmospheric Science 50:
with the rear-inflow current in (C) denoted by the thick solid vector. 645-670.
The shading denotes the surface cold pool. The thin, circular Weisman ML (2001) Bow echoes: A tribute to T.T. Fujita.
arrows depict the most significant sources of horizontal vorticity, Bulletin of the American Meteorological Society 82:
which are either associated with the ambient shear or which are
97-1 16.
generated within the convective system, as described in the text.
Regions of lighter or heavier rainfall are indicated by the more
Weisman ML and Davis C (1998) Mechanisms for the
sparsely or densely packed vertical lines, respectively. The generation of mesoscale vortices within quasi-linear
scalloped line denotes the outline of the cloud. (Reproduced with convective systems. Journal of Atmospheric Science 5 5 :
permission from Weisman, 1993.) 2603-2622.
BUOYANCYAND BUOYANCYWAVES I Optical Observations 323
Contents
Optical Observations
Theory
sion, which is indicative of upward transport of energy

OpUcal Observations from lower-atmospheric sources. In the upper meso-
sphere and lower thermosphere, processes involving
M J Taylor and W R Pendleton Jr, Utah State viscosity and thermal conduction attenuate and dissi-
University, Logan, UT, USA pate energy for waves in the saturation region of
growth. In Figure 1, a simplified sketch captures much
Copyright 2003 Eisevier Science Ltd. All Rights Reserved.
of the basic physics embodied in the modern concept
of a realistic atmosphere permeated by packets of
Introduction buoyancy waves.
During the past four decades, a wide variety of remote
and in situ observational techniques have revealed a
rich spectrum of wave activity in the upper atmos-
-
It is now widely accepted that buoyancy waves with
periods less than 1h achieve major importance in
the MLT region because of the relatively large quan-
phere, with horizontal wavelengths ranging from a tities of momentum (and to a lesser extent energy) that
few kilometers (km) to global scales and temporal they transport to this region. In this connection,
periodicities ranging from a few minutes to about 2 growth of many of the tropospherically generated
weeks. In terms of decreasing spatial-temporal scales, buoyancy waves between their sources and the
the observed spectrum of freely propagating atmos- mesopause region is expected to increase their energy
pheric waves is subdivided into planetary, inertia- (per unit mass) by a factor of -100. A number of
gravity, and buoyancy (or acoustic-gravity) waves. comprehensive experimental and theoretical evalua-
Although physically similar to inertia-gravity waves, tions of gravity-wave forcing have resulted in the
the buoyancy waves are of sufficiently small temporal conclusion that this process has a profound impact on
and spatial scales that the rotation and curvature
of the Earth play minor roles in determining their
properties. Optical observations of these mesoscale
(-10-1000 km horizontal wavelengths) waves are the
primary focus of this article.
A complete spatial-temporal description of the
global distribution of buoyancy wave sources has yet
to be made, but a growing body of evidence indicates
copious sources in the lower atmosphere. This is not
particularly surprising since any perturbation of a
stably stratified, low-dissipation region with frequen-
cy components between the high (Brunt-Vaisala) and
low (Coriolis) frequency limits will generate buoyancy
waves. The propagation of such waves in a realistic
atmosphere will be strongly impacted by the back-
ground winds and temperature structure, producing
such effects as refraction, reflection (complete or
partial), and ducting (partial or complete). Hence,
propagation of tropospherically generated buoyancy
waves to the mesosphere and lower thermosphere Figure 1 Surrealistic representation of several key physical
processes associated with the propagationof internal atmospheric
(MLT) region (altitude range -80-100km) is not gravity waves (CO Hines et a/. The Upper Atmosphere in Motion,
ensured, but the mesoscale waves observed in this Geophysical Monograph 18, American Geophysical Union, Part 2,
region frequently exhibit downward phase progres- Atmospheric Gravity Waves, 194, Figure 1).
324 BUOYANCY AND BUOYANCY WAVES I Optical Observations
the large-scale circulation in the MLT, as well as on the near-perpendicular, ‘billow’ waves. The average
thermal and minor-constituent structures of the re-
gion. In addition, such forcing is expected to impact
variability in the MLT on many spatial and temporal
horizontal wavelengths associated with such defor-
mations lie in the range 5-10 km (billows) and 20--
50 km (bands), the latter of which are typical of the
scales as a result of various types of interactions such as lower buoyancy range. It is interesting to note that
wave-tide and wave-mean flow. current theories for the formation of NLC draw
heavily on gravity-wave forcing from below to provide
the requisite extremely cold near-mesopause environ-
Optical Observations ment during the summer months for successful ice
nucleation and growth.
Optical observations of buoyancy waves in the terres-
It is convenient to distinguish between active and
trial upper atmosphere have been greatly facilitated by
passive optical methods for studying buoyancy waves.
‘nature’s gifts’ of several vertically distinct airglow
The active methods are usually associated with Ray-
layers, alkali metal layers, and noctilucent clouds
leigh and resonant-scatter lidars that probe the vertical
(NLC). Some of the earliest evidence for atmospheric
structure of temperature and/or density fields by
buoyancy waves was provided by wavelike deforma-
studying the return signal from short, very intense,
tions frequently observed during NLC displays. These
gated pulses of monochromatic light scattered by the
-
are very tenuous mesospheric ice clouds (altitude
82 km) that form during the summer months at high
latitudes (typically > 55’) when the mesopause cools
upper-atmospheric medium. In contrast, most passive
methods (with the exception of NLC) utilize the wave-
induced spatial-temporal modulation of the airglow
to the lowest temperature on Earth (110-15OK). As
emissions to detect and study certain characteristics of
the clouds are tenuous, they can only be seen from the
the waves. The passive optical methods include
ground by the scattering of sunlight during the hours
various types of narrowband imaging, photometry1
of twilight, when the observer and the atmosphere
radiometry, and spectrometry/interferometry. Clearly,
below the cloud layer are in darkness while the clouds
the lidar techniques yield direct information on
themselves remain illuminated. (This condition occurs
vertical structure and motion, including the sense of
for solar depression angles between -6“ and 16”.)
vertical phase progression. In contrast, many of the
Optimum locations for observing mesospheric clouds
passive techniques yield direct information on hori-
now and over the past 100 years are Scandinavia1
zontal structure and motion and limited direct infor-
northern Europe, central Asia, and Canada in the
mation on vertical structure when two or more layer
Northern Hemisphere and the southern tip of South
observations are employed.
America in the Southern Hemisphere, where they
are readily seen during the prolonged hours of
twilight. An excellent example illustrating the Lidar Measurements
wave forms frequently observed in NLC is shown
With the development of powerful lidar systems in the
in Figure 2. The photograph clearly shows three
1980s, a new era of active remote sensing of the middle
large-scale NLC ‘bands’ with many smaller-scale,
and upper atmosphere was born. Studies of buoyancy
wave dynamics using lidars have emphasized Rayleigh
scatter methods for investigating structure and wave
motions in the stratosphere and lower mesosphere
(-25-70 km) and resonant scatter (e.g. using sodium)
for studying waves in the MLT region (-80-105 km).
The low-altitude limit (-25 km) for the stratospheric
measurements is determined principally by signal
distortion associated with Mie scattering from aero-
sols and particulates, whereas the upper-altitude limit
(-70 km) results from limitations imposed by signal
photon noise. In practice, each limit is determined
by the characteristics of the particular lidar system
(primarily its power-aperture product) and by
the information retrieval methods used in the data
analysis.
-
Figure2 High-latitude summertwilight photograph of noctilucent
clouds at 82 km altitude showing band-type buoyancywavesand
many smaller-scale billows (Copyright of P. Parviainen, Finland,
Lidar studies of wave dynamics in the mesopause
region have utilized the alkali metal layers (e.g., Na
e-mail: pekka.parviainen@ppl.inet.fi). and K) that are created primarily by the ablation of
BUOYANCYAND BUOYANCYWAVES I Optical Observations 325
meteors in this region of the atmosphere. As Na is 6.9 km and a vertical phase velocity of - 0.33 m s - l ,
much more abundant and has a large back-scatter where the minus sign signifies downward phase
cross-section, it is the most well suited for buoyancy progression, as illustrated in the figure. An inferred
wave studies. In all cases, the lower- and upper- horizontal wavelength of about 450 km follows from
altitude limits of the measurements are determined by these directly measured parameters. More sophisti-
signal photon noise and/or information retrieval cated wind/temperature lidar systems have recently
uncertainties. It is usually assumed that the Na atoms been developed that are capable of measuring temper-
act as a passive tracer of wave dynamics under typical ature perturbations induced by long-period waves and
measurement conditions. However, extensive model- tides (which are forced rather than freely propagating
ing studies of the chemistry of the Na layer and its waves) and studying the background wind field
response to wave forcing suggest that this basic through which the waves progress using thermal
assumption may be invalid for altitudes < 85 km. In broadening and Doppler shifting of the resonance
addition, this assumption can be compromised under line. Together with the wave measurements these data
the special conditions that apply when sporadic layers can be used to estimate the sensible heat flux and
form within the Na layer. This said, lidar studies vertical flux of horizontal momentum transported into
represent one of the most powerful and advanced tools the MLT region.
for sounding the atmosphere.
Current lidar studies of buoyancy waves provide a
direct measure of vertical structure in Na number
density induced by well-developed quasi-monochro-
Image Measurements
-
matic events. Vertical wavelengths in the measurement
range, typically 1 to 20 km (limited by layer thick-
ness), in the mesopause region and -3-30 km in the
Over the past 25 years the capability of imaging
instrumentation for remote sensing faint structures in
the upper-atmospheric nightglow emissions has
stratopause region are common. Quasi-monochro- evolved considerably. Early photographic observa-
matic waves with observed periods in the range of tions demonstrating the existence of wavelike motions
several minutes (close to the local Brunt-Vaisala value j (akin to those seen in NLC) have been superseded by
to several hours (usually limited by the data-record low-light, intensified TV cameras and more recently
length) have been measured. However, in practice by solid state CCD imaging systems that now provide
lidars have proven to be most sensitive to a class of an exceptional capacity for quantitative studies of
waves with short vertical wavelengths (<10 kmj mesospheric wave motions.
which exhibit relatively long observed periods and Images of the naturally occurring nightglow emis-
hence slow vertical phase progression. These waves sions afford an excellent method for investigating the
tend to follow the so-called diffuse-damping limit for horizontal morphology and dynamics of short-period
wave growth. Figure 3 shows a sequence of Na layer (typically <1h ) buoyancy waves. To date, most
profiles ( - 3 min spacing), tracing the downward imaging studies have utilized the bright near infrared
progression of a well-developed long-period (5.8 h) (NIR) hydroxyl (OH) Meinel band emissions that
buoyancy wave observed in the Na layer. This typical
observation yielded a mean vertical wavelength of - -
originate from a well-defined layer (typical half-width
8 km) centered at 87 km. However, there are a
growing number of observations of the NIR 0 2 (0,lj
atmospheric band emission at -865nm and the
visible wavelength 0 1 (557.7nmj and Na (589.2nmj
line emissions. Although these emissions are consid-
erably weaker than the broadband O H emission, they
each exhibit well-defined nighttime profiles at differ-
ent, but closely spaced altitudes in the MLT region.
Table 1lists the properties of the nightglow emission
layers together with the metal layers that are most
frequently used in resonant lidar studies and the NLC
1900 LST 2140 LST
layer characteristics. As the waves propagate and
Figure 3 Sequence of Na lidar profiles (3-min spacing) illustra- dissipate within the MLT region they induce signifi-
ting the downward phase progression of a coherent gravity wave of cant modulations in the line-of-sight brightness (and
6.9-km vertical wavelength and observed period of -5.8-hr. (RL
rotational temperature j of these emission layers which
Collins eta/.(1996) Gravity wave activity in the upper mesosphere
over Urbana, Illinois: lidar observations and analysis of gravity is detected as ‘structure’. Measurements of two (or
wave propagation models. J. Atmos. Terr. Phys., 58, 1905-1 926, more) nightglow emissions therefore provide an
Figure 2). important additional method for investigating the
Table 1 Layer characteristics (full-width at half-maximum), curved waves (like the segments of an orange) due to
except the Na layer which is of full width the format of the all-sky lens. However, when they are
Emission/layer Peak altitude (km) Width (km) mapped into geographic coordinates they often (but
certainly not always) appear as an extensive series of
01 (557.7 nm) 96 6-1 0 quasi-planar waves. In this example the wave pattern
0 2 (0,1) 94 8-1 0
-10
appeared linear and exhibited a horizontal wavelength
Na (589.2nm) 90
NIR OH 87 8-1 0 of 3 8 f 2 k m , and an observed phase speed of
Na layer 92 80-1 05 34 k 3 m s - indicating an observed period of
NLC 82 2 19 -t 2 rnin. The histogram plots of Figure 5 illustrate
the range of values typically observed for bands.
In contrast, the second type of wave motion (termed
ripples) is quite distinct from the bands, exhibiting
much smaller spatial and temporal scales. Ripples
vertical as well as horizontal propagation of short- usually occur in localized wave packets that occupy
period buoyancy waves. much smaller regions of the sky, typically
In particular, all-sky image data (180” field of view) < 5 x l o 3 km2. They have relatively short lifetimes (a
yield unique information on the occurrence frequency, few minutes to -45 min) and almost always exhibit
two-dimensional horizontal spatial characteristics, periods close to the local Brunt-Vaisala period
and the prevailing directionality of small-scale waves ( - 5 min). Two ripple patterns are also evident in
over an exceptionally large geographic area Figure 4A superimposed on the band-type wave
-
(-400 km radius) corresponding to 500 000 km2
at MLT heights with high temporal and spatial
pattern. The similarity between the wave forms seen
in this image and the narrow angle NLC of Figure 2 is
resolution. These cameras are most sensitive to rela- striking.
tively fast-moving waves exhibiting vertical wave- The importance of multilayer measurements is
lengths somewhat greater than the layer thickness (i.e., illustrated in Figure 4B, which shows wave structure
-
> 8 km) and horizontal wavelengths ( I + ) 5-200 km
(Le., significantly less than the maximum field of
imaged in the Na emission layer at approximately the
same time as 4A. (Note that the bright lines are due to
view). In general, lidar and image measurements an Na lidar beam probing two regions of sky during
therefore sample different (but overlapping) regions the 120 s image exposure.) The same band pattern is
of the buoyancy wave spectrum. clearly evident in this emission (and in the OH and 0 2
Optical measurements of the airglow emissions can emissions) indicating that this wave motion was
be made at any latitude and season providing a global, coherent and extended vertically throughout the
all year round capability. Such studies, once limited to MLT region whereas the ripples are absent revealing
the realm of NLC, have revealed a wealth of small- their limited horizontal and vertical extent.
scale wave activity at equatorial, mid- and high A likely source of ripples and billows is the chance
latitudes from many sites around the world, and it is combination of wind and long-period wave motions
not uncommon to observe several different wave (including tides) creating localized regions of strong
patterns during the course of a night suggesting wind shear which then generate small-scale waves in
copious sources. Figure 4 illustrates a variety of situ through the Kelvin-Helmholtz instability. Alter-
wave patterns that are most commonly observed in natively, the waves may be generated by a three-
the MLT emissions. These observations have mainly dimensional convective-type instability which pre-
been made from mid- and low-latitude mountain sites, dicts that the waves should form near-orthogonal to
and distinct spatial and temporal properties have the perturbing wave. In each case, the wave patterns
emerged which suggest the existence of two dominant will be short lived and spatially localized as evident
types of short-period waves, termed ‘bands’ and from the image data. In contrast, the band-type waves
‘ripples’ (akin to those seen in NLC). Figure 4A have been shown to be due to freely propagating or
illustrates the most prominent quasi-monochromatic ducted buoyancy waves most probably of tropospher-
pattern which usually appears as an extensive, coher- ic origin.
ent series of waves that exhibit horizontal wavelengths Figure 4C, D depicts different class of band-type
of a few to several tens of kilometers and horizontal motion termed a ‘frontal event’. Unlike most band
phase speeds up to Band displays are patterns this type of wave is characterized by a sharp
persistent, usually lasting for a few to several hours, leading edge followed by a discrete number of (typi-
and are spatially extensive, often occupying an area of cally < 10)trailing wave crests, similar in morphology
sky much larger than the instantaneous all-sky field of to a bore on a river. This type of wave motion is much
view. In this image the bands appear as a series of less common and is thought to be the signature of a
BUOYANCY AND BUOYANCY WAVES I Optical Obsetvations 327
Figure 4 Example all-sky (180") images illustrating the dominant spatial signatures of short-period buoyancy waves observed
frequently in the MLT nightglow emissions (adaptedfrom MJ Taylor et a\. (1995) All-sky measurements of short-periodwaves imaged in
the Ol(557.7 nm), Na(589.2 nm) and the near infrared OH and 02(0,1) nightglow emissions during the ALOHA-93 campaign. Geophys.
Res. Lett, 22, 2833-2836, Figure 1).
trapped (or ducted) wave propagating near-horizon- the image is due to auroral precipitation. Joule heating
tally at mesospheric heights. In this case, the effect of and other forcing associated with solar-induced mag-
the bore intrusion on the airglow layers was to netic storms are a known source of large-scale buoy-
simultaneously lower the height of the NIR O H ancy waves that are often detected in the high-altitude
(image D) and raise the 01 (557.7nm) layer (image ionosphere as traveling ionospheric disturbances
C) which is manifested as a reversal in contrast to the (TIDs).However, in this case the bands were moving
structuring. towards the auroral zone suggesting other, tropo-
Figure 4E shows another example of band-type spheric-type sources.
wave structure (& = 35.5 k 1.0 km) but this time These examples have been chosen for their clarity to
imaged in the high-latitude (-65" N) O H emission illustrate the types of short-period wave motions that
from central Alaska. The bright arc in the lower part of exist at MLT heights. However, as one would expect
21° N
20" N
'I
0
0 10 20 30 40 50
15WW 157"W 156"W 155"W
Figure 6 Map illustrating a relatively complex wave display

imaged in the Ol(557.7 nm) and 02(0,1) nightglow emissions and
Horizontal wavelength (km) consisting of two quasi-linear wave patterns and one curved wave
train (MJ Taylor eta/. (1995) All-sky measurements of short-period
waves imaged in the 01(557.7nm), Na(589.2nm) and the near
infraredOH and 02(0,1)nightglowemissionsduringthe ALOHA-93
campaign, Geophys. Res. Lett., 22, 2833-2836, Figure 2).
from the quasi-random ensemble of sources and waves

propagating into the upper atmosphere (as depicted in
the sketch of Figure 1)there are oftentimes many more
$ 4 waves evident, resulting in a complex, time-varying
airglow pattern. Such a situaton is shown in the 0 1
z (557.7 nm) image of Figure 4F. In this case two band-
2 type motions progressing in almost orthogonal direc-
tions are evident. A map illustrating the geographical
location and orientations of an even more complex
wave display observed over the Hawaiin Islands is
0 20 40 60 80 100 shown in Figure 6 . The display consists of two quasi-
Horizontal phase speed (ms-l) linear wave patterns and one curved wave train. The
9 temporal evolution of these events reveals the true/
exceptionally dynamic nature of the MLT region.
Figure 7A, B shows two narrow angle images
depicting exciting new evidence of wave break-
down creating turbulent structures. The two images
show a portion of a well-developed band pattern
(%h = 27 km) which appears to increase in contrast, as
the wave becomes nonlinear (image A) and then
breaks into a number of much smaller scale turbulent
features (image B). Such observations are currently
rare, yet it is expected that the breakdown of these
waves to smaller scale sizes eventually resulting in
turbulence and the associated transfer of energy and
0 4 8 12 16 20
momentum into the background medium is a major
Observed period (min)
driver of the MLT region dynamics.
Ground-based, airborne, and, most recently, satel-
Figure 5 Histogram plots illustrating the typical ranges of lite-borne image measurements have been employed
horizontal wavelength, phase speed and observed period associ- to study buoyancy waves. The advantage of satellite-
ated with short-period gravity wave patterns (adapted from Taylor
et a/. (1997) Image measurements of short-period gravity waves
based measurements is their ability to study waves and
at equatorial latitudes. J. Geophys. Res., 102, 26283-26299, their potential sources over remote, inaccessible areas
Figures 4, 5 and 6). on a global scale. This new capability is illustrated in
BUOYANCY AND BUOYANCY WAVES / Optical Observations 329
Figure 7 (A&) Narrow-angle images illustrating evidence of wave breaking in the OH emission leading to turbulent structures (adapted
from Y Yamada et a/. (2001) Breaking of small-scale gravity wave and transition to turbulence observed in OH airglow, Geophys. Res.
Lett., 28, 2153-2156, Figure 1).
Figure 8 MSX satellite observations of upper stratospheric COP emissions illustrating a near circular buoyancy wave pattern likely
produced by a ‘point-like’ thunderstorm located at its center of curvature (EM Dewan et a/. (1998) MSX satellite observations
of thunderstorm-generated gravity waves in mid-wave infrared images of upper stratosphere, Geophys. Res. Lett., 25, 2809-2812,
Figure 2b).
Figure 8, which shows data from the Midcourse Space be obtained directly from the observed temperature
Experiment (MSX) providing the first observations variation. However, in practice, an accurate determi-
from space of gravity waves generated by a thunder- nation of the wave amplitude requires a knowledge of
storm over Indonesia. In this case, the data are nadir the background temperature structure, the distribu-
pointing from the midwave infrared COZ emission at tion of emitters in the layer, and the vertical wave-
4.3 Fm in the upper stratosphere (altitude 40 km) and length of the wave. This information set is seldom, if
show circular wave fronts (1.h =25 km) launched by a ever, available. Furthermore, the column emission rate
‘pointlike’ thunderstorm at their center. Until recently
most satellite measurements have focused on much 10
larger horizontal scale size (&-a few thousand 2 8
3
kilometers), longer period waves that can be readily 3 6
discerned in limb airglow measurements. However, 83 4
E-
this new nadir capability is poised to revolutionize our g 2
understanding of wave generation and propagation in
the middle atmosphere on a global scale. ,g5
.i o
-2
Interferometry, Photometry, and

Spectrometry Measurements
-1 0
Interferometers, photometers (radiometers),and spec- 0 10 20 30 40 50 60 70
trometers continue to play major roles in quantifying Sample number
the line-of-sight-integrated intensity and temperature
Figure 9 Ground-based measurements of OH Meinel (3,l) band
responses Of the layers to buoyancy waves‘ For emission illustrating the intensity and rotational temperature
quasi-monochromatic events, the ‘layer-averaged’ signatures induced by the passage of a short-period (-6-min)
amplitude of the perturbing wave can, in principle, buoyancy wave (courtesy Utah State University).
330 BUOYANCY AND BUOYANCY WAVES / Theory
(CER)response is less directly re,lated to the amplitude I 1%), precise measurements pose a significant ex-
of the induced perturbation. As a result, ratio of the perimental challenge.
fractional perturbations in CER and temperature In their more sophisticated multifield or imaging
(which is independent of wave amplitude) has become modes, these instruments have provided useful infor-
the key parameter in comparing experiment with mation on the horizontal spatial scales and motions of
theory. the waves, in addition to defining the integrated layer
In their simplest operating mode (static, single field), responses to perturbations induced by the waves.
these instruments continue to provide much useful Furthermore, as with imaging studies, measurements
information for testing the predictions of sophisticated using two (or more) vertically distinct airglow emis-
chemical-dynamical models for the layer response. sions (e.g., OH and 0 2 (0,l)bands) provides a means
For example, Figure 9 shows the response of the OH for assessing horizontal and vertical scales of motion
Meinel airglow layer to short-period ( E 5.8 min) and the amplitude of wave growth.
buoyancy waves near the Brunt-Vaisala limit as
measured by a high-throughput Michelson interfer-
ometer. A train of six oscillations clearly illustrates See also
that the high-frequency wave-induced fractional tem- Buoyancy and Buoyancy Waves: Theory. Lidar: At-
perature perturbations are nearly an order of magni- mospheric Sounding Introduction. Middle Atmosphere:
tude smaller than the fractional CER perturbations. As Gravity Waves. Noctilucent Clouds. Optics, Atmos-
the induced perturbation amplitudes for such high- pheric: Airglow Instrumentation; Optical Remote Sensing
frequency waves are expected to be small (typically Instruments.
Theory
T J Dunkerton, Northwest Research Associates,
Bellevue, WA, USA 111
where w is the vertical component of velocity, d/dt
denotes the material derivative following the parcel, g
Introduction is the acceleration of gravity, p is density and O, is
pressure. Linearizing the second and third terms of
Under the influence of gravity, a fluid of variable
eqn [l]about a basic state in hydrostatic balance,
density settles into a configuration in which surfaces of
such that
constant density are approximately horizontal and
density decreases with height. Such a fluid is said to be p=fi+p’ Pal
stably stratified. Stratification is an important proper-
ty of geophysical flows, affecting motions over a wide
range of spatial and temporal scales. The buoyancy of
a fluid parcel (or of a solid object) suspended within a
fluid describes the tendency of the parcel or object to and
move upward or downward in response to gravita-
131
tional and pressure-gradient forces. Buoyancy might
be regarded as an effective ‘force’, but it actually
the equation may be written
represents a combination of these two fundamental
forces. Coherent oscillations resulting from the restor-
ing force of buoyancy are known as buoyancy waves, [41
or gravity waves; these waves play an important role in
atmospheric dynamics. The second term describes the buoyancy of a parcel:
b ’ - g~ rP f
Buoyancy P
The vertical acceleration of a parcel of fluid is implying a tendency for lighter parcels to accelerate
governed by upward, and vice versa, under the influence of gravity.
BUOYANCYAND BUOYANCYWAVES I Theory 331
The buoyancy variable in eqn [4] acts as an effective Buoyancy Waves

‘force’ in the vertical direction.
In a nondiffusive incompressible fluid, density is Gravity waves exist in two forms: internal and
conserved following the motion: external. The most common type of external wave
propagates horizontally along the interface between
two fluids of different density, with amplitude dimin-
ishing away from the interface. Internal waves,
which may be linearized about a state of rest and on the other hand, propagate horizontally and
combined with the definition of buoyancy variable to vertically in a fluid of continuously varying density
give or potential temperature. Internal waves are more
ab’ important in the atmosphere. While propagating
--fW’N2 =0 [71 vertically, gravity waves transport horizontal momen-
at
tum vertically, often over many density scale heights.
where To conserve wave energy, the amplitude of a gravity
wave generally grows with altitude because of the
density decrease, so that a small perturbation (perhaps
undetectable in the lower atmosphere) becomes large
is the Brunt-Vaisala frequency squared, a quantitative in the middle and upper atmosphere. Momentum is
measure of stratification (see Instability: Symmetric irreversibly deposited in a region where the waves
Stability). N 2 is positive in a stably stratified fluid. In attain large amplitude and become locally unstable
this case, N is the frequency of oscillation of a fluid (e.g., at a critical level (see later)), breaking via
parcel about its equilibrium level of neutral buoyancy, convective or shear instability. Vertical mixing of
as can be seen, for example, by combining eqn [7]with heat and constituents also can be attributed to gravity-
the fully linearized version of eqn [4] and neglecting wave breaking.
the vertical pressure gradient force. Gravity waves assume the form of plane waves,
In an ideal gas, potential temperature 0 is conserved consisting of alternating slabs of fluid sliding upward
following the motion (apart from diabatic effects) and or downward at an angle, as illustrated in Figure 1.
may be used as a surrogate for density in the definition They also appear with surfaces of constant phase
of buoyancy. This approximation is valid if the speed radiating diagonally from a two-dimensional source
of sound is large and the effect of the pressure (e.g., a vibrating horizontal rod) or as conical phase
perturbation on density is thereby minimized. The surfaces in three dimensions emanating from a point
corresponding definition of N 2 is source. Waves originating from a point source appear
as concentric rings in a horizontal cross-section, and
191 have been observed in the stratosphere above isolated
thunderstorms.
so that potential temperature increases with height in a Figure 1 depicts a plane wave with phase propaga-
stably stratified atmosphere. In this case, once again, tion downward and to the right. For this wave the
N is the frequency of oscillation of an air parcel about group propagation is upward and to the right,
its equilibrium level of neutral buoyancy. orthogonal to the direction of phase propagation as
Unlike solid objects that are shape-preserving, fluid is generally the case with gravity waves. A downward
parcels are stretched and deformed by the motion; propagating phase is commonly observed in the
buoyant plumes entrain or detrain fluid while mixing middle atmosphere, suggesting upward group propa-
with the environment. The concept of buoyancy, as gation and a tropospheric source. An upward propa-
applied to individual parcels, is therefore of limited gating phase is sometimes observed in the troposphere,
use except in special situations where the integrity suggesting downward group propagation and an
of parcels is maintained. Moreover, individual parcels upper tropospheric source. In other cases, upward
are rarely able to move about without affecting group propagation is observed throughout the tropo-
neighboring ones. Buoyancy remains important, sphere, indicating a source near the surface. Vertical
nonetheless, because it influences the collective motion transport of horizontal momentum is in the direction
of parcels coupled together by the fundamental forces. of vertical group propagation, as implied by the
Stratification affects virtually all classes of atmospher- positive correlation of horizontal and vertical velocity
ic motion. It is particularly important for buoyancy components in Figure 1.
waves - more commonly known as gravity waves - In a simple mean flow without shear, or slowly
which arise in a stably stratified flow as a result of the varying in time and space, the intrinsic frequency
restoring force of buoyancy. of a vertically propagating gravity wave must lie in
332 BUOYANCY AND BUOYANCY WAVES I Theory
E
t
.-m
1
Longitude -.
Figure 1 Structure of an internal gravity wave with phase
propagation downward and to the right. Contours indicate zonal Figure 2 Hines's critical circle for plane waves with horizontal
velocity, vertical velocity, or pressure, with negative values shaded. component of wavevector, k , normalized to unit length in a mean
Rising and sinking motions produce adiabatic cooling and warm- flow U, assumed horizontal. Along the circle, the mean flow speed
ing, respectively. The resulting temperature anomalies are in projected onto the direction of wave propagation is equal to the
quadrature with the pressure anomalies, leading the pressure wave's horizontal phase speed, c.
anomalies by a quarter-cycle.
speed are equal. Waves with upward group propaga-

the range tion, approaching a critical level from below, grow to
large amplitude, overturn, and break in a critical layer
If I < IGI < N POI of finite depth lying below the critical level. In a
compressible atmosphere where density decreases
where f is the local Coriolis frequency. The intrinsic
with height, waves generally grow in amplitude with
frequency is
height, with or without shear, and eventually break.
(2, = w -k.C P11
This situation is illustrated in Figure 3, showing
potential temperature surfaces distorted to such an
where w is the wave frequency relative to the coordi- extent that some of these surfaces are about to
nate system, k is the wavevector, and u is the mean overturn and create conditions for local static insta-
flow, assumed horizontal. The special situation (2, = 0 bility. Parcels within a convectively unstable region
(relevant if f = 0) is illustrated in Figure 2, and is experience exponential growth of displacement am-
referred to as the Hines critical circle, named after its plitude with time, quickly restoring a locally stable
inventor Colin Hines, an atmospheric scientist of the configuration. The process of wave breaking leads to
latter half of the 20th century who has written turbulence generation, momentum deposition, mix-
extensively on the role of gravity waves in the middle ing, and diminution of wave amplitude, and in some
atmosphere. This circle traces out a locus of points for cases, excitation of waves not present in the original
which the component of mean flow velocity in the spectrum.
direction of horizontal wave propagation equals the Waves with intrinsic frequency near f are known as
horizontal phase speed of the wave, c = w / k , where k inertia-gravity waves. The effect of positive f is
is the horizontal wavenumber. Equivalently, for any illustrated in Figure 4.A vertical profile of horizontal
phase velocity vector extending from the origin to a velocity would look something like the hodograph in
point on the circle, the mean flow speed IU1 matches the Figure 5 , with phase increasing downward (m<O).
apparent speed of wave crests in the direction of the Horizontal wind components obtained from a balloon
mean flow. ascending rapidly through this wave would exhibit
When the wavevector and mean-flow vector are clockwise rotation with height, indicating upward
parallel, the wave is said to have a critical level at the group propagation. Velocity data displayed in this
altitude where mean-flow speed and horizontal phase format are informative. The eccentricity of the ellipse
BUOYANCY AND BUOYANCY WAVES I Theory 333
'/ f / / A
zb
'/ f /,
Longitude -
Figure 4 Structure of an inertia-gravity wave in the Northern
Hemisphere, where the Coriolis parameter f is positive. Features
resemble those of the internal gravity wave shown in Figure 1, but
with a meridional component of velocity included. Eastward and
westward zonal wind anomalies generate southward and north-
ward meridional wind anomalies, respectively, in quadrature with
Longitude the zonal wind anomalies, and positively correlated with the
Figure 3 Potential temperaturesurfaces displaced by an upward temperature anomalies.
propagatinginternal gravity-wavepacket experiencingexponential
growth with height in a compressible atmosphere. The wave is
about to overturn and break in the upper part of the domain, approaches f , the vertical wavenumber increases to
beginning at the breaking level, zb.
infinity, i.e., the vertical wavelength approaches zero.
The presence of f 2 in eqn [ 121 implies a larger vertical
( a / b )indicates the ratio of wave intrinsic frequency to wavenumber than would otherwise be the case with-
local Coriolis frequency, while the orientation of the out the Coriolis force. One implication of this result is
major axis indicates the direction of horizontal phase that inertia-gravity waves may break down via shear
propagation, to within 180". This directional ambi-
guity can be resolved if simultaneous temperature data
are available, as implied by Figure 4. In addition to a
vertical flux of momentum, the inertia-gravity wave
contributes to a horizontal flux of heat transverse to
the direction of horizontal phase propagation.
Inertia-gravity wave parameters are related through
a dispersion relation
(I52 - f 2 ) d = (N2 - I52)kZ [I21

where k and m are horizontal and vertical wavenum-
bers, respectively. This formula is valid for waves on an
f-plane, in a constant or slowly varying mean flow,
under the incompressible approximation; it is also
approximately valid for atmospheric waves with
vertical wavelength 2, <<471Hwhere H = - ( I / p ) x
(dpldz) is the density scale height where p is the Figure 5 Hodograph of horizontal wind components for the
horizontally averaged density. Most waves of interest inertia-gravity wave of Figure 4. The horizontal component of the
wavevector points in the positive x-direction, along the major axis
satisfy this inequality. It can be seen from the disper- of the ellipse. The vertical wavenumber rn is negative, so that
sion relation that when the intrinsic frequency phase decreases with height.
334 BUOYANCY AND BUOYANCY WAVES I Theory
instability rather than via convective instability as of frontal zones and regions of strong ageostrophic
wave amplitude increases. acceleration.
According to eqn [12], waves with large intrinsic Gravity waves have several important effects on the
frequency have relatively steep surfaces of constant general circulation of the atmosphere. They can
phase, while waves with small intrinsic frequency have accelerate the flow spontaneously, giving rise to such
relatively flat surfaces of constant phase. That phase phenomena as the quasibiennial and semiannual
surfaces become vertical in the limit G + N is intui- oscillations of the tropical middle atmosphere. Grav-
tively reasonable since N is the frequency of oscillation ity waves also decelerate the flow, reducing the speed
for vertical displacements in a stably stratified fluid. In of the winter polar night jet, summer mesospheric jet,
this limit, the vertical group velocity and vertical and the topside of the tropospheric jet stream. These
wavenumber go to zero, and the waves are reflected waves stimulate weather events in the troposphere,
from the surface where 6j = N. In the opposite limit such as precipitation bands and new convective
6j --t f , the vertical group velocity also approaches elements. Convection therefore is not only a source
zero, but the vertical wavenumber approaches infinity, of gravity waves, but is also triggered by gravity waves
and the waves are likely to break and be absorbed. generated by distant convection.
Gravity waves are sometimes trapped in vertical Simulation of the atmospheric general circulation
waveguides or ‘ducts’ and are able to traverse a large requires either a proper parametrization of gravity-
horizontal distance. Long-period waves have relative- wave effects, or explicit simulation of the waves.
ly small vertical group velocity and are also able to Because gravity waves span a wide range of horizontal
propagate over a range of latitudes, so that the value of scales, it is impractical to simulate the entire spectrum.
f seen by the waves changes slowly with time (see In most models, some combination of explicit
Tropical Meteorology: Equatorial Waves). Near the modeling and parametrization is utilized.
equator, inertia-gravity waves may become trapped
within an equatorial waveguide, forming a modal
structure, or equatorially trapped inertia-gravity
wave. These modes constitute the ‘fast manifold’ of See also
waves on an equatorial beta plane. The intrinsic period Buoyancy and Buoyancy Waves: Optical Observations.
of equatorial inertia-gravity waves can be longer than Convection: Laboratory Models of. Density Currents.
a day, in contrast to mid-latitude gravity waves which Dynamic Meteorology: Waves. Instability: Symmetric
are typically restricted to a range of intrinsic periods Stability. Kelvin Waves. Kelvin-Helmholtz Instability.
from several minutes to several hours. Lee Waves and Mountain Waves. Middle Atmos-
There are several sources of gravity waves in the phere: Gravity Waves. Mountain Meteorology. Oro-
atmosphere, including flow over topography (see graphic Effects: Mountain Waves and Stratospheric
Valley Winds), convection, shear instability (see Kel- Chemistry. Tropical Meteorology: Equatorial Waves.
vin-Helmholtz Instability), and geostrophic adjust- Valley Winds. Wave Mean-Flow Interaction.
ment. Orographic excitation is favored when surface
winds are strong and directed across undulations of
the topography. Convection generates gravity waves Further Reading
through pulsations in forcing or through a ‘topo- Andrews DG, Holton JR and Leovy CB (1987) Middle
graphic’ effect as the convective plume impinges on a Atmosphere Dynamics. Orlando, FL: Academic Press.
stably stratified shear layer. Organized clusters of Baines PG (1995) Topographic Effects in Stratified Flows.
convection also generate gravity waves, but on a much Cambridge: Cambridge University Press.
larger scale than that of individual clouds. Sometimes Gill AE (1982)Atmosphere-Ocean Dynamics. Orlando, FL:
the clusters themselves are organized by inertia- Academic Press.
gravity waves to form a coupled moist dynamical Gossard EE and Hooke WH (1975) Waves in the Atmos-
system. This behavior is observed in the tropics, over phere. New York: Elsevier Scientific Publishing Co.
open ocean (see Tropical Meteorology: Equatorial Hines CO, et al. (1974) The Upper Atmosphere in Motion.
Waves). Shear instability generates larger-scale gravity Worcester: Heffernan Press.
LeBlond PH and Mysak LA (1978) Waves in the Ocean.
waves through envelope radiation. Geostrophic ad-
New York: Elsevier Scientific Publishing Co.
justment has long been regarded as a source of gravity Lighthill J (1978) Waves in Fluids. Cambridge: Cambridge
waves radiating away from a state of initial imbalance. University Press.
A similar process, the spontaneous emission of gravity Turner JS (1973) Buoyancy Effects in Fluids. Cambridge:
waves, is now recognized for its role in evolving large- Cambridge University Press.
scale flows, such as baroclinic instabilities, in which Wallace JM and Hobbs PV (1977) Atmospheric Science: an
regions of imbalance are generated through formation Introductory Survey. New York: Academic Press.
CARBON DIOXIDE 335
C L Sabine, University of Washington, Seattle, WA,

USA 380
R A Feely, NOAA Pacific Marine Environmental

-E,360
340
-
-
Laboratory, Seattle, WA, USA
Copyright 2003 Elsevier Science Ltd. All Rights Resewed.
-
Q
C
320
300
-
-
$ 280 - vbir)+v#+v%v*
C
260 - - Mauna Loa
Introduction 8 240 - A LawDome
0” 220 - v Adelie Land
200 - * Siple
Carbon dioxide (C02) is considered a trace gas in the * South Pole
180 I I I
atmosphere, with contemporary concentrations of I I
approximately 370 parts per million by volume (ppm).

Despite its low concentrations relative to those of
nitrogen or oxygen, C 0 2 plays a significant role in the Figure 1 COP concentrations in Antarctic ice cores (symbols)
Earth’s life cycle and in controlling the global climate. and annual mean concentrations from direct atmospheric meas-
COZ is released as a by-product of aerobic respira- urements (line) for the past millenium. Prior to the industrial
revolution atmospheric COPvalues were very near 280 ppm. For
tion. Plants take up C02 and release oxygen as a part the past 150 years atmospheric concentrations have been
of photosynthesis. Variations in the global balance increasingexponentially. (Adapted from Prentice eta/. (2001).)
between photosynthesis and respiration result in
seasonal variations in atmospheric COZ of up to
15 ppm. For example, atmospheric COZ concentra-
tions in the Northern Hemisphere are generally lower concentrations will be discussed along with a brief
in the summer, when many plant species enter a new introduction to some of the global policy issues
growth stage and photosynthesis predominates over regarding regulation of future CO2 emissions.
respiration. COz is also a greenhouse gas that absorbs
long-wavelength radiation in the atmosphere, attenu-
ating its escape into space. The trapping of radiation Geological History of Atmospheric C02
by C02 and other greenhouse gases (e.g. water vapor,
methane, nitrous oxide, and chlorofluorocarbons) Atmospheric C 0 2 varied considerably over the Earth’s
helps keep the planet warmer than it would be without early history. The balance of several geochemical
an atmosphere. It is this greenhouse warming that processes, including organic carbon and calcium
makes life, as we know it, possible on planet Earth. carbonate burial, silicate rock weathering, and vol-
Mankind is currently in the process of altering the canism, controls the concentration of atmospheric
chemistry of the global atmosphere. Atmospheric C02 C 0 2 over long (millennial) time scales. When one
concentrations have been increasing as a direct result process dominates over the others, such as during
of human activities such as deforestation and the intensified periods of active volcanism, atmospheric
burning of fossil fuels (e.g., coal and oil). Over the past C02 concentrations can change significantly over
150 years, C 0 2 concentrations in the atmosphere have time. For example, during times of high volcanism in
increased by as much as 30% (from 280 to 370 ppm; the Jurassic period about 200-150 million years ago,
Figure 1).This has been accompanied by an increase in C 0 2 concentrations in the atmosphere rose to more
global mean surface temperature of between 0.4 and than 3000ppm. The high COz levels made the rain
023°C. The present rate of increase in CO2 is unprec- more acidic by reacting with the water to form
edented over the last 20 000 years. carbonic acid. The high acid content of the rainwater
This article briefly describes the complicated role eventually caused higher rates of chemical breakdown
that COZ plays on the planet Earth, the different pools (weathering)in carbonate and silicate rocks, releasing
where carbon is stored, and the ways in which carbon basic ions (e.g., bicarbonate and silicate). These basic
is transferred between pools over various periods. We ions traveled via rivers and ground water to the
will focus on those pools (reservoirs) and transfers oceans, eventually increasing the ocean’s capacity to
(fluxes) with time scales relevant to the human absorb C02 from the atmosphere. The patterns of the
alteration of the natural carbon cycle. In addition, large-scale geological processes and the responses of
scientific issues relevant to future atmospheric C 0 2 the global carbon cycle during the early stages of the
336 CARBON DIOXIDE
(1Pg = 1015g), has the smallest total pool of carbon.

The annual atmospheric exchange of C 0 2 with both
the oceans and the terrestrial biosphere is on the order
of 80-130 Pg C y-l. Although these fluxes are very
c
.O_
c
300 large, the mass of carbon in each of these reservoirs
280 may not change over time. Ice core records suggest that
C
260 the atmospheric C 0 2 concentrations were very close
C
8 240 to 280 ppm for the 1000 years prior to the industrial
0"
0
220 era (see Figure 1). This constancy suggests that the
200 carbon pools were more or less in equilibrium, and the
180 net transfer over sufficiently large areas was close to
400 300 200 100 0 zero.
Age (ky before present)
Atmosphere
Figure 2 C02concentrations from the Vostok Antarctic ice core.
The data indicate that atmospheric C o n concentrations were Of the three rapidly exchanging major carbon pools,
generally low (<200ppm) during glacial times and higher during the atmosphere is the most well mixed. Since most of
interglacial periods. (Adapted from Prentice et a/. (2001).) the current C 0 2 sources to the atmosphere are in the
Northern Hemisphere and C 0 2 exchange across the
Earth's history caused large swings in atmospheric Equator is estimated to take about one year, the C 0 2
c02. concentrations in the north are higher than in the
The relatively high atmospheric C 0 2 concentra- southern latitudes (see Figure 3). The fact that most of
tions present over most of the Earth's history were the terrestrial biosphere is located in the Northern
substantially decreased with the evolution of terres- Hemisphere also results in larger seasonal variations
trial vegetation and the subsequent enhancement of relative to the Southern Hemisphere. Data from the
silicate weathering from the decomposition of plant present Global C 0 2 Atmospheric Sampling Stations
matter in soils. For approximately the past 20 million (Figure 4) indicate that distinct C 0 2 sources and sinks
years, the geochemical evidence from the sedimentary are difficult to identify by examining atmospheric
record suggests that C 0 2 concentrations in the concentrations alone because the C 0 2 is quickly
atmosphere remained beiow 300 ppm until the begin- distributed around the globe by the winds. East-west
ning of the twentieth century. Recent studies of C 0 2 gradients of atmospheric C 0 2 concentration are an
trapped in air bubbles preserved in ice cores from order of magnitude smaller than north-south gradi-
Greenland and Antarctica have provided scientists ents. C 0 2 concentrations over the continents are only
with high-resolution records of atmospheric C 0 2 a few ppm higher than corresponding locations over
concentrations for the past 400 000 years. The longest the oceans.
record, from the Vostok ice cores in Antarctica, shows
that atmospheric C 0 2 fluctuations were only about Terrestrial Biosphere
100 ppm, associated with transitions from glacial to
interglacial periods (Figure 2). The Vostok ice core Each year approximately 18% of the C 0 2 in the
record, which spans four glacial-interglacial cycles, atmosphere is cycled through terrestrial plants. This
reveals that atmospheric C 0 2 was low (-180ppm) amount of carbon, or about 120PgC, is converted
during the glacial periods and higher (-280ppm) each year to carbohydrates and plant material during
during the interglacial periods. Natural changes in the photosynthesis. About half of this material forms
global carbon cycle between glacial and interglacial plant tissue (leaves, stems, roots, and wood) and the
periods have maintained the atmospheric C 0 2 levels other half is returned to the atmosphere by autotro-
between these two extremes for at least the past phic respiration ( C 0 2 released by the plants). Global
420 000 years. net primary production (photosynthesis - autotro-
phic respiration) is estimated to be approximately
60 Pg C y-'. Of this amount, approximately 83% is
Controls of Modern Atmospheric COP returned to the atmosphere via heterotrophic respira-
There are only three major reservoirs with exchange tion (decomposition of plant material by bacteria,
rates fast enough to vary significantly on the time scale fungi, and herbivores) and combustion due to forest
of decades to centuries: the atmosphere, the terrestrial fires. The carbon that remains is called the net
biosphere, and the oceans. Of this three-component ecosystem production. Estimates of changes in carbon
system, approximately 93% of the carbon is located in stocks by the US Forest Service suggest that about
the oceans. The atmosphere, at about 750PgC 10 Pg Cy-' is fixed into new plant matter globally
CARBON DIOXIDE 337
3 94' 95' 9 6 7 98'199;

Year
Figure 3 Three-dimensional representationof changes in the latitudinal distribution of atmospheric carbon dioxide as afunction of time.
COP concentrations increase from south to north. The surface shows greater variability in the northern latitudes because of seasonal
changes in the terrestrial biosphere. The increase in COPconcentration with time is the result of anthropogenic emissions. Data are from
the NOANCMDL cooperative air sampling network.
90" N
60" N
GARY KAZAKSTAN ( 2 )
30" N
CAPE KUMUKAHI
0" HRISTMAS ISLAND ,
30" S
EASTER ISLAND AMSTERDAM
TIERRA DEL FUEG

60" S
90" s
100"E 140"E 180" 140"W 1OO"W 60"W 20"W 20"E 60"E 100"E
Figure 4 Map of Global COP Atmospheric Sampling Stations organized through NOANCMDL. Network and observatory stations
provide time-series measurements. Tower and aircraft measurements provide vertical resolution. (Figure provided by P. Tans, N O A N
CMDL.)
Next Page
338 CARBON DIOXIDE
each year. The net ecosystem production suffers Another process that transfers C02 away from the
further losses of carbon from harvesting and erosion surface ocean is termed the biological pump. Photo-
of soil carbon and plant materials, which are eventu- synthetic production of marine plants (phytoplank-
ally transported via streams to rivers, and to the ton) incorporates C02 and nutrients from sea water
oceans. What remains behind, the net biome produc- into living plant tissue and detritus. Microscopic
tion, is the amount of carbon that is accumulated by marine animals, called zooplankton, consume the
the terrestrial biosphere each year. It is this carbon that phytoplankton and provide the basis for the food web
is ultimately lost from the atmosphere as a terrestrial for all animal life in the sea. The gross primary
sink. Estimates of the net biome production averaged production by marine phytoplankton is estimated to
0.2PgCy-' (i.e,, small net carbon uptake from the be about 110 Pg C y-'. Most of this carbon is recycled
atmosphere) during the decade of the 1980s and in the upper ocean via autotrophic respiration, simi-
1.0 Pg C y-l in the decade of the 1990s. larly to what occurs on land. Some of the organic
carbon is transformed into dissolved organic carbon,
which is transported by currents and diffusion to
deeper depths and oxidized by marine bacteria. The
Oceans
remainder of the organic carbon sinks as particulate
The oceans are the largest of the three main C02 matter. The downward transport of dissolved organic
reservoirs, containing about 50 times more C02 than carbon, particulate organic carbon, and detritus
the atmosphere and 19 times more than the terrestrial makes up the bulk of the downward export flux of
biosphere. On an annual basis, the two-way exchange carbon into the ocean interior. Estimates of this global
of C02 between the atmosphere and the surface ocean export production range from 10 to 2OPgCy-'.
is approximately 90 Pg Cy-'. Net exchange of C02 Heterotrophic respiration converts most of this or-
occurs by diffusion when there is a difference in the ganic carbon back into dissolved inorganic carbon
C 0 2 partial pressure (pC02) between the atmosphere (dissolved C02, bicarbonate, and carbonate) at depth.
and oceans. For example, when the atmospheric pCO2 Only about 0.1 Pg Cy-' reaches the sea floor to be
is higher than the surface ocean, C02 diffuses across buried in the sediments. The C02 that is recycled at
the air-sea boundary into the oceans. The oceans are depth is slowly transported large distances by currents
able to hold much more carbon than the other to areas where the waters return to the surface
reservoirs because most of the C02 that diffuses into (upwelling regions). When the waters regain contact
the oceans reacts with the water to form carbonic acid with the atmosphere, the C02 originally sequestered
and its dissociation products, bicarbonate and car- by the phytoplankton is returned to the atmosphere.
bonate ions. The conversion of dissolved C02 to This reequilibration process helps to regulate atmos-
bicarbonate and carbonate ions effectively reduces the pheric C02 concentrations over decadal and longer
pC02 in the water, promoting more diffusion from the time scales. The major upwelling regions are along the
atmosphere. equatorial belt, the Antarctic Circumpolar region, and
The oceans are mixed much more slowly than the various localized coastal upwelling regions. It is this
atmosphere, so there are large horizontal and vertical biological pump that primarily maintains the strong
gradients in C02 concentration. For more than 40 vertical gradient of dissolved carbon in the oceans.
years marine scientists have been measuring the
distribution of pCO2 in the surface waters of the
oceans. A summary of the global data set is presented
COP in the Industrial Era
in Figure 5 . C02 uptake from the atmosphere was Vast amounts of carbon are stored in the Earth's crust
generally found to occur in the high-latitude oceans as as coal, natural gas, and oil. If not for the mining
a result of two factors. First, C02 is more soluble in activities of mankind, this carbon would not be
cold water, so as ocean currents (such as the Gulf recycled back into the atmosphere for millennia.
Stream) transport water from the tropics to the poles Combustion of coal, natural gas, and oil released an
they are cooled and can absorb more C02 from the estimated 5.5 Pg Cy-' into the atmosphere in the
atmosphere. Second, the high-latitude zones are also decade of the 1980s and 6.3PgCy-' in the 1990s
regions where intermediate and bottom waters are (Table 1). Other human activities, such as cement
formed. As the waters are cooled, they become denser manufacturing, also contribute significant quantities
and sink into the ocean's interior taking with them the of C02 into the atmosphere each year. C02 generated
COS accumulated at the surface. This process of as a direct result of human activities is called
transporting C02 from the surface ocean to the deep anthropogenic C02. The release of anthropogenic
because of the cooling and sinking of water masses is C 0 2 increases the atmospheric concentrations by
known as the solubility pump. about 1 % each year. This increase has been well
CHEMISTRY OF THE ATMOSPHERE / Chemical Kinetics 343
Contribution of working group I to the Third Assessment Takahashi T, Wanninkhof RH, Feely RA, et al. (1999)Net
Report of the Intergovernmental Panel on Climate air-sea C02 flux over global oceans: An improved
Change. Cambridge: Cambridge University Press. estimate based on sea-air pC02 difference. In: Proceed-
Sarmiento JL, Hughes TMC, Stouffer RJ and Manabe S ings of 2nd International Symposium on COz in the
(1998) Simulated response of the ocean carbon cycle to Oceans, CGER-1037-99, CGER/NIES, Tsukuba, Japan,
anthropogenic climate warming. Nature 393: 245-249. pp. 9-15.
See PREDICTABILITY AND CHAOS
Contents
Chemical Kinetics
Gas Phase Reactions
Ion Chemistry
Laboratory Studies
Principles of Chemical Change
Rate Laws
Chemical Kinetics Consider the hypothetical chemical reaction [I].
R P Wayne, University of Oxford, Oxford, UK A + B + . . . + products [I1

Experimentally, the rate is found t o be proportional to
the concentrations of A, B, ... each raised to some
power (eqn [l]).
Reaction Kinetics and Atmospheric
Chemistry
Rate = -__ 0 1 = k[A]X [B]B
441 = --
111
dt dt
Chemistry in the atmosphere often consists of several
consecutive and parallel steps that compete with each
other. Interpretation of the rates of chemical change,
Order, Molecularity and Rate Constant
and the concentrations and lifetimes of atmospheric The powers a and p in eqn [ 11are the order of reaction
constituents, requires a knowledge of the rates of the +
with respect to reactants A and B, and a p is the
elementary reaction steps that make up the complex overall order; the constant of proportionality, k, is the
scheme. The kinetic data embodied in this knowledge rate coefficient (rate ‘constant’). The molecularity of a
is best obtained by laboratory experiment, although reaction is the number of reactant molecules written in
theory may have to be used if the experiments cannot the stoichiometric equation. Order is thus an exper-
be performed. Theory is useful in another way as well, imental quantity; molecularity is an arbitrary theoret-
because it can provide a rational basis for the extra- ical one. An elementary reaction step is conceived as
polation of laboratory data to temperatures, pres- one that cannot be split into any chemically simpler
sures, and concentrations that exist in the atmosphere processes. For truly elementary steps, order and
but that cannot be used in the laboratory studies. molecularity are in general identical. Thus, if reaction
344 CHEMISTRY OF THE ATMOSPHERE I Chemical Kinetics
[I] is elementary, and the only reactants are A and B, it In eqn [3], o, is the cross-sectional area for collision
is both bimolecular and overall second-order: first- (collision cross-section), given by eqn [4], and Z is the
order in each of the components A and B. However, a mean relative velocity of molecules for temperature T ,
special case often arises in atmospheric chemistry. If given by eqn [5].
the second reactant, B, is in great excess over A, then its
concentration is effectively constant throughout the oc = n(TA + rBC)2 141
reaction. We can then combine the concentration with
the rate coefficient, k, and write the rate of reaction as 8 k ~ T' I 2 ~ A ~ B C
; p= 151
k'[A], where k' = k[B]. Such a process is termed a mA + mBC
pseudo-first-order reaction, and k' is the pseudo-first-
order rate coefficient. The quantities nA and Y ~ B Cin eqn [3] are the number
densities of A and BC (concentrations in molecular
units such as molecule ~ m - ~Equation
). [3] certainly
The Arrhenius Equation has the correct concentration dependence for an
Many rate constants are found to follow a temperature elementary bimolecular reaction, so that the rate
law embodied in the Arrhenius expression of eqn [2], coefficient can be written
where E , is the activation energy and A is the pre-
exponential factor.
k = o , exp
~ (- j$)
k=Aexp -_
( 2) It is clear that the Arrhenius equation (eqn [2]) and
eqn [6] bear a close resemblance, and E, is commonly
identified with E,, so that the question may be asked
The Arrhenius equation is entirely empirical, but whether a,Z is to be compared with A . However,
several theories of kinetics yield expressions for the it should not be forgotten that 2 is dependent on T1I2
rate coefficient that are similar in form. (cf. eqn [.SI), while A, in the simplest formulation,
is not temperature dependent. A more telling difficulty
The Theories of Elementary Gas-Phase concerns the absolute magnitudes of A and o,c.
For typical atmospheric reactants, with collision
Bimolecular Reaction Steps radii N 400 pm and relative molecular masses
Bimolecular processes are probably the most impor- N 30, ocE is N 3 ~ 1 0 - ~ ' c mmolecule-Is-'
~ at
tant class of reaction and, as we shall see later, 300 K. The product cr,E is called the collision frequen-
termolecular and many very important unimolecular cy factor. Except for the very simplest of reactants,
reactions involve several bimolecular elementary steps. experimental A factors are usually less than, and often
The obvious starting point in discussing the theories of much less than, the collision frequency factor. An
reaction is thus with bimolecular reactions. Two explanation for the lack of agreement is sought in
simplifications are commonly adopted in discussions terms of molecular complexity, with the existence of
of these theories. The first is the collision theory (CT), special geometric arrangements that are needed during
and the second is the transition state theory (TST). the collision to bring reactive parts of the molecules
together (steric requirements), and of special needs for
Collision Theory the distribution of internal energy. That explanation
In simple collision theory, reactant molecules are takes us well away from the idea of hard-sphere
assumed to be hard spheres (radii TA and rBC, say, for reactants.
reactants A and BC), and reaction is taken to be
Transition State Theory
possible only if two conditions are met: (1)a collision
must occur, and (2) the energy of collision along the The alternative simplification adopted in the interpre-
line of centers must equal or exceed the energy tation of bimolecular reactions is that of the transition
required, E,, to reach a critical configuration (ABC#, state theory (TST)or activated complex theory (ACT).
the transition state in Figure 1 ( D ) in the article on The reactants and the critically configured ABC
principles of chemical change; see Chemistry of the molecule are assumed to be in 'quasi-equilibrium'.
Atmosphere: Principles of Chemical Change). The rate Equilibrium constants can be expressed in statistical
of reaction according to this theory is readily shown to thermodynamic terms, and if the formulation is also
be given by eqn [3]. valid for the quasi-equilibrium, where the system is at
a (free) energy maximum rather than minimum, then
-- dnA -
dt
- -- dnBc = nAnBC o,Z exp
dt
(- j$) [3] concentrations of ABC# (the transition state) may be
calculated. Rates of reaction can then be obtained
CHEMISTRY OF THE ATMOSPHERE I Chemical Kinetics 345
from the rate at which ABC' passes to products (as a negative. The most sensible procedure in temperature
result of translational or vibrational motions along the extrapolation thus seems to be first to predict n from a
reaction coordinate). The resultant rate coefficient, k, model of the reaction, and then to fit the experimental
is given by eqn [7]. data to eqn [8] with that value of n.
Activation Energies and Long-Range Forces
Rates of reaction are, in part, controlled by the energy
Partition functions are written as q i , qLc, qlBC+for of a critical (transition-state)configuration, an energy
reactants and transition state, the primes showing that that has as its counterpart the activation energy of
the quantities are volume-independent. The double experimental kinetics. The energy barrier arises be-
prime indicates that the motion along the reaction cause the reactant molecules are forced close together
coordinate has been factorized out (and a numerical (closer than the sum of their radii in the hard-sphere
constant has been introduced). In TST, then, the collision approximation), and reactant bonds have to
internal motions neglected in collision theory (CT)are be broken while product bonds are made. The energy
required is less than that required first to break
expressly taken into account through the use of the
partition functions. TST concentrates only on that reactant bonds and then to form product molecules
region of the potential energy surface around the in separate steps. The energy does not decrease at any
transition state (for the calculation of the partition stage in this picture as the system passes from
function qiBc+)while CT is interested only in the separated reactants to the transition state. Such a
height of the energy barrier at the transition state. It is decrease in energy would correspond to long-range
the calculation of qiBc+that offers most difficulty in attractive forces, and might lead to an increased
the practical implementation of TST. Spectroscopic collision frequency, and to an A factor that exceeded
parameters for the reactant molecules are usually
ocE.Many examples of this type of behavior are in fact
available, so that q;, qLc are readily estimated. known, even with neutral reactants, but the effects are
However, a knowledge of the shape of, and the forces strongest and most common with charged reactants.
acting at, the transition state would imply that the In ion-molecule reactions such as reaction [11], the ion
can induce a dipole in the neutral reactant, and the
potential energy surface is itself known, at least in the
region of ABC'. The usual practice is to make an resultant attractive force can both balance the ordi-
'informed guess' at the magnitude of qiBc+based on a nary chemical activation barrier as well as make the
hypothetical interaction mechanism and a corre- real encounter rate greater than the gas-kinetic colli-
sion frequency factor for neutral molecules.
sponding model for the transition state. Considerable
differences in predicted pre-exponential factors are
obtained from models of the transition state that are,
o++ co2 -+ 0; + co PI1
for example, linear, bent, or cyclic. In a more limited Near-zero activation energies are thus often found
way, TST can suggest a sensible order of magnitude for
the pre-exponential factor. The three total partition
functions in eqn [7]are each the product of transla-
(typically -
in this type of reaction, and the pre-exponential factors
io9cm3 molecule- s - l ) are several
times larger than the values for neutral reactants.
tional, rotational, and vibrational partition functions. Because the long-range attractive forces dominate the
The translational parts can all be calculated, and potential energy, high velocities of approach are
orders of magnitude for rotational and vibrational counterproductive in promoting reaction, and some
parts can be employed in accordance with the number negative temperature coefficient of rate constant may
of each of these modes that exist in A, BC, and ABC# . be observed. The stronger (or longer-range) the
The temperature dependence for every partition func- interaction, the larger the rate coefficient. For ion
tion can be evaluated as a power law, so that eqn [7] reactions with neutral molecules possessing perma-
can be rewritten in the form of eqn [8], where A' is the nent (rather than induced) dipoles, pre-exponential
temperature-independent part of the pre-exponential factors are increased by another two or three times.
function, and n is some exponent chosen from the Thus charge transfer from O+ to the dipolar mole-
nature of the reactants (monatomic, diatomic, etc.) cule H20 (reaction [III]), has a rate coefficient of
and a model of the transition state. 2.3 x 10-9cm3 molecule-ls-l at 298K, and the
(- g)
activation energy is essentially zero.
k = A'T" exp
For the hard sphere (CT) case, n = 0.5, from eqns [5] The long-range interactions are yet larger, of course,
and [6]. In the more general case, n can be positive or for two reactants both of which are charged. Positive
ion-negative ion, or positive ion-electron reactions formation. A steady state for [XI has been reached,
are characterized by rate coefficients 3-4 orders of and d[X]/dt + 0. To illustrate the stationary-state
magnitude larger than typical gas-kinetic collision method, consider the pair of reactions [VI and [VI].
frequency factors. For example, the rate coefficient
(298K) for neutralization of NO+ by an electron A+B +X kf formation [VI
(reaction [IVI)is 4.5 x 1 0 - ' c m ~ m o l e c u l e - ~ s - ~ .
X+C+D k, loss [VI1
NO' + e -+ N +0 [IVI
For simplicity, let [B] and [C] be in great excess, so
that we may write pseudo-first-order rate coefficients
k; = kf[B] and k[ = k,[C]. Equation [9] is the kinetic
Multistep Reactions and the Stationary equation that describes reactions [VI and [VI].
State Hypothesis
Atmospheric chemistry consists of complex interac- = k;[A] - k:[X] 191
dt
tions of elementary reactions. Some of the processes to
be described in the following section on unimolecular If X is in a stationary state, then we set the differential
and termolecular reactions also involve several steps, equal to zero and obtain eqn [IO], where [XI,, indicates
which is why the subject of multistep reactions is a steady-state concentration of X.
introduced at this stage.
Consecutive and parallel steps involve reactive
intermediates in competitive processes. Reaction
intermediates of particular interest include atoms, The problem is to know whether the concentration of
radicals, ions, and excited species. Most of these X calculated using the SSH bears any relationship to
intermediates are highly reactive and, with one or two actual concentrations. Our two-reaction example has
exceptions, cannot be 'stored' in a laboratory for long been chosen because it can also be solved analytically.
periods because they are lost on the walls of the So long as [A] and [C] are independent of time, eqn [9]
containing vessel or react with each other. Such can be integrated to yield eqn [ 111,where t is the time
intermediates are not necessarily unstable, and chem- for which the system has been reacting.
ical lifetimes of isolated atoms or radicals in the
absence of surfaces can be virtually infinite. Many 441
[XI = -[I - exp(-k:t)]
excited-state species are unstable, since they may k:
possess enough internal energy to fragment, and they This expression for [XI approaches the steady-state
may also be able to lose their energy by emission of expression so long as k ; t >> 1,the error in applying the
radiation. An excited species that cannot undergo loss SSH being less than one per cent for kit > 4.6.The SSH
by an allowed radiative transition is said to be can thus be applied so long as [A] and [C] remain
metastable. constant over a period long enough for this inequality
Multistep reaction schemes are interpreted kinetic- to be reached. It is evident that the circumstances
ally by writing down the differential equations, such as under which the SSH is most likely to be valid are thus
eqn [l],for all the species of interest, including the those where k; is large: that is, if the species X is highly
intermediates. Solution of these equations then allows reactive. One example of a species at steady state in the
prediction of the concentration-time variation of each atmosphere is the highly reactive state of atomic
of the species. Unfortunately, analytical solution of the oxygen, O('D), throughout the troposphere, strato-
many simultaneous differential equations is rarely sphere, and probably the mesosphere. Ground-state
possible. Numerical solution has become a widely atomic oxygen, O(3P),however, cannot generally be
used alternative since the advent of high-speed com- treated in the atmosphere by steady-state methods
puters and the development of good techniques for because of its relatively small reactivity.
dealing with differential equations. For some highly
reactive intermediates, the stationary-state hypothesis
(SSH) (often alternatively called the steady-state Theories of Unimolecular and
hypothesis) provides a simplification that will permit
algebraic solution of the kinetic equations. Consider
Termolecular Reactions
an intermediate X that is created in a process whose We are now in a position to consider thermal uni-
rate is constant, and whose loss-rate increases with molecular reactions, and their close counterpart,
increased [XI. After the reaction is started, [XI will termolecular reactions. If chemical reaction requires
increase until the rate of loss is equal to the rate of collision between, or at least close proximity of, the
reactants, then it might seem that all thermal processes first-order kinetics follow. It is obvious, however, that
ought to be kinetically of second order. Unimolecular, at sufficiently low concentrations of AB, there comes a
first-order, elementary processes appear to lack the point at which reaction [IX] is rate-limiting, and the
necessary approach of reactants, and termolecular, kinetic behavior will be second-order. Transition from
third-order, steps suffer from the impossibility of a first- to second-order behavior is, indeed, seen at low
simultaneous collision between three hard-sphere enough pressures in this kind of thermal unimolecular
reactants. The explanation for first- and third-order reaction. Quantitative expression of these ideas can be
thermal kinetics shares common ground, and a simple obtained by a steady-state treatment for the concen-
introduction is provided here. tration of ABt as described in the previous section. The
N o obstacle exists to understanding how single-step result for the rate of loss of AB is eqn [12], where k~ is
unimolecular and first-order decomposition occurs in the experimentally defined pseudo-first-order rate
a molecule AB that already has more than enough coefficient.
energy in it to break one of its bonds. An obvious case
is the photodissociation of polyatomic molecules,
where optical dissociation or predissociation popu-
lates vibrational levels of AB sufficiently to cause
fragmentation (see Chemistry of the Atmosphere: So long as kd[AB] >> k,, the reaction is first-order, but
Principles of Chemical Change). The rate of fragmen- if [AB] is reduced to the point at which the reverse
tation may depend on the rate at which energy can inequality holds, then the reaction becomes second-
accumulate in the bond to be broken, but the reaction order. At high concentration, the limiting value of k1
will be kinetically of first order. Chemical activation (referred to as k,”) is equal to ( k a k r / k d )and is thus
offers another route to high vibrational excitation. For truly first-order, being independent of [AB]. The low-
example, the reaction of HO2 with N O (reaction [VII]) pressure limit, kp, is equal to ka[AB] and is itself first-
can produce a highly excited H 0 2 N O t molecule (the order in pressure, or second-order overall.
dagger representing vibrational excitation). Considerations of high- and low-pressure extrapo-
lations of rate data are most frequently met in
HO2 + N O + HOINOt [VI11 atmospheric chemistry in connection with termolecu-
lar reactions. As with unimolecular reactions, termo-
This excited H 0 2 N O t can then either split up to the lecular processes have orders that are variable with
reactants again, or form O H andNO2 (reaction [VIII]). pressure, being third-order at ‘low’ pressure and
second-order at ‘high’ pressure. Such reactions
H02NOt 4 HO + NO2 [VIII]
are extremely important in combination processes
Reaction of the excited H 0 2 N O t is a unimolecular, (sometimes erroneously called ‘recombination’ proc-
first-order, elementary reaction. esses), and we can see why by first looking at the
It is in interpreting thermal unimolecular reactions reaction of two atoms to form a diatomic molecule. A
that some difficulty arises, since the formation of an typical case is the combination of two O(3P)atoms.
excited ABt molecule involves collisions between the Curve ‘X’ in Figure 2 of the article on principles of
AB species, and might therefore be expected to show chemical change (see Chemistry of the Atmosphere:
second-order kinetics. A basic understanding was Principles of Chemical Change) illustrates this com-
provided by Lindemann, who suggested that thermal bination. Even if the combining atoms have no relative
first-order reactions were not true elementary steps, translational energy, the newly formed 0 2 molecule
but rather involved at least three elementary processes has the 0+0 combination energy stored in it:
that energy is the 0-0 bond energy, and the 0 2 is
AB + AB 4 AB^ + AB chemically activated 0; at its dissociation limit.
Unless some energy is removed within one vibrational
collisional activation 11x1 period, the molecule will fall apart again as
the internuclear distance increases on the first oscilla-
ABt + AB 2 AB .tAB deactivation [XI tion. Energy can be removed in collisions; the species
that dissipates energy is often represented by the
symbol M. In the atmosphere of Earth, M is usually
AB^ 4 A + B reaction [XI1 the ‘bath’ mixture of N2 and 0 2 . The overall reaction
If reaction [XI dominates as a loss process for ABt over is now written as eqn [XII], which is a termolecular
[XI], then the concentration of ABt is almost at its step.
thermal equilibrium value, while the rate-determining
step for reaction is the first-order process [XI]. Overall 0+0+ M -+ 0 2 +M
348 CHEMISTRY OF THE ATMOSPHERE / Chemical Kinetics
The redissociation that has been prevented is the atmospheric pressure range.
unimolecular dissociation of 0; equivalent to step
[XI], and the process deactivating 0; is the equivalent OH + NO2 + M H N 0 3 + M
4 [XVII
of step [XI. We shall see shortly that internal energy
seems to flow fairly freely between different vibra- 0+0 2 +M 0 3 +M 4 [XVII]
tional modes of a polyatomic molecule. If the newly Equations [ 121 and [ 131 represent the variations of
formed molecule is larger than diatomic, there are such experimentally determined rates of reaction with
modes into which the bond-combination energy can pressure. The pseudo-first-order or pseudo-second-
flow. The lifetime of the newly formed molecule can order rate coefficients k~ or k11 can be conveniently
thus correspond to many vibrational periods before expressed in terms of the high- and low-pressure
the energy flows back to the critical bond. With a large limiting values k r , kp, or k r , kp,. For example, k11 in
enough polyatomic molecule, the lifetime can be so eqn [13] can be expressed as in eqn [14].
great that collisional removal of excess energy (stab-
ilization) is no longer rate-determining, and combina-
tion then exhibits second-order kinetics. Because
reactions [XI and [XI] are common to both unimolec-
ular and termolecular reactions, the same general Remembering that kpl is itself first-order in pressure, it
considerations about flow of energy apply to both can be seen that eqn [14] represents in outline the
types of process. variation of k11with pressure that is found experimen-
The analogue of eqn [12] can be derived from the tally. Unfortunately, however, the equation does not
single excitation level kinetic scheme of reactions match experimental data in detail, so that it cannot be
[XIIII-[XV]. applied directly to the calculation of rates at interme-
diate pressures. The reasons for the failure are known.
A + B 5 ABt combination [XIII] The reactions and the rate coefficients k, or k,, k d or k,,
and k, should have been defined for each individual
ABt + M 4AB + M stabilization [XIV] quantized vibrational level of ABt, and the individual
rates summed to give the total rate. It is, perhaps, easy
to see that the more energy available (beyond the
AB+5A + B reaction [XVI critical amount needed to break a particular bond), the
more rapid will be the fragmentation (i.e. the larger
The result is eqn [13], where ,411 is the experimentally
will be k,). Related to this point is the implication that
defined pseudo-second-order rate coefficient (analo-
energy stored in any vibrational mode can be made
gous to k~ in eqn [12]).
available to the critical bond. Experimental evidence
largely favors the flow of energy between modes as
being fairly free, and the distribution as being near
statistical. An additional complication involves the
We see straightaway that, if k, >> k,[M], the reaction interconversions of vibrations and rotations in the
is third-order, with kFI = (k,k,/k,)[M]. If, how- fragmenting molecule. The theory has been extended,
ever, k, << k,[M], then the reaction is second-order, modified, and manipulated over the years by Rice,
with k E = k,. Increased complexity in the molecule Ramsperger, Kassel, and Marcus, and the familiar
AB reduces the value of k,, because the combina- initials RRKM are used to designate their formulation.
tion energy is distributed among more vibrational With sufficient sophistication of the input informa-
modes. The concentration, or pressure, of third- tion, very good agreement can be obtained between
body M at which third-order behavior turns theory and experiment. Correspondingly, one could
over to second-order kinetics is thus lower the more have confidence in the extrapolation of data obtained
complex the molecule produced. ‘Complex’ is only in an intermediate concentration regime to either high-
a relative term here: combination of two hydrogen pressure (first-order) or low-pressure (second-order)
atoms to form H2 is third-order up to 104atm, limits. However, application of RRKM theory to real
while combination of two CH3 radicals to form processes of atmospheric importance is in practice
CZH6 is second-order at all but the lowest pressures. rather difficult, and an alternative, much simpler,
However, it so happens that the reactants in several approach is now almost universally adopted. This
combination reactions of great atmospheric impor- approach has its origins in work by Troe on the
tance, such as reactions [XVI] and [XVII] are of theoretical prediction of unimolecular reaction rate
just such molecular size that complex intermediate- parameters. However, with k r , k:, known, Troe has
order kinetics are displayed at some point in the shown that a simplification of his theory allows the
CHEMISTRY OF THE ATMOSPHERE /Chemical Kinetics 349
right-hand side of eqn [14] to be multiplied by a explicitly the steps involved in a heterogenous reaction
broadening factor, F, that is a function of (kpI/kr). For of a gas-phase species either with the bulk constituent
many atmospherically important termolecular reac- of a liquid droplet or with another species that is
tions, F may be calculated from a simple mathematical already dissolved in it. The uptake of gas-phase
expression (eqn [15]),in which Fcentis the broadening molecules can be either reactive or nonreactive.
factor for the center of the fall-off curve; typically, Chemical change corresponds to loss of the gas-phase
FCent= 0.6. molecule; uptake that is nonreactive can arise from
physical dissolution or from reversible chemistry.
Individual steps that can be envisaged are (1)gas-
phase transport of the reactant to the surface of the
Third-order reactions often show decreasing rate droplet; (2) accommodation at the surface; ( 3 )diffu-
with increasing temperature: they have a negative sion into the liquid; (4)chemical reaction; ( 5 )diffusion
temperature coefficient. The reason is that the larger of unreacted molecules and products to the surface;
the thermal kinetic energy possessed by the reactants A and (6) desorption of species from the interface.
and B in process [XIII], the more internal vibrational Characterization of each of these individual steps is
energy will be stored in the ABt molecule produced. As obviously a formidable task, although one that may be
pointed out earlier, the chance of the critical bond simplified - as often happens in kinetics - by one of the
energy finding its way back to a breakable bond is thus steps being rate-determining.
increased, and k, is larger. Since k, and k, are only In many respects, the kinetics of liquid, surface, and
slightly affected by temperature, it follows from eqn heterogeneous reactions are governed by the same
[13] that the rate of reaction will decrease with principles that we have established for gas-phase
increasing temperature. Thermal energy in effect processes. There are, however, some key differences.
assists the newly formed molecule to split up again, Reactions occurring inside particles are really confined
thus slowing the rate of combination. In the third- to the liquid phase, since diffusion coefficients within
order limit, kp, is inversely proportional to k, (see solids are too small to allow significant reaction rates.
above). Theory suggests that the temperature varia- On the other hand, reactions on solid surfaces are
tions of k, should be better expressed in terms of a thought to be of very considerable atmospheric
power, T", rather than as a conventional activation significance. The PSCs involved in stratospheric
energy. Hence experimental measurements of kf:as a chemistry may consist, for example, in part, of solid
function of temperature should be fitted against a T-" water-ice and solid nitric acid trihydrate. A convenient
law to allow rational interpolation or extrapolation to starting point in the present discussion will thus be an
atmospheric temperatures. Typical measured values of examination of surface reactions themselves. An
n are 2.5 to 3.1 for reaction [XVI] and 1.7 for reaction added degree of complexity arises when the particle
[XVII]. Models of the transition state for bond is liquid, as is the case for droplets in the troposphere,
association reactions also suggest that, at the high- and possibly for stratospheric sulfate aerosol, which
pressure limit, k11 should possess a negative exponent may be in the form of supercooled liquid sulfuric acid.
of temperature. Formal kinetic equations for reactions on the
surface of atmospheric particles often start from the
Condensed-Phase, Surface, and Langmuir adsorption isotherm, which is the simplest
of equations that expresses the partitioning of gas
Heterogeneous Reactions between surface and gas phases. The isotherm makes
Reactions within the liquid droplets of clouds and fogs several assumptions, including one that states that all
are important in several aspects of tropospheric surface sites are equivalent and that there are no
chemistry, such as the oxidation of sulfur dioxide. interactions between molecules adsorbed on them.
Liquid or solid particles can play a critical role in the With these assumptions, it may be shown that the
chemistry of the stratosphere under certain conditions. surface coverage, O X , is given by eqn [16].
Particles such as those of sulfate aerosol or clouds
formed from water-ice and hydrates of nitric acid
(polar stratospheric clouds, PSCs) are implicated in
such processes. This chemistry may involve surface
reactions or reactions within the bulk material, but the Here b is a constant equal to the ratio of rate
interface between gas and condensed phases is in- coefficients for adsorption onto the surface and
volved in some way, and the reactions are thus known desorption from it. For the low partial pressures of
as heterogeneous reactions. The possible complexity adsorbates present as minor constituents in the
of the kinetics can be illustrated by considering atmosphere, Ox is likely to be very nearly a linear
function of p x , although the full equation might be Uptake coefficients may be determined by a variety of
needed for very strongly adsorbed reactants or for high experimental methods. Regardless of whether the
partial pressures in the atmosphere. An extended molecule is removed by reaction on or within the
treatment for the situation where two different particle or by dissolving in it, eqn [21] provides the link
species, X and Y, are adsorbed is straightforward, between the kinetics of the uptake process and the
with the surface coverages OX and By both entering uptake coefficient. It will be evident that the reactive
into the equation. uptake coefficient is equivalent to the reaction
The kinetics of reaction for a single reactant are then probability.
developed by including a loss (probably decomposi- A complication obviously arises if a molecule does
tion) of the adsorbed molecules. If the rate coefficient not react irreversibly, but can desorb again from a
for this first-order loss process is k ~ the , rate of surface, or come out of solution to re-enter the gas
chemical change given by eqn [ 171. phase. In such cases, y can apparently be time-
dependent, and the measurement of the variation of
Rate = k L O X N kLbxpx 1171 y with time provides one way of examining these
reversible processes. In the case of the atmosphere,
The second (approximate) equality applies for low Ox. the most important aspect concerns the partitioning
The overall kinetics are first-order in p x (and thus [XI ).
of molecules between gaseous and liquid phases.
Analogous equations can obviously be developed for Solubilities of gases at low solute concentrations
the case of X and Y both adsorbed on a surface, and
obey Henry’s law (eqn [22]).
interacting there. If both species are weakly adsorbed
(small surface coverage), the kinetics will be second-
order, with a rate proportional to pxpy. If one (or
both) of the reactants is adsorbed too strongly to use [X(s)]is the concentration of X in solution, p x is its
the low-pressure limiting equation, then the full form pressure in the gas phase, and H X is the Henry’s law
of the adsorption isotherm must be employed. coefficient (which is a function of temperature).
A key parameter in the discussion of surface Henry’s law expresses an equilibrium situation, in
processes is the uptake coefficient, y, which is the which the fluxes of molecules into and out of the liquid
ratio of molecules lost to a surface to the number of are equal. However, it is straightforward to calculate
gas-surface collisions that occur. If the rate of collision the forward and reverse fluxes, and thus the net flux
of a molecule X with an area A of the surface is o, into the liquid, under nonequilibrium conditions. The
then the rate of loss of X per unit volume, -d[X]/dt, is first term comes immediately from eqn [20], while
equal to y w / V , where V is the volume of the system. the second requires use of the diffusion equation
The kinetic theory of gases shows that w is given by for transport of the molecules from the bulk liquid to
eqn [18]. the interface. If the coefficient of diffusion for this
latter process is D, then it may be shown that eqn [23]
EA [XI holds, where yo and y t are the uptake coefficients at
w=-
4 time 0 and time t.
Equation [ 191 follows from this expression for o.
4x1 -
-- --YWXI
dt 4v The equation shows how HX can be calculated from
measurements of uptake coefficient as a function
Now the loss of X may also be described in terms of time, or, conversely, how the variation of up-
of phenomenological rate equations of the type of take coefficient with time may be estimated from a
equation [20], where IS} represents the number of knowledge of the solubility of the gas. Note that at
active surface sites per unit area and ks and k’,are the
‘infinite’ time, yt becomes zero: the system has reached
second-order and the corresponding pseudo-first- equilibrium.
order rate coefficients for the surface loss process. The material developed so far is applicable both to
physical processes - adsorption, absorption, or solu-
d[X1= kS{S}[X] E
-- ki[X] 1201 tion - and to chemical change. In the particular case of
dt
chemical change, we can envisage two possibilities.
From a comparison of eqns 1191 and [20] Either the reaction may involve an interaction of the
gas-phase reactant with the surface or the bulk
constituent of the particle, or it may involve reaction
1211 with some second species already adsorbed on, or
dissolved in, the particle. The concepts set out earlier For many of the liquid-phase reactions of interest in
remain applicable in the second, 'bimolecular', situa- atmospheric chemistry, the intrinsic reactivity of the
tion, but the value of {S} at the surface (eqn [20])or the partners is, indeed, very high, leading to diffusion-
concentration of the partner reactant Y, in solution, controlled kinetic behavior. A very elementary treat-
will be determined by factors similar to those already ment of the diffusion-controlled rate constant, k d ,
determining the adsorption or solubility of Y. leads to eqn [24].
Finally, it is necessary to examine the kinetics
of reaction within the liquid phase itself. The solvent k d = ~WAB DAB ~ 4 1
obviously has the potential to exert a considerable
where TAB is a hypothetical encounter distance at
influence on the course of chemistry in the liquid
which two partners A and B will react, and DABis the
phase. In air at l a t m pressure, and at ambient diffusion coefficient for the reactants.
temperature, the molecules themselves occupy only The encounter distance may be roughly the sum of
roughly 0.2% of the total volume; in liquids, the gas-kinetic radii of the partners for neutral
the molecules can make up half the volume.
reactants, while the appropriate diffusion coefficient
At pressures of 1atm and below, we have been able may be similar to a mean bulk diffusion coefficient of
to assume that the reactant molecules undergo essen-
the reactants in the solvent. Making these assumptions
tially unhindered motion, and that assumption lies
and with typical values of TAB = 0.5nm and DAB=
behind the various formulations of kinetics that we 1 . 3 ~ 1 m2 0 ~ s-l~ (for Na+ in H20), kD is calculated
have discussed in previous sections. In distinction, in as N 8 x m3 molecule-' s-' or, in the units that
liquids the reactive molecules must squeeze past the we have been using for rate coefficients so far,
solvent molecules (or each other, if one species is also
8x cm3 molecule-' s-'. In liquid-phase kinetics,
the bulk liquid) if they are to reach each other and it is more conventional to use molar units for concen-
undergo reaction. Reactants, activated complexes or
trations, so that the equivalent figure is ( 6 ~ 1 0 ~ ~
intermediates, and products can also all interact with 1 0 3 x 8 x 1 0 - ' s ) z 5 x 109dm3mol-ls-'. In whatever
the solvent. One manifestation of the interaction with units this rate coefficient is expressed, it is evidently
intermediates is that energy removal in association about 40 times lower than the maximum gas-kinetic
reactions, such as the combination processes [XVI] or rate coefficient. In general, a rate coefficient of
[XVII], is virtually instantaneous and the systems > IO9dm3 mol-' s-' for an aqueous-phase reaction
always display pure second-order kinetics in the liquid is taken to be indicative of a diffusion-controlled
phase, in contrast to the behavior described in the last mechanism.
section for gas-phase reactants. Interactions of the One of the largest known rate coefficients for a
reactants and the solvent (especially water) may make condensed-phase process is that for the very important
the formation of ions energetically more favorable reaction [XVIII] (1.4 x 10" drn3mol-'s-' at 298 K).
than in the gas phase. New reaction channels may thus
become accessible, and the kinetics of the processes H+ +OH- + H20 [XVIII]
can be influenced by the attractive or repulsive
electrostatic interactions between the reactants, The magnitude mainly reflects the large diffusion
among many other factors. coefficients in water of OH- and, especially, of H'; the
Simple treatments of liquid-phase kinetics often rapid diffusion is itself a consequence of the special
start from the concept of the encounter pair of reactant mechanisms by which these ions migrate in liquid
molecules that find themselves together within a H20.
solvent cage. Two extreme cases can be envisaged. In Although the diffusion coefficient is most important
the first, the two species are very highly reactive in making reaction [XVIII] so fast, there is another
toward each other, and undergo chemical transforma- factor operating that may be dominant in other
tion within a very few 'collisions' within the cage. The reactions. The positive and negative ions attract each
rate-determining process is then the diffusion of the other, so that the effective encounter distance can be
reactants through the solvent to form the encounter much greater than the gas-kinetic collision distance;
pair, and the process is a diffusion-controlled reaction. that is, TAB has to be replaced by reffin eqn [24]. For
At the other extreme, the activation energy for TAB = 0.5 nm, straightforward electrostatic calcula-
reaction may require the partners to pick up appreci- tions indicate that reffought to be about 0.2 nm for
able amounts of energy as they shake against each oppositely charged ions (and 0.7 nm for like charges)
other within the cage, so that the kinetics are control- in water with a relative permittivity of 78. However, it
led by the rate of reaction within the cage, rather than seems that this bulk permittivity is inappropriate to the
by the rate at which they reach it. Activation control- highly ordered solvent molecules in the immediate
led reaction kinetics then result. vicinity of the ions, and that more realistic values of r,ff
352 CHEMISTRY OF THE ATMOSPHERE / Gas Phase Reactions
ought to be 10 n m and l o p 9 nm for unlike and like and Sodeau JR (eds)Low Temperature Chemistry of the
charges. These values mean, of course, that oppositely Atmosphere, pp. 235-262. Berlin: Springer-Verlag.
charged ions will react twenty times faster than their Kolb CE, Worsnop DR and Zahniser MS, et al. (1995)
neutral analogues, under similar conditions, while Laboratory studies of atmospheric heterogeneous chem-
similarly charged ions can be assumed not to react at istry. In: Barker JR (ed.) Progress and Problems in
all. Atmospheric Chemistry, pp. 771-875. Singapore: World
Scientific.
While this discussion has so far centered o n the
Laidler KJ (1987) Chemical Kinetics, 3rd edn. London:
behavior of the atmospherically dominant class of Harper & Row.
diffusion-controlled reactions, some processes of Molina MJ, Molina LT and Kolb CE (1996)Gas-phase and
interest are activation-controlled. One characteristic heterogeneous kinetics of the troposphere and strato-
of such reactions is that the activation energy may be sphere. Annual Review of Physical Chemistry 47:
smaller than for the equivalent gas-phase reaction, 327-367.
because the reactant pair undergoes many individual Pilling MJ (1996)Radical-radical reactions. Annual Review
‘collisions’ at each encounter, whereas, in the gas of Physical Chemistry 47: 81-108.
phase, the collision and the encounter are the same Pilling MJ and Seakins PW (1995) Reaction Kinetics.
thing. A particularly interesting property shown by Oxford: Oxford University Press.
activation-controlled ionic reactions is that of the Pilling MJ and Smith IWM (eds) (1987) Modern Gas
Kinetics. Oxford: Blackwell Scientific Publications.
kinetic salt effect. Rate coefficients are affected by the
Sims IR and Smith IWM (1995) Gas-phase reactions and
presence of other ionic species present in the solution energy transfer at very low temperatures. Annual Review
that do not themselves participate in the reaction. of Physical Chemistry 46: 109-137.
Interactions between oppositely charged partners are Smith IWM and Rowe BR (2000)Reaction kinetics at very
slowed down by the presence of such salts. In the low temperatures: laboratory studies and interstellar
atmosphere such effects may be of significance, since chemistry. Accounts of Chemical Research 33: 261-268.
water droplets may contain substantial amounts of Su Tand Bowers MT (1979)Classical ion-molecule collision
sea-salt or other similar species. theory. In: Bowers MT (ed.) Gas Phase Ion Chemistry,
vol. 1, pp. 83-118. New York: Academic Press.
Tolbert MA (1994) Laboratory studies of heterogeneous
See also reactions. In: Moortgat GK, Barnes AJ, Le Bras G, and
Sodeau JR (eds) Low Temperature Chemistry of the
Chemistry of the Atmosphere: Laboratory Studies; Atmosphere, pp. 263-286. Berlin: Springer-Verlag.
Principles of Chemical Change. Laboratory Kinetics. Troe J (1979) Predictive possibilities of unimolecular rate
theory. Journal of Physical Chemistry 83: 114-126.
Troe J (1983)Specific rate constants k ( E ,J ) for unimolecular
Further Reading bond fissions. Journal of Chemical Physics 79:
Cox BG (1994) Modern Liquid Phase Kinetics, Oxford: 6017-6029.
Oxford University Press. Warneck P (1994)Aqueous solution chemistry.In: Moortgat
Golden DM (1994) Gas phase homogeneous kinetics. In: GK, Barnes AJ, Le Bras G, and Sodeau JR (eds) Low
Moortgat GK, Barnes AJ, Le Bras G and Sodeau JR (eds) Temperature Chemistry of the Atmosphere, pp. 175-196.
Low Temperature Chemistry of the Atmosphere, Berlin: Springer-Verlag.
pp. 69-92. Berlin: Springer-Verlag. Wayne RP (2000) Photochemistry and kinetics applied
Golden DM and Williams LR (1994)Heterogeneous chem- to atmosphere. Chemistry of Atmospheres, 3rd edn,
istry and kinetics. In: Moortgat GK, Barnes AJ, Le Bras G, pp. 97-137. Oxford: Oxford University Press.
Gas Phase Reactions

G Le Bras, CNRS, Orleans, France reactions together with multiphase chemical processes
Copyright 2003 Elsevier Science Ltd. All Rights Resewed. and transport of the component species. Gas-phase
reactions play a dominant role in determining the
concentrations of key species such as reactive green-
house gases (methane, hydrochlorofluorocarbons
Introduction
(HCFCs), hydrofluorocarbons (HFCs), tropospheric
It has been known for only 50 years that the Earth’s ozone, etc.), stratospheric ozone, or pollutants (sulfur
atmosphere is not a chemically inert system but that its dioxide (SOZ), nitrogen oxides (NO, NOz), ozone,
chemical composition is determined by gas-phase etc.), in the lower troposphere. In addition to oxygen,
CHEMISTRY OF THE ATMOSPHERE I Gas Phase Reactions 353
a large variety of trace species are involved in gas- concentration of OH ([OH]x l o 6 r a d i c a l ~ c m - ~ ) .

phase reactions. Their volume mixing ratios range Typical lifetimes are 1.5 days, 2 weeks, 2 months, and
from -1.7parts per million (ppmv) for 10 years for N02, SOZ, CO, and CH4, respectively.
methane (CH4) to fractions of parts per trillion The minimum lifetime for a species that reacts at the
(pptv) for free radicals such as the hydroxyl fastest possible rate with the OH radical - i.e., at every
radical (OH). Most gas-phase reactions occurring in collision - is a few hours.
the atmosphere involve free radicals since these In the remote troposphere, where the effects of
reactions have low activation energies, and are thus human activity are negligible, the OH radical reacts
fast enough to participate at the relatively low mainly with methane and carbon monoxide, follow-
temperatures of the atmosphere (200-300 K). Some ing the primary reactions [IV] and [VI:
of the radicals are generated by photochemical proc-
esses activated by sunlight. Other radicals are gener- OH + CH4 + CH3 + H20 [IVI
ated as products of chemical reactions. This article
describes the key gas-phase reactions involved in the OH + CO + H + C02 [VI
major atmospheric chemical systems that define the
properties of the atmosphere in relation to climate, UV The CH3 and H free radicals and atoms produced
radiation at the surface, and air quality. combine with oxygen, as do many atmospheric
free radicals, to form peroxy radicals RO2 (ROO),
where R=CH3 or H in this case (reactions [VI]
and [VII]):
Hydroxyl Radical-Initiated Oxidation of
Methane and Carbon Monoxide, and CH3 + 0 2 (+M) 4 CH30 2 (+M) [VI]
Ozone Chemistry in the Remote
Troposphere H + 0 2 (+M) + HO2 (+M) [VI11
The major oxidant in the troposphere is the hydroxyl In tropospheric regions where the NO concentra-
radical (OH),which removes many trace gases. The tions are very low (less than lopptv), the peroxy
OH radical is produced by the photochemical pro- radicals recombine to produce peroxides as in reac-
cesses [I] and [11]: tions [VIII] and [IX]:
0 3 + sunlight(), = 300nm) + O(lD) + 0 2 [I] CH3O2 + H02 + CH3OOH +0 2 [VIII]
O(lD) $-H2O 4 2 0 H [I11 HO2 +€402 + H202 +0 2 [IXI

Ozone is photodissociated by solar radiation at These reactions represent a net loss for HO, (OH and
wavelengths around 300 nm that reaches the Earth's H02) radicals if the hydrogen peroxide (H202) and
surface (shorter-wavelength radiation is absorbed by methyl hydroperoxide ( CH300H)do not regenerate
the stratospheric ozone layer). Excited atomic oxygen, HO, by solar photodissociation.
O(lD), is produced and a small fraction reacts with
water vapor to yield OH radicals, the major part of
-
If NO concentrations are higher than 10 pptv, the
peroxy radicals react with NO as in reactions [XI
O(lD) being deactivated by collisions with N2 or 0 2 . and [XI]:
Oxygen atoms combine with 0 2 to produce ozone
according to [111]: CH302 + NO + CH30 + NO2 [XI
0+0 2 + M(N2 or 0 2 ) -+ 0 3 +M [111]
HO2 + NO + OH + NO2 [XI1
0 2 is therefore the primary precursor of the O H
radical, and in this way it indirectly participates in The methoxy radical (CH30) further reacts with
atmospheric oxidation. oxygen as in [XII]:
The OH radical reacts with many species such as
sulfur dioxide (SOZ),nitrogen dioxide (N02),hydro-
CH3O +0 2 + CH20 + HO2 [XI11
carbons, and other volatile organic compounds Formaldehyde (CH20)is formed and its photodisso-
(VOCs). Reaction with OH determines the atmos- ciates or reacts with OH. CH20 is a source of
pheric lifetime or persistence of species that are HO,, mainly through the photodissociation channel
predominantly removed by this route. The lifetime is [XIII]:
expressed as 5 = l/k[OH], k being the rate coefficient
for the reaction and [OH] the mean tropospheric CH20 + sunlight(+202) + 2H02 + CO [XIII]
354 CHEMISTRY OF THE ATMOSPHERE I Gas Phase Reactions
The fact that CH20 is a source of HO, makes the OH-Initiated Oxidation of VOCs
oxidation of methane a net source of HO, at ‘high’NO
The degradation of VOCs is initiated mainly by
concentrations.
reaction with the O H radical. The reaction of satu-
The OH-initiated oxidations of CH4 and CO are the
rated VOCs, including hydrocarbons (RH),proceeds
major processes that control the ozone budget in the
by H-atom abstraction (reaction [XVII]), as already
remote troposphere. The photochemical source of
shown in reaction [IV] for methane:
tropospheric ozone is reaction [111], the oxygen atoms
needed for it being produced by photodissociation of OH + RH 4 R + H20 [XVII]
NO2 by sunlight according to [XIV]:
Unsaturated hydrocarbons react through addition of
NO2 + sunlight(i 5 400 nm) + N O + 0 [XIV] O H to the unsaturated bond, as illustrated in [XVIII]
for propene:
The sources of NO2 that lead to a net production of
ozone are reactions of NO with H02 and CH302 ([XI O H + CH3CH=CH2 -+ CH3CHCH20H [XVIII]
and [XI]).NO2 is also produced by reaction [XV], but
this reaction has no net effect on ozone since an ozone The organic radicals produced in the addition reac-
molecule is lost in reaction [XV] for every ozone tions, also represented here by R (reaction [XIX]), add
molecule created in reaction [111]: to oxygen to form peroxy radicals, RO2:
NO+03 +NO2+02 [XVI R + 0 2 (+M) -+ RO2 (+M) [XIXI

Reactions [XI and [XI] occur only at ‘high’ NO In a ‘NO,-rich’ (NO, = N O + NO2) polluted con-
concentrations, the HO2 and CH302 radicals other- tinental troposphere, the RO2 radicals react predom-
wise combining as indicated in reactions [VIII] and inantly with NO, as in [XX], [XXI], and [XXII]:
[IX]. HO2 itself can also react with ozone according
to [XVI]:
RO2 +NO --t NO2 + RO WI
H02+ 0 3 -+ O H + 202 [XVII + RON02 (organic nitrates) [XXII
In the remote troposphere, therefore, the oxidation of RO2 + N02(+M) +R02N02(+M) [xxIIl
CH4 and CO, which produces the CH3O2 and H02 (peroxy nitrates)
radicals, is a source of ozone through reactions [XI and
[XI] a t ‘high’ N O concentrations, and a sink through Reaction [XX], followed by photodissociation of NO2
reaction [XVI] at ‘low’ N O concentrations. The other (reaction [XIV]) and addition of 0 to 0 2 (reaction
important gas-phase sink of ozone is its photodisso- [111]) is a source of ozone. Reaction [XXI] to form
ciation in reaction [I]. organic nitrates is significant only for RO2 radicals
involving relatively ‘large’ (complex) groups R. Per-
oxy nitrates are generally thermochemically unstable,
Oxidation of Volatile Organic
and decompose back to the reactants. Peroxyacyl
Compounds (VOCs) and Photooxidant nitrates (RCO(02)02N02) behave somewhat differ-
Formation in the Polluted Troposphere ently, since they are fairly stable at low atmospheric
Human activities give rise to atmospheric emissions of temperatures.
VOCs and NO, that may generate photochemical The oxy radicals (RO) produced in reaction [XX]
smog. Species present in this form of photochemical can react with 0 2 , for simple RO, but also decompose
pollution include ozone, NO2, peroxides, carbonyl or isomerize for more complex radicals such as the
compounds such as aldehydes, and organic nitrates. pentoxy radical produced in the OH-initiated oxida-
The major classes of anthropogenic VOCs are tion of n-pentane ([XXIII] to [XXV]):
saturated hydrocarbons (alkanes),unsaturated hydro-
carbons (alkenes, and aromatic species such as
CH3C(O)HCH2CH2CH3 +0 2
benzene, toluene, and xylenes), mainly emitted + HO2 + CH3C(O)CH2CH2CH3 [XXIII]
by road traffic, and oxygenated organic compounds
(ketones, alcohols, ethers, esters, etc.) used as
fuel additives or solvents. Biogenic emissions of CH3C(O)HCH2CH2CH3 (+M)
unsaturated hydrocarbons (isoprene and terpenes) -+ CH3CHO + CH3CH2CH2(+M)
from vegetation can also contribute to photooxidant
formation. (decomposition) [XXIV]
CH3C(O)HCH2CH2CH3
-+ CH3 CH(OH) CH2 CH2 CHI
(isomerization) [XXVI
"rnW
NO
The products of these reactions are HO2, carbonyl

compounds ( 2-pentanoneY CH3C(O)CH2CH2CH3;
acetaldehyde, CH3CHO), and organic radicals R
( CH3CH2CH2, CH3CH(OH)CH2CH2CH2), which
further add 0 2 to form new peroxy radicals, RO2,
these latter species producing the corresponding R O
Ro2 7s
NO Np2
Ro decomposition or
isomerization
by reaction with NO. HO2 produced by reaction of L

RO with 0 2 leads to the regeneration of O H through
reaction with N O (reaction [XI]). This mechanism is W
therefore a catalytic cycle that oxidizes the VOC under Figure 1 Schematic representation of free radical-catalyzed
consideration into a first generation of carbonyl oxidation of a VOC into its first-generation carbonyl products in the
products, with ozone generated as a by-product presence of NO,, and the associated generation of ozone.
(Figure 1).
The carbonyl compounds formed as intermediates
in these oxidation steps themselves react with O H
or are photodissociated, and they thus provide, NO3- and 03-lnitiated Oxidation of VOCs
as already stated for formaldehyde, an addi- In addition to OH, two oxidants - the nitrate radical
tional source of HO, (OH, HO2) radicals. Acetalde- (NO3)and ozone - also contribute to the initiation of
hyde is also a source of peroxyacetyl nitrate the oxidation of certain VOCs, mainly unsaturated
(PAN, CH3C(0 ) 0 2 N 0 2 ) , an important component hydrocarbons. The NO3 radical is produced by
of photochemical smog (reactions [XXVI] to reaction [XXX]:
[XXVIII] ):
NO2 +0 3 + No3 +0 2 [XXXI
O H + CH3CHO -+ CH3CO + H20 [XXVI]
The NO3 radical is essentially a nighttime species
CH3CO + Oz(+M) -+ CH3C(0)02(+M) [XXVII] because of its fast removal by photolysis during
daytime. In addition to reacting with VOCs, NO3
also reacts with NO2 ,forming dinitrogen pentoxide in
CH3C(0)02 N02(+M) + a reversible, equilibrium reaction [XXXI]:
.+CH3C(O)02NOz (+M) [XXVIII] N o 3 NO2(+M) .+ NlOs(-kM) + [XXXI]
The decomposition of PAN (reaction - [XXVIII]) is The concentrations of NO3 at night range between
strongly temperature dependent, with lifetimes of a few pptv to a few hundred pptv. The reaction of
PAN from 1hour at 25°C to 1 year at - 60°C. PAN the NO3 radical with alkenes proceeds through
can thus act as a reservoir of NO, that can be addition of the radical to the carbon-carbon double
transported over long distances into cold regions bond in a similar way to the behavior of OH.
before releasing NO, through decomposition in The oxidation chain also includes intermediate peroxy
warmer regions. NO, itself cannot be transported (R02)and oxy (RO)radicals, leading to the formation
directly far from its emission location since the lifetime
of carbonyls and nitro-oxy carbonyls, such as
of the oxides (1.5 days) is limited by the sink reactionthe nitro-oxy ketone CH3C(O)CH20N02, produced
[XXIX]: by the NO3-initiated oxidation of propene. NO3
O H + NO2 (+M) -+ H N 0 3(+M) [XXIX] also reacts quite rapidly with some important un-
saturated biogenic VOCs such as isoprene and the
In highly polluted urban areas (NO,25ppbv by terpenes.
volume), this reaction competes with the reaction of Although ozone is less reactive toward alkenes than
O H with VOCs in the cycle of Figure 1, thereby is OH, it contributes significantly to their oxidation
limiting the formation of ozone through this cycle. The because of the much larger tropospheric ozone
removal of ozone by reaction with N O in reaction concentrations (a few tens to a few hundreds ppbv)
[XV] also limits the concentrations ozone can reach in compared to O H concentrations (0.01-1 x pptv).
urban areas. The initial step of ozonolysis is addition of O3
across the double bond to form a primary ozonide The active halogen species, X and XO (particularly C1,
[XXXII]: Br, and BrO), have been suggested as potential
h
v oxidants in the troposphere that operate in addition
O3 + R, R,C =CR3R4 -+ ’0 ‘ 0 (ozonide) to OH, NO3, and ozone. Active halogens have also
\ I
R, R,C-CR,R,
been suggested as having a potential influence
on tropospheric ozone. In particular, evidence has
[XXXII] been reported that the halogens deplete ozone in a few
days in the Arctic troposphere in spring. The catalytic
The unstable ozonide fragments to form an aldehyde cycle involving the self-reaction of BrO (reaction
or a ketone and a biradical called the Criegee [XXXVIII]) followed by reaction [XXXVI] converts
intermediate ([XXXIII] and [XXXIV]): two O3 molecules into three 0 2 molecules for every
cycle completed, and is likely to be the major process
ozonide + R1R2C=O + [XXXIII] that depletes ozone in polar spring conditions:
+ R3R4C=O + R1R2COO [XXXIV] BrO + BrO + Br + Br +0 2 [XXXVIII]
The Criegee biradicals undergo complex reactions Another catalytic cycle involving heterogeneous pro-
that are so far not well understood. The re- cesses on wet sea salt could also be important since it
actions lead to carboxylic acids (RC(O)OH),alde- both depletes ozone and liberates additional active
hydes (RC(O)H),and hydroperoxides (ROOH).The bromine from sea salt. This process has been called
Criegee radicals when first formed are energy-rich; ‘bromine explosion’, and consists of the steps
before they are stabilized by collision, they can [XXXVI] and [XXXIX] to [XLI]:
decompose to produce O H radicals. This source of
O H has recently been considered to be especially 2(Br + 0 3 -+ BrO + 0 2 ) [XXXVI]
important at night when there are no photolytic
sources of the radical. 2(Br0 + H 0 2 + HOBr +02) [XXXIX]
Chemistry of Halogen-Containing
HOBr + (Br- + H+ in sea salt) + . . .
Species in the Marine Troposphere -+ Br2 + H20 [XLI
Organic halogen compounds - methyl halides Br2 + sunlight + Br + Br WII
(CH3C1, CH3Br, and CH31), produced mainly in the
oceans - react with O H by H-atom abstraction, as
exemplified for CH3C1 in [XXXV]: Net: 2H02 + 2 0 3 + (Br- + H+ in sea salt)
O H + CH3C1-+ CH2C1+ H2O [XXXV] + sunlight HOBr + 402 + H20
+
Methyl halides are oxidized mostly in the troposphere,

although some CH3C1 and CH3Br, with tropospheric
Oxidation of Dimethyl Sulfide in the
lifetimes in excess of 1 year, reach the stratosphere.
The tropospheric reactions of the halomethyl radicals Marine Troposphere
CH2X (X = C1, Br, I) lead to halogen-atom formation. Dimethyl sulfide (DMS, CH3SCH3) plays an impor-
Halogen atoms are also produced from photodissoci- tant role in atmospheric sulfur chemistry because its
ation of photolabile species: alkyl iodides ( CH31, gas-phase oxidation leads to formation of aerosols
CH212)are sources of iodine atoms; ClN02 and C12 are that serve as cloud condensation nuclei, so that DMS
sources of chlorine atoms; and BrCl is a source of both can have a climatic impact. DMS, which is emitted by
chlorine and bromine atoms. ClN02, Cl2, and BrCl oceans, reacts predominantly with O H radicals, fol-
are formed at the surface of sea-salt particles, which lowing two channels ([XLII] and [XLIII]):
supply halogens. Chlorine atoms react predominantly
with hydrocarbons in the troposphere, whereas reac- O H + CH3SCH3 + CH3SCH2 + H20
tion with ozone is favored for Br (reaction [XXXVI])
(H-atom abstraction) [XLII]
and essentially exclusive for I (reaction [XXXVII]):
Br+O3 - + B r O + 0 2 [XXXVI] O H + CH3SCH3 + CH3S(OH)CH3
I+o3 +10+02 jXXXVIIj (addition) [XLIII]

The abstraction channel dominates at higher tropo- by gas-phase steps of the chain oxidation of DMS, or
spheric temperatures, and the addition channel at by aqueous-phase reactions of soluble species such as
lower temperatures. The sulfur-containing radical DMSO and MSIA. Such processes have not yet been
formed in the abstraction channel adds to 0 2 to identified or quantified.
produce the peroxy radical CH3SCH202.This radical
reacts with HO2 to form the (hydro)peroxide
CH3SCH200H if NO, concentrations are very low. Ozone Chemistry in the Stratosphere
In the presence of sufficient NO,, the peroxy radical is Depletion of the Ozone Layer
converted to the oxy radical CH3SCH20. This latter
radical decomposes as in [XLIV]: The stratospheric ozone layer is for a large part
controlled by chemistry, although atmospheric mo-
CH3SCH20(+M) + CH3S + CH20(+M) [XLIV] tions play an important role in redistributing the ozone
in the stratosphere. The Chapman mechanism was
The CH3S radical is further oxidized to CH3S0, and first proposed in 1930 to explain the maintenance of
CH3SO is oxidized to CH3S02 in the presence of the ozone in a photostationary state. The steps include [L],
atmospheric oxidants 0 2 , 03,and/or NO2. CH3S02 [III], [LI], and [LII]:
very likely decomposes to CH3 and S02. SO2 is
oxidized by reaction with O H in the gas phase, leading 0 2 + sunlight(A<240 nm) + 0 + 0 [L]
to the formation of gaseous sulfuric acid ([XLV] to
[XLVII]):
SO2 + OH(+M) -+ HOS02(+M) WLVI
O3 + sunlight(A< 1.14 pm) +0 +0 2 [LI]
HOS02 + 0 2 -+ HOz + SO3 [XLVI]
0 + 0 3 202
-+ [LIII
SO3 + H2O + H2SO4 [XLVII] ‘Odd’ oxygen ([O] + [03])
is producedonly by
reaction [L] and is consumed only by reaction [LIV],
SO2 can also be oxidized to H2SO4 in aqueous while reactions [111] and [LI] serve to interconvert 0
aerosols, and these two processes are responsible to and 03.Although this mechanism predicts the forma-
acid-rain formation in SO2-polluted continental areas. tion of a layer of ozone, concentrations of ozone
In the remote marine atmosphere, gaseous H2S04 calculated using it are much higher than those meas-
produced from DMS can aggregate with other gaseous ured. Calculated ozone concentrations can be brought
H2SO4 and H20 molecules to generate new aerosols into line with measured values by postulating the
serving as condensation nuclei for clouds. occurrence of catalytic cycles of the form [LIII]
In the addition channel for the reaction of O H with and [LIV]:
DMS, the adduct formed reacts with 0 2 according to X+03-+XO+02 [LIII]
[XLVIII]:
CHjS(OH)CH3 +0 2 xo + 0 x + 0 2
-+
[LIVI
+ CH3S(O)CH3 + HO2 [XLVIII]

Net: 0 +0 3 + 202
In such cycles, the catalytic species X and X O are not
Dimethyl sulfoxide (DMSO, CH3S(O)CH3) is the consumed. The species X include H, OH, NO, C1, and
product of this process, and itself has recently been Br (withthe corresponding X O being OH, HO2, NO2,
found to react with O H to form methanesulfinic acid C10, and BrO).
(MSIA) as in [XLIX]:
O H + CH3S(O)CH3 + CH3S(O)OH + CH3 Hydroxyl species For X = OH, where O H is prima-
rily produced by reaction of O ( l D ) with H20, CH4,
[XLIX] and H2, the cycle [LV], [LVI] occurs:
MSIA is also likely to react with O H by H-atom OH+03 +H02+02 LVI
transfer to generate the CH3S02 radical, which in turn
will decompose to SO2 and CH3. The O H + DMSO
addition channel would therefore produce high yields
of SO2 and possibly of H2SO4, as does the abstraction
channel. However, these high yields might be limited Net: 0 + 0 3 -+ 202
This cycle is important in the upper stratosphere where Several reactions convert C10 back to C1, thus leading
the concentrations of oxygen atoms are relatively high to catalytic cycles that deplete ozone (Figure2), such as
as a result of fast photochemical production from [LXI] to [LXIV]:
reaction [L]. In the middle and lower stratosphere,
HO2 reacts with O3instead of oxygen atoms, resulting ClO + 0 c1+0 2
-+
[LXII
in the cycle [LVII], [XVI]:
O H + 0 3 -+H02+02 [LVII]
C10 + N O * C1+ NO2 [LXII]
H02 + O3 + O H + 2 0 2 C10 + HO2 + HOCl +0 2 [LXIII]

[XVII
HOCl + sunlight + OH + C1 [LXIV]
Net: O3 + 3 0 2
The reservoirs of C10, are HCl and C10N02. HC1,
The efficiency of the cycles is reduced by reactions that for instance, is produced by reaction [LXV]:
store the catalytic species in reservoir molecules. For
example, the HO, reservoir, H20, is produced by C1+ CH4 --f HC1+ CH3 WVI
reaction [LVIII]:
It regenerates C1 by reaction [LXVI]:
O H + HO2 + H2O +0 2 [LVIII]
OH + HC1+ C1+ H20 [LXVI]
HC1 is the major inorganic chlorine species because of
NO, species For the NO, species, which mainly
its rather low reactivity (lifetime of 10 days), the active
result from nitrous oxide (N20) transported to the
chlorine (Cl, C10) representing only a few percent of
stratosphere, the cycle [XV], [LIX] involving oxygen
the total inorganic chlorine.
atoms depletes ozone in the upper stratosphere:
NO + 0 3 + NO2 +0 2 [XVI Bromine species For the BrO, (Br and BrO) species,
bromine atoms are primarily formed by photodisso-
ciation of halons (CF3Br, CF2BrC1, etc.) and CH3Br,
and partly from the OH-initiated oxidation of CH3Br.
Net: 0 + 0 3 + 202
H N 0 3 , ClON02, and N205 are reservoirs for NO2.
H N 0 3 , for instance, is produced by reaction of NO2
with O H (reaction [XXIX]), and slowly regenerates
NO2 and O H by photolysis. hv
In the lower stratosphere, NO reacts mostly with
HO2 and CH3O2, produced by oxidation of methane
that survives tropospheric oxidation and is transport-
ed to the stratosphere. The NO2 produced is photo-
dissociated, rather than reacting with oxygen atoms,
whose concentration in this region is low. Conse-
quently, in the lower stratosphere NO, is a source of
ozone from the sequence of reactions [XI, [XI], [XIV],
and [111] in the same way as in generation of 0 3 in the
troposphere.
Chlorine species For the C10, (C1and C10) species, M

the primary route into the chains involves chlorine
atoms, which are produced from halocarbons, mainly
by photodissociation of chlorofluorocarbons (CFCs;
CF2C12, CFC13, etc.) and by OH-initiated oxidation of
CH3C1. C1 atoms are thenconverted to C10 as in [LX]:
figure 2 Chernicai cycles of ozone depletion catalyzed by CIO,
c1+0 3 -+ c10 + 0 2 ILXI species.
Next Page
CHEMISTRYOF THE ATMOSPHERE I Ion Chemistry 359
The chemistry of BrO, is similar to that of C10,. The and CION02 to photolabile reservoirs Cl2 and HOC1
major cycles (below) that deplete ozone occur in the that rapidly generate chlorine atoms even when
lower troposphere: sunlight intensities are very low. The major catalytic
cycle (below) involves the C10 dimer (C10)2:
BrO + C10 --f Br + C1+ 0 2 [LXVII]
C10 + ClO(+M) + (ClO),(+M) [LXXIII]
Br+O3 -+BrO+02
C1+ 0 3 --f C10 +0 2

[LXVIII]
[LXIX]
(ClO), + sunlight(/, 5 350 nm) - C1+ ClOO
[LXXIV]
Net: 203 -+ 302 ClOO( + M) 4 C1+ O2(+M) [LXXV]
BrO + H02 4 HOBr +0 2 [LXXI 2(C1+ O3 -+ C10 +0 2 ) ILXI
HOBr + sunlight OH + Br [LXXI] Net: 2 0 3 4 3 0 2

The cycle involving C10, and BrO, radicals in
B r + 0 3 -+BrO+02 [LXVIII] reactions [LXVII] to [LXIX] also contributes to ozone
hole formation.
OH+03 +HO2+02 [LVII]
Net: 2 0 3 - 302
See also
Boundary Layers: Surface Layer. Chemistry of the
An important difference from C10, chemistry is that Atmosphere: Chemical Kinetics; ion Chemistry; Labora-
bromine atoms do not react with CH4 to produce the tory Studies; Principles of Chemical Change. Land-At-
HBr reservoir. The reactions of Br with H2CO and mosphere Interactions: Trace Gas Exchange. Ozone:
HOz, which do produce HBr, are much slower than Photochemistry of Ozone. Satellite Remote Sensing:
the reaction of C1 with CH4 [LXV] forming HC1. In Surface Wind. Stratospheric Chemistry and Compo-
sition: Hydroxyl Radical. Ultraviolet, Surface.
addition, Br is more rapidly regenerated from HBr
[LXXII] than is C1 from HCl through reaction [LXVI]:
OH + HBr - Br + H20 [LXXII] Further Reading
Atkinson R (1994) Gas-phase tropospheric chemistry of
Moreover, the other potential bromine reservoir, organic compounds. Journal of Physical and Chemical
BrON02, is less stable than the analogous chlorine Reference Datu, Monograph 2: 11-216.
reservoir, ClON02. The partitioning of inorganic Finlayson-Pitts BJ and Pitts JN Jr (1999) Chemistry of the
bromine is therefore very different from the partition- Upper and Lower Atmosphere. San Diego: Academic
ing of chlorine, with a higher fraction of active Press.
bromine (20-60%). Le Bras G (ed.) (1997) Chemical Processes in Atmospheric
Oxidation. Berlin: Springer-Verlag.
Ozone Hole Formation Seinfeld JH and Pandis SN (1998)Atmospheric Chemistry
and Physics. New York: Wiley.
Gas-phase reactions operating in catalytic cycles are Wayne RP (2000) Chemistry of Atmospheres, 3rd edn.
also responsible for the formation of the ozone hole in Oxford: Oxford University Press.
the polar stratosphere in spring, although heterogene- Zellner R (ed.) (1999) Global Aspects of Atmospheric
ous processes also need to be invoked. The heteroge- Chemistry. DarmstaddSteinkopf, New York: Springer-
neous processes convert the chlorine reservoirs HC1 Verlag.
Ion Chemistry
J L Fox, State University of New York at Stony Brook, Introduction
Stony Brook, NY, USA Most of the ionosphere is to be found in the outer
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. reaches of the atmosphere, the thermosphere, where
Contribution of working group I to the Third Assessment Takahashi T, Wanninkhof RH, Feely RA, et al. (1999)Net
Report of the Intergovernmental Panel on Climate air-sea C02 flux over global oceans: An improved
Change. Cambridge: Cambridge University Press. estimate based on sea-air pC02 difference. In: Proceed-
Sarmiento JL, Hughes TMC, Stouffer RJ and Manabe S ings of 2nd International Symposium on COz in the
(1998) Simulated response of the ocean carbon cycle to Oceans, CGER-1037-99, CGER/NIES, Tsukuba, Japan,
anthropogenic climate warming. Nature 393: 245-249. pp. 9-15.
See PREDICTABILITY AND CHAOS
Contents
Chemical Kinetics
Gas Phase Reactions
Ion Chemistry
Laboratory Studies
Principles of Chemical Change
Rate Laws
Chemical Kinetics Consider the hypothetical chemical reaction [I].
R P Wayne, University of Oxford, Oxford, UK A + B + . . . + products [I1

Experimentally, the rate is found t o be proportional to
the concentrations of A, B, ... each raised to some
power (eqn [l]).
Reaction Kinetics and Atmospheric
Chemistry
Rate = -__ 0 1 = k[A]X [B]B
441 = --
111
dt dt
Chemistry in the atmosphere often consists of several
consecutive and parallel steps that compete with each
other. Interpretation of the rates of chemical change,
Order, Molecularity and Rate Constant
and the concentrations and lifetimes of atmospheric The powers a and p in eqn [ 11are the order of reaction
constituents, requires a knowledge of the rates of the +
with respect to reactants A and B, and a p is the
elementary reaction steps that make up the complex overall order; the constant of proportionality, k, is the
scheme. The kinetic data embodied in this knowledge rate coefficient (rate ‘constant’). The molecularity of a
is best obtained by laboratory experiment, although reaction is the number of reactant molecules written in
theory may have to be used if the experiments cannot the stoichiometric equation. Order is thus an exper-
be performed. Theory is useful in another way as well, imental quantity; molecularity is an arbitrary theoret-
because it can provide a rational basis for the extra- ical one. An elementary reaction step is conceived as
polation of laboratory data to temperatures, pres- one that cannot be split into any chemically simpler
sures, and concentrations that exist in the atmosphere processes. For truly elementary steps, order and
but that cannot be used in the laboratory studies. molecularity are in general identical. Thus, if reaction
[I] is elementary, and the only reactants are A and B, it In eqn [3], o, is the cross-sectional area for collision
is both bimolecular and overall second-order: first- (collision cross-section), given by eqn [4], and Z is the
order in each of the components A and B. However, a mean relative velocity of molecules for temperature T ,
special case often arises in atmospheric chemistry. If given by eqn [5].
the second reactant, B, is in great excess over A, then its
concentration is effectively constant throughout the oc = n(TA + rBC)2 141
reaction. We can then combine the concentration with
the rate coefficient, k, and write the rate of reaction as 8 k ~ T' I 2 ~ A ~ B C
; p= 151
k'[A], where k' = k[B]. Such a process is termed a mA + mBC
pseudo-first-order reaction, and k' is the pseudo-first-
order rate coefficient. The quantities nA and Y ~ B Cin eqn [3] are the number
densities of A and BC (concentrations in molecular
units such as molecule ~ m - ~Equation
). [3] certainly
The Arrhenius Equation has the correct concentration dependence for an
Many rate constants are found to follow a temperature elementary bimolecular reaction, so that the rate
law embodied in the Arrhenius expression of eqn [2], coefficient can be written
where E , is the activation energy and A is the pre-
exponential factor.
k = o , exp
~ (- j$)
k=Aexp -_
( 2) It is clear that the Arrhenius equation (eqn [2]) and
eqn [6] bear a close resemblance, and E, is commonly
identified with E,, so that the question may be asked
The Arrhenius equation is entirely empirical, but whether a,Z is to be compared with A . However,
several theories of kinetics yield expressions for the it should not be forgotten that 2 is dependent on T1I2
rate coefficient that are similar in form. (cf. eqn [.SI), while A, in the simplest formulation,
is not temperature dependent. A more telling difficulty
The Theories of Elementary Gas-Phase concerns the absolute magnitudes of A and o,c.
For typical atmospheric reactants, with collision
Bimolecular Reaction Steps radii N 400 pm and relative molecular masses
Bimolecular processes are probably the most impor- N 30, ocE is N 3 ~ 1 0 - ~ ' c mmolecule-Is-'
~ at
tant class of reaction and, as we shall see later, 300 K. The product cr,E is called the collision frequen-
termolecular and many very important unimolecular cy factor. Except for the very simplest of reactants,
reactions involve several bimolecular elementary steps. experimental A factors are usually less than, and often
The obvious starting point in discussing the theories of much less than, the collision frequency factor. An
reaction is thus with bimolecular reactions. Two explanation for the lack of agreement is sought in
simplifications are commonly adopted in discussions terms of molecular complexity, with the existence of
of these theories. The first is the collision theory (CT), special geometric arrangements that are needed during
and the second is the transition state theory (TST). the collision to bring reactive parts of the molecules
together (steric requirements), and of special needs for
Collision Theory the distribution of internal energy. That explanation
In simple collision theory, reactant molecules are takes us well away from the idea of hard-sphere
assumed to be hard spheres (radii TA and rBC, say, for reactants.
reactants A and BC), and reaction is taken to be
Transition State Theory
possible only if two conditions are met: (1)a collision
must occur, and (2) the energy of collision along the The alternative simplification adopted in the interpre-
line of centers must equal or exceed the energy tation of bimolecular reactions is that of the transition
required, E,, to reach a critical configuration (ABC#, state theory (TST)or activated complex theory (ACT).
the transition state in Figure 1 ( D ) in the article on The reactants and the critically configured ABC
principles of chemical change; see Chemistry of the molecule are assumed to be in 'quasi-equilibrium'.
Atmosphere: Principles of Chemical Change). The rate Equilibrium constants can be expressed in statistical
of reaction according to this theory is readily shown to thermodynamic terms, and if the formulation is also
be given by eqn [3]. valid for the quasi-equilibrium, where the system is at
a (free) energy maximum rather than minimum, then
-- dnA -
dt
- -- dnBc = nAnBC o,Z exp
dt
(- j$) [3] concentrations of ABC# (the transition state) may be
calculated. Rates of reaction can then be obtained
CHEMISTRY OF THE ATMOSPHERE I Chemical Kinetics 345
from the rate at which ABC' passes to products (as a negative. The most sensible procedure in temperature
result of translational or vibrational motions along the extrapolation thus seems to be first to predict n from a
reaction coordinate). The resultant rate coefficient, k, model of the reaction, and then to fit the experimental
is given by eqn [7]. data to eqn [8] with that value of n.
Activation Energies and Long-Range Forces
Rates of reaction are, in part, controlled by the energy
Partition functions are written as q i , qLc, qlBC+for of a critical (transition-state)configuration, an energy
reactants and transition state, the primes showing that that has as its counterpart the activation energy of
the quantities are volume-independent. The double experimental kinetics. The energy barrier arises be-
prime indicates that the motion along the reaction cause the reactant molecules are forced close together
coordinate has been factorized out (and a numerical (closer than the sum of their radii in the hard-sphere
constant has been introduced). In TST, then, the collision approximation), and reactant bonds have to
internal motions neglected in collision theory (CT)are be broken while product bonds are made. The energy
required is less than that required first to break
expressly taken into account through the use of the
partition functions. TST concentrates only on that reactant bonds and then to form product molecules
region of the potential energy surface around the in separate steps. The energy does not decrease at any
transition state (for the calculation of the partition stage in this picture as the system passes from
function qiBc+)while CT is interested only in the separated reactants to the transition state. Such a
height of the energy barrier at the transition state. It is decrease in energy would correspond to long-range
the calculation of qiBc+that offers most difficulty in attractive forces, and might lead to an increased
the practical implementation of TST. Spectroscopic collision frequency, and to an A factor that exceeded
parameters for the reactant molecules are usually
ocE.Many examples of this type of behavior are in fact
available, so that q;, qLc are readily estimated. known, even with neutral reactants, but the effects are
However, a knowledge of the shape of, and the forces strongest and most common with charged reactants.
acting at, the transition state would imply that the In ion-molecule reactions such as reaction [11], the ion
can induce a dipole in the neutral reactant, and the
potential energy surface is itself known, at least in the
region of ABC'. The usual practice is to make an resultant attractive force can both balance the ordi-
'informed guess' at the magnitude of qiBc+based on a nary chemical activation barrier as well as make the
hypothetical interaction mechanism and a corre- real encounter rate greater than the gas-kinetic colli-
sion frequency factor for neutral molecules.
sponding model for the transition state. Considerable
differences in predicted pre-exponential factors are
obtained from models of the transition state that are,
o++ co2 -+ 0; + co PI1
for example, linear, bent, or cyclic. In a more limited Near-zero activation energies are thus often found
way, TST can suggest a sensible order of magnitude for
the pre-exponential factor. The three total partition
functions in eqn [7]are each the product of transla-
(typically -
in this type of reaction, and the pre-exponential factors
io9cm3 molecule- s - l ) are several
times larger than the values for neutral reactants.
tional, rotational, and vibrational partition functions. Because the long-range attractive forces dominate the
The translational parts can all be calculated, and potential energy, high velocities of approach are
orders of magnitude for rotational and vibrational counterproductive in promoting reaction, and some
parts can be employed in accordance with the number negative temperature coefficient of rate constant may
of each of these modes that exist in A, BC, and ABC# . be observed. The stronger (or longer-range) the
The temperature dependence for every partition func- interaction, the larger the rate coefficient. For ion
tion can be evaluated as a power law, so that eqn [7] reactions with neutral molecules possessing perma-
can be rewritten in the form of eqn [8], where A' is the nent (rather than induced) dipoles, pre-exponential
temperature-independent part of the pre-exponential factors are increased by another two or three times.
function, and n is some exponent chosen from the Thus charge transfer from O+ to the dipolar mole-
nature of the reactants (monatomic, diatomic, etc.) cule H20 (reaction [III]), has a rate coefficient of
and a model of the transition state. 2.3 x 10-9cm3 molecule-ls-l at 298K, and the
(- g)
activation energy is essentially zero.
k = A'T" exp
For the hard sphere (CT) case, n = 0.5, from eqns [5] The long-range interactions are yet larger, of course,
and [6]. In the more general case, n can be positive or for two reactants both of which are charged. Positive
ion-negative ion, or positive ion-electron reactions formation. A steady state for [XI has been reached,
are characterized by rate coefficients 3-4 orders of and d[X]/dt + 0. To illustrate the stationary-state
magnitude larger than typical gas-kinetic collision method, consider the pair of reactions [VI and [VI].
frequency factors. For example, the rate coefficient
(298K) for neutralization of NO+ by an electron A+B +X kf formation [VI
(reaction [IVI)is 4.5 x 1 0 - ' c m ~ m o l e c u l e - ~ s - ~ .
X+C+D k, loss [VI1
NO' + e -+ N +0 [IVI
For simplicity, let [B] and [C] be in great excess, so
that we may write pseudo-first-order rate coefficients
k; = kf[B] and k[ = k,[C]. Equation [9] is the kinetic
Multistep Reactions and the Stationary equation that describes reactions [VI and [VI].
State Hypothesis
Atmospheric chemistry consists of complex interac- = k;[A] - k:[X] 191
dt
tions of elementary reactions. Some of the processes to
be described in the following section on unimolecular If X is in a stationary state, then we set the differential
and termolecular reactions also involve several steps, equal to zero and obtain eqn [IO], where [XI,, indicates
which is why the subject of multistep reactions is a steady-state concentration of X.
introduced at this stage.
Consecutive and parallel steps involve reactive
intermediates in competitive processes. Reaction
intermediates of particular interest include atoms, The problem is to know whether the concentration of
radicals, ions, and excited species. Most of these X calculated using the SSH bears any relationship to
intermediates are highly reactive and, with one or two actual concentrations. Our two-reaction example has
exceptions, cannot be 'stored' in a laboratory for long been chosen because it can also be solved analytically.
periods because they are lost on the walls of the So long as [A] and [C] are independent of time, eqn [9]
containing vessel or react with each other. Such can be integrated to yield eqn [ 111,where t is the time
intermediates are not necessarily unstable, and chem- for which the system has been reacting.
ical lifetimes of isolated atoms or radicals in the
absence of surfaces can be virtually infinite. Many 441
[XI = -[I - exp(-k:t)]
excited-state species are unstable, since they may k:
possess enough internal energy to fragment, and they This expression for [XI approaches the steady-state
may also be able to lose their energy by emission of expression so long as k ; t >> 1,the error in applying the
radiation. An excited species that cannot undergo loss SSH being less than one per cent for kit > 4.6.The SSH
by an allowed radiative transition is said to be can thus be applied so long as [A] and [C] remain
metastable. constant over a period long enough for this inequality
Multistep reaction schemes are interpreted kinetic- to be reached. It is evident that the circumstances
ally by writing down the differential equations, such as under which the SSH is most likely to be valid are thus
eqn [l],for all the species of interest, including the those where k; is large: that is, if the species X is highly
intermediates. Solution of these equations then allows reactive. One example of a species at steady state in the
prediction of the concentration-time variation of each atmosphere is the highly reactive state of atomic
of the species. Unfortunately, analytical solution of the oxygen, O('D), throughout the troposphere, strato-
many simultaneous differential equations is rarely sphere, and probably the mesosphere. Ground-state
possible. Numerical solution has become a widely atomic oxygen, O(3P),however, cannot generally be
used alternative since the advent of high-speed com- treated in the atmosphere by steady-state methods
puters and the development of good techniques for because of its relatively small reactivity.
dealing with differential equations. For some highly
reactive intermediates, the stationary-state hypothesis
(SSH) (often alternatively called the steady-state Theories of Unimolecular and
hypothesis) provides a simplification that will permit
algebraic solution of the kinetic equations. Consider
Termolecular Reactions
an intermediate X that is created in a process whose We are now in a position to consider thermal uni-
rate is constant, and whose loss-rate increases with molecular reactions, and their close counterpart,
increased [XI. After the reaction is started, [XI will termolecular reactions. If chemical reaction requires
increase until the rate of loss is equal to the rate of collision between, or at least close proximity of, the
reactants, then it might seem that all thermal processes first-order kinetics follow. It is obvious, however, that
ought to be kinetically of second order. Unimolecular, at sufficiently low concentrations of AB, there comes a
first-order, elementary processes appear to lack the point at which reaction [IX] is rate-limiting, and the
necessary approach of reactants, and termolecular, kinetic behavior will be second-order. Transition from
third-order, steps suffer from the impossibility of a first- to second-order behavior is, indeed, seen at low
simultaneous collision between three hard-sphere enough pressures in this kind of thermal unimolecular
reactants. The explanation for first- and third-order reaction. Quantitative expression of these ideas can be
thermal kinetics shares common ground, and a simple obtained by a steady-state treatment for the concen-
introduction is provided here. tration of ABt as described in the previous section. The
N o obstacle exists to understanding how single-step result for the rate of loss of AB is eqn [12], where k~ is
unimolecular and first-order decomposition occurs in the experimentally defined pseudo-first-order rate
a molecule AB that already has more than enough coefficient.
energy in it to break one of its bonds. An obvious case
is the photodissociation of polyatomic molecules,
where optical dissociation or predissociation popu-
lates vibrational levels of AB sufficiently to cause
fragmentation (see Chemistry of the Atmosphere: So long as kd[AB] >> k,, the reaction is first-order, but
Principles of Chemical Change). The rate of fragmen- if [AB] is reduced to the point at which the reverse
tation may depend on the rate at which energy can inequality holds, then the reaction becomes second-
accumulate in the bond to be broken, but the reaction order. At high concentration, the limiting value of k1
will be kinetically of first order. Chemical activation (referred to as k,”) is equal to ( k a k r / k d )and is thus
offers another route to high vibrational excitation. For truly first-order, being independent of [AB]. The low-
example, the reaction of HO2 with N O (reaction [VII]) pressure limit, kp, is equal to ka[AB] and is itself first-
can produce a highly excited H 0 2 N O t molecule (the order in pressure, or second-order overall.
dagger representing vibrational excitation). Considerations of high- and low-pressure extrapo-
lations of rate data are most frequently met in
HO2 + N O + HOINOt [VI11 atmospheric chemistry in connection with termolecu-
lar reactions. As with unimolecular reactions, termo-
This excited H 0 2 N O t can then either split up to the lecular processes have orders that are variable with
reactants again, or form O H andNO2 (reaction [VIII]). pressure, being third-order at ‘low’ pressure and
second-order at ‘high’ pressure. Such reactions
H02NOt 4 HO + NO2 [VIII]
are extremely important in combination processes
Reaction of the excited H 0 2 N O t is a unimolecular, (sometimes erroneously called ‘recombination’ proc-
first-order, elementary reaction. esses), and we can see why by first looking at the
It is in interpreting thermal unimolecular reactions reaction of two atoms to form a diatomic molecule. A
that some difficulty arises, since the formation of an typical case is the combination of two O(3P)atoms.
excited ABt molecule involves collisions between the Curve ‘X’ in Figure 2 of the article on principles of
AB species, and might therefore be expected to show chemical change (see Chemistry of the Atmosphere:
second-order kinetics. A basic understanding was Principles of Chemical Change) illustrates this com-
provided by Lindemann, who suggested that thermal bination. Even if the combining atoms have no relative
first-order reactions were not true elementary steps, translational energy, the newly formed 0 2 molecule
but rather involved at least three elementary processes has the 0+0 combination energy stored in it:
that energy is the 0-0 bond energy, and the 0 2 is
AB + AB 4 AB^ + AB chemically activated 0; at its dissociation limit.
Unless some energy is removed within one vibrational
collisional activation 11x1 period, the molecule will fall apart again as
the internuclear distance increases on the first oscilla-
ABt + AB 2 AB .tAB deactivation [XI tion. Energy can be removed in collisions; the species
that dissipates energy is often represented by the
symbol M. In the atmosphere of Earth, M is usually
AB^ 4 A + B reaction [XI1 the ‘bath’ mixture of N2 and 0 2 . The overall reaction
If reaction [XI dominates as a loss process for ABt over is now written as eqn [XII], which is a termolecular
[XI], then the concentration of ABt is almost at its step.
thermal equilibrium value, while the rate-determining
step for reaction is the first-order process [XI]. Overall 0+0+ M -+ 0 2 +M
The redissociation that has been prevented is the atmospheric pressure range.
unimolecular dissociation of 0; equivalent to step
[XI], and the process deactivating 0; is the equivalent OH + NO2 + M H N 0 3 + M
4 [XVII
of step [XI. We shall see shortly that internal energy
seems to flow fairly freely between different vibra- 0+0 2 +M 0 3 +M 4 [XVII]
tional modes of a polyatomic molecule. If the newly Equations [ 121 and [ 131 represent the variations of
formed molecule is larger than diatomic, there are such experimentally determined rates of reaction with
modes into which the bond-combination energy can pressure. The pseudo-first-order or pseudo-second-
flow. The lifetime of the newly formed molecule can order rate coefficients k~ or k11 can be conveniently
thus correspond to many vibrational periods before expressed in terms of the high- and low-pressure
the energy flows back to the critical bond. With a large limiting values k r , kp, or k r , kp,. For example, k11 in
enough polyatomic molecule, the lifetime can be so eqn [13] can be expressed as in eqn [14].
great that collisional removal of excess energy (stab-
ilization) is no longer rate-determining, and combina-
tion then exhibits second-order kinetics. Because
reactions [XI and [XI] are common to both unimolec-
ular and termolecular reactions, the same general Remembering that kpl is itself first-order in pressure, it
considerations about flow of energy apply to both can be seen that eqn [14] represents in outline the
types of process. variation of k11with pressure that is found experimen-
The analogue of eqn [12] can be derived from the tally. Unfortunately, however, the equation does not
single excitation level kinetic scheme of reactions match experimental data in detail, so that it cannot be
[XIIII-[XV]. applied directly to the calculation of rates at interme-
diate pressures. The reasons for the failure are known.
A + B 5 ABt combination [XIII] The reactions and the rate coefficients k, or k,, k d or k,,
and k, should have been defined for each individual
ABt + M 4AB + M stabilization [XIV] quantized vibrational level of ABt, and the individual
rates summed to give the total rate. It is, perhaps, easy
to see that the more energy available (beyond the
AB+5A + B reaction [XVI critical amount needed to break a particular bond), the
more rapid will be the fragmentation (i.e. the larger
The result is eqn [13], where ,411 is the experimentally
will be k,). Related to this point is the implication that
defined pseudo-second-order rate coefficient (analo-
energy stored in any vibrational mode can be made
gous to k~ in eqn [12]).
available to the critical bond. Experimental evidence
largely favors the flow of energy between modes as
being fairly free, and the distribution as being near
statistical. An additional complication involves the
We see straightaway that, if k, >> k,[M], the reaction interconversions of vibrations and rotations in the
is third-order, with kFI = (k,k,/k,)[M]. If, how- fragmenting molecule. The theory has been extended,
ever, k, << k,[M], then the reaction is second-order, modified, and manipulated over the years by Rice,
with k E = k,. Increased complexity in the molecule Ramsperger, Kassel, and Marcus, and the familiar
AB reduces the value of k,, because the combina- initials RRKM are used to designate their formulation.
tion energy is distributed among more vibrational With sufficient sophistication of the input informa-
modes. The concentration, or pressure, of third- tion, very good agreement can be obtained between
body M at which third-order behavior turns theory and experiment. Correspondingly, one could
over to second-order kinetics is thus lower the more have confidence in the extrapolation of data obtained
complex the molecule produced. ‘Complex’ is only in an intermediate concentration regime to either high-
a relative term here: combination of two hydrogen pressure (first-order) or low-pressure (second-order)
atoms to form H2 is third-order up to 104atm, limits. However, application of RRKM theory to real
while combination of two CH3 radicals to form processes of atmospheric importance is in practice
CZH6 is second-order at all but the lowest pressures. rather difficult, and an alternative, much simpler,
However, it so happens that the reactants in several approach is now almost universally adopted. This
combination reactions of great atmospheric impor- approach has its origins in work by Troe on the
tance, such as reactions [XVI] and [XVII] are of theoretical prediction of unimolecular reaction rate
just such molecular size that complex intermediate- parameters. However, with k r , k:, known, Troe has
order kinetics are displayed at some point in the shown that a simplification of his theory allows the
CHEMISTRY OF THE ATMOSPHERE /Chemical Kinetics 349
right-hand side of eqn [14] to be multiplied by a explicitly the steps involved in a heterogenous reaction
broadening factor, F, that is a function of (kpI/kr). For of a gas-phase species either with the bulk constituent
many atmospherically important termolecular reac- of a liquid droplet or with another species that is
tions, F may be calculated from a simple mathematical already dissolved in it. The uptake of gas-phase
expression (eqn [15]),in which Fcentis the broadening molecules can be either reactive or nonreactive.
factor for the center of the fall-off curve; typically, Chemical change corresponds to loss of the gas-phase
FCent= 0.6. molecule; uptake that is nonreactive can arise from
physical dissolution or from reversible chemistry.
Individual steps that can be envisaged are (1)gas-
phase transport of the reactant to the surface of the
Third-order reactions often show decreasing rate droplet; (2) accommodation at the surface; ( 3 )diffu-
with increasing temperature: they have a negative sion into the liquid; (4)chemical reaction; ( 5 )diffusion
temperature coefficient. The reason is that the larger of unreacted molecules and products to the surface;
the thermal kinetic energy possessed by the reactants A and (6) desorption of species from the interface.
and B in process [XIII], the more internal vibrational Characterization of each of these individual steps is
energy will be stored in the ABt molecule produced. As obviously a formidable task, although one that may be
pointed out earlier, the chance of the critical bond simplified - as often happens in kinetics - by one of the
energy finding its way back to a breakable bond is thus steps being rate-determining.
increased, and k, is larger. Since k, and k, are only In many respects, the kinetics of liquid, surface, and
slightly affected by temperature, it follows from eqn heterogeneous reactions are governed by the same
[13] that the rate of reaction will decrease with principles that we have established for gas-phase
increasing temperature. Thermal energy in effect processes. There are, however, some key differences.
assists the newly formed molecule to split up again, Reactions occurring inside particles are really confined
thus slowing the rate of combination. In the third- to the liquid phase, since diffusion coefficients within
order limit, kp, is inversely proportional to k, (see solids are too small to allow significant reaction rates.
above). Theory suggests that the temperature varia- On the other hand, reactions on solid surfaces are
tions of k, should be better expressed in terms of a thought to be of very considerable atmospheric
power, T", rather than as a conventional activation significance. The PSCs involved in stratospheric
energy. Hence experimental measurements of kf:as a chemistry may consist, for example, in part, of solid
function of temperature should be fitted against a T-" water-ice and solid nitric acid trihydrate. A convenient
law to allow rational interpolation or extrapolation to starting point in the present discussion will thus be an
atmospheric temperatures. Typical measured values of examination of surface reactions themselves. An
n are 2.5 to 3.1 for reaction [XVI] and 1.7 for reaction added degree of complexity arises when the particle
[XVII]. Models of the transition state for bond is liquid, as is the case for droplets in the troposphere,
association reactions also suggest that, at the high- and possibly for stratospheric sulfate aerosol, which
pressure limit, k11 should possess a negative exponent may be in the form of supercooled liquid sulfuric acid.
of temperature. Formal kinetic equations for reactions on the
surface of atmospheric particles often start from the
Condensed-Phase, Surface, and Langmuir adsorption isotherm, which is the simplest
of equations that expresses the partitioning of gas
Heterogeneous Reactions between surface and gas phases. The isotherm makes
Reactions within the liquid droplets of clouds and fogs several assumptions, including one that states that all
are important in several aspects of tropospheric surface sites are equivalent and that there are no
chemistry, such as the oxidation of sulfur dioxide. interactions between molecules adsorbed on them.
Liquid or solid particles can play a critical role in the With these assumptions, it may be shown that the
chemistry of the stratosphere under certain conditions. surface coverage, O X , is given by eqn [16].
Particles such as those of sulfate aerosol or clouds
formed from water-ice and hydrates of nitric acid
(polar stratospheric clouds, PSCs) are implicated in
such processes. This chemistry may involve surface
reactions or reactions within the bulk material, but the Here b is a constant equal to the ratio of rate
interface between gas and condensed phases is in- coefficients for adsorption onto the surface and
volved in some way, and the reactions are thus known desorption from it. For the low partial pressures of
as heterogeneous reactions. The possible complexity adsorbates present as minor constituents in the
of the kinetics can be illustrated by considering atmosphere, Ox is likely to be very nearly a linear
function of p x , although the full equation might be Uptake coefficients may be determined by a variety of
needed for very strongly adsorbed reactants or for high experimental methods. Regardless of whether the
partial pressures in the atmosphere. An extended molecule is removed by reaction on or within the
treatment for the situation where two different particle or by dissolving in it, eqn [21] provides the link
species, X and Y, are adsorbed is straightforward, between the kinetics of the uptake process and the
with the surface coverages OX and By both entering uptake coefficient. It will be evident that the reactive
into the equation. uptake coefficient is equivalent to the reaction
The kinetics of reaction for a single reactant are then probability.
developed by including a loss (probably decomposi- A complication obviously arises if a molecule does
tion) of the adsorbed molecules. If the rate coefficient not react irreversibly, but can desorb again from a
for this first-order loss process is k ~ the , rate of surface, or come out of solution to re-enter the gas
chemical change given by eqn [ 171. phase. In such cases, y can apparently be time-
dependent, and the measurement of the variation of
Rate = k L O X N kLbxpx 1171 y with time provides one way of examining these
reversible processes. In the case of the atmosphere,
The second (approximate) equality applies for low Ox. the most important aspect concerns the partitioning
The overall kinetics are first-order in p x (and thus [XI ).
of molecules between gaseous and liquid phases.
Analogous equations can obviously be developed for Solubilities of gases at low solute concentrations
the case of X and Y both adsorbed on a surface, and
obey Henry’s law (eqn [22]).
interacting there. If both species are weakly adsorbed
(small surface coverage), the kinetics will be second-
order, with a rate proportional to pxpy. If one (or
both) of the reactants is adsorbed too strongly to use [X(s)]is the concentration of X in solution, p x is its
the low-pressure limiting equation, then the full form pressure in the gas phase, and H X is the Henry’s law
of the adsorption isotherm must be employed. coefficient (which is a function of temperature).
A key parameter in the discussion of surface Henry’s law expresses an equilibrium situation, in
processes is the uptake coefficient, y, which is the which the fluxes of molecules into and out of the liquid
ratio of molecules lost to a surface to the number of are equal. However, it is straightforward to calculate
gas-surface collisions that occur. If the rate of collision the forward and reverse fluxes, and thus the net flux
of a molecule X with an area A of the surface is o, into the liquid, under nonequilibrium conditions. The
then the rate of loss of X per unit volume, -d[X]/dt, is first term comes immediately from eqn [20], while
equal to y w / V , where V is the volume of the system. the second requires use of the diffusion equation
The kinetic theory of gases shows that w is given by for transport of the molecules from the bulk liquid to
eqn [18]. the interface. If the coefficient of diffusion for this
latter process is D, then it may be shown that eqn [23]
EA [XI holds, where yo and y t are the uptake coefficients at
w=-
4 time 0 and time t.
Equation [ 191 follows from this expression for o.
4x1 -
-- --YWXI
dt 4v The equation shows how HX can be calculated from
measurements of uptake coefficient as a function
Now the loss of X may also be described in terms of time, or, conversely, how the variation of up-
of phenomenological rate equations of the type of take coefficient with time may be estimated from a
equation [20], where IS} represents the number of knowledge of the solubility of the gas. Note that at
active surface sites per unit area and ks and k’,are the
‘infinite’ time, yt becomes zero: the system has reached
second-order and the corresponding pseudo-first- equilibrium.
order rate coefficients for the surface loss process. The material developed so far is applicable both to
physical processes - adsorption, absorption, or solu-
d[X1= kS{S}[X] E
-- ki[X] 1201 tion - and to chemical change. In the particular case of
dt
chemical change, we can envisage two possibilities.
From a comparison of eqns 1191 and [20] Either the reaction may involve an interaction of the
gas-phase reactant with the surface or the bulk
constituent of the particle, or it may involve reaction
1211 with some second species already adsorbed on, or
dissolved in, the particle. The concepts set out earlier For many of the liquid-phase reactions of interest in
remain applicable in the second, 'bimolecular', situa- atmospheric chemistry, the intrinsic reactivity of the
tion, but the value of {S} at the surface (eqn [20])or the partners is, indeed, very high, leading to diffusion-
concentration of the partner reactant Y, in solution, controlled kinetic behavior. A very elementary treat-
will be determined by factors similar to those already ment of the diffusion-controlled rate constant, k d ,
determining the adsorption or solubility of Y. leads to eqn [24].
Finally, it is necessary to examine the kinetics
of reaction within the liquid phase itself. The solvent k d = ~WAB DAB ~ 4 1
obviously has the potential to exert a considerable
where TAB is a hypothetical encounter distance at
influence on the course of chemistry in the liquid
which two partners A and B will react, and DABis the
phase. In air at l a t m pressure, and at ambient diffusion coefficient for the reactants.
temperature, the molecules themselves occupy only The encounter distance may be roughly the sum of
roughly 0.2% of the total volume; in liquids, the gas-kinetic radii of the partners for neutral
the molecules can make up half the volume.
reactants, while the appropriate diffusion coefficient
At pressures of 1atm and below, we have been able may be similar to a mean bulk diffusion coefficient of
to assume that the reactant molecules undergo essen-
the reactants in the solvent. Making these assumptions
tially unhindered motion, and that assumption lies
and with typical values of TAB = 0.5nm and DAB=
behind the various formulations of kinetics that we 1 . 3 ~ 1 m2 0 ~ s-l~ (for Na+ in H20), kD is calculated
have discussed in previous sections. In distinction, in as N 8 x m3 molecule-' s-' or, in the units that
liquids the reactive molecules must squeeze past the we have been using for rate coefficients so far,
solvent molecules (or each other, if one species is also
8x cm3 molecule-' s-'. In liquid-phase kinetics,
the bulk liquid) if they are to reach each other and it is more conventional to use molar units for concen-
undergo reaction. Reactants, activated complexes or
trations, so that the equivalent figure is ( 6 ~ 1 0 ~ ~
intermediates, and products can also all interact with 1 0 3 x 8 x 1 0 - ' s ) z 5 x 109dm3mol-ls-'. In whatever
the solvent. One manifestation of the interaction with units this rate coefficient is expressed, it is evidently
intermediates is that energy removal in association about 40 times lower than the maximum gas-kinetic
reactions, such as the combination processes [XVI] or rate coefficient. In general, a rate coefficient of
[XVII], is virtually instantaneous and the systems > IO9dm3 mol-' s-' for an aqueous-phase reaction
always display pure second-order kinetics in the liquid is taken to be indicative of a diffusion-controlled
phase, in contrast to the behavior described in the last mechanism.
section for gas-phase reactants. Interactions of the One of the largest known rate coefficients for a
reactants and the solvent (especially water) may make condensed-phase process is that for the very important
the formation of ions energetically more favorable reaction [XVIII] (1.4 x 10" drn3mol-'s-' at 298 K).
than in the gas phase. New reaction channels may thus
become accessible, and the kinetics of the processes H+ +OH- + H20 [XVIII]
can be influenced by the attractive or repulsive
electrostatic interactions between the reactants, The magnitude mainly reflects the large diffusion
among many other factors. coefficients in water of OH- and, especially, of H'; the
Simple treatments of liquid-phase kinetics often rapid diffusion is itself a consequence of the special
start from the concept of the encounter pair of reactant mechanisms by which these ions migrate in liquid
molecules that find themselves together within a H20.
solvent cage. Two extreme cases can be envisaged. In Although the diffusion coefficient is most important
the first, the two species are very highly reactive in making reaction [XVIII] so fast, there is another
toward each other, and undergo chemical transforma- factor operating that may be dominant in other
tion within a very few 'collisions' within the cage. The reactions. The positive and negative ions attract each
rate-determining process is then the diffusion of the other, so that the effective encounter distance can be
reactants through the solvent to form the encounter much greater than the gas-kinetic collision distance;
pair, and the process is a diffusion-controlled reaction. that is, TAB has to be replaced by reffin eqn [24]. For
At the other extreme, the activation energy for TAB = 0.5 nm, straightforward electrostatic calcula-
reaction may require the partners to pick up appreci- tions indicate that reffought to be about 0.2 nm for
able amounts of energy as they shake against each oppositely charged ions (and 0.7 nm for like charges)
other within the cage, so that the kinetics are control- in water with a relative permittivity of 78. However, it
led by the rate of reaction within the cage, rather than seems that this bulk permittivity is inappropriate to the
by the rate at which they reach it. Activation control- highly ordered solvent molecules in the immediate
led reaction kinetics then result. vicinity of the ions, and that more realistic values of r,ff
ought to be 10 n m and l o p 9 nm for unlike and like and Sodeau JR (eds)Low Temperature Chemistry of the
charges. These values mean, of course, that oppositely Atmosphere, pp. 235-262. Berlin: Springer-Verlag.
charged ions will react twenty times faster than their Kolb CE, Worsnop DR and Zahniser MS, et al. (1995)
neutral analogues, under similar conditions, while Laboratory studies of atmospheric heterogeneous chem-
similarly charged ions can be assumed not to react at istry. In: Barker JR (ed.) Progress and Problems in
all. Atmospheric Chemistry, pp. 771-875. Singapore: World
Scientific.
While this discussion has so far centered o n the
Laidler KJ (1987) Chemical Kinetics, 3rd edn. London:
behavior of the atmospherically dominant class of Harper & Row.
diffusion-controlled reactions, some processes of Molina MJ, Molina LT and Kolb CE (1996)Gas-phase and
interest are activation-controlled. One characteristic heterogeneous kinetics of the troposphere and strato-
of such reactions is that the activation energy may be sphere. Annual Review of Physical Chemistry 47:
smaller than for the equivalent gas-phase reaction, 327-367.
because the reactant pair undergoes many individual Pilling MJ (1996)Radical-radical reactions. Annual Review
‘collisions’ at each encounter, whereas, in the gas of Physical Chemistry 47: 81-108.
phase, the collision and the encounter are the same Pilling MJ and Seakins PW (1995) Reaction Kinetics.
thing. A particularly interesting property shown by Oxford: Oxford University Press.
activation-controlled ionic reactions is that of the Pilling MJ and Smith IWM (eds) (1987) Modern Gas
Kinetics. Oxford: Blackwell Scientific Publications.
kinetic salt effect. Rate coefficients are affected by the
Sims IR and Smith IWM (1995) Gas-phase reactions and
presence of other ionic species present in the solution energy transfer at very low temperatures. Annual Review
that do not themselves participate in the reaction. of Physical Chemistry 46: 109-137.
Interactions between oppositely charged partners are Smith IWM and Rowe BR (2000)Reaction kinetics at very
slowed down by the presence of such salts. In the low temperatures: laboratory studies and interstellar
atmosphere such effects may be of significance, since chemistry. Accounts of Chemical Research 33: 261-268.
water droplets may contain substantial amounts of Su Tand Bowers MT (1979)Classical ion-molecule collision
sea-salt or other similar species. theory. In: Bowers MT (ed.) Gas Phase Ion Chemistry,
vol. 1, pp. 83-118. New York: Academic Press.
Tolbert MA (1994) Laboratory studies of heterogeneous
See also reactions. In: Moortgat GK, Barnes AJ, Le Bras G, and
Sodeau JR (eds) Low Temperature Chemistry of the
Chemistry of the Atmosphere: Laboratory Studies; Atmosphere, pp. 263-286. Berlin: Springer-Verlag.
Principles of Chemical Change. Laboratory Kinetics. Troe J (1979) Predictive possibilities of unimolecular rate
theory. Journal of Physical Chemistry 83: 114-126.
Troe J (1983)Specific rate constants k ( E ,J ) for unimolecular
Further Reading bond fissions. Journal of Chemical Physics 79:
Cox BG (1994) Modern Liquid Phase Kinetics, Oxford: 6017-6029.
Oxford University Press. Warneck P (1994)Aqueous solution chemistry.In: Moortgat
Golden DM (1994) Gas phase homogeneous kinetics. In: GK, Barnes AJ, Le Bras G, and Sodeau JR (eds) Low
Moortgat GK, Barnes AJ, Le Bras G and Sodeau JR (eds) Temperature Chemistry of the Atmosphere, pp. 175-196.
Low Temperature Chemistry of the Atmosphere, Berlin: Springer-Verlag.
pp. 69-92. Berlin: Springer-Verlag. Wayne RP (2000) Photochemistry and kinetics applied
Golden DM and Williams LR (1994)Heterogeneous chem- to atmosphere. Chemistry of Atmospheres, 3rd edn,
istry and kinetics. In: Moortgat GK, Barnes AJ, Le Bras G, pp. 97-137. Oxford: Oxford University Press.
Gas Phase Reactions

G Le Bras, CNRS, Orleans, France reactions together with multiphase chemical processes
Copyright 2003 Elsevier Science Ltd. All Rights Resewed. and transport of the component species. Gas-phase
reactions play a dominant role in determining the
concentrations of key species such as reactive green-
house gases (methane, hydrochlorofluorocarbons
Introduction
(HCFCs), hydrofluorocarbons (HFCs), tropospheric
It has been known for only 50 years that the Earth’s ozone, etc.), stratospheric ozone, or pollutants (sulfur
atmosphere is not a chemically inert system but that its dioxide (SOZ), nitrogen oxides (NO, NOz), ozone,
chemical composition is determined by gas-phase etc.), in the lower troposphere. In addition to oxygen,
a large variety of trace species are involved in gas- concentration of OH ([OH]x l o 6 r a d i c a l ~ c m - ~ ) .

phase reactions. Their volume mixing ratios range Typical lifetimes are 1.5 days, 2 weeks, 2 months, and
from -1.7parts per million (ppmv) for 10 years for N02, SOZ, CO, and CH4, respectively.
methane (CH4) to fractions of parts per trillion The minimum lifetime for a species that reacts at the
(pptv) for free radicals such as the hydroxyl fastest possible rate with the OH radical - i.e., at every
radical (OH). Most gas-phase reactions occurring in collision - is a few hours.
the atmosphere involve free radicals since these In the remote troposphere, where the effects of
reactions have low activation energies, and are thus human activity are negligible, the OH radical reacts
fast enough to participate at the relatively low mainly with methane and carbon monoxide, follow-
temperatures of the atmosphere (200-300 K). Some ing the primary reactions [IV] and [VI:
of the radicals are generated by photochemical proc-
esses activated by sunlight. Other radicals are gener- OH + CH4 + CH3 + H20 [IVI
ated as products of chemical reactions. This article
describes the key gas-phase reactions involved in the OH + CO + H + C02 [VI
major atmospheric chemical systems that define the
properties of the atmosphere in relation to climate, UV The CH3 and H free radicals and atoms produced
radiation at the surface, and air quality. combine with oxygen, as do many atmospheric
free radicals, to form peroxy radicals RO2 (ROO),
where R=CH3 or H in this case (reactions [VI]
and [VII]):
Hydroxyl Radical-Initiated Oxidation of
Methane and Carbon Monoxide, and CH3 + 0 2 (+M) 4 CH30 2 (+M) [VI]
Ozone Chemistry in the Remote
Troposphere H + 0 2 (+M) + HO2 (+M) [VI11
The major oxidant in the troposphere is the hydroxyl In tropospheric regions where the NO concentra-
radical (OH),which removes many trace gases. The tions are very low (less than lopptv), the peroxy
OH radical is produced by the photochemical pro- radicals recombine to produce peroxides as in reac-
cesses [I] and [11]: tions [VIII] and [IX]:
0 3 + sunlight(), = 300nm) + O(lD) + 0 2 [I] CH3O2 + H02 + CH3OOH +0 2 [VIII]
O(lD) $-H2O 4 2 0 H [I11 HO2 +€402 + H202 +0 2 [IXI

Ozone is photodissociated by solar radiation at These reactions represent a net loss for HO, (OH and
wavelengths around 300 nm that reaches the Earth's H02) radicals if the hydrogen peroxide (H202) and
surface (shorter-wavelength radiation is absorbed by methyl hydroperoxide ( CH300H)do not regenerate
the stratospheric ozone layer). Excited atomic oxygen, HO, by solar photodissociation.
O(lD), is produced and a small fraction reacts with
water vapor to yield OH radicals, the major part of
-
If NO concentrations are higher than 10 pptv, the
peroxy radicals react with NO as in reactions [XI
O(lD) being deactivated by collisions with N2 or 0 2 . and [XI]:
Oxygen atoms combine with 0 2 to produce ozone
according to [111]: CH302 + NO + CH30 + NO2 [XI
0+0 2 + M(N2 or 0 2 ) -+ 0 3 +M [111]
HO2 + NO + OH + NO2 [XI1
0 2 is therefore the primary precursor of the O H
radical, and in this way it indirectly participates in The methoxy radical (CH30) further reacts with
atmospheric oxidation. oxygen as in [XII]:
The OH radical reacts with many species such as
sulfur dioxide (SOZ),nitrogen dioxide (N02),hydro-
CH3O +0 2 + CH20 + HO2 [XI11
carbons, and other volatile organic compounds Formaldehyde (CH20)is formed and its photodisso-
(VOCs). Reaction with OH determines the atmos- ciates or reacts with OH. CH20 is a source of
pheric lifetime or persistence of species that are HO,, mainly through the photodissociation channel
predominantly removed by this route. The lifetime is [XIII]:
expressed as 5 = l/k[OH], k being the rate coefficient
for the reaction and [OH] the mean tropospheric CH20 + sunlight(+202) + 2H02 + CO [XIII]
The fact that CH20 is a source of HO, makes the OH-Initiated Oxidation of VOCs
oxidation of methane a net source of HO, at ‘high’NO
The degradation of VOCs is initiated mainly by
concentrations.
reaction with the O H radical. The reaction of satu-
The OH-initiated oxidations of CH4 and CO are the
rated VOCs, including hydrocarbons (RH),proceeds
major processes that control the ozone budget in the
by H-atom abstraction (reaction [XVII]), as already
remote troposphere. The photochemical source of
shown in reaction [IV] for methane:
tropospheric ozone is reaction [111], the oxygen atoms
needed for it being produced by photodissociation of OH + RH 4 R + H20 [XVII]
NO2 by sunlight according to [XIV]:
Unsaturated hydrocarbons react through addition of
NO2 + sunlight(i 5 400 nm) + N O + 0 [XIV] O H to the unsaturated bond, as illustrated in [XVIII]
for propene:
The sources of NO2 that lead to a net production of
ozone are reactions of NO with H02 and CH302 ([XI O H + CH3CH=CH2 -+ CH3CHCH20H [XVIII]
and [XI]).NO2 is also produced by reaction [XV], but
this reaction has no net effect on ozone since an ozone The organic radicals produced in the addition reac-
molecule is lost in reaction [XV] for every ozone tions, also represented here by R (reaction [XIX]), add
molecule created in reaction [111]: to oxygen to form peroxy radicals, RO2:
NO+03 +NO2+02 [XVI R + 0 2 (+M) -+ RO2 (+M) [XIXI

Reactions [XI and [XI] occur only at ‘high’ NO In a ‘NO,-rich’ (NO, = N O + NO2) polluted con-
concentrations, the HO2 and CH302 radicals other- tinental troposphere, the RO2 radicals react predom-
wise combining as indicated in reactions [VIII] and inantly with NO, as in [XX], [XXI], and [XXII]:
[IX]. HO2 itself can also react with ozone according
to [XVI]:
RO2 +NO --t NO2 + RO WI
H02+ 0 3 -+ O H + 202 [XVII + RON02 (organic nitrates) [XXII
In the remote troposphere, therefore, the oxidation of RO2 + N02(+M) +R02N02(+M) [xxIIl
CH4 and CO, which produces the CH3O2 and H02 (peroxy nitrates)
radicals, is a source of ozone through reactions [XI and
[XI] a t ‘high’ N O concentrations, and a sink through Reaction [XX], followed by photodissociation of NO2
reaction [XVI] at ‘low’ N O concentrations. The other (reaction [XIV]) and addition of 0 to 0 2 (reaction
important gas-phase sink of ozone is its photodisso- [111]) is a source of ozone. Reaction [XXI] to form
ciation in reaction [I]. organic nitrates is significant only for RO2 radicals
involving relatively ‘large’ (complex) groups R. Per-
oxy nitrates are generally thermochemically unstable,
Oxidation of Volatile Organic
and decompose back to the reactants. Peroxyacyl
Compounds (VOCs) and Photooxidant nitrates (RCO(02)02N02) behave somewhat differ-
Formation in the Polluted Troposphere ently, since they are fairly stable at low atmospheric
Human activities give rise to atmospheric emissions of temperatures.
VOCs and NO, that may generate photochemical The oxy radicals (RO) produced in reaction [XX]
smog. Species present in this form of photochemical can react with 0 2 , for simple RO, but also decompose
pollution include ozone, NO2, peroxides, carbonyl or isomerize for more complex radicals such as the
compounds such as aldehydes, and organic nitrates. pentoxy radical produced in the OH-initiated oxida-
The major classes of anthropogenic VOCs are tion of n-pentane ([XXIII] to [XXV]):
saturated hydrocarbons (alkanes),unsaturated hydro-
carbons (alkenes, and aromatic species such as
CH3C(O)HCH2CH2CH3 +0 2
benzene, toluene, and xylenes), mainly emitted + HO2 + CH3C(O)CH2CH2CH3 [XXIII]
by road traffic, and oxygenated organic compounds
(ketones, alcohols, ethers, esters, etc.) used as
fuel additives or solvents. Biogenic emissions of CH3C(O)HCH2CH2CH3 (+M)
unsaturated hydrocarbons (isoprene and terpenes) -+ CH3CHO + CH3CH2CH2(+M)
from vegetation can also contribute to photooxidant
formation. (decomposition) [XXIV]
CH3C(O)HCH2CH2CH3
-+ CH3 CH(OH) CH2 CH2 CHI
(isomerization) [XXVI
"rnW
NO
The products of these reactions are HO2, carbonyl

compounds ( 2-pentanoneY CH3C(O)CH2CH2CH3;
acetaldehyde, CH3CHO), and organic radicals R
( CH3CH2CH2, CH3CH(OH)CH2CH2CH2), which
further add 0 2 to form new peroxy radicals, RO2,
these latter species producing the corresponding R O
Ro2 7s
NO Np2
Ro decomposition or
isomerization
by reaction with NO. HO2 produced by reaction of L

RO with 0 2 leads to the regeneration of O H through
reaction with N O (reaction [XI]). This mechanism is W
therefore a catalytic cycle that oxidizes the VOC under Figure 1 Schematic representation of free radical-catalyzed
consideration into a first generation of carbonyl oxidation of a VOC into its first-generation carbonyl products in the
products, with ozone generated as a by-product presence of NO,, and the associated generation of ozone.
(Figure 1).
The carbonyl compounds formed as intermediates
in these oxidation steps themselves react with O H
or are photodissociated, and they thus provide, NO3- and 03-lnitiated Oxidation of VOCs
as already stated for formaldehyde, an addi- In addition to OH, two oxidants - the nitrate radical
tional source of HO, (OH, HO2) radicals. Acetalde- (NO3)and ozone - also contribute to the initiation of
hyde is also a source of peroxyacetyl nitrate the oxidation of certain VOCs, mainly unsaturated
(PAN, CH3C(0 ) 0 2 N 0 2 ) , an important component hydrocarbons. The NO3 radical is produced by
of photochemical smog (reactions [XXVI] to reaction [XXX]:
[XXVIII] ):
NO2 +0 3 + No3 +0 2 [XXXI
O H + CH3CHO -+ CH3CO + H20 [XXVI]
The NO3 radical is essentially a nighttime species
CH3CO + Oz(+M) -+ CH3C(0)02(+M) [XXVII] because of its fast removal by photolysis during
daytime. In addition to reacting with VOCs, NO3
also reacts with NO2 ,forming dinitrogen pentoxide in
CH3C(0)02 N02(+M) + a reversible, equilibrium reaction [XXXI]:
.+CH3C(O)02NOz (+M) [XXVIII] N o 3 NO2(+M) .+ NlOs(-kM) + [XXXI]
The decomposition of PAN (reaction - [XXVIII]) is The concentrations of NO3 at night range between
strongly temperature dependent, with lifetimes of a few pptv to a few hundred pptv. The reaction of
PAN from 1hour at 25°C to 1 year at - 60°C. PAN the NO3 radical with alkenes proceeds through
can thus act as a reservoir of NO, that can be addition of the radical to the carbon-carbon double
transported over long distances into cold regions bond in a similar way to the behavior of OH.
before releasing NO, through decomposition in The oxidation chain also includes intermediate peroxy
warmer regions. NO, itself cannot be transported (R02)and oxy (RO)radicals, leading to the formation
directly far from its emission location since the lifetime
of carbonyls and nitro-oxy carbonyls, such as
of the oxides (1.5 days) is limited by the sink reactionthe nitro-oxy ketone CH3C(O)CH20N02, produced
[XXIX]: by the NO3-initiated oxidation of propene. NO3
O H + NO2 (+M) -+ H N 0 3(+M) [XXIX] also reacts quite rapidly with some important un-
saturated biogenic VOCs such as isoprene and the
In highly polluted urban areas (NO,25ppbv by terpenes.
volume), this reaction competes with the reaction of Although ozone is less reactive toward alkenes than
O H with VOCs in the cycle of Figure 1, thereby is OH, it contributes significantly to their oxidation
limiting the formation of ozone through this cycle. The because of the much larger tropospheric ozone
removal of ozone by reaction with N O in reaction concentrations (a few tens to a few hundreds ppbv)
[XV] also limits the concentrations ozone can reach in compared to O H concentrations (0.01-1 x pptv).
urban areas. The initial step of ozonolysis is addition of O3
across the double bond to form a primary ozonide The active halogen species, X and XO (particularly C1,
[XXXII]: Br, and BrO), have been suggested as potential
h
v oxidants in the troposphere that operate in addition
O3 + R, R,C =CR3R4 -+ ’0 ‘ 0 (ozonide) to OH, NO3, and ozone. Active halogens have also
\ I
R, R,C-CR,R,
been suggested as having a potential influence
on tropospheric ozone. In particular, evidence has
[XXXII] been reported that the halogens deplete ozone in a few
days in the Arctic troposphere in spring. The catalytic
The unstable ozonide fragments to form an aldehyde cycle involving the self-reaction of BrO (reaction
or a ketone and a biradical called the Criegee [XXXVIII]) followed by reaction [XXXVI] converts
intermediate ([XXXIII] and [XXXIV]): two O3 molecules into three 0 2 molecules for every
cycle completed, and is likely to be the major process
ozonide + R1R2C=O + [XXXIII] that depletes ozone in polar spring conditions:
+ R3R4C=O + R1R2COO [XXXIV] BrO + BrO + Br + Br +0 2 [XXXVIII]
The Criegee biradicals undergo complex reactions Another catalytic cycle involving heterogeneous pro-
that are so far not well understood. The re- cesses on wet sea salt could also be important since it
actions lead to carboxylic acids (RC(O)OH),alde- both depletes ozone and liberates additional active
hydes (RC(O)H),and hydroperoxides (ROOH).The bromine from sea salt. This process has been called
Criegee radicals when first formed are energy-rich; ‘bromine explosion’, and consists of the steps
before they are stabilized by collision, they can [XXXVI] and [XXXIX] to [XLI]:
decompose to produce O H radicals. This source of
O H has recently been considered to be especially 2(Br + 0 3 -+ BrO + 0 2 ) [XXXVI]
important at night when there are no photolytic
sources of the radical. 2(Br0 + H 0 2 + HOBr +02) [XXXIX]
Chemistry of Halogen-Containing
HOBr + (Br- + H+ in sea salt) + . . .
Species in the Marine Troposphere -+ Br2 + H20 [XLI
Organic halogen compounds - methyl halides Br2 + sunlight + Br + Br WII
(CH3C1, CH3Br, and CH31), produced mainly in the
oceans - react with O H by H-atom abstraction, as
exemplified for CH3C1 in [XXXV]: Net: 2H02 + 2 0 3 + (Br- + H+ in sea salt)
O H + CH3C1-+ CH2C1+ H2O [XXXV] + sunlight HOBr + 402 + H20
+
Methyl halides are oxidized mostly in the troposphere,

although some CH3C1 and CH3Br, with tropospheric
Oxidation of Dimethyl Sulfide in the
lifetimes in excess of 1 year, reach the stratosphere.
The tropospheric reactions of the halomethyl radicals Marine Troposphere
CH2X (X = C1, Br, I) lead to halogen-atom formation. Dimethyl sulfide (DMS, CH3SCH3) plays an impor-
Halogen atoms are also produced from photodissoci- tant role in atmospheric sulfur chemistry because its
ation of photolabile species: alkyl iodides ( CH31, gas-phase oxidation leads to formation of aerosols
CH212)are sources of iodine atoms; ClN02 and C12 are that serve as cloud condensation nuclei, so that DMS
sources of chlorine atoms; and BrCl is a source of both can have a climatic impact. DMS, which is emitted by
chlorine and bromine atoms. ClN02, Cl2, and BrCl oceans, reacts predominantly with O H radicals, fol-
are formed at the surface of sea-salt particles, which lowing two channels ([XLII] and [XLIII]):
supply halogens. Chlorine atoms react predominantly
with hydrocarbons in the troposphere, whereas reac- O H + CH3SCH3 + CH3SCH2 + H20
tion with ozone is favored for Br (reaction [XXXVI])
(H-atom abstraction) [XLII]
and essentially exclusive for I (reaction [XXXVII]):
Br+O3 - + B r O + 0 2 [XXXVI] O H + CH3SCH3 + CH3S(OH)CH3
I+o3 +10+02 jXXXVIIj (addition) [XLIII]

The abstraction channel dominates at higher tropo- by gas-phase steps of the chain oxidation of DMS, or
spheric temperatures, and the addition channel at by aqueous-phase reactions of soluble species such as
lower temperatures. The sulfur-containing radical DMSO and MSIA. Such processes have not yet been
formed in the abstraction channel adds to 0 2 to identified or quantified.
produce the peroxy radical CH3SCH202.This radical
reacts with HO2 to form the (hydro)peroxide
CH3SCH200H if NO, concentrations are very low. Ozone Chemistry in the Stratosphere
In the presence of sufficient NO,, the peroxy radical is Depletion of the Ozone Layer
converted to the oxy radical CH3SCH20. This latter
radical decomposes as in [XLIV]: The stratospheric ozone layer is for a large part
controlled by chemistry, although atmospheric mo-
CH3SCH20(+M) + CH3S + CH20(+M) [XLIV] tions play an important role in redistributing the ozone
in the stratosphere. The Chapman mechanism was
The CH3S radical is further oxidized to CH3S0, and first proposed in 1930 to explain the maintenance of
CH3SO is oxidized to CH3S02 in the presence of the ozone in a photostationary state. The steps include [L],
atmospheric oxidants 0 2 , 03,and/or NO2. CH3S02 [III], [LI], and [LII]:
very likely decomposes to CH3 and S02. SO2 is
oxidized by reaction with O H in the gas phase, leading 0 2 + sunlight(A<240 nm) + 0 + 0 [L]
to the formation of gaseous sulfuric acid ([XLV] to
[XLVII]):
SO2 + OH(+M) -+ HOS02(+M) WLVI
O3 + sunlight(A< 1.14 pm) +0 +0 2 [LI]
HOS02 + 0 2 -+ HOz + SO3 [XLVI]
0 + 0 3 202
-+ [LIII
SO3 + H2O + H2SO4 [XLVII] ‘Odd’ oxygen ([O] + [03])
is producedonly by
reaction [L] and is consumed only by reaction [LIV],
SO2 can also be oxidized to H2SO4 in aqueous while reactions [111] and [LI] serve to interconvert 0
aerosols, and these two processes are responsible to and 03.Although this mechanism predicts the forma-
acid-rain formation in SO2-polluted continental areas. tion of a layer of ozone, concentrations of ozone
In the remote marine atmosphere, gaseous H2S04 calculated using it are much higher than those meas-
produced from DMS can aggregate with other gaseous ured. Calculated ozone concentrations can be brought
H2SO4 and H20 molecules to generate new aerosols into line with measured values by postulating the
serving as condensation nuclei for clouds. occurrence of catalytic cycles of the form [LIII]
In the addition channel for the reaction of O H with and [LIV]:
DMS, the adduct formed reacts with 0 2 according to X+03-+XO+02 [LIII]
[XLVIII]:
CHjS(OH)CH3 +0 2 xo + 0 x + 0 2
-+
[LIVI
+ CH3S(O)CH3 + HO2 [XLVIII]

Net: 0 +0 3 + 202
In such cycles, the catalytic species X and X O are not
Dimethyl sulfoxide (DMSO, CH3S(O)CH3) is the consumed. The species X include H, OH, NO, C1, and
product of this process, and itself has recently been Br (withthe corresponding X O being OH, HO2, NO2,
found to react with O H to form methanesulfinic acid C10, and BrO).
(MSIA) as in [XLIX]:
O H + CH3S(O)CH3 + CH3S(O)OH + CH3 Hydroxyl species For X = OH, where O H is prima-
rily produced by reaction of O ( l D ) with H20, CH4,
[XLIX] and H2, the cycle [LV], [LVI] occurs:
MSIA is also likely to react with O H by H-atom OH+03 +H02+02 LVI
transfer to generate the CH3S02 radical, which in turn
will decompose to SO2 and CH3. The O H + DMSO
addition channel would therefore produce high yields
of SO2 and possibly of H2SO4, as does the abstraction
channel. However, these high yields might be limited Net: 0 + 0 3 -+ 202
This cycle is important in the upper stratosphere where Several reactions convert C10 back to C1, thus leading
the concentrations of oxygen atoms are relatively high to catalytic cycles that deplete ozone (Figure2), such as
as a result of fast photochemical production from [LXI] to [LXIV]:
reaction [L]. In the middle and lower stratosphere,
HO2 reacts with O3instead of oxygen atoms, resulting ClO + 0 c1+0 2
-+
[LXII
in the cycle [LVII], [XVI]:
O H + 0 3 -+H02+02 [LVII]
C10 + N O * C1+ NO2 [LXII]
H02 + O3 + O H + 2 0 2 C10 + HO2 + HOCl +0 2 [LXIII]

[XVII
HOCl + sunlight + OH + C1 [LXIV]
Net: O3 + 3 0 2
The reservoirs of C10, are HCl and C10N02. HC1,
The efficiency of the cycles is reduced by reactions that for instance, is produced by reaction [LXV]:
store the catalytic species in reservoir molecules. For
example, the HO, reservoir, H20, is produced by C1+ CH4 --f HC1+ CH3 WVI
reaction [LVIII]:
It regenerates C1 by reaction [LXVI]:
O H + HO2 + H2O +0 2 [LVIII]
OH + HC1+ C1+ H20 [LXVI]
HC1 is the major inorganic chlorine species because of
NO, species For the NO, species, which mainly
its rather low reactivity (lifetime of 10 days), the active
result from nitrous oxide (N20) transported to the
chlorine (Cl, C10) representing only a few percent of
stratosphere, the cycle [XV], [LIX] involving oxygen
the total inorganic chlorine.
atoms depletes ozone in the upper stratosphere:
NO + 0 3 + NO2 +0 2 [XVI Bromine species For the BrO, (Br and BrO) species,
bromine atoms are primarily formed by photodisso-
ciation of halons (CF3Br, CF2BrC1, etc.) and CH3Br,
and partly from the OH-initiated oxidation of CH3Br.
Net: 0 + 0 3 + 202
H N 0 3 , ClON02, and N205 are reservoirs for NO2.
H N 0 3 , for instance, is produced by reaction of NO2
with O H (reaction [XXIX]), and slowly regenerates
NO2 and O H by photolysis. hv
In the lower stratosphere, NO reacts mostly with
HO2 and CH3O2, produced by oxidation of methane
that survives tropospheric oxidation and is transport-
ed to the stratosphere. The NO2 produced is photo-
dissociated, rather than reacting with oxygen atoms,
whose concentration in this region is low. Conse-
quently, in the lower stratosphere NO, is a source of
ozone from the sequence of reactions [XI, [XI], [XIV],
and [111] in the same way as in generation of 0 3 in the
troposphere.
Chlorine species For the C10, (C1and C10) species, M

the primary route into the chains involves chlorine
atoms, which are produced from halocarbons, mainly
by photodissociation of chlorofluorocarbons (CFCs;
CF2C12, CFC13, etc.) and by OH-initiated oxidation of
CH3C1. C1 atoms are thenconverted to C10 as in [LX]:
figure 2 Chernicai cycles of ozone depletion catalyzed by CIO,
c1+0 3 -+ c10 + 0 2 ILXI species.
Next Page
CHEMISTRYOF THE ATMOSPHERE I Ion Chemistry 359
The chemistry of BrO, is similar to that of C10,. The and CION02 to photolabile reservoirs Cl2 and HOC1
major cycles (below) that deplete ozone occur in the that rapidly generate chlorine atoms even when
lower troposphere: sunlight intensities are very low. The major catalytic
cycle (below) involves the C10 dimer (C10)2:
BrO + C10 --f Br + C1+ 0 2 [LXVII]
C10 + ClO(+M) + (ClO),(+M) [LXXIII]
Br+O3 -+BrO+02
C1+ 0 3 --f C10 +0 2

[LXVIII]
[LXIX]
(ClO), + sunlight(/, 5 350 nm) - C1+ ClOO
[LXXIV]
Net: 203 -+ 302 ClOO( + M) 4 C1+ O2(+M) [LXXV]
BrO + H02 4 HOBr +0 2 [LXXI 2(C1+ O3 -+ C10 +0 2 ) ILXI
HOBr + sunlight OH + Br [LXXI] Net: 2 0 3 4 3 0 2

The cycle involving C10, and BrO, radicals in
B r + 0 3 -+BrO+02 [LXVIII] reactions [LXVII] to [LXIX] also contributes to ozone
hole formation.
OH+03 +HO2+02 [LVII]
Net: 2 0 3 - 302
See also
Boundary Layers: Surface Layer. Chemistry of the
An important difference from C10, chemistry is that Atmosphere: Chemical Kinetics; ion Chemistry; Labora-
bromine atoms do not react with CH4 to produce the tory Studies; Principles of Chemical Change. Land-At-
HBr reservoir. The reactions of Br with H2CO and mosphere Interactions: Trace Gas Exchange. Ozone:
HOz, which do produce HBr, are much slower than Photochemistry of Ozone. Satellite Remote Sensing:
the reaction of C1 with CH4 [LXV] forming HC1. In Surface Wind. Stratospheric Chemistry and Compo-
sition: Hydroxyl Radical. Ultraviolet, Surface.
addition, Br is more rapidly regenerated from HBr
[LXXII] than is C1 from HCl through reaction [LXVI]:
OH + HBr - Br + H20 [LXXII] Further Reading
Atkinson R (1994) Gas-phase tropospheric chemistry of
Moreover, the other potential bromine reservoir, organic compounds. Journal of Physical and Chemical
BrON02, is less stable than the analogous chlorine Reference Datu, Monograph 2: 11-216.
reservoir, ClON02. The partitioning of inorganic Finlayson-Pitts BJ and Pitts JN Jr (1999) Chemistry of the
bromine is therefore very different from the partition- Upper and Lower Atmosphere. San Diego: Academic
ing of chlorine, with a higher fraction of active Press.
bromine (20-60%). Le Bras G (ed.) (1997) Chemical Processes in Atmospheric
Oxidation. Berlin: Springer-Verlag.
Ozone Hole Formation Seinfeld JH and Pandis SN (1998)Atmospheric Chemistry
and Physics. New York: Wiley.
Gas-phase reactions operating in catalytic cycles are Wayne RP (2000) Chemistry of Atmospheres, 3rd edn.
also responsible for the formation of the ozone hole in Oxford: Oxford University Press.
the polar stratosphere in spring, although heterogene- Zellner R (ed.) (1999) Global Aspects of Atmospheric
ous processes also need to be invoked. The heteroge- Chemistry. DarmstaddSteinkopf, New York: Springer-
neous processes convert the chlorine reservoirs HC1 Verlag.
Ion Chemistry
J L Fox, State University of New York at Stony Brook, Introduction
Stony Brook, NY, USA Most of the ionosphere is to be found in the outer
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. reaches of the atmosphere, the thermosphere, where
CLEAR AIR TURBULENCE 393
Modern Gas Kinetics, pp. 99-134. Oxford: Blackwell Wayne RP (1988) Principles and Applications of Photo-
Scientific Publications. chemistry. Oxford: Oxford University Press.
Smith IWM (1997) Collisional energy transfer, intramolec- Wayne RP (2000) Photochemistry and kinetics
ular vibrational relaxation and unimolecular reactions. applied to atmosphere. Chemistry of Atmospheres,
Journal of the Chemical Society, Faraday Transactions 3rd edn, pp. 97-137. Oxford: Oxford University
93: 3741-3750. Press.
See CLOUDS: Classification
G P Ellrod, National Oceanographic and Atmospheric turbulence intensity is based on deviations in vertical
Administration/National Environmental Satellite, Data, acceleration from the normal acceleration of gravity
and Information Service, Camp Springs, MD, USA (1g). These categories and their threshold devia-
P F Lester, San Jose State University, San Jose, CA, tions are: light (f0.2g), moderate (f0.5 g), severe
USA ( f1.0 g), and extreme (f2.0g). The physical impact
of CATon crew and passengers varies from discomfort
L J Ehernberger, National Aeronautics and Space
Administration Dryden Flight Research Center,
for the lighter turbulence categories to loss of flight
Edwards, CA, USA control during the rare extreme turbulence event. In
the most intense episodes, injuries, and in some very
rare cases, fatalities have occurred. Unrestrained crew
and passengers are especially vulnerable. Flight
through turbulent conditions also produces stresses
Introduction on the airframe. Repeated turbulence encounters over
Since the first aircraft flight, pilots have been aware of the lifetime of the aircraft may lead to metal fatigue
in-flight turbulence. Because the known turbulence of and, in extremely rare cases, structural failure.
the time was tied to strong, low-level winds, rough
terrain, and convection, some of the early pilots
predicted that, with the exception of thunderstorms, Causative Mechanisms and
the ability to attain higher flight altitude with pressu-
rized cabins would be accompanied by a marked
Characteristics
decrease in turbulence. This was not to be. In the Aircraft turbulence (including CAT) differs from
1940s, as fighter aircraft attained tropopause alti- turbulence in fluids as usually conceived. Aircraft
tudes, they experienced a previously unknown phe- turbulence is defined in terms of the aircraft response
nomenon: clear air turbulence (CAT), so called while classical turbulence is defined in terms of the
because initial encounters occurred in areas devoid state of the fluid. Simply put, aircraft turbulence is
of clouds. As aircraft were designed to fly higher and ‘bumpiness in flight’. The difference is critical in that
faster during the last half of the twentieth century, bumpiness depends on aircraft design, weight, speed,
CAT became the focus of many organized research and pilot input, in addition to the state of the
efforts. Our knowledge of CAT has grown substan- atmosphere (gusts, windshears, or waves), which
tially as a result. may or may not be turbulent in the classical sense.
CAT is now defined as aircraft turbulence that The discussion that follows considers atmospheric
occurs at altitudes of 5.6 km (about 500 hPa) or higher, phenomena that contribute to CATon a wide range of
either in cloud-free conditions or within stratiform scales. This section examines briefly the primary
clouds. The critical influence of CAT is on flight safety. physical characteristics and causes of microscale
One of the oldest schemes for the characterization of CAT. It then considers macroscale and mesoscale
394 CLEAR AIR TURBULENCE
forcing mechanisms that create a favorable environ- Ri becomes 1.0 or more, and KHI are present, they will
ment for CAT development. decrease in amplitude. The use of Ri as a practical
index for CAT diagnosis and prediction is discussed in
Kelvin-Helmholtz Waves and CAT a later section.
Although sheared layers in which unstable KHI
Our understanding of the production of CAT and its
occur are, typically, less than 1km in depth, horizontal
characteristics is rooted in theoretical studies of fluid dimensions are much larger (Le., from -10km
mechanics, in laboratory and numerical experiments, to a few hundred kilometers, often elongated in
and in field studies. The evidence from these investi- the direction of the wind). Thus, an individual
gations is that Kelvin-Helmholtz instability (KHI) KHI develops over a considerably smaller horizontal
episodes are the cause of a large fraction of CAT. scale than the sheared stable layers in which it
Growth and intensification of atmospheric turbulence is embedded. This observation has four important
occur when the destabilizing effect of wind shear is ramifications:
sufficient to overcome the dampening influence of
static stability. KHI produces shearing-gravity waves
with typical horizontal wavelengths of a few tens of 1. An aircraft flying in a sheared layer may experience
meters to a few kilometers, precisely the range of eddy CATover a distance much greater than the scale of
sizes to which most aircraft will have the maximum an individual KHI.
response. 2. CAT encounters are often burst-like (highly inter-
KHI arises from micro- and mesoscale wind shear mittent) as an aircraft crosses a thin sheared layer
intensification, when smooth, wavelike oscillations or as an aircraft flying within a sheared layer
within a sheared, statically stable layer grow in intercepts individual KHI elements in different
amplitude to the point where the wave crests overturn, stages of development.
or ‘break’. Wave-breaking at wavelengths of hundreds 3 . Regular, comprehensive observations of CAT are
of meters is highly effective in producing CAT, with a difficult to acquire because of its small scale and
rapid cascade of energy from the KHI to smaller-scale intermittent nature.
turbulence and dissipation. With respect to CAT, some 4. Conditions conducive to the development of
of the more important characteristics of KHI are the CAT are rooted in larger-scale processes that
following: produce and perturb the extensive, sheared stable
layers.
1. The typical lifetime of an individual KHI is about
5 minutes. Internal Gravity Waves and CAT
2. The length of the dominant wave (most unstable
Internal gravity waves (IGW)with horizontal scales of
KHI mode) is proportional to the depth of the
a few kilometers to a few hundred kilometers are also
sheared layer (i.e., about six times the depth).
mechanisms for CAT production. IGW may become
However, as KHI-induced turbulence and mixing
significant in the production of CAT in several ways.
modify the local wind shear structure and stability
Wave amplitude variations with height are caused by
stratification, variations in the KHI wavelengths
the decrease in density with altitude and by the
can be expected.
variation of background stability and wind with
3 . The intensity of the turbulence produced by KHI is
height. Aircraft intersecting large-amplitude IGW
in direct proportion to the initial wind shear across
may be exposed to ‘sharp-edged’ gusts or periodic
the layer. If the turbulent mixing caused by KHI
vertical motions which may be interpreted as CAT.
sufficiently weakens the background wind shear,
Vertical displacements due to gravity wave motions
the turbulence will decay and the flow will again
will modulate background wind shear, leading to the
become laminar. However, if turbulent mixing
production of microscale KHI. Near or slightly below
strengthens the wind shears near the boundaries
the ‘critical layer’ (where phase speed and background
of the old turbulence layer then new KHI may
wind speed are the same), wave-breaking may produce
develop.
intense turbulence. Finally, the occurrence and inten-
Another relevant result from studies of KHI has been sity of CAT are also affected by the excitation of IGW
the development of a basic dynamic instability prin- from different sources as well as by resonant nonlinear
ciple for a simple model of shearing gravity waves. interactions between different IGW modes and be-
Based on linear theory, the Miles-Howard criterion tween IGW and KHI in turbulent layers, can affect the
states that unstable wave modes resulting from verti- occurrence and intensity of CAT (see Dynamic Mete-
cal shear is likely to occur when the local gradient orology: Waves. Lee Waves and Mountain Waves.
Richardson number, Ri,is less than 0.25. In contrast, if Middle Atmosphere: Gravity Waves).
Macroscale Forcing Macroscale flow patterns conducive to CAT are

shown schematically in Figure 2. They can be classified
Any atmospheric circulation system that produces
into four basic types: (A)col pattern, (B)sharp trough,
deformation and/or convergence in the flow field,
(C) ridge, and (D) baroclinic instability. Based on
differential horizontal temperature advection, or sec-
observational studies, flow conditions resulting in
ondary (vertical) circulations has the potential to
anticyclonic shear or curvature (such as pattern (C),
create and strengthen sloping stable layers and their
Figure 2) produce CAT most frequently, while cyclonic
associated atmospheric stability and wind shear.
conditions (pattern (B), Figure 2) produce the most
Because of thermal wind requirements, such layers
intense CAT. While patterns (B) and (D) are both
(also called baroclinic layers, or frontal zones) are
associated with troughs, the orientation of the troughs
vertically sheared. Frontal zones are particularly
results in significant differences in the locations of
common with extratropical cyclones and associated
CAT and prevalent cloud conditions. A north-east to
upper tropospheric jet streams, thus presenting opti-
south-west oriented trough axis (pattern (B))results in
mal conditions for development of KHI and CAT.
CAT to the rear of the upper trough in relatively cloud-
Additionally, developing extratropical cyclones and
free conditions. A north-west to south-east trough
jet streaks, as well as sharply curved anticyclonic
(pattern (D)),typically associated with cyclogenesis,
flows, are characterized by unbalanced (ageostrophic)
results in CAT in advance of the trough, often
flow. Under these conditions, accelerations generate a embedded in thick cirrus clouds. As suggested by
broad spectrum of not only IGW but also longer,
Figure 2, sharply curved jet stream segments associ-
gravity-inertia waves which may also be effective in
ated with upper-level troughs or ridges are typically
producing CAT by modulating shallow shear layers, as
more turbulent than most zonal (predominately west-
well as by their own instabilities. Because the jet
to-east) jet streams.
stream environment is optimum for the production of
sheared stable layers, about two-thirds of CAT occur-
rences are found near the jet stream. As a function of Mesoscale Forcing
height, CAT frequencies reach a maximum near the
The processes by which gravity waves can produce
tropopause, and in the baroclinic zone below the jet
CAT have been studied extensively for a well-known
core (Figure 1).
type of IGW known as a mountain lee wave (typical
-
wavelength: 10 km). Lee waves are generated when
stable airflow passes over a topographical barrier. In
most cases associated with strong turbulence, energy
from these waves is 'trapped' at or below a strong
temperature inversion near mountaintop level. On the
lee side of the ridge the airflow initially descends and
then rebounds to generate wave updrafts. As the flow
reaches its peak altitude and begins to descend again, a
wave crest is formed which may be made evident by
the formation of lenticular clouds. Under the crests of
strong waves, turbulent rotor circulations are often
-Cyclonic *Anticyclonic4 found in lee wave systems. Occasionally, with large
mountain ranges, rotor-produced turbulence will
reach the altitude range of CAT.
"Altitude (km)
In some situations, depending on the wavelength
- 9 and the distribution of wind velocity and temperature
- 8 with altitude, lee waves may be vertically propagating,
Cold Warm
- 7 such that the wave energy is transferred to great
Ridge 6 heights to produce CAT by one or more of the
-500 -250 0 250 500 processes described earlier. Atmospheric conditions
Distance from jet (km) usually associated with such mountain waves are: (1)a
high, cold tropopause, (2)a thermal inversion near the
Figure 1 Schematic cross-sectionsthrough the core of the polar mountain top, ( 3 ) low level winds >15rns-l at
jet stream for a trough (A) and a ridge (B).Typical locations of
mountain top level, (4)weak positive vertical wind
CAT (dashed lines) are shown relative to the jet core (J), the
tropopauses (heavy solid lines) and frontal zones (thin solid lines). shear, and ( 5 )weak stability above the inversion.
For a straight jet stream (not shown). Horizontal shear conditions The increase of lee wave amplitude with height leads
(cyclonic or anticyclonic) are indicatedat top of each cross section. to greater vertical displacement and tilting of the
Figure 2 Idealized streamlines showing macroscale flow patterns that are most conducive to the occurrence of CAT. Hatched areas
show where CAT is most likely. Heavy arrows represent approximate jet stream locations. The patterns are defined as: (A) col or
deformation zone, (B) sharp trough, (C) ridge, and (D) baroclinic instability. With (D), the CAT would most likely occur in or near dense
cirrus clouds, since this is a favorable flow pattern for cyclogenesis. In day-to-day situations, there are many possible hybrid combinations
of these patterns.
stable layers, and often reduces the local Ri enough to mountainous areas provide conditions favorable for
produce strong CAT. Furthermore, very strong wave the development of mountain lee waves. The enhance-
forcing occasionally produces a lee wave hydraulic ment of CAT-producing mechanisms by mountain
jump condition, resulting in a deep, extremely turbu- waves accounts for the higher frequency of CAT over
lent layer in the lee of the mountain, which may extend midlatitude mountains than elsewhere. However,
from near the surface through the tropopause. An owing to the set of conditions required for breaking
example of an aircraft encounter with severe moun- lee waves mentioned earlier, mountain wave-related
tain wave turbulence is shown in Figure 3. CAT is not usually found beneath the jet stream core,
but mainly on the anticyclonic side.
-
In some cases, a large mountain range will give
rise to unusually long (wavelength 50 km), nearly
hydrostatic lee waves. When favorable atmospheric
Another phenomenon that may produce conditions
favorable for CAT is deep mesoscale convection. If
wind and temperature profiles exist, shorter wave- convective elements penetrate a capping stable layer
length perturbations induced by these waves may such as the tropopause, and the stable layer is sheared,
destabilize, owing to partial reflection of lee wave then KHI may result. Also, longer IGW may be
energy from levels near the tropopause. Aircraft flying produced. If stability and wind conditions are favor-
through such wave action may experience extreme able then the IGW will propagate to altitudes where
vertical gusts. they may contribute to CAT.
As described previously, cyclones, strong cold fron- In cases of exceptionally strong convection, tops of
tal systems, and jet streams occurring over midlatitude thunderstorm cells penetrate the tropopause with
0.0 30.0 60.0 90.0 120.0 150.0180.0 210.0240.0270.0 300.0

Time (s)
Figure3 Time history of vertical velocities associatedwith mountainwaves and severe turbulence as measuredby an aircraft at 12.4km
(39000ft) altitude near Morton, WY, on 16 July 1982.Long-period, relatively smooth mountain waves along the flight path (from left to
right) deteriorate into extreme CAT conditions associated with KHI.
'.
vertical velocities as high as 20-30 m s - As a result, commercial jetliner for a convectively forced CAT
IGW develop in the stable stratosphere and propagate encounter that resulted in injuries to passengers.
away from the thunderstorm, generating conditions CAToccurrence near thunderstorms in midlatitudes
favorable to CAT production. If strong winds are occurs more often during spring and early summer in
present, flow regimes similar to mountain lee waves late afternoon and evening. However, large mesoscale
develop in the stratosphere, over and downwind of the convective systems (MCS) that form typically near
convection. Therefore, at tropopause levels, an air- the axis of an upper level ridge at night tend to enhance
craft flying near thunderstorm tops is vulnerable not the strength of the jet stream on the poleward side of
only to turbulence produced within a thunderstorm the MCS. This often results in unstable flows which
but also to CAToutside the thunderstorm. Figure 4 is a generate a significant area of CAT downstream
record of vertical acceleration and altitude from a that may continue for several hours after sunrise.
37.5 -
--
37.4 11.40
37.3 -
- 11.35
37.2 -
~
37.1-.
- 11.30
37.0 -
36.9 -- l 1 '25
Figure 4 Time history of vertical accelerations (solid) and altitude (dashed) along the flight path of a commercial jet aircraft near
Hannibal, MO, from 0121 to 0127 UTC on 4 April 1981.The point at which the aircraft flew downwind over the line of thunderstorms is
labeled 'Squall Line'.
Similar conditions may occur on the poleward side of of maximum CAT occurrence tends to weaken and
tropical cyclones as they encounter strong westerlies in shift poleward during the warm season in each
the midlatitudes. hemisphere, as one would expect. Large year-to-year
changes have also been observed in these data, that are
likely related to the El Niiio Southern Oscillation
Climatology Studies (ENSO) phenomenon. During strong El Niiio condi-
Some advances in our knowledge of the global tions, the subtropical jet stream in the Northern
distribution of CAT along heavily traveled airways Hemisphere is typically much stronger than normal,
have been derived from programs to collect pilot resulting in a higher potential for CAT occurrence.
reports (PIREPs), such as one conducted by the
International Civil Aviation Organization in the mid- Observation of CAT
1960s. A more comprehensive global ‘climatology’ of
large-scale, upper-level conditions favorable for CAT Observations of CAT are critical for research, diagno-
has been constructed by using a numerical model to sis, and prognosis of CAT. However, adequate resolu-
determine the distribution of a globally averaged CAT tion of CAT requires microscale measurements. Such
index equal to the product of horizontal deformation measurements are not regularly available via the
and vertical wind shear (described below, under ‘CAT standard surface and upper-air weather observation
Prediction Techniques’). This approach describes CAT networks. Occasionally, instrumented and radar-
produced by jet streams and upper fronts (e.g. cyclo- tracked balloons have been used for turbulence
genesis), but does not account for mountain waves and measurements, but aircraft have been the most
convection. Based on long-term averages of model frequent platforms of choice for the direct measure-
output, regions of relatively high CAT risk have been ment of CAT.
identified, including central and eastern Asia, the
Aircraft Measurements
southern United States, North Africa, and southern
Australia and New Zealand (Figure 5). The seasonal Subjective PIREPs include a description of CAT inten-
variation of the mean CAT index shows that the zone sity, aircraft position and altitude, and appropriate
Figure 5 Global mean turbulence index for the winter months (Decemberthrough Februaryin NorthernHemisphere, and June through
August in the Southern Hemisphere) showing relativeoccurrenceof CAT at approximately9.5 to 10.8 km (30 000 to 34 OOOft), basedon 9
years of data for the Northern Hemisphere, and 2 years for the SouthernHemisphere. The indexwas derivedfrom the NationalCenters for
EnvironmentalPredictionAviation Model at 0000 UTC daily. Turbulence areas are likely to be underestimated,since the index does not
account for mountain waves or deep convection.
remarks. Reports of turbulence intensity for Ground-Based Measurements

most aircraft are based on the pilot’s estimate of
flight control difficulty, of the movement of objects Although lacking the mobility and limited in number,
within the aircraft, or readings from available ground-based systems have generally had the advan-
instrumentation, for example, airspeed fluctuations, tage over airborne systems in size, power, and data
rate of climb in otherwise level flight, and g-meter processing capability to enable the sensing of turbu-
excursions. Bias is frequently introduced in these lence and turbulence-related parameters at greater
reports as a function of aircraft type, suddenness of ranges.
onset, and pilot experience, among several factors. Ground-tracked sounding balloon systems, such as
Despite these problems, pilot reports of turbulence radiosondes, are used ordinarily for the determination
are an important day-to-day source of direct CAT of indirect turbulence indicators such as stability, wind
measurements. velocity, wind shear, and Ri. An ongoing problem
Improvements in instrumentation and communica- with the twice-daily (0000 UTC and 1200 UTC) radio-
tions have made it possible for automated pilot reports sonde observations is the poor detection of thin,
from some commercial airliners to be acquired very potentially turbulent layers, which may vary rapidly in
quickly by international aviation weather forecast time and space. A few specialized sounding balloon
centers, increasing the timeliness and volume of CAT systems, such as those used for the support of the
reports. launch of space vehicles (e.g., radar tracking of the
Some aeronautical and atmospheric research rigid, Jimsphere balloons at Cape Canaveral), are
aircraft are equipped to directly measure true capable of resolving layers of 5 0 m or less. Some
turbulence gust velocities with high accuracy and studies have used balloon rate-of-rise fluctuations to
sample rates (e.g., about 0.3 m s and 50 samples per find a direct correlation between large-amplitude IGW
second, respectively). These measurements require vertical motions and turbulence in CAT regions.
specially calibrated sensors, high-capacity data re- Sensitive, vertically pointing, scanning Doppler
cording systems and judicious, post-flight engineering radars, known as wind profilers, are used to acquire
analysis. time sequences of wind speed and direction as a
Most commercial airliners carry onboard inertial function of altitude. VHF wind profilers are capable of
navigation and digital recording systems that are making a complete sounding about every 6minutes
capable of sampling vertical and horizontal accelera- with samples at 250 m intervals over an altitude range
tions plus several other parameters including aircraft of 2 to 16 km above ground level (AGL). Typically
attitude, engine status, position, and altitude. Such averaged over a period of an hour to eliminate noise,
onboard measurement systems provide information profiler data are most useful for determining the
for research into the nature and impact of CAT; also, altitude, intensity, and temporal behavior of larger
they have laid the foundation for the development of scale features associated with CAT, such as shear
onboard turbulence metrics that lend themselves to zones, trough lines, and jet streams.
automated and standardized CAT reporting, minimiz- Very powerful ground-based research radars are
ing the bias of subjective CAT reports. Two metrics capable of detecting clear air echoes associated with
which suit these requirements are the turbulent kinetic KHI. Modern Doppler weather radar systems (e.g.,
energy (TKE)dissipation rate and the integrated TKE. the National Weather Service ( N W S ) Weather Sur-
The latter metric is derived from that part of the TKE veillance Radar-1988 Doppler (WSR-88D)) in the
spectrum where aircraft have the greatest sensitivity to United States can measure turbulence in clear air,
turbulence. based on the ‘spread’ of the velocity distribution,
Beyond using an aircraft as a platform for instru- although that capability is generally limited to the
ments to measure CAT directly, an ideal airborne boundary layer. However, most operational Doppler
CAT detection system should be able to detect radar systems may also be used to determine higher-
the location and intensity of CAT far enough ahead altitude wind information similar to that from wind
of the aircraft so that the pilot has sufficient time to profilers.
take evasive flight path deviations and/or to Ground-based research lidar has been used to
warn cabin crew and passengers to fasten seatbelts accurately measure wind profiles and observe violent
to minimize the effects of the turbulence. Al- wave-breaking episodes at jet stream altitudes. How-
though airborne remote sensing of CAT can be ever, such observations cannot be acquired under all
achieved partially with forward-looking infrared sky conditions; this is because the air at the turbulence
radiometers and lidars, the best approach may locations must have enough particles to reflect some of
be the integration of data from multiple detection the lidar energy, and the intervening path between the
systems. lidar and the turbulence must not have so many
particulates or droplets that these would block the denoted by pronounced cirrus cloud bands that are
lidar beam or its return signal. oriented nearly perpendicular (transverse) to the flow,
possibly caused by inertial instability. An example of
Remote Sensing from Space this cloud feature is shown by the IR image in Figure 6 .
The advent of the geostationary meteorological satel- Wider, thicker transverse cloud bands have been
lites in the latter half of the twentieth century provided associated with a strong likelihood of moderate to
an additional tool for monitoring regions of potential severe CAT.
turbulence. Although pixel resolutions are only mar- Pronounced warming observed over a period of a
ginal for this purpose (1-5 km), the image frequency few hours in a series of water vapor images has been
(15min-1 h) and spatial coverage (global except near associated with strong subsidence and tropopause
the poles) of geostationary satellites can be useful, ‘folds’ in the vicinity of upper-level fronts, and a
when combined with other data, for the detection and corresponding increase in the risk of CAT. These
short-range forecasting of CAT. The spectral bands regions are usually associated with deformation zones
that have been found most useful for CAT detection (pattern (A), Figure 2 ) , sharp upper troughs (pattern
are visible, infrared (IR),and water vapor, described in (B)),or in the ‘dry slot’ portion of intensifying cyclones
Table 1. Specific applications of these images will be (pattern (D)).These synoptic patterns also correlate
described in more detail in the next section. Cloud with high concentrations of stratospheric ozone and
motion vectors derived from a sequence of geosta- large values of potential vorticity.
tionary satellite images can provide valuable data over Satellite images in several spectral bands (visible,IR,
remote regions for assimilation into numerical predic- and water vapor) may also show classic ‘washboard’
tion models. cloud patterns associated with mountain waves
Visible and IR images from polar orbiting satellites (Figure 7). The water vapor images often depict a
are available less frequently (every 2-6 h, depending greater area coverage of mountain wave conditions
on latitude), but can be used to identify CAT patterns than either visible or IR, owing to their greater
associated with long-lasting, large-scale systems, and sensitivity to moisture. Warm subsidence zones ori-
to corroborate features observed in geostationary ented along and just downstream from the mountain
satellite data. They are especially useful at high ridges and slightly upstream from the lee cirrus plume,
latitudes where geostationary coverage is poor be- sometimes referred to as ‘Fohn gaps’, are indicative of
cause of extreme parallax. possible intense turbulence.
Analysis of satellite imagery, compared with colo- High-resolution visible imagery (0.5-1 l m ) can
cated PIREPs, indicates that CAT is found not only in sometimes detect very small-scale (I1km wave-
clear air but also in cirrus clouds, and along borders of length) wave cloud patterns (referred to as billows)
large-scale cloud systems. Zones of turbulence asso- that correspond to areas of KHI. Billow clouds are
ciated with the subtropical jet stream are often embedded typically within or near large-scale cloud
systems, or in the vicinity of convective storms when
wind shears are present.
Table 1 Spectral channels used in detecting CAT-related phe-
nomena from space
Channel type Spectral range Applications CAT Prediction Techniques

bm)
As illustrated in Figure 2 , certain synoptic-scale upper
Visible 0.5-1 .O Small scale ( 51 km) flow patterns have been empirically related to CAT
wave clouds occurrence through many years of operational expe-
Infrared (IR) 10-12 Cloud top
temperatures
rience. By applying these pattern types to predicted
(heights), large- flow patterns from numerical forecast models, esti-
scale cloud mates can be made of the likelihood of CAT in certain
borders, cloud regions for longer forecast time periods.
bands Objective prediction of CAT has become more
Water vapor 6-7 Regions of sinking/
commonplace and accurate, owing to increased com-
rising associated
with mountain puting speed, diminished cost of high-speed compu-
waves, upper ters, improved numerical model physics, and better
level fronts, techniques for assimilation of wind and temperature
deformation data from aircraft and satellites. A number of
zones, and
diagnostic and predictive indices for CAT have
cyclones
been developed and used over the years. The non-
Figure 6 Infrared image from the geostationary operational environmental satellite GOES-8 at 0045 UTC on 3 March 2000, showing
transverse cirrus cloud bands on the Equatorward side of the subtropical jet stream near the Leeward Islands. Moderate to severe
turbulence was reported bya B767aircraft atthe locationshown. Wind barbs are from radiosondesitesfor250 hPaat 0000 UTC, or aircraft
reports for 8.9 to 11.8 km (28 000 ft to 37 000 ft) between 0000 and 0300 UTC.
dimensional Richardson number (Ri)(see eqn [l])is Significant, prolonged turbulence is most likely in
adequate in many situations, although it often exhibits regions of the atmosphere where larger-scale processes
a much wider range of critical values in the free are continually acting to decrease Ri, despite the
atmosphere than in the laboratory. Ri is most reliable influence of turbulent mixing. The primary processes
when generated from high resolution observational include: horizontal deformation, convergence, and
data, therefore it is principally used as a diagnostic differential thermal advection. Specifically, the Ri
tool. Richardson number is defined as: tendency technique calculates the large-scale forcing
necessary to overcome the kinetic energy dissipation
resulting from the turbulence process in order to
PI i.
maintain the Ri at values 5 A simple formulation of
Richardson number tendency equation is
where 8 is potential temperature (K), V the vector
horizontal wind (m s--'), g acceleration of gravity 121
) , z is height (m).
( m ~ - ~and
A more successful index for CAT prediction em- where @ is nonturbulent, large-scale forcing by defor-
ploys the Richardson number tendency equation. mation, and E is turbulent dissipation.
Figure 7 Mountain wave patterns east of the Rocky Mountainsobserved by GOES-8 in the (A) water vapor, (8) infrared, and (C) visible
band at 1915 UTC on 5 April 2000. Considerable severe turbulence was reported below 6.4 km (20 000 ft). The watervapor image depicts
more extensivecoverage of mountainwaves than the othertwo images. Cirrus plumes observed in upper left are indicativeof high-altitude
mountain waves downwind of the Salt River and Wind River ranges in Wyoming.
A similar approach can generate a diagnostic or mechanisms, and (3) most numerical models cannot
predictive index by simply calculating the product of accurately account for the intense, subgrid-scale
horizontal, resultant deformation and the vertical vertical shears and strong horizontal forcing present
vector wind shear at each grid point. The basis of this during severe CAT. More specifically, model problems
index is similar to Ri tendency in that it considers are due to resolution limitations and systematic under-
large-scale forcing from frontogenesis, in addition to forecasting of maximum wind speeds within the jet
the presence of vertical shear in the mesoscale. One stream.
difficulty with this approach is that deformation can
also lead to frontolysis in some cases, which tends to Statistical Approaches
reduce vertical shear. This turbulence index (TI) is
defined as Because of the sporadic, microscale nature of CAT, it
would seem that statistical approaches would be
useful for forecasting its occurrence. The first such
TI = [(au/ax - aV/ayl2 + (av/ax + a ~ / a y ) ~ ] l / ~
Resultant deformation efforts were completed in the United Kingdom in the
late 1970s. Turbulence data from about 4500 aircraft
x avpz 131 reports, compared with eleven colocated numerical
Vertical shear
parameters derived from a coarse-resolution predic-
where u and v are wind components (ms-I), V the tion model, revealed that the best correlation was
vector wind, and z height (m). between CAT and vertical and horizontal wind shears.
The turbulent kinetic energy (TKE) technique for Similar studies were completed in the United States in
the prediction of CAT attempts to show areas of the 1980s using higher-resolution numerical model
turbulence generation through the processes of pro- data that showed CAT to be highly correlated with
duction by vertical shear, production or destruction by horizontal deformation and scalar wind speed.
buoyancy, energy advection, and loss via dissipation. A technique that statistically integrates information
The simplified TKE equation can be expressed as: from many numerical turbulence indices, known as
the integrated turbulence forecast algorithm (ITFA),
d (TKE) -
- au - av + -
w l e -g was developed in the late 1990s. ITFA first assigns a
dt
Energy production
az
Shear production
az Buoyancy e score to each diagnostic index, based on comparison
with available observations at the initial time period
+ A - & throughout the numerical grid domain. A weighted
Advection Energy dissipation [41
sum is then determined from all of the indices to arrive
TKE = (.I2 + v’* + w r 2 ) / 2or , total kinetic energy per
at a final ITFA index value at each grid point. The same
unit mass, where the terms u t , v’, and w‘ are pertur- weights are then assigned to each index in deriving
bation (gust) velocities. The TKE approach can predictions for each forecast time period. This process
account for turbulence from a wide variety of mech- is repeated with every forecast cycle.
anisms such as convection and mountain waves, as Although useful, statistical approaches to CAT
well as jet stream CAT, and provides a direct estimate prediction are not always able to correctly depict
of possible turbulence intensity through the energy turbulence events accurately because of limitations in
dissipation rate. While TKE presents the most rigorous climatological or regionally derived thresholds.
depiction of turbulent processes of all the objective
approaches, it is more effective when numerical See also
-
models have very high vertical ( 100 m) and hori-
zontal resolution (- 10 km). In solving eqn [4] using Aviation Weather Hazards. Convective Storms:
such high-resolution models, the need for accurate Convective Initiation. Dynamic Meteorology: Waves.
Frontogenesis. Instability: Inertial Instability. Jet
observations is critical, and the best analysis and Streaks. Kelvin-Helmholtz Instability. Lee Waves
prediction results are attained in regions of dense and Mountain Waves. Satellite Remote Sensing:
reporting networks that incorporate aircraft data, Temperature Soundings; Wind, Middle Atmosphere. Tur-
chiefly over continental areas. bulence and Mixing. Turbulence, Two Dimensional.
The indices described above do a reasonably good Turbulent Diffusion.
job of predicting large outbreaks of CAT, but tend to
predict CATover a region that is larger than the area of
actual occurrence (also known as ‘over-forecasting’). Further Reading
Their inadequacies are due largely to the following Anderson RK (1995) Synoptic scale cloud and moisture
factors: (1)they only account for large scale conditions patterns; and: Clouds generated by mountains. In: Bader
favorable for CAT, (2)they do not consider triggering MJ, Forbes GS, Grant JR, Lilley RB and Waters AJ (eds),
CLIMATE I Overview 403
Images in Weather Forecasting: A Practical Guide forGossard EE and Hooke WH (1975) Waves in the Atmos-
phere. Amsterdam: Elsevier.
Interpreting Satellite and Radar Imagery, pp. 96-137 and
472477. Cambridge: Cambridge University Press. Holton JR (1992) Introduction to Dynamic Meteorology.
Atlas D (ed.) (1990) Radar in Meteorology: Battan New York: Academic Press.
Memorial and 40th Anniversary Radar Meteorology Lilly D (1986) Instabilities. In: Ray P (ed.) Mesoscale
Conference. Boston, MA: American Meteorological Analysis and Forecasting, pp. 259-271. Boston, MA:
Society. American Meteorology Society.
Blumen W (ed.) (1990) Mountain waves and down- Panofsky HA and Dutton JA (1984) Atmospheric Turbu-
slope winds. Atmospheric Processes over Complex lence: Models and Methods for Engineering Application.
Terrain. ch. 4. Meteorological Monographs, 23.45, New York: Wiley.
pp. 59-8 1. Boston, MA: American Meteorological Soci-
Pao Y and Goldberg A (1969) Clear Air Turbulence and Its
ety. Detection. New York: Plenum Press.
Durran DR (1986) Mountain waves, In: Ray P (ed.) Vinnechenko NK, Pinus NZ, Shmeter SM and Shur GN
Mesoscale Analysis and Forecasting, pp. 472492. Bos- (1980) Turbulence in the Free Atmosphere. New York:
ton, MA: American Meteorology Society. Consultants Bureau.
Earth’s biological and geochemical resources. Because

Overview of the large human population of the Earth and the
adoption of technology by societies, humans now have
D L Hartmann, University of Washington, Seattle, WA, the ability to make relatively rapid changes in the
USA Earth’s global climate.
Introduction Description of Climate

Climate is the composite or generalization of weather The average surface temperature of the Earth is about
conditions of a region as a function of season. It can be 288K or 15°C. The global average precipitation is
expressed in terms of the expected values of meteor- about 1meter per year. The climate of a given location
ological variables such as temperature, precipitation, varies with latitude, altitude, and geographical condi-
pressure, humidity, cloudiness, sunshine, and winds. tions (Figure 1).Seasonal variations are greater in
The expected values are usually obtained by averaging higher latitudes and in continental rather than mari-
observations over a number of years. A complete time areas (Figure 2). Oceans have a large capacity to
description of the climate would also include infor- store heat, so that seasonal variations in surface
mation on the year-to-year variability of averages over temperature are tempered by heat exchange with the
shorter time intervals such as months. Climatology is ocean.
the scientific study of climate. A complete understand- The distribution of precipitation is more complex.
ing of climate requires a thorough understanding of Precipitation is greatest near the Equator, is reduced in
the atmosphere and its physical and chemical interac- subtropical latitudes (15-30 degrees), and increases
tions with the ocean and the land surface. Life forms in again in middle latitudes (Figure 3 ) . The rainfall
the ocean and on the land influence the composition of maximum near the Equator is associated with the
the air, the color of the ocean, and the exchange of general convergence of wind at low levels and the high
energy and moisture between the land surface and the water vapor content of tropical air. This precipitation
atmosphere. maximum moves north and south across the Equator
Climate is important for humanity and life on Earth, following the position of the Sun relative to the
since it has set the context for human evolution and Equator. The very low precipitation values that occur
subsequent social, political, and historical develop- in the tropics are associated with downward motion
ments. In the modern world it influences agriculture, and divergence of the surface winds. Precipitation
water resources, human health, and energy use. It maxima occur in midlatitudes where cyclonic storms
continues to play an important role in natural ecology are frequent and produce heavy rainfall. Two rela-
and the interaction of human endeavor with the tively strong precipitation maxima in the Northern
Modern Gas Kinetics, pp. 99-134. Oxford: Blackwell Wayne RP (1988) Principles and Applications of Photo-
Scientific Publications. chemistry. Oxford: Oxford University Press.
Smith IWM (1997) Collisional energy transfer, intramolec- Wayne RP (2000) Photochemistry and kinetics
ular vibrational relaxation and unimolecular reactions. applied to atmosphere. Chemistry of Atmospheres,
Journal of the Chemical Society, Faraday Transactions 3rd edn, pp. 97-137. Oxford: Oxford University
93: 3741-3750. Press.
See CLOUDS: Classification
G P Ellrod, National Oceanographic and Atmospheric turbulence intensity is based on deviations in vertical
Administration/National Environmental Satellite, Data, acceleration from the normal acceleration of gravity
and Information Service, Camp Springs, MD, USA (1g). These categories and their threshold devia-
P F Lester, San Jose State University, San Jose, CA, tions are: light (f0.2g), moderate (f0.5 g), severe
USA ( f1.0 g), and extreme (f2.0g). The physical impact
of CATon crew and passengers varies from discomfort
L J Ehernberger, National Aeronautics and Space
Administration Dryden Flight Research Center,
for the lighter turbulence categories to loss of flight
Edwards, CA, USA control during the rare extreme turbulence event. In
the most intense episodes, injuries, and in some very
rare cases, fatalities have occurred. Unrestrained crew
and passengers are especially vulnerable. Flight
through turbulent conditions also produces stresses
Introduction on the airframe. Repeated turbulence encounters over
Since the first aircraft flight, pilots have been aware of the lifetime of the aircraft may lead to metal fatigue
in-flight turbulence. Because the known turbulence of and, in extremely rare cases, structural failure.
the time was tied to strong, low-level winds, rough
terrain, and convection, some of the early pilots
predicted that, with the exception of thunderstorms, Causative Mechanisms and
the ability to attain higher flight altitude with pressu-
rized cabins would be accompanied by a marked
Characteristics
decrease in turbulence. This was not to be. In the Aircraft turbulence (including CAT) differs from
1940s, as fighter aircraft attained tropopause alti- turbulence in fluids as usually conceived. Aircraft
tudes, they experienced a previously unknown phe- turbulence is defined in terms of the aircraft response
nomenon: clear air turbulence (CAT), so called while classical turbulence is defined in terms of the
because initial encounters occurred in areas devoid state of the fluid. Simply put, aircraft turbulence is
of clouds. As aircraft were designed to fly higher and ‘bumpiness in flight’. The difference is critical in that
faster during the last half of the twentieth century, bumpiness depends on aircraft design, weight, speed,
CAT became the focus of many organized research and pilot input, in addition to the state of the
efforts. Our knowledge of CAT has grown substan- atmosphere (gusts, windshears, or waves), which
tially as a result. may or may not be turbulent in the classical sense.
CAT is now defined as aircraft turbulence that The discussion that follows considers atmospheric
occurs at altitudes of 5.6 km (about 500 hPa) or higher, phenomena that contribute to CATon a wide range of
either in cloud-free conditions or within stratiform scales. This section examines briefly the primary
clouds. The critical influence of CAT is on flight safety. physical characteristics and causes of microscale
One of the oldest schemes for the characterization of CAT. It then considers macroscale and mesoscale
forcing mechanisms that create a favorable environ- Ri becomes 1.0 or more, and KHI are present, they will
ment for CAT development. decrease in amplitude. The use of Ri as a practical
index for CAT diagnosis and prediction is discussed in
Kelvin-Helmholtz Waves and CAT a later section.
Although sheared layers in which unstable KHI
Our understanding of the production of CAT and its
occur are, typically, less than 1km in depth, horizontal
characteristics is rooted in theoretical studies of fluid dimensions are much larger (Le., from -10km
mechanics, in laboratory and numerical experiments, to a few hundred kilometers, often elongated in
and in field studies. The evidence from these investi- the direction of the wind). Thus, an individual
gations is that Kelvin-Helmholtz instability (KHI) KHI develops over a considerably smaller horizontal
episodes are the cause of a large fraction of CAT. scale than the sheared stable layers in which it
Growth and intensification of atmospheric turbulence is embedded. This observation has four important
occur when the destabilizing effect of wind shear is ramifications:
sufficient to overcome the dampening influence of
static stability. KHI produces shearing-gravity waves
with typical horizontal wavelengths of a few tens of 1. An aircraft flying in a sheared layer may experience
meters to a few kilometers, precisely the range of eddy CATover a distance much greater than the scale of
sizes to which most aircraft will have the maximum an individual KHI.
response. 2. CAT encounters are often burst-like (highly inter-
KHI arises from micro- and mesoscale wind shear mittent) as an aircraft crosses a thin sheared layer
intensification, when smooth, wavelike oscillations or as an aircraft flying within a sheared layer
within a sheared, statically stable layer grow in intercepts individual KHI elements in different
amplitude to the point where the wave crests overturn, stages of development.
or ‘break’. Wave-breaking at wavelengths of hundreds 3 . Regular, comprehensive observations of CAT are
of meters is highly effective in producing CAT, with a difficult to acquire because of its small scale and
rapid cascade of energy from the KHI to smaller-scale intermittent nature.
turbulence and dissipation. With respect to CAT, some 4. Conditions conducive to the development of
of the more important characteristics of KHI are the CAT are rooted in larger-scale processes that
following: produce and perturb the extensive, sheared stable
layers.
1. The typical lifetime of an individual KHI is about
5 minutes. Internal Gravity Waves and CAT
2. The length of the dominant wave (most unstable
Internal gravity waves (IGW)with horizontal scales of
KHI mode) is proportional to the depth of the
a few kilometers to a few hundred kilometers are also
sheared layer (i.e., about six times the depth).
mechanisms for CAT production. IGW may become
However, as KHI-induced turbulence and mixing
significant in the production of CAT in several ways.
modify the local wind shear structure and stability
Wave amplitude variations with height are caused by
stratification, variations in the KHI wavelengths
the decrease in density with altitude and by the
can be expected.
variation of background stability and wind with
3 . The intensity of the turbulence produced by KHI is
height. Aircraft intersecting large-amplitude IGW
in direct proportion to the initial wind shear across
may be exposed to ‘sharp-edged’ gusts or periodic
the layer. If the turbulent mixing caused by KHI
vertical motions which may be interpreted as CAT.
sufficiently weakens the background wind shear,
Vertical displacements due to gravity wave motions
the turbulence will decay and the flow will again
will modulate background wind shear, leading to the
become laminar. However, if turbulent mixing
production of microscale KHI. Near or slightly below
strengthens the wind shears near the boundaries
the ‘critical layer’ (where phase speed and background
of the old turbulence layer then new KHI may
wind speed are the same), wave-breaking may produce
develop.
intense turbulence. Finally, the occurrence and inten-
Another relevant result from studies of KHI has been sity of CAT are also affected by the excitation of IGW
the development of a basic dynamic instability prin- from different sources as well as by resonant nonlinear
ciple for a simple model of shearing gravity waves. interactions between different IGW modes and be-
Based on linear theory, the Miles-Howard criterion tween IGW and KHI in turbulent layers, can affect the
states that unstable wave modes resulting from verti- occurrence and intensity of CAT (see Dynamic Mete-
cal shear is likely to occur when the local gradient orology: Waves. Lee Waves and Mountain Waves.
Richardson number, Ri,is less than 0.25. In contrast, if Middle Atmosphere: Gravity Waves).
Macroscale Forcing Macroscale flow patterns conducive to CAT are

shown schematically in Figure 2. They can be classified
Any atmospheric circulation system that produces
into four basic types: (A)col pattern, (B)sharp trough,
deformation and/or convergence in the flow field,
(C) ridge, and (D) baroclinic instability. Based on
differential horizontal temperature advection, or sec-
observational studies, flow conditions resulting in
ondary (vertical) circulations has the potential to
anticyclonic shear or curvature (such as pattern (C),
create and strengthen sloping stable layers and their
Figure 2) produce CAT most frequently, while cyclonic
associated atmospheric stability and wind shear.
conditions (pattern (B), Figure 2) produce the most
Because of thermal wind requirements, such layers
intense CAT. While patterns (B) and (D) are both
(also called baroclinic layers, or frontal zones) are
associated with troughs, the orientation of the troughs
vertically sheared. Frontal zones are particularly
results in significant differences in the locations of
common with extratropical cyclones and associated
CAT and prevalent cloud conditions. A north-east to
upper tropospheric jet streams, thus presenting opti-
south-west oriented trough axis (pattern (B))results in
mal conditions for development of KHI and CAT.
CAT to the rear of the upper trough in relatively cloud-
Additionally, developing extratropical cyclones and
free conditions. A north-west to south-east trough
jet streaks, as well as sharply curved anticyclonic
(pattern (D)),typically associated with cyclogenesis,
flows, are characterized by unbalanced (ageostrophic)
results in CAT in advance of the trough, often
flow. Under these conditions, accelerations generate a embedded in thick cirrus clouds. As suggested by
broad spectrum of not only IGW but also longer,
Figure 2, sharply curved jet stream segments associ-
gravity-inertia waves which may also be effective in
ated with upper-level troughs or ridges are typically
producing CAT by modulating shallow shear layers, as
more turbulent than most zonal (predominately west-
well as by their own instabilities. Because the jet
to-east) jet streams.
stream environment is optimum for the production of
sheared stable layers, about two-thirds of CAT occur-
rences are found near the jet stream. As a function of Mesoscale Forcing
height, CAT frequencies reach a maximum near the
The processes by which gravity waves can produce
tropopause, and in the baroclinic zone below the jet
CAT have been studied extensively for a well-known
core (Figure 1).
type of IGW known as a mountain lee wave (typical
-
wavelength: 10 km). Lee waves are generated when
stable airflow passes over a topographical barrier. In
most cases associated with strong turbulence, energy
from these waves is 'trapped' at or below a strong
temperature inversion near mountaintop level. On the
lee side of the ridge the airflow initially descends and
then rebounds to generate wave updrafts. As the flow
reaches its peak altitude and begins to descend again, a
wave crest is formed which may be made evident by
the formation of lenticular clouds. Under the crests of
strong waves, turbulent rotor circulations are often
-Cyclonic *Anticyclonic4 found in lee wave systems. Occasionally, with large
mountain ranges, rotor-produced turbulence will
reach the altitude range of CAT.
"Altitude (km)
In some situations, depending on the wavelength
- 9 and the distribution of wind velocity and temperature
- 8 with altitude, lee waves may be vertically propagating,
Cold Warm
- 7 such that the wave energy is transferred to great
Ridge 6 heights to produce CAT by one or more of the
-500 -250 0 250 500 processes described earlier. Atmospheric conditions
Distance from jet (km) usually associated with such mountain waves are: (1)a
high, cold tropopause, (2)a thermal inversion near the
Figure 1 Schematic cross-sectionsthrough the core of the polar mountain top, ( 3 ) low level winds >15rns-l at
jet stream for a trough (A) and a ridge (B).Typical locations of
mountain top level, (4)weak positive vertical wind
CAT (dashed lines) are shown relative to the jet core (J), the
tropopauses (heavy solid lines) and frontal zones (thin solid lines). shear, and ( 5 )weak stability above the inversion.
For a straight jet stream (not shown). Horizontal shear conditions The increase of lee wave amplitude with height leads
(cyclonic or anticyclonic) are indicatedat top of each cross section. to greater vertical displacement and tilting of the
Figure 2 Idealized streamlines showing macroscale flow patterns that are most conducive to the occurrence of CAT. Hatched areas
show where CAT is most likely. Heavy arrows represent approximate jet stream locations. The patterns are defined as: (A) col or
deformation zone, (B) sharp trough, (C) ridge, and (D) baroclinic instability. With (D), the CAT would most likely occur in or near dense
cirrus clouds, since this is a favorable flow pattern for cyclogenesis. In day-to-day situations, there are many possible hybrid combinations
of these patterns.
stable layers, and often reduces the local Ri enough to mountainous areas provide conditions favorable for
produce strong CAT. Furthermore, very strong wave the development of mountain lee waves. The enhance-
forcing occasionally produces a lee wave hydraulic ment of CAT-producing mechanisms by mountain
jump condition, resulting in a deep, extremely turbu- waves accounts for the higher frequency of CAT over
lent layer in the lee of the mountain, which may extend midlatitude mountains than elsewhere. However,
from near the surface through the tropopause. An owing to the set of conditions required for breaking
example of an aircraft encounter with severe moun- lee waves mentioned earlier, mountain wave-related
tain wave turbulence is shown in Figure 3. CAT is not usually found beneath the jet stream core,
but mainly on the anticyclonic side.
-
In some cases, a large mountain range will give
rise to unusually long (wavelength 50 km), nearly
hydrostatic lee waves. When favorable atmospheric
Another phenomenon that may produce conditions
favorable for CAT is deep mesoscale convection. If
wind and temperature profiles exist, shorter wave- convective elements penetrate a capping stable layer
length perturbations induced by these waves may such as the tropopause, and the stable layer is sheared,
destabilize, owing to partial reflection of lee wave then KHI may result. Also, longer IGW may be
energy from levels near the tropopause. Aircraft flying produced. If stability and wind conditions are favor-
through such wave action may experience extreme able then the IGW will propagate to altitudes where
vertical gusts. they may contribute to CAT.
As described previously, cyclones, strong cold fron- In cases of exceptionally strong convection, tops of
tal systems, and jet streams occurring over midlatitude thunderstorm cells penetrate the tropopause with
0.0 30.0 60.0 90.0 120.0 150.0180.0 210.0240.0270.0 300.0

Time (s)
Figure3 Time history of vertical velocities associatedwith mountainwaves and severe turbulence as measuredby an aircraft at 12.4km
(39000ft) altitude near Morton, WY, on 16 July 1982.Long-period, relatively smooth mountain waves along the flight path (from left to
right) deteriorate into extreme CAT conditions associated with KHI.
'.
vertical velocities as high as 20-30 m s - As a result, commercial jetliner for a convectively forced CAT
IGW develop in the stable stratosphere and propagate encounter that resulted in injuries to passengers.
away from the thunderstorm, generating conditions CAToccurrence near thunderstorms in midlatitudes
favorable to CAT production. If strong winds are occurs more often during spring and early summer in
present, flow regimes similar to mountain lee waves late afternoon and evening. However, large mesoscale
develop in the stratosphere, over and downwind of the convective systems (MCS) that form typically near
convection. Therefore, at tropopause levels, an air- the axis of an upper level ridge at night tend to enhance
craft flying near thunderstorm tops is vulnerable not the strength of the jet stream on the poleward side of
only to turbulence produced within a thunderstorm the MCS. This often results in unstable flows which
but also to CAToutside the thunderstorm. Figure 4 is a generate a significant area of CAT downstream
record of vertical acceleration and altitude from a that may continue for several hours after sunrise.
37.5 -
--
37.4 11.40
37.3 -
- 11.35
37.2 -
~
37.1-.
- 11.30
37.0 -
36.9 -- l 1 '25
Figure 4 Time history of vertical accelerations (solid) and altitude (dashed) along the flight path of a commercial jet aircraft near
Hannibal, MO, from 0121 to 0127 UTC on 4 April 1981.The point at which the aircraft flew downwind over the line of thunderstorms is
labeled 'Squall Line'.
Similar conditions may occur on the poleward side of of maximum CAT occurrence tends to weaken and
tropical cyclones as they encounter strong westerlies in shift poleward during the warm season in each
the midlatitudes. hemisphere, as one would expect. Large year-to-year
changes have also been observed in these data, that are
likely related to the El Niiio Southern Oscillation
Climatology Studies (ENSO) phenomenon. During strong El Niiio condi-
Some advances in our knowledge of the global tions, the subtropical jet stream in the Northern
distribution of CAT along heavily traveled airways Hemisphere is typically much stronger than normal,
have been derived from programs to collect pilot resulting in a higher potential for CAT occurrence.
reports (PIREPs), such as one conducted by the
International Civil Aviation Organization in the mid- Observation of CAT
1960s. A more comprehensive global ‘climatology’ of
large-scale, upper-level conditions favorable for CAT Observations of CAT are critical for research, diagno-
has been constructed by using a numerical model to sis, and prognosis of CAT. However, adequate resolu-
determine the distribution of a globally averaged CAT tion of CAT requires microscale measurements. Such
index equal to the product of horizontal deformation measurements are not regularly available via the
and vertical wind shear (described below, under ‘CAT standard surface and upper-air weather observation
Prediction Techniques’). This approach describes CAT networks. Occasionally, instrumented and radar-
produced by jet streams and upper fronts (e.g. cyclo- tracked balloons have been used for turbulence
genesis), but does not account for mountain waves and measurements, but aircraft have been the most
convection. Based on long-term averages of model frequent platforms of choice for the direct measure-
output, regions of relatively high CAT risk have been ment of CAT.
identified, including central and eastern Asia, the
Aircraft Measurements
southern United States, North Africa, and southern
Australia and New Zealand (Figure 5). The seasonal Subjective PIREPs include a description of CAT inten-
variation of the mean CAT index shows that the zone sity, aircraft position and altitude, and appropriate
Figure 5 Global mean turbulence index for the winter months (Decemberthrough Februaryin NorthernHemisphere, and June through
August in the Southern Hemisphere) showing relativeoccurrenceof CAT at approximately9.5 to 10.8 km (30 000 to 34 OOOft), basedon 9
years of data for the Northern Hemisphere, and 2 years for the SouthernHemisphere. The indexwas derivedfrom the NationalCenters for
EnvironmentalPredictionAviation Model at 0000 UTC daily. Turbulence areas are likely to be underestimated,since the index does not
account for mountain waves or deep convection.
remarks. Reports of turbulence intensity for Ground-Based Measurements

most aircraft are based on the pilot’s estimate of
flight control difficulty, of the movement of objects Although lacking the mobility and limited in number,
within the aircraft, or readings from available ground-based systems have generally had the advan-
instrumentation, for example, airspeed fluctuations, tage over airborne systems in size, power, and data
rate of climb in otherwise level flight, and g-meter processing capability to enable the sensing of turbu-
excursions. Bias is frequently introduced in these lence and turbulence-related parameters at greater
reports as a function of aircraft type, suddenness of ranges.
onset, and pilot experience, among several factors. Ground-tracked sounding balloon systems, such as
Despite these problems, pilot reports of turbulence radiosondes, are used ordinarily for the determination
are an important day-to-day source of direct CAT of indirect turbulence indicators such as stability, wind
measurements. velocity, wind shear, and Ri. An ongoing problem
Improvements in instrumentation and communica- with the twice-daily (0000 UTC and 1200 UTC) radio-
tions have made it possible for automated pilot reports sonde observations is the poor detection of thin,
from some commercial airliners to be acquired very potentially turbulent layers, which may vary rapidly in
quickly by international aviation weather forecast time and space. A few specialized sounding balloon
centers, increasing the timeliness and volume of CAT systems, such as those used for the support of the
reports. launch of space vehicles (e.g., radar tracking of the
Some aeronautical and atmospheric research rigid, Jimsphere balloons at Cape Canaveral), are
aircraft are equipped to directly measure true capable of resolving layers of 5 0 m or less. Some
turbulence gust velocities with high accuracy and studies have used balloon rate-of-rise fluctuations to
sample rates (e.g., about 0.3 m s and 50 samples per find a direct correlation between large-amplitude IGW
second, respectively). These measurements require vertical motions and turbulence in CAT regions.
specially calibrated sensors, high-capacity data re- Sensitive, vertically pointing, scanning Doppler
cording systems and judicious, post-flight engineering radars, known as wind profilers, are used to acquire
analysis. time sequences of wind speed and direction as a
Most commercial airliners carry onboard inertial function of altitude. VHF wind profilers are capable of
navigation and digital recording systems that are making a complete sounding about every 6minutes
capable of sampling vertical and horizontal accelera- with samples at 250 m intervals over an altitude range
tions plus several other parameters including aircraft of 2 to 16 km above ground level (AGL). Typically
attitude, engine status, position, and altitude. Such averaged over a period of an hour to eliminate noise,
onboard measurement systems provide information profiler data are most useful for determining the
for research into the nature and impact of CAT; also, altitude, intensity, and temporal behavior of larger
they have laid the foundation for the development of scale features associated with CAT, such as shear
onboard turbulence metrics that lend themselves to zones, trough lines, and jet streams.
automated and standardized CAT reporting, minimiz- Very powerful ground-based research radars are
ing the bias of subjective CAT reports. Two metrics capable of detecting clear air echoes associated with
which suit these requirements are the turbulent kinetic KHI. Modern Doppler weather radar systems (e.g.,
energy (TKE)dissipation rate and the integrated TKE. the National Weather Service ( N W S ) Weather Sur-
The latter metric is derived from that part of the TKE veillance Radar-1988 Doppler (WSR-88D)) in the
spectrum where aircraft have the greatest sensitivity to United States can measure turbulence in clear air,
turbulence. based on the ‘spread’ of the velocity distribution,
Beyond using an aircraft as a platform for instru- although that capability is generally limited to the
ments to measure CAT directly, an ideal airborne boundary layer. However, most operational Doppler
CAT detection system should be able to detect radar systems may also be used to determine higher-
the location and intensity of CAT far enough ahead altitude wind information similar to that from wind
of the aircraft so that the pilot has sufficient time to profilers.
take evasive flight path deviations and/or to Ground-based research lidar has been used to
warn cabin crew and passengers to fasten seatbelts accurately measure wind profiles and observe violent
to minimize the effects of the turbulence. Al- wave-breaking episodes at jet stream altitudes. How-
though airborne remote sensing of CAT can be ever, such observations cannot be acquired under all
achieved partially with forward-looking infrared sky conditions; this is because the air at the turbulence
radiometers and lidars, the best approach may locations must have enough particles to reflect some of
be the integration of data from multiple detection the lidar energy, and the intervening path between the
systems. lidar and the turbulence must not have so many
particulates or droplets that these would block the denoted by pronounced cirrus cloud bands that are
lidar beam or its return signal. oriented nearly perpendicular (transverse) to the flow,
possibly caused by inertial instability. An example of
Remote Sensing from Space this cloud feature is shown by the IR image in Figure 6 .
The advent of the geostationary meteorological satel- Wider, thicker transverse cloud bands have been
lites in the latter half of the twentieth century provided associated with a strong likelihood of moderate to
an additional tool for monitoring regions of potential severe CAT.
turbulence. Although pixel resolutions are only mar- Pronounced warming observed over a period of a
ginal for this purpose (1-5 km), the image frequency few hours in a series of water vapor images has been
(15min-1 h) and spatial coverage (global except near associated with strong subsidence and tropopause
the poles) of geostationary satellites can be useful, ‘folds’ in the vicinity of upper-level fronts, and a
when combined with other data, for the detection and corresponding increase in the risk of CAT. These
short-range forecasting of CAT. The spectral bands regions are usually associated with deformation zones
that have been found most useful for CAT detection (pattern (A), Figure 2 ) , sharp upper troughs (pattern
are visible, infrared (IR),and water vapor, described in (B)),or in the ‘dry slot’ portion of intensifying cyclones
Table 1. Specific applications of these images will be (pattern (D)).These synoptic patterns also correlate
described in more detail in the next section. Cloud with high concentrations of stratospheric ozone and
motion vectors derived from a sequence of geosta- large values of potential vorticity.
tionary satellite images can provide valuable data over Satellite images in several spectral bands (visible,IR,
remote regions for assimilation into numerical predic- and water vapor) may also show classic ‘washboard’
tion models. cloud patterns associated with mountain waves
Visible and IR images from polar orbiting satellites (Figure 7). The water vapor images often depict a
are available less frequently (every 2-6 h, depending greater area coverage of mountain wave conditions
on latitude), but can be used to identify CAT patterns than either visible or IR, owing to their greater
associated with long-lasting, large-scale systems, and sensitivity to moisture. Warm subsidence zones ori-
to corroborate features observed in geostationary ented along and just downstream from the mountain
satellite data. They are especially useful at high ridges and slightly upstream from the lee cirrus plume,
latitudes where geostationary coverage is poor be- sometimes referred to as ‘Fohn gaps’, are indicative of
cause of extreme parallax. possible intense turbulence.
Analysis of satellite imagery, compared with colo- High-resolution visible imagery (0.5-1 l m ) can
cated PIREPs, indicates that CAT is found not only in sometimes detect very small-scale (I1km wave-
clear air but also in cirrus clouds, and along borders of length) wave cloud patterns (referred to as billows)
large-scale cloud systems. Zones of turbulence asso- that correspond to areas of KHI. Billow clouds are
ciated with the subtropical jet stream are often embedded typically within or near large-scale cloud
systems, or in the vicinity of convective storms when
wind shears are present.
Table 1 Spectral channels used in detecting CAT-related phe-
nomena from space
Channel type Spectral range Applications CAT Prediction Techniques

bm)
As illustrated in Figure 2 , certain synoptic-scale upper
Visible 0.5-1 .O Small scale ( 51 km) flow patterns have been empirically related to CAT
wave clouds occurrence through many years of operational expe-
Infrared (IR) 10-12 Cloud top
temperatures
rience. By applying these pattern types to predicted
(heights), large- flow patterns from numerical forecast models, esti-
scale cloud mates can be made of the likelihood of CAT in certain
borders, cloud regions for longer forecast time periods.
bands Objective prediction of CAT has become more
Water vapor 6-7 Regions of sinking/
commonplace and accurate, owing to increased com-
rising associated
with mountain puting speed, diminished cost of high-speed compu-
waves, upper ters, improved numerical model physics, and better
level fronts, techniques for assimilation of wind and temperature
deformation data from aircraft and satellites. A number of
zones, and
diagnostic and predictive indices for CAT have
cyclones
been developed and used over the years. The non-
Figure 6 Infrared image from the geostationary operational environmental satellite GOES-8 at 0045 UTC on 3 March 2000, showing
transverse cirrus cloud bands on the Equatorward side of the subtropical jet stream near the Leeward Islands. Moderate to severe
turbulence was reported bya B767aircraft atthe locationshown. Wind barbs are from radiosondesitesfor250 hPaat 0000 UTC, or aircraft
reports for 8.9 to 11.8 km (28 000 ft to 37 000 ft) between 0000 and 0300 UTC.
dimensional Richardson number (Ri)(see eqn [l])is Significant, prolonged turbulence is most likely in
adequate in many situations, although it often exhibits regions of the atmosphere where larger-scale processes
a much wider range of critical values in the free are continually acting to decrease Ri, despite the
atmosphere than in the laboratory. Ri is most reliable influence of turbulent mixing. The primary processes
when generated from high resolution observational include: horizontal deformation, convergence, and
data, therefore it is principally used as a diagnostic differential thermal advection. Specifically, the Ri
tool. Richardson number is defined as: tendency technique calculates the large-scale forcing
necessary to overcome the kinetic energy dissipation
resulting from the turbulence process in order to
PI i.
maintain the Ri at values 5 A simple formulation of
Richardson number tendency equation is
where 8 is potential temperature (K), V the vector
horizontal wind (m s--'), g acceleration of gravity 121
) , z is height (m).
( m ~ - ~and
A more successful index for CAT prediction em- where @ is nonturbulent, large-scale forcing by defor-
ploys the Richardson number tendency equation. mation, and E is turbulent dissipation.
Figure 7 Mountain wave patterns east of the Rocky Mountainsobserved by GOES-8 in the (A) water vapor, (8) infrared, and (C) visible
band at 1915 UTC on 5 April 2000. Considerable severe turbulence was reported below 6.4 km (20 000 ft). The watervapor image depicts
more extensivecoverage of mountainwaves than the othertwo images. Cirrus plumes observed in upper left are indicativeof high-altitude
mountain waves downwind of the Salt River and Wind River ranges in Wyoming.
A similar approach can generate a diagnostic or mechanisms, and (3) most numerical models cannot
predictive index by simply calculating the product of accurately account for the intense, subgrid-scale
horizontal, resultant deformation and the vertical vertical shears and strong horizontal forcing present
vector wind shear at each grid point. The basis of this during severe CAT. More specifically, model problems
index is similar to Ri tendency in that it considers are due to resolution limitations and systematic under-
large-scale forcing from frontogenesis, in addition to forecasting of maximum wind speeds within the jet
the presence of vertical shear in the mesoscale. One stream.
difficulty with this approach is that deformation can
also lead to frontolysis in some cases, which tends to Statistical Approaches
reduce vertical shear. This turbulence index (TI) is
defined as Because of the sporadic, microscale nature of CAT, it
would seem that statistical approaches would be
useful for forecasting its occurrence. The first such
TI = [(au/ax - aV/ayl2 + (av/ax + a ~ / a y ) ~ ] l / ~
Resultant deformation efforts were completed in the United Kingdom in the
late 1970s. Turbulence data from about 4500 aircraft
x avpz 131 reports, compared with eleven colocated numerical
Vertical shear
parameters derived from a coarse-resolution predic-
where u and v are wind components (ms-I), V the tion model, revealed that the best correlation was
vector wind, and z height (m). between CAT and vertical and horizontal wind shears.
The turbulent kinetic energy (TKE) technique for Similar studies were completed in the United States in
the prediction of CAT attempts to show areas of the 1980s using higher-resolution numerical model
turbulence generation through the processes of pro- data that showed CAT to be highly correlated with
duction by vertical shear, production or destruction by horizontal deformation and scalar wind speed.
buoyancy, energy advection, and loss via dissipation. A technique that statistically integrates information
The simplified TKE equation can be expressed as: from many numerical turbulence indices, known as
the integrated turbulence forecast algorithm (ITFA),
d (TKE) -
- au - av + -
w l e -g was developed in the late 1990s. ITFA first assigns a
dt
Energy production
az
Shear production
az Buoyancy e score to each diagnostic index, based on comparison
with available observations at the initial time period
+ A - & throughout the numerical grid domain. A weighted
Advection Energy dissipation [41
sum is then determined from all of the indices to arrive
TKE = (.I2 + v’* + w r 2 ) / 2or , total kinetic energy per
at a final ITFA index value at each grid point. The same
unit mass, where the terms u t , v’, and w‘ are pertur- weights are then assigned to each index in deriving
bation (gust) velocities. The TKE approach can predictions for each forecast time period. This process
account for turbulence from a wide variety of mech- is repeated with every forecast cycle.
anisms such as convection and mountain waves, as Although useful, statistical approaches to CAT
well as jet stream CAT, and provides a direct estimate prediction are not always able to correctly depict
of possible turbulence intensity through the energy turbulence events accurately because of limitations in
dissipation rate. While TKE presents the most rigorous climatological or regionally derived thresholds.
depiction of turbulent processes of all the objective
approaches, it is more effective when numerical See also
-
models have very high vertical ( 100 m) and hori-
zontal resolution (- 10 km). In solving eqn [4] using Aviation Weather Hazards. Convective Storms:
such high-resolution models, the need for accurate Convective Initiation. Dynamic Meteorology: Waves.
Frontogenesis. Instability: Inertial Instability. Jet
observations is critical, and the best analysis and Streaks. Kelvin-Helmholtz Instability. Lee Waves
prediction results are attained in regions of dense and Mountain Waves. Satellite Remote Sensing:
reporting networks that incorporate aircraft data, Temperature Soundings; Wind, Middle Atmosphere. Tur-
chiefly over continental areas. bulence and Mixing. Turbulence, Two Dimensional.
The indices described above do a reasonably good Turbulent Diffusion.
job of predicting large outbreaks of CAT, but tend to
predict CATover a region that is larger than the area of
actual occurrence (also known as ‘over-forecasting’). Further Reading
Their inadequacies are due largely to the following Anderson RK (1995) Synoptic scale cloud and moisture
factors: (1)they only account for large scale conditions patterns; and: Clouds generated by mountains. In: Bader
favorable for CAT, (2)they do not consider triggering MJ, Forbes GS, Grant JR, Lilley RB and Waters AJ (eds),




404 CLIMATE I Overview
January surface temperature January precipitation
July surface temperature July precipitation
Figure 1 Global surface temperature mapsfor Januaryand July. Figure 3 Global precipitation maps for January and July.
Contour interval is 4 K. Data are 2-meter temperatures from the Contour interval 2cm month-'. Data are from the Global Precip-
analysis products of The European Centre for Medium-Range itation Climatology Project.
Weather Forecasts for the period 1985-94.
Hemisphere are associated with the storm tracks over much larger seasonal variation of temperature. The
the western Pacific and Atlantic Oceans. variations of precipitation in these two regions are also
Annual variations of temperature and precipitation very different. Seattle has a wintertime precipitation
vary greatly with location. In the tropics, temperature maximum associated with the midlatitude storm
variations are usually modest, but precipitation can activity. Minneapolis has a summertime rainfall max-
vary from complete drought in some seasons to imum that comes largely from thunderstorms. The
torrential rains in another. Land areas that are annual temperature variation of New York City is
downwind of large water masses generally have influenced by the large continent upwind to the west,
smaller seasonal temperature variations than loca- and so is quite large. New York gets both wintertime
tions in the center of large continents. Good examples storm precipitation and summertime convective pre-
of maritime and continental climates are given by cipitation, so that monthly precipitation is more
Seattle and Minneapolis (Figure 4): Minneapolis has a nearly evenly spread over the year. The annual total
precipitation in New York City (120cm) is greater
than that of Seattle (95cm).
July minus January sulface temperature difference
Global Energy Balance

The mean temperature at the surface of the Earth is
determined by the flow of energy through the climate
system, which consists of the atmosphere, ocean, and
land surface. The source of energy for the planet is
radiation emitted by the Sun; although the Earth has
an internal energy source from radioactive decay, this
source is too small to influence the global mean surface
Figure 2 July minus January difference in surface air tempera- temperature. At the average position of the Earth
ture. Contour interval is 5 K. in its orbit about the Sun, the Sun provides about
Figure 4 Annual cycles of climatological monthly temperature (solid line) and precipitation (dashed line) at four cities.
1367WmP2 of total solar irradiance; the solar irra- Boltzmann black body emission constant. If this
diance varies as the inverse square of the distance from equation is solved for the blackbody emission temper-
the Sun. Because Earth is approximately spherical in ature of Earth, a value of about 255K, or -18"C, is
shape, the ratio of its area to its surface shadow area is obtained, which is much less than the global mean
4. So to get the solar energy available per unit of surface temperature of 288 K, or 15°C. The emission
surface area, the total solar irradiance must be divided temperature of the Earth is equal to the average
by four, yielding about 342Wm-2, or the energy temperature of the atmosphere about 5 km above the
equivalent of about 3.4 100-watt light bulbs for each surface, and indeed, most of the energy that the Earth
square meter of the Earth. Averaged over the whole emits to space is emitted from the atmosphere, rather
Earth, about 70% of this flux is absorbed by Earth and than the surface.
about 30% is reflected back to space without heating A diagram showing the energy flow through the
the Earth. The fraction that is reflected is called the global climate system is given in Figure 5 . Although
albedo, from a Greek word meaning whiteness. nearly half of the solar energy that enters the climate
The energy that is absorbed is converted into heat system is absorbed at the surface, very little of the
and later emitted back to space as thermal infrared infrared radiation emitted from the Earth's surface
radiation. escapes directly to space. The atmosphere absorbs
The simplest model for the global mean temperature most of the infrared radiation emitted from the
T of the Earth equates the absorbed solar radiation surface, primarily through water vapor, clouds, and
with the emitted terrestrial radiation, assuming that carbon dioxide gas. Moreover, the atmosphere emits
the Earth emits like a black body (eqn [l]). infrared radiation downward toward the surface, and
S the energy supplied to the surface by this downward
$1 - a ) = aT4 infrared flux is nearly twice as great as the amount of
energy supplied to the surface from the Sun. The
In eqn [ 11S is the total solar irradiance, a is the albedo, transparency of the atmosphere to solar radiation
and o = 5 . 6 7 ~ 1 0 - ~ W m - ~ K is- ~ the Stefan- combined with the opaqueness of the atmosphere to
406 CLIMATE / Overview
Figure 5 Energy flow through the global climate system. Values are W m - *
infrared radiation results in a heating effect that raises mean precipitation rate of 1 meter per year corre-
the surface temperature above the value that it would sponds to an atmospheric heat input of 80 WmP2.
have in the absence of the atmosphere. This is often Continuous heating of the surface and cooling of the
called the atmospheric greenhouse effect. The green- atmosphere by radiative processes drives convective
house effect also reduces the amplitude of the diurnal instability and the hydrological cycle within the
cycle in surface temperature, the daily variation of Earth’s climate system. Evaporation is greater than
temperature associated with the rising and setting of precipitation over the world’s oceans. The excess
the Sun. Because the downward long-wave emission water is transported to the land areas, where the
from the atmosphere continues after the Sun sets, the average precipitation exceeds the evaporation. The
fall of temperature after sunset is much less than it excess of precipitation over evaporation in land areas
would be in the absence of the atmosphere. At high returns to the oceans as runoff in rivers (Figure 6 ) .The
altitudes the daily variation of temperature is greater supply of water from the ocean supports life on the
because the mass of atmosphere above the surface is land, and the return of minerals and other elements of
less and the greenhouse effect is therefore reduced. life to the ocean in rivers supports life in the ocean. The
Because water vapor is the principal greenhouse gas, hydrological cycle is also a key element of the chemical
the strength of the greenhouse effect increases with and biological cycling of carbon through the Earth
temperature and relative humidity. system that regulates atmospheric carbon dioxide on
A net radiative input heats the surface of the Earth, time scales of millennia and longer.
and the net effect of radiation on the atmosphere is to
cool it at the rate of about 1 3 ° C day-’. Heat is
transferred from the surface to the atmosphere by Seasonal and Latitudinal Insolation
atmospheric motions that carry heat and moisture
upward. The release of latent heat of vaporization
Variations
stored in water vapor during condensation is the The seasonal and latitudinal distribution of insolation
largest heating term in the atmosphere and offsets is an important determinant of climate. The instanta-
atmospheric cooling by radiation emission. The global neous insolation per unit of surface area is given by the
CLIMATE / Overview 407
World water balance
Ocean 70% of area
Figure 6 Global cycling of water between ocean and land. (After Hartmann (1994).)
total solar irradiance times the cosine of the solar On time scales of millennia, the distribution of
zenith angle, the angle between local vertical and the insolation with latitude and season changes as the
Sun. The daily average insolation available at the top parameters of the Earth's orbit vary in response to
of the atmosphere is given as a function of latitude and dynamical interactions with the orbits of other plan-
season in Figure 7. The insolation decreases with ets. The tilt of the axis of rotation with respect to the
increasing latitude, except in summer, and the annual plane of the Earth's orbit varies with a period of 4 1 000
variation of insolation is greatest near the poles, where years. In the last few million years the tilt angle, or
six months of darkness alternate with six months of obliquity, has varied between 22 and 24.5 degrees. It is
daylight. In polar regions during summer, the available currently 23.45 degrees. The eccentricity or degree to
insolation is greater than that at the Equator, because, which the Earth's orbit differs from a perfect circle
although the Sun is near the horizon, it shines 24 hours varies with periods of 100000 years and 400000
a day at the poles during the summer half-year. The years. The season when the Earth makes its closest
insolation available during Southern Hemisphere approach to the Sun, the perihelion of the orbit, varies
summer is about 7% greater than that available during with periods near 20 000 years.
Northern Hemisphere summer, because the Earth's
orbit is not perfectly circular and at the present time
the Earth is closer to the Sun during Southern Transport of Energy and the
Hemisphere summer. Circulation of the Atmosphere and
Ocean
Much more solar energy is available to heat tropical
latitudes than high latitudes. In the annual average, net
radiative energy is input into tropical latitudes and
high latitudes lose energy (Figure 8).The heating of the
tropics and cooling of the polar regions drives circu-
lations in the atmosphere and ocean that transport
heat from tropical to polar regions. The ocean and the
atmosphere have similarly important roles in pole-
ward transport, with the ocean transport larger in
subtropical latitudes (20" N and 20" S) and the atmos-
phere dominating at middle and high latitudes (50"N
and 50" S). Transport in the atmosphere comprises
latent, thermal, and potential energy transports. In
tropical latitudes, upward motion near the Equator
and downward motion in subtropical latitudes are
critical features of the atmospheric circulation. At low
levels this requires equatorward winds that are turned
westward by the Earth's rotation to form the
Figure 7 Daily insolation at the top of the atmosphere as a Trade Winds. At upper levels the poleward flow is
function of season and latitude. (After Hartmann (1994).) turned eastward by the Earth's rotation to form the
408 CLIMATE I Overview
in middle latitudes and have a significant influence on

the downstream climate. Because of the narrowness
and far northward extent of the Atlantic Ocean, the
Gulf Stream has a dramatic effect on the climate of
Europe.
a
- 200 The western boundary currents are generally shal-
E low surface currents. In the North Atlantic Ocean
L
8 100
warm, salty water travels very far north, where it is
c cooled without substantial dilution of the salt content.
.-m
U Because the density of sea water depends mostly on
-E o salinity near the freezing point, the cooled Gulf Stream
water becomes some of the densest water in the ocean
radiation and sinks to near the bottom. It then returns equator-
ward and may travel as far as the North Pacific before
l -200
- -90" -60" -30" 0" 30" 60" 90"
it rises to the surface again. The density-driven
circulation of the deep ocean is called the thermo-
Latitude haline circulation. The formation of deep water in the
North Atlantic Ocean creates a convergence of surface
Figure 8 Annually averaged absorbed solar radiation, emitted
water, which allows more warm water to flow in from
long-wave radiation, and net radiation at the top of the atmosphere
as functions of latitude. Data are from the Earth Radiation Budget the south and heat the high-latitude regions near
Experiment. the North Atlantic. Deep water is not formed in the
North Pacific because the Pacific is not as salty as the
Atlantic. The salinity of the oceans differs because
the atmospheric circulation carries fresh water from
the Atlantic basin to the Pacific. Major variations in
subtropical jet stream, which is a band of high winds the intensity of the thermohaline circulation in
around 30 degrees north and south at around 12 km the North Atlantic are known to have occurred during
altitude. The general rising near the Equator and ice ages. Evidence from Greenland ice cores suggests
sinking in the subtropical latitudes of the winter that the formation of deep water has switched on and
hemisphere is called the Hadley circulation, after a off on time scales of centuries or less during glacial
seventeenth-century meteorologist. The Hadley circu- periods.
lation extends only to about 30 degrees of latitude.
Beyond that the circulation becomes unstable and
breaks down into eddies. These eddies are very
Topographic Effects
efficient at transporting energy poleward, and in Air pressure and density decrease rapidly with alti-
middle latitudes the poleward transport of heat is tude. The air pressure at 5 km above sea level is only
accomplished mostly by atmospheric eddies or storms. half that at the surface. The air temperature at 5 km
The structure of these storms is such that warm, humid above the surface of the ocean is about 30°C colder
parcels of air move poleward and cold, dry parcels of than the ocean surface. When the land extends higher
air move toward the Equator. Storms thus result in a into the atmosphere, the surface pressure and temper-
net transport of heat and moisture toward the poles. ature both decrease. If the land is heated by the Sun, it
Because warm, humid parcels within these storms are will be warmer than the air at the same pressure that is
pushed upward as they are pushed poleward, midlat- not in contact with the land. This results in the
itude storms also produce heavy precipitation. generation of buoyancy of the air in contact with the
Poleward movement of warm water parcels and land, which may cause it to rise. A dramatic example
equatorward movement of cold water parcels produc- of this is the seasonal variation in winds and precip-
es poleward heat transport in the ocean. The most itation over southern Asia. When the Sun heats the
visible examples of this are the warm western bound- Himalayas during Northern Hemisphere summer, air
ary currents of the Atlantic and Pacific oceans - the at low levels is drawn toward the Himalayas, resulting
Gulf Stream and the Kuroshio Current. These are in heavy precipitation in the mountains and the
narrow streams of warm water that flow poleward at adjoining lowlands. During winter the situation is
the western margins of the oceans. The Gulf Stream reversed and air flows from the continent toward the
can carry warm, salty water all the way from the ocean. During the Northern Hemisphere winter,
tropics to the vicinity of Northern Norway. These precipitation near the Himalayas is reduced and
currents account for the relative warmth of the oceans precipitation over the equatorial Indian Ocean is
increased. This seasonal reversal in winds and associ- the ice piled on the northern continents and leads to
ated precipitation changes is called the Asian Mon- interglacial climatic conditions.
soon. A muted form of the same seasonal variation The effect of insolation variations is amplified by ice
occurs over the tropical and subtropical Americas. albedo feedback and biogeochemical feedbacks
When air approaches a topographic obstruction and affecting the concentration of carbon dioxide and
is forced to pass over it, much of the moisture in the air other greenhouse gases in the atmosphere. As the
is condensed out on the windward side of the topog- Earth cools, more of the surface of the Earth is covered
raphy, and the leeward side can be quite arid if it is with ice. Because ice has a higher albedo than other
consistently blocked from maritime sources of water surfaces, this leads to less solar absorption by the
vapor. The Cascade and Sierra Nevada Mountains, for planet and further cooling. The carbon cycle feed-
example, block the flow of moisture from the Pacific backs are less well known, but changes in the ocean
Ocean to the Great Basin between the coastal moun- circulation and changes in the supply of trace metals to
tains and the Rocky Mountains farther inland. The the ocean have been suggested as means of coupling
Himalayas prevent moisture from the Indian Ocean decreased temperature to decreased carbon dioxide.
from reaching central Asia. On time scales of thousands of years, the atmospheric
Topographic barriers also divert the flow of air by carbon dioxide concentration is tied closely to the
generating waves in the atmosphere. During winter carbon dioxide in the ocean. In the surface of the
the mountains of western North American generate a oceans, carbon dioxide is removed by photosynthetic
stationary upstream high-pressure system and down- life. Photosynthetic life in the oceans is limited by the
stream low-pressure system. This is associated with amount of nutrients and trace metals that are present
equatorward flow on the lee of the Rocky Mountains, in the illuminated zone near the surface. Ice ages are
which makes the center of North America colder than known to be dustier than interglacial epochs like the
it would be in the absence of topography. current one. During ice ages stronger winds mix the
ocean more effectively and can bring more nutrients to
the surface from depth. Stronger winds also loft more
continental dust containing trace metals and carry it to
Climate Variability and Change regions of the ocean far removed from continents
It is known that the climate of the Earth has where trace metals required for photosynthesis would
varied greatly in the past. During the Cretaceous otherwise be lacking.
Period 65-135 million years ago, the climate Climate also experiences significant year-to-year
was much warmer. Tropical plants and dinosaurs and decadal variations that result from the natural
ranged far north above the Arctic Circle. On these internal variability of the climate. The El Niiio-
long time scales continental drift has altered the Southern Oscillation (ENSO) phenomenon is a cou-
arrangement of the continents, which influences pled ocean-atmosphere mode of variability with a
the climate. It is likely that concentrations of atmos- time scale from 2 to 7 years that is centered in the
pheric greenhouse gases such as carbon dioxide or equatorial Pacific. During ENSO warm events, the
methane were much higher during the Cretaceous equatorial sea surface temperature in the east Pacific
Period. Changes in atmospheric greenhouse gas con- rises several degrees above normal and convection that
centrations equivalent to a factor of 5 or 10 increase in is normally present in the far west Pacific may extend
carbon dioxide seem to be required to explain the all the way to coastal South America. The rainfall that
similarity of tropical and polar surface temperatures normally occurs in the western equatorial Pacific and
during the Cretaceous. Indonesia follows the warm water eastward into the
Twenty thousand years ago much of North America central Pacific. The movement of the localized con-
and Europe was covered with sheets of ice several vective heating in the atmosphere drives atmospheric
kilometers thick, and a similar glacial age occurred waves that may influence weather in middle latitudes.
about 140 000 years ago. Recent variations in climate ENSO variability may be coupled to longer-term
such as the succession of ice ages over the last several variability in the North Pacific Ocean. Year-to-year
million years appear to have been triggered by changes variations in surface temperature may be caused by
in the Earth’s orbital parameters. A large tilt of the explosive volcanic eruptions that inject large amounts
Earth’s axis of rotation relative to the plane of the of sulfur-bearing gases into the stratosphere, where
Earth’s orbit about the Sun favors an interglacial fine aerosol particles of sulfuric acid can reflect
climate, because summertime and annual mean inso- sunlight.
lation in high latitudes increase with larger tilt. When Decadal variations of climate have been observed,
Northern Hemisphere summer solstice occurs when which seem to be related to interactions between
the Earth is closest to the Sun, this also acts to reduce atmosphere and ocean. Weather anomalies during
410 CLIMATE / Overview
winter in middle latitudes can drive thermal anomalies Because carbon dioxide is an important greenhouse
in the oceanic mixed layer and thermocline structure gas, its increase can lead to increases in global mean
which are then sealed under a shallow, warm oceanic surface temperature. In addition to carbon dioxide,
mixed layer during the following summer. With the the concentrations of nitrous oxide, methane, and
onset of surface cooling and turbulent mixing the some industrially created greenhouse gases are also
following winter, these thermal anomalies can be increasing in the atmosphere. The effects of these and
uncovered and influence the climate, thus giving projected future changes can be studied with global
climate anomalies a year-to-year persistence in middle climate models. The known physics of the climate
latitudes. Ocean current circulations in middle lati- system can be incorporated into these computer
tudes can also transport these anomalies from place to models. Because of the wide range of spatial and
place. temporal scales in the climate system, current limita-
tions on computer power, and some gaps in our
understanding of the climate system, these projections
remain somewhat uncertain. Key feedback processes
Human-Induced Climate Change such as the cloud formation need to be treated with
The concentrations of atmospheric greenhouse gases very approximate methods. None the less, the basic
and aerosol particles in the atmosphere have been physics of the greenhouse effect can be calculated
changing over time in response to human activities. precisely and scientists believe that human-induced
From ice bubbles trapped in ice cores and other changes in atmospheric composition have caused
evidence we know that the concentration of carbon much of the recent warming of the surface temperature
dioxide just prior to the Industrial Revolution was of the Earth. Many of the warmest years in the global
about 275 ppmv. The concentration has increased at a temperature record have occurred in the last two
rate of about 0.5% per year, so that the difference decades (Figure 10).
between the preindustrial value and the current value Changes in the Earth's climate will occur in the
is nearly as large as the differences between preindus- future as a result of human activities. Because of the
trial and ice age conditions (Figure 9 ) .Recent changes large heat capacity of the oceans, the warming
in carbon dioxide concentration are known to be resulting from greenhouse gases lags several decades
related to fossil fuel use by humans, because the behind the greenhouse gas changes. It is estimated that
isotopic composition of carbon derived from fossil about another 0.5"C of warming should result from
fuels such as coal, oil, and natural gas is different from changes in the composition of the atmosphere that
carbon that has not been stored in sediments for humans have already caused, and human modification
thousands of years. of the atmosphere is continuing at an increasing rate.
400
-
1 1 1 1 1 1 1 1 1 1 1 1 , 1 1 1 1 1
- 14.8 , I I I I I I I I I I I I
-
-
Previous
Carbon dioxide
concentration
vs time
2000
1990 +
+,
-
-
-
-
-
-
14.6 1 Global surface air temperature
from 1867-2000
1970+ -
interglacial
1950 -+
-
age -
-AJ
-
-
f 14.2
Industrial Revolution+ I
- "\
-
-
-
-
-
Glacial y Glacial -
age age -
-
150 1 1 1 ' 1 1 1 1 1 1 1 1 1 1 ' 1 1 1 13.4 i ' I I I I I I I I I
160 120 80 40 0 1860 1880 1900 1920 1940 1960 1980 2000
Year
Figure 9 Carbon dioxide for last 130000 years. Data are from Figure 10 Global mean surface air temperature as a function of
Vostok, Antarctica and Siple Station, Greenland ice cores and time, 1867-2000. Data are from Goddard Institute of Space
modern instrumental data from Mauna Loa. Hawaii. Studies.
CLIMATE PREDICTION (EMPIRICAL AND NUMERICAL) 41 1
See also Houghton JT (1994) Global Warming: The Complete

Briefing. Oxford: Lion Publishing.
Aerosols: Role in RadiativeTransfer. Antarctic Climate. Houghton JT (2001) Climate change 2001: The Scientific
Arctic Climate. Carbon Dioxide. Climate Variability:
Basis: Contribution of Working Group I to the Third
Decadal to Centennial Variability; North Atlantic and Arctic Assessment Report of the Intergovernmental Panel on
Oscillation; Seasonal to lnterannual Variability. El Niiio Climate Change. Cambridge: Cambridge University
and the Southern Oscillation: Observation. Energy
Press.
Balance Model, Surface. General Circulation: Energy
Imbrie J and Imbrie KP (1979)Ice Ages: Solving the Mystery.
Cycle. Global Change: Surface Temperature Trends.
Short Hills, NJ: Enslow Publishers.
Kump LR, Kasting JF and Crane RG (1999) The Earth
System. Upper Saddle River, NJ: Prentice Hall.
Further Reading Peixoto JP and Oort AH (1992) Physics of Climate. New
Crowley TJ and North GR (1991) Paleoclimatology. York: American Institute of Physics.
Oxford: Oxford University Press. Trenberth KE (1992)Climate System Modeling. Cambridge:
Graedel TE and Crutzen PJ (1995) Atmosphere, Climate, Cambridge University Press.
and Change. New York: Freeman. Washington WM and Parkinson CL (1986)A n Introduction
Hartmann DL (1994) Global Physical Climatology. San to Three-Dimensional Climate Modeling. Mill Valley,
Diego, CA: Academic Press. CA: University Science Books.
S Hastenrath; University of Wisconsin, Madison, WI, illustrated in Figure 1.From experience it seems more
USA fruitful to pursue empirical and numerical modeling
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. approaches concurrently. A 'code of good conduct'
requires documentation of the method, assessment of
method performance on an independent data set, and
Introduction regular verification of real-time forecasts.
Attempts at climate prediction span more than a

century and a renewed interest has developed since the
Methods
early 1980s. On theoretical grounds it has been In the endeavors at climate prediction over the past
suggested that prospects for the prediction of annual two decades, three broad categories of approaches can
variability should be better for the lower than the be recognized: (1)empirical methods based on the
higher latitudes, Method development has been combination of general circulation diagnostics and
directed at a wide variety of regional targets, as statistical techniques; ( 2 )numerical modeling; and ( 3 )
I l l
40" N
ELNINO
SOUTHERN OSCILLATION
20" s
I I I I I I I I I I
90" w 0" 90" E

Longitude
Figure 1 Orientation map showing location of forecasting targets referred to in this review: India, eastern Africa, southern Africa, Sahel,
Northeast Brazil (Nordeste), North Atlantic storms, El NiAo and Southern Oscillation, Australia, North America, Europe.
See also Houghton JT (1994) Global Warming: The Complete

Briefing. Oxford: Lion Publishing.
Aerosols: Role in RadiativeTransfer. Antarctic Climate. Houghton JT (2001) Climate change 2001: The Scientific
Arctic Climate. Carbon Dioxide. Climate Variability:
Basis: Contribution of Working Group I to the Third
Decadal to Centennial Variability; North Atlantic and Arctic Assessment Report of the Intergovernmental Panel on
Oscillation; Seasonal to lnterannual Variability. El Niiio Climate Change. Cambridge: Cambridge University
and the Southern Oscillation: Observation. Energy
Press.
Balance Model, Surface. General Circulation: Energy
Imbrie J and Imbrie KP (1979)Ice Ages: Solving the Mystery.
Cycle. Global Change: Surface Temperature Trends.
Short Hills, NJ: Enslow Publishers.
Kump LR, Kasting JF and Crane RG (1999) The Earth
System. Upper Saddle River, NJ: Prentice Hall.
Further Reading Peixoto JP and Oort AH (1992) Physics of Climate. New
Crowley TJ and North GR (1991) Paleoclimatology. York: American Institute of Physics.
Oxford: Oxford University Press. Trenberth KE (1992)Climate System Modeling. Cambridge:
Graedel TE and Crutzen PJ (1995) Atmosphere, Climate, Cambridge University Press.
and Change. New York: Freeman. Washington WM and Parkinson CL (1986)A n Introduction
Hartmann DL (1994) Global Physical Climatology. San to Three-Dimensional Climate Modeling. Mill Valley,
Diego, CA: Academic Press. CA: University Science Books.
S Hastenrath; University of Wisconsin, Madison, WI, illustrated in Figure 1.From experience it seems more
USA fruitful to pursue empirical and numerical modeling
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. approaches concurrently. A 'code of good conduct'
requires documentation of the method, assessment of
method performance on an independent data set, and
Introduction regular verification of real-time forecasts.
Attempts at climate prediction span more than a

century and a renewed interest has developed since the
Methods
early 1980s. On theoretical grounds it has been In the endeavors at climate prediction over the past
suggested that prospects for the prediction of annual two decades, three broad categories of approaches can
variability should be better for the lower than the be recognized: (1)empirical methods based on the
higher latitudes, Method development has been combination of general circulation diagnostics and
directed at a wide variety of regional targets, as statistical techniques; ( 2 )numerical modeling; and ( 3 )
I l l
40" N
ELNINO
SOUTHERN OSCILLATION
20" s
I I I I I I I I I I
90" w 0" 90" E

Longitude
Figure 1 Orientation map showing location of forecasting targets referred to in this review: India, eastern Africa, southern Africa, Sahel,
Northeast Brazil (Nordeste), North Atlantic storms, El NiAo and Southern Oscillation, Australia, North America, Europe.
41 2 CLIMATE PREDICTION (EMPIRICAL AND NUMERICAL)
Figure 2 Scheme of climate prediction based on general circulation diagnostics and statistics. (Adapted from Hastenrath, 1995a.)
empirical but purely statistical techniques, although ical models for the operational prediction of El Niiio.
there are combinations of and transitions between Similarly, the application of general circulation mod-
the categories. Categories 1 and 2 imply a diagnostic els (GCM) to seasonal rainfall forecasting for the
understanding of circulation mechanisms. In the West African Sahel and Brazil’s Nordeste have a broad
empirically based approaches 1 and 3 , a clear distinc- empirical-diagnostic basis. In approach 2 the
tion must be made between the dependent portion prediction model is developed from basic circulation
of the record (or training period) from which the theory rather than from past observations as in
method was developed and the independent data approach 1. Accordingly, the separation into depend-
set reserved for the verification of forecast per- ent and independent portions of the record (as
formance (verification period). An independent for approaches 1 and 2 ) does not arise. Values
verification period is essential to protect against calculated from previously observed conditions
noise-fitting, a severe risk especially where a large are again compared with the observed predictand
number of predictors is used after excessive screening. values for verification. In principle, there are
In category 2, verification of performance against prospects of coupling the atmosphere, ocean,
observations over a couple of decades is also in and land surface portions of the system. Purely
order. statistical techniques (approach 3 ) may seem unsatis-
Approach 1 - general circulation and statistics - is factory because of the lack of insight into the circu-
illustrated in Figure 2. Studies of general circulation lation mechanisms involved. Some of these endeavors
diagnostics are aimed at understanding the mecha- merit attention, however, because of their remarkable
nisms of climate anomalies. On this basis, indices practical performance.
are selected that appear promising as predictors.
Effective combinations of predictors and formulations
of quantitative predictor-predictand relationships Regional Targets
are ascertained by appropriate statistical methods, Methods for the forecasting of interannual variability
such as stepwise multiple regression (SMR), 1’inear have been developed for numerous regions, especially
discriminant analysis (LDA), and neural networks. in the low latitudes. A series of examples are addressed
Other statistical procedures widely used include here, namely, Indian monsoon, eastern and southern
canonical correlation analysis (CCA), singular Africa, Sahel, Northeast Brazil (Nordeste), North
spectrum analysis (SSA), and empirical orthogonal Atlantic storms, El NiAo and Southern Oscillation,
function (EOF) analysis. The quantitative predictor- Australia, North America and Europe, as identified
predictand relationship represents the prediction in Figure 1.
model. This then serves for calculating predictand
values for a portion of the record that had not
Indian Monsoon
been used in the development of the prediction
model. Quantitative comparison of the calculated Empirical-diagnostic research over more than a
versus the observed values of the predictand provides century has led to a considerable understanding of
verification of forecast performance. The evolu- the general circulation mechanisms of Indian mon-
tion from general circulation diagnostics to prognosis soon rainfall anomalies and to the identification of
is exemplified by a series of papers on Brazil’s numerous viable predictors. These can be loosely
Nordeste. grouped into three classes, pertaining to the upper-
An understanding of the general circulation mech- air flow over India, to heat low development
anisms of climate anomalies from empirical-diagnos- over southern Asia and establishment of meridional
tic studies is also basic to approach numerical pressure gradient and cross-equatorial flow over
modeling. This is exemplified for the Pacific El Nifio the Indian Ocean, and to the Southern Oscillation
phenomenon in the progression from the pioneering (SO, high phase defined by anomalously highllow
empirical oceanographic work over the first diagnostic pressure at Tahiti/Darwin), although there are
ocean numerical studies to the application of numer- interrelations between the three classes. Figure 3
the latter is strongly (and inversely) related to the

loo0Model
] 1 SO. On purely statistical grounds, it has been shown
that a portion of the variance of boreal autumn rains is
predictable from the Tahiti minus Darwin pressure
index alone. The general circulation causes of eastern
L, DPT
E
h
I
600 African rainfall anomalies are related to the
v
- 1940 50 60 70 80 90 SO through a combination of cooperative mecha-
.- nisms, which most effectively function in the boreal
Model 2
d 1000 autumn rainy season of eastern Africa. (a) Westerly
winds along the Indian Ocean Equator are conducive
to lower-tropospheric divergence over equatorial
East Africa; in the high SO phase these are accele-
I d
600 rated, especially in October-November owing
to the anomalous eastward pressure gradient. (b)
Figure 3 Prediction of all-India summer monsoon rainfall from The Equatorial westerlies drive the Eastward Equato-
two models, using as predictors the latitude position of a 500 hPa
ridge along 75"E in April (L); April minus January pressure
rial Jet in the upper hydrosphere, which entails cold
tendency at Darwin (DPT); and an index of January-February water upwelling in the western extremity of the
Northern Hemisphere temperature (NHT). Solid dots basin, where sea surface temperature further hydro-
denote regressed values for years up to 1968 and forecast values statically affects the zonal pressure gradient and
from 1969 onward, as separated by the vertical broken line. Open thus feeds back into the equatorial westerly winds.
circles indicate observed rainfall. (Adapted from Hastenrath,
1995a.)
(c) In addition, cold water anomalies in the western
Indian Ocean, most pronounced in October-
November during the high SO phase, also suppress
convection. (d) In the high SO phase, the Indian
illustrates two simple prediction models based on summer monsoon tends to be strong, leaving behind
variously used predictors and stepwise multiple an anomalously cold western Indian Ocean,
regression. Conspicuously absent in the prediction of which in turn feeds into the mechanisms (a) to
the Indian monsoon rainfall anomalies is the use (c). These processes are strong and definite in
of numerical models in the operational work of October-November, when the equatorial westerlies
the India Meteorological Department or in the pub- correlate at -0.85 with the rains at the East
lished literature. This is all the more remarkable African coast, but reveal themselves little in
because the sustained empirical work has afforded effective precursors. Relationships for April-May
a sound understanding of the functioning of the are weak.
general circulation of the monsoon and has shown
Indian monsoon rainfall anomalies to be amenable to Southern Africa
prediction. Along with these overall encouraging
Various groups endeavor to forecast the austral
prospects for Indian monsoon forecasting, the
summer rains of Southern Africa from purely statis-
reservation should be noted that the performance
tical methods and more recently also using numerical
of various predictors has been known to vary in
modeling. Rains tend to be more abundant during the
the long term. Predictability was found to be high
high/cold SO phase. A better diagnostic understanding
for the 1960s and 1970s, to be followed by drastically
of the circulation mechanisms of rainfall anomalies is
reduced performance in the 1980s, as is apparent
needed.
in Figure 3. Such changes in predictability should
be seen in the context of decadal-scale evolutions in
Sahel
the general circulation setting, but the direct
causalities await clarification. Numerical model The climate problems in the sub-Saharan zone of West
applications to the prediction of Indian monsoon Africa are particularly complex because interannual
rainfall anomalies will have to face the decadal- variability is superimposed on drastic decadal-scale
scale changes in intrinsic predictability as an added changes of rainfall. These share some common mech-
challenge. anisms involving variations of the meridional sea
surface temperature (SST) gradient in the tropical
Atlantic and latitudinal displacements of the Inter-
Eastern Africa
tropical Convergence Zone (ITCZ), as well as SST
Eastern Africa features two rainy seasons centered changes in the western Indian Ocean, but the reasons
around April-May and October-November, and only for the persistence of the present drought conditions
are not yet understood. Thus, it seems essential structed from a network of quality-controlled rain
to separate year-to-year variations from trendlike gauge stations with continuous record. Effective
developments, a challenge both in terms of general predictions can be made from observations through
circulation mechanisms and the proper statistical January. The predictors are the preseason rainfall
treatment. in the Nordeste itself, indices of the fields of
Real-time forecasts are published regularly. the meridional wind component and of SST in the
The Hadley Centre of the UK Meteorological Office tropical Atlantic, and, less importantly, an index
contributes forecasts from both empirical and numer- of equatorial Pacific SST. This information serves as
ical modeling approaches. The empirical method input to stepwise multiple regression, linear discrimi-
entails empirical orthogonal functions of SST, nant analysis, and neural networking. Figure 4 illus-
where the interhemispheric SST gradient in the trates the performance of two models using stepwise
tropical Atlantic is most important. Correlation multiple regression and of one based on neural
between forecast and observed rainfall is around networks. An approach by the UK Meteorological
0.6, mostly related to persistence. The numerical Office uses as input information SST in the tropical
modeling has global SST as input and correlations Atlantic and equatorial Pacific. The diagnostic re-
of predicted versus observed rainfall are between search at the University of Wisconsin has also served as
0.3 and 0.6. The NOAA-Climate Prediction Center in basis for numerical modeling experiments by other
the United States reports results from an empirical groups.
method using canonical correlation analysis on global
SST and gridded rainfall data for Africa. Most
important is the interhemispheric difference in
SST anomaly, particularly in the Atlantic, and ENS0
is not the dominating aspect of the relationship; R32 I
I
correlation between predicted and observed +2- P
rainfall is around 0.3. Thus, there are indications
for a moderate predictability of Sahel rainfall
anomalies. Given the peculiar evolution of Sahel -2 I
- T31-NDJ, PAC i
climate, closer attention to the contributions from
year to year compared to trendlike developments
seems desirable.
North-east Brazil
Northern Northeast Brazil has its rainy season nar- a
rowly concentrated around March-April, when
the near-equatorial trough reaches its southernmost
position in the course of the annual cycle. Rainfall
I
anomalies have a severe human impact, their general +2-
circulation mechanisms are definite and well under-
stood, and they offer themselves as a prime target
of opportunity for climate prediction. During drought
years, the southward SST gradient in the tropical
Atlantic is enhanced, the near-equatorial low-pressure 1920 30 40 50 60 70 80 1990
trough and embedded wind confluence and ITCZ Year
are displaced northward, and the North Atlantic
Trade Winds are reduced while the cross-equatorial Figure 4 Prediction of an index of March-June rainfall in
northern Northeast Brazil from three models, using as predictors
flow from the Southern Hemisphere is accelerated.
October-January Nordeste rainfall (ONDJ); index of January
Warm water anomalies in the equatorial Pacific also meridional wind component over the tropical Atlantic (V); index of
tend to be associated with dry conditions in Brazil's January SST field in the tropical Atlantic (ST29); index of
Nordeste. November-January SST field in the tropical Atlantic (T31-NDJ);
General circulation diagnostics combined with and SSTanomaly in theequatorial Pacific (PAC). Solid dots denote
statistical techniques form the basis for effective regressed (trained for model N) values for years up to 1957 and
forecast values from 1958 onward, as separated by the vertical
prediction methods. In the work at the University broken line. Open circles indicate observed MAMJ values in all
of Wisconsin, the predictand is an index of the panels. Models 25 and 32 use stepwise multiple regression, and
March-June rainfall in the northern Nordeste, con- model N neural networking.
North Atlantic Storms surface waters to the north of Australia and in

the Indonesian seas tend to be anomalously warm,
Since the mid 1 9 8 0 ~seasonal
~ forecasts have been
with low pressure and relatively abundant rainfall.
issued for North Atlantic hurricane activity with
In accordance with the prevailing SST anomalies,
remarkable success. Input information includes
tropical storm activity also tends to be enhanced.
stratospheric zonal wind component, El Niiio devel-
Such processes in the large-scale circulation setting
opment in the Pacific, pressure over the Caribbean,
are being used to advantage for real-time fore-
and rainfall over West Africa.
casting of rainfall, tropical cyclone activity, and crop
El NiAo and Southern Oscillation (ENSO) yields.
Following the considerable understanding of the under- North America
lying atmospheric-hydrospheric mechanism offered
by extensive empirical-diagnostic investigations El Seasonal forecasts of the spatial patterns of tempera-
Nifio has been an early target of efforts in extended- ture and precipitation anomalies over the United
range forecasting. It was the prediction of El Niiio to States are regularly issued by the Climate Prediction
which numerical modeling was first applied. Methods Centre of NOAA, based on statistical methods and
developed from a numerical ocean model driven by numerical modeling. The forecast system consists of
recent observed surface winds, over a coupled ocean- tools that forecast the tropical Pacific SST and tools
atmosphere model using the observed history of the that forecast the US surface temperature and precip-
Pacific wind field during the preceding years, to a two- itation. Forecasts of tropical Pacific SST are produced
tiered approach in which a model for predicting from canonical correlation analysis of patterns of
tropical Pacific SST is used in tandem with an global sea level pressure and tropical Pacific SST, and
atmospheric GCM. It may be anticipated that even from coupled ocean-atmosphere models run with full
though the upper-air flow patterns seem realistically coupling in the tropical Pacific region. These statistical
depicted, the prediction of regional rainfall anomalies and numerical modeling results are then combined
remains a challenge. into a forecast of tropical Pacific SST. This along with
Purely statistical techniques have also proven effec- the SST values observed elsewhere in the global ocean
tive for the prediction of El Nifio, including canonical serve as boundary condition to force an atmospheric
correlation analysis of SST and wind fields, SSA of GCM. Note that this two-tiered numerical modeling
the Tahiti minus Darwin index, and linear inverse system is in principle applicable anywhere over the
modeling of SST. globe, depending on any skill which could be derived
To the extent that warm water anomalies in from it, but it is used operationally only for US
extended areas of the equatorial Pacific feed back seasonal forecasts. In addition to this two-tiered
into the atmospheric circulation, the prediction of El numerical modeling system, statistical techniques are
Nifio is of interest beyond the Pacific domain proper. It also applied, namely, optimal climate normals (OCN)
must be noted, however, that the SO contributes for the region and canonical correlation analysis of the
substantially to the interannual climate variability global SST, northern hemispheric 700 hPa patterns,
only in limited regions of the tropics. With a view to and US temperature and precipitation. A statistical
implications for the global atmospheric circulation, at consolidation of the diverse tools is made to aid the
the Climate Prediction Center of NOAA observed forecasters in objectively combining the information.
surface and subsurface thermal conditions are input to In the course of the annual cycle, skill is highest at the
an ocean model and an anomaly coupling scheme is end of winter and summer, and poorest in between. In
used for interaction with the atmospheric model. The this context, lead times are less relevant, so that
Pacific El Nifio phenomenon is clearly related to the forecasts are issued out to a year; at certain seasons,
SO and this is commonly described by the Tahiti minus skill is low even at the shortest lead times; at other
Darwin pressure index. Signals of the SO are perva- seasons, whatever little is known is known well in
sive, but only in limited domains of the tropics does advance. In a similar vein, experiences for India and
this contribute substantially to the regional climate north-east Brazil have shown predictability a couple of
variability. months ahead of the season, with no improvement a t
shorter lead times.
Australia
Europe
Lying as it does near a dipole of the SO, it is not
surprising that the greater Australian region has its Attempts at seasonal forecasting of temperature and
climatic variability strongly tied into the phases of this precipitation by the UK Meteorological Office con-
large-scale pressure seesaw. In the high SO phase, tinue, based on statistical methods and numerical
modeling. Wintertime SST anomalies in the North

Atlantic serve as input to linear regression to predict
summer temperature for central England. An exten- AFRICA
sion of this work is devoted to the prediction of ClRC
summer temperature over Europe. Associations with
ENSO are found to be weak.
A major collaborative European effort involving 11
institutions in six countries, PROVOST, aims at
exploring the potential of numerical modeling for
seasonal forecasting for the continent. Tentative
so
+0.3
+n SAHEL
+0.3 +0.5
results have been reported from the contribution of ClRC
the UK Meteorological Office. ENSO has not fur-
nished skill for the prediction of the winter or spring
conditions. Interest also focuses on the possibility of
predicting when a forecast will be skillful. Work
continues. so
+0.3
+0.3
-0.7
)In
~
NORDESTE
ClRC P
Caveats Figure 5 Schemes of long-term mean diagnostic relationships,

showing coefficients of concurrent correlation between indices of
It is tempting to believe that the tropical climate the Southern Oscillation (SO), the regional circulation (CIRC), and
prediction problem can be reduced to ENSO and that rainfall at the coast of eastern Africa (October-November), in the
numerical modeling is the obviously superior tool in West African Sahel (July-August), and in Brazil’s Nordeste
the long run. These widely held beliefs are briefly (March-April).
addressed in the following.
Regarding ENSO, it should be realized that inter-
annual variability in the regional climate is directly
due to anomalous behavior of the quasi-permanent
circulation systems in the region; to the extent that the casts. Five years later they confirmed this assessment
SO is associated with variability in the regional and speculated about the possibility that numerical
circulation, it may show a correlation with regional models will never be able to materially outperform
climatic conditions. Figure 5 offers some illustrations empirical models.
pertaining merely to diagnostics rather than progno- In this context Nicholls’ cautioning in 1999 against
sis. Thus, the October-November rains at the coast of cognitive illusions in climate prediction is pertinent.
equatorial East Africa are correlated at better than 0.8 With reference to a recent experience he noted:
with the concurrent surface westerlies over the equa-
torial Indian Ocean, as compared to correlations of The general view among scientists of the
only about 0.6 between the SO and the rains or winds. accuracy of forecasts of the El Nifio of 1997
In a similar vein, the summer rainfall in the West appears to illustrate hindsight bias. None
African Sahel has correlations of around 0.6 with of the climate forecast systems ... predicted
indices of the circulation but only of 0.3 with the SO. anything more than slight warming ... . Yet
The rains of Brazil’s Nordeste are correlated at about the forecasting of the 1997 El Nifio with
0.7 with the regional circulation, whereas the SO has large models is now regarded as a stunning
correlations of only about 0.3 with either the rain or success .., . Some of the model predictions for
the circulation. the 1997/98 event were very poor. It seems
The development of both empirical and numerical these misfortunes are being ignored when the
modeling methods for operational climate prediction stunning success of the El Nifio model forecasts
has been pioneered by two large governmental is assessed .... People seek confirmatory evi-
institutions: the Hadley Centre of the UK Meteoro- dence and avoid the search for disconfirming
logical Office and the Climate Prediction Center evidence.
of NOAA. A group from the latter institution
has critically examined the potential of the two
Outlook
approaches in review papers half a decade apart.
In the earlier review they found that the two There has been encouraging progress in the seasonal
approaches delivered roughly equally skillful fore- prediction of interannual climate variability. While
CLIMATE PREDICTION (EMPIRICALAND NUMERICAL) 417
forecasting methods are being developed for ever logy: Overview and Theory. Walker Circula-
more target regions, the impression remains tion. Weather Prediction: Seasonal and lnterannual
that climate anomalies may be highly predictable Weather Prediction. Wind Chill. World Climate
only for limited areas of the tropics where the bulk Research Program.
of the rainfall is prevailingly derived from a single well-
organized quasi-permanent circulation system (such Further Reading
as the ITCZ), especially at the extremes of
the planetary-scale annual cycle. Fortunately, how- Anderson J, Van den Do01 H, Barnston A, et al. (1999)
ever, these regions tend naturally to coincide Present-day capabilities of numerical and statistical
with areas that are most prone to climatic anomalies models for atmospheric extratropical seasonal simula-
tion and prediction. Bulletin of the American Meteoro-
and where the human impact is most severe. For
logical Society 80: 1349-1361.
other vast domains the prospects of seasonal predic- Barnston AG, Van den Do01 H, Zebiak SE, et al. (1994)
tion may be more remote. The extent t o which Long-lead seasonal forecasts - where do we stand?
moderate intrinsic predictability may be practically Bulletin of the American Meteorological Society 75:
useful in agricultural and economic planning is an 2097-2114.
issue that remains to be explored in continuing Barnston AG, Leetmaa A, Kousky VE, et al. (1999)NCEP
discourse between the forecasting community and forecasts of the El Nifio of 1997-98 and its US impacts.
the potential user community. In this context, climate Bulletin of the American Meteorological Society 80:
monitoring should be cultivated in tandem with 1829-1 852.
climate prediction. Carson DJ (1998)Seasonal forecasting. Quarterly Journal of
It seems desirable to pursue in parallel the general the Royal Meteorological Society 124: 1-26.
COLA (1998-99) C O L A Experimental Long-Lead Forecast
circulation-based empirical approach and the numer-
Bulletin, vols 7-8.
ical modeling, because this combination should be Hastenrath S (1985)Climate and Circulation of the Tropics.
mutually fruitful and offer insight into the atmos- Dordrecht: Reidel.
phere-ocean mechanisms involved. It should be real- Hastenrath S (1986) On climate prediction in the tropics.
ized that, although the Southern Oscillation signal is Bulletin of the American Meteorological Society 67:
pervasive, climate anomalies in many regions are not 692-702.
prevailingly related t o the SO. Decadal-scale changes Hastenrath S (1990)Tropical climate prediction: a progress
in predictability are poorly understood, an impedi- report 1985-90. Bulletin of the American Meteorological
ment to prediction regardless of technique. A broad- Society 71: 819-825.
based effort is needed, combining empirical and Hastenrath S (1995a) Recent advances in tropical climate
modeling approaches and coupling diagnostics with prediction. Journal of Climate 8 : 1519-1532.
Hastenrath S (1995b) Climate Dynamics of the Tropics.
prognosis. Documentation of the method, verifica-
Dordrecht: Kluwer.
tion of performance on an independent data set, Latif M, Anderson D, Barnett T, et al. (1998)A review of the
and evaluation of real-time forecasts are always predictability and prediction of ENSO. JGR-Oceans
imperative. 14375-14393.
Palmer TN and Anderson DLT (1994) The prospect
for seasonal forecasting - a review paper. Quarterly
See also Journal of the Royal Meteorological Society 120:
Climate: Overview. Climate Variability: Seasonal 755-793.
to lnterannual Variability. El Niiio and the Southern Nicholls N (1999) Cognitive illusions, heuristics, and
Oscillation: Theory. General Circulation: Over- climate prediction. Bulletin of the American Meteoro-
view. Hurricanes. Monsoon: Prediction. Numerical logical Society 80: 1365-1397.
Models: Methods. Ocean Circulation: General Proc- NOAA-CPC (1992-97) N O A A - C P C Experimental Long-
esses. Predictability and Chaos. Tropical Meteoro- Lead Forecasting Bulletin, vols 1-6.
41 8 CLIMATE VARIABILITY / Decadal to CentennialVariability
Contents
Decadal to Centennial Variability

Glacial, Interglacial Variations
Nonlinear Aspects
North Atlantic and Arctic Oscillation
Seasonal to Interannual Variability
over these time scales is referred to as ‘dec-cen’

Decadal to Centennial climate variability.
Variability From a practical standpoint, dec-cen climate vari-
ability involves climate change that occurs once in a
while over the course of a human lifetime (e.g., the
D G Martinson, Columbia University, Palisades, ‘real’ winters your grandfather remembers), or in
NY, USA general it occurs slowly relative to our year-to-year
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. experience of climate. This is distinctly different from
shorter-term change that occurs rapidly, and often
apparently dramatically within a year or several years
Introduction (see Climate Variability: Seasonal to Interannual
Variability). The difference extends well beyond our
In the final decade of the twentieth century, recovery of perception of the change. It also has considerable
high-resolution paleoclimate records of natural cli- implications regarding how the change or variations
mate variability (see Paleoclimatology: Ice Cores; influence society and how we study them.
Varves) improved dramatically our perception of the
long-term behavior of the Earth’s climate. Immediate-
Dec-Cen Variability and Society
ly apparent from these records was the fact that the
once implicit notion that the modern-day climate From a societal perspective, the time scale over which
system (following termination of the last ice age (see climate variability manifests itself is important from
Ice Ages (Milankovitch Theory)) was relatively stable, several perspectives. Short-term variability, such as
was no longer tenable. This notion quickly yielded to that related to El Niiio events (see El Niiio and the
one in which the Earth’s climate is continually chang- Southern Oscillation: Observation; Theory), influ-
ing over all time scales. Presumably it will continue to ences everything from agriculture to recreation. But if
change, with or without human-induced (anthropo- such short-term change persists for dec-cen time
genic) influences. scales, the implications are considerably different.
Climate, as measured by the averaged value of any For example, in the 1990s the Midwestern United
characteristic of weather, such as temperature or States was twice hit with devastating floods (1993 and
precipitation, can show considerable differences in 1997).As rare and extreme short-term climate events,
how it varies through time and across the globe. For society must deal with their impact through disaster
example, it might undergo a smooth or abrupt relief and flexible adaptive measures. But if they are a
transition from one quasi-stable state to another; it consequence of changed climate conditions during the
may vary cyclically, not unlike the familiar daily or latter part of the twentieth century and represent a
annual cycles, but with cycles lasting tens or hundreds regular characteristic of the new climate state, then
of years or longer; and it may vary through changes in such flooding must be dealt with through policy
extreme states (e.g., colder winters), or in the magni- decisions and investments in infrastructure.
tude or degree of fluctuations (e.g., more storms, or The latter requires considerable foresight and the
larger differences from one year to the next). When best possible information. This is particularly impor-
such variations occur over, or persist for, tens to tant for dec-cen variability since the magnitude of
hundreds of years, we classify them as decadal-to- climate change is often proportional to the length of
centennial variability. For convenience, variability time over which it operates. Changes on dec-cen time
CLIMATE VARIABILITY / Decadal to Centennial Variability 41 9
scales can involve potential shifts in agriculture belts, and dynamically consistent, comprehensive ‘reanaly-
in droughdflooding frequency, magnitude and extent, sis’ data set through the use of models. The paleocli-
significant rises in sea level, fundamental adjustments mate records, our greatest hope, are still limited
in energy usage, etc. It can also modify the larger in their spatial distribution, and sometimes accuracy
background climate state that influences our ability to and precision, though they are improving rapidly
predict shorter-term climate events, such as El Nifio. (see Paleoclimatology: Ice Cores; Varves).
An ability to recognize or anticipate such change in Contrast this to the remarkably successful paradigm
order to minimize the negative impacts and optimize used to advance short-term climate prediction. In that
the positive ones often requires advance action and case, numerical models or statistical methods for
decisions. Consequently, an ability to forecast such prediction can be immediately calibrated against the
change, recognize its signs, or understand its potential past few decades of observations. They can then be
is a fundamental goal of modern climate studies. tested against an upcoming year, and soon thereafter
Unfortunately, the potentially huge impacts (posi- modified according to their success or failure. This
tive and negative) of dec-cen variations are typically, allows a very effective and rapid advancement in
though not always, realized only slowly with time. short-term prediction.
Such slow, often imperceptible, change disguises the For prediction of dec-cen climate variability, this
potential long-term implications of the change, while paradigm is clearly impractical except for cases where
tending to undermine the immediacy of the problem the change is realized at a fast enough rate to allow
and our resolve to address it. So too does the some testing in the near distant future (this is the
knowledge that any experienced change may be part paradigm being used to predict greenhouse warming;
of a longer cycle that will be returning to its previous see Global Change: Biospheric Impacts and Feed-
(acceptable) state in due time. Regardless, climate will backs; Human Impact of Climate Change; Ozone
change and infrastructure/policy decisions will ulti- Trends; Surface Temperature Trends; Upper Atmos-
mately be made either in response to the change or in pheric Change). Consequently, we are forced into a
anticipation based on imperfect information. Thus, different paradigm, dependent upon the collection of
understanding dec-cen climate change in order to comprehensive and widespread paleoclimate data sets
provide sound information regarding the potential or and the new model reanalysis products. We are equally
likelihood of change is fundamental to our long-term, dependent upon sustained acquisition of observations
social and individual, well-being. that will eventually provide that comprehensive data
set necessary for quick hindcast calibration and
evaluation of models for future generations.
Furthermore, because dec-cen variability involves
The Study of Dec-Cen Variability such long time scales, its study is faced with a number
The study of dec-cen climate variability is new, and of additional difficulties: (1)Even the relatively high-
is at a distinct disadvantage relative to that of shorter- order physics, those that describe and control proc-
term climate variability. In fact, the differences esses that do not impose any significant influence over
are significant enough that the paradigm by which short time scales, have enough time to introduce
climate studies, particularly prediction, have hereto- systematic biases or feedbacks into the system which
fore relied on must be changed to study dec-cen must be accounted for over long time scales. In models,
climate variability. Because dec-cen variability in- it is typically necessary that all of the physics of a given
volves slow change, long data records are required order (i.e., relative level of importance) be included
for model calibrations, predictive skills (the ultimate since they often work to mitigate or enhance the
test of understanding), and observational analyses. influence of one another. ( 2 )With longer time scales,
That is, records must contain enough occurrences changes can be communicated over farther spatial
(realizations) of the phenomena being examined to distances and have time to interact with different
give a reasonable statistical foundation for analysis or components of the climate system. For example,
model-data comparison. For dec-cen variability, only predictions of El Nifio have been quite successful
the bare minimum of such data sets currently exist. without taking into consideration the current state of
It will take decades into the future to obtain long the polar sea ice fields, the level of atmospheric CO2,
and comprehensive enough modern records for or the state of vegetation on land. However, if one
dec-cen studies. The historical record from modern wishes to predict how the frequency, intensity, or other
instruments does not extend far enough back in time characteristics of El Nifios may vary over decade-to-
and is typically too sparse in most locations. Invalu- century time scales, it is possible that any and all of
able advances are being made in our ability to these may impart some influence that ultimately
interpolate these historical data into an internally impacts the tropical Pacific and its evolution.
420 CLIMATE VARIABILITY / Decadal to Centennial Variability
The implications of these are tremendous. Numer- granted that the expression climate ‘change’ repre-
ical models examining dec-cen climate variability sents variations in climate due to anthropogenic
must therefore include some treatment of detailed reasons, where climate ‘variability’ refers to natural
physical processes not required in shorter-term climate variability. In this article, both change and variability
models. These detailed processes typically operate on are used interchangeably, with the differentiation
the smallest spatial scales, and may have their largest between natural and anthropogenic change explicitly
influence in regions quite remote from the region of stated when required.
interest (for example, some aspects of the ocean
circulation originate in the subpolar regions as a result
of small-scale local interactions, and from there
transport heat and salt to the remainder of the world’s
Modes of Dec-Cen Variability
oceans). This requires that the models either break Climate variability on decade-to-century time scales
the Earth into a great many very small grid cells has manifested itself in historic times through a
(see Numerical Models: Methods), or include param- number of fairly well-known climate events (see the
eterizations of how larger areas may respond in Further Reading section). Some of the more notable
average to the small (‘subgrid’) scale processes that ones include the prolonged drought of the Great Plains
cannot be explicitly resolved. The former allows for of the United States responsible for the dust bowl of the
better treatment of the physics but at tremendous 1930s, and the crippling drought of the Sahel in
computational burden (making a computer simula- northern Africa during the latter decades of the
tion of the model extremely slow and expensive); the twentieth century, which killed over half a million
latter relieves some of the computational burden but at people in the mid-1970s. The global warming of the
the cost of possibly overlooking some of the additional twentieth century or the enhanced warming started in
details that may be important. In addition to these the mid-1970s. Changes in the Earth’s ozone layer
detailed processes, the models must contain other have led to increases in the level of ultraviolet radiation
parts of the climate system not required in the short- at the Earth’s surface in high southern and, more
term climate models, such as the biosphere, cryo- recently, northern latitudes. Dec-cen changes in cloud
sphere, and atmospheric composition, or more aspects cover also have led to increased surface radiation in
of the ocean, such as the very slow deep ocean Australia, North America, India, and Europe through-
circulation. Not only do the additional physics, global out the twentieth century. The number of major
scale, and finer model resolution add computational hurricanes varies on dec-cen time scales, as do the
burden, but the very nature of the problem being number of Nor’ Easters ravaging the north-east coast
decade-to-century time scale variability requires that of North America. Sea level has been rising throughout
such burdened models run for inordinately long the last century (-20 cm) and changes in ecosystems
simulation periods as well, Consequently, model directly related to dec-cen climate have been dramatic
simulations are slow, many of the physics missing or in fisheries: the North Atlantic cod and eastern North
uncertain, and the models are so complex that Pacific salmon. Paleoclimate evidence suggests that
relatively few exist. This limits our ability to assess dec-cen climate variability is also responsible for
dec-cen climate variability by multiple experiments the fall of civilizations (e.g., the Classic Maya) and
under a variety of conditions with many different mass migrations of societies (e.g., the Nordestinos of
models involving different parameterizations (though Brazil).
this is still done to the extent possible, but it is greatly While the dramatic nature of these events makes
hindered by these extreme demands). them rather conspicuous examples of dec-cen varia-
Finally, the dec-cen problem is also faced with bility, much of our focus is directed toward a broader
making predictions based on changes in greenhouse view of dec-cen variability. Specifically, observations
gases (the composition of the atmosphere) that are a made during the last several decades suggest that
function of highly uncertain future emission scenarios. climate variability over large expanses of the Earth
Therefore, even if we succeed in understanding natural seems to organize itself into patterns that preserve a
dec-cen climate variability, future predictions must be general shape in space, but whose amplitude may
compromised by including uncertain estimates of how change in time. Such coherent structure is referred to
the atmospheric composition will change. Here we are as a mode of variability or, to admit a slightly broader
forced to depend on multiple forecasts using a variety class of structures, they are more generally referred to
of different emission scenarios, but again the large as climate patterns or spatial-temporal patterns.
computational burden of such comprehensive models Our study of such patterns is relatively new and,
limits the practical number of experiments that can be because of the lack of historical global data (though,
run and examined. Note that it is sometimes taken for here reanalysis data have proven invaluable), we do
CLIMATE VARIABILITY / Decadal to Centennial Variability 421
not have a comprehensive inventory of the global see Climate Variability: Seasonal to Interannual Var-
patterns, nor do we understand their mechanisms, iability) and the Pacific-North American Teleconnec-
couplings, longevity, or full implications for climate tion (PNA; 0400). In addition to these, there is a
prediction. However, we are encouraged by the study pattern in the tropical Atlantic (referred to as tropical
of the most thoroughly investigated pattern: the El Atlantic SST variability), in which SST often shows
Niiio Southern Oscillation (ENSO) pattern. El Nifio anomalous warmth (referred to as a ‘warm pool’) in
and La NiAa are extreme states of the ENSO pattern the tropical North Atlantic and a complementary cool
(see Monsoon: ENSO-Monsoon Interactions). The pool in the tropical South Atlantic, or vice versa. These
ENSO phenomenon is characterized by a pattern of seem to vary coherently over decadal time scales,
tropical Pacific sea surface temperature (SST)relative though they vary independently on shorter time scales
to the mean SST. Studies of it revealed that the state of in these regions. These low-frequency SST phenomena
the pattern (e.g., periods when the eastern tropical show concurrent anomalies in the rainfall over Brazil
Pacific SST was abnormally high) was related to and northern Africa. It has also been suggested that the
regional climate in various regions around the world. decadal changes in the SST in the subtropical North
Further investigation revealed that the pattern was Atlantic may be responsible for changes in the distri-
predictable according to some simple laws involving bution and intensity of hurricanes in that region.
the interaction of the atmosphere and ocean (over Likewise there is a decadal ENSO-like pattern,
short time scales). This led to considerable insights where low-frequency covarying changes in the trop-
regarding the nature of climate, the coupling between ical Pacific atmosphere and ocean strongly resemble
climate components (in this case, the ocean and the pattern of the interannual ENSO phenomenon,
atmosphere), scales of influence and, most important- including teleconnected anomalies in the midlatitude
ly, our first successful climate predictions (over atmosphere and ocean of the North Pacific. These
seasonal to interannual time scales). We are hopeful decadal ENSO-like anomalies are also teleconnected
that additional predictions for other regions of throughout the tropics, with large concurrent changes
the Earth may be realized in analogous fashion for in tropical Atlantic and Indian Ocean SST, in addition
dec-cen varying patterns. to the North Pacific.
These other patterns are not as well documented or This anomaly pattern has shown an extended
studied, but indeed they do appear to be related to ‘warm’ phase throughout the last few decades of
regional climate. Some are also related to the frequen- the twentieth century, which preceded a significant
cy of hurricanes, Nor’ Easters, crop yields, and reduction in the alpine glaciers throughout the tropics.
fisheries. The covariation of two patterns dominated The frequency of precipitation, stream flow, and
global temperature variations since the mid-1970s. snowpack in the north-west and south-west of North
Others display regional or global teleconnections, and America are also well correlated with this time series
they may serve to focus different climate forcings and describing the decadal ENSO-like climate pheno-
processes into single coherent responses. Because of menon variability.
these attributes and covarying relationships, it is A number of regional atmospheric patterns have
hoped that their further study may ultimately yield been analyzed such as the North Pacific Oscillation
benefits similar to those obtained through the study of (NPO),West Pacific Oscillation (WPO),West Atlantic
ENSO. Patterns also provide an obvious means for Pattern, and Pacific Decadal Oscillation (PDO). It is
breaking the complex climate system down into a not clear how these are related to the PNA or each
finite set of manageable, and hopefully predictable, other (if at all). A completely different kind of pattern,
components. involving sea ice, has been found in the Southern
Most modes are defined by statistical classifications Ocean, and is called the Antarctic Circumpolar Wave
of the observed variability in surface temperature, sea- (ACW).This pattern is characterized by deviations in
level pressure, or other quantities. The precise defini- the Antarctic sea ice extent from monthly climatolog-
tion may vary according to the statistical methodology ical averages, though it is also apparent in surface
employed to define them (see Data Analysis: Empirical wind, SST, and sea-level pressure anomalies near the
Orthogonal Functions and Singular Vectors). Statisti- winter ice edge. It is also highly coherent with
cal patterns may ultimately prove to be related to temporal variations in ENSO and the Indian Ocean
physical laws or to the distribution of land and ocean, monsoons. Other atmospheric patterns have been
of mountains, etc. They may also be artifacts of nature, identified in the Southern Hemisphere, though the
whereby they are not stable over long periods of time, data are typically too sparse in time and space to allow
or they may be statistical artifacts. more detailed analyses of these.
The two most important patterns that show dec-cen In addition to the above, there are structures which
variability are the North Atlantic Oscillation (NAO; may, or may not, be considered climate patterns,
422 CLIMATE VARIABILITY / Decadal to Centennial Variability
though they are often related to the other patterns or response of the ocean serves to integrate the high-
presented in a similar manner. For example, the Asian frequency atmospheric forcing resulting in an ocean
monsoon, though predominantly a seasonal signal, that also varies but over considerably slower time
is strongly correlated to ENSO and shows decadal scales. This theory, in combination with the influence
variability as indexed by precipitation and wind of land-sea contrasts and distribution of mountain
speeds over India. The global thermohaline circula- belts, allows for spatially varying patterns of dec-cen
tion (see Ocean Circulation: Thermohaline Circula- variability. It represents the most basic mechanism of
tion) has been tied to distinct changes in the ocean dec-cen variability (effectively, our ‘null hypothesis’ in
surface conditions and NAO in the North Atlantic the absence of more complex mechanisms).
Ocean. Other possible internal mechanisms include
the interactions between the slow variations of
the slow components (e.g., the ocean and the cryo-
sphere), and the coupling of system components
Mechanisms of Dec-Cen Variability that individually may not show slow variability, but
The mechanisms responsible for dec-cen climate together can. A considerable amount of attention
variability are conveniently separated into those that has been given to mechanisms of decadal ENSO
arise as a consequence of changes in the external variability whereby tropical SSTanomalies are quickly
forcing of the system and those that arise due to propagated through the atmosphere to the extra-
internal variability within the system (independent of tropics and midlatitudes where they introduce local
changes in the forcing). The external forcing on the anomalies to the ocean. The ocean retains the anomaly
Earth’s climate consists of solar radiation impinging while slowly transporting it back to the tropics
on the Earth’s atmosphere, aerosols (particles or liquid (via surface currents, subsurface currents, etc.)
suspended in air) from volcanic eruptions, and the where it moderates the tropical SST causing
chemical composition of the atmosphere (controlled further anomalies and continuation of the cycle.
by natural and anthropogenic sources/sinks of green- Considerable progress has been made toward identi-
house gases). The ice ages and the more recent Little fying potential mechanisms, though more work is still
Ice Age have been attributed to changes in the intensity required. Presently, these hypotheses help to focus
of incoming solar radiation. Such changes are typically model experiments and observational studies.
very small (of the order of a couple percent for the ice
age changes associated with changes in the Earth’s
orbital geometry; and even smaller for the Little Ice
Age associated with changes in sunspot activity.
Future Directions and Needs
Unfortunately, while the observed climate variations As stated, the study of decadal-to-centennial climate
of the ice ages and Little Ice Age are consistent with the variability is in its infancy. New discoveries are being
variations in solar activity, we still do not understand made each year, and at the time of this writing we are
how such minuscule changes can drive such significant clearly on the steep slope of the learning curve. The
responses. task is onerous given the considerable demands on the
Clearly changes in climate associated with changes models for including all components of the climate
in the external forcing require that we can predict the system (atmosphere, ocean, cryosphere, biosphere,
changes in the forcing. Therefore, climate predictions and land surface), resolving processes operating at the
and mechanisms of change associated with changes in smallest spatial scales to the largest, and having to
external forcing require studies that work under given integrate (run) the models for decades to centuries of
scenarios of change. model years to realize a single simulation. More
The most fundamental theory regarding internal powerful computers and their broader availability to
dec-cen climate variability (Le., independent of climate researchers will aid considerably in this
changes in external forcing) was presented in an respect, as will close collaboration between scientists
elegant theory by Hasselmann, whereby the day-to- of multiple disciplines, and between observationalists,
day weather, representing high-frequency climatic theorists and modelers.
noise, works to drive a slow component of the climate We are faced with changing atmospheric concen-
system such as the ocean. Relative to the atmosphere, trations of radiatively active gases and we need to
the ocean has an enormous heat capacity and mixes obtain more accurate records of the actual emission
slowly. These characteristics regulate the speed with rates to help constrain models simulating past condi-
which it can respond to changes in the forcing (thus, it tions. We have limited observations of dec-cen climate
is considered a slow component relative to the variability, and those from the twentieth century may
atmosphere which is a fast component). The ‘muted’ already be contaminated by anthropogenic climate
CLIMATE VARIABILITY / Decadal to Centennial Variability 423
change (masking the signal of natural variability that is between several regions of ac-
required in order to ultimately recognize an ant- tivity, linked to several smaller
hropogenic change from the natural variability back- patterns in Pacific and appar-
ground). There are additional demands on the data ently to regional and hemi-
quality required to efficiently study dec-cen variabil- spheric climate in general.
ity. This reflects the fact that dec-cen change pro- Teleconnection: apparent links between broad-
ceeds at such a small pace on a year-to-year basis, ly separated regions of the
and that any such change is easily lost within the Earth, identifying locations
diurnal and seasonal cycle as well as standard inter- that covary in some climatic
annual variability, all of which are large relative to the characteristic and presum-
annual dec-cen change. Thus, we need high-precision ably indicating some causal or
instruments to pick up dec-cen change as early as perhaps predictable relation-
possible. ship.
At present, there is no long-term climate observing Cryosphere: That portion of the climate
system for dec-cen variability in place. Consequently, system related to ice or extreme
if we are to provide an observational basis from which cold (cryo means very cold),
future generations will be able to more reliably such as glacial ice, sea ice,
diagnose their model (hindcast) predictions, and snow, lake and river ice, cold
analyze comprehensive records of dec-cen variability, ocean regions and alpine belts,
we must begin a systematic collection of key variables but typically not including ice
now. This requires close coordination between re- clouds (cirrus), or the frigidly
search and operational groups. cold stratosphere.
Despite these impediments, we are encouraged
by the fact that climate patterns may ultimately
allow us to predict some aspects of dec-cen See also
climate variability. The problem is of considerable
Climate Variability: North Atlantic and Arctic Oscillation;
importance and will yield invaluable insights regard-
Seasonal to lnterannual Variability. Data Analysis: Em-
ing the nature and sensitivities of our planet’s climate
pirical Orthogonal Functions and Singular Vectors. El NiAo
system. and the Southern Oscillation: Observation. Ice Ages
(Milankovitch Theory). Numerical Models: Methods.
Nomenclature Ocean Circulation: Thermohaline Circulation. Paleocli-
matology: Ice Cores. Teleconnections.
Dec-cen variability: climate variability on decade to
century time scales
Mode (or pattern): statistically consistent spatial Further Reading
patterns in observed climate
variability in which patterns Bradley RS and Jones PD (eds)(1992)Climatesince AD 1500.
preserve a general shape in London: Bradley, Routledge.
Hasselmann K (1976)Stochastic climate models. 1. Theory.
space, but whose amplitude
Tellus 28: 473-485.
varies through time. Kalnay E, Kanamitsu M, Kistler R, et al. (1996)The NCEP/
NAO: North Atlantic Oscillation, NCAR 40-year reanalysis project. Bulletin of the Amer-
representing a large spatial pat- ican Meteorological Society 77: 437471.
tern that influences much of the Mann ME, Bradley RS and Hughes MK (1998)Global-scale
North Atlantic and bordering temperature patterns and climate forcing over the past six
continental regions, indexed centuries. Nature 392: 779-787.
by pressure difference between National Research Council (1998)In: Martinson DG et al.
Iceland and Lisbon, and linked (eds) Decade-to-Century-Scale Climate Variability and
to storms, winds, precipitation, Change: A Science Strategy. Washington, DC: National
and other regional climate. Academy Press.
National Research Council (1998)In: Moore B et al. (eds)
PNA: Pacific-North American Tele- Overview of Global Environmental Change: Research
connection, representing a Pathways for the N e x t Decade. Washington, DC: Na-
large spatial pattern that influ- tional Academy Press.
ences much of the North Pacific National Research Council (1998) In: Karl T et al. (eds)
and North America, indexed Capacity of US Climate Modeling. Washington, DC:
by 500 mb height anomalies National Academy Press.
424 CLIMATE VARIABILITY / Glacial, Interglacial Variations
National Research Council (1998) In: Dutton J et al. (eds) the N e x t Decade. Washington, DC: National Academy
The Atmospheric Sciences Entering the Twenty-First Press.
Century. Washington, DC: National Academy Press. National Research Council (1999) In: Karl T et al. (eds)
National Research Council (1999) In: Moore B et al. (eds) Adequacy of Climate Observing Systems. Washington,
Global Environmental Change: Research Pathways for DC: National Academy Press.
Glacial, Interglacial Variations

P Yiou and D Paillard, Laboratoire des Sciences du generally implicitly assumed that changes in climate
Climat et de I’Environnement, Gif-sur-Yvette, France should also be gradual ones. But recurrent and abrupt
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. climatic changes of large amplitude have been found in
many locations, without any possible link with orbital
variations. This ‘sub-Milankovitch’ variability inval-
Introduction and Basic Principles idates the idea that climate changed continuously in
Instrumental records of climatic parameters only the past. The climate appears to be a much more
cover one or, at best, a few centuries. Since climate variable system than previously suspected. With all
is often defined as the mean state of the atmosphere this new information on Quaternary climatic changes,
over several decades, this represents a very poor the understanding of glacial-interglacial cycles is
sampling interval for the study of climate natural currently rapidly progressing.
variability. If direct observations are not available,
reconstructions over much longer periods of time are
necessary, in particular to better understand the Glacial-Interglacial Variations
stability properties of the climate system. Many
paleoclimatic indicators have thus been developed
Glacial Times
and compiled, in particular t o document the late
Pleistocene era, the last few hundred thousands of During glacial times, the Earth was quite a different
years, which also covers the time of human develop- environment. Huge ice sheets, up to 3 k m high,
ment. The largest observed environmental changes covered Canada and Scandinavia. The water mass
are undoubtedly the succession of glacial and inter- that was thus stored over the continents represents
glacial climates. about 3 % of the oceanic reservoir. Consequently, the
Geologists in the mid-nineteenth century have sea level was about 12Om lower, and the coastlines
found many traces of glacial deposits far away from were quite different from the present day: England was
present-day glaciers. The only possible explanation no more an island, the Indonesian throughflow was
was a much colder climate, and the theory of Quater- much reduced, and several marginal seas were no more
nary glaciations was born. From the beginning, it was open to the ocean. Because of the dramatic increase in
suspected that the orbital variations of the Earth may the extent of continental and sea ice, the planetary
be responsible for the succession of glacial and albedo was significantly larger, and the global mean
interglacial epochs. Still, Milankovitch was the first, temperature was several degrees below the present-
in the early twentieth century, t o provide a convincing day value. Climate was generally much drier. The
astronomical theory of paleoclimates based on the atmospheric COZ concentration was around 200 ppm,
detailed computation of the solar energy received at compared to the preindustrial value of 280 ppm.
the top of the atmosphere, or insolation. But a major Ecosystems were shifted Equatorwards by up to a
difficulty of geology at this time was the impossibility thousand kilometers.
to provide an absolute temporal framework. The How such drastic changes are induced is still a
recent advances in geochemistry, the use of radioactive matter of debate. As early as in the nineteenth century,
chronometers, and the increasing accuracy of analy- variations in the Earth’s orbital parameters were
tical technologies have largely confirmed the role of suspected to have a fundamental role. However, since
orbital variations on the Earth’s climate. Still, many most components of the Earth system are involved (the
questions remain unanswered. atmosphere, the ocean, the cryosphere, the biosphere,
In the 1990s the discovery of widespread sub- and many geochemical cycles), the detailed mecha-
Milankovitch variability was a big surprise. Since the nisms leading to ice age cycles are still largely
orbital changes of the Earth are very slow, it was unknown.
CLIMATEVARIABILITY / Glacial, InterglacialVariations 425
Glacial Cycles and the Astronomical Theory annual mean and only change the seasonal distribu-
tion of insolation (10-100 W m P 2 ) . Obliquity and
In the 1970~-80s,advances in analytical technologies
precessional changes are usually assumed to be the
and in geochemistry provided a clear demonstra-
main external forcing parameters, while eccentricity is
tion that the glacial-interglacial changes are induced
important mostly in modulating the amplitude of the
by the variations of the orbital parameters of
precessional parameter e sin z i ~ .
our planet. Indeed, the Earth's orbit is an ellipse
The theory of Milankovitch stresses the role
whose characteristics are influenced by the other
of summer insolation at high northern latitudes
planets (Figure 1).Its eccentricity e undergoes changes
on the persistence of snow fields throughout the
with pseudo-periodicities around 100 and 400 kyr
year, and therefore on the mass balance of ice
(1kyr = 1000 years). Furthermore, the moon and
sheets in these areas. This is indeed probably a critical
the sun act on its equatorial bulge to change the
parameter and climate appears to respond almost
position of the Earth's axis: the tilt, or obliquity, E of
linearly to this insolation in the 23 and 4 1 kyr
the Earth's axis with respect to the orbital plane frequency bands. Nevertheless, the mechanisms
changes between about 22" and 24.5" with a main and feedbacks leading to the building and melting
periodicity of 41 kyr, and the position m of the
of the huge ice caps present during glacial times,
astronomical seasons (solstices and equinoxes) rela-
or those leading to large changes in atmospheric
tive to the perihelion is also changing. Its energetic
C 0 2 levels, are poorly known. The most important
effect, measured by the precessional parameter e sin m,
changes occur about every 100 kyr (Figure 3). These
varies with periodicities around 23 and 1 9 kyr. All
large changes cannot be directly related to the eccen-
these astronomical periodicities have been found in
tricity changes, both from data analysis and
the geological record (Figure 2), which is a clear
from simple energetic considerations. Furthermore,
demonstration that astronomical changes are driving
probably the largest ice volume (or sea-level) change in
climate variations.
the Quaternary occurred around 430 kyr before
Eccentricity changes only affect the global mean
present (BP): sea level during stage 1 2 may have been
annual insolation very slightly (0.1-0.5 W m P 2 ) . Ob-
20 m lower than during the last glacial maximum
liquity changes, by shifting the location of the tropics
(LGM), while during stage 11 sea level might
and polar circles, affect the latitudinal distribution
have been 2 0 m higher than the present, which
of annual insolation by several W m - 2 , the effect
probably implies the melting of Greenland and
at high latitude being maximum in summer (10-
parts of Antarctica. This deglaciation (termination
1 0 0 W m - 2 at the pole) and null in winter. The V) occurred at a time of low eccentricity, and there-
precessional changes are, on the contrary, zero in
fore of minimal insolation changes. The reverse
situation also happens to be true: termination I11
(between stages 8 and 7 ) is rather small, while
I a I the insolation changes are maximal at this time. A
1
I
I
4
If 2 I
simple linear relationship between high northern
I I I
latitude summer insolation and glacial cycles is
consequently not possible, at least for the main
terminations, and strongly nonlinear mechanisms
need to be invoked to account for these observed
peculiarities in the glacial cycles. Another difficulty is
PeI the 'mid-Pleistocene' transition about 1 million years
ago, from climatic variations dominated by the 41 kyr
periodicity, to the more classical 100 kyr cycles that
are dominant since. This evolution can be understood
NH Spring Earth
as the last stage of the slow cooling trend all along the
Cenozoic era, from tropical climates all over the Earth
Figure 1 Orbital parameters of the Earth involved in the
insolation. The eccentricity e is defined as e = c / a (c = distance
during the Cretaceous to glacial cycles in the Quater-
between focus and center of the ellipse, a = semimajor axis). The nary. Still, no particular geological event can be
present-day value is e = 0.0167, The tilt of the Earth's axis with associated with this mid-Pleistocene transition, and
respect to the orbital plane is the obliquity E (current value the Earth climate appears to switch along, rather
E = 23.44") which defines the location of the tropics and polar
abruptly, from one variability mode to another one.
circles on Earth. The intersection of the Earth's equatorial plane
and its orbital plane defines the line "J'and the position of solstices
Here again, the concept of 'quasi-linear' oscillations
and equinoxes. The position of the perihelion with respect to y around some basic equilibrium state appears to be
(the spring equinox) is called m,the climatic precession. inadequate.
426 CLIMATE VARIABILITY / Glacial, InterglacialVariations
SPECMAP Stack 6 ' * 0
0 200 400 600 800

Time (kyr BP)
Spectral power estimation with

100 i 30 kyr the Blackman-Tukey method
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
Frequency (cycles per 1000 years)
Figure 2 A 'reference record' of glacial-interglacial cycles has been provided by the SPECMAP group (spectral mapping and prediction
project). The curve has been obtained by averaging several records of the oxygen isotopic composition of the calcite of foraminifera. It
reflects, to first order, the variations of the mean ocean oxygen isotopic composition, and thus the variations of global ice volume. The
spectral analysis of this record, as well as of most paleoclimatic records, reveals the astronomical periodicities.
Variability During the Last Ice Age from the Laurentide ice sheet eroding the floor, and
transporting the material into the middle of the
Phenomenology, Data, and Models
Atlantic Ocean off the margin of Portugal. Analyses
The presence of large ice sheets over North America, of oxygen-18 (see Isotopes, Stable) in these cores
Scandinavia, and Siberia during the last ice age showed spikes interpreted in terms of salinity anom-
(approximately 110-15 kyr BP) lowered the average alies due to discharges of fresh water. These events
temperature over the Northern Hemisphere, with a were numbered and named after Heinrich who first
spatial variability caused by orography changes. reported the debris layers in a marine core (Figure 4).
Those large ice sheets also had an impact on the In parallel, millennia1 variations - which are con-
climate stability itself, and enhanced the magnitude of sidered rapid compared to Milankovitch typical fre-
temperature variations, as well as other factors, such quencies -were also observed in a Greenland ice core.
as precipitation, sea level and the ocean circulation. These oscillations in isotopic content were proved to
The contrast between the relatively calm Holocene be climatically relevant when the high-resolution cores
and the variable ice age is shown in Figure 4 (GRIP and from Dome Summit (Greenland) came, and were
Byrd isotopic records). The wide oscillations that correlated with Northern Atlantic marine cores, as
punctuated the last ice age were discovered jointly in well as some terrestrial records. Such oscillations have
marine and ice core records, and were given particular an approximate periodicity of 1500 years, and were
attention as soon as a common driving mechanism was also found in high-resolution marine records. Many
proposed for the overall picture. research groups have reported this kind of variability
The presence of layers of debris in Northern Atlantic in various remote places on the planet (Antarctica,
cores suggested the existence of iceberg discharges Santa Barbara Basin, China, etc.), hence suggesting a
CLIMATE VARIABILITY / Glacial, Interglacial Variations 427
Marine isotopic stage numbers
0"280
'.-1
260
t
c
240
E 220
180
3
0 100 200 300 400 500
Age (kyr BP)
Figure 3 The topmost curve is the summer solstice insolation at 65"N normalized to zero mean and unit variance. According to the
classical Milankovitch theory, this represents the main external forcing. Below is a record of global ice volume obtained in a similar way as
in Figure 2. It is worth noting that the largest sea-level change, around 420 kyr BP, corresponds to minimal insolation changes. The
converse also appears to be true about 240 kyr BP, with large insolation changes associated with minimal ice volume variations. The next
curve is a record of sea surface temperatures in the Southern Ocean. The next two curves are the temperature changes and the
atmospheric concentration in COPas recorded by the Vostok ice core. It is interesting to note that, in some sense, the 'main transitions' are
more clearly defined at Vostok or in the Southern Ocean, than using the ice volume record.
global phenomenon. The associated temperature var- which are, on occasion, accompanied by a Heinrich
iations can have an amplitude of 5-10°C in the North event (HE)(Figure 4).
Atlantic; they can take place in less than a century Modeling studies focused on climate instabilities
(hence the term 'abrupt climate change'), and last for driven by ice sheet oscillations. The ice sheet oscilla-
several centuries. Methane contents of ice cores also tions were controlled by albedo and precipitation
record these millennia1 variations, which indicates feedbacks, or by basal melting, and the models could
large changes in climate conditions at the surface of the predict the typical periodicity of HEs, i.e., around
planet. 7000 years. Several modeling groups have also pointed
When the marine and ice records are put into to the existence of multiple states of the ocean
perspective, they highlight the features of Dansgaard- thermohaline circulation, which can be destabilized
Oeschger (DO) oscillations during the last ice age, by freshwater perturbations. Thus it is very probable
Next Page
428 CLIMATE VARIABILITY / Glacial, Interglacial Variations
1
-34
-36
0 -38
m
“i -40
-42
-44
3 3000
-3 2500
2000
5:
1500
1000
E
0) 500
5 0
. . ._
10 15 20 25 30 35 40 45 50
Time (kyr BP)
Figure 4 Climate variations during the last ice age, recorded from two ice cores and a marine core. The isotopic variations (6”O in
permil) measured in the GRIP core are shown in green; the corresponding temperature range is about 10°C. The purple curve represents
6”O in the planktonic foraminifera of a North Atlantic core (ENAM 93-21); the isotopic variations mostly account for salinity anomalies due
to freshwater pulses. The red curve represents magnetic susceptibility in the ENAM marine core, which is a proxy forthe intensity of bottom
currents. The blue curve represents the 6”O in Antarctica measured in the Byrd ice core.
that the ‘fast’ climate oscillations during the last ice age phere Models). Up to now, only simplified models are
involved interactions between: able to simulate sequences of D O cycles, mainly due to
computer limitations.
ice sheets, because they are the main suspect for a
sudden freshwater input and have a large impact on
Dynamics
planetary albedo;
the oceans; the thermohaline circulation is a main The prominent cycles around 7 and 1.5 kyr of the last
conveyor of heat from the Southern Hemisphere to ice age were revealed by spectral analyses (see Data
the Northern Hemisphere; shutting it off would Analysis: Time Series Analysis) of high-resolution
have a large impact on climate; records. It is important to note that the distances
the atmosphere, which responds almost instantly, between two successive DO oscillations or HEs do
but feeds the ice sheets with snow, so that they can vary so that there is no strict periodicity in any record.
grow again, and affects the ocean surface conditions However, the fast events are more frequent towards
with the hydrologic cycle. the end of the last ice age. Thus, this oscillatory
behavior is not stationary through - time, and the
The variety (or hierarchy) of models for such interac- frequency of the DO events seems to be connected to
tions is described elsewhere (Coupled Ocean-Atmos- the size of the ice sheets.
CLOUD CHEMISTRY 451
patterns and ocean temperatures in the tropical Pacific seasonal climate predictability is the ENSO cycle,
are a particular focus of real-time monitoring. Areas of which has been shown to correlate with climate
the planet experiencing extreme, persistent anomalies anomalies in many regions throughout the world.
of temperature and precipitation are also monitored Ocean temperatures and surface winds across the
closely. Drought indices, which can be based on tropical Pacific are now observable in real time thanks
different combinations of precipitation, soil moisture, to the development of an extensive monitoring system
and vegetation status indicators, are particularly that combines in situ and satellite observations.
closely monitored and disseminated to the public. Climate forecasters know that if the equatorial Pacific
If any of the causes of interannual variability listed Ocean and atmosphere develops a significant cold or
above persist for longer than a season (or if the causes warm anomaly in the boreal autumn then it is highly
themselves could be predicted a season or more in likely that such anomalous conditions will persist
advance), and if the magnitude of interannual varia- through the following winter season. A winter sea-
bility forced thereby is large compared to unpredict- sonal forecast featuring enhanced probabilities of
able interannual variability such as weather noise, then large-scale anomalies of precipitation and tempera-
in principle it should be possible to make skillful ture associated with ENSO extrema can then be
predictions of short-term climate anomalies at lead disseminated in time for precautionary action.
times extending beyond the deterministic limit of a
weather forecast. Such seasonal forecasts of short-term See also
climate anomalies cannot predict the evolution of
individual weather systems, and thus are incapable of Air-Sea Interaction: Sea Surface Temperature. Cli-
predicting the weather on specific days. The goal mate Variability: Decadal to Centennial Variability; North
Atlantic and Arctic Oscillation. Coupled Ocean-Atmos-
instead is to forecast how the average weather for a
phere Models. Drought. El Niiio and the Southern
month or a season will differ from the expected value Oscillation: Theory. Kelvin Waves. Middle Atmos-
derived from a straightforward climatological average. phere: Quasi-Biennial Oscillation. Monsoon: Overview.
Seasonal hurricane forecasts for the north-west Predictabilityand Chaos. Teleconnections.Tropical
Atlantic region, for example, are based on knowledge Meteorology: Tropical Climates. Walker Circulation.
of the ENSO cycle, Atlantic ocean temperature Weather Prediction: Seasonal and lnterannualWeather
anomalies, the phase of the QBO in tropical strato- Prediction.
spheric winds, and several other factors that can be
observed prior to the start of the hurricane season. At
the beginning of the North Atlantic hurricane season Further Reading
(in late summer), forecasts are issued for the number of Barry RG and Chorley RJ (1998)Atmosphere, Weather,and
tropical cyclones that are expected to make landfall in Climate. London: Routledge.
North America based on historical associations be- Hartmann D (1994) Global Physical ClirnatologSanDiego:
tween the factors listed above and subsequent hurri- Academic Press.
cane activity. It should be emphasized that no attempt Peixoto JP and Oort A (1992)Physics ofclimate. New York:
is made to forecast the timing or tracks of individual American Institute of Physics.
cyclones, simply the aggregate hurricane activity for Philander SGH (1990)El Nirio, La Niria, and the Southern
Oscillation. San Diego: Academic Press.
the entire season. Shea D (1986)ClimatologicalAtlas: 2950-1979. SurfaceAir
Active operational prediction efforts are aimed at Temperature,Precipitation,Sea-Level Pressure, and Sea-
producing skillful and useful estimates of the proba- Surface Temperature. NCAR Technical Note NCAW
bilities of seasonal climate anomalies of temperature TN-269+STR. Boulder, CO: NCAR.
and precipitation up to about a year in advance. At Trenberth KE (ed.) (1992) Climate System Modeling.
present the principal well-demonstrated source of Cambridge: Cambridge University Press.
J L Collett and P Herckes,Colorado State University, Introduction

Fort Collins, CO, USA
Clouds occupy a small but important fraction of the
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. total volume of the troposphere. In addition to playing
452 CLOUD CHEMISTRY
a key role in the global hydrologic cycle and influenc- are required for the homogeneous nucleation of water
ing atmospheric radiative transfer, clouds interact vapor. Atmospheric particles often contain mixtures
with a variety of chemical species. Together with gases of organic and inorganic components. Hygroscopic
and particles, clouds comprise a complex multiphase particles can take up or lose water in response to
system. Clouds act both as reactors for the production changes in the ambient relative humidity (see Aero-
of new chemical species and as vectors for particle and sols: Role in Cloud Physics). At high humidity these
trace gas removal, via wet deposition and direct particles can take up water vapor. The point at which a
deposition of cloud and fog drops to the surface. dry particle spontaneously takes up water vapor to
Figure 1provides an overview of several important form a saturated solution is known as the relative
processes in the multiphase atmospheric system. humidity of deliquescence (RHD).The RHD depends
Clouds and fogs (clouds in contact with Earth’s on particle composition.
surface) interact with both aerosol particles and trace Particles that have deliquesced are referred to as
gases. The incorporation of particles and gases into haze particles. As the humidity increases above the
cloud drops are key steps in determining the initial RHD, haze particles take on additional water to
chemical composition of the cloud. The sections below maintain equilibrium with the partial pressure of
discuss several aspects of tropospheric cloud chemis- water vapor in the atmosphere. The equilibrium
try. It is also important to keep in mind that clouds partial pressure of water vapor above a haze drop
occur in the stratosphere too, where they play depends on the drop size and the drop composition.
an important role in promoting heterogeneous reac- Increased drop curvature raises the equilibrium water
tions important to stratospheric ozone depletion and vapor pressure via the Kelvin effect. Increased drop
other processes, though this subject is not discussed solute content lowers the equilibrium vapor pressure
here. by displacement of water molecules near the drop
surface. The combined effects of drop curvature and
solute content are often described using Kohler theory
Cloud Drop Formation
(see Aerosols: Role in Cloud Physics). Application of
Cloud drops form via condensation of water vapor Kohler theory indicates that for a given particle size
onto a subset of particles termed cloud condensation there is a critical supersaturation above which drops
nuclei (CCN), a process known as heterogeneous will spontaneously take up water vapor and grow
nucleation. In the absence of suitable particles, clouds unstably. Such drops have been activated. The process of
would not form readily, as enormous supersaturations activation is also referred to as nucleation scavenging.
Figure 1 Schematic representation of the multiphase cloud-particle-trace gas system in the atmosphere. Included are processes of dry
and wet deposition, particle and gas scavenging by cloud drops, chemical reaction, and precipitation formation in a mixed-phase
(ice-liquid water) cloud.
CLOUD CHEMISTRY 453
Growth of activated drops is limited primarily by the C, is the aqueous phase concentration inside the
availability of water vapor. Haze particles which do droplets resulting from particle scavenging, Ca
not reach their critical supersaturation will not acti- the initial mass concentration of the particles, and E,
vate and will remain in equilibrium with the ambient the mass fraction incorporated in the drops: the
water vapor concentration. scavenging efficiency. LWC is the cloud liquid water
Many types of atmospheric particles are capable of content. E,, E B , and EC represent the mass fraction of
serving as CCN. In many clouds, particles as small as matter incorporated by nucleation ( E ~ ) , Brownian
0.1 pm in diameter can activate and grow into cloud motion ( E B ) , and collision ( E C ) . Overall mass scaveng-
drops. In other cases, particularly when supersatura- ing efficiencies for soluble aerosol species, such
tions are low and/or particle concentrations are high, as sulfate or chloride, may be only 20-30% for
the minimum particle size activated may be 0.5 to polluted radiation fogs, but can approach 100%
l p m . CCN have traditionally been thought to be for clouds formed by vigorous updrafts in pristine
comprised mainly by sulfate particles, sea salt parti- environments.
cles, and, in some environments, nitrate particles.
Recent studies, however, indicate that many carbon-
aceous particles are also capable of acting as CCN.
Further work is needed to elucidate interactions of Dissolution of Soluble Gases
carbonaceous particles with clouds and fogs. In addition to particle scavenging, the composition of
The soluble fraction of the cloud condensation cloud drops can be significantly affected by dissolution
nucleus determines the initial chemical composition of of soluble gases. These processes are represented in
the cloud drop. In addition to nucleation scavenging, Figure 2, where several key soluble gases are repre-
unactivated (interstitial) aerosol particles can be sented. Even in remote areas, cloud drops can be
incorporated into cloud drops by a variety of mech- partially acidified by dissolution of carbon dioxide. In
anisms including interception, inertial impaction, and polluted areas, further acidification can occur by
diffusion. While nucleation scavenging can often be uptake of nitric acid. In certain environments hydro-
quite efficient, scavenging of interstitial particles tends chloric acid can also be important. Low-molecular-
to be relatively inefficient. weight carboxylic acids, especially formic and acetic
The concentration of a compound in the aqueous acids, can be important contributors to drop acidity in
phase resulting from particle scavenging can be both polluted and pristine environments. Sulfur diox-
expressed as ide is moderately soluble in most cloud drops. Its
uptake is of particular interest because of the potential
Cw=- Ea Ca (En + +
EB EC) for rapid oxidation to sulfate in the aqueous phase.
111 Gaseous oxidants such as hydrogen peroxide and
LWC = ca LWC
Figure 2 Schematic representationof the influence of particles and gases on cloud/fog drop composition. Reactions between several of
the dissolved species shown here can also be important.
454 CLOUD CHEMISTRY
ozone are important because they serve as effective trace gases in aqueous solution varies strongly, from
oxidants of dissolved sulfur dioxide. A number of slightly soluble species ( e g , 0 3 ) to moderately soluble
other organic gases can also dissolve into drops. species (e.g., SOZ), to very soluble species (e.g., H202
Dissolution of formaldehyde is depicted in Figure 2. and H N 0 3 ) .
Formaldehyde is often of interest because it may be
present at high concentrations and because it can react
with dissolved sulfur dioxide, as discussed below. Drop Composition
The extent to which a gas partitions into a cloud
drop at equilibrium depends on its solubility. In some The combined effects of particle and gas scavenging,
cases the time required to achieve equilibrium is long along with chemical reactions (discussed below),
relative to the rate of reaction of the dissolved gas or determine the chemical composition of cloud
relative to the drop’s lifetime, so that phase equilib- drops. Absolute concentration levels of individual
rium may not be achieved. When the liquid and gas species are also influenced in part by condensational
phases are in equilibrium, the concentration in the growth or evaporation of drops. Major species
drop is given by Henry’s law: commonly found to dominate fog and cloud compo-
sition include a number of ions, especially nitrate,
sulfate, and ammonium. In coastal regions, sea
salt ions are important contributors to the
where [Xaq]is the concentration of the species X in drop composition, while in some locations contribu-
solution (mol.L-’), p , the partial pressure of this tions of soil dust components can be important.
species in the atmosphere (atm) and H , the Henry’s Although measurements are not widely available,
Law constant for the species (mo1.L-1 atm-I). The it is apparent that organic compounds can also
solubility of most gases increases with decreasing be important contributors to fog and cloud composi-
temperature. tion. In some polluted environments, concentra-
As written above, Henry’s law expresses the phys- tions of formaldehyde, acetate, and formate can
ical solubility of a gas. For many species, the overall reach levels similar to those observed for major
solubility is further enhanced by dissociation or inorganic ion species. Table 1 provides approximate
reaction in solution. An important example is the concentration ranges of various chemical species
dissolution of sulfur dioxide: observed in clean and polluted clouds and in polluted
radiation fogs.
Experimental observations reveal that only a
fraction of the total organic carbon (TOC) content
of fogs or cloud consists of low-molecular-weight
compounds such as formate, acetate, and formalde-
hyde. To date, only modest efforts have been made to
HS0;bSO;- + H+ PI1 identify and quantify higher-molecular-weight com-
If the compounds are reactive in solution, it is useful to pounds in fogs and clouds, although some studies have
define an effective Henry’s law constant which takes focused on measuring particular compound families
into account the chemical reactions and includes
the total amount of incorporated compound (e.g.,
+
H z S 0 3 HSO; + SO:-). Table 1 Typical range of solute concentrations in remote and
polluted clouds and fogs
Solute Polluted Polluted Remote
cloud fog cloud
PH 2-5 3-7 4-6

In eqn [3], K1 and K2 are the acid dissociation SO$ (peq L-’) 50-2 000 50-5 000 5-50
constants corresponding to reactions [I13 and [111], NOS (peqL-’) 10-2 000 50-20 000 0-20
respectively. In the case of acid-base reactions in NH: (peq L-‘1) 50-1 000 100-20 000 10-50
solution, the effective Henry’s law constant varies with cr (peq L-’)
Na* (peq L-’)
0-500
0-500
0-1 00
0-200
0-500
0-500
pH. In the case of sulfur dioxide dissolution, the HCOO- (peq L-’) 0-1 00 0-1 000 0-20
formation of the ionized forms provides extra reser- CH3COO- (peq L-‘1) 0-100 0-500 0-1 0
voirs for sulfur dioxide in solution, thereby increasing HCHO (pM) 10-50 5-500 0-1 0
the effective solubility of sulfur dioxide substantially TOC (PPmC) 2-30 5-50 0-2
~ ~
when the drop pH exceeds one or both pKa’s for peq L-’, microequivalents per liter; pM, micromolar; ppm C, parts
dissolved sulfur dioxide. The solubility of atmospheric per million carbon by mass.
CLOUD CHEMISTRY 455
because of their specific sources or their toxicity. 1 .o

Organic compounds measured in fogs and clouds
include carboxylic acids (formic, acetic, pyruvic,
propionic), aldehydes (formaldehyde, acetaldehyde,
glyoxal, benzaldehyde), polycyclic aromatic hydro-
carbons (chrysene, pyrene, fluoranthene, benzo(e)-
pyrene, benzo(b)-fluoranthene), ketones (acetone),
pesticides (carbaryl, diazinon, malathion, lindane, -
C
.-0
0.5
atrazine), and various phenols (phenol, 2- and 4- I
I
nitrophenol, 2,4-dinitrophenol, 3/4-methylphenol
(m/p-cresol), guaiacol, syringol). This list of com-
pounds is illustrative but not exhaustive. Many
organic compounds found in cloud drops are emitted
during combustion, while many are formed as sec-
ondary products of atmospheric reactions. With the
"
exception of the low-molecular-weight carboxylic 0 2 4 6 8 10
acids and aldehydes, cloud drop concentrations of
PH
most individual organic compounds are typically
much smaller than those of the major inorganic Figure 3 The speciation of dissolved sulfur dioxide as a function
species. of pH.
Although clouds have traditionally been assumed to
be comprised of populations of chemically homoge-
neous drops, there is increasing evidence that a wide
between 2 and 7 ) ,the dominant form of dissolved
variety of drop compositions often exists within a
sulfur dioxide is bisulfite. As the pH climbs above 7,
given cloud or fog. These differences arise in part from the speciation shifts toward sulfite.
the variability in the composition of the underlying
Several S(1V) oxidation mechanisms occur in
CCN. Other differences arise from variations in
the aqueous phase; their relative importance depends
condensational growth rates as a function of drop
on the drop p H and the availability of oxidants
size and from nonequilibrium absorption of highly
and catalysts. Three important aqueous S(1V) oxida-
soluble trace gases. Increasingly, cloud chemistry tion pathways are oxidation by hydrogen per-
models and measurements are being designed to oxide, oxidation by ozone, and oxidation by oxygen
investigate chemical heterogeneity among cloud drop (autooxidation) catalyzed by Fe(II1) and Mn(I1).
populations. Figure 4 depicts the rate of sulfate production by these
three pathways for l p p b ( v ) sulfur dioxide in the
Reactions Occurring in Cloud Drops presence of 30 ppb(v) ozone and 1ppb(v) hydrogen
peroxide. The rate of production is shown as a
Many chemical species incorporated in cloud drops function of drop pH.
can react in the aqueous phase. Considerable attention In many clouds hydrogen peroxide is assumed to be
has been focused on the aqueous oxidation of dis- the dominant S(1V)oxidant. As shown in Figure 4,this
solved sulfur dioxide to sulfate, although numerous oxidation pathway tends to be much faster than the
other reactions have also been shown to be important. others at lower pH values. While the ozone and
Several key chemical reactions are discussed briefly autooxidation paths become slower with decreasing
below. pH, the rate of the hydrogen peroxide pathway is
essentially independent of pH over the range of
Aqueous-Phase Sulfur Oxidation and Complexation interest. The lack of p H dependence results from the
opposing p H dependence of two contributing factors.
Oxidation of SO2 to H2S04 is a key reaction in the
While the effective solubility of sulfur dioxide de-
atmosphere. The resulting sulfate contributes to acid
creases with decreasing pH, the intrinsic rate of sulfate
deposition, visibility degradation, respiratory prob-
production increases at lower pH. It is believed that
lems, and climate modification. Sulfate production oxidation occurs by nucleophilic displacement of a
occurs both in the gas phase and in the aqueous phase,
water molecule by hydrogen peroxide attack on
but the latter is much faster.
bisulfite:
As described above, dissolution of SO2 in cloud
drops is a function of the drop p H (Figure 3 ) . In the pH
range usually encountered in cloud and fog drops (pH HSO; +H 2 0 2 + S 0 2 0 0 H - + H20 [IV]
456 CLOUD CHEMISTRY
1E-2 is oxidized much more rapidly than bisulfite, which is

oxidized more rapidly than sulfurous acid, the rate of
sulfate production by this pathway increases strongly
1E-4 with increasing pH. As shown in Figure 4, the rate of
this pathway can exceed even that of the hydrogen
peroxide pathway at pH values greater than 5 for
1E-6 typical conditions. The rate of this pathway becomes
so rapid, in fact, that it tends to become mass-
r
h
transport-limited for larger drops at high pH. This is
(0
1E-8 especially likely when formaldehyde is present as a
E.
copollutant, since dissolved sulfur dioxide is rapidly
e
c
2
consumed by reaction with both ozone and formalde-
5 1E-10 hyde at high pH.
m
I Aqueous S(1V)oxidation by oxygen is also possible
but is very slow in the absence of catalysts. The
1E-12 reaction can be fast enough to be important when
catalyzed by certain trace metals including Fe(II1)and
Mn(I1). Of particular interest is the synergistic cata-
1E-14 lysis afforded by the simultaneous presence of Fe(111)
and Mn(I1).While several investigators have reported
oxidation rate expressions for this pathway, there is far
from uniform agreement among the reported expres-
1 2 3 4 5 6 7 8 sions. Evaluation of the importance of this pathway in
PH actual clouds and fogs is further complicated by the
Figure 4 The pH dependence of the rate of aqueous S(IV) difficulty of making accurate measurements of iron
oxidation by hydrogen peroxide, by ozone and by oxygen (cataly- and manganese speciation in the field.
zed by Fe(1ll) and Mn(l1)). Conditions represented in the figure are In some circumstances, S(IV) oxidation by radicals,
T = 298K, pso, = 1 ppb(v), PO, = ~ O P P ~ PH*o* V , = 1 PPb(v),
Fe(lll)=2.5 x 10-6M,andMn(II)=1.0 x 104M.
including OH, Cl;, and Br;, may also be important,
especially in polluted conditions where the pH is low
and the sulfur dioxide concentration far exceeds the
hydrogen peroxide concentration.
followed by reaction of the peroxymonolsulfurous Field studies have found S(1V)to be present in cloud
acid intermediate with hydrogen ion to yield sulfuric droplets at much higher concentrations than predicted
acid: by Henry's law, even accounting for ionization of
dissolved sulfur dioxide in solution. An important
reason accounting for this apparent discrepancy is the
tendency for sulfite and bisulfite to form complexes
Because the second step is rate-limiting, the intrinsic
with various aldehydes, especially formaldehyde:
rate of oxidation increases with increasing hydrogen
ion concentration. Oxidation of S(1V) is also possible HCHO + HSO;*CH2(0H)SO; [VI]
by organic peroxides, but those reactions are of lesser
importance owing to their lower atmospheric concen-
trations and their lower aqueous solubility. HCHO + SOi-+.CH2(O)SO:- [VII]
At higher pH, or after available hydrogen peroxide
The product of these reactions is hydroxy-
has been consumed, S(1V)oxidation by ozone can be
methanesulfonate (HMS). Formation of HMS is
important. The rate of reaction of S(1V)with ozone in
favored at high pH. Significant concentrations of
aqueous solution can be expressed as
HMS have been measured in high-pH fogs in polluted
environments, including California's San Joaquin
Valley. Formation of HMS is of interest because it
where ko, kl, and k2 are the reaction rate constants for represents an additional sink for sulfur dioxide in
reaction of ozone with H2S03, HS03, and SO:-, high-pH drops and because it is fast enough under
respectively; and 60,al,and a2 represent the fractions those conditions to limit the amount of aqueous phase
of dissolved sulfur dioxide present as H2S03,HSO;, sulfate production. Although S(IV) complexation by
and SO;-, respectively. Because the effective solubility other aldehydes can also occur, these reactions are
of sulfur dioxide increases with pH and because sulfite generally of lesser importance than reaction with
CLOUD CHEMISTRY 457
formaldehyde, owing to the lower solubilities of O H and HO2 can also be formed in solution as a
higher-molecular-weight aldehydes and their lower result of bimolecular reactions and radical intercon-
concentrations in the atmosphere. HMS is stable with version processes. One reaction of particular interest is
respect to oxidation by 0 3 and H202, but may be the oxidation of hydrated formaldehyde (methylene
oxidized by OH. glycol) by O H to produce HO2 and formic acid. The
overall stoichiometry for this multistep reaction is
Oxidant Chemistry
H2C(OH), +OH +0 2
A number of other reactions also occur in cloud drops
involving radicals and other oxidants. Several oxi- + HCOOH + HO2 + H20 [XV]
dants, including H202, 0 3 , HOzradical, and O H
radical, contribute importantly to aqueous-phase Formic acid is commonly observed in clouds and
atmospheric chemistry. Organic peroxides and organ- fogs formed in both pristine and polluted environ-
ic peroxy radicals do not play as important a role ments. Similar reactions are possible for oxidation of
because of their lower atmospheric concentrations hydrated formaldehyde by other radicals, including
and/or lower aqueous solubilities. NO3, SO,, Cl,, and COX. Unlike the radicals
H202,03, HO2, and O H can all be transferred to
discussed above, which can be formed in situ or
cloud or fog drops from the gas phase. H202, H02 and transferred from the gas phase, SO;, Clz, and COY
OH can also be generated photolytically in solution. are formed only in the aqueous phase.
Several photolytic sources exist for O H production, Formic acid and other higher-molecular-weight
including photodissociation of H202 , iron-hydroxo- organic acids are also subject to oxidation by aqueous
complexes, nitrate, and nitrite. O H can also be radicals. Reaction between formic acid and O H
produced by the photo-Fenton reaction O H + HCOOH -+ H20 + COOH [XVI]
Fe(III)L, + hv + Fe(II)L,, [VIII]
COOH + 0 2 + C02 + HO2 [XVII]
Fe(I1) + H202 + H+ -+ Fe(II1) + H20 + O H [IX] represents an efficient source of HO2 and a strong sink
for OH. O H is also capable of oxidizing formate
where L denotes an organic ligand such as oxalate.
anion, so that HO2 formation by formic acid oxida-
Experimental studies confirm that HO2 is formed
tion is not expected to vary strongly with pH.
photochemically upon the illumination of cloud and
In addition to reacting with aldehydes and organic
fog samples, subsequently yielding H202. Superoxide
acids, aqueous-phase radicals are capable of reacting
(0;)is believed to be formed from reaction of Fe(I1)-
with other organic substrates containing abstractable
oxalate complexes
hydrogen atoms. Because we know relatively little
+ + +
[Fe(II)(C204)]+ 0 2 + Fe(I1) 0, 2C02 [XI about the organic speciation of fog and cloud drops, it
is difficult to predict accurately how such reactions
followed by protonation to yield HO2: might impact the lifetimes of aqueous-phase radicals.
Several aqueous radicals are believed to undergo
+
0 2 H++H02 [XI1 reaction with one or more transition metals. In
Superoxide can form peroxide in solution via addition to reaction with iron, reactions are also
possible with manganese, cobalt, and copper. Copper
2 0 2 + 2H+ + H202 0 2 + 1x111 is of particular interest because it reacts quickly
enough with H 0 2 / 0 ? to compensate for its generally
Fe(I1) has also been postulated to react with H02 (or low concentrations in atmospheric waters, while
0: to form H202: manganese and cobalt are not expected to significantly
Fe(I1) + HO2 + H' + Fe(II1) H202 + [XIII] influence aqueous-phase radical concentrations.
Reactions with nitrite may also represent important
Overall, the redox pair of Fe(II)/Fe(III) catalyzes sinks for aqueous radicals including OH. The mech-
degradation of oxalate into C 0 2 and Hz02: anism for the reaction of radical species X with nitrite
is given by
C20:- + + +
2H' + 0 2 hv + 2C02 H202 [XIV]
Laboratory studies of the effect of sunlight illumi-
X + NO, + X- NO2 + [XVIII]
nation on cloud samples suggest these reactions can be Cloud drop scavenging of HO2 from the gas phase
important both in producing H202 and in influencing and HO2 production in situ via the mechanisms
the daytime cloud pH. discussed above have received attention because of
458 CLOUD CHEMISTRY
their potential effects on tropospheric ozone concen- collision and coalescence of large cloud drops leads to
trations. One expects that depletion of HO2 from the formation of rain drops.
gas phase should influence ozone concentrations in the In some environments cloud and fog drops can be
gas phase because of the suppression of the reaction: deposited directly to Earth’s surface. Drops are
removed by inertial impaction, interception, and
sedimentation. The latter process dominates removal
in radiation fogs while the former processes can be
Further, dissociation of HO2 in solution to form quite important for capture of cloud drops by montane
superoxide can promote ozone destruction in the forest canopies. Various studies in the US and in
aqueous phase via Europe have shown that for high-elevation sites with a
high cloud interception frequency, hydrologic input by
0 3 +o, + 0, + 0 2 [XXI direct cloud deposition is measurable but often low
compared with the input by rain and snow. However,
because cloud solute contents are often far higher than
precipitation solute contents, cloud drop deposition
can significantly enhance total wet deposition fluxes of
The overall effect of clouds on tropospheric 0 3 many chemical species from the atmosphere to terres-
concentrations remains a topic of discussion, with trial ecosystems. Likewise, studies of radiation fogs
some authors arguing for a potentially significant reveal that they can represent an important vector for
effect while others suggest it is probably not a major deposition of accumulation-mode aerosol particles in
factor. Differing conclusions stem at least in part from polluted environments.
differences in the reaction mechanisms considered. In certain parts of the world, including coastal
northern California and parts of the Hawaiian islands,
the hydrologic input from cloud drop deposi-
Deposition
tion appears to be an important contributor to
The chemical and microphysical properties of clouds the survival of some plant species. Likewise, cloud
exert a significant influence on the composition of drop deposition has been harnessed in some arid
precipitation and the resulting wet deposition. Precip- regions as a means of supplementing the local water
itation (raindrops and snow crystals) can form in a supply for small towns and villages. Large screens of
variety of ways. Outside the tropics, much precipita- mesh are erected on ridge tops, where nonprecipitat-
tion is produced in mixed-phase (ice-liquid water) ing clouds are frequently intercepted, to capture water
clouds. In this environment precipitation is formed in for domestic use.
the ice phase in the presence of supercooled cloud
drops. Ice crystal growth can occur by water vapor
deposition, by inertial capture of cloud drops (accre-
tion or riming), and by ice crystal aggregation. Because See also
the saturation vapor pressure of water is higher over Aerosols: Physics and Chemistry of Aerosols; Role in
liquid water than over ice at the same temperature, ice Cloud Physics. Biogeochemical Cycles: Sulfur Cycle.
crystals often take up water vapor at the expense of the Cloud Microphysics. Clouds: Classification. Fog.
cloud drops. As the cloud drops evaporate, their solute Tropospheric Chemistry and Composition: Aero-
concentrations increase. Meanwhile, the condensa- sols/Particles; Sulfur Chemistry, Organic.
tion of water vapor on the ice crystal surface tends to
dilute the concentrations of those species present.
Consequently, it is common for precipitation to
Further Reading
possess much lower solute concentrations than are Bator A and Collett JL (1997) Cloud chemistry varies
found in cloud drops. For larger ice crystals and larger with drop size. Journal of Geophysical Research 102:
cloud drops, accretional growth of precipitation 28071-28078.
becomes more important. It has been shown that Collett JL, Prevot ASH, Staehelin J and Waldvogel A (1991)
Physical factors influencing winter precipitation chemis-
when crystals are heavily rimed their composition
try. Environmental Science and Technology 25: 782-789.
closely resembles that of the accreted cloud drops. Collett Jr JL, Bator A, Sherman DE, et a!. (2001) The
Because accretional growth favors inertial capture of chemical composition of fogs and intercepted clouds in
larger cloud drops, any drop-size-dependence of cloud the United States. Atmospheric Research 64: 29-40.
drop composition can influence the precipitation Faust BC (1994) Photochemistry of clouds, fogs and
scavenging efficiencies of individual solute species. aerosols. Environmental Science and Technology 28:
This is also true in warm rain formation, where 217A-222A.
CLOUD MICROPHYSICS 459
Faust BC, Anastasio C, Allen JM and Arakaki T (1993) Munger JW, Collett J, Daube BC and Hoffmann MR
Aqueous-phasephotochemical formation of peroxides in (1989) Carboxylic acids and carbonyl compounds
authentic cloud and fog waters. Science 260: 73-75. in southern California clouds and fogs. Telltas 41B:
Finlayson-Pitts BJ and Pitts JN (2000) Chemistry of the 230-242.
Upper and Lower Atmosphere. San Diego: Academic Pruppacher HR and Klett JD (1997)Microphysics of Clouds
Press. and Precipitation. Dordrecht: KIuwer.
Herrmann H, Ervens B, Jacobi H-W, Wolke R, Nowacki P Ravishankara AR (1997) Heterogeneous and multiphase
and Zellner R (2000)CAPRAM2.3: A chemical aqueous chemistry in the troposphere. Science 276: 1058-1065.
phase radical mechanism for tropospheric chemistry. Seinfeld JH and Pandis SN (1998)Atmospheric Chemistry
Journal of Atmospheric Chemistry 36: 231-284. and Physics. New York: Wiley.
Jacob DJ (2000)Heterogeneous chemistry and tropospheric Weathers KC, Likens GE, Bormann FH, et al.
ozone. Atmospheric Environment 34: 2131-2159. (1988) Cloudwater chemistry from ten sites in North
Junge CE (1963) Air Chemistry and Radioactivity. New America. Environmental Science and Technology 22:
York: Academic Press. 1018-1026.
Levsen K, Behnert S, Mussmann P, Raabe M and Priess B Zellner R and Herrmann H (1995)Free radical chemistry of
(1993) Organic compounds in cloud and rain water. the aqueous atmospheric phase. In: Clark RJH and
International Journal of Environmental Analytical Hester RE (eds)Spectroscopy in Environmental Science,
Chemistry 52: 87-97. pp. 381-451. Chichester: Wiley.
D Lamb, Pennsylvania State University, PA, USA of the updraft speed. This state of disequilibrium is
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. gradually relieved as the vapor condenses out onto
suitable aerosol particles to form the liquid and solid
(ice) particles of the cloud.
A cloud, especially during its early stages of forma-
Introduction tion, often exhibits the properties of a colloidal system,
Cloud microphysics is the branch of the atmospheric a suspension of tiny particles that follow the airflow
sciences concerned with the many particles that make and interact only weakly with one another. Whereas
up a cloud. Relative to the cloud as a whole, the the individual aqueous particles may form, grow and
individual particles are very small and so exist on the subsequently disappear, the system as a whole remains
‘microscale’, that is, over distances from fractions of a microphysically stable for a time and well character-
micrometer to several centimeters. The microscale ized in terms of the number concentrations of liquid
‘structure’ of a cloud, a specification of the number drops and ice particles. The discipline of cloud
concentrations, sizes, shapes, and phases of the vari- microphysics helps us to understand the specific
ous particles, is important to the behavior and lifetime mechanisms needed to break such colloidal stability
of the cloud. The abilities of clouds to produce rain or and to form precipitation.
snow, generate lightning, and alter the radiation
balance of the earth, for instance, stem in large part
from their individual microstructures. Cloud physi-
cists attempt to characterize the diverse microstruc-
Microphysical Descriptions
tures of atmospheric clouds and t o understand the The microstructure of a cloud may be categorized and
phenomena that cause them to change with time. described statistically in a number of ways. Empirical
Clouds typically form in response to changes in descriptions, typically derived from in situ or remote
atmospheric conditions on scales much larger than the measurements of clouds, facilitate communications
particles, indeed, often much larger than the cloud among atmospheric scientists and provide the first
itself. Most commonly, the upward motions of moist glimpses of the physical processes likely to have been
air, driven by synoptic-scale disturbances or convec- operative at time of measurement. The microstructure
tion, cause decreases in the local pressure and temper- may differ substantially from one part of a cloud to
ature, which lead to a lowering of the equilibrium another, and it evolves with time in ways that depend
vapor pressures of the liquid and solid phases of water. on the environmental setting and the physical phe-
‘Excess’ vapor, that amount above the equilibrium nomena that are active. Mathematical and numerical
value, develops in rough proportion to the magnitude models, depending on their purpose, may employ the
Faust BC, Anastasio C, Allen JM and Arakaki T (1993) Munger JW, Collett J, Daube BC and Hoffmann MR
Aqueous-phasephotochemical formation of peroxides in (1989) Carboxylic acids and carbonyl compounds
authentic cloud and fog waters. Science 260: 73-75. in southern California clouds and fogs. Telltas 41B:
Finlayson-Pitts BJ and Pitts JN (2000) Chemistry of the 230-242.
Upper and Lower Atmosphere. San Diego: Academic Pruppacher HR and Klett JD (1997)Microphysics of Clouds
Press. and Precipitation. Dordrecht: KIuwer.
Herrmann H, Ervens B, Jacobi H-W, Wolke R, Nowacki P Ravishankara AR (1997) Heterogeneous and multiphase
and Zellner R (2000)CAPRAM2.3: A chemical aqueous chemistry in the troposphere. Science 276: 1058-1065.
phase radical mechanism for tropospheric chemistry. Seinfeld JH and Pandis SN (1998)Atmospheric Chemistry
Journal of Atmospheric Chemistry 36: 231-284. and Physics. New York: Wiley.
Jacob DJ (2000)Heterogeneous chemistry and tropospheric Weathers KC, Likens GE, Bormann FH, et al.
ozone. Atmospheric Environment 34: 2131-2159. (1988) Cloudwater chemistry from ten sites in North
Junge CE (1963) Air Chemistry and Radioactivity. New America. Environmental Science and Technology 22:
York: Academic Press. 1018-1026.
Levsen K, Behnert S, Mussmann P, Raabe M and Priess B Zellner R and Herrmann H (1995)Free radical chemistry of
(1993) Organic compounds in cloud and rain water. the aqueous atmospheric phase. In: Clark RJH and
International Journal of Environmental Analytical Hester RE (eds)Spectroscopy in Environmental Science,
Chemistry 52: 87-97. pp. 381-451. Chichester: Wiley.
D Lamb, Pennsylvania State University, PA, USA of the updraft speed. This state of disequilibrium is
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. gradually relieved as the vapor condenses out onto
suitable aerosol particles to form the liquid and solid
(ice) particles of the cloud.
A cloud, especially during its early stages of forma-
Introduction tion, often exhibits the properties of a colloidal system,
Cloud microphysics is the branch of the atmospheric a suspension of tiny particles that follow the airflow
sciences concerned with the many particles that make and interact only weakly with one another. Whereas
up a cloud. Relative to the cloud as a whole, the the individual aqueous particles may form, grow and
individual particles are very small and so exist on the subsequently disappear, the system as a whole remains
‘microscale’, that is, over distances from fractions of a microphysically stable for a time and well character-
micrometer to several centimeters. The microscale ized in terms of the number concentrations of liquid
‘structure’ of a cloud, a specification of the number drops and ice particles. The discipline of cloud
concentrations, sizes, shapes, and phases of the vari- microphysics helps us to understand the specific
ous particles, is important to the behavior and lifetime mechanisms needed to break such colloidal stability
of the cloud. The abilities of clouds to produce rain or and to form precipitation.
snow, generate lightning, and alter the radiation
balance of the earth, for instance, stem in large part
from their individual microstructures. Cloud physi-
cists attempt to characterize the diverse microstruc-
Microphysical Descriptions
tures of atmospheric clouds and t o understand the The microstructure of a cloud may be categorized and
phenomena that cause them to change with time. described statistically in a number of ways. Empirical
Clouds typically form in response to changes in descriptions, typically derived from in situ or remote
atmospheric conditions on scales much larger than the measurements of clouds, facilitate communications
particles, indeed, often much larger than the cloud among atmospheric scientists and provide the first
itself. Most commonly, the upward motions of moist glimpses of the physical processes likely to have been
air, driven by synoptic-scale disturbances or convec- operative at time of measurement. The microstructure
tion, cause decreases in the local pressure and temper- may differ substantially from one part of a cloud to
ature, which lead to a lowering of the equilibrium another, and it evolves with time in ways that depend
vapor pressures of the liquid and solid phases of water. on the environmental setting and the physical phe-
‘Excess’ vapor, that amount above the equilibrium nomena that are active. Mathematical and numerical
value, develops in rough proportion to the magnitude models, depending on their purpose, may employ the
460 CLOUD MICROPHYSICS
empirical descriptions directly, or they may calculate

the time evolution of the microstructure if the relevant
processes are included.
Cloud particles vary in both phase and composition.
The phase, whether solid or liquid, is the traditional
descriptor of the aqueous particles, but the nonaque-
ous aerosol particles in the atmosphere are also crucial
to cloud development and evolution. Liquid conden-
sate forms preferentially on the soluble ‘cloud con-
densation nuclei’ (CCN),which often contain sulfates
or nitrates. Ice particles, by contrast, typically form on
‘ice nuclei’ (IN), insoluble aerosol particles that
contain crustal components or biogenic matter having
crystalline structures related to that of ice. The
dominant phase of the aqueous particles forms the
basis for classifying clouds as ‘warm’, when only liquid
drops are present, or ‘cold’, when ice is involved (with
or without liquid drops). The ‘mixed-phase’ region of
a cloud, throughout which both the liquid and solid
(ice)phases of water may be present simultaneously, is
that vertical zone between the melting level (0°C) and
the -40°C isotherm, the practical lower limit for liquid
water to exist in the metastable (i.e., ‘supercooled’)
state. The relative abundance of each phase in a given
cloud depends on the prevailing meteorological con-
Figure 1 Various categories of liquid drops found in clouds. The
ditions and the microphysical processes active indicated drop radii ( R ) are drawn roughly to scale, as are the
throughout the lifecycle of the cloud. arrows representing the terminal fallspeeds ( v ) of the various drop
The sizes and shapes of the aqueous particles play categories. ‘CCN’ represents a ‘cloud condensation nucleus’, a
important roles in cloud development. Whereas ice solution droplet that serves as the initial site of condensation. The
large raindrop is shown distorted to represent the effect of a large
particles can and do appear in a wide variety of shapes,
dynamic pressure on its underside.
all but the largest liquid drops tend to remain spherical
because of surface tension effects. Figure 1depicts the
various categories of liquid drops based on their sizes.
Note that the size of a particle is an important (length along the principal or ‘c’ axis divided by the ‘d
determinant of its terminal fallspeed and hence of axis, the width across the corners of the hexagon) that
its ability to fall against typical updrafts speeds depend systematically on the temperature. As shown
(-1Ocms-’ in stratiform clouds; -10ms-’ in in Figure 2, ‘plates’ ( c / aaxial ratios less than unity) are
convective storms). Drizzle drops represent the tradi- found when the temperature is either between 0°C and
tional transition between the small ‘droplets’ that about -3°C or between about -8 and -22°C. On
remain suspended in the air and the larger drops that the other hand, ‘columns’ ( c / a >1) appear in the
may reach the ground as rain. The shapes of bigger approximate temperature ranges - 3 to - 8°C and less
raindrops tend to become distorted because of the than - 22°C. Deviations from simple hexagonal
large dynamic pressure on the lower side, giving the prisms are common and depend on the excess vapor
rough appearance of a ball of dough pressed lightly density, as suggested by the various symbols in Figure
onto a table top. This flattening of the underside causes 2. In addition to the many single crystals, a number of
the aerodynamic resistance to increase relative to that polycrystalline forms of ice are found in cold clouds.
of spherical drops of equivalent volume, thus limiting For instance, several to hundreds of single crystals may
the fallspeeds of raindrops to little more than about clump together to form ‘aggregates’ (Le., snowflakes),
10m s-’. and supercooled cloud droplets may freeze onto ice
The ice particles in a cloud vary enormously in both particles, giving rise to rimed crystals, graupel, and
size and shape. The first ice to appear in many clouds hail. Individual crystals seldom grow to more than a
tends to be small ( 10 km across) and monocrystalline
N few millimeters across, but hailstones can sometimes
in structure. Single crystals of ice subsequently grow exceed 10 cm in diameter.
into hexagonal prisms (each prism being bounded by The cloud microstructure is best viewed as a multi-
two ‘basal’faces and six ‘prism’faces) with axial ratios dimensional specification of the number concentrations
Figure 2 Schematic representation of the wide variations in the shapes (i.e., habits) of ice crystals found in clouds at the indicated
temperaturesand excess vapor densities (relativeto the ice phase).The excess vapor density of air saturated with respectto liquid water is
shown by the lightly dashed curve. The heavy dashed vertical lines identifythe temperatures of the primary-habittransitions. (Reproduced
with permission from Lamb D (1999) Atmospheric ice. In: Trigg GL (ed.) Encyclopedia of Applied Physics. Weinheim: Wiley-VCH.)
of the various particles in a cloud. In the case of liquid A more general form, the modified gamma distribu-
drops, one need consider only the size of the drops in tion,
addition to the usual dependence of number concen-
tration on the three-dimensional location within the
n ( D ) = CDP exp (-3.0)
cloud and time. The non-spherical ice particles, by attempts to correct this deficiency, although at the
contrast, require some additional specification, such expense of requiring an additional parameter, p. At
as axial ratio, to account for the shapes of the particles. least for the data shown in Figure 3, a log-normal
‘Spectral’ descriptions of the cloud particles tell us in function of the form
effect how many of what kinds and sizes of particles
are present at given locations within a cloud. n ( D )= NT exp (- In2 ( D / D , ) / 2 ln2 D) [3]
Often, it is useful to characterize the spectra in v%D In D
mathematical terms to minimize the number of vari-
ables needed to represent the microstructure. The size works well with appropriate choices of the three
distributions of raindrops from convective storms, for parameters, N T ,o,and D,. It is important to recognize
example, can be described by analytical functions that that all size spectra simply describe the cloud micro-
have been fitted to observational data by specifying structure without regard to the mechanisms that
two or three parameters, as shown in Figure 3. An produced it.
exponential function, one subclass of which is referred
to as a ‘Marshall-Palmer’ distribution, has the form
Warm-Cloud Microphysics
n ( D ) = C exp (-AD) The liquid drops in ‘warm’ clouds evolve spectrally via
sets of microphysical processes that interact in com-
where C and 3, are the parameters fitted to the plicated ways with the larger-scale cloud environment.
measured number concentration n ( D )of drops within The process of condensation, for instance, starts
a unit size interval about diameter D. Exponential during upward motions of the moist air and continues
distributions are used frequently because only two even as the drops interact with themselves and grow
parameters need to be specified. However, as the into raindrops. As the cloud updraft entrains dry
dashed line in Figure 3 shows, exponential distribu- environmental air, ceases or possibly reverses, evapo-
tions often over estimate the number of smaller drops. ration may dominate for a time and change the
8 \ I I I I equilibrium saturation ratio SK as a function of the

\
\
\ Rain rate: 39 mm h-’ droplet radius r:
\
\
7
\
\
SK = a, exp (A/Y) [41
\
\
\
\
Here, the first factor on the right-hand side of eqn [4]
\ describes the vapor pressure-lowering effect of the
6 \
\ solute in terms of the water activity a, = 1 - ix,,
where x , is the mole fraction of nonvolatile solute that
h
effectively dissociates into i molecular or ionic com-
7
ponents. The second factor in eqn [4] accounts for the
‘E
E vapor pressure-raising effect of droplet curvature, in
m
E
which A = 2cqv/(nwRT)is a function of the physi-
Y
cally relevant variables, the liquid-vapor surface free
.-5 4
4-
_m
energy CLV( = 7 2 m J m P 2 ) , the liquid-water density
+
C n, (= 5 . 5 ~ 1 m 0 ~ l m - ~ the
) , universal gas constant
a,
0
C R ( = 8.31 J m ~ l - ~ K - l ) and
, the temperature T .
8 3
Y
Because each droplet grows by the simple addition of
-C water, its total solute content m, remains constant.
Equation [4] is thus most conveniently expressed in
2 terms of the supersaturation needed to maintain
equilibrium with the solution droplet:
where B = 3/(4m,) and approximations suitable for

0 initial cloud formation have been made.
The properties of the aerosol involved in cloud
Diameter (mm) formation are often depicted graphically. Equation [5]
Figure 3 The size distribution of rain drops from a convective can be seen to represent a family of ‘Kohler’ curves of
storm. Circles connected by light dotted lines: field data based on a constant solute content, as shown in perspective form
one-minute sample from a 20-channel disdrometer. Dashed line: in Figure 4. The competing effects of solute and
datafittedwith an exponential function (eqn [l]).Dot-dashed curve: curvature yield a clear maximum in the equilibrium
data fitted with a modified gamma function (eqn [2]). Solid curve:
Data fitted with a log-normal function (eqn [3]). (Adapted with
saturation ratio, which must be overcome by the
permission from Feingold G and Levin 2 (1986) The lognormal fit to ambient supersaturation before the particle can ‘acti-
raindrop spectra from frontal convective clouds in Israel. Journalof vate’ and grow spontaneously as a cloud droplet. The
Climate and Applied Meteorology 25: 1346-1 363.) larger the salt particle, the smaller is the ‘critical
supersaturation’ that needs to be exceeded. Note that
microstructure in important ways. The ever-changing

population of drops often influences the very atmos-
pheric motions that spawned the drops in the first
place.
Droplet Formation
The many small droplets in a cloud form initially as
excess vapor condenses onto the larger, more soluble
aerosol particles (i.e., onto the CCN).The solute in the
CCN lowers the equilibrium vapor pressure of the
liquid droplets through molecular-scale effects, where-
as the droplet curvature increases it. Figure 4 Perspective view of Kohler curves expanded along the
These two opposing effects of solute and curvature solute axis. Each curve is valid for a constant mass of NaCI, the
are typically combined in Kohler theory to give the nonvolatile solute in this case.
the critical supersaturation is especially pronounced

for particles with the smallest solute contents, so small
particles may never experience sufficiently high ambi-
ent supersaturations to become activated. Such par-
ticles would remain as submicrometer haze droplets
interstitially within the cloud.
Growth by Condensation
Individual cloud droplets that are actively growing in

an updraft act in effect as tiny sinks of water vapor.
During net condensation, the concentration of vapor
immediately over each droplet surface (within a few
mean free paths of the air molecules) is reduced
relative to the average vapor concentration far from
the droplet. The radial gradients of vapor concentra-
tion thus established give rise to a net flux of vapor
Figure 5 Results from numerical computations of droplet growth
molecules toward the drop by the process of molecular in an air parcel ascending adiabatically at the rate of 15 cm s -’.
diffusion. Even though the water molecules must also The total concentration of particles in the population is 5 0 ~ m - ~ .
be transported across the liquid-vapor interface, it is Solid curves: droplet radii (lower scale) at various heights above
the vapor diffusion step that tends to limit the mass cloud base for droplets each containing the indicated number of
moles of nonvolatile solute. Dashed curve: the ambient supersat-
transport under most cloud conditions. Nevertheless,
uration (upper scale). (Based on calculations of Mordy (1959);
the change of phase from vapor to liquid results in a figure reproduced with permission from Lamb D (2000) Rain
slight warming of the droplet owing to the added production in convective systems. In Doswell CA. (ed.) Severe
enthalpy of condensation, energy that must be con- Local Convection. Boston: American Meteorological Society.)
ducted through the air and away from the droplet. This
energy consequence of condensation raises the equi-
librium vapor pressure of the liquid and imposes an
additional limitation to the growth rate.
The theory that simultaneously accounts for the maritime environment. The supersaturation (dashed
exchanges of vapor and energy between a growing curve) builds up until vapor is removed at a rate
droplet and the surrounding air was first developed by comparable to the rate that excess vapor is generated
Maxwell in the nineteenth century. The resulting due to adiabatic ascent. Once the maximum in the
expression for the linear growth rate is ambient supersaturation is reached, no new particles
can be activated, so the initial number concentration
dr 1 of cloud droplets is established low in the cloud. The
- = G(s - SK) - close packing of the curves on the right-hand side of
dt r
Figure 5 indicates that the droplets tend to bunch
where G is a growth parameter that varies slowly with together in radius. This narrowing of the drop spec-
the temperature and pressure. Note that a droplet trum is an inherent property of adiabatic condensation
grows only to the extent that the ambient super- and poses a hindrance to the formation of precipita-
saturation, s, exceeds the equilibrium value, s ~ .tion.
As r becomes large, SK + 0 and the growth rate
drldt oc llr, indicating that the droplets grow rela-
Collisional Interactions
tively more slowly as they become bigger. Calculations
based on eqn [6] show that individual droplets Individual pairs of cloud drops occasionally collide
experiencing a supersaturation of 1% require hun- with one another. If two drops ‘coalesce’ during a
dreds of seconds to grow to radii much beyond 10 pm. particular collision, a single, larger drop replaces the
A population of growing cloud droplets derives its two parent drops in the cloud. Repeated collision-
water from a common supply, namely the vapor coalescence events lead eventually to large drops that
initially carried with the rising air parcel. Competition fall rapidly and become raindrops.
for the available vapor among all the droplets sets up a The growth of drops through collisional interac-
strong interplay between the condensation kinetics tions may be quantified by consideration of the
and the vapor field. Results from numerical computa- separate probabilities for collision and for coales-
tions of droplet growth within an adiabatic parcel are cence. Most commonly, collisions result when a lar-
shown in Figure 5 for the case of a relatively clean ger drop (the collector drop) overtakes a smaller drop
(the collected drop) during its fall through the air. a new mode in the drop size distribution emerges, as
However, not all drops in the geometrical path of the shown by the maxima toward the right-hand side of
collector experience collisions, for the simple reason Figure 6. Once the drops in this large-drop mode
that the air deviating around the collector drop exceed a few hundred micrometers, they grow rapidly
‘pushes’the smallest drops out of the way. The fraction in mass at a more or less continuous rate given by
of drops in the path of the collector that do collide with
it is the ‘collision efficiency’, E , a complicated function
[71
of both the collector- and collected-drop sizes. The
maximum collision efficiency can approach unity
when the collector-drop radii are greater than about where K(r1, r,) = n(r1 + T ~ ) ~ E ,-( vV, )~ is the collec-
tion kernel and OL is the liquid water content (massper
40 pm and the collected-drop radii exceed about 5 pm.
unit volume of cloudy air).The collection kernel is best
However, for collector radii less than about 20 pm, the
collision efficiency becomes very small. The ‘coales- viewed as the effective volume of cloudy air (contain-
cence efficiency’, E , the fraction of drop-drop colli- ing small droplets of radius rs, each falling at rate v,)
sions actually resulting in the formation of a larger that is swept out in unit time by the collector drop
drop, is often less than unity for larger collected drops of radius rl having fallspeed vl. In this continu-
ous-growth regime (in which rs << YI, v, << V I , V I 3: YI,
because of drop distortion and the trapping of air at
the point of collision. The ‘collection efficiency’ ( E,), and E , = l), one finds to first approximation
the product of the collision and coalescence efficien- dm/dt 3: r: K m. Thus, once the collision-coales-
cies, E, = EE, typically reaches a weak maximum at cence process gets started, the large drops increase in
intermediate collected-drop radii. The collection effi- mass (and size) exponentially with time, until the
ciency provides an overall indication of the effective- supply of cloud droplets is exhausted or the drops
ness of drop growth by collision-coalescence. rupture.
Collision-coalescence becomes a powerful mecha-
nism for generating raindrops under appropriate Cold-Cloud Microphysics
microphysical conditions. Because of the de facto
thresholds that exist on both the collected and Clouds become ‘cold’ once ice particles form and
collector drop sizes, collision-coalescence tends to become active players in the cloud microphysics.
begin in the tail of the cloud drop size distribution, as Whereas ice particles are necessary components of
seen in Figure 6. Initially, only a tiny fraction of the cold clouds, the liquid drops are nevertheless often
bigger droplets will collide and coalescence with present and important to the evolution of the cloud
neighboring droplets, yielding slightly larger droplets microstructure. The mixed-phase zone of a cloud,
that then have enhanced probabilities of collecting where the ice particles and liquid drops interact, is
additional droplets. The growth process accelerates as microphysically the most active portion of a cloud.
the collection efficiencies increase and other drops join ‘Glaciation’, the transformation of a cloud from
this favored subset of the drop population. Eventually, supercooled liquid drops to ice particles, is compli-
cated by the diversity of the interactions that can
take place.
Ice Formation
Ice can form once the liquid drops have supercooled by
at least 5”C,although the supercooled state can persist
in some clouds to temperatures as low as -40°C. The
first ice particles in a supercooled cloud appear most
commonly when the temperature is between -10 and
-15°C following ‘primary nucleation’, a process by
which submicrometer, insoluble aerosol particles cat-
alyze ice formation by acting as molecular templates
for the crystal lattice. Such primary ice particles may
form directly from the vapor phase (via ‘deposition
Figure 6 Evolution of the size distribution of drops due to nucleation’), but more commonly they arise from the
collision+oalescence. (Based on computations by Berry and
Reinhardt (1974); figure reproduced with permission from Lamb D
freezing of supercooled cloud droplets (via ‘freezing
(2000) Rain production in convective systems. In Doswell CA. (ed.) nucleation’). The freezing of droplets at relatively high
Severe Local Convection. Boston: American Meteorological temperatures (greater than about -18°C) tends to
Society.) yield single crystals that grow subsequently into
crystallographically aligned double plates. At lower set of secondary habit features superimposed on
temperatures, however, the probability of forming the primary habit. Good physical reasons exist why
multiple crystals within a single droplet becomes large, one seldom if ever finds any two ice crystals alike in
giving rise to polycrystalline forms, such as bullet nature.
rosettes.
Some ice particles form because of the prior exist- Riming and Aggregation
ence of other ice particles. Such ‘secondary’ ice
particles arise via several mechanisms, none of which Riming and aggregation are both processes involving
are fully understood. In some situations, crystals may collection. As in warm clouds, collisions between
‘fragment’, such as when the delicate arms of dendrites particles must first occur, then the colliding particles
break off, thereby increasing the ice particle number must stick together to form a combined particle. In the
concentration. When conditions are just right, as when case of riming, an ice particle collects supercooled
the temperature is between about -3 and -8”C, and droplets that freeze on contact with the ice surface. By
‘graupel’ particles are actively growing by riming (i.e., contrast, aggregation involves the collisional interac-
accreting supercooled cloud water), tiny splinters of tion of two ice particles, with no change of phase.
ice may be released that subsequently grow into Aggregation is a complicated process, in part because
columnar ice crystals. The self-breeding, or ‘multipli- snowflakes fall erratically and because two solid
cation’, of ice seems to be important in clouds that particles may simply bounce apart after colliding.
glaciate rapidly. Snow crystals are most likely to stick together at
temperatures within a few degrees of the melting point
(because of sintering) and when the arms of dendrites
Growth by Vapor Deposition can interlock.
Individual ice particles grow initially by the deposition Growth via the riming process occurs by stages that
of vapor onto their surfaces. As with the growth depend on particle size and the rate at which super-
of a cloud droplet by condensation, water molecules cooled droplets are accreted. Initially, during
must first diffuse to the particle surface from the the ‘crystal stage’, the rate of accretion is slow, the
supersaturated vapor field surrounding the particle. collision efficiency becoming appreciable only once
However, the transport of molecules across the the vapor-grown crystal attains an a-axis dimension
vapor-solid interface cannot be ignored, if for no of about 150ym for plates, 25ym for columns.
reason other than the need to account for facets The crystal becomes lightly to moderately rimed,
and the nonspherical shapes of the ice particles. but the crystal morphology remains identifiable.
The molecular-scale processes on the crystal surface The growth is termed ‘dry’ because each droplet
involve migration of molecules across the surface, freezes rapidly on the spot of impingement. The
from the point of first contact to a step that may ‘graupel stage’ begins once the crystal identity
be a micrometer or more away. As the step gathers becomes obscured by the shroud of ‘dry’ rime ice on
adsorbed molecules, it propagates across the surface, the particle. During this and the subsequent ‘hail
contributing an additional layer (of thickness equal stage’, the ice particle grows in mass at rates described
to the step height) of molecules to the lattice. The reasonably well by eqn [7],with suitable adjustments
rate of advancement of the crystal face is determined in the parameters. The hail stage is distinguished from
largely by the frequency with which the steps the graupel stage by the formation of one or more
are generated, a factor that depends on whether the layers of clear ice, which results when the rate of
steps originate from two-dimensional layer nucleation accretion exceeds the ability of the particle to dissipate
or from the emergence of screw dislocations on the the enthalpy added by the freezing of the supercooled
surface. water. Such ‘wet’ growth occurs when the surface
The aspect ratio of an ice crystal reflects the relative temperature rises to 0°C and the accreted liquid
rates of growth of the basal and prism faces. For spreads across the surface before freezing. Hailstones
reasons that remain largely unknown, the linear represent one extreme to which the microstructure of
growth rates vary with the temperature in complicated clouds can evolve.
ways, giving rise to the observed alternation of
primary habit between plates and columns with
temperature (as depicted in Figure 2). At relatively
Precipitation
large supersaturations with respect of ice, the vapor Precipitation, whether as rain, snow, sleet, or
gradients in the vicinity of a given crystal face become hail, generally results once the aqueous particles in a
important, leading to the bunching of steps, the cloud have grown sufficiently large to fall against the
‘hollowing’ of the face near its center, and a myriad local updraft. In the case of a stratiform cloud, one
characterized by rather weak and uniform updrafts clean, maritime environments or during turbulent
over a broad area, the precipitating particles may mixing), alternatively from ‘giant nuclei’ in dusty
simply fall out through the base of the cloud, in the regions. Once the warm-rain mechanism is estab-
process depleting condensate from the cloud and lished, the rain drops grow rapidly by sweeping out the
depositing it on the ground. On the other hand, in smaller cloud droplets until they themselves become
convective storms, in which large local updraft speeds unstable and break into fragments during their fall to
can aerodynamically support big particles, the precip- earth.
itation itself may influence the motions of air though In the cold parts of a cloud, the colloidal stability of
the cloud. The large mass associated with the precip- the cloud is broken once the ice phase has been
itation commonly initiates a downdraft along the edge nucleated in the presence of supercooled droplets. This
of the updraft, causing the microphysical and dynam- ‘ice crystal’ mechanism, often termed the Bergeron
ical aspects of cloud evolution to become intertwined process, arises from the inherent difference in the
in complicated ways. As an aid to the discussion below, equilibrium vapor pressures of liquid and solid water.
Figure 7 offers a summary of the various processes The relatively low vapor pressure of ice compared with
operating in the ‘warm’ and ‘cold’ parts of a repre- that of the droplets at any given temperature gives the
sentative convective cloud during precipitation ice crystals a growth advantage by causing water
formation. vapor to transfer (via diffusion) from the many cloud
In the warm parts of clouds, large drops can emerge droplets to the fewer ice crystals. The process proceeds
out of the stable population of cloud droplets only rapidly, especially in the temperature range between
through collisional interactions. Condensational - 12 and - 15”C,permitting the ice crystals to attain
growth alone is too slow, but it must be recognized sizes sufficient to initiate the other cold-cloud growth
as a necessary process, for the numerous cloud mechanisms, such as aggregation and riming. These
droplets serve as the feedstock for the growth of the large ice particles eventually fall into the warm part of
larger drops. The collision-coalescence process be- the cloud, where they may melt and join the popula-
comes an effective mechanism for breaking the tion of raindrops formed by the collision-coalescence
colloidal stability of the cloud once the threshold size process. The ice process can be an effective initiator of
(-2.5 pm in diameter) for collection has been over- precipitation in both stratiform clouds and summer
come. The needed ‘coalescence embryos’ can arise thundershowers.
from the droplet population itself (most commonly in The efficiencies with which clouds develop
precipitation depend partly on the types of micro-
physical processes that are active and partly on the
environmental settings in which the clouds form.
For instance, the relative ease with which ‘maritime’
clouds release precipitation compared with ‘continen-
tal’ clouds most likely stems from the differences in
aerosol abundance found in the different airmasses.
The relative absence of active sources of aerosol
particles over the open oceans allows the CCN
concentrations to become low, which in turn means
that those droplets that do form in maritime
clouds tend to be larger on average than those in
continental clouds. The collision-coalescence process
thus gets started early in the life cycle of maritime
clouds, providing such clouds with a decisive
microphysical mechanism for developing precipita-
tion. Within a given climatic regime, storm organiza-
tion on the mesoscale seems to be an important
contributor to precipitation efficiency. Low magnitu-
des of vertical wind shear at the time of cloud
formation tend to favor vertically erect storms
with high precipitation efficiencies, presumably
because the incipient precipitation particles can then
Figure 7 Summary depiction of the microphysical processes fall directly through the condensate-rich inflow of the
operating during the formation of precipitation in a deep convective storm. At the same time, however, such systems tend to
cloud. be short-lived and yield relatively small total amounts
CLOUDS I Classification 467
of precipitation. Storms that form in environments in Further Reading

which the wind varies modestly with height in both Lamb D (1999) Atmospheric ice. In: Trigg GL (ed.) Ency-
speed and direction last longer and yield the most clopedia of Applied Physics, pp. 3-25. Weinheim:
precipitation, for then synergism ensues between Wiley-VCH.
the dynamical time scales of the storm and the time Lamb D (2000) Rain production in convective systems. In:
scales for the microphysical processes to operate Doswell CA (ed.) Severe Convective Storms. Meteor.
effectively. Mongr., 28, no. 50. Boston: American Meteorological
Society, pp. 299-321.
Pruppacher HR and Klett JD (1997)Microphysics of Clouds
and Precipitation, 2nd edn. Dordrecht: Kluwer.
See also Rogers RR and Yau MK (1989) A Short Course in Cloud
Cloud Chemistry. Clouds: Classification; Climatology; Physics, 3rd edn. Oxford: Pergamon Press.
Measurement Techniques In Situ. Convective Cloud Young KC (1993) Microphysical Processes in Clouds. New
Systems: Modelling. York: Oxford University Press.
Contents
Classification
Climatology
Cumuliform
Measurement Techniques In Situ
What is a Cloud?
Classification As defined by the World Meteorological Organization,
a cloud is an aggregate of minute, suspended particles
A L Rangno, University of Washington, Seattle, WA, of water or ice, or both, that are in sufficient
USA concentrations to be visible: a collection of ‘hydrome-
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. teors’, a term that also includes in some cases, due to
the distance of the observer, the precipitation particles
that fall from them. Today the term ‘cloud’ also
Introduction
includes those clouds that are nearly invisible to the
The classification of clouds and reports by observers of human eye but are readily detectable in satellite
their coverage have taken on considerable value in thermal imagery.
recent years. Those reports now provide a quite long- Clouds are tenuous and transitory; no single cloud
term record in which changes in cloud coverage and exists for more than a few hours, and most small
type may be discerned that might not be detectable in clouds in the lower atmosphere exist for only a few
the relatively short record of satellite data. minutes. In precise numbers, the demarcation between
The classification system, was, of course, formulat- a cloud and clear air is even hard to define: how many
ed and revised on the basis of visual attributes, without cloud drops per liter constitute a cloud? When are ice
the benefit of in-cloud measurements. Airborne meas- crystals and snow termed ‘clouds’ rather than precip-
urements would have delineated the internal compo- itation? When are drops too large to be considered
sition of clouds that appear somewhat similar and ‘cloud’ drops, but rather raindrops? These questions
have common labels but are very different in compo- are difficult for scientists to answer in unanimity
sition (e.g., stratus and altostratus). In this article, because the difference between cloud particles and
airborne measurements of cloud composition will be precipitation particles, for example, is not black and
used where possible to augment the composition of white; rather they represent a continuum of fallspeeds.
clouds. For some scientists, a 50 pm diameter drop represents
of precipitation. Storms that form in environments in Further Reading

which the wind varies modestly with height in both Lamb D (1999) Atmospheric ice. In: Trigg GL (ed.) Ency-
speed and direction last longer and yield the most clopedia of Applied Physics, pp. 3-25. Weinheim:
precipitation, for then synergism ensues between Wiley-VCH.
the dynamical time scales of the storm and the time Lamb D (2000) Rain production in convective systems. In:
scales for the microphysical processes to operate Doswell CA (ed.) Severe Convective Storms. Meteor.
effectively. Mongr., 28, no. 50. Boston: American Meteorological
Society, pp. 299-321.
Pruppacher HR and Klett JD (1997)Microphysics of Clouds
and Precipitation, 2nd edn. Dordrecht: Kluwer.
See also Rogers RR and Yau MK (1989) A Short Course in Cloud
Cloud Chemistry. Clouds: Classification; Climatology; Physics, 3rd edn. Oxford: Pergamon Press.
Measurement Techniques In Situ. Convective Cloud Young KC (1993) Microphysical Processes in Clouds. New
Systems: Modelling. York: Oxford University Press.
Contents
Classification
Climatology
Cumuliform
Measurement Techniques In Situ
What is a Cloud?
Classification As defined by the World Meteorological Organization,
a cloud is an aggregate of minute, suspended particles
A L Rangno, University of Washington, Seattle, WA, of water or ice, or both, that are in sufficient
USA concentrations to be visible: a collection of ‘hydrome-
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. teors’, a term that also includes in some cases, due to
the distance of the observer, the precipitation particles
that fall from them. Today the term ‘cloud’ also
Introduction
includes those clouds that are nearly invisible to the
The classification of clouds and reports by observers of human eye but are readily detectable in satellite
their coverage have taken on considerable value in thermal imagery.
recent years. Those reports now provide a quite long- Clouds are tenuous and transitory; no single cloud
term record in which changes in cloud coverage and exists for more than a few hours, and most small
type may be discerned that might not be detectable in clouds in the lower atmosphere exist for only a few
the relatively short record of satellite data. minutes. In precise numbers, the demarcation between
The classification system, was, of course, formulat- a cloud and clear air is even hard to define: how many
ed and revised on the basis of visual attributes, without cloud drops per liter constitute a cloud? When are ice
the benefit of in-cloud measurements. Airborne meas- crystals and snow termed ‘clouds’ rather than precip-
urements would have delineated the internal compo- itation? When are drops too large to be considered
sition of clouds that appear somewhat similar and ‘cloud’ drops, but rather raindrops? These questions
have common labels but are very different in compo- are difficult for scientists to answer in unanimity
sition (e.g., stratus and altostratus). In this article, because the difference between cloud particles and
airborne measurements of cloud composition will be precipitation particles, for example, is not black and
used where possible to augment the composition of white; rather they represent a continuum of fallspeeds.
clouds. For some scientists, a 50 pm diameter drop represents
468 CLOUDS I Classification
~~~
a ‘drizzle’ drop because it must have formed from the concept of low, middle, and high cloud groupings
collisions with other drops with coalescence; but for into which Howard’s cloud types were more system-
others it may be termed a ‘cloud’ drop because it falls atically placed by H. Hildebrandsson, Director of the
too slowly. Also, the farther an observer is from fallingUppsala Observatory, and Abercromby in their 1887
precipitation, the more it appears to be a ‘cloud’ due toclassification. Also by then cumulus and cumulonim-
perspective. For example, many of the higher ‘clouds’ bus clouds became a fourth distinct category repre-
above us, such as cirrus and altostratus are composed senting ‘convective’ clouds, those with appreciable
mainly of ice crystals and snowflakes that are settling upcurrents. Howard’s now modified classification
toward the Earth, and would not be considered a system was reexamined at the International Meteor-
‘cloud’ by an observer inside them, but rather a light ological Conference at Munich in 1891, followed by
snowfall. the publication of a color cloud atlas in 1896. At this
point, the definitions of clouds had reached nearly
their modern forms. Additional international com-
Origin of the Present-Day Cloud mittees made minor modifications to the system in
1926 that were realized with the publication of the
Classification System 1932 International Cloud Atlas. For example, a
The process of classifying objectively the many shapes precipitating stratiform cloud became nimbostratus
and sizes of something as ephemeral as a cloud was rather than just nimbus. Little change has been made
first accomplished at about the beginning of the since that time. The most comprehensive version of the
nineteenth century by English chemist Luke Howard, classification system was published by the World
in 1803, and French naturalist Jean Baptiste Lamarck Meteorological Organization in 1956, and pictorially
(1802).Both published systems of cloud classification. updated in 1989 with the introduction of photographs
However, because Howard used Latin descriptors of from more disparate locations than in the previous
the type that scientists were already using in other Atlas.
fields, his descriptions appeared to resemble much of There are ten types or ‘genera’ into which clouds are
what people saw, and because he published his results classified: cirrus, cirrostratus, cirrocumulus, altostra-
in a relatively well-read journal, Tilloch’s Philosoph- tus, altocumulus, nimbostratus, stratocumulus, stra-
ical Magazine, Howard’s system became accepted and tus, cumulus, and cumulonimbus. Table 1 is a partial
was reproduced in books and encyclopedias during the list of the nomenclature involving the most commonly
ensuing years. seen species and varieties of these clouds. Figures 1 to
Howard observed, as had Lamarck, that there were 10 illustrate these main forms.
three basic cloud regimes. There were fibrous and The three cloud height regimes are ‘high’ clouds,
wispy clouds that he called ‘cirrus’ (in Latin meaning those with bases generally above 7 km above ground
‘hair’); sheetlike laminar clouds that covered much or level (AGL); ‘middle-level’ clouds, those with bases
all of the sky that he referred to as ‘stratus’ (meaning between 2 and about 7km AGL; and ‘low’ clouds,
flat); and clouds that were less pervasive but had a those with bases at or below 2 km AGL. ‘About’is used
strong vertical architecture that he called ‘cumulus’ because clouds with certain visual attributes that make
(meaning ‘heaped up’). Howard used an additional them, for example, a ‘middle-level’cloud may actually
Latin term ‘nimbus’ (Latin for cloud) to refer to a have a base that is above 7 km. Similarly,in wintertime
cloud or system of clouds from which precipitation or in the arctic, ‘high’ clouds with cirriform attributes
fell. Today, nimbus itself is not a cloud, but rather a (fibrous and wispy) may be found at heights below
prefix or suffix to denote the two main precipitating 7 km. Also, some clouds that are still considered low
clouds, nimbostratus and cumulonimbus. The ques- clouds (e.g., cumulus clouds) can have bases that are a
tion of clouds and their types generated such enthu- kilometer or more above the general ‘low cloud’ upper
siasm among naturalists in the nineteenth century that base limit of 2 km AGL. Hence, these cloud base height
an ardent observer, Ralph Abercromby, also a member boundaries should be considered somewhat flexible.
of the British Royal Meteorological Society, took two Note that what is classified as an altocumulus layer
voyages around the world to make sure that no cloud when seen from sea level will be termed a stratocu-
type had been overlooked! mulus layer when seen by an observer at the top of a
Later, the emerging idea that clouds tended to high mountain.
occupy just two or three levels in the atmosphere was The classification of clouds is also dependent on
verified by measurements in Uppsala, as well as at sites their composition, because the composition of a cloud
in Germany and in the United States in the 188Os, - all liquid, all ice, or a mixture of both - determines its
using theodolites and photogrammetry to measure visual attributes on which the classifications are
cloud heights. These measurements eventually led to founded (e.g., luminance, texture, color, opacity, and
CLOUDS / Classification 469
Table 1 The ten cloud types and their most common species and varieties (after the World MeteorologicalOrganization.) The letters in
parentheses denote accepted abbreviations
Genera Species Varieties
Cirrus (Ci) uncinus, fibratus, spissatus, castellanus intortus, radiatus, vertebratus

Cirrostratus (Cs) nebulosus, fibratus
Cirrocumulus (Cc) castellanus, floccus, lenticularis undulatus
Altocumulus (Ac) castellanus, floccus, lenticularis translucidus, opacus, undulatus, perlucidus
Altostratus (As) none translucidus, opacus
Nimbostratus (Ns) none none
Stratocumulus (Sc) castellanus, lenticularis perlucidus, translucidus opacus
Stratus (St) fractus, nebulosus
Cumulonimbus (Cb) calvus, capillatus,
Cumulus (Cu) fractus, humilis, mediocris, congestus
the level of detail of the cloud elements). For example, from it concentrated in a narrow shaft, suggesting
an altocumulus cloud cannot contain too many ice heaped cloud tops above, or is the precipitation
crystals and still be recognizable as an altocumulus widespread with little gradation, a characteristic that
cloud. It must always be composed largely of water suggests uniform cloud tops? Answering these ques-
drops to retain its sharp-edged compact appearance. tions will allow the best categorization of clouds into
Thus, it cannot be too high and cold. On the other their ten basic types.
hand, wispy trails of ice crystals comprising cirrus
clouds cannot be too low (and thus, warm). Therefore,
having the ability to assess the composition of clouds High Clouds
(i*e*>ice liquid water) visually can in the Cirrus, cirrostratus, and cirrocumulus clouds (Figures
determination of a cloud’s height.
1, 2 , and 3, respectively) comprise ‘high’ clouds. By
There are other important attributes for identifying
WMO definition, they are not dense enough to
a cloud: How much of the sky does it cover? Does it
produce shading except when the sun is near the
obscure the Sun’s disk? If the Sun’s position is visible, is
horizon, with the single exception of a thick patchy
its disk sharply defined or diffuse? Does the cloud
cirrus variety called cirrus spissatus in which gray
display a particular pattern such as small cloud
shading is allowable. Cirrus and cirrostratus clouds are
elements, rows, billows, or undulations? Is rain or
composed of ice crystals with, perhaps, a few momen-
snow falling from it? If so, is the rain or snow falling
tary exceptions just after forming and when the
temperature is higher than -40°C. In this case,
droplets may be present briefly at the instant of
formation. The ‘bases’ of cirrus and cirrostratus
clouds, composed of generally low concentrations of
Figure 1 Cirrus. Figure 2 Cirrostratus.

470 CLOUDS / Classification
Figure 5 Altostratus.
Figure 3 Cirrocumulus.
ice crystals that are about to evaporate, are usually than are the droplet concentrations in liquid water
colder than - 20°C. The coldest cirriform clouds (Le., clouds. In contrast, droplet clouds look hard and
cirrus and cirrostratus) can be - 80°C or lower in deep sharp-edged with the details of the tiniest elements
storms with high cloud tops ( >15 km above sea level), clearly visible. The long filaments often comprising
such as in anvils associated with thunderstorm outflow. cirriform clouds are due to larger ice crystals falling
Because of their icy composition, cirrus and cirro- out in narrow, sloping shafts because of wind shear.
stratus clouds are fibrous, wispy, and diffuse. This Due to the slow settling of ice crystals soon after they
‘wispy’ and diffuse attribute arises because the ice form, mature cirrus and cirrostratus clouds are often
crystals that comprise them are overall in much lower I k m or more thick, though the Sun may not be
concentrations (often an order of magnitude or more) appreciably dimmed.
Figure 4 Altocumulus. Figure 6 Nimbostratus.

CLOUDS / Classification 471
Figure 7 Stratus.
Figure 9 Cumulus.
Cirrus and cirrostratus clouds often produce haloes are at the same altitudes and temperatures as cirriform
when viewed from the ground, whereas thicker mainly clouds, haloes are often observed near the tops of
ice clouds such as altostratus clouds (see below) altostratus clouds when aircraft ascend through them.
cannot. This is because the cirriform clouds consist In this upper cloud region the ice crystals are virtually
of small, hexagonal crystals such as thick plates, short the same as those in halo-producing cirriform clouds.
solid columns, that refract the Sun’s light as it passes Parhelia (‘sun dogs’) can be observed in thin portions
through them. Altostratus clouds, in contrast, are of altostratus when the Sun is low in the sky. Parhelia
much deeper and are therefore composed of much occur as platelike ice crystals fall with their faces
larger, complicated ice crystals and snowflakes that downward.
do not permit simple refraction of the Sun’s light. Cirrocumulus clouds are patchy, finely granulated,
However, since altostratus clouds often have tops that and, because of a definition that allows no shading, are
Figure 8 Stratocumulus. Figure 10 Cumulonimbus.

472 CLOUDS I Classification
Figure 11 Altocumulus castellanus. Figure 13 Cumulonimbus calvus.
very thin (less than 2 0 0 m thick). The largest of the irization (also called iridescence) is produced due to
visible cloud elements can be no larger than the width the diffraction of sunlight by the cloud’s tiny (<10 pm
of a finger held skyward when observed from the diameter) droplets.
ground; an apparent size larger than this classifies the
cloud as an altocumulus. Cirrocumulus clouds are
Middle-Level Clouds
composed mostly or completely of water droplets.
(Some especially high and cold cirrocumulus clouds Altocumulus, altostratus, and nimbostratus clouds
are only momentarily comprised of water drops before (Figures 4, 5, and 6 , respectively) are considered
evolving into cirrus clouds composed entirely of ice ‘middle-level’ clouds because their bases are located
crystals.) The liquid phase of these clouds can usually between about 2 and 7 k m AGL (see discussion
be deduced when they are near the Sun; a corona or concerning the variable bases of nimbostratus clouds
below.) These clouds are generally the product of slow
upcurrents (centimeters per second) taking place in the
middle troposphere over an area of thousands of
square kilometers. Gray shading is expected in alto-
stratus, and is generally present in altocumulus clouds.
It is this property of shading that immediately differ-
entiates them from being ‘high’ clouds. Nimbostratus
clouds by definition are dark gray and the Sun cannot
be seen through them.
Altostratus and altocumulus are different from one
another in the same way that cirrus and cirrostratus
clouds are different from cirrocumulus clouds; in
altocumulus clouds, droplets predominate that gives
them a crisp, sharper-edged look. In altostratus clouds
ice crystals and snowflakes dominate or comprise the
entire cloud, giving it a diffuse, fibrous look. The
appearance of altocumulus and altostratus clouds has
long been a marker for deteriorating weather in the
hours ahead.
In spite of their name, altocumulus clouds are
generally rather flat clouds that strongly resemble
Figure 12 Altocumulus lenticularis. stratocumulus clouds. An exception to this overall
~~~
laminar architecture is in those species of altocumulus ing and shrinking in response to fluctuations in the
called castellanus and floccus. In this form altocumu- relative humidity of the air mass being lifted over the
lus clouds resemble miniature, lofted cumulus clouds terrain. Because the conditions under which these
in rows or patches. Altocumulus clouds are distin- clouds form are most often associated with advancing
guished from cirrocumulus because they are lower and short wave troughs in the middle and upper atmos-
the cloud elements in altocumulus are, or appear from phere and their accompanying regions of low pressure,
the ground to be, several times larger than those in lenticularis clouds are usually precursors to deterio-
cirrocumulus clouds. For example, the elements of an rating weather.
altocumulus cloud are typically the width of three In altostratus clouds (Figure5 ) ,the dominance of ice
fingers held skyward at the ground. Also, shading causes a diffuse, amorphous appearance with striat-
toward the center of the thicker elements is usually ions or fallstreaks on the bottom because an observer is
present in altocumulus clouds, a property that is not viewing relatively low concentrations of precipitation
allowed in the classification of cirrocumulus clouds. particles rather than a ‘cloud’ per se. Altostratus
Altocumulus clouds are distinguished from stratocu- clouds are rarely less than 2 km thick and often have
mulus because they are higher above ground level than tops at the same heights as cirrus and cirrostratus
stratocumulus (at least 2 km) and because the individ- clouds. Therefore, they are considerably colder than
ual cloud elements in altocumulus are, or appear from altocumulus clouds and span a much greater temper-
the ground to be, smaller than those in stratocumulus. ature range. They also, by definition, cover the sky or
In spite of the gray shading that may be present, at least wide portions of it, and produce a dull, gray,
altocumulus clouds are rarely more than 1km thick. and often relatively featureless base or one with trails
This is because the concentrations of drops in droplet of noticeable virga. Precipitation is usually imminent
clouds are relatively high (typically 50 000 to 200 000 when altostratus clouds cover the sky.
per liter). This density of particles produces an ‘optical The relatively low concentrations of large particles
depth’ in which the Sun’s disk can be obscured (optical in some altostratus clouds (tens per liter) can allow the
depth of 4 or more) by an altocumulus cloud layer only Sun’s position to be seen as though looking through
300-500 m thick. The same is true for stratocumulus ‘ground glass’, such that the position of the Sun is
clouds. apparent but the outline of its disk is not. This often
Altocumulus clouds sometimes sport patchy ‘virga’. occurs even when the cloud layer is, contrary to its thin
Virga is light precipitation that falls from a cloud but appearance, two or more kilometers thick (transluci-
evaporates before it reaches the ground. Because virga dus variety).
is almost always due to falling snow, it appears fibrous, When the top of an altostratus cloud layer is warmer
often with striations or long filaments that often far than about - 30” to - 35”C, it is not uncommon to
surpass the depth of the cloud from which it is falling, find, somewhat surprisingly, a thin droplet cloud
and can appear to be quite dense. Altocumulus clouds similar to an altocumulus cloud layer where the ice
with virga are predominantly those clouds whose below was initiated. The survival, growth, aggrega-
temperatures are lower than - 10°C. However, at the tion, and breakup of ice crystals over a great depth
same time, they are rarely colder than about - 30°C usually obscures the ice-producing droplet cloud top
because at very low temperatures they are likely to from an observer at the ground, who sees only the dark
take on the attributes of ice clouds such as cirrus or its gray fibrous ‘base’ of the altostratus cloud.
thicker brethren, altostratus. Optical phenomena associated with altostratus
The species of altocumulus clouds called altocumu- clouds are limited to parhelia (‘sun dogs’). However,
lus castellanus has always had a special significance in because the composition of the uppermost regions of
meteorology (Figure 11) because these clouds reveal the deepest altostratus clouds is virtually identical
an unstable lapse rate in the middle troposphere. with cirriform ice clouds with simpler, smaller ice
Instability of this sort has been viewed as a marker for crystals, haloes often appear to those in an aircraft
releases of deeper convection in the hours ahead. nearing the top of altostratus clouds.
Occasionally, the castellanus clouds group and enlarge Nimbostratus clouds (Figure 6) are virtually iden-
into cumulus and cumulonimbus clouds. tical to altostratus clouds in their composition except
When the winds are relatively strong aloft (greater that their ‘bases’ are usually perceived from the ground
than about 20 m s - ’) and moderately moist, but stable as lower than in altostratus, from which it has usually
lapse rate conditions are present, a species of altocu- derived by a downward thickening. They therefore
mulus called ‘lenticularis’ (lens or almond-shaped) often appear somewhat darker. The perceived ‘base’ of
may form over or downwind of mountains. Altocu- nimbostratus is due to snowflakes that are melting into
mulus lenticularis clouds (Figure 12) can hover over rain drops. This is because the greater opacity of snow
the same location for minutes to hours while expand- gives the impression of a ‘base’ or sharp increase in
474 CLOUDS / Classification
thickness at the melting level. Thus, without lower cause they have a rather lumpy appearance at cloud
clouds such as stratocumulus to obscure nimbostratus, base with darker and lighter regions due to embedded
it can be perceived at 'mid-levels' on a day when the weak convection. Also their bases tend to be higher,
freezing level is relatively high (above 2 km), such as in and more irregular in height than those of stratus
southern latitudes or the tropics, and be seen as a 'low' clouds. Stratus presents a smoother, lower, more
cloud when the freezing level is low, as in northern uniform sky than does stratocumulus cloud because
latitudes in the winter. the internal convective overturning that produces
Nimbostratus clouds produce relatively steady pre- lighter and darker regions is slight. Drizzle (precipita-
cipitation that often continues for hours at a time. tion comprising drops < 500 pm diameter that almost
They are not clouds responsible for passing showers float in the air) is likely to form in clean environments
with periods of sunshine between. As with other sheet in which cloud droplet concentrations are about
clouds, they are due to a gradual ascent of the air over a 100cm-3 or lower. For example, drizzle is common
large region, usually in advance of traveling low- from both stratus and stratocumulus clouds at sea and
pressure areas associated with upper-level short wave along coastlines in onshore flow.
troughs.
The tops of dark, steadily precipitating nimbostra-
tus clouds can be as shallow as 2-3 km and even be
above freezing in temperature, or they may reach
Convective Clouds
above 13 km (at cirriform cloud levels) and be as cold Cumulus and cumulonimbus clouds (Figures 9 and 10,
as -80°C. At elevations above the freezing level, respectively) are convective clouds brought about
nimbostratus is largely composed of ice crystals and when the temperature decreases rather rapidly with
snowflakes, though embedded thin (supercooled) increasing height. Differential heating and converging
droplet cloud layers similar to altocumulus clouds air currents in this circumstance can therefore send
are relatively common. Also, similarly to altostratus plumes of warmer air skyward with relative ease.
clouds, when the temperature at the top of nimbo- Convective clouds are limited in coverage compared
stratus clouds is above about - 30" to - 35"C, a thin with stratiform clouds and, except for the anvil
droplet cloud layer may be found in which the first ice portions of cumulonimbus clouds, rarely cover the
crystals are spawned and released. entire sky or do so only for short periods. This
A broken-to-overcast layer of shallow stratus or coverage characteristic differentiates, for example,
stratocumulus clouds often resides at the bottom of stratocumulus clouds, with their linked cloud bases
nimbostratus clouds. Such lower layers may obscure covering large portions of the sky, and similar-sized
the deep cloud producing the precipitation. However, cumulus clouds that by definition must be relatively
while usually not precipitating themselves, these lower scattered into isolated clouds or small clusters with
cloud layers are important in enhancing the amount of large sky openings.
rain or snow that falls from nimbostratus clouds. This Cumulus clouds have a size spectrum of their own
is because of the accretion or riming of the relatively that ranges from cumulus fractus, those first cloud
small cloud drops in the lower clouds by the rapidly shreds that appear at the top of the convective
falling precipitation-sized particles from the nimbo- boundary layer, to congestus size (more than about
stratus layer above. This enhancement is especially 2 km deep). Between these sizes are cumulus humilis
evident in hilly or mountainous regions. and cumulus mediocris clouds, which range between
Cumulonimbus clouds (see below) may also be about 1 and 2 km in depth, respectively. The tops of
embedded in nimbostratus clouds. The presence of these larger clouds are marked by sprouting portions
such clouds within nimbostratus is evident by sudden called turrets that represent the growing and usually
gushes of much heavier rain and sometimes lightning warmer parts of the cloud. Individual turrets are
within a context of relatively steady rain. generally one to a few kilometers wide, though in
strong storms individual turrets may coalesce into
groups of many turrets to form a large, tightly packed,
and hard-appearing cauliflower mass that roils up-
Low Stratiform Clouds
ward with little turret differentiation.
Stratus and stratocumulus clouds (Figures 7 and 8, Prior to reaching the cumulonimbus stage, cumulus
respectively) are shallow stratiform clouds, almost clouds are therefore composed of droplets and contain
always less than 1km thick. They are composed of very few if any precipitation particles. Precipitation,
droplets unless the cloud top is cooler than about - 5 however, usually begins to develop in cumulus con-
to - 10°C, in which case ice crystals may form. gestus clouds if they are more than about 3 km thick
Stratocumulus clouds differ from stratus clouds be- over land and about 2 km thick over the oceans. The
precipitation that falls may be due to collisions, with (soft hail) are usually found, if not at the ground
coalescence of the larger cloud drops in the upper then aloft, in virtually all cumulonimbus clouds.
portions of the cloud (particularly when the cloud base Updrafts may reach tens of meters per second in
temperature is above 5-10°C), or it may be due to the cumulus and cumulonimbus clouds, particularly in
formation of ice particles in clouds with cooler bases. warm air masses. These updrafts lead to large amounts
However, in the wintertime, even small cumulus of condensation and liquid water content. Depending
clouds with tops colder than about - 10 to - 15°C on how warm cloud base is, the middle and upper
can produce virga, snow flurries, or even accumulating building portions of deep cumulus clouds might
amounts of snow. These kinds of small, cold, and contain 1-5 grams per cubic meter of condensed
precipitating cumulus clouds are found in wintertime water in the form of droplets and rain drops, When
in such locations as the Great Lakes of the United ‘supercooled’, such liquid water concentrations are
States, off the east coasts of the continents, or over high sufficient to cause a buildup of about 1cm or more of
mountains. ice on an aircraft frame for every minute in cloud and
If significant precipitation begins to develop in deep thus pose a great hazard for aircraft. Cumulonimbus
cumulus clouds, they quickly take on the visual are the only clouds, by definition, that produce
attributes of cumulonimbus clouds (Figure 10); a lightning. That is, if lightning is observed, the cloud
strong precipitation shaft is seen below cloud base type producing it is automatically designated a
with a cloud top that is fibrous, fraying, or wispy. The cumulonimbus cloud.
visual transition to a softer, fibrous appearance in the
upper portion of cumulus clouds is due to the lowering
of the concentrations of the particles from hundreds of See also
thousands per liter of relativ ely small cloud droplets Cloud Microphysics. Clouds: Cumuliform.
( < 50 Km diameter), to only tens to hundreds per liter
of much larger (millimeter-sized) precipitation-sized
particles (r Further Reading
n drops or ice particles). These larger
particles tend to fall in filaments and help produce aHamblyn R (2001) The Invention of Clouds. New York:
striated appearance. Farrar, Straus and Giroux.
In the period while this transformation is taking Heymsfield AJ (1993)Microphysical structures of stratiform
place and the fibrosity is just becoming visually and cirrus clouds. In: Hobbs PV (ed.) Aerosol-Cloud-
apparent in the upper portions of a cumulus congestus Climate Interactions, pp. 97-121. New York: Academic
cloud as this particle spectrum change is underway, the Press.
cloud is entering a short-lived period of its life cycle Howell WE (1951) The classification of cloud forms. In:
when it is referred to as a cumulonimbus calvus Malone TF (ed.) Compendium of Meteorology, pp.
1161-1 166. Boston: American Meteorological Society.
(‘bald’) cloud (Figure 13). At this same time, a
Ludlum FH (1980) Clouds and Storms: The Behavior and
concentrated precipitation shaft may be just emerging Effect of Water in the Atmosphere, pp. 123-128.
below cloud base. When the fibrosity of the upper University Park: Pennsylvania State University Press.
portion of the cloud is fully apparent, the cumulonim- Schaefer VJ and Day JA (1981) A Field Guide to the
bus cloud has transitioned to a cumulonimbus cap- Atmosphere, pp. 45-134. Boston: National Audubon
illatus (‘hairy’) in which most or all of its upper portion Society.
consists of ice crystals and snowflakes. In the tropics, Scorer RS and Verkaik A (1989) Spacious Skies. London:
this visual change also occurs but can be due solely to David and Charles.
drizzle and raindrops in smaller cumulonimbus Wallace JM and Hobbs PV (1977)Atmospheric Science: An
clouds. If a pronounced flattening of the top develops Introductory Survey, pp. 216-218. New York: Academic
into a spreading anvil, then the cloud has achieved the Press.
World Meteorological Organization (1956) International
status of a cumulonimbus capillatus incus (‘anvil’).
Cloud Atlas (Complete Atlas), vols. I and 11. Geneva:
Cumulonimbus capillatus clouds span a wide range WMO.
of depths, from miniature versions only about 2 km World Meteorological Organization (1975) International
deep in polar air masses over the oceans, to as much as Cloud Atlas, vol. I, Manual on the Observations of
20 km in the most severe thunderstorms in equatorial Clouds and Other Meteors. Geneva: WMO.
regions, the plains of eastern China, and the plains and World Meteorological Organization (1987) International
south-east regions of the United States. Hail or graupel Cloud Atlas, vol. 11. Geneva: WMO.
476 CLOUDS / Climatology
Climatology
S G Warren, University of Washington, Seattle, WA, heating of the surface can initiate convection and cause
USA cumulus (Cu) clouds to form at the lifting condensa-
C J Hahn, University of Arizona, Tucson, AZ, USA tion level. Cumulus clouds are small; they may develop
further into large cumulonimbus (Cb) clouds.
Cumulonimbus can extend vertically to the tropo-
pause and often contain ice crystals in their upper
Introduction parts. Cumulonimbus are associated with thunder,
lightning, and showery rain or snow.
Clouds are an important component of the Earth’s Stratus (St) and stratocumulus (Sc) are both hori-
climate system. They reflect solar radiation back to zontally extensive low clouds. They are distin-
space, they absorb thermal infrared radiation emitted guished in that stratocumulus contains convective
from below, and they produce rain and snow. A cloud elements but stratus does not. Fog is a cloud at the
climatology describes the time-averaged geographical ground surface, usually in the form of stratocumulus.
distribution of cloud properties and the diurnal, Stratus and stratocumulus cover large regions of the
seasonal, and interannual variations of those
oceans. In the subtropics they are found over the
properties. eastern parts of the oceans, where subsidence is
Cloud climatologies are used to determine the
occurring in the free atmosphere above the boundary
radiative effects of clouds on climate and to determine layer.
the extent to which interannual and multidecadal Nimbostratus (Ns) clouds are much thicker than Sc
changes in the Earth’s radiation budget can be attrib-
and St, extending vertically through several kilometers
uted to changes in clouds. Cloud climatologies also of the atmosphere. Nimbostratus clouds form as a
find applications in assessing the prediction of clouds
result of large-scale uplift of moist air in frontal storms
by climate models, assessing the significance of chem-
at middle and high latitudes, and they precipitate rain
ical reactions in clouds, quantifying climatic feedbacks
and snow.
involving clouds, estimating the radiative forcing
Clouds with bases 2-6km above the surface are
by anthropogenic aerosols, selecting sites for astro- termed ‘middle’ clouds, and are classified altostratus
nomical observatories and atmospheric field experi- (As) or altocumulus (Ac) by their resemblance to
ments, and assessing the potential for solar energy stratus or cumulus.
development. Clouds with bases between 6 km and the tropopause
The properties of clouds most important for climate are the ‘high’ clouds: cirrus (Ci),cirrostratus (Cs) and
are those that affect radiation and precipitation, cirrocumulus (Cc).They consist of ice crystals and as a
namely, cloud height, thickness, horizontal extent group are called ‘cirriform’ clouds. They can result
and horizontal variability, water content, phase (liquid from gradual uplift in large-scale storms in mid-
or ice), and droplet and crystal sizes. It is therefore latitudes, or can be sheared off the tops of cumulo-
important to distinguish different types of clouds. The nimbus in the tropics.
climatic effects of clouds further depend on the Clouds above the tropopause are rare, but they can
geographical location of the clouds, the albedo and occur in the polar regions in the stratosphere at
temperature of the underlying surface, and the season 15-25 km height as polar stratospheric clouds (PSCs;
of the year and time of day. The effect of clouds on nacreous clouds), and in the mesosphere a t 8 0 k m
the Earth’s radiation budget, called the ‘cloud height as polar mesospheric clouds (PMCs; noctilu-
radiative forcing’ is generally negative in the daytime cent clouds). Those two types of clouds are discussed
but positive at night (that is, clouds cool the surface in other articles in the Encyclopedia; this article is
in the day but warm the surface at night), so an concerned only with tropospheric clouds (see Nocti-
accurate determination of the diurnal cycle of each lucent Clouds).
cloud type is an important component of a cloud
climatology.
Satellite Observations
Cloud Types
Cloud climatologies have been developed from two
Clouds are classified according to their form and their kinds of data: (1) using radiances measured by
height. Low clouds, with bases in the atmospheric satellites in polar and geostationary orbits, and (2)
boundary layer less than 2 km above the surface, using visual observations of clouds from the Earth’s
are influenced by their proximity to the surface. Solar surface, as coded in weather reports from stations on
CLOUDS I Climatology 477
land and ships in the ocean. Satellites detect clouds hidden from the satellite’s view by higher clouds.
principally at visible and thermal-infrared wave- Multiple cloud layers often occur together, so the
lengths. At visible wavelengths cloudy scenes appear views from above and below are complementary.
brighter than cloud-free scenes when viewed from Some clouds are difficult to detect from satellites
above. Clouds are usually colder than the underlying (clouds over snow, low clouds at night) because they
surface, so the emission of thermal-infrared radiation provide little contrast in albedo or temperature with
to space is less than for clear scenes. During the the underlying surface.
daytime clouds can be detected in both wavelength Surface observers are close to the clouds, so they can
regions, but at night only in the thermal infrared. The identify clouds by type, including clouds smaller
altitude of the cloud top is inferred by relating the than a satellite’s pixel size, which is typically at least
infrared emission temperature to the vertical profile of 1km.
temperature obtained from radiosondes (carried by The cloud types defined morphologically by surface
weather balloons) or satellite sounders. Cloud optical observers are directly related to meteorology and
thickness (opacity) is inferred from reflectance in the cloud processes, whereas the satellite climatology
visible channel. If a second solar channel (in the near defines cloud types by their radiative properties.
infrared) is available, then the vertically integrated Weather reports of clouds are available for several
liquid water content, and the effective radius of the decades with no change in official observing in-
droplets, can also be inferred. structions, so interdecadal variations and trends can
Measurements from satellites can be used to probe studied.
duce a cloud climatology if the following criteria are
satisfied: (a)pixel size is at most a few kilometers; (b) This review emphasizes the climatology obtained
temporal sampling is conducted at regular intervals from surface observations, because that is the subject
throughout the day and night; (c) the coverage is of the authors’ own research.
global; and (d)a long period of record (many years) is
maintained. To satisfy these requirements, the Inter-
national Satellite Cloud Climatology Project (ISCCP) Cloud Information in Surface Weather
uses five geostationary satellites that hover over the
Reports
Equator at five longitudinal locations, and two polar-
orbiting satellites. This project began in 1983 and is Cloud observations are coded into weather reports
continuing. using the ‘synoptic code’ of the World Meteorological
More detailed information about clouds can be Organization (WMO). In some countries the obser-
obtained from satellite instruments with finer spatial vations are reported both in the synoptic code and in
resolution (e.g., Landsat), and from satellites with another code used locally. Reports in these other codes
more spectral channels (e.g., the Moderate Resolution have been used to develop climatologies in some
Imaging Spectrometer (MODIS) on the Earth Obser- individual countries, but the synoptic code is the only
vation Satellite (EOS)). Three-dimensional informa- one used worldwide. The information about clouds in
tion about clouds can be obtained from satellites that the synoptic weather report consists of total cloud
look at the same scene from different angles (e.g., the cover, low or middle cloud amount, low cloud type,
Multi-angle Imaging Spectroradiometer (MISR)). middle cloud type, high cloud type, present weather,
These instruments are useful for studying cloud and base height of the lowest cloud.
properties but do not yet offer sufficiently long periods About 6500 land stations routinely report cloud
of record to produce a climatology (see Satellite observations in the synoptic code. Usually they report
Remote Sensing: Cloud Properties). The principal every three hours beginning at 0000 UT, but about
satellite cloud climatology in use now is therefore that one-quarter of them report only every six hours. About
of ISCCP. 20% of the stations do not make observations at night.
The average spacing of land stations is about 180 km,
Surface Observations but it is far from uniform. Europe has more stations
Surface observations of clouds are made less frequent- than needed for a cloud climatology, and Antarctica
ly than satellite observations in many areas, and they has too few. Some parts of the Sahara Desert and
have variable spatial density, but they offer a useful Western Australia are also inadequately sampled.
adjunct to satellite observations for the following Most ships make weather observations four times
reasons. per day; the observations are recorded in logbooks and
also transmitted by radio to world meteorological
0 The surface observer views clouds from below, and centers. In a recent typical year, reports from an average
thus can observe the low clouds that are often of 1150 ships were received at 0000,0600,1200,1800
478 CLOUDS I Climatology
UTand from 160 ships at 0300,0900,1500,2100 UT. The amount, or even the presence, of a middle or
Most of these are merchant ships with volunteer high cloud may be indeterminate when a lower cloud
weather observers; some are military ships and re- nearly or completely covers the sky. The average
search ships, and a few (<10) are dedicated weather amounts of middle and high cloud types can be
ships. Unlike on land, there is little tendency for fewer estimated by assuming that the frequency and
observations at night, but the nighttime observations amount-when-present are the same in observations
may not be transmitted promptly by radio so it is where they cannot be calculated as in observations
important to have the complete logbook records. The where they can be calculated. Also, to obtain amount-
average spacing of ships that report clouds is 600 km, when-present, the clouds at different levels are as-
much greater than for weather stations on land, but sumed to be randomly overlapped. The amounts
since the ships are moving they sample most parts of the directly visible from below (the 'nonoverlapped'
ocean. A project to compile ship-based weather obser- amounts) may also be calculated.
vations from all maritime nations, including many For the climatology, the Earth is divided into an
logbook reports, has resulted in the Comprehensive array of boxes on a geographical grid, and cloud cover
Ocean-Atmosphere Data Set (COADS),which is being is computed for each box. Several possible biases may
used for research on air-sea interaction and climatic affect computed cloud cover but they may be reduced
change throughout the world ocean. or eliminated with appropriate analysis procedures.
In many parts of the ocean the accuracy of comput- Two small biases that oppose each other and are
ed mean cloud amounts is limited by the scarcity of unique to ship observations are the fair-weather bias
observations. This is not the case on land, where the (the tendency for more ships to enter a grid-box on
random error due to inadequate temporal sampling is days of fair weather) and the foul-weather bias (the
very small. Statistical tests performed on observations tendency of ships to oversample stormy or foggy
from weather ships indicate that 100 observations weather because they are traveling more slowly). Two
taken at random times during a three-month period other biases that may affect both ship and land data are
will represent the seasonal mean cloud cover to within the diurnal sampling bias (somewhat more reports are
3% in an oceanic grid-box of size 5" latitude by 5" transmitted by ships during the day than at night, and
longitude. If 1 % accuracy is desired, then 1000 some land stations in a box with several stations may
observations are needed. not report at night) and the trend bias (a box may be
The synoptic code was defined in 1929 but was sampled by more ships in later years than in earlier
changed in 1949; the reporting procedures became years, or a land station may change location during the
adopted worldwide in the early 1950s. Synoptic period of record). These situations can cause biases if
observations are available with global coverage for the cloud amount undergoes a diurnal cycle or exhibits
all oceans since 1952 and for all continents since 1971, a long-term trend, but such biases can be eliminated by
about 400 million observations to date. appropriate analysis procedures.
The most serious bias, on both land and ocean, is the
'night-detection bias'. Visual observations of clouds
Computation of Average Cloud are hindered at night by inadequate illumination of the
clouds. This usually leads to an underestimation of the
Amounts
average cloud cover at night, especially for the
For low clouds the computation of average amount is amounts of middle and high clouds, in climatologies
straightforward, but for middle and high clouds the based on surface observations. The diurnal cycles of
question of overlap must be considered. The 'amount' cloud amounts, if based on all the surface observa-
of a cloud type is defined as the fraction of the sky tions, are therefore in error, but the cycles can be
covered by that type, whether visible or hidden behind obtained more accurately if the nighttime observa-
another cloud. The time-averaged amount can be tions are screened to select those made under sufficient
obtained as the product of frequency-of-occurrence moonlight or twilight. A criterion for adequacy of
(fraction of weather observations in which a cloud of moonlight or twilight has been established; it permits
this type is present, whether visible or hidden) and the use of about 38% of the nighttime observations. By
amount-when-present (the average fraction of the sky this criterion, adequate illumination is provided by a
covered by this cloud type when it is present, whether full moon at an elevation angle of 6" or a partial moon
visible or hidden). For example, if cumulus is present at higher elevation, or by twilight from the Sun less
in 30% of the weather observations from a station, than 9" below the horizon.
and if it covers on average 40% of the sky when it is A complete description of the climatology of clouds
present, then the average amount of cumulus at that is the subject of atlases such as those listed in Further
station is 12%. Reading, which give the average amounts of each
CLOUDS I Climatology 479
Table 1 Annual average cloud cover from surface observations For the low clouds, Table 3 also shows the obser-
(1982-1 991) vers' estimate of the height of the cloud base above the
Land Ocean Globe ground surface. The bases are on average twice as high
over land as over ocean, and the heights increase with
Average total cloud amount (%) 54 68 64 distance inland from the ocean.
Day-night difference (%) +3 0 +1
Geographical Variations
cloud type for each season in grid-boxes of 5" latitude
by 5" longitude, as well as their diurnal cycles and What the averages of Table 3 cannot show is that there
interannual variations. A few illustrative examples are striking geographical variations. Fog is rare over
from the climatology are shown in this article. most of the globe but its frequency exceeds 10% over
the North Atlantic and North Pacific oceans in
Global Averages summer poleward of 40" N, and reaches 2 0 4 0 % in
the Sea of Okhotsk. Nimbostratus is likewise rare in
The annual average total cloud cover as determined the tropics but common in middle and high latitudes.
from surface observations is summarized in Table 1. Cumulonimbus amount exceeds 10% in a narrow
Average cloud cover is greater over the ocean than over band along the Intertropical Convergence Zone
land. Cloud cover over land tends to be greater in (ITCZ) near the Equator and over a much broader
daytime than at night, but the ocean shows little day- region of warm water in the western Pacific called the
night difference. 'warm pool'.
Cloud properties from eight years (1986-1993) of Completely clear sky, also given in Table 3, is
ISCCP are summarized in Table 2. The average cloud common over land but rare over the ocean. Over most
cover and the day-night differences are slightly of the open ocean there are almost always some low
different from those obtained from surface observa- clouds visible from ships; the reports of clear sky are
tions (Table 1).The optical thickness (opacity) and mostly confined to coastal regions. The Red, Mediter-
cloud-water path (vertically integrated liquid-water ranean, and Arabian Seas are the most cloud-free parts
content) inferred from the satellite radiances are of the world ocean.
smaller than those usually obtained from aircraft in The geographical distribution of total cloud cover
field experiments. This difference is probably due to for December-January-February (DJF) is shown in
horizontal inhomogeneity of the clouds; the ISCCP
optical thickness is an effective optical thickness for a
hypothetical horizontally homogeneous cloud. Table 3 Cloud type amounts and heights from surface observa-
Global average amounts for nine different cloud tions
types defined in the surface observations are shown in ~ ~ ~~
Cloud type Annual average Base height (meters

Table 3. Globally, the most common types are strato- amount ("A) above surface)
cumulus, altocumulus, and high (cirriform)clouds. All
the low cloud types are more common over the ocean Land Ocean Land Ocean
than over land. The middle types altostratus and 1 1 0 0
Fog
altocumulus together cover the same fraction of the Stratus 5 11 500 400
sky over land as over ocean; cirriform cloud is the only Stratocumulus 12 22 1000 600
type that is less common over ocean. Cumulus 5 12 1100 600
Cumulonimbus 4 6 1000 500
Nimbostratus 6
Table 2 Annual average cloud properties from the International
Altostratus 22
Satellite Cloud Climatology Project (1986-1 993)
Altocumulus 17
Land Ocean Globe High (cirriform) 22 13
Clear sky 22 3
Average total cloud amount (Yo) 58 72 68 (frequency)
Day-night difference (Yo) +5 -2 0
Cloud top temperature ("C) - 20 -7 - 11 Land values are new analyses for the years 1971-1996. Ocean
Day-night difference ("2) +12 +2 +5 values are for the years 1952-1981, adapted from Table 9 of
Cloud top pressure (hPa) 490 620 580 Warren SG, Hahn CJ, London J, Chervin RM and Jenne RL (1988)
Cloud optical thickness 4 4 4 Global Distribution of Total Cloud Cover and Cloud Type Amounts
Cloud-water path (g m -') 76 61 66 over the Ocean, Tech. Note TN-317+STR, National Center for
Atmospheric Research, Boulder, Colorado. For the ocean analy-
Adapted from Tables 4 and 5 of Rossow WB and Schiffer RA sis, altostratusand altocumuluswere combined into a single group.
(1999) Bulletin of the American Meteorological Society 80: 2261- The amounts of all the cloud types add up to more than the total
2287, American MeteorologicalSociety, Boston, with permission. cloud cover because of overlap.
Figure 1 Percent total cloud cover for Decemberdanuary-February from surface observations (weather stations on land, ships in the
ocean) for the 10-year period 1982-1991, average of day and night.
Figure 1. This is the only season in which complete

global coverage is available from surface observations, 90
because ships avoid the Antarctic Ocean in other 80
seasons when it is ice-covered. The largest cloud
70
amounts are found in the high-latitude oceans,
-8 60
particularly in summer, exceeding 90% in the sub-
Antarctic in DJF. The North Atlantic and North
-5 50
Pacific cloud amounts reach similarly high values 8' 40
during the northern summer (June-July-August; JJA). ;30

Values of cloud cover below 40% are found in the 8 20
deserts of Australia, Mongolia, Arabia, North Africa, (A)
-8
r. ,!
Southern Africa, Argentina, and Mexico. Most of the 100 1""'""'"""'i
Indian subcontinent has cloud cover of 20-30% on
1
this map, during the winter-monsoon dry season. The
eastern Sahara is the clearest large region on Earth,
with cloud cover below 20% on this map for DJF, but
it is even clearer in summer when a large area of less 50 , , . l 1 '? , /oP
d,;; ,
than 5 % cloud cover extends on across northern
40 OB.""
Arabia.
30 OOOO
The ITCZ appears in Figure 1 as a latitudinal
20
maximum near the Equator in the Atlantic and north
90" S 60" S 30" S 0" 30" N 60" N 90" N
of the Equator in the eastern Pacific, then south of the
Latitude
Equator through Indonesia, and at about 10" S across
the Indian Ocean, Africa and South America. This is in Figure 2 Zonal average total cloud cover (average of day and
agreement with the location of the ITCZ as deter- night) for 5' latitude zones. Separate averages are formed for the
land and ocean parts of each zone. (A) December-January-
mined by measurement of winds and pressure.
February 1982-1 991. (B) June-July-August 1982-1 991. (Adapt-
The total cloud cover averaged around latitude ed from Figure 4 of Hahn CJ, et a/. (1995) Journal of Climate 8:
zones is shown in Figure 2 for the two extreme seasons. 1429-1 446, American Meteorological Society, Boston, with
The figure shows that the average cloud cover is less Dermission.)
,
CLOUDS / Climatology 481
over land than over ocean, and that the latitudinal

variation of cloud cover is greater over land than over
ocean. The peak cloudiness in the ITCZ moves from
7" N in JJA only to 2" N in DJF over the ocean, but to as
far as 12"S over land. The latitudes of maximum cloud
cover near 60"N, 60" S, and the Equator correspond to
the latitudes of maximum precipitation, and the
+---I
30
40
latitudes of minimum cloud cover on land are the

latitudes of the great deserts.
Diurnal Variations
The amounts of many cloud types vary from day to
night. Two examples of diurnal variations for oceanic 6 8 10 12 14 16 18 20 22 0 2 4 6
regions are shown in Figure 3. The central North Local time (h)
Atlantic in winter exhibits no diurnal variation, with
Figure 4 Diurnal cycles of cumulus and cumulonimbus amounts
total cloud cover averaging about 75% at all hours.
reported from weather stations in Central America (10-15"N,
The largest oceanic diurnal variations are in the 85-90"W) in summer (June-July-August 1971-1981).
stratocumulus regions of the eastern subtropical
Atlantic and Pacific. The region displayed, in the
Atlantic Ocean west of Namibia, exhibits a strong
further develop into cumulonimbus, which continue
diurnal cycle in total cloud cover with a peak of 73% at
precipitating into the evening.
07.00 and minimum of 53% at 16.00. This cycle is
paralleled by the diurnal cycle of low stratiform
clouds, indicating that these cloud types are the types
Seasonal Variations
responsible for the diurnal cycle here. These bound-
ary-layer clouds develop during the night and dissipate The largest seasonal variations of cloud cover are
during the day under the influence of solar heating. associated with the subtropical monsoons of Africa,
Figure 4 shows an example of diurnal cycles on land South America, India, and Australia. Cloud variations
in Central America during the summer rainy season. in the Indian Ocean region are shown in Figure 5A. In
Solar heating of the surface begins at sunrise, leading south-west India the average total cloud cover in-
to convection that produces cumulus clouds in the creases from 16% in February to 89% in July. During
morning. In the afternoon, many of these clouds India's dry winter, northern Australia experiences its
cloudy and rainy summer.
In contrast to the sinusoidal pattern of the Indian
and Australian monsoons, the central Arctic Ocean
(Figure 5B)exhibits a peculiar boxlike seasonal cycle,
: l Central North Atlantic in which cloud cover increases rapidly during May.
The greater cloud cover from June to September is due
h
80 -I Tntal cloiid t mainly to the low, thin 'Arctic summer stratus' clouds
8
1
E 70 that form over the perennial sea ice during the melting
3
season. A still different pattern is exhibited in parts of
$ 60 the North Atlantic (Figure SB), which have almost no
U
seasonal variation of cloud cover.
ua 50
\o Total cloud
40 South-east Atlantic
St+Sc amount lnterannual Variations and Trends
30
I ' I ~ I ~ I ~ I ~ I ~ I Clouds
~ I ' interact
I ~ with
I ~ other
I components of the climate
~
0 2 4 6 8 10 12 14 16 18 20 22 24 system, so changes in cloud amounts can be expected
Local time (h) to accompany changes in other climatic variables, and
also to feed back on those other variables. The
Figure3 Diurnal cycles of oceanic cloud, from ship observations
in December-January-February (1 982-1 991 for the North Atlantic
magnitudes, and even the nature, of the possible
( 3 0 4 0 c N , 40-60"W); 1952-1981 for the Southeast Atlantic climatic feedbacks involving clouds are far from
(2O-3O0S, 0-20" E)). understood, but the long historical climatic record
Next Page
100 circulation) and effects (e.g., diurnal temperature

90 range) of the cloud changes and to correlate these
80
related climatic variables to the cloud changes. Some
-ae
v
70
examples are shown in Figure 6 . Interannual varia-
tions of the amount of marine stratiform cloudiness
60 (St+Sc) commonly correlate negatively with inter-
>
8 50 annual variations of SST. Many of these interannual
?0 40 variations are related to cycles of El Nifio and the
-
2 30 Southern Oscillation. Figure 6 shows how St+Sc and
d SST covary at two locations in the Pacific Ocean: one
20
location that exhibits a long-term trend and another
(A)
0
lo
J
2 F M A M J J A S O N D J
that does not. Both sets of data came from shipboard
observations. The fact that clouds and SST are
measured by different methods but correlate well
100 argues for the reality of both time-series. The good
90 correlations in Figure 6 also suggest that the error in a
80
seasonal mean due to random sampling of weather
-ae
v
70
situations during a particular season is small. In other
grid-boxes traversed by fewer ships, seasonal means
2 6o suffer from sampling error and the correlation of
8 50 St+Sc with SST is not as strong.
U
3 40 The global average reported cloud cover over the
-0
2m 30 oceans has apparently increased by 1-2% since 1952.
w
$ 20
The largest increase is in the tropical regions. How
10
0
1 June-July-August 30-40 N 160-180 W l
J F M A M J J A S O N D J
(B) Month
Figure 5 Examples of seasonal cycles of total cloud cover,

1982-1991, from surface observations. (A) Land: 0 , South-west
India, 15-17.5" N, 72.5-75" E; Northern Australia, 12.5-1 5" S,
130-132.5" E. (B) Ocean: 0 , North Atlantic, 45-50" N, 20-25" W;
Arctic Ocean, 80-90" N. The observations in the Arctic Ocean were
made from drifting stations established on perennial sea ice.
may help to identify them. The degree to which the (A) 3 1950 1960 1970 1980 1990 2000
actual variations of the amounts of the different cloud 5
0
types are faithfully recorded in the analysis of visual 9
m
70 June-July-August 0-10 S 80-100 W Y
e
observations is itself variable, depending on the spatial 60
and temporal density of observations, the ability to 8
c 50
detect and remove biases, and the spatial scale of the E
analysis. Real interannual variations of cloud amount
5 40
in a 10" x 10"box, for example, are often large enough 30

to overwhelm any subtle progressive changes in 20
observing procedure. However, interannual variations 10
of zonal average cloud amount are smaller than those 6
0
of grid-box cloud amount because of partially com- 1950 1960 1970 1980 1990 2000
pensating positive and negative changes in different (B) Year
parts of the zone. For zonal averages it is therefore
more difficult to dissect the observed changes into Figure 6 Seasonal average daytime amounts of stratus plus
climatic and nonclimatic causes. stratocumulus, and seasonal average sea surface temperatures
(SST) for two grid-boxes in the Pacific Ocean. (A) June-July-
A powerful way to assess the validity of observed August, 30-40"N, 160-180"W. (B) June-July-August, O-lO"S,
cloud changes is to identify-likely causes (e.g., changes 80-100" W. The SST is plotted on a reversed scale to illustrate the
in sea surface temperature (SST) or atmospheric correlation. (Data supplied by Dr Joel Norris.)
498 COASTAL METEOROLOGY
sample volume is on the order of 20%. As in the single- niques, vol. 2, ASTM STP 1083, pp. 115-127. Philadel-
particle light scattering probes, derived integral prop- phia: American Society for Testing and Materials.
erties like LWC will have a n uncertainty proportional Gayet JF, Crepe1 0,Fournol JF and Oshchepkov S (1997)A
to the errors in concentration and sizing. The estimat- new polar nephelometer for measurements of optical and
ed uncertainty in derived LWC from OAPs is approx- microphysical cloud properties. Part I: Theoretical
design. Annales Geophysicae 15: 451-459.
imately 33%. A much larger source of error with
Gerber H (1991)Direct measurement of suspended partic-
OAPs arises when hydrometeors are ice, with aspher- ulate volume concentration and far-infrared extinction
ical geometry and unknown bulk densities. In these coefficient with a laser-diffraction instrument. Applied
cases the uncertainty in derived ice water content may Optics 30: 48244831.
easily exceed 100%. King WD, Parkin DA and Handsworth RJ (1978)A hot wire
water device having fully calculable response character-
istics. Journal ofApplied Meteorology 17: 1809-1813.
See also Knollenberg RG (1981) Techniques for probing cloud
microstructure. In: Hobbs PV and Deepak A (eds)
Cloud Chemistry. Cloud Microphysics. Clouds: Cli-
Clouds, Their Formation, Optical Properties, and
matology. Convective Cloud Systems: Modelling. Me-
Effects. New York: Academic Press.
soscale Meteorology: Cloud and Precipitation Bands.
Korolev AV, Strapp JW, Isaac GA and Nevzorov AV (1998)
Parameterization of Physical Processes: Clouds.
The Nevzorov airborne hot wire LWC/TWC probe:
Satellite Remote Sensing: Cloud Properties.
principle of operation and performance characteristics.
Journal of Atmospheric and Oceanic Technology 15:
1495-1510.
Further Reading
Noone KJ, Ogren JA, Heintzenberg J, Charlson RJ and
Baumgardner D and Rodi A (1989) Laboratory and wind Covert DS (1988) Design and calibration of a counter-
tunnel evaluations of the Rosemount icing detector. flow virtual impactor for sampling of atmospheric fog
Journal of Atmospheric and Oceanic Technology 6 : and cloud droplets. Aerosol Science and Technology 8:
971-9 79. 235-244.
Baumgardner D, Cooper WA and Dye JE (1990)Optical and Twohy CH, Schanot AJ and Cooper WA (1997) Measure-
electronic limitations of the Forward-Scattering Spec- ment of condensed water content in liquid and ice clouds
trometer Probe. In: Hirlman ED, Bachalo WD and Felton using an airborne counterflow virtual impactor. Journal
PG (eds) Liquid Hydrometeor Size Measurements Tech- of Atmospheric Oceanic Technology 14: 197-202.
W A Nuss, Naval Postgraduate School, Monterey, CA, comprise coastal meteorology, these circulations are
USA united through general dynamic principles associated
Copyright 2003 Elsevier Science Ltd. All Rights Reserved with several basic forcing mechanisms and their
interaction with larger-scale processes that occur a t
coastlines.
As suggested in the above definition of coastal
Introduction meteorology, a broad range of spatial and temporal
Coastal meteorology is the study of smaller-scale scales is important in coastal weather events. Typi-
weather systems that can be attributed to the presence cally, the mesoscale or smaller-scale aspect of the
of a coastline. The detailed distributions of winds, coastal circulations is of primary interest, which often
temperature, clouds and precipitation, visibility, arises from the smaller-scale structure of the coastline
boundary layer structure, and other weather elements itself. Mesoscale coastal structures arise in two distinct
in regions of complex coastal geometry are of ways. First, mesoscale dynamic forcing of the atmo-
fundamental interest in coastal meteorology. Both sphere by the coast can generate mesoscale circula-
the development of smaller-scale circulations forced tions. In these situations the synoptic-scale forcing is
by the change in surface characteristics at the coastline typically rather weak and the coastal properties
as well as the modification of larger-scale weather generate structure differences that dynamically force
systems by interaction with a coastline are a part of the atmosphere. Second, mesoscale structures develop
coastal meteorology. Although there are a myriad along coastlines through the interaction of the coast
specific coastal circulations around the world that and a synoptic-scale circulation. These mesoscale
COASTAL METEOROLOGY 499
structures are an artifact of the interaction for a given atmosphere that occurs at a coastline. Fundamentally,
synoptic forcing and depend upon the synoptic forcing the only difference at the coastline is a change in the
in order to exist at all. For these types of circulations surface properties, which impact directly only on
the synoptic forcing tends to be strong and the coastal the atmospheric boundary layer. The structure of
properties alter the synoptic forcing to introduce or the boundary layer is determined by a combination of
organize features within the synoptic scale system. The surface forcing through the surface fluxes and the
coastal forcing may not be fundamental to their larger-scale flow established by the synoptic-scale
existence but simply a modifying factor. Examples of weather pattern.
mesoscale forced circulations and synoptically forced The surface energy balance that determines the
circulations are the sea breeze and coastally forced surface temperature on either side of the coast
precipitation structures in a front, respectively. typically establishes different heat fluxes over the
The development and evolution of coastal circula- water versus the land. At times the overlying atmo-
tions are inherently dependent on the nature of the sphere may be warmer than the surface and other
coastal boundary. The coastal boundary may be times it may be colder than the underlying surface.
characterized either by a change in the surface Consequently, the boundary layer stratification and
properties from water to land that result in heating the associated transfer of heat upward from the
differences or by coastal mountains or slope changes surface usually result in differing boundary layer
that can block or channel the flow. These surface structures over the land and the water as well as a
properties vary along the coast, which adds additional surface temperature gradient at the coast. This
complexity to the nature of the boundary. These difference in low-level thermal structure can force
characteristics of the coastal boundary allow a flow acceleration across the coast and produce
simplification of coastal circulations into three idea- differing depths to the boundary layer on either side
lized coastal meteorological problems: Thermally of the interface, which give rise to a variety of coastal
forced circulations, Topographically forced circula- circulations (see Energy Balance Model, Surface).
tions, and Synoptic-scale coastal interactions. These The coast also represents an important moisture
describe the fundamental processes that shape coastal boundary that arises because of the differing surface
weather phenomena. Pure thermally driven circula- fluxes of moisture over the water versus the land.
tions arise due to heating differences that occur across These differences in low-level moisture contribute to
a flat or gently sloped coastal boundary and are often the occurrence of and modifications to coastal clouds
considered without along-shore variations. Examples and precipitation. Coastal fog and low-level stratus
of these types of circulations are the classic sea or land are a direct result of the moisture supplied by the ocean
breeze and coastal fronts. Pure topographically forced that has cooled either through radiative or dynamic
circulations arise from a significant change in surface processes. Coastal thunderstorms depend heavily on
slope across the coast and do not depend upon the the supply of moisture from the ocean that gets forced
surface characteristic changes that may also occur. upward in coastal circulations such as the sea breeze.
Examples of topographically driven circulations While the coastal moisture distribution is not funda-
include lee vortices, flow-blocking effects, and cold mental to orographic enhancements to coastal rain,
air damming. Synoptic-scale coastal interactions the interaction of various coastal flows with coastal
represent a class of smaller-scale phenomena that arise mountains is a very important process that determines
from the passage of a larger-scale circulation across a details in coastal clouds and precipitation.
coastline. These smaller-scale circulations occur when Friction differences also occur across the coastline
the surface characteristics are able to modify larger- and influence the character of the coastal atmosphere.
scale processes in a manner that spawn a circulation The frictional turning of the flow toward lower
that may depend upon both the coastline and pressure that normally occurs near the surface may
the synoptic scale system. Examples include the be more pronounced over the rough land surface
modification of fronts approaching or crossing a compared with the relatively smooth water surface.
coastline to produce a coastal barrier jet and tropical These frictional differences can result in coastal
cyclones making landfall to spawn severe coastal convergence or divergence, depending upon the
thunderstorms. direction of the flow (offshore, onshore, or coast-
parallel). Given time for the flow to adjust, these
frictional differences are often not important for
spawning coastal circulations, but they can greatly
Character of the Coastal Atmosphere shape the character of the flow along a coast.
The development of coastal circulations depends upon The nature of the coastal atmosphere is also
the structure and characteristics of the low-level determined by the direction of the large-scale flow
direction relative to the coastal boundary. Synoptic- in numerous dynamics textbooks. The key point is that
scale weather patterns are very important for setting if a thermal gradient develops (baroclinic structure)
the stage for different types of coastal circulations orthen the air must respond to the pressure gradient to
weather events. The larger-scale weather system force a circulation. If the thermal gradient stays fixed
determines the flow direction relative to the coastline over time then the atmosphere will simply adjust to the
to be directed either offshore, onshore, or coast- pressure gradient and the winds tend toward geos-
parallel. These differing flow directions as well as thetrophic balance. Hence, the initially cross-coast
strength of the flow are important for determining forcing would result in geostrophically balanced
what type of coastal circulation may develop. For along-coast winds. The time scale of this adjustment
example, a sea breeze can be completely masked by is given by the inertial period, assuming that the
strong onshore flow. Another aspect of the larger-scale thermal structure does not change by the induced
weather system that is important in coastal meteorol- circulation.
ogy is the processes that establish the low-level static One of the most recognizable thermally forced
stability. Large-scale subsidence can be very important coastal circulations is the sea-land breeze circulation.
This circulation owes its existence to the differential
for establishing a strong low-level inversion that limits
the ability of the air to flow over coastal topography. heating that occurs across the land-sea boundary. As
Differential warm advection and its time variation can the land warms up relative to the adjacent ocean, a
be important for altering the boundary layer stratifica-temperature gradient is established at the coast, as
tion to allow the excitation of low-level mountain shown in Figure 1. As noted in the above description of
waves ahead of an approaching frontal system. In thermal forcing, this leads to the acceleration of the
short, the large-scale weather system establishes flow across the coastal boundary, owing to the induced
the environment within which mesoscale coastal pressure gradient. The circulation is confined to the
circulations arise and evolve. layer through which an identifiable thermal gradient
exists. This is usually below any inversion or stable
layer capping the boundary layer at the coast, and the
Thermally Forced Coastal Circulations entire circulation typically occurs below 2-3 km. A
Surface heating differences are produced across a complete circulation is formed through conservation
coastline owing to the differing surface energy of mass as surface wind divergence (convergence) is
balances that occur for water and land surfaces, and established offshore (inland), which forces vertical
these differences initiate the most common coastal motions as well as a return flow aloft, as depicted in
circulation, the thermally driven circulation. Ther- Figure 1. The upward motion associated with the sea
mally driven circulations are very well understood breeze is an important contributor to the development
from a theoretical basis, and on larger-scales essen- of coastal convection and thunderstorms. The sea
tially form the basis for adjustments in quasi- breeze as depicted in Figure 1 may reverse its sense of
geostrophic flows. A simple dynamic understanding circulation at night when the land becomes cooler than
of the development of thermally forced circulations the adjacent ocean, although this is not an entirely
can be found in the following equation that describes
the time rate of change of a closed-path circulation in
the horizontal (cross-coast) and vertical plane:
The term on the right-hand side is referred to as the

solenoidal term and represents the generation of
circulation (or vorticity) by the fact that the pressure
and temperature structures are different across the
coastline. The right-hand side can be integrated
between a Po (lower) and P I (upper)pressure level to
yield the thickness between the pressure levels. This
results in a height gradient or pressure gradient across
Figure 1 Idealized sea breeze circulation. The solid thin lines are
the coast due to the cross-coast temperature differ-
isolines of constant potential temperature of the boundary layer,
ence. Consequently, the air accelerates towards the cool over water and warm over land. The heavy arrows depict the
lower height or pressure over the land area to force a sense of the circulation or winds. The approximate vertical scale in
cross-coast circulation. This is a classic result covered kilometers is given on the left.
symmetric response, because the boundary layer Topographically Forced Coastal

structure often becomes decoupled owing to the Circulations
surface cooling. These basic circulations are highly
modified by numerous factors, such as the direction of In considering the interaction of a flow with coastal
the larger-scale flow, distribution of both low and high topography, several general factors must be taken into
clouds, coastline shape, and boundary layer or deeper account. First is the nature of the topography itself - an
static stability (see Land-Sea Breeze). isolated mountain, a long, high mountain barrier, a
Another type of thermally forced coastal circulation gradually sloping coastal plain, etc. Each type of
is the wintertime coastal front that develops along coastal topography can dictate a particular response
parts of the eastern coast of the United States as well as for the same synoptic-scale flow regime. Second, the
elsewhere around the world. In this case, the ocean nature of the incident flow plays a big role in
areas are heated from below through surface heat determining the response to a given topographic
fluxes from the relatively warm ocean into overlying feature. Whether the air will go up and over a
cold air. This process establishes a coastal temperature mountain or be forced around it is critical in
gradient much like the land breeze situation. However, determining the type of topographically forced circu-
the coastal front forms as the large-scale winds change lation that may arise. Fundamentally, this depends
to a more coast-parallel direction, allowing frictional upon whether the airflow has sufficient kinetic energy
convergence to concentrate the coastal temperature to overcome the gravitational potential energy, which
gradient into a front. Figure 2 depicts an example of a are respectively associated with the incident wind
coastal front after the winds have shifted to an onshore velocity and the static stability, to flow over the
direction. Crucial to this process is the maintenance of mountain. For the statically stable atmosphere, the
the cold air over the land as the winds shift, which is restoring force acting on a lifted parcel is the difference
typically linked to the process of cold air damming due between the ambient lapse rate and the dry adiabatic
to inland mountains. The coastal front is a shallow lapse rate after lifting a parcel from its initial height to
boundary layer front unlike other fronts, and typically some new height such as the top of the mountain, hm.
behaves like a warm or stationary front, where warm, This restoring force is given directly from the Brunt-
moist, south-easterly flow comes off the ocean and is Vaisala frequency, N = ( g / & ) (dd/dz)1’2. Conse-
forced up and over a wedge of cold air inland from the quently the condition of whether a given air parcel
coastal front. This lifting is responsible for low clouds will go up and over a mountain depends on a Froude
and light precipitation on the inland side of the coastal number, Fr = U/Nh,, associated with the mountain.
front. Freezing rain, freezing drizzle, and even snow If Fr > 1 then the airflow will easily go over the
may occur on the cold side of the coastal front, and can mountain, which can result in a variety of mountain
greatly impact coastal metropolitan areas. wave responses. If Fr< 1 then the airflow will not be
able to go over the mountain, which suggests that the
flow is essentially blocked by the topography and must
either go around or be turned back.
For realistic three-dimensional atmospheric flows
characterized by a low Froude number, the extent and
onset of flow blocking depends upon the character of
the mountain barrier. Generally, steep mountains are
more effective at producing blocked flow, and the
effective steepness of the mountain depends upon the
stability, N, and location through the Coriolis para-
meter, f . A dynamically defined steepness, Nhm/fL,
can be defined and indicates that for larger values of N
a less steep mountain can cause blocking. Conse-
quently, the dynamic mountain steepness is funda-
mental in initiating flow blocking, and in higher wind
speeds blocking occurs only with steeper topography
or greater stratification. In addition to the occurrence
Figure 2 Depiction of an idealized coastal front along the New of blocking, the flow and mountain characteristics also
England coast. The solid lines depict the sea level pressure, with determine a distance upstream from a mountain
the axis of high pressure extending south over the inland area.
Dashed lines show the surface temperature in ’C,with cold barrier where the flow is blocked. The height of the
temperatures inland and a well-defined coastal temperature mountain and stratification along with the Cori-
gradient. olis parameter define a mountain Rossby radius,
L = Nh,/f, which identifies the distance upstream bances, while relatively infrequent (about 1-2 per
where the flow is influenced by the blocking. Within month in the summer), can rapidly change coastal
this Rossby radius the blocked flow is highly ageos- conditions from clear to foggy so as to impact on
trophic and the winds tend to point down the pressure airport operations as well as on beachgoers.
gradient. This Rossby radius also identifies a distinct Another example of a summertime coastal circula-
coastal zone where low-Froude-number, topographi- tion that arises owing to flow blocking by a strong
cally forced flows occur (see Mountain Meteorology). low-level inversion and coastal mountains is the
One example of a low-Froude-number flow that development of a coastal jet and associated regions
produces a significant interaction between the coastal of enhanced low-level winds. Figure 4 illustrates the
topography and the atmospheric boundary layer situation that occurs along the US West Coast, where
occurs along the West Coast of the United States as the strong boundary layer inversion slopes down
well as other locations characterized by a cold ocean toward the coast to establish a strong cross-coast
and significant coastal topography. Figure 3 shows a thermal gradient. While coastal jets occur where there
narrow tongue of coastal stratus and southerly winds are persistent coastal temperature gradients, in this
that is observed to propagate up the coast. This stratus case the coastal jet develops below this inversion as a
surge occurs when the summertime marine boundary coast-parallel flow because of both a strong thermal
layer off the coast is characterized by a strong low- gradient and blocking by the coastal mountains. The
level inversion and a well-mixed boundary layer. The thermal gradient sets up a strong cross-coast pressure
coastal disturbance is initiated after a period of gradient that may initially result in a cross-coast sea
significant offshore flow, in which the marine bound- breeze circulation, but as the pressure gradient persists
ary layer is cleared out and the inversion base is beyond a single diurnal cycle strong along-coast winds
lowered to the surface, resulting in cloud-free condi- are established in a more quasi-geostrophic response.
tions. As the offshore flow continues, a lee trough As a consequence of the blocking, this strong, along-
develops at a particular location along the coast as a coast, low-level jet typically behaves in a manner
weak mesoscale low. This sets up a south-to-north analogous to hydraulic flow through a channel where
pressure gradient force that forces the air to flow up the changes in channel width result in flow speed increases
coast from the south because the coastal mountains or decreases. For the coastal jet, the strong inversion,
block the airflow. Evidence suggests that these inland blocking by topography, and an onshore-
disturbances propagate as coastally trapped Kelvin directed pressure gradient force dynamically produce
waves as long as the south to north pressure gradient this channel and regions of enhanced winds in the lee
remains. The air cannot move inland, owing to the of capes and points where the channel characteri-
blocking by the coastal mountains. These distur- stics change. These regions of enhanced winds are a
Figure 3 Visible satellite image showing a narrow tongue of stratus clouds along the California coast. Ship and buoy reports are also
shown and indicate northerly winds in the clear air and southerly winds in the cloudy region. This cloud feature and the associated winds
propagate north over time.
Froude number is low. In the presence of relatively

strong stable stratification in the lower atmosphere,
the flow must be blocked and subsequently channeled
around topographic features. Figure 5 illustrates
increased wind speeds downwind from gaps in a
topographic barrier. These gaps in coastal mountain
ranges act as channels through which the air can flow
that would otherwise be blocked upstream. Low
pressure occurs offshore from these gaps, and the
flow undergoes acceleration owing to the pressure
gradient through the gap. Very high winds can occur
near the mouths of these coastal gaps, and these
regions of high winds may extend considerable
distances offshore, as seen in Figure 5.
The other extreme in terms of flow interaction with
topography is when Fr >> 1 and the winds are strong
enough for a given stratification to easily make it up
over the mountain. In this case, the air is forced up the
mountain on the incident side by mechanical lifting
Figure 4 Three-dimensional view of the California topography
and down the other side by negative buoyancy. While
viewed from the south-west. The solid, sloping plane depicts the
hypotheticaltopology of the marine boundaly layer inversion base. this simple reasoning suggests a symmetric response
The axis of the strongest low-level winds (the coastal jet) is up- and downstream from the mountain, the actual
depicted by the arrow below the inversion and along the coast. response of a flow across a mountain barrier depends
upon a variety of factors, including the width of the
mountain, stratification distribution, and incident
manifestation of the flow adjustment to the coastal wind speed. When the wind is either strong a n d o r
topography under conditions where the flow speed is the mountain is narrow, the flow cannot adjust
faster than the propagation of gravity waves. geostrophically, and vertically propagating gravity
Flow around topography and through gaps in the waves are excited by the cross-barrier flow. The
topography along coastlines are other examples of detailed structure of the wave response depends
topographically forced coastal circulations where the critically on the stratification and wind profiles around
Figure 5 Satellite-derivedsurface wind speeds near Japan. Gaps through the Japanese mountains and Islands are noted by the heavy
arrows. The winds are from the north-west and regions of higher winds are shown downstream from these gaps.
504 COLD AIR DAMMING
the mountain barrier. However, the adiabatic cooling circulations produce unique local flow, cloud,
due to lifting on the upstream side of the mountain and and precipitation patterns near the coast. The dynamic
warming on the downstream side often cause the processes associated with thermal forcing and
development of a windward ridge and leeward trough topographic forcing, while fundamental drivers of
in the sea-level pressure. When the wind is either weak many atmospheric circulations, generate the many
and/or the mountain is broad, the flow tends to adjust unique coastal circulations owing to the complex
geostrophically as it crosses the barrier and a pressure geometry, topography, and boundary layer structure
ridge develops over the mountain owing to the that characterize the coastal environment on small
adiabatic cooling as the air is lifted over the barrier. scales.
The flow in this case is deflected anticyclonically on the
windward side and back cyclonically on the lee side.
Since many coastal mountains are relatively narrow, Nomenclature
coastal lee troughs and windward ridges are most f Coriolis parameter (units: s - l )
often associated with this topographically forced flow Fr Froude number (non-dimensional)
regime. g Acceleration of gravity (units: m sP2)
An example of a high-Froude-number, topographi- him Height of the topography (units: m)
cally forced coastal circulation is the development of L Upstream length scale (units: m)
mesoscale lee vortices or eddies along the coast. N Brunt Vaisala frequency (units: s - ')
Catalina eddies along the southern California coast, U Flow speed (units: m s - ')
Olympic lee lows in western Washington state, and 0 Potential temperature (units: "C)
other coastal lows around the world arise owing to the 00 Reference potential temperature (units:
flow over the associated coastal mountain features.
Downslope flow and associated lee troughing are
primarily responsible for the development of these
See also
types of mesoscale eddies. These mesoscale coastal
lows are often responsible for regions of enhanced Energy Balance Model, Surface. LandSea Breeze.
winds (Olympic low) or flow reversals as noted above Mountain Meteorology.
for stratus surges. Other smaller-scale vortices that
occur in the wakes of islands represent a different
Further Reading
dynamic process and are associated with shear along
the edges of faster flow around a mountainous island Holton JR (1992) Introduction to Dynamic Meteorology.
when stronger stratification is present. New York: Academic Press.
The variety of coastal weather is the result of National Research Council (1992) Coastal Meteorology:
the complex interaction between cross-coast differ- A Review of the State of the Science, Washington, DC:
National Academy Press.
ences in surface properties, coastal mountains, and Ray P (ed.) (1988)Mesoscale Meteorology and Forecasting.
larger-scale weather patterns. Circulations arise Boston, MA: American Meteorological Society.
within a narrow zone along the coast owing to thermal Simpson JE (1994)Sea Breeze and Local Winds. New York:
and moisture gradients, flow blocking, trapping, Cambridge University Press.
orographic waves, and other dynamic processes Whiteman CD (2000) Mountain Meteorology: Fundamen-
due to flow interaction with topography. These tals and Application. New York: Oxford University Press.
B A Colle, State University of New York at Stony Brook, range. This results in a cold dome adjacent to the
Stony Brook, NY, USA barrier, which can often be identified by a pronounced
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. 'nose' in the sea-level pressure pattern and an en-
hancement in the terrain-parallel flow, commonly
referred to as a 'barrier jet'. The temperatures within
Introduction the damming region can be more than 20°C cooler
than surrounding areas of equal elevation. Damming
Cold air damming is the process whereby cold air events typically occur during the cool season when an
becomes entrenched along the slopes of a mountain along-barrier pressure gradient is created by a cold
the mountain barrier. However, the adiabatic cooling circulations produce unique local flow, cloud,
due to lifting on the upstream side of the mountain and and precipitation patterns near the coast. The dynamic
warming on the downstream side often cause the processes associated with thermal forcing and
development of a windward ridge and leeward trough topographic forcing, while fundamental drivers of
in the sea-level pressure. When the wind is either weak many atmospheric circulations, generate the many
and/or the mountain is broad, the flow tends to adjust unique coastal circulations owing to the complex
geostrophically as it crosses the barrier and a pressure geometry, topography, and boundary layer structure
ridge develops over the mountain owing to the that characterize the coastal environment on small
adiabatic cooling as the air is lifted over the barrier. scales.
The flow in this case is deflected anticyclonically on the
windward side and back cyclonically on the lee side.
Since many coastal mountains are relatively narrow, Nomenclature
coastal lee troughs and windward ridges are most f Coriolis parameter (units: s - l )
often associated with this topographically forced flow Fr Froude number (non-dimensional)
regime. g Acceleration of gravity (units: m sP2)
An example of a high-Froude-number, topographi- him Height of the topography (units: m)
cally forced coastal circulation is the development of L Upstream length scale (units: m)
mesoscale lee vortices or eddies along the coast. N Brunt Vaisala frequency (units: s - ')
Catalina eddies along the southern California coast, U Flow speed (units: m s - ')
Olympic lee lows in western Washington state, and 0 Potential temperature (units: "C)
other coastal lows around the world arise owing to the 00 Reference potential temperature (units:
flow over the associated coastal mountain features.
Downslope flow and associated lee troughing are
primarily responsible for the development of these
See also
types of mesoscale eddies. These mesoscale coastal
lows are often responsible for regions of enhanced Energy Balance Model, Surface. LandSea Breeze.
winds (Olympic low) or flow reversals as noted above Mountain Meteorology.
for stratus surges. Other smaller-scale vortices that
occur in the wakes of islands represent a different
Further Reading
dynamic process and are associated with shear along
the edges of faster flow around a mountainous island Holton JR (1992) Introduction to Dynamic Meteorology.
when stronger stratification is present. New York: Academic Press.
The variety of coastal weather is the result of National Research Council (1992) Coastal Meteorology:
the complex interaction between cross-coast differ- A Review of the State of the Science, Washington, DC:
National Academy Press.
ences in surface properties, coastal mountains, and Ray P (ed.) (1988)Mesoscale Meteorology and Forecasting.
larger-scale weather patterns. Circulations arise Boston, MA: American Meteorological Society.
within a narrow zone along the coast owing to thermal Simpson JE (1994)Sea Breeze and Local Winds. New York:
and moisture gradients, flow blocking, trapping, Cambridge University Press.
orographic waves, and other dynamic processes Whiteman CD (2000) Mountain Meteorology: Fundamen-
due to flow interaction with topography. These tals and Application. New York: Oxford University Press.
B A Colle, State University of New York at Stony Brook, range. This results in a cold dome adjacent to the
Stony Brook, NY, USA barrier, which can often be identified by a pronounced
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. 'nose' in the sea-level pressure pattern and an en-
hancement in the terrain-parallel flow, commonly
referred to as a 'barrier jet'. The temperatures within
Introduction the damming region can be more than 20°C cooler
than surrounding areas of equal elevation. Damming
Cold air damming is the process whereby cold air events typically occur during the cool season when an
becomes entrenched along the slopes of a mountain along-barrier pressure gradient is created by a cold
COLD AIR DAMMING 505
anticyclone passing to the north (south) of the barrier limit (i.e., when Coriolis effects are to be considered),
slope in the Northern (Southern) Hemisphere. When theoretical studies have shown that the Burger number
the favorable synoptic conditions persist, the cold air ( B = H N / L f , where L is the mountain half-width and
mass is often observed to remain in a quasi-steady f is the Coriolis parameter) is also an important
state. parameter to diagnose potential blocking. For situa-
Long, continuous mountain ranges obstruct tions of equal stratification and latitude, the Burger
the low-level flow of cold air and can result in number is proportional to the steepness of the moun-
damming. The structure and evolution of cold tain slope ( H I L ) . When B < 1 , the flow is quasi-
air damming is determined by the ambient flow geostrophic/semi-geostrophic and the flow can pass
speed and direction approaching the barrier, terrain over the mountain; however, when B > 1, the barrier is
height, low-level density stratification, and latitude. considered to be ‘hydrodynamically steep’ and the
The terrain-normal wind speeds, stratification, and flow is blocked by the barrier.
terrain height determine whether the low-level flow is When Fr < 1 or B > 1, blocked ageostrophic down-
blocked by the barrier. Since damming events typically gradient flow is forced by an along-barrier pressure
last more than several hours, planetary rotation limits gradient imposed by the synoptic-scale pressure field.
the horizontal extent of the damming and determines Using scale analysis on the momentum equations,
the evolution of the momentum balances near the when I/L (cross-barrier length scale/along-barrier
terrain. length scale) is small and the Rossby number R, =
Cold air damming is common along many moun- V / f l (where V is the magnitude of the along-barrier
tainous regions around the world, such as the flow) is approximately equal to or greater than one,
Appalachians and Rockies in North America. The the winds are primarily terrain-parallel within
development of the cold dome during these events approximately a Rossby radius of deformation,
results in several operational forecast challenges such 1~ = N H / f , of the barrier. For example, for the
as freezing rain, heavy snow, ‘coastal fronts’, and cold Appalachian mountains of North America, 1~ is
or cool surges. Numerical models need to have typically around 150 km (where H N 1000 m,
significant horizontal resolution (i.e., < 40 km grid N N 0.015 s-l, and f N s-l). Within a Rossby
spacing) in order to resolve many cold air damming radius of the terrain, an antitriptic balance typically
events. develops in the along-barrier direction between fric-
tion and the pressure gradient. Meanwhile, in the
cross-barrier direction the mass field adjusts under the
Nondimensional Flow Parameters influence of rotation (e.g., geostrophic adjustment)
Cold air damming is dependent on the amount of cold such that the pressure gradient normal to the terrain
air blocked by an elongated mountain range. Flow balances the Coriolis force associated with the terrain-
blocking by a barrier is related to the ratio of the parallel winds. Since the Rossby radius depends on the
kinetic energy of the ambient flow to the potential latitude (as defined by f ) ,cold air damming is able to
energy needed to get the air parcel above the mountain extend further away from the barrier closer to the
top. This ratio is given by the Froude number, Fr, Equator given identical stratification and mountain
where height.
Fr := U / N H PI Structure and Evolution

In eqn [l], U is the speed of the average flow Antecedent Conditions
impinging towards the barrier below crest level, N is
the Brunt-Vaisala frequency, and H is the mountain The interaction of the synoptic-scale flow with an
height. F? may be interpreted as the ratio of inertial elongated mountain barrier acts to initiate cold air
forces, U ( U / H ) ,to buoyancy forces, g(AO/B) = N2H. damming. Typically, significant damming events are
Large Froude numbers (Fr >1)indicate flow that is associated with an upper-level trough and an associ-
able to surmount the barrier while low Froude ated cold surface anticyclone crossing immediately to
numbers ( F r < l ) indicate blocked flow. Many cold the north (south) of the barrier in the Northern
air damming episodes are characterized by an Fr less (Southern) Hemisphere (Figure 1A). (For purposes of
than 0.5, which is indicative of significant flow illustration, the evolution of cold air damming in the
blocking. Northern Hemisphere is shown). For damming along
The interaction of flow with mountains is not the Appalachians, there is often split westerly flow
uniquely determined by the ratio of the kinetic to aloft, with a secondary upper-level trough and an
potential energy as expressed by Fr. In the rotating associated surface cyclone situated to the south west of
200km I\ I
Figure 1 Horizontal evolution of cold air damming for the (A) initiation, (B) development, (C) mature, and (D) decay stages. Sea-level
pressure (solid evely 4 mb), temperature (dashed every 5"C), and surface winds are shown. Line segment AB in (A) is the location of the
cross-section shown in Figure 2.
the barrier. For other barriers, such as the Rocky cold advection is typically much larger than upslope
Mountains and Andes, significant cross-barrier flow at adiabatic cooling during damming events. The rapid
crest level to the south of the anticyclone often results decrease in cold advection with height results in the
in downslope warming and a lee trough (Figure 1A). shallow cold air being capped by an isothermal or
For either situation this high-low couplet enhances the inversion layer. Meanwhile, a Coriolis torque acting
along-barrier pressure gradient. on this donwgradient flow results in an upslope
flow component near the barrier slope, which helps
push cold air up the mountain slope. For some cases
Initiation and Mature Phases
the synoptic-scale low-level flow will also have a
The along-barrier pressure gradient results in ageo- geostrophic wind component directed towards the
strophic nearly terrain-parallel flow adjacent to the barrier.
barrier (Figure lA, B), which advects cold air south- Because of the significant low-level stability and
ward along the barrier slope, while above the moun- significant mountain slope, the Froude number (Fr) is
tain there may be neutral or even warm advection. typically<l and Burger number ( B ) is>>1, which
Because of the significant cold air source to the north, results in blocking of the cold air before it reaches the
COLD AIR DAMMING 507
~ ~~
mountain top. This blocking prevents the develop-

ment of geostrophic balance in the along-barrier
direction and a wedge of cold air (cold dome) develops
adjacent to the mountain. This cold dome hydrostat-
ically results in a pressure ridge along the barrier which
extends a Rossby radius ( 1 ~ upstream
) of the crest
(Figure 1C). In the along-barrier direction the
flow becomes balanced by the pressure gradient
and friction processes. Because of the strong static
stability and weak vertical mixing, most of the friction
is the result of surface drag. In the cross-barrier
direction a geostrophic balance develops between
the cross-barrier pressure gradient and the Coriolis
force.
To demostrate this adjustment process and vertical
structure of the damming process, Figure 2 shows
the horizontal flow and thermal evolution in a
cross-section taken normal to the mountain range.
Prior to the damming event, the isentropes are
nearly horizontal adjacent to the mountain with fairly
weak stability near crest level. For many cases there
may be a preexisting cold stable layer from a previous
damming event or a depression in the isentropes
from adiabatic warming associated with downslope
flow (Figure 2A). At this intial time there is little flow
in the along-barrier direction. Once the along-
barrier pressure gradient becomes more established,
ageostrophic northerlies develop and the Coriolis
force turns the flow in an upslope direction (Figure Figure 2 Vertical evolution of cold air damming for the
2B). The isentropes begin to tilt in response to the (A) initiation, (B) development, and (C) mature stages for cross-
cold air being advected up the barrier. The low- section AB showing the potential temperature (solid every 4 K) and
horizontal winds in the section. The location of AB is shown in
level cold advection and weak upslope cooling in Figure 1A. The letter 'J' shows the position of the terrain-parallel
concert with neutral or warm advection above results barrier jet.
in a stable layer intensifying near the top of the cold
dome.
During the next few hours the mass field associated The cold dome and barrier jet associated with
with the blocked flow adjusts to planetary rotation damming persists as long as there is a cold-air source
and extends approximately a Rossby radius upstream feeding the cold dome. Meanwhile, strong warm
of the barrier (Figure 2C). By this time cold air advection may occur to the east and over the cold
damming and the cold dome are firmly established dome (not shown), which results in air parcels
adjacent to the barrier. The cold dome is deepest near ascending up and over the cold dome. For some
the bottom of the slope, which hydrostatically corre- events, such as Appalachian damming, precipitation
sponds to the axis of the sea-level pressure ridge falling from the saturated flow over the cold dome
(Figure 1B). The damming regime is often character- results in evaporative cooling and further strengthen-
ized by neutral or conditionally unstable lapse rates in ing or maintenance of the cold dome.
the lower layer (cold dome) and is capped by an
inversion. The terrain-parallel flow within this cold Dissipation Phase
dome is maximized where the greatest tilt of
the isentropes and resulting cross-barrier pressure Cold air damming dissipates when the cold surface
gradient exists. The core of maximum winds (labeled J anticyclone moves away from the barrier, and the cold
in Figure 2C), which can frequently exceed 20 m s-', air source is lost (Figure 1D).The pressure falls to the
is often referred to as a barrier jet. The development north decrease the along-barrier pressure gradient and
of the barrier jet feeds more cold air into the dam- the flow within the cold dome weakens. The cold dome
ming region, therefore helping to sustain the cold slowly weakens through diurnal heating and mixing of
dome. warmer air from aloft. The weakening may be
expedited if surface pressure falls occur to the east of along Vancouver Island British Columbia, the east
the barrier, which results in a wind component and slope of the Washington Cascades, and along the
advection of cold air away from the mountain. In some eastern Sierras. During these cases cold air is pooled up
locations, particularly for large mountain barriers against the barrier from a previous excursion of
such as the Rockies and Himalayas, the cold dome cold air from the interior of the continent. The
may be eroded by significant cross-barrier and down- along-barrier pressure gradient is often enhanced
slope flow near crest level (e.g., ‘Chinook’ or ‘Foehn’ with the approach of a surface front, resulting in an
winds). enhancement of the terrain-parallel ageostrophic flow
(barrier jet).
For maritime climates, such as coastal California
Cold Air Damming Examples around and Australia, damming occurs when cool marine
air is rotationally blocked against the coastal terrain.
the World Typically, these events are associated with a transition
There have been many documented studies of cold air from warm dry offshore flow to terrain-blocked
damming around the world. The most widely studied cool marine adjacent to the coast. These coastal
area has been along the eastern side of the Appalachian marine surges often result in a narrow stratus tongue
Mountains, where the ‘wedge ridge’ sea-level pressure that extends approximately a Rossby radius off the
pattern is sometimes referred to as the ‘Baker’ ridge. coast.
Because the Appalachians are within a few hundred For many years cold air damming could not
kilometers of the relatively warmer Atlantic Ocean, be resolved by operational numerical weather models,
the low-level temperature gradient near the coast can such as NCEP’s (National Centers for Environmental
be enhanced through confluence between the nearly Prediction) Nested Gridded Model (NGM) and
terrain-parallel flow associated with cold air damming Global Spectral (MRF),since horizontal grid spacings
and the warmer more geostrophic flow near the coast. of these models are around 80km. More recent
This frontogenetical process helps in the development operational or research mesoscale models (i.e.,
of an enhanced temperature gradient near the coast NCEP’s Eta model at 32 km resolution and the Penn
called a ‘coastal front’. The low-level temperature State-National Center for Atmospheric Research
advections associated with these shallow baroclinic MM5) are able to realistically forecast different
zones result in mixed precipitation, which can be in the damming events.
form of freezing rain (ice storms) and sleet within areas
of the cold dome.
Cold air damming has recently been documented for See also
other large mountain barriers, such as the Rockies and Anticyclones. Aviation Weather Hazards. Boundary
the Andes. For these extremely long N-S barriers, Layers: Complex Terrain. Coastal Meteorology.
relatively cold air can propagate into tropical latitudes Coriolis Force. Dynamic Meteorology: Balanced
(10”NE). This channeling of cold air is often referred Flows. Mountain Meteorology. Static Stability.
to as a ‘cold surge’, in which a cold anticyclone is Weather Prediction: Regional Prediction Models.
moving Equatorward adjacent to the mountains. The
primary mechanism for driving the surge southward is
the development of an along-barrier pressure gradient, Further Reading
cold advection, and the damming evolution described Baker DG (1970) A Study of High Pressure Ridges to the
above. The very high orography associated with the East of the Appalachian Mountains. PhD thesis, Massa-
Rockies, Andes and Himalayas often results in a chusetts Institute of Technology.
horizontal scale of damming that is typically much Bell GD and Bosart LF (1988) Appalachian cold-air dam-
larger (>300 km) than Appalachians and other more ming. Monthly Weather Review 116: 137-161.
narrow coastal barriers. Since the Rocky Mountains Bluestein HB (1993) Synoptic-Dynamic Meteorology in
are adjacent to the sloping Great Plains, flow blocking Midlatitudes; Observations and Theory of Weather
is also favored along the slope, resulting in a cold dome Systems, vol. 11, pp. 359-362. Oxford: Oxford University
that extends >500km east of the Rockies. As the Press.
Colle BA and Mass CF (1995)The structure and evolution of
North American surge moves southward along the cold surges east of the Rocky Mountains. Monthly
mountainous Mexican coast and the large-scale slope Weather Review 123: 2577-2610.
is lost, the scale of the damming collapses to a few Forbes GS, Anthes RA and Thomson DW (1987) Synoptic
hundred kilometers. and mesoscale aspects of an Appalachian ice storm
Cold air damming and associated barrier jets can associated with cold air damming. Monthly Weather
occur for other mountainous coastal regions, such as Review 115: 564-591.
CONTRAILS 509
Overland JE (1984) Scale analysis of marine winds in straits Stauffer DR and Warner TT (1987) A numerical study of
and along mountainous coasts. Monthly Weather Review Appalachian cold-air damming and coastal frontogene-
112: 2532-2536. sis. Monthly Weather Review 115: 799-821.
Pierrehumbert RT and Wyman B (1985) Upstream effects Xu Q (1990) A theoretical study of cold air
of mesoscale mountains. Journal of the Atmospheric damming. Journal of the Atmospheric Sciences 47:
Sciences 42: 977-1003. 2969-2985.
P Minnis, NASA Langley Research Center, Hampton, point or by precipitation into unsaturated layers below
VA, USA the flight level.
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. Another type of contrail that forms briefly at
warmer temperatures is composed of water droplets
that form behind the tips or the leading edges of
Introduction aircraft wings. These are commonly seen emanating
One of the most visible anthropogenic effects on the from fighter aircraft in high-speed maneuvers in a
atmosphere is the condensation trail, or contrail. humid atmosphere. In these cases, the ambient air is
These aircraft-induced clouds have become a common compressed at the wing tip and then expands quickly
sight since the 1960s because of increasing jet traffic, during adiabatic expansion within the low-pressure
but they were observed as early as 1919. Contrails area above the wing tip. The expansion temporarily
were frequently seen and filmed in World War I1 during cools the air sufficiently that it falls below the dew
bombing raids or dogfights. They were briefly studied point, resulting in condensation. Because ice contrails
in Germany during the war but drew little scientific are the more common variety, the liquid water
interest again until the early 1950s when the use of jet contrails are not considered further here.
aircraft by military and commercial aviation acceler- The basic concepts for determining the conditions
ated. Interest waned, with only sporadic studies until for contrail formation were developed independently
the 1990s when aircraft effects and contrails became by E. Schmidt in Germany during 1941 and H.
the foci of numerous research efforts. Concerns over Appleman in the United States during 1954. The lines
their impact on climate and aircraft visibility have in Figure 1 schematically illustrate the ice contrail
been the primary motivation for the recently intensi- formation process for several scenarios with the
fied research into contrails. Understanding their ambient temperatures T , and water vapor partial
effects requires knowledge of their physical and pressures e, indicated by the points at the lower end of
optical characteristics and how, when, and where each line. Each line extends to the temperature T, and
they form. water vapor partial pressure e, of the exhaust exiting
the engine. In cases defined by the lines I, 11, and IV, the
ambient water vapor pressure is less than the ice
saturation partial pressure ei, while in case 111, e, > ei.
Contrail Formation
In case I, the partial pressure exceeds ei during the
Contrails are generally composed of ice crystals with mixing but never reaches water saturation and a
trace amounts of exhaust products such as soot and contrail does not develop. A short-lived contrail would
sulfates. The contrail ice crystals form because the develop in case I1 because, at point F, the mixture
relative humidity with respect to liquid water, U,, temperature TFcoincides with the liquid water satu-
temporarily reaches the saturation point in the plume ration partial pressure e,. The contrail would form
mixture of ambient air and hot exhaust gases. Tiny when the plume temperature reached TFand would
droplets develop on background aerosols or on aero- persist until the plume partial pressure decreased to a
sols formed by exhaust compounds. Because the value below ei at approximately - 42°C. A long-lived,
ambient temperatures required for formation of con- persistent contrail would form in case I11 because the
trails are generally less than -40°C, the small water ambient air is supersaturated with respect to ice.
droplets instantly freeze and grow via vapor-to-ice Because saturation conditions cover a greater range of
deposition as long as the relative humidity with respect temperatures after initial formation, the contrail
to ice, Vi, remains above the saturation point. They formed in case IV would probably last longer than
dissipate via sublimation if Ui is below the saturation that in case 11.
CONTRAILS 509
Overland JE (1984) Scale analysis of marine winds in straits Stauffer DR and Warner TT (1987) A numerical study of
and along mountainous coasts. Monthly Weather Review Appalachian cold-air damming and coastal frontogene-
112: 2532-2536. sis. Monthly Weather Review 115: 799-821.
Pierrehumbert RT and Wyman B (1985) Upstream effects Xu Q (1990) A theoretical study of cold air
of mesoscale mountains. Journal of the Atmospheric damming. Journal of the Atmospheric Sciences 47:
Sciences 42: 977-1003. 2969-2985.
P Minnis, NASA Langley Research Center, Hampton, point or by precipitation into unsaturated layers below
VA, USA the flight level.
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. Another type of contrail that forms briefly at
warmer temperatures is composed of water droplets
that form behind the tips or the leading edges of
Introduction aircraft wings. These are commonly seen emanating
One of the most visible anthropogenic effects on the from fighter aircraft in high-speed maneuvers in a
atmosphere is the condensation trail, or contrail. humid atmosphere. In these cases, the ambient air is
These aircraft-induced clouds have become a common compressed at the wing tip and then expands quickly
sight since the 1960s because of increasing jet traffic, during adiabatic expansion within the low-pressure
but they were observed as early as 1919. Contrails area above the wing tip. The expansion temporarily
were frequently seen and filmed in World War I1 during cools the air sufficiently that it falls below the dew
bombing raids or dogfights. They were briefly studied point, resulting in condensation. Because ice contrails
in Germany during the war but drew little scientific are the more common variety, the liquid water
interest again until the early 1950s when the use of jet contrails are not considered further here.
aircraft by military and commercial aviation acceler- The basic concepts for determining the conditions
ated. Interest waned, with only sporadic studies until for contrail formation were developed independently
the 1990s when aircraft effects and contrails became by E. Schmidt in Germany during 1941 and H.
the foci of numerous research efforts. Concerns over Appleman in the United States during 1954. The lines
their impact on climate and aircraft visibility have in Figure 1 schematically illustrate the ice contrail
been the primary motivation for the recently intensi- formation process for several scenarios with the
fied research into contrails. Understanding their ambient temperatures T , and water vapor partial
effects requires knowledge of their physical and pressures e, indicated by the points at the lower end of
optical characteristics and how, when, and where each line. Each line extends to the temperature T, and
they form. water vapor partial pressure e, of the exhaust exiting
the engine. In cases defined by the lines I, 11, and IV, the
ambient water vapor pressure is less than the ice
saturation partial pressure ei, while in case 111, e, > ei.
Contrail Formation
In case I, the partial pressure exceeds ei during the
Contrails are generally composed of ice crystals with mixing but never reaches water saturation and a
trace amounts of exhaust products such as soot and contrail does not develop. A short-lived contrail would
sulfates. The contrail ice crystals form because the develop in case I1 because, at point F, the mixture
relative humidity with respect to liquid water, U,, temperature TFcoincides with the liquid water satu-
temporarily reaches the saturation point in the plume ration partial pressure e,. The contrail would form
mixture of ambient air and hot exhaust gases. Tiny when the plume temperature reached TFand would
droplets develop on background aerosols or on aero- persist until the plume partial pressure decreased to a
sols formed by exhaust compounds. Because the value below ei at approximately - 42°C. A long-lived,
ambient temperatures required for formation of con- persistent contrail would form in case I11 because the
trails are generally less than -40°C, the small water ambient air is supersaturated with respect to ice.
droplets instantly freeze and grow via vapor-to-ice Because saturation conditions cover a greater range of
deposition as long as the relative humidity with respect temperatures after initial formation, the contrail
to ice, Vi, remains above the saturation point. They formed in case IV would probably last longer than
dissipate via sublimation if Ui is below the saturation that in case 11.
510 CONTRAILS
80
I
' / /
/
-65 -60 -55 -50 -45 -40 -35 -30 -25

Temperature ("C)
Figure 1 Phase diagram with mixing lines for aircraft exhaust in different ambient conditions.
Although contrail formation has been observed The mixing line slope depends on the specific plume
at temperatures as great as -36"C, it is clear from enthalpy h, and the water vapor mixing ratio q, which,
Figure 1 that contrails form more easily at lower in turn, are related to the emission index EI,, mass
temperatures. The threshold temperature TT for specific combustion heat Q, and the overall engine
contrail formation is defined as the warmest ambient efficiency y. The relation is given specifically as eqn [3],
temperature that will support contrail formation for a where cp is the specific heat capacity, p is the pressure,
given value of e, and the exhaust parameters T, and e,. and E = 0.622. The emission index, the mass of water
The latter quantities determine the mixing line slope, produced per mass of combusted fuel, accounts for Aq
G, and are functions of engine type, operating condi- since e, >> e,.
tions, and fuel, while the value of e, can be determined
from vertical profiles of atmospheric and dew point
temperatures. In case 11, the ambient temperature at
point T is the contrail formation threshold tempera-
ture for the given values of e, and the mixing line slope
G. That is, the ambient temperature enabling contrailThe enthalpy differential is also determined almost
formation would have to change if either e, or G entirely by Q and y because the ambient heat is
varied and, therefore, TT is unique for each pair of e,
negligible compared to that produced by the engine.
and G. The threshold temperatures are greater than T,Since Q and E I , can be determined for a given fuel, the
for cases IV and 111, and less than T , for case I. To find
overall efficiency, the ratio of propulsion energy to
TT for a particular slope and e,, it is necessary to total combustion energy is the primary variable
determine the tangent point TFfor a line having slopeaffecting the mixing line slope. The slope of the line
G with the curve describing the variation of e , with T .
increases with increasing efficiency. Each type of
Given a value of G, the threshold temperature can be engine has a nominal efficiency that is based on
computed for TFbetween - 10°C and - 60°C using stationary operating conditions. The overall efficien-
eqn [I]. cy, however, may vary for a given engine because of
different airframes, maintenance, and operating con-
+
TF = - 46.46 9.43 In( G - 0.053) + 0.720 ditions. Figure 2 illustrates the impact of efficiency for
x [In(G - 0.053)12 a given set of ambient conditions. In this instance y2 is
[11
slightly less than y l , resulting in a contrail from the
where G is given in PaK-'. The threshold tempera- plane with y1 and no contrail from the one with yz.
ture for any value of U , or e, can be determined Thus, two planes flying in the same environment can
iteratively with eqn [2]. produce two different results. Similarly, a plane might
produce a contrail when it is cruising but not when it is
ew ( T F ) - U w ew ( TT) ascending, depending on the effect of acceleration on
TT = TF - 121 the efficiency.
G
CONTRAILS 511
80 1
/I
O} I I I 1
-50 -45 -40 -35 -30 -25

Temperature ("C)
Figure 2 Hypothetical mixing lines for different propulsion efficiencies.
Contrails typically form at a distance of about Contrail Growth and Structure

3 0 m or less behind the aircraft engines where the
turbulent mixing sufficiently reduces the temperature. Once formed, a contrail develops or dissipates in the
The latest research results indicate that the initial same manner as a naturally generated cirrus cloud.
condensation of the supercooled droplets takes Growth and spreading of contrails depend on the
place on a wide variety of particles, including exhaust thickness of the supersaturated layer, the degree of ice
products such as sulfate aerosols, soot, and metallic supersaturation, and the wind speed and shear. When
particles as well as ambient mineral aerosols. When contrails persist, the particles typically grow to
the contrails are about 1minute old, the mean particle 30-1000 pm, sizes usually associated with natural
radius is around 2pm. A wide variety of particle cirrus clouds. Ice particle growth is rapid in highly
shapes have been observed in young contrails, includ- supersaturated layers and results in fall streaks that
ing hexagonal columns and plates, triangles, irregular spread horizontally in lower layers according to the
forms, and spheroids. Young contrails often appear wind shear. Figure 3A shows a cross-section view of a
saw-toothed or appear to have a cellular structure hypothetical persistent contrail growing and spread-
that results from the vortices formed by the aircraft. ing in the absence of vertical wind shear. It spreads
This structure provides irregularities for formation mostly by turbulent mixing induced by the aircraft
of local convective cells or radiative cooling vortex or by radiative processes. When wind shear is
gradients that aid mixing of the contrail with the present (Figure 3B), it will also spread horizontally by
ambient air. precipitation into the lower layers. If the crystals fall
ea < ei
t2 t3 t4 t5
512 CONTRAILS
into supersaturated air below, they will continue to Because water vapor and temperature are not
grow or, possibly, split into additional crystals. The homogeneously distributed, even at relatively small
linear shape of the contrail will be distorted and the scales (-loom), contrails may form or persist in an
contrail will soon look like a natural cirrus cloud to the apparently erratic fashion, as shown in Figure 4.For
observer. Well-aged contrails are often indistinguish- example, an on-off pattern can occur as an aircraft
able from natural cirrus clouds regardless of shear flies through a moist layer disturbed by a vertical wave
conditions. or even weak convective plumes. The contrails in
Most studies indicate that the number of crystals in Figure 4A form in the ascending parts of the wave or
a contrail remains constant after formation in super- plume where the temperature of the rising air falls
saturated conditions. Thus, if the contrail precipitates, below the threshold temperature, while in the de-
the contrail cloud at flight level might gradually fade as scending portions the air warms and dries, resulting in
its particles are depleted. If e, is just above ice no contrail formation. Similar patterns can result from
supersaturation, then the crystal growth will be a plane ascending or descending through several thin
limited and little precipitation will occur. In this layers that are near saturation but separated by dry
case, the contrail may spread slowly by diffusion, layers as in Figure 4B. The persistence of a contrail or
maintaining its linear shape for a relatively long time. parts of it depends on the value of e, relative to ei along
Because the crystals grow by deposition, the amount of the contrail line. Thus, parts of a contrail may dissipate
ice water in the contrail increases until the particles fall rapidly while other portions may linger and even grow.
out or equilibrium is reached between the ice water The local turbulence induced by the airframe, the
content and ei. Such equilibrium conditions generally atmospheric stability, and the wind vector also affect
do not last very long and the contrail eventually the morphology of the contrail.
dissipates. Although most persistent contrails have Photographs of the most familiar type, the short-
visible optical depths between 0.1 and 0.4, the values lived new contrail, are shown in Figure 5 . In both
are highly variable, ranging between 0.03 and 1. The cases, the pair of trails forming behind the aircraft
lifetimes of contrails are also extremely variable. gradually faded. In those situations, e, is only slightly
Short-lived contrails may only last a few seconds, less than ei. When e, exceeds ei, less familiar shapes
while some contrail-generated cirrus clouds have been can occur. Figure 6 shows examples of contrails at
tracked for more than 17 h. The shape, size, optical different stages of growth or persistence at the same
properties, and life cycle of contrails are highly time in different parts of the sky. To the north of the
dependent on their environment, so that a multitude observer (Figure 6A), contrails remain very thin and
of contrail morphologies can occur. Contrail-cirrus wispy at one end and dense and distorted at the other.
clouds are generally like natural cirrus clouds within a To the south-east (Figure 6B), a succession of slowly
few hours after their formation. spreading contrails appears off to the horizon.
(A) T1> T2> T3> T4 ea2< e,,
Figure 4 Schematic depiction of contrails forming in an on-off pattern.

CONTRAILS 513
Figure 5 Short-lived contrails.
Although these contrails appear to have little vertical satellite imagery (Figure 8) that was taken about
development, they persisted for at least several hours an hour before the photograph in Figure 7 .Subsequent
before advecting out of view. To the south-west (Figure imagery shows that these contrails dissipated
GC),many of the contrails are older and appear more downstream to the east while additional contrails
like natural cirrus clouds. Avery young thin contrail is formed within or beneath the advancing thin cirrus
evident in the upper left quadrant. clouds.
Condensation trails often form ahead of advancing Contrails can form within cirrus clouds, where they
fronts in the poleward flow of an upper level trough are manifest by reduced particle sizes or local thick-
where conditions are not quite saturated enough for ening of the cloud. Aircraft exhaust can also affect
natural cirrus development. In these instances they supercooled liquid water clouds. When a plane flies
can occur at multiple levels in the atmosphere because through this type of cloud, it introduces ice nuclei that
the formation conditions often cover a large depth cause freezing of the cloud droplets. The thermody-
of the atmosphere and air traffic uses a wide range of namic equilibrium shifts from a vapor-to-liquid to
altitudes. In Figure 7, contrails at the higher altitudes vapor-to-ice process, causing a rapid depletion of the
spread more than those below owing either to age, available water vapor onto the frozen droplets. The
the amount of wind shear, or the angle of the contrail newly formed ice crystals quickly grow large enough
to the ambient wind. Note the complex linear shadows to fall out of the cloud, resulting in a fall streak below
cast by the thin contrail on the left. These crossing the cloud and a gap within the cloud. This gap, called a
contrails are common in areas where air traffic lanes distrail, is linear when the plane flies for an extended
intersect. The contrails seen in Figure 7 are part distance within the cloud or oval shaped when the
of a larger contrail cluster that is easily observed in aircraft is briefly inside the cloud as it ascends or
Figure 6 Persistent contrails in various stages of growth and decay.

514 CONTRAILS
Figure 7 Persistent contrails observed over eastern Virginia, USA during 26 January 2001. Photo courtesy of L. Nguyen, NASA LaRC.
descends. Depending on the conditions, especially the Remote Sensing O f Contrails

original cloud thickness, a distrail will either persist or
be filled in with a new water droplet cloud. Distrails Although contrails are most often identified by
are most frequently observed in altostratus or their linear shapes both from ground observa-
altocumulus clouds. tions (Figure 7) and satellite imagery (Figure 8), these
Figure 8 Infrared 1-km resolution image of contrails over Virginia and Maryland, USA from the NOAA-16 Advanced Very High
Resolution Radiometer at 1832 UTC, 26 January 2001.
CONTRAILS 515
man-made clouds can take on other geometric shapes contrail, the observed radiance can be modeled simply
according to the particular flight patterns and winds. as in eqn [4].
For instance, spiral shapes result from a plane
in a circular holding pattern within an advecting [41
supersaturated layer, while a figure-of-eight can
form in similar layer if the plane flies a linear holding where Lc is the radiance emitted at the cloud temper-
pattern. The linear structure is most common ature T,,Lb is the upwelling radiance at the cloud
and forms the basis for identifying contrails. Because base with an equivalent brightness temperature Tb,
detection of contrails is important for various scientific and E>. is the cloud emissivity. In general, Lb > L,, so
applications, methods have been developed for that an increase in cloud transmissivity, (1 - E A ) ,
differentiating contrails from other linear clouds in results in more transmission of L b and a larger value
satellite imagery, the only plausible data source of LI,. Thus, Ti will be greater at some wavelengths
for studying the global effects of contrails. Automated than at others as long as the cloud is optically thin ( E
techniques for contrail detection typically create less than 0.9 or so). This effect can be seen in Figure 9,
an image of a parameter most likely to be associated which shows the 11pm image and an image of
with a contrail, then apply a variety of image brightness temperature difference between the 11pm
processing methods to find linear structures within and 1 2 p m channels on the NOAA-12 Advanced
that image. Very High Resolution Radiometer. For small ice
Such methods, which are still being researched, crystals, the extinction efficiency and, therefore,
usually take advantage of the relatively distinct infra- the optical depth at 1 2 p m can be as much as twice
red optical properties of younger contrails to compute that at 11pm. Thus, for small optical depths (<1.0),
a parameter that could have a distinctive contrail the transmission at 11pm is greater and the contrails
signal for image processing. Because of their relatively appear ‘warmer’ than at 12pm. Although some of
small size (effective diameters between 5 and 50 pm), the contrails are readily apparent in Figure 9A,
the ice crystals in contrails have extinction efficiencies many are obscured by other cirrus clouds. The
in the thermal infrared window region (8-14 pm) that temperature-difference image in Figure 9B reveals
vary much more with wavelength than the extinction many contrails that were not evident in the standard
efficiencies of larger particles typically found in most infrared image and highlights others more clearly.
cirrus clouds (effective diameters greater than 30 pm). Because the actual temperature difference contrast
Thus, young contrails transmit more radiation at depends on the effective particle sizes and optical
certain wavelengths than a cirrus cloud of equivalent depths of the surrounding clouds, and those quantities
optical depth, resulting in a signal that often reveals a are naturally variable, the contrails are not always
contrail. detected. Furthermore, other features such as cirrus
To better understand this effect, consider that streaks, coastlines, or cloud edges may produce similar
the satellite measures a spectral radiance LA that is signals.
recorded as an equivalent blackbody temperature When Tb is not very different from T,, such
TI, using the Planck function BA. For a cloud or techniques do not readily reveal the contrails because
Figure 9 1-km resolution infraredand infraredtemperature difference images of contrails over Virginia and North Carolina, USA from
the NOAA-12 Advanced Very High Resolution Radiometer at 2312 UTC, 29 October 1996.
516 CONTRAILS
the signal is so small. Therefore, contrails embedded in the 11pm image in Figure 10 shows no sign
relatively thick cirrus clouds cannot be seen in most of contrails, while the 3.7 pm image and the 11pm
temperature-difference imagery. However, during the and 3.7 pm temperature-difference image reveal a
daytime, contrails can be detected using temperature number of linear contrails. The 3.7pm image shows
differences between a channel near 11 pm and one in dark or ‘warm’ lines, while the temperature-difference
the ‘solar-infrared’ wavelength range (3.5-4.5 pm). At image reveals white contrails. Additional enhance-
those wavelengths, the satellite imagers measure an ment of the photograph would reveal more contrails
emission component and a solar-reflected component. in the temperature-difference image. The ability
The smaller contrail ice crystals reflect more sunlight to detect contrails in a thick cirrus cloud depends
than the surrounding cirrus crystals, resulting in on many factors, including the contrail age and its
a relatively large brightness temperature. For example, relative depth in the cloud as well as the particle sizes in
Figure 10 Contrails imbedded in thickcirrus over Kansas, USAfrom NOAA-14AdvancedVery High Resolution Radiometer data taken
at 2122 UTC, 23 March 2000.
CONTRAILS 517
the cirrus cloud, and the viewing and illumination favorable most often during winter and early spring
conditions. when the troposphere is coldest. During summer, the
Contrails can also be detected in high-resolution temperatures at flight levels are often too high to
visible and near-infrared imagery in certain condi- enable contrail initiation. Over areas poleward of
tions. For example, when not embedded in cirrus or about 50" latitude, the tropopause is often below flight
over lower clouds, a young contrail is often reflective level during winter, so that a significant number of
enough to be seen as a bright line in a 1-km resolution planes fly in the stratosphere, resulting in contrail
visible channel image. Sometimes, the contrail will suppression. Conditions are more favorable for con-
cast a shadow on lower clouds and can be detected trails during the summer and autumn in the subarctic
from its shadow. Near-infrared channels on or near regions. In the tropics, the altitude for contrails is
water vapor absorption lines, like one near 1.38 ym, generally above 11km year round, so the potential for
do not receive any significant reflectance from lower contrail formation by many commercial planes is
clouds, so that even very thin high clouds like contrails
reduced. However, persistent contrails are likely to
can be detected. Near-infrared methods are relatively occur more frequently in the tropics than at other
new and have not been developed as much as the latitudes at altitudes above l l k m because of the
infrared techniques. greater abundance of water vapor.
Determination of contrail properties such as Surface observations over the United States during
temperature, height, optical depth, and effective the 1990s indicated that persistent contrails occur,
particle size is accomplished with the same methods on average, approximately 9 % of the time, but
applied to the remote sensing of cloud properties. the frequency varies from less than 5% in low
Such techniques typically require multispectral traffic areas to 25% in the main air corridors.
imagers that can be used to simultaneously solve Approximately 80% of these persistent contrails
for E , Tc, and the particle size. When an insufficient are embedded in, extending from, or near natural
multispectral set is available, one or more of cirrus clouds. Contrail coverage has been derived
the parameters must be assumed in order to obtain from satellite imagery only for those contrails that are
a solution for the other parameters. These methods linear and large enough to observe in 1-km resolution
generally provide results consistent with in situ infrared data. Initial satellite-based estimates of
measurements. Both infrared and solar methods mean contrail amounts over Europe, the North
are also applicable to high-resolution imagery created Atlantic, and the conterminous United States are
by instruments on research aircraft. Active sensors 0.8%, 0.5%, and 1.8%, respectively, and roughly
such as lidar have been used to study the fine- 0.1% globally for the early 1990s. Similar values have
scale structures of contrails and their micro- been derived from theoretical calculations using real-
physical properties. They are often used to validate istic air traffic patterns, numerical analyses of mete-
the retrievals of contrails from passive satellite orological fields, and specified engine efficiencies.
imagers. Later studies suggest that the coverage may not be
as large as the initial satellite estimates owing to
possible false identification of natural clouds as
contrails.
Contrail-Cirrus Coverage Detection and assessment of contrail coverage
An increase in cirrus cloudiness due to contrail has been confined to contrails that are identifiable by
formation has been hypothesized since the beginning their linear structure and small particle sizes. Because
of the commercial jet age. The possibility of enhanced these identifying features are often lost as the contrails
cirrus coverage resides in the frequency and extent spread, the linear contrail coverage estimates repre-
of areas that are ice-supersaturated. In situ measure- sent the minimum amount of the sky that is covered
ments and numerical model analyses have shown by contrails. Geostationary satellite data are used
that e, exceeds ei 10-20% of the time in air at flight to track contrails as they grow and change in shape
altitudes (8-12 km). Thus, the potential exists and composition. Studies based on geostationary
for substantial increases in cirrus coverage over data indicate that the actual cirrus coverage generated
areas crossed by air traffic. Because the stratosphere by persistent contrails might be as large as a factor
is generally very dry, aircraft flying above the tropo- of 4 times the coverage estimated for younger,
pause generate few contrails, especially persistent linear contrails. However, the actual factor is probably
ones. somewhere between 1 and 3 . Determination of
The conditions necessary for supporting contrail contrail coverage and the resulting changes in
formation at flight altitudes change with the seasons. cirrus cloud amounts remains a topic of ongoing
Over mid-latitude areas, contrail conditions are research.
518 CONTRAILS
Climate Effects of Contrails of the system. For instance, if the contrail forms over a
dark background during midday, the amount of
Contrails, like other cirrus clouds, can affect both the reflected sunlight may exceed the amount of infrared
hydrological and radiation budgets. Many of the radiation blocked and reradiated by the cloud. Con-
possible contrail effects have only been the subject of versely, if it develops over a bright, hot surface (i.e.,
educated speculation, although some have been esti- desert) during the day, a contrail may reflect little
mated to some degree. Some of these potential effects additional radiation, but trap a significant amount
are mentioned here. because it is much colder than the surface. Its overall
By freezing out water vapor prior to the natural impact would be substantially different than that over
formation of cirrus clouds, contrails may alter the dark surface. At night, contrails warm the atmos-
the overall distribution of cirrus. General circulation phere. However, even during the day when solar and
model studies have shown that, if additional cirrus infrared forcing can almost cancel each other, the
cloud is specified in the air corridors, cirrus coverage contrail will still impact the radiation field because loss
decreases in other areas. Contrail formation of most of the blocked sunlight results in cooling of the
may decrease precipitation in some clouds by reducing surface, while much of the infrared or longwave
the average particle size in the affected clouds. radiation ‘trapped’ by the contrail warms the upper
Conversely, the precipitation induced by persistent troposphere and has little immediate impact on the
contrails in otherwise clear air (e.g., Figure 3A) may surface.
result in moistening of the middle layers of the These radiative forcing effects, estimated with
troposphere and drying of the atmosphere at flight several different models and assumptions, may result
altitudes. in a minor amount of global warming when averaged
Contrails reflect some solar or short-wave radiation over a long period or in some slight cooling on an
that would otherwise warm the surface, and absorb instantaneous basis. For example, Figure 11 shows the
outgoing infrared radiation that cools the surface- distribution of net contrail radiative forcing assuming
atmosphere system. The overall radiative impact or random contrailkloud overlap, an average contrail
forcing depends on the contrast between the contrail particle effective diameter of 24pm, and an optical
and its background, the lifetime and optical properties depth of 0.1. This estimate, based on air traffic
of the contrail, and the solar zenith angle when it is analyses for 1992, shows the areas of strongest
present. Depending on the solar zenith angle and the warming over the north-eastern United States and
contrast between the contrail and surface tempera- Europe. The maximum forcing of 0.35 W mP2is found
tures, the net forcing can result in cooling or warming over Europe, while the overall global net forcing for
Figure 11 Estimate of radiative forcing from linear contrails with a mean optical depth of 0.1 at 200 hPa for 1992.
CONTRAILS 519
Figure 12 Estimate of radiative forcing from linear contrails with a mean optical depth of 0.1 at 200 hPa for 2050 air traffic scenario.
this case is 0.0083 W m-2. Other scenarios yield values association with natural cirrus. Thus, current
between 0.0004 and 0.0203 Wm-2 for a range of estimates of their impact are highly uncertain. Never-
contrail coverage and altitudes with mean optical theless, their potential for affecting global climate
depths varying between 0.1 and 0.5 (Figure 12). and providing military intelligence has spurred
Airplane fuel consumption is expected to increase more interest and focused research into their forma-
between 1992 and 2050 by a factor ranging from 1.8 tion, dissipation, microphysical and morphological
to 14. The best-case scenario yields a factor of 4.3 characteristics, and methods for suppressing them.
for traffic above 500 hPa with a concomitant rise in Removal of fuel sulfur or use of liquid hydrogen
efficiency to q = 0.5. The greatest increases in fuels have been suggested as means for diminishing
air traffic are expected over eastern Europe and Asia. the number of cloud nuclei and, hence, the number of
The combination of engine efficiency and air traffic contrails. Tests and theoretical studies have shown
increase yields an estimate of global contrail coverage that such measures would probably not reduce
of 0.5% and a 0.0488 Wm-2 global net radia- the frequency of contrails. Hydrogen fuels would
tive forcing, with a maximum regional value of cause larger increases in local relative humidity in
1.46 W rn-’ over Europe. Other scenarios using the exhaust plume, causing higher supersatura-
different contrail radiative properties and fuel use tions than would occur with hydrocarbon fuels.
projections produce both smaller and larger estimates Thus, liquid hydrogen would probably cause
of contrail cover and radiative forcing for the future. more contrails to form, but possibly with greater
Current uncertainties in contrail coverage, optical particle sizes and fallout rates, resulting in shorter
depth, lifetimes, and overlap with lower clouds pre- lifetimes and smaller radiative impacts. It is possible
clude a definitive assessment of the overall impact of that a propulsion source that does not produce
contrails. Despite these uncertainties, it is clear that water vapor will be necessary to effectively eliminate
whatever effect they currently have on climate will the generation of contrails from high-flying air-
increase in the future. craft.
Other methods for minimizing contrail formation
would involve changes in flight altitude or
path. Contrail coverage could be reduced dramatically
The Future by flying in the stratosphere, where formation condi-
Contrails are difficult to study because of their tions are rare. However, other effects from the exhaust
high altitude, large advection rates, and frequent and increased fuel usage may limit the amount
520 CONTRAILS
of stratospheric traffic. Flying at lower altitudes would Temperature corresponding to tangent point
diminish the number of contrails in tropical areas, of mixing line with e, ("C or K)
but would cause additional coverage in the Contrail formation threshold temperature
mid-latitudes and polar regions. Conversely, higher ("C or K)
mean flight altitudes would decrease contrails over Spectral equivalent blackbody temperature
the poles and temperate zones while causing (K)
more contrails in the equatorial areas. Ideally, numer- Relative humidity with respect to ice
ical weather predictions and contrail formation prog- Relative humidity with respect to liquid water
nostication programs could be used together with Water vapor partial pressure difference (Pa)
flight planners to map out for each destination Mixing ratio difference (gkg - I)
a sequence of flight altitudes that best avoids contrail Exhaust-ambient air temperature difference
formation conditions. Such sophisticated planning ("C or K)
would require more accurate temperature and humid- Ratio of dry air and water vapor gas con-
ity data and contrail prediction schemes than stants; emissivity
are currently available as well as a more complex Spectral cloud emissivity
air traffic control network. Future research may Overall aircraft propulsion efficiency
provide the tools to minimize the climatic effects of Wavelength
contrails, but it is likely that these artificial clouds will
be a common feature in the sky for many years to
come.
See also
Aerosols: Role in Cloud Physics. Aircraft Emissions.
Nomenclature
Cloud Microphysics. Clouds: Classification. Convec-
Planck function tion: Laboratory Models of. Global Change: Human
Specific heat capacity (J kg - K - impact of Climate Change. Optics, Atmospheric: Optical
Ambient water vapor partial pressure (Pa) Phenomena; Optical Remote Sensing Instruments. Ra-
Exhaust water vapor partial pressure (Pa) diative Transfer: Cloud-radiative Processes. Satellite
Remote Sensing: Cloud Properties. Thermodynamics:
Ice saturation partial pressure (Pa)
Saturated Adiabatic Processes.
Liquid water saturation partial pressure (Pa)
Water vapor emission index (kg kg - ')
Exhaust-ambient air mixing line slope
(PaK-l) Further Reading
Specific plume enthalpy (J kg - ') Brasseur GP, Hauglustaine D, Cox RA, et al. (1998)
Pressure (hPa) European scientific assessment of the atmospheric effects
Water vapor mixing ratio (g kg- ') of aircraft emissions. Atmospheric Environment 32 (13):
Upwelling radiance at cloud base 2329-24 18.
(Wm-2sr-1) Penner JE, et al. (1999) Aviation and the Global Atmos-
Radiance emitted by cloud ( W m-2sr-1) phere. Cambridge: Cambridge University Press.
Spectral radiance (W m - sr - ') Schumann U (1996) On conditions for contrail formation
Mass specific combustion heat (MJk g - l ) from aircraft exhausts. Meteorologische Zeitschrift 5 :
Time (s) 4-23.
Schumann U and Amanatidis GT (2001)Aviation, aerosols,
Temperature ("C or K)
contrails, and cirrus clouds (A2C3).Air Pollution Re-
Ambient temperature ("C or K) search Report 74. European commission, Brussels,
Equivalent blackbody temperature of upwell- Belgium.
ing radiance at cloud base (K) Toon OB, et al. (1996) Subsonic aircraft: contrail and cloud
Cloud temperature (K) effects special study. Geophysical Research Letters 25(8):
Exhaust temperature ("C or K) 1109-1 168.
CONVECTION/ Convection in the Ocean 521
Contents
Convection in the Ocean

Laboratory Models
Thermal convection is associated with the cooling

Convedlon In the Ocean of the ocean surface due to sensible (QT),latent (QL),
and effective Iong-wave radiation (QE)heat fluxes. QT
A Soloviev and B Klinger, Nova Southeastern may have either sign; its magnitude, however, is much
University, Dania Beach, FL, USA less than that of QE or QL (except perhaps in some
extreme situations). The top of the water column
becomes colder and denser than the water below, and
convection begins. In this way, cooling is associated
with the homogenization of the water column and the
Introduction deepening of the mixed layer. Warming due to solar
Free convection is fluid motion due to buoyancy radiation occurs in the surface layer of the ocean and
forces. Free convection, also referred to simply as is associated with re-stratification and reductions in
convection, is driven by the static instability that the depth of the mixed layer. The most prominent
results when relatively dense fluid lies above relatively examples of this mixingre-stratification process are
light fluid. In the ocean, greater density is associated the diurnal cycle (nighttime cooling and daytime
with colder or saltier water, and it is possible to have warming) and the seasonal cycle (winter cooling
thermal convection due to the vertical temperature and summer warming).
gradient, haline convection due to the vertical salinity There are also important geographical variations in
gradient, or thermohaline convection due to the convection, with net cooling of relatively warm water
combination. occurring more at higher latitudes and a net warming
Since sea water is about 1000 times denser than air, of water occurring closer to the Equator. For this
the air-sea interface from the water side can be reason, mixed layer depth generally increases toward
considered a free surface. So-called thermocapillary the poles, though at very high latitudes ice-melt can
convection can develop near this surface owing to the lower the surface salinity enough to inhibit convec-
dependence of the surface tension coefficient on tion. Over most of the ocean, annual average mixed
temperature. There are experimental indications that layer depths are in the range 30-100 m, though very
in the upper ocean layer more than 2 cm deep, buoyant dramatic convection in such places as the Labrador
convection dominates. Surfactants, however, may Sea, Greenland Sea, and western Mediterranean Sea
affect in the surface renewal process. This article will can deepen the mixed layer to thousands of meters.
mainly consider convection without these capillary This article discusses convection reaching no deeper
effects. than a few hundred meters. Dynamically, the convec-
Over most of the ocean, the near-surface region tion discussed here differs from deep convection in
is considered to be a mixed layer in which turbulent being more strongly affected by surface wind stress
mixing is stronger than at greater depth. The strong and much less affected by the rotation of the Earth.
mixing causes the mixed layer to have very small Convection directly affects several aspects of the
vertical variations in density, temperature, and near-surface ocean. Most obviously, the velocity
other properties compared to the pycnocline region patterns of the turbulent flow are influenced by the
below. Convection is one of the key processes driving presence of convection, as is the velocity scale. The
mixed layer turbulence, though mechanical stirring convective velocity field then controls the vertical
driven by wind stress and other processes is transport of heat (or, more correctly, internal energy),
also important. Therefore, understanding convection salinity, momentum, dissolved gases, and other prop-
is crucial to understanding the mixed layer as erties, and the vertical gradients of these properties
well as property fluxes between the ocean and the within the mixed layer. Convection helps to determine
atmosphere. property exchanges between the atmosphere and
522 CONVECTION I Convection in the Ocean
ocean and the upper ocean and the deep ocean. The 0.1"C makes Ra > Racr as long as h is greater than a
importance of convection for heat and gas exchange centimeter.
has implications for climate studies, while convective For Ra > Rata the Rayleigh number still serves a
influence on the biologically productive euphotic zone useful purpose as a guide to the nature of the
has biological implications as well. convective activity (although the problem also de-
pends on the Prandtl number, Pr = v / ~ T ) .For a fixed
Pr and for Ra only slightly larger than Ra,,, motion
Phenomenology occurs in regular, steady cells. As Ra is increased, the
motion becomes time-dependent. Regular oscillations
The classical problem of free convection is to deter- occur, and these increase in number and frequency for
mine the motion in a layer of fluid in which the top higher Ra. At sufficiently high Ra, the flow is turbulent
surface is kept colder than the bottom surface. This is and intermittent. The value of Ra in the ocean is very
an idealization of such geophysical examples as an large (typically greater than 1014 for a temperature
ocean being cooled from above or the atmosphere difference of 0.1"C over 10 m), so convection is usually
being heated from below. The classical problem turbulent.
ignores such complications as wind stress on Turbulent convection is usually characterized by the
the surface, waves, topographic irregularities, and formation of descending parcels of cold water. In
the presence of a stably stratified region below the laboratory experiments, it has been found that water
convection region. The study of convection started in from the cooled surface layer collects along lines,
the early twentieth century with the experiments of producing thickened regions that become unstable and
Benard and the theoretical analysis of Rayleigh. One plunge in vertical sheets (Figure 1).In analogy to the
might expect that heavier fluid would necessarily atmospheric convection, we will here call these parcels
exchange places with lighter fluid below owing to thermals, although, in contrast to the atmosphere, in
buoyancy forces. This happens by means of convective the ocean they are colder than the surrounding fluid.
cells or localized plumes of sinking dense fluid and In 1966, Howard formulated a phenomenological
rising light fluid. However, such cells or plumes are theory that represented turbulent convection as the
retarded by viscous forces and are also dissipated by following cyclic process. The thermal boundary layer
thermal diffusion as they fall or sink into an environ- forms by diffusion, grows until it is thick enough to
ment with a different density. When the buoyancy start convecting, and is destroyed by convection,
force is not strong enough to overcome the inhibitory which in turn dies down once the boundary layer is
effects, the heavy-over-light configuration is stable and
no convection forms. The relative strengths of these
conflicting forces is measured by the Rayleigh number,
a nondimensional number given by eqn [I].
gaATh3
Ra = ~
kTV
In eqn [l], g is the acceleration of gravity, a is

the thermal expansion coefficient of sea water
0 ~ ~at T = 20°C and S = 35 psu),
( a = 2 . 6 ~ 1 "C-l
AT is the temperature difference between the top and
bottom surfaces, h is the convective layer thick-
ness, and v and k T are the molecular coefficients of
viscosity and thermal diffusivity, respectively
(v = 1.1x10-6m2s-1 and k T = 1 . 3 ~ 1 0 - ' m ~ s -at~
T = 20°C and S = 35 psu). The term aAT = A p / p
represents the fractional density difference between
top and bottom.
Convection only occurs if Ra is greater than a
critical value, Ra,,, which depends somewhat on Figure I Orthogonal views of convective streamers in the warm
geometrical and other details of the fluid. For the water that is cooling from the surface. The constantly changing
patterns appear as intertwining streamers in the side view.
classical problem of water bounded above and below (Reproduced from Spangenberg WG and Rowland WR (1961)
by solid plates, Racr = 657. For sea water under Convective circulation in water induced by evaporative cooling.
typical conditions, even a temperature difference of Physics of Fluids 4:743-750.)
destroyed. Then the cycle begins again. This phen- layer, the mixed layer density increases to p z , which is
omenological theory has implications for the devel- slightly denser than p1 (Figure 3B). The static insta-
opment of parameterizations for the air-sea heat bility now allows convection to act on the pycnocline
and gas exchange under low wind speed conditions down to density pz (Figure3C), so that the mixed layer
(see later). grows at the expense of the pycnocline. This is known
The descending parcels of water have a mushroom- as nonpenetrative convection.
like appearance. In the process of descending to deeper In reality, the largest thermals acquire enough
layers, the descending parcels developing as a result of kinetic energy, as they fall through the mixed layer,
the local convective instability of the thermal molec- that they can overshoot the base of the mixed layer,
ular sublayer join and form larger mushroomlike working against gravity. This is penetrative convec-
structures. The latter descend faster and eventually tion. The penetrative convection produces a counter-
form bigger structures. This cascade process produces gradient flux that is not properly accounted for if we
a hierarchy of convective scales, which is illustrated in model convective mixing as merely a very strong
Figure 2 on the example of the haline convection. vertical diffusion. Unlike the smooth density profile at
the base of a mixed layer that is growing by nonpen-
etrative convection (Figure 3C), penetrative convec-
Penetrative Convection tion is characterized by a density jump at the base of
the mixed layer (Figure 3D).
The unstable stratification of the mixed layer is usually The cooling of the ocean from its surface is
bounded below by a stratified pycnocline. One can compensated by the absorption of solar radiation.
imagine the mixed layer growing in depth, with The latter is a volume source for the upper meters of
thermals confined to the statically unstable depth the ocean. The thermals from the ocean surface, as
range. Suppose the density at the top of the pycnocline they descend deeper into the mixed layer, produce heat
is p1 (Figure 3A). As surface buoyancy loss and flux that is compensated by the volume absorption of
convection increase the average density of the mixed solar radiation. This is another type of the penetrative
convection in the upper ocean, which will be consid-
ered in more detail in a later section.
Relative Contributions of Convection

and Shear Stress to Turbulence
For the limiting case in which the only motion in the
mixed layer is due to convection, there are simple
estimates of average speed and temperature fluctua-
tions associated with the plumes. When the Rayleigh
number is high enough that the flow is fully turbulent,
the plume characteristics should be largely independ-
ent of the viscosity and diffusivity throughout most of
the mixed layer. In that case, ignoring the Earth's
rotation and influences from the pycnocline, the
governing parameters of the system are simply the
mixed layer depth h and the surface buoyancy flux Bo.
Bo is based on the surface heat fluxes according to
eqn 121.
In eqn [2], p is the water density, cp is the specific heat

capacity of water ( M 4 x l o 3J kg-I K'),L is the latent
heat released by evaporation (x2 . 5 ~ 1 Jkg-l),
0~ S is
the surface salinity, and j?is the coefficient of salinity
Figure 2 Shadowgraph picture of the development of secondary - ~ T = 20°C and
expansion (/3 = 7 . 4 ~ 1 0 - ~ p s u at
haline convection. (Reproduced from Foster TD (1975). The
hierarchy of convection. In: Processus de Formation Des Eaux S = 35psu). The first term in the square bracket in
Oceaniques Profondes, pp. 235-241. Colloques Internationaux du the right-hand side of eqn [2] relates to the buoyancy
C.N.R.S. No. 215.) flux due to surface cooling; the second term relates to
Original density After cooling

I I I
I
Density
Nonpenetrative
-
I
Density
Penetrative
-
I I I I I I 1
5
a
8
J i
7--------
Figure 3 Schematic diagram of nonpenetrative and penetrating convection.

Density -
the buoyancy flux due to the surface salinity increase Laboratory experiments have shown that these scales
because of evaporation. are in good agreement with actual fluctuations during
Given all the above restrictions, the velocity scale w, convection. For typical oceanic parameters (for in-
is then given by the Priestly formula (eqn [3]). stance, heat flux of Qo = 100 W m-2 and h = 100 m),
weis a few centimeters per second and b, is equivalent
W* = ( ~ ~ h ) ~ / ~ 131 to temperature fluctuations of about 0.01"C.
Two major sources of turbulent kinetic energy in the
This is the only combination of Bo and h that will give upper ocean are the wind stress and buoyant convec-
the proper units for velocity. Similarly, if we define the tion. Upper ocean convection is usually accompanied
buoyancy to be b = g A p / p , the buoyancy scale b, is by near-surface currents induced by wind and wind
given by eqn [4]. waves. The near-surface shear is then an additional
b,= c) 113
141
source of near-surface turbulent mixing. In the 1950s,
Oboukhov proposed the buoyancy length scale
LO = K M ~ / B owhere
, K is the von Karman constant
( K = 0.4), Bo is the surface buoyancy flux (e.g., defined diffusion in the absence of convection.
by eqn [2]),and u* is the boundary layer velocity scale
(friction velocity) defined as u* = ( r / p ) ‘ j 2 , where z
represents the bottom stress in the atmospheric case
and the wind stress in the oceanic case ( p is the density
of air or water, respectively). Later, Monin and This quantity must be a function of the given nondi-
Oboukhov suggested the stability parameter, ( = mensional parameters of the system, which for ther-
z / L o (where z is the height in the atmosphere or the mal convection in the absence of other driving
depth in the ocean), to characterize the relative mechanisms are just Ra and the Prandtl number Pr
importance of shear and buoyant convection in the (here we ignore the Earth’s rotation and entrainment
planetary boundary layer. Experimental studies con- from the pycnocline). A further simplifying assump-
ducted in the atmospheric boundary layer show that at tion is that for high Ra (greater than lo’) typical of the
(< - 0.1 the flow is primarily driven by buoyant mixed layer, the convection is fully turbulent and does
convection. Owing to the analogy between the atmos- not depend on the mixed layer thickness, h, which
pheric and oceanic turbulent boundary layers, the implies that eqn [6] holds, where A(Pr) is a dimen-
Monin-Oboukhov theory is often applicable to the sionless coefficient depending on the Prandtl number
analysis of the oceanic processes as well. In particular, (according to laboratory measurements, A FZ
it provides us with a theoretical basis on which to 0.16-0.25).
separate the layers of free and forced convection in the
upper ocean turbulent boundary layer. Nu = A ( P ~ ) R u ‘ / ~ [61
For a 5 m s-l wind speed and Qo = 100 W m-2, the Given the definitions of Ra and Nu,this relation can be
Oboukhov scale is LO 15m. This means that the
N rearranged to yield the temperature difference AI
shear-driven turbulent flow is confined within a 1.5m across the cool skin as a function of the surface heat
thick near-surface layer of the ocean. In a 5Om deep + +
flux QO= QL QE QT as in eqn [7].
mixed layer, 97% of its depth will be driven by the
buoyant convection during nighttime, with the rate of
dissipation of turbulent kinetic energy there being
about equal to the surface buoyancy flux, Bo, as shown
by Shay and Gregg. AT is 0.2-0.4”C under typical oceanic conditions but
can be as much as 1°C in regions of very high heat loss
to the atmosphere (e+, the Gulf Stream at high
latitudes).
Convection and Molecular Sublayers While the term ‘sea surface temperature’ (SST) is
Convection is driven by the horizontal-mean vertical often used to represent the temperature of the mixed
density gradient. At high Ra, typical vertical velocities layer as a whole, the existence of a cool skin means that
are much lower near the top and bottom boundaries the temperature of the literal surface of the ocean can
than they are in the bulk of the water column. Since the be somewhat lower than the rest of the mixed layer.
vertical density gradient is reduced by the convective Satellite measurements of SST are based on infrared
motion, the velocity distribution causes most of the emissions from a layer several micrometers thick, so
vertical density gradient to occur near the boundaries. that these measurements can be somewhat different
Indeed, under low-wind, low-wave conditions in from ship-based ‘surface’ measurements, which are
which convection dominates, the mixed-layer temper- generally based on sampling within the upper several
ature gradient is largely confined to a region only meters of the ocean. Indeed, while the first experi-
about 1mm deep. Because the vertical heat flux at the mental evidence of the cool skin was obtained in the
base of the convection region is typically much smaller 1920s, this phenomenon was not widely recognized by
than at the surface, the large temperature gradient only the oceanic community until sophisticated methods,
occurs at the surface, where this thermal sublayer is including remote sensing by infrared techniques,
often referred to as the cool skin. began helping to incorporate the cool skin into
The temperature jump across the cool skin can be modern oceanography.
related to the vertical flux of heat at the air-sea The accuracy of current satellite remote sensing
interface and constants of molecular viscosity and heat techniques is, nevertheless, still below that level at
diffusion in water using convection laws. The vertical which the cool skin becomes of crucial importance.
heat flux, Qo, can be written in nondimensional form The effect of the cool skin on the heat exchange
as the Nusselt number Nu (eqn [ 5 ] ) ,in which the heat between ocean and atmosphere is also basically below
flux is normalized by the heat flux due to vertical the resolution of widely used bulk-flux algorithms.
However, one interesting practical application of the reaches a value of approximately - 1 . 5 ~ 1 0 - (here
~ p
cool skin phenomenon emerged in the 1990s. Similar is the water density).
laws govern the thermal sublayer of the ocean (the cool
skin) and diffusive sublayers associated with air-
sea gas exchange. Such gas exchange is a key biogeo-
chemical variable, and for greenhouse gases such as Diurnal and Seasonal Cycle
C02 is of climatological importance as well. The rate of Convection
at which gases cross the air-sea interface is measured For much of the year, much of the ocean experiences a
by the piston velocity, K (see Air-Sea Interaction: cycle of daytime heating and nighttime cooling that
Gas Exchange). Boundary layer laws relate K to AT leads to a strong diurnal cycle in convection and mixed
according to eqn [8].
layer depth. Such behavior is illustrated in Figure 4. At
night, when there is cooling, the convective plumes
reach the base of the mixed layer, which deepens as the
mixed layer grows colder and denser. During the day,
In eqn [8], A0 is a dimensionless constant (E 1.85) convection is inhibited within the bulk of the mixed
and p is the molecular gas diffusion coefficient in water layer but may still occur near the surface of the mixed
- l COZ at T = 20°C and layer, even if the mixed layer experiences a net heat
( p = 1 . 6 ~ 1 0 - ~ m ~ sfor
S = 35 psu). The more readily available cool skin gain. This is because the vertical distributions of
data can then be used for an adjustment of the gas cooling and heating are somewhat different. Heat loss
transfer parameterization. is dominated by latent heat flux associated with
The convective parameterizations for the cool skin evaporation, hence this forcing occurs at the top
and air-sea gas exchange are valid within the range of surface. Heat gain is dominated by solar radiation that
wind speed from 0 to 3-4 m s - l . Under higher wind is absorbed by the water over a range of depths that
speed conditions, the cool skin and the interfacial air- can extend tens of meters in many parts of the ocean.
sea gas exchange are controlled by the wind stress and For example, one can have surface heat loss of
surface waves. The transition is observed when the 100 W m-2 occurring at the surface and net radiative
surface Richardson number, Rfo = ocgQo/(cppui), heat gain of 500 W mP2 distributed over the top 30 m
Figure 4 Diurnal cycles in the outer reaches of the California Current (34” N, 127” W). Each day the ocean lost heat and buoyancy
starting several hours before sunset and continuing until afew hours after sunrise. These losses are shown by the shaded portions of the
surface heat and buoyancy fluxes (A). In response, the surface turbulent boundary layer slowly deepened (B). The solid line marks the
depth of the surface turbulent boundary layer, and the lightest shading shows 1O-’W kg-I < E < lO-’W kg-I. The shading increases by
decades, so that the darkest shade is E > W kg-I. Note that 1 MPa in pressure p corresponds to approximately lOOm in depth,
Jg = -6,and J: = -(a+ QR), where QR is the solar radiation flux penetrating ocean surface. (Reproduced from Lombard0 CP and
Gregg MC (1989) Similarity scaling during nighttime convection. Journal of Geophysical Research 94: 6273-6214.)
CONVECTIONI Convection in the Ocean 527
of the ocean. Calculating the rate of change of heat due Open ocean convection is a mechanism effectively
to the forcing between the surface and a depth z , we controlling the seasonal cycle in the ocean as well.
find that there is actually heat loss for small z down to a Resolution of diurnal changes is usually uneconomical
depth known as the thermal compensation depth. when the seasonal cycle is considered. Because of
Below this depth, the mixed layer re-stratifies and nonlinear response of the upper ocean to the atmos-
convection occurs only through the mechanism of pheric forcing, simply averaged heat fluxes cannot be
penetrative convection. For most of the world ocean, used to estimate the contribution of the convection on
the thermal compensation depth is less than l m the seasonal scale. The sharp transition between the
between sunrise and sunset. nocturnal period, when convection dominates mixing
Usually, the rate of turbulent kinetic energy pro- in the surface layer, and the daytime period, when the
duction in the mixed layer is dominated by the Sun severely limits the depth of convection, leaving the
convective term at night but by the wind stress term wind stress to control mixing, may simplify the design
during most of the day. Because the thermal compen- of models for the seasonal cycle of the upper ocean.
sation depth is generally quite small, turbulent kinetic Incorporation of convection adjustment schemes into
energy generated by convection makes no contribu- the oceanic component of the global circulation
tion to turbulent entrainment of water through the models leads to an appreciable change of troposphere
bottom of the mixed layer, which lies much deeper. temperature in high latitudes, which affects the global
Under conditions of low wind speed and strong solar ocean and atmospheric circulation. Parameterization
insolation, the thickness of the surface convective of the convection on the seasonal and global scale is
layer of the ocean may reduce to only several centi- therefore an important task for the prediction of
meters. In that case, convection in the upper ocean climate and its changes.
may be of a laminar or transitional nature.
Stable stratification inhibits turbulent mixing below
the relatively thin near-surface convection layer. Ver- Conclusions
tical mixing of momentum is confined to the shallow Observation of the open ocean convection is a difficult
daytime mixed layer, so that during the day flow driven experimental task. Although convective processes
directly by the wind stress is confined to a similarly have been observed in several oceanic turbulence
thin current known as the diurnal jet. In the evening, studies, most of our knowledge of this phenomenon in
when convection is no longer confined by the solar the ocean is based on the analogy between atmospher-
radiation effect, convective plumes penetrate deeper ic and oceanic boundary layers and on laboratory
into the stratified part of the mixed layer, increasing studies. Many intriguing questions regarding the
the turbulent mixing of momentum at the bottom of convection in the open ocean remain. Some of them,
the diurnal jet. The diurnal jet then releases its kinetic such as the role of penetrative convection in mixed
energy during a relatively short time. This process is so layer dynamics, are of crucial importance for im-
intensive that the released kinetic energy cannot be provement of the global ocean circulation modeling.
dissipated locally. As a result, a Kelvin-Helmholtz Others, like the role of surfactants in the surface
type instability is formed, which generates billows - a renewal process, are of substantial interest for study-
kind of organized structure. The billows intensify the ing the air-sea exchange and global balance of
deepening of the diurnal mixed layer. greenhouse gases such COz.
Although the energy of convective elements is
relatively small, it serves as a catalyst for the release
of the kinetic energy by the mean flow. In the See also
equatorial ocean, the shear in the upper ocean is
Air-Sea Interaction: Gas Exchange; Sea Surface Tem-
intensified by the Equatorial Undercurrent; the eve-
perature. Ocean Circulation: Surface-Wind Driven
ning deepening of the diurnal jet is therefore some- Circulation; Thermohaline Circulation. Parameterization
times so intense that it resembles a shock, which of Physical Processes: Turbulence and Mixing. Turbu-
radiates very intense high frequency internal waves in lence and Mixing.
the underlying thermocline.
The diurnal cycle is often omitted from numerical
ocean models for reasons of computational cost. Further Reading
However, the mixed layer response to daily-averaged Busse FH and Whitehead JA (1974) Oscillatory and collec-
surface fluxes is not necessarily the same as the average tive instabilities in large Prandtl number convection.
response to the diurnal cycle. Neglecting the diurnal Journal of Fluid Mechanics 66: 67-79.
cycle replaces periodic nightly convective pulses with Caldwell DR, Lien R-C, Moum JN and Gregg MC (1997)
chronic mixing that does not reach as deep. Turbulence decay and restratification in the equatorial
528 CONVECTION I Laboratory Models
ocean surface layer following nighttime convection. Kraus EB and Rooth CGH (1961)Temperature and steady
Journal of Physical Oceanography 27: 1120-1 132. state vertical heat flux in the ocean surface layers. Tellus
Foster TD (1971) Intermittent convection. Geophysical 13: 231-238.
Fluid Dynamics 2: 201-217. Shay TJ and Gregg MC (1986) Convectively driven turbu-
Fru NM (1997) The role of organic films in air-sea lent mixing in the upper ocean. Journal of Physical
gas exchange. In: Liss PS and Duce RA (eds) The Sea Oceanography 16: 1777-1 79 1.
Surface and Global Change, pp. 121-172. Cambridge: Soloviev AVand Schluessel P (1994)Parameterization of the
Cambridge University Press. cool skin of the ocean and of the air-ocean gas transfer on
Gregg MC, Peters H, Wesson JC, Oakey NS and Shay TJ the basis of modeling surface renewal.Journal ofPhysica1
(1984)Intense measurements of turbulence and shear in Oceanography 24: 1339-1346.
the equatorial undercurrent. Nature 3 18: 140-144. Thorpe SA (1988)The dynamics of the boundary layers of
Holland WR (1977) The role of the upper ocean as the deep ocean. Science Progress Oxford 72: 189-206.
a boundary layer in models of the oceanic general Turner JS (1973) Buoyancy Effects in Fluids. Cambridge:
circulation. In: Kraus EB (ed.) Modelling and Prediction Cambridge University Press.
of the Upper Layers of the Ocean. Oxford: Pergamon Woods JD (1980) Diurnal and seasonal vibration of
Press. convection in the wind-mixed layer of the ocean. Quar-
Katsaros KB (1980) The aqueous thermal boundary layer. terly Journal of the Royal Meteorological Society 106:
Boundary-Layer Meteorology 18: 107-127. 379-394.
Laboratory Models
H J S Fernando, Arizona State University, Tempe, O n the planetary scale, atmospheric convection is
AZ, USA largely driven by the meridional imbalance of net solar
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. radiation, contributed by a net radiative gain in low
latitudes and a loss in polar regions as well as heat loss
at upper levels by radiative cooling. In the absence of
air circulation, such an imbalance would lead t o a
Introduction
continuous increase in temperature at low latitudes
Convection occurs when a fluid is heated or cooled and vice versa, but convection acts t o prevent such
either at its boundaries (e.g., daytime heating of changes by realizing meridional transfer of heat. For a
ground or nighttime cooling of ocean surface) or in nonrotating Earth, this would cause the rise of low-
its interior (e.g., clouds). It is of profound importance latitude warm air followed by sinking of cold air in
in atmospheric and oceanic dynamics, particularly in high latitudes, forming a single meridional overturn-
transporting heat from one location t o another. ing cell (Hadley circulation). Because the Earth
Convection may be driven solely by the buoyancy rotates, the circulation cell is greatly modified by
forces (“buoyant convection”) or by a combination of Coriolis forces, the compounding effects of which
buoyancy and mechanical (e.g., pressure gradient) make the Hadley circulation unstable. As a result, the
forcing known as “forced convection.” Convective planetary convection consists of three circulation cells,
phenomena abound in the atmosphere, spanning from communicating with each other through a complex set
planetary scale to small scales, with each scale playing of processes, as illustrated in Figure 1. The first
an important role in maintaining delicate balances of (Hadley)cell is dominated by upper radiative cooling,
heat, moisture, and momentum in the atmosphere with little influence of Earth’s rotation, in much the
conducive for life. Atmospheric convection is strongly same way as in the previously described Hadley
coupled with oceanic processes though air-sea inter- circulation. The rising equatorial air reaches very
action, the combined action of which largely deter- high altitudes (deep convection), of the order of tens of
mines the climate on the Earth. Over the history of the kilometers, thus forming a belt of clouds (Intertropical
Earth (and other planets), convective processes have Convergence Zone). Sinking air parcels in the Hadley
undergone large changes that have affected the cell feed the equatorial Trade Winds and drive the
oceanic and atmospheric general circulation patterns. Ferrel cell, in which poleward-moving warm air
Further changes are expected in the future as a result encounters colder-high latitude air t o form the polar
of anthropogenic activities, which, some believe, front. Such a front can attain geostrophic equilibrium,
could even be to some extent detrimental to human with the cross-frontal pressure gradient balanced by
existence. the Coriolis forces of zonal flows (polar frontal jets).
CONVECTION / LaboratoryModels 529
Figure 1 A schematic of the atmospheric general circulation driven by meridional variation of the heat flux. The three-cell structure and
the deep and slantwise (sloping) convection are indicated. The daytime convection in complex and flat terrain is depicted in the inset.
(Adapted from Houghton (1989).)
The warm and cold air tend to be separated by a Sloping Convection

sloping (frontal) surface, which becomes unstable
Much of the initial understanding of sloping convec-
(baroclinic instabilities) and degenerates into wavelike
tion associated with the general circulation of
disturbances that allow the interchange of air parcels
the atmosphere has come from laboratory experi-
along slanted paths (known as the slantwise or slope
ments. A typical laboratory flow configuration
convection). Without such slanting convection, the
consists of a fluid annulus with inner and outer radii
meridional flow would be symmetric and would spiral
toward the poles, with inefficient heat transport
a and 6 and temperatures T, and Tb, respectively,
subjected to background rotation of frequency SZ
capabilities. Slantwise convection allows fluid parcels
(Figure 2 ) . To mimic the atmosphere, the outer
to rise and fall along inclined paths, releasing the
annulus is maintained colder than the inner annulus,
potential energy of the front in the form of kinetic
AT = T , - T b > 0. The governing parameters for
energy of baroclinic waves. Symmetric spiraling
the problem are a, 6, Q, gaAT ( a is the thermal
convection can only transport about one-eighth of
expansivity and g the gravity), v (kinematic viscosity),
the heat flux that is needed to be transported
IC (thermal diffusivity), and d (depth of the fluid layer).
meridionally, but slantwise convection with six
In most cases, the dependence on a and b can
baroclinic waves can accomplish this task.
be replaced by (6 -a). Note that, in the absence
In addition to the planetary-scale phenomena,
of rotation, the radial temperature differential of
convective processes of smaller scales are prevalent
AT of a fluid column of depth d can initiate a
in the atmosphere: from synoptic-scale and meso-scale
meridional (radial) gravity current of speed of
convection in the troposphere (powered by latent heat
released during condensation in clouds) to boundary
-
U (gzATd)'I2, but this current is rotationally
arrested when the Rossby number Ro = U/SZ(b - a)
layer-scale convection driven by the heating at land
or the parameter gctATd/SZ2(b- a)' is sufficiently
and ocean surfaces. Convection over land may arise
small. Prior to the initiation of rotational effects,
and be modified by nonuniform spatial heating, for
the convection tends to assume the Hadley circula-
example, due to isolated sources of widely different
tion, with the flow confined to upper and lower
scales (e.g., thermals released from the ground, anvil
boundary layers; but as the rotation is imposed
clouds, microbursts, and urban heat islands). The
the fluid parcels deflect in the zonal direction, with
nature of the terrain, whether it is complex or flat
the upper level flow spiraling toward the poles.
terrain, also affects convection. All of these types of
The nature of convection therein is determined
convection are complex and often defy detailed
by the governing parameters for the problem, given
theoretical treatment. To this end, laboratory models
in eqns [1]-[4].
(mostly conducted with liquid working fluids) have
played an important role in understanding atmo-
gaATd
spheric convection, and a very brief review of some ROT =
relevant laboratory modeling efforts is presented here. R2(b - a)'
530 CONVECTION I Laboratory Models
Figure 2 Flow regimes observed in a ‘dishpan experiment’ with ( b - a ) / d = 2 and Pr RZ 21. These experiments mimic baroclinic
waves and slantwise convection observed in the atmosphere. (Compiled from Hide and Mason (1975)and Buzyna G,Pfeffer RL and
Kung R (1 984) Transition to geostrophic turbulence in a rotating differentially heated annulus of fluid. Journal of Fluid Mechanics 145:
377-403.)
values such that the motions are nonaxisymmetric.

In this case, baroclinic waves with a dominant
wavenumber are generated in a background of myriad
4Q2d4 other wavenumbers. This breakdown process paral-
Tu2 = - 131 lels the baroclinic instability of a stratified fluid with
V2
a buoyancy gradient N 2 in a rotating fluid, in which
V
the flow becomes un-stable when the Eady number
Pr =- [41 E = N2d2/4Cl2(6- a ) 2 drops below a critical value.
k Widely different behaviors are possible in this
where Tal and Tu2 are the Taylor numbers baroclinic wave regime (Figure2 ) ,as discussed below.
with Tal/Taz= [(b- a)/d]’. Figure 2 shows a regime At somewhat higher ROT,a well-defined azimuthal
diagram of an experiment carried out at constant wave pattern dominated by a single wavenumber with
(6 - a ) / d and Prandtl number Pry wherein the periodic time dependence (amplitude vacillations) is
important governing parameters become ROT possible. This (doubly) periodic regime allows large
and Tal. At no rotation or very low rotation speeds, growth and decay rates of energy of the dominant
the flow belongs to the Hadley regime, but as mode as well as modulations of its low-amplitude
the rotation is gradually increased the flow assumes a sidebands and higher harmonics. The lower ROT
‘symmetric flow regime’ where mean flow is largely region of the wave regime is characterized by
azimuthal (zonal).For Tal < 2 x lo5or so, the viscosity structural vacillations, wherein the wave structure
ensures that the flow is axisymmetric for all values becomes more complicated with semiperiodic time
of ROT,but at higher Tal there exists a range of ROT dependence of wave shape and energy distribution.
CONVECTIONI LaboratoryModels 531
It is dominated by a single zonal wavenumber and does not predict specific horizontal planforms for
much less active higher harmonics. Further reduction the marginally stable flow, but allows the flow to
of ROT (and/or increase of Tal) leads to accept solutions with tessellating symmetric plan-
a regime with spatially and temporally chaotic flow, forms. In reality, the wavelength selection is quite
classified as geostrophic turbulence. The changes sensitive to the initial and boundary conditions used;
between each of these regimes, especially from the for example, constant-flux boundary conditions make
wave to the geostrophic regimes, are usually asso- cell patterns much flatter than constant-temperature
ciated with well-defined transitional regimes, but boundary conditions. Changes in Ra, however, lead to
abrupt transitions (e.g. amplitude to structural vacil- a host of flow patterns (Figure 3). With increasing Ra
lation regimes) are also possible. The baroclinic waves beyond the critical value, the updrafts, which emerge
developed in this ‘dishpan’ flow configuration can along the boundaries of polygon-shaped planforms
realize substantial amounts of meridional heat trans- (with size 2-3d) for marginally stable flows, tend to
N
port through sloping convection and play an impor- concentrate along the hubs of the warm spokes of
tant role in the mid-latitude circulation in the polygons. This is also called the ‘spoke’ convection.
atmosphere. Because of the rich variety of flow and Rising warm plumes/thermals from these ‘hot spots’
transitional regimes possible, the mid-latitude atmo- rise to the upper boundary and spread horizontally,
sphere is the most dynamic and complicated (like a while downdrafts make up for the lost fluid near the
war zone!). lower boundary. With a further increase of Ra, the
flow becomes turbulent. Convective turbulent flows
have been thoroughly studied in the laboratory,
Convection in Wide Horizontal Fluid particularly the case of constant buoyancy (or heat)
flux qo (or Qo) at the bottom boundary in view of its
Layers direct applications to the atmospheric daytime
The archetype of convection in flat terrain is the boundary layer. In this case, the flux Rayleigh number
case where an unstably stratified fluid layer with Raf = q 0 d 4 / ~replaces 2~ Ra, where qo = gctQo/poCp,
a vertical temperature gradient T, is sandwiched p o is a reference density, and C pis the specific heat. The
between two horizontal boundaries separated by flow appears to become turbulent for Raf > lo8. The
a distance d. Overall, this problem has two governing Rayleigh number convection in the regime Raf N
parameters: the Rayleigh number Ra = gaT,d4 / K V 107-108 occurs in intermittent bursts in which blobs of
and Pr. However, the linear stability of the problem heated fluid elements (thermals) are sporadically
is determined by Ra alone, and the onset of steady detached, enter the interior of the fluid and, are
convection occurs when Ra exceeds a critical value. homogenized therein (Figure 4).Motions directly
Near this critical value, the flow consists of stationary induced by these thermals and fluid that sweeps in to
-
cells with the horizontal scale on the order of fill the void so generated determine the structure of
the vertical scale (wavelength d ) . Linear analysis convection. The sizes of the thermals so produced have
Turbulent flow
Time-dependent 3-dimensional flow
Ill
3-dimensional flow
/ Steady 2-dimensional flow

I
1o3 I I No motion , I I
532 CONVECTION/ Laboratory Models
Figure 4 The formation of isolated thermals above a heated plate. At higher Rayleigh numbers, these thermals are effaced by the
turbulent eddies upon establishment of full convection. (Reproduced from Sparrow EM,Hussar RB and Goldstein RJ (1970) Observations
and other characteristics of thermals. Journal of Fluid Mechanics 41 : 793-807.)
length and time scales of ( ~ ~ v / q o ) ' and

/ ~ (v/qo)1'2, such flow configurations, also known as nonpenetra-
respectively. tive convection, neither a buoyancy jump nor internal
The role of thermals leaving the heated surfaces waves exist at the interface. In many geophysical
changes with increasing Raf ( > l o 9 ) ,in response to the situations, however, an interfacial buoyancy jump
intensification of convection. Although thermals near develops at the interface (penetrative convection),
the surface may still be developing, they are largely allowing plumes to cause mechanical entrainment or
effaced by strong convective eddies in the fluid layer, internal waves to grow and break. The rate of growth
paving the way for updrafts to carry heat from near the of the convective mixed layer is an important quantity
surface, which are compensated by downdrafts that in atmospheric boundary layer modeling and is usually
occupy a relatively larger area. The root mean square correlated with Ri or a related quantity. The condi-
(rms) turbulent fluctuations within the convective tions under which the transition from nonpenetrative
layer, evaluated by horizontally averaging the mea- to penetrative convection occurs are yet to be
sured fluctuations at a given distance z from the determined.
surface, show that they follow the scaling proposed by Since the presence of strong background rotation
Deardorff, in that the velocity and length scales of constrains inviscid fluid motions parallel to the axis
turbulence for z > O.ld become, respectively, the of rotation (Taylor-Proudman theorem), it suppresses
convective velocity w* = ( q 0 d ) l l 3and the convective eddying motions and delays the onset of convection.
layer depth d.
In the atmospheric daytime convection, the con-
vective layer is bounded by a density stratified
(inversion) layer aloft, and the structure of convective
motions near the boundary (interface) is governed by
the 'strength' of the inversion or the Richardson
number Ri = Abd/wg where Ab is the buoyancy jump
across the interface. The updrafts created during
convection penetrate into and generate internal waves
in the stratified layer (Figure 5 ) . Similar imprints of
plume impingement have been observed in radar
studies of atmospheric penetrative convection. The
ensuing vortical perturbations can excite outward- Figure 5 The growth of a turbulent convective layer into a linearly
propagating internal waves, at times carrying a stratified environment. The flow visualization has been performed
substantial amount of energy out of the convective by introducing fluorescent dye into the bottom convective layer. As
region. rising updrafts impinge on the stratified layer, they are deflected
In the Ri range attainable in the laboratory, and flow along the interface, causing local overturning events as
indicated by A. Thin inward cusp like events indicated by B occurs
interfacial mixing appears to occur by a mechanism when two horizontal counterflows meet each other. (Reproduced
whereby the interface is maintained at a marginally from Fernando HJS and Little LJ (1990) Molecular-diffusiveeffects
(statically) unstable state Ab = 0. Consequently, in in penetrative convection. Physics of Fluids A 2(9):1592-1 596.)
CONVECTION/ Laboratory Models 533
At the onset of convection, however, the flow is shown in Figure 7 indicates possible states as well as
dominated by viscous effects, and hence departures empirical and theoretical demarcation criteria separ-
from the Taylor-Proudman behavior can be observed. ating them. At slight supercriticalities, the biconical
For example, convective plumes emanating from structure of linear motion breaks down to form a
the buoyancy source (e.g., a heated plate) can rise regular vortex grid. The vortical structures emanating
vertically in a spiraling path along the axis of rotation, from the heated boundary, therefore, consist of thin
which can be described by linear stability theory: the annular columns of rising warm fluid and cores of
important parameters here are Ra and Pr and the descending colder fluid. These vortices are quite
Taylor number Ta2. The fluid parcels leaving the stationary, except for the occasional vortex merger.
boundaries possess vertical vorticity of background During the merger process, respective vortices rotate
rotation, amplify this vorticity owing to stretching, around each other while becoming closer, and initiate
and follow biconical particle trajectories as shown in connection at the top of the vortices while interlacing
Figure 6. with each other. The lateral motion and merger of
Laboratory experiments show wide departure from vortices becomes pronounced with increasing super-
linear theory predictions under varying degrees criticality (nonlinearity), although the general regu-
of supercriticality (the ratio Taz/Raf has been larity of the vortex grid is retained. Further increase of
proposed as a measure of superciticality, with the supercriticality causes the regular grid to transform
linear regime occurring at Tal/ Raf < 25). A wide into an irregular vortex grid, where the vortices are
variety of supercritical flow states have been docu- somewhat weaker, move randomly, and merge. Unlike
mented, though much is yet to be learned about their the regular vortex grid case, these vortices do not
dynamics and transitions. The Raf-Ta2 diagram extend to the heated bottom boundary, the flow in the
vortex core is absent, and the vortices are not strictly
vertical (geostrophic vortices). At larger Rayleigh
numbers, the irregular vortex regime transitions to
a regime with distinct ephemeral vortices, also known
as ‘isolated plume vortices’. Unlike geostrophic
vortices, these structures appear spontaneously,
are intense, extend to the heated boundary, abide for
a short period of time, and then fade away, much
like the ‘dust devils’ in the atmosphere. These vortices
undergo intense stretching and merging, indicating
strong nonlinearity of the motion field. Away from
the boundaries, the vortices assume gradient wind
balance due to strong curvature of flow paths.
Interposed between these isolated vortices are (inco-
herent) turbulent motions, which are influenced by
the vortices and perhaps affected by Coriolis forces.
Measurements show that convective turbulent
motions are affected by background rotation
at a height h, M 4 . 5 ( q 0 / C l ~ ) ~from
’ ~ the source,
whence the rms velocity and integral length scale,
respectively, are given by u, = 1.7(qo/R)li2 and
e, = l . l ( q ~ / C l ~ ) ~thus
’ ~ , maintaining a constant
Rossby number Ro = ur/2Cl4?,= 0.75. In typical
oceanic deep convection regions, h, is of the order of
10 km, and its value can be much higher for atmo-
spheric deep convection. Therefore, turbulence gen-
erated during deep convection can be considered as
unaffected by Earth’s rotation.
Convection in Complex Terrain

Figure 6 A perspective sketch of the path of a fluid parcel in a
hexagonal cell in rotating convection. (Reproducedfrom Veronis G
(1959) Cellular convection with finite amplitude in a rotating fluid. The flow in complex terrain differs greatly from
Journal of Fluid Mechanics 5 : 401-435.) its flat-terrain counterpart in that a host of
534 CONVECTIONI Laboratory Models
Figure 7 A regime diagram of flux Rayleigh number Rat versus Taylor number Tan showing various flow structures possible during
rotating thermal convection. (Based on laboratoly experiments of Boubnov BM and Golitsyn GS (1986) Journalof Fluid Mechanics 167:
503-531); Fernando HJS, Chen RR and Boyer DL (1991), Journal of Fluid Mechanics 228: 513-547); and David Fultz (personnel
communication.)
new physical phenomena emerge owing t o topo- shown in Figure 9. The bottom of this model is
graphic influences, as illustrated in Figure 8 using a subjected to heating and cooling, mimicking the
generic configuration of an open valley adjoining diurnal thermal forcing. In the morning, heating of
a plane and the response thereof to diurnal varia- the ground generates a shallow convective boundary
tion of thermal forcing. The nighttime flow consists of layer (CBL) at the bottom and the slopes of the valley.
the drainage of dense air formed on slopes into Since the contours of mean temperature are parallel to
the valley (downslope, drainage, or katabatic winds) the slope, an upslope mean pressure gradient force is
and the channeling of air pooled at the bottom of generated, thus inducing a weak upslope mean flow
the valley to the nearby plane (downvalley winds). (a case of Coanda effect) within a thin boundary layer
The extent of pooling in a valley depends on a number (Figure 9A). Lateral mixing of this upslope flow with
of factors, including topographic barriers, valley background air generates air parcels of varying
geometry, and hydraulic control at the valley mouth. densities, which can intrude into ambient air if the
Drainage winds, which are driven purely by basin is stably stratified (Figure 9B). Such intrusions
downward buoyancy forces, play an important role can form a layered density structure in the basin,
in the development of stable stratification in producing a variable lapse rate. By midday, with the
urban valleys at night. During the day, the flow intensification of heating, the upslope flow strength-
consists of upvalley and upslope flows driven by ens, removing substantial amounts of turbulent air
upward buoyancy forces. Both upslope and from the CBL and creating a compensatory subsidence
downslope winds, therefore, are examples of con- in the middle of the valley. The competition between
vection. the subsidence of the stable core and the growth of the
A laboratory model of a simplified complex terrain CBL determines the daytime flow and transport in the
consisting of a two-dimensional sinusoidal basin is basin. In reality, the subsidence of warm air into a
a given ridge-level plan area), leads to a down-

valley hydrostatic pressure gradient and hence an
upvalley flow. The nighttime cooling of the slopes
creates katabatic flow that pools in the basin.
The shear between the katabatic flow and the
ambient air lead to the development of overturning
instabilities and turbulent mixing along the slopes
(Figure 9C).
Convection from Isolated Sources

Convection occurring in nature is delimited by the
finite spatial extent of forcing, and in some cases the
aspect ratio A (defined as the ratio of the width W to
the depth d of convection) is small enough for lateral
Figure 8 A schematic of flow phenomena occurring in complex motions to play an important role in the dynamics of
terrain consisting of a simple valley and an adjoining plane.
Downslopeand downvalley winds during the day and upslope and convection. The idealized case of W + 0 (A + 0)
upvalley winds during the night are indicated. corresponds to p o i n t h e thermals (blobs of buoyant
fluid; Figure 10A) or plumes (continuous buoyancy
sources; Figure lOB), studies on which are useful
building blocks in understanding natural flows. Point
valley, compounded with topographic amplifica- (line) plumes are described by the total buoyancy
tion (higher temperatures in the valley than the nearby flux (buoyancy flux per unit length) and point
plane due to a lesser amount of air contained below (line) thermals are specified by the total buoyancy
Figure 9 A laboratoty model of a two-dimensional basin that illustrates upslope and downslope winds in complex terrain. (A) Upslope
winds generated during the heating of the basin bottom; dye has been introduced at the basin bottom for flow visualization. (B) The same
as in (A), but the basin fluid is stably stratified with temperature. Note the formation of horizontal intrusions into the stable core. (C) The
formation of a downslope flow upon cooling of the basin surface. The shear between the katabatic flow and the ambient fluid generates
overturning instabilities.
536 CONVECTIONI Laboratory Models
sively studied in the laboratory. Much of the literature

on cases with background rotation is recent, however,
and can be attributed to the increased emphasis on
deep convection as well as oceanic leads, polynyas,
and hydrothermal vents, in recognition of their
climatic implications.
Laboratory studies show that in nonrotating and
nonstratified environments, the travel distance of
plume and thermal fronts h, increases with time t
according to simple laws dictated by dimensional
analysis, h, 1.7(q,t3)1'4and h, = 1.8qT ,
1Pt1/2
where q, and qT, respectively, are the buoyancy flux
and total buoyancy of the plume and thermal. In
the presence of rotation, the Coriolis effects tend
to constrain the growth of the plumehhermal at a
time t N E'; the subsequent growth is characterized
by the inhibition of the lateral growth and somewhat
slow vertical descent. Eventually, the plume/thermal
transforms into a cylindrical cyclonic vortex, within
which smaller, slender vortices are also evident. Owing
to Coriolis effects, the exterior entrainment flow
deflects to form a cyclonic vortex (Figure l l A ) , which,
after some time, breaks down by barotropic instabil-
ities to form cyclonic eddies (Figure l l B ) , the size of
which scales with the width of original cyclonic
circulation.
Atmospheric and oceanic plumes/thermals typically
interact with either a boundary or a density interface,
Figure 10 A point thermal (A) and a plume (B) descending in a whence interesting phenomena arise. Figure 12 shows
homogeneous environment toward a solid surface. Note the the impingement of a descending line plume on
formation of well-defined vortices upon plume impingement on a density interface, followed by the generation
the surface, illustrating the structure of horizontal wind shear of a gravity current. When the interface is strongly
generated during microburst impacts.
stable, the evolution following the impingement
is similar to that at a solid boundary (Figure 12A),
but substantial penetration of the plume through
(buoyancy per unit length) released at the source. The the interface and turbulent mixing is possible
dynamics of point and line thermals and plumes in when the interface is weak (Figure 12B). Back-
homogeneous and stratified flows have been exten- ground rotation induces an anticyclonic circulation
Figure II The plan view of a plume descending into a homogeneous fluid. (A) Note the generation of a rim current (indicated by an
arrow) surrounding the source. (B) Breakdown of the rim current into eddies. (Reproduced from Fernando HJS, Chen RR and Ayotte BA
(1998) Development of point plumes in the presence of background rotation. Physics of Fluids lO(9):2369-2383.)
Figure 12 Impingement of a dense line plume on a density

interface. (A) A ‘strong’ interface deflects the current, Rip = 53.(B)
A moderately stable interface allows the plume to penetrate into the
lower layer and then spread horizontally, Rip = 1 ; the arrow shows
the interfacial location. Here the interfacial ‘strength’ is measured in
terms of a Richardson number Rip = A b ! , / $ based on the Figure 13 The impingement of a (dyed) descending point plume
buoyancy jump across the interface A6 and the velocity up and with a boundary in a rotating fluid (cf Figure 1OB). The plume is in
length scale tPof the impinging plume. (Adapted from Ching CY, the center. Note the formation of a frontal surface (A) and its
Fernando HJS and Noh Y (1993) Interaction of a negatively breakdown to form anticyclonic eddies. (6) The cyclonic eddies are
buoyant line plume with a density interface. Dynamics of Afmo- marked by particle streaks and the anticyclonic eddies are seen as
spheres and Oceans 19:367-388.) colored regions. They are located one above the other, forming the
structures known as hetons. (Reproduced from Fernando HJS,
Chen RR and Ayotte BA (1998)Development of point plumes in
the presence of background rotation. Physics of Fluids lO(9):
within this outflow, thus generating frontal shear 2369-2383.)
and ultimately breaking down the current to form
a series of anticyclonic vortices (Figure 13A, B).
The size of the current at the onset of instabilities
and the eddy sizes scale with the Rossby deforma- than about lODo), however, the memory of source
tion radius, pointing to the dominance of baro- dimensions is lost and the plume evolution resembles
clinic instability. When eddies form away from that of a point plume. Near the source, Do
the solid boundaries, they can achieve approximate is important, and thus Deardorff’s free convection
geostrophic balance, with the ratio of the eddy radius scaling fails. For example, the velocity scale near
Re to the height he being proportional to N/2!2, the source is given by (40D0)1/3,based on the
yieldingal’randtlratioof Nhe/2Re!2E (0.45-0.8). At buoyancy flux per unit area 40. As in the case of point
solid boundaries, viscous effects become important plumes, the presence of rotation leads to a rim current
and eddies are strongly ageostrophic. that breaks down to form cyclonic vortices, and
Cyclonic vortices generated by source-level rim the outflow at greater depths carrying dense fluid
current and anticyclonic vortices formed at greater becomes unstable to form anticyclonic eddies.
depths by the breakdown of radial outflow interact The evolution of distributed plumes in rotating fluids
to form a system of vortices in which the rota- depends on the relative magnitudes of Do and
tion varies from anticyclonic to cyclonic over the characteristic Rossby deformation radius L R .
the depth (Figure 13B); these belong to the class When Do < LR, the plume evolution is qualitatively
of vortex structures known as ‘hetons’. The thermal similar to the point source case described above.
wind balance maintains the vertically varying Conversely, when Do > LR, the distributed plume
circulation of these vortices, and the interface be- is baroclinically unstable from the outset, and the
tween lighter upper and dense lower parts of plume evolution is dominated by the lateral eddy
the hetons domes to maintain the geostrophic shedding.
balance. Interesting phenomena arise when the vertical
When the buoyancy source is distributed (say development of plumes is affected by background
over the diameter D O ) ,then point plume approxima- stratification. For example, in the absence of rota-
tion becomes invalid. Far from the source (greater tion, a point plume released into a uniformly stratified
538 CONVECTION/ Laboratory Models
fluid of buoyancy frequency N rises to its neutrally Taylor number

buoyant level h N cz 3q;l3/N and forms a horizontal Taylor number
intrusion. When background rotation is present, Vertical temperature gradient
the plume can be affected by background rotation, Velocity scale of rotating con-
either in the rising or horizontal spreading phase. vective turbulence (subscript p
The latter scenario is possible in deep convection denotes plumes)
of high-latitude oceans in which horizontally spread- Gravity current speed
ing gravity currents become unstable and shed Convective velocity of Dear-
off anticyclonic (geostrophic) eddies (as in dorff
Figure 13). W Width of the convective source
z Distance from the heated sur-
face
a Thermal expansivity
Nomenclature V Kinematic viscosity
a Inner radius of the dishpan IC Thermal diffusivity
A Aspect ratio ( W / d ) Po Reference density
b Outer radius of the dishpan Ab Buoyancy jump across a den-
CP Specific heat sity interface
d Depth of the fluid layer
DO Diameter of the buoyancy Rate of rotation
source
E = N2d2/ Eady number
4 n 2 ( b- a)2 See also
i? gravitational acceleration
Baroclinic Instability. Boundary Layers: Convective
Height of the eddies
Boundary Layer. Convection: Convection in the Ocean.
Neutral depth of a plume in a Coriolis Force. Frontogenesis.General Circulation:
stratified layer Overview. Hadley Circulation. Katabatic Winds.
hP Travel distance of a plume Laboratory Geophysical Fluid Dynamics. Micro-
front bursts. Parameterization of Physical Processes:
Depth where rotation affects Turbulence and Mixing. Tropical Meteorology: Inter
the convective turbulence Tropical Convergence Zones (ITCZ).
Length scale of the convective
turbulence (subscript p denotes
plumes) Further Reading
LR Rossby deformation radius
N Background buoyancy fre- Blumen W (ed.) (1990) Atmospheric Processes Over
quency Complex Terrain. Meteorological Monographs 23 (45).
American Meteorological Society.
Prandtl number
Bodenschatz E, Pesch W and Ahlers G (2000) Recent
Buoyancy flux developments in Rayleigh-Benard convection. Annual
Buoyancy flux of a plume Review of Fluid Mechanics 32: 709-778.
Total buoyancy of a thermal Chandrasekhar S (1961) Hydrodynamic and Hydromag-
Qo Heat flux netic Stability. New York: Oxford University Press.
Ra = g a T z d 4 / ~ v Rayleigh number Emanuel KA (1994) Atmospheric Convection. New York:
Raf = qod4 / K 2 v Flux Rayleigh number Oxford University Press.
Re Radius of eddies Fernando HJS and Smith DC IV (2001)Vortex structures in
Ri = Abd/w: Richardson number (subscript geophysical convection. European Journal of Mechanics
p denotes plumes) B -Fluids 20: 347-470.
Rossby number Fischer H, List J, Koh R, Imberger J and Brooks N (1979)
Mixing in Inland and Coastal Waters. London: Academic
Thermal Rossby number of
Press.
dishpan experiments Houghton J (1989) The Physics of the Atmosphere.
t Time Cambridge: Cambridge University Press.
Ta Inner temperature of the dish- Hide R and Mason PJ (1975) Sloping convection in a
pan rotating fluid. Advances in Physics 24: 47-100.
Tb Outer temperature of the dish- Hunt JCR (1998) Eddy dynamics and kinematics
pan of convective turbulence. In: Plate EJ et al. (eds),
CONVECTIVE CLOUD SYSTEMS MODELING 539
Buoyant Convection in Geophysical Flow. Dordrecht: Tritton DJ (1977) Physical Fluid Dynamics. Reading: Van
Kluwer. Nostrand Reinhold.
Maxworthy T (1997) Convection into domains with open Siggia ED (1994)High Rayleigh number convection. Annual
boundaries. Annual Review of Fluid Mechanics 29: Review of Fluid Mechanics 26: 137-168.
327-3 71. Turner JS (1979) Buoyancy Effects in Fluids. Cambridge:
McIlveen R (1992) Fundamentals of Weather and Climate. Cambridge University Press.
London: Chapman and Hall. Whiteman CD (2000) Mountain Meteorology, Fundamen-
Scorer RS (1978)Environmental Aerodynamics. New York: tals and Applications. New York: Oxford University
Ellis Horwood. Press.
W-K Tao, NASA Goddard Space Flight Center, parameterizations, a quest that continues to this day.
Greenbelt, MD, USA The effect of ice processes on cloud formation and
M W Moncrieff, Mesoscale & Microscale Meteorology evolution, stratiform rain processes and their relation
Division, Boulder, CO, USA to deep convective rainfall were focal points during
this period, as was the organizing effect of environ-
mental windshear on mesoscale convective systems.
The impact of radiative processes on cloud develop-
ment was investigated in the late 1980s. In the 1990s,
Introduction cloud-resolving models (with grid resolutions of 1-3
Numerical cloud models, which are based on the kilometers) began to quantify multiscale interactions,
nonhydrostatic equations of motion, have been ex- cloud chemistry interactions, cloud-climate interac-
tensively applied to cloud-scale and mesoscale pro- tion and surface processes. An important application
cesses during the past four decades. Because cloud- was the design of satellite rainfall-retrieval algori-
scale dynamics are treated explicitly, uncertainties thms for the Tropical Rainfall Measuring Mission
stemming from convection that have to be parame- (TRMM). Table 1 lists the major highlights of cloud-
terized in (hydrostatic)large-scale models are obviat- resolving modeling over the past four decades.
ed, or at least mitigated, in cloud models. Global Theoretical studies have advanced our basic know-
models will require a nonhydrostatic framework when ledge of convection dynamics, and also how convec-
their horizontal resolution approaches 10 km, the tion interacts with large scales of motion. These
theoretical limit of the hydrostatic approximation. studies enable the complexity of numerically simulat-
This juncture will be reached in one or two decades. ed clouds to be reduced to first principles, which is
The earliest kind of cloud model, the one-dimen- essential to understand the role of moist processes in
sional entraining bubble or plume that simply param- the Earth’s weather and climate at a basic level.
eterizes the lateral entrainment of environmental air, During the past generation, voluminous data sets on
was used extensively in cloud-seeding research. It is atmospheric convection have accumulated from radar,
still used in convective parameterization schemes. In instrumented aircraft, satellites, and rawinsonde
the 1 9 6 0 two-dimensional
~~ cloud models (with grid measurements in field campaigns, enabling detailed
sizes of a few hundred meters) were developed to study evaluation of models. Improved numerical methods
cloud evolution in idealized environments. Three- have resulted in more accurate and efficient dynamical
dimensional cloud models (with grid sizes of 1-2 cores in models. Improvements have been made in the
kilometers), developed in the early 1970s, quantified parameterizations of microphysical processes, radia-
the effects of wind shear on convection, for example, tion, boundary layer effects, and turbulence; how-
squall lines and mid-latitude supercell thunderstorms, ever, microphysical parameterizations remain a major
which are associated with tornado genesis. source of uncertainty in all classes of atmospheric
During the late 1970s and early 1980s, cumulus models.
ensemble models (also known as cloud-resolving In recent years, exponentially increasing computer
models with few hundreds meter grid resolutions) power has extended cloud model integrations from
began to quantify the collective effect of convection on hours to months, and the number of computational
the large-scale environment, with emphasis on the grid points from less than a thousand to close to
Tropics. A primary objective was to improve cumulus ten million. Three-dimensional models are now more
Buoyant Convection in Geophysical Flow. Dordrecht: Tritton DJ (1977) Physical Fluid Dynamics. Reading: Van
Kluwer. Nostrand Reinhold.
Maxworthy T (1997) Convection into domains with open Siggia ED (1994)High Rayleigh number convection. Annual
boundaries. Annual Review of Fluid Mechanics 29: Review of Fluid Mechanics 26: 137-168.
327-3 71. Turner JS (1979) Buoyancy Effects in Fluids. Cambridge:
McIlveen R (1992) Fundamentals of Weather and Climate. Cambridge University Press.
London: Chapman and Hall. Whiteman CD (2000) Mountain Meteorology, Fundamen-
Scorer RS (1978)Environmental Aerodynamics. New York: tals and Applications. New York: Oxford University
Ellis Horwood. Press.
W-K Tao, NASA Goddard Space Flight Center, parameterizations, a quest that continues to this day.
Greenbelt, MD, USA The effect of ice processes on cloud formation and
M W Moncrieff, Mesoscale & Microscale Meteorology evolution, stratiform rain processes and their relation
Division, Boulder, CO, USA to deep convective rainfall were focal points during
this period, as was the organizing effect of environ-
mental windshear on mesoscale convective systems.
The impact of radiative processes on cloud develop-
ment was investigated in the late 1980s. In the 1990s,
Introduction cloud-resolving models (with grid resolutions of 1-3
Numerical cloud models, which are based on the kilometers) began to quantify multiscale interactions,
nonhydrostatic equations of motion, have been ex- cloud chemistry interactions, cloud-climate interac-
tensively applied to cloud-scale and mesoscale pro- tion and surface processes. An important application
cesses during the past four decades. Because cloud- was the design of satellite rainfall-retrieval algori-
scale dynamics are treated explicitly, uncertainties thms for the Tropical Rainfall Measuring Mission
stemming from convection that have to be parame- (TRMM). Table 1 lists the major highlights of cloud-
terized in (hydrostatic)large-scale models are obviat- resolving modeling over the past four decades.
ed, or at least mitigated, in cloud models. Global Theoretical studies have advanced our basic know-
models will require a nonhydrostatic framework when ledge of convection dynamics, and also how convec-
their horizontal resolution approaches 10 km, the tion interacts with large scales of motion. These
theoretical limit of the hydrostatic approximation. studies enable the complexity of numerically simulat-
This juncture will be reached in one or two decades. ed clouds to be reduced to first principles, which is
The earliest kind of cloud model, the one-dimen- essential to understand the role of moist processes in
sional entraining bubble or plume that simply param- the Earth’s weather and climate at a basic level.
eterizes the lateral entrainment of environmental air, During the past generation, voluminous data sets on
was used extensively in cloud-seeding research. It is atmospheric convection have accumulated from radar,
still used in convective parameterization schemes. In instrumented aircraft, satellites, and rawinsonde
the 1 9 6 0 two-dimensional
~~ cloud models (with grid measurements in field campaigns, enabling detailed
sizes of a few hundred meters) were developed to study evaluation of models. Improved numerical methods
cloud evolution in idealized environments. Three- have resulted in more accurate and efficient dynamical
dimensional cloud models (with grid sizes of 1-2 cores in models. Improvements have been made in the
kilometers), developed in the early 1970s, quantified parameterizations of microphysical processes, radia-
the effects of wind shear on convection, for example, tion, boundary layer effects, and turbulence; how-
squall lines and mid-latitude supercell thunderstorms, ever, microphysical parameterizations remain a major
which are associated with tornado genesis. source of uncertainty in all classes of atmospheric
During the late 1970s and early 1980s, cumulus models.
ensemble models (also known as cloud-resolving In recent years, exponentially increasing computer
models with few hundreds meter grid resolutions) power has extended cloud model integrations from
began to quantify the collective effect of convection on hours to months, and the number of computational
the large-scale environment, with emphasis on the grid points from less than a thousand to close to
Tropics. A primary objective was to improve cumulus ten million. Three-dimensional models are now more
540 CONVECTIVE CLOUD SYSTEMS MODELING
Table 1 Major highlights of cloud modelingdevelopment overthe

Dast four decades
1960s Water loading, buoyancy and entrainment
1970s Slab- vs. axis-symmetric models

Cloud seeding
Cloud dynamics and warm rain
1980s Ensemble of clouds - cumulus parameterization

Cloud interactions and mergers
Ice processes
Squall line
Convective and stratiform clouds
Wind shear and cool pools
Gravity wave and density current
Large-scale and cloud-scale interactions
Cloud-radiation interaction
1990s Super cell dynamics

2D vs. 3D
Landandocean processes
Multiscale interactions
Figure 1 Representation of the three-class ice scheme used in
Cloud chemistry
the cloud model.
Process modeling - climate variation implications
GEWEX Cloud System Study (GCSS)
Coupling with microwave radiative model for satellite
cloud retrieval (TRMM)
be spherical. The warm cloud microphysics assumes
the population of water particles is bimodal, consist-
ing of small cloud water droplets whose terminal
prevalent. Much attention is devoted to precipitating velocity is minute compared to typical vertical
cloud systems where the crucial 1km scales are air velocities, and large rain droplets that obey certain
resolved in horizontal domains as large as 10 000 km size distributions based on limited observations.
in two dimensions, and lo6km2 in three dimensions. Condensation, evaporation, and autoconversiordcol-
Cloud models (cloud-resolving models) are thus a lection processes (from small cloud droplets to
comprehensive facility whether used in an idealized large rain droplets) are parameterized. The ice micro-
framework or in close conjunction with observations. physics assumes three types of particles: small cloud
Observations provide both the initial conditions and ice whose terminal velocity is also minute compared to
data for model validation; model results provide typical vertical air velocities, snow whose terminal
statistical information useful for developing physically velocity is about 1-3 m s-', and large sized graupel or
based parameterization for climate models and nu- hail with faster terminal velocities. Graupel has a
merical weather prediction models. low density and a high intercept (i.e., high number
concentration). In contrast, hail has a high density
and a small intercept. The choice of graupel or hail
Physical Processes in Cloud Models depends on where the clouds or cloud systems devel-
Cloud microphysical processes (phase changes of oped. For tropical clouds, graupel is more represent-
water and precipitation) must be parameterized in ative than hail. For mid-latitude clouds, hail is more
cloud models, as must atmospheric turbulence (dissi- representative. More than 25 transfer processes
pation of kinetic energy), turbulent processes at between water vapor, liquid and ice particles are
oceanic or terrestrial boundaries (latent and sensible included. These include the growth, of ice crystals
heat fluxes into the atmosphere), and radiative transfer by riming, the aggregation of ice crystals, the forma-
processes (complex in the presence of clouds). tion of graupel and hail, the growth of graupel and hail
by the collection of supercooled rain drops,
the shedding of water drops from hail, the rapid
Microphysics and Precipitation
growth of ice crystals in the presence of supercooled
Figure 1 depicts the widely used two-class liquid water, the melting of all forms of ice, and the
(cloud water and rain droplet) and three-class ice deposition and sublimation of ice. Only large rain
(cloud ice, snow and graupel/hail) microphysics droplets, snow and graupel/hail fall toward the ground
schemes. The shapes of liquid and ice are assumed to as precipitation.
Only recently have some cloud-resolving models Radiation

adopted a two-moment four-class ice scheme that
Emission and absorption by water vapor and cloud
combines the main features of the three-class
droplets are represented by two-stream long-wave
ice schemes by calculating the mixing ratios of
radiative transfer schemes. Broadband methods for
both graupel and frozen dropdhail. Additional model
long-wave radiation combine the effects of reflection,
variables include the number concentrations of
emission, and transmission by cloud droplets and air
all ice particles (small ice crystals, snow, graupel,
molecules. The treatment of short-wave radiation is
and frozen drops), as well as the mixing ratios of
also based on broadband approximations. One key
liquid water for each of the precipitation ice
issue is how to parameterize cloud optical properties
species during wet growth and melting for purposes
(optical thickness), especially in the presence of the ice
of accurate active and passive radiometric calcula-
phase, in view of the important impact of radiative
tions.
heating and cooling profiles within clouds. Note that
In addition, explicit bin-microphysical schemes
only limited observations are available upon which to
have been developed for cloud models for the study
base parameterizations for ice clouds.
of cirrus development and cloud-aerosol interaction.
The use of a fully explicit microphysics scheme
The formulation of the explicit bin-microphysical
(liquid and ice) and a fine horizontal resolution can
processes is based on solving stochastic kinetic equa-
provide relatively realistic cloud optical properties,
tions for the size distribution functions of water
which are crucial for determining the radiation bud-
droplets (cloud droplets and raindrops), and ice
gets. With high spatial resolution, each atmospheric
particles of different habits (columnar, platelike,
layer is considered either completely cloudy (overcast)
dendrites, snowflakes, graupel, and frozen drops).
or clear. N o partial cloudiness is assumed.
Each type is described by a special size distribution
function containing over 30 categories (bins). Nucle- Ocean Surface Fluxes
ation (activation)processes are also based on the size
distribution function for cloud condensation nuclei Two types of surface flux schemes are typically used.
(also over 30 size categories). Because of the numerous The first is a simple bulk aerodynamic formula where
interactions involved in bin-microphysical schemes, the transfer coefficients for momentum, sensible heat,
computational domains are small and simulation and latent heat fluxes are functions of wind speed only.
times are short. These detailed microphysics calcula- The second type is more complex but, nevertheless,
tions can provide a useful framework for evaluating primarily a bulk approach. The transfer coefficients
and ultimately improving bulk microphysical for momentum, sensible heat, and latent heat fluxes
schemes. Of particular interest in this regard is are based on the Monin-Obukhov similarity theory of
convectively generated cirrus, which has a major the atmospheric surface layer. The parameters, such as
impact on the radiative properties of the tropical the roughness lengths, are closely related to the sea
atmosphere. surface characteristics and the turbulence characteris-
tics. At very low wind speeds the similarity profile is
singular, a problem addressed by adding a convective
Turbulence velocity to yield nonzero fluxes under windless con-
While large eddies are resolved in cloud models, eddies ditions. The exchange coefficients in the simple bulk
much smaller than the grid-scale must be parameter- aerodynamic formula method and in the second bulk
ized. An implicit assumption is that the small scales flux algorithm differ in two ways. First, in the lower
approximate to an inertial subrange where the energy wind speed regime (less than 4ms-'), the exchange
spectrum is in statistical equilibrium, with an energy coefficients in the complex bulk scheme increase with
cascade from the resolved scales to the dissipation decreasing wind speed in order to account for the
scales. The most sophisticated turbulence parameter- convective exchange at low wind speeds. Second, the
ization presently use is a third-moment closure. coefficients in the simple bulk aerodynamic formula
Typical cloud models used simple k-type (first-order) linearly increase with respect to the wind speed, but
turbulence closure or determine the coefficient k, decrease if the wind speed is greater that 5 m s-l in the
diagnostically or prognostically, from the turbulence more complex bulk schemes. These differences in the
kinetic energy (TKE)equation (one-and-a-half order). exchange coefficients can affect rainfall amounts and
In the prognostic TKE method, thermodynamic sta- boundary layer structure.
bility, deformation, shear stability, diffusion, dissipa-
Land Surface Processes
tion and transport of subgrid energy are included. In
the diagnostic method, deformation and stability are Detailed interactive land surface process models of the
used for computing the k coefficient. heterogeneous land surface (soil and vegetation) and
adjacent near-surface atmosphere have recently been vection). This approach has been used extensively in
applied in cloud models to study the effect of soil the west Pacific warm pool region (TOGA COARE)
moisture distribution and atmospheric boundary and the eastern Atlantic region (GATE).In the second
conditions on cloud structure, rainfall, and soil mois- initial-value problem (cf., the classical cloud model),
ture distribution. A land surface model usually has convective evolution over periods of hours is simulat-
three elements: (1)a soil module that includes at least ed and the initiation (ortriggering) of convection is the
several water reservoirs (i.e., plant internal storage, primary issue: cold pools, surface fluxes, or stochastic
dewhntercepted precipitation, surface material, a perturbation excite locally forced convection. Such
topsoil root layer, a subsoil root layer, and two deeper simulations are very useful for model development,
layers that regulate seasonal and interannual varia- especially when conducted in conjunction with field
bility of the soil hydrology); (2) a surface slab of campaigns (e.g., ilz situ surface-based and aircraft
vegetation, litter, and other loose material that shades observations; ground-, aircraft-, and space-based re-
the soil and acts as the source for sensible heat flux, and mote sensing) that provide high-resolution data for
intercepts precipitation and dew; and ( 3 ) the surface model validation. Henceforth, the terms cloud model,
layer of the atmosphere (up to the lowest grid level of cloud-resolving model, and cumulus ensemble model
the model to which it is coupled), within which the will be used interchangeably.
fluxes of sensible heat and water vapor are calculated.
Organization of Convection on Mesoscales
Modeling Tropical Convective Systems The past few decades have witnessed advances in the
Tropical convection affects the large-scale circulation understanding of organized convection, with convec-
of the atmosphere, and in particular atmosphere- tion over the tropical oceans being a focus. Figures 2A
ocean coupling, in important ways. The multiscale and B show the evolution of numerically simulated
organization of tropical convection is manifested as convective cloud systems in the west Pacific warm
cloud streets in the shallow trade-wind convection, pool region and eastern Atlantic region, respectively.
cumulonimbus, cloud clusters, squall lines, tropical In the former, cloud systems travel in one direction and
cyclone rainbands, convection in westerly wind bursts embedded convection in the opposite direction. Syn-
and intraseasonal oscillations, and enhanced convec- optic-scale easterly waves strongly modulate convec-
tion in the ITCZ (i.e., the rising branch of the Hadley tion in the eastern Atlantic because they strongly affect
circulation) and in planetary-scale Walker circula- environmental shear and large-scale forcing, and
tions. Within the next few decades, computers will be thereby convective organization and intensity. Shal-
powerful enough to resolve this entire multiscale low convection evolves to nonsquall (slow-moving)
hierarchy, but until this juncture is reached many basic deep precipitating cloud systems that travel westward,
issues need to be addressed, not least in parameter- steered by the mean wind. Fast-moving squall systems
ization (e.g., convective organization and its implica- subsequently develop as the lower-tropospheric shear
tions for the basic scale-separation principle). intensifies. The simulated cloud systems become less
However, it is important to realize that resolving organized and produce less surface precipitation as the
convection will not obviate parameterization, but forcing decreases.
rather move the problem downscale, (e.g., to the The simulated domain-averaged surface rainfall
boundary layer) and make greater demands on the (mm), and stratiform amount (percentage) for
microphysical parameterizations. both the west Pacific warm pool and the east Atlantic
regions are shown in Table 2. The ratios between
evaporation and condensation, sublimation and
Ensemble Versus Local Convection
deposition, and deposition and condensation illustrate
Convective cloud models can be broadly categorized the relative importance of liquid vs. ice processes
as follows. First, a quasi-statistical approach in which and source and sink terms associated with water
clouds of different types in various stages of evolution vapor. The microphysical processes are decomposed
are simulated in large domains (viz., cumulus ensem- according to convective organization: slow-moving,
ble models or cloud-resolving models). A key objective fast-moving, less organized convective episodes
is to quantify how convection interacts with the large from the east Atlantic region, vigorous deep convec-
scales of motion. Large-scale ‘forcing’, the primary tion and weaker convective events during a westerly
source of convective available potential energy, can be wind burst episode in the western Pacific accompany-
derived from an objective analysis of sounding net- ing the convectively active phase of the intraseasonal
works and applied as domain-mean tendencies of oscillation wherein the stratiform component was
temperature and moisture (continuously forced con- larger. The dominance of warm-rain processes in
Figure2 Time-sequence of the two-dimensional cloud model estimated domain mean surface rainfall rate (103 m h-’) for (A) the west
Pacific warm pool region and (6)the east Atlantic region.
the east Atlantic region squall and nonsquall convec- Figure 3 shows numerically simulated three-dimen-
tive systems explains the smaller stratiform rain sional cloud systems over the west Pacific warm pool.
amounts. The depleted ice processes on 6 and 8 Organized mesoscale convective systems consist
September are indicative of shallow convection. In of families of leading-edge, quasi-linear, heavily
contrast, ice processes are important for both active precipitating cumulonimbus followed by an extensive
and relatively inactive convective periods over the area of trailing light (stratiform) precipitation.
west Pacific. The cumulonimbus slope with height owing to the
544 CONVECTIVECLOUD SYSTEMS MODELING
Table 2 Cloud model-simulated domain-average surface rainfall

(mm), stratiform amount (percentage), and microphysical pro-
cesses (ratios between evaporation and condensation, sublima-
tion and deposition, and deposition and condensation) for (a) the
west Pacific warm pool region and (b) the east Atlantic region. For
west Pacific warm pool region, the cloud model results are also
separated into subperiods, deep strong convection during 20-23
and 24-25 December and weaker convection prior to, between,
and after the deep convection (19-20,23-24, and 25-26 Decem-
ber 1992). Slow-moving (nonsquall, 2 - 4 September), fast-moving
(squall, 4-6 September), and less-organized (6-8 September)
periods for the cloud model-simulated east Atlantic region results
are also shown
(a) West Pacific Warm Pool Region (19-26 December 1992)
After WWB Pre and

(2Days) during
WWB
(4 days)
Total surface rainfall 50.8 85.1
(mm)
Stratiform amount (“A) 42.0 48.0
Evaporation/ 62.0 73.0
condensation (Yo)
Sublimation/ 46.0 50.0
deposition (Yo)
Deposition/ 34.0 41 .O
condensation (“A)
(b) East Atlantic Region (2-8 September 1974)

Figure3 (A) Horizontal and (B) vertical cross-sections of vertical
Slow-moving Fast-moving Random velocity (filled contours) and total cloud mixing ratio (solid contour)
taken from a three-dimensional cloud model simulation of the west
Total surface rainfall 43.34 39.62 8.50
Pacific warm pool region precipitating system (during a Westerly
(mm) Wind Burst episode). The location of the vertical cross-section
Stratiform amount (“A) 27.0 26.0 44.0
shown by the vertical line in (A).
Evaporation/ 58.0 44.0 68.0
condensation (Yo)
Sublimation/ 36.0 27.0 24.0
deposition (“A)
by the mean flow. This multiscale organization occurs
Deposition/ 23.0 25.0 9.5 even with constant sea surface temperature, indicating
condensation (Yo) that it is self-organizing.
The above modeling studies have quantified many
observed properties of convection and its multiscale
environmental wind shear. Mesoscale descent behind organization. The challenge now is to understand how
the leading edge develops mainly at low levels where this organization affects and is controlled by atmos-
evaporative cooling is strongest. Descent occurs above pheric properties, which is a fundamental issue of fluid
the mesoscale ascent, typically separated by the 0°C dynamics at large.
(melting) level. Evaporatively cooled downdraft out-
flows (density currents) provide localized uplift that
Heat and Moisture Budgets
triggers convection. Convectively generated gravity
waves occur in the upper troposphere. The collective effects of convection need to be repre-
How organized convection interacts with large- sented in parameterizations, rather than structural
scale tropical waves and intraseasonal oscillations is details. In this context, the budgets of heat, moisture,
only beginning to be quantified using multiscale cloud convective mass flux, and convective momentum
resolving models. Figure 4 shows a cloud-resolving transport derived from observations are useful.
model realization of Madden-Julian-like tropical Heat and moisture transports by convection strong-
intraseasonal variability Figure 4A shows an east- ly affect the general circulation of the atmosphere.
ward-traveling convective envelope, while Figures Transports are estimated from the objective analysis
4B-D show the westward-traveling organized con- of sounding networks as thermodynamic budget
vective systems embedded in the envelope and steered residuals - the apparent sources of heat (Q1) and
0 Condensate and precipitation (g kg-')
U
(B) Perturbation horizontal velocity (m s-')
20 --.3
Streamfunction
18
-E
s
E 9
.-cn
I"
40 0
-250 0 250
(D) Distance from x = 8300 km
Figure 4 (A) The space-time distribution of surface precipitation from a cloud-resolving 40-day, two-dimensional 20 000 km domain
simulation of tropical convection on a constant sea surface temperature aquaplanet. Dotted line shows the eastward-propagating
convective envelope. Parts (B) to (D), along the continuous line, show the highly organized, westward-traveling mesoscale convective
systems within the convective envelope: (B) condensate and precipitation; (C) perturbation velocity; and (D) streamfunction.
moisture ( Q 2 ) . Figure SA shows that convec- difference (balance) between the cloud processes
tive heating has a maximum in the 600-650 hPa layer. (response/feedback) and large-scale forcing is indica-
In the stratiform region (Figure 5B), heating is tive of the quasi-equilibrium state of the tropical
maximized in the upper troposphere (around atmosphere.
400 hPa), with cooling prevailing below the melting
level. The stratiform heating is smoother because
Convective Mass Flux
convective bursts have a more rapid evolution than the
mesoscale processes that dominate in the stratiform While convective mass flux is an important quantity
region. Also, stronger heating occurs aloft, and in the parameterization of convection in large-
stronger cooling below in the stratiform region owing scale models, it is difficult to observe accurately.
to the evaporation of rain produced by melting ice Figure 6 shows the 7-day evolution of simulated
particles. cloud mass fluxes (total condensate exceeding
Figures 5C and 5D show the corresponding appar- 0.1 g kg-I). The larger mass fluxes trace the organized
ent moisture sink Q2. Drying in the convective region cloud systems (nonsquall clusters, days 2 and 5;
is caused by the condensation processes associated and squall line, day 4).Evaporative cooling associated
with cloud updrafts, and its maximum is lower than with the downdrafts is about half of the condensatio-
the apparent heat source. In the stratiform region, nal heating in the updrafts (Table 2). Convective
there is strong moistening (by evaporative cooling) updrafts account for approximately 75% of the cloud
below the 600 hPa level with weak drying aloft. Cloud updraft mass flux, yet they occupy a mere 12-14% of
models also showed that the eddy transport of heat is the total area; these so-called 'hot towers' play a
one order smaller than the effects of the microphysical critical role in the heat and moisture budgets in the
processes, contrasting with the eddy transport of tropics, despite the small fractional area they occupy.
moisture, which is of the same order. These distinctive Downdrafts account for about 30% of downdraft
heating/moistening patterns are consistent with mass flux, suggesting they are active only over
observed mesoscale convective systems. The small relatively small areas.
Figure5 Evolution of the simulated apparent sources averaged over the west Pacific warm pool region for the period 19-27 December
for the stratiform region: (C) Q2 for the convective region; and (D) 4 for the stratiform region.
1992. (A) QIforthe convective region; (B) 01
Convective Momentum Transport organized flow. Cloud models are instrumental in

quantifying convective momentum transport and
The vertical transport of momentum by convection in deriving a physically based parameterization.
affects the conversion of kinetic energy from sub-grid- Convectively generated gravity waves affect the
scale eddies to the mean flow, the rate of frictional momentum balance of the atmosphere, and are
dissipation and, therefore, the atmospheric energy particularly important in the tropical stratosphere
spectrum. However, the large-scale effects of momen- and mesosphere. The wave generation mechanism and
tum transport are poorly understood. The horizontal its vertical propagation to the deep atmosphere are
pressure gradient force within cloud systems, which is being studied using cloud models.
the primary quantity affecting momentum transport,
is difficult to measure observationally. Moreover,
Diurnal Variability of Precipitation
the representativeness of measurements is an issue
because the pressure field is a strong function of Cloud models have quantified the mechanisms re-
convective dynamics. Theoretical models show that sponsible for the diurnal cycle of precipitation pro-
convective momentum transport can either be upgra- cesses over the tropical oceans associated with the
dient, which enhances the mean flow, or downgradient diurnal variation of radiation. Variability of rainfall
in which case it is a mixing process. The existence of occurs even if there is no diurnal variation of sea
these (opposing) effects is supported by observations. surface temperature; however, the maximum rainfall
Entropy considerations would suggest that, on aver- is shifted from nighttime (0200 LST) to early morning
age, momentum transport must be downgradient, (about 0500 LST). Thus, the diurnal variation of sea
with upgradient transport occurring only in special surface temperature modulates the rainfall, but may
conditions (e.g., in highly organized squall systems). play a secondary role in its diurnal variation. Cloud
However, the strongest momentum fluxes occur in models also indicate that convection is modulated by
until the temperature and water vapor fields reach a

quasi-equilibrium state. They can produce different
h
E 12 quasi-equilibrium states (warm and humid versus cold

s and dry), even though similar initial thermodynamic
Em 8
.- profiles and fixed sea surface temperature are used.
$ 4 Stronger surface winds tend to produce a warmer and
(A) 0 t t t t more humid thermodynamic equilibrium state. The
moist static energy budget further indicates that the
large-scale forcing of water vapor is another process
responsible for warmer and more humid equilibrium
states.
Cloud-resolving models have also been used to
quantify hypotheses relating to global warming. Key
results to date are (1)conversion of ice-phase water
Total cloud mass flux Mc (hPa h-’) 3D
into the vapor phase associated with the dissipation of
5 12 upper-level stratifordcirrus clouds contributes to
v
upper tropospheric moisture on the same order as
Em 8
.- moisture transport from deep convection; (2) cloud
$ 4 activity is much more sensitive to convergence in the
0 t t t I large-scale atmospheric circulation over an oceanic
1 2 3 4 ’ 5 6 7 ’ warm pool than it is to the local sea surface temper-
(C) Day (September 1974) ature; and ( 3 ) organization of cloud systems can
Figure 6 Evolution of domain-averaged mass fluxes (hPa h-’), largely determine the magnitude of upper-level cloud-
(A) updraft, (B) downdraft, and (C) total mass fluxes, for simulated iness and moisture profiles. The above conclusions do
east Atlantic cloud systems. not say whether or not global warming is occurring,
only that if cloud processes are neglected or poorly
formulated, the consequences could lead to substan-
tial errors in important climate hypotheses.
the diurnal change in available water as a function of
temperature and is responsible for the nighttime
maximum in rainfall. This implies that the increase
(decrease) in surface precipitation associated with
Conclusion
long-wave cooling (solar heating) may be due t o an Observations together with cloud-resolving models
increase (decrease) in relative humidity. However, the are a promising approach to developing and testing
interaction of radiation with organized convection can representations of cloud processes in numerical
affect the diurnal variability of rainfall. Well- (less-) weather prediction models and climate models. Ob-
organized cloud systems can have strong (weak) servations provide both the initial conditions
diurnal variations in rainfall, but ice processes en- and model-validation data. Modern cloud models
hance the diurnal variation of precipitation. operate reasonable, albeit not complete, microphysi-
Prediction of the diurnal cycle of convective precip- cal parameterizations and simulate the evolution,
itation has very low skill over land, for reasons not structure, and life cycles of cloud systems. They also
well understood. The development of the diurnal explicitly calculate interactions between clouds and
evolution of the convective boundary layer, the role of long-wave and solar radiation that are difficult, if not
orography, and the effects of land surface processes are impossible, to measure observationally.
all involved. Concerning the diurnal cycle of precip- During the past four decades, cloud models have
itation over tropical islands, seabreezes and land- advanced from simple process models to full multi-
breezes and their interaction with coastlines and scale cloud-system simulators that span a dynamic
orography are key mechanisms. Cloud models have range of 1km to planetary scale. They are helping
addressed these aspects in considerable detail. improve our understanding of the interaction between
convection, radiation and the large-scale environ-
ment. They address basic issues in global and regional
Water and Energy Cycles
prediction, as well as fundamental problems associat-
Cloud-resolving models are beginning to be used in ed with the Earth’s water and energy cycles. They
studies of the tropical water and energy cycles. In this are now being used extensively to develop physi-
context, the models are typically run for several weeks cally based parameterizations. Because of the range
548 CONVECTIVE STORMS / Overview
of scales resolved by modern cloud models, space- ties; Precipitation. Thermodynamics: Saturated Adia-
based remote sensing is an ever more necessary part of batic Processes. Tropical Meteorology: Overview and
model validation. Theory. World Climate Research Program.
Finally, a new approach is to use cloud-resolving
models in place of traditional cumulus parameterizat-
ions in large-scale models. This explicit approach,
Further Reading
called cloud-resolving convection parameterization or Cotton WR and Anthes RA (1989) Storm and Cloud
super-parameterization, alleviates several uncertain- Dynamics. International Geophysics Series, Vol. 44.
ties associated with parameterization. Experimental San diego: Academic Press.
at present, it has the potential t o be viable in Grabowski WW and Moncrieff MW (2001) Large-scale
operational prediction models as advances continue organization of tropical convection in two-dimensional
explicit numerical simulations. Quarterly Journal of the
to be made in computer technology and processing
Royal Meteorological Society 127: 445-468.
speed. Houze RA Jr (1993) Cloud Dynamics. International Geo-
physics Series, Vol. 53. San Diego: Academic Press.
Ludlam FH (1980) Clouds and Storms: The Behavior and
See also Effect of Water in the Atmosphere. University Park: The
Air-Sea Interaction: Momentum, Heat and Vapor Flux- Pennsylvania State University Press.
es. Boundary Layers: Modeling and Parameterization. Moncrieff MW and Tao W-K (1999) Cloud-resolving
Cloud Microphysics. Convection: Convection in the models. In: Browning K and Gurney RJ (eds) Global
Ocean. Convective Storms: Overview. Density Cur- Water and Energy Cycles, pp. 200-209. Cambridge:
rents. Diurnal Cycle. Mesoscale Meteorology: Meso- Cambridge University Press.
scale Convective Systems; Models; Overview. Numerical Smith RK (1997) The Physicsatid Parameterizatio~ofMoist
Models: Methods. Parameterizationof Physical Proc- Atmospheric Convection. NATO Advanced Study Insti-
esses: Clouds. Radiative Transfer: Cloud-radiative tute Series C: Mathematical and Physical Sciences,
Processes. Satellite Remote Sensing: Cloud Proper- vol. 505.
Contents
Overview
Convective Initiation
lifetimes ranging from 30-40 min to greater than 6 h.

Ovetview Furthermore, groups of convective cells can become
organized into larger mesoscale convective systems,
M L Weisman, National Center for Atmospheric such as squall lines, bow echoes, and mesoscale
Research, Boulder, CO, USA convective complexes, which can extend over hun-
dreds of kilometers and, in some cases, can last for
several days.
Convective storms exist under a wide variety of
conditions and evolve in an equally wide variety of
Introduction ways. Storm behavior is inherently dependent on the
Convective storms, also commonly referred t o as environment in which the storm grows, including
thunderstorms, produce some of the fiercest weather thermodynamic stability, vertical wind profiles, and
on earth, including flooding rains (with rain rates up to mesoscale forcing influences. In the following, we
several inches or 100 mm h - I ) , severe surface winds review the properties of the most basic storm types,
(sometimes reaching magnitudes greater than 100 kn), including the ordinary cell, multicell, and supercell,
hail (reaching the size of grapefruit), frequent light- and explain the fundamental physical processes that
ning, and tornadoes. Individual convective cells are promote the various storm behaviors. Our knowledge
generally observed on scales of 5-30 km, and can have of convective storms is based largely on extensive
of scales resolved by modern cloud models, space- ties; Precipitation. Thermodynamics: Saturated Adia-
based remote sensing is an ever more necessary part of batic Processes. Tropical Meteorology: Overview and
model validation. Theory. World Climate Research Program.
Finally, a new approach is to use cloud-resolving
models in place of traditional cumulus parameterizat-
ions in large-scale models. This explicit approach,
Further Reading
called cloud-resolving convection parameterization or Cotton WR and Anthes RA (1989) Storm and Cloud
super-parameterization, alleviates several uncertain- Dynamics. International Geophysics Series, Vol. 44.
ties associated with parameterization. Experimental San diego: Academic Press.
at present, it has the potential t o be viable in Grabowski WW and Moncrieff MW (2001) Large-scale
operational prediction models as advances continue organization of tropical convection in two-dimensional
explicit numerical simulations. Quarterly Journal of the
to be made in computer technology and processing
speed. Houze RA Jr (1993) Cloud Dynamics. International Geo-
physics Series, Vol. 53. San Diego: Academic Press.
Ludlam FH (1980) Clouds and Storms: The Behavior and
See also Effect of Water in the Atmosphere. University Park: The
Air-Sea Interaction: Momentum, Heat and Vapor Flux- Pennsylvania State University Press.
es. Boundary Layers: Modeling and Parameterization. Moncrieff MW and Tao W-K (1999) Cloud-resolving
Cloud Microphysics. Convection: Convection in the models. In: Browning K and Gurney RJ (eds) Global
Ocean. Convective Storms: Overview. Density Cur- Water and Energy Cycles, pp. 200-209. Cambridge:
rents. Diurnal Cycle. Mesoscale Meteorology: Meso- Cambridge University Press.
scale Convective Systems; Models; Overview. Numerical Smith RK (1997) The Physicsatid Parameterizatio~ofMoist
Models: Methods. Parameterizationof Physical Proc- Atmospheric Convection. NATO Advanced Study Insti-
esses: Clouds. Radiative Transfer: Cloud-radiative tute Series C: Mathematical and Physical Sciences,
Processes. Satellite Remote Sensing: Cloud Proper- vol. 505.
Contents
Overview
Convective Initiation
lifetimes ranging from 30-40 min to greater than 6 h.

Ovetview Furthermore, groups of convective cells can become
organized into larger mesoscale convective systems,
M L Weisman, National Center for Atmospheric such as squall lines, bow echoes, and mesoscale
Research, Boulder, CO, USA convective complexes, which can extend over hun-
dreds of kilometers and, in some cases, can last for
several days.
Convective storms exist under a wide variety of
conditions and evolve in an equally wide variety of
Introduction ways. Storm behavior is inherently dependent on the
Convective storms, also commonly referred t o as environment in which the storm grows, including
thunderstorms, produce some of the fiercest weather thermodynamic stability, vertical wind profiles, and
on earth, including flooding rains (with rain rates up to mesoscale forcing influences. In the following, we
several inches or 100 mm h - I ) , severe surface winds review the properties of the most basic storm types,
(sometimes reaching magnitudes greater than 100 kn), including the ordinary cell, multicell, and supercell,
hail (reaching the size of grapefruit), frequent light- and explain the fundamental physical processes that
ning, and tornadoes. Individual convective cells are promote the various storm behaviors. Our knowledge
generally observed on scales of 5-30 km, and can have of convective storms is based largely on extensive
CONVECTIVE STORMS / Overview 549
radar studies (using both conventional and Doppler conditionally unstable atmosphere, producing large
radars) as well as numerical cloud modeling studies. amounts of liquid water and ice. When the raindrops
More information on convective storms can also be or ice particles become too heavy for the updraft to
obtained from related chapters on lightning, hail, support, they begin to fall, creating a downdraft that
tornadoes, mesoscale convective systems, bow echoes, quickly replaces the updraft. The downdraft is initially
convective storm modeling, and severe weather nearly saturated, but as it falls into the lower tropo-
forecasting. sphere and mixes with drier air, strong evaporational
cooling may occur. This cooling accelerates the
downdraft (because of negative buoyancy), which
Observed Convective Storm Types spreads out horizontally as a cold pool (gust front) on
The concept of the convective cell is fundamental to a reaching the surface. If the diverging outflow winds
discussion of convective storms. The convective cell reach severe levels (greater than about 50kn),
will be regarded as a region of strong updraft (greater the event is referred to as a downburst or microburst.
than 5 m s - ') and associated precipitating downdraft This life cycle (Figure 1) usually takes 30-50min to
having a horizontal cross-section of 10-100 km2, and complete, and generally severe weather such as high
extending in the vertical through most of the tropo- winds or hail tends to be short-lived. Relatively weak,
sphere. Intense convective cells can have updrafts short-lived tornadoes do occasionally occur with
greater than 60 m s -',with downdrafts sometimes ordinary cells, and are sometimes referred to as
greater than 3 0 m s - l . Research has shown that landspouts or non-supercell tornadoes.
convective cells as observed on radar often evolve in
identifiable, repeatable patterns. On the basis of these Multicell Storm
radar characteristics, conceptual models have been
proposed for the most commonly observed storm The multicell storm can be thought of as a cluster of
types. These include the short-lived ordinary cell, short-lived ordinary cells. The cold outflows from each
multiple cell systems or 'multicell', and supercell. cell, however, combine to form a large gust front, the
convergence and lifting along its leading edge being
generally strongest in the downshear direction relative
Ordinary-Cell Storm
to the low-level (0-3 km agl) vertical wind shear
The ordinary cell represents the most basic convective vector. In most cases, this also happens to be in the
storm type (Figure 1). It consists of a single updraft, direction of storm motion. This convergence and
which rises rapidly through the troposphere in a lifting can trigger new updraft development along and
~~ kilometers
. .
(8-16 krn) (6-J
Mature staae Dissipatin
(A)
Figure 1 (A) Towering cumulus stage, (B) mature stage, and (C) dissipating stage of an 'ordinary' convective cell. (Courtesy of CA
Doswell. Adapted with permissionfrom Byers HR and Braham RR Jr (1949) The Thunderstorm. Supt of Documents. Washington DC: US
Government Printing Office.
550 CONVECTIVESTORMS / Ovewiew
just behind the gust front, and new cells evolve as several small-scale rain centers embedded within a
described in the previous subsection. Figure 2 shows larger encompassing cellular structure. However,
this process in a vertical cross-section through a the general structure and evolution of the supercell
multicellular hailstorm observed during the National suggest that it is dynamically different from ordinary
Hail Research Experiment. The new cell growth often convection.
appears disorganized, but occasionally occurs on a A schematic of a supercell is presented in Figure 3.
preferred storm flank. Because of their ability to renew Unlike ordinary cells or multicell systems, supercells
themselves constantly through new cell growth, multi- are often characterized by a persistent separation
cell storms often last many hours, affecting areas between the primary updraft and downdraft currents.
thousands of square kilometers. If the storm motion is The updraft region is generally found on the upshear
very slow, heavy local rainfall may occur, presenting side of the cloud, and is characterized by a well-defined
the possibility of flooding. Severe surface winds in the cloud base with rapidly growing cloud turrets above.
form of downbursts or microbursts can occur with This portion of the storm often exhibits pronounced
multicell storm systems, with hail and tornadoes also cyclonic rotation. The downdraft region is found
possible in the vicinity of strong updraft centers. primarily downshear of the updraft region, appearing
more diffuse due to the heavy precipitation. The anvil
spreads predominantly downshear aloft, but in strong-
Supercell Storm
er storms also extends upshear somewhat as the
The supercell is potentially the most dangerous con- divergence near the storm top is able to force itself
vective storm, often producing high winds, large hail, upstream against the strong upper-level flow. Over-
and long-lived tornadoes. In its purest form it consists shooting tops are quite common for the stronger
of a single, quasi-steady, rotating updraft and associ- storms. Many supercells also display a stair-step-
ated downdraft, which may have a lifetime of several shaped flanking line extending upshear from the
hours. It often evolves from multicell storm systems, storm's main updraft region. A persistent lowering of
and even during its quasi-steady phase may comprise the cloud base, referred to as a wall cloud, is also often
Figure 2 Vertical cross-section through a multicell hailstorm, along the storm's direction of travel through a series of evolving cells
( n - 2,n - 1, n, n + 1). The solid lines are streamlines of flow relative to the moving system; on the left their broken ends represent flow
into and out of the plane, and on the right they represent flow remaining within a plane a few kilometers closer to the reader. Light shading
represents the extent of the cloud, and the three darker shades represent radar reflectivities of 35, 45, and 50dBZ. (Reproduced with
permission from Browning KA, Fankhauser JC, Chalon J-P, et a/. (1976) Structure of an evolving hailstorm. Part V: Synthesis and
implications for hail growth and hail suppression. Monthly Weather Reviews 104: 603-610.)
CONVECTIVESTORMS / Overview 551
Penetrating top
Anvil,
I \
s
3 / 1 2
. WYW,
T
Mamma
fl ,Vriga
ciuuu
Upwind side-
CIOUU ciuuu nail iiaii iaiii rairi iaiii
-10 km
Approximate
horizontal
scale
f- Downwind side
(usually north-east)
(usually south-west) Side view
Figure 3 Schematic visual view of a mature supercell thunderstorm. (Reproduced with permission from Bluestein HB and Parks CR
(1983) A synoptic and photographic climatology of low-precipitation severe thunderstorms in the southern plains. Monthly Weather
Reviews 111: 2034-2046.)
observed beneath the main updraft region of the A time series of radar reflectivity structure for a
storm, and is often a precursor to the development of storm that occurred on 19 April 1972 near Norman,
tornadoes. OK (Figure 6 ) portrays a commonly observed trait of
The structure of a mature supercell as it might supercell storms. About 1h into the storm’s lifetime,
be observed on radar is depicted in Figure 4. The the rain center appears to split into two diverging echo
reflectivity field tends to be elongated in the direction masses: the more intense southern storm veers to the
of the mean vertical wind shear, with a hooklike right and slows its motion, while the northern storm
appendage often appearing on the south-west flank of moves more quickly to the north-east. Such storm
the storm. The midlevel reflectivity often overhangs splitting is common in association with supercell
the low-level echo, and often a bounded weak echo storms. The right mover (right relative to the direction
region (BWER) appears at middle levels above the of the ambient shear vector) is associated with a
edge of the low-level reflectivity gradient. A BWER cyclonically rotating updraft while the left mover is
usually indicates the presence of both strong up- associated with an anticyclonically rotating updraft.
draft and strong rotation about a vertical axis in its Both right and left movers of a splitting storm are apt
vicinity. to produce severe weather such as hail and high winds,
Figure 5 presents the significant surface features but tornadoes are rarely associated with left-moving
commonly observed during a supercell’s mature phase. storms.
The main updraft region is found straddling the hook
or notch in the rain field, with two primary downdraft
regions, referred to the forward flank downdraft Physical Mechanisms Controlling
and rear flank downdraft, located on the downshear Convective Storm Growth and
and upshear sides of the updraft, respectively. A
surface gust front separates the cool, rainy air from
Evolution
the warm ambient air, with the gust front often Convective storm type and severity are strongly
wrapped around the southern flank of the storm due to dependent on the environmental conditions in which
the circulation associated with a surface mesocyclone. the storm grows. Of particular importance is the
This rear flank gust front can overtake the gust frontal thermodynamic instability (buoyancy) and vertical
boundary associated with the forward flank downwind shear. Thermodynamic instability exerts a fun-
draft, creating an occlusion of these frontal features. A damental control on convective storm strength, as it
tornado, if present, often forms at the tip of this controls the vertical acceleration of air parcels. Verti-
occlusion (on the edge of the hook echo) on the cal wind shear, however, influences strongly the form
gradient between updraft and downdraft (but within that the convection might take, i.e., whether
the updraft). the convection evolves as short-lived ordinary cells,
552 CONVECTIVESTORMS / Overview
Figure 5 Schematic plan view of a tornadic supercell thunder-

storm at the surface. The thick line encompasses radar echo. The
thunderstorm wavelike 'gust front' structure is also depicted by use
of a solid line and frontal symbols. Surface positions of the updraft
are finely stippled: forward flank downdraft (FFD) and rear flank
downdraft (RFD) are coarsely stippled; along with associated
streamlines (relative to the storm). Likely tornado locations are
shown by encircled T's. The major cyclonic tornado is most
probable at the wave apex, while a minor cyclonic tornado may
occur at the bulge in the cold front (southern T), which also marks
the favored location for new mesocyclone. Anticyclonic tornadoes,
if any, are found even farther south along the cold front. (Repro-
duced with permission from Davies-Jones RP (1985) Tornado
Dynamics. Kessler E (1986) Thunderstorms: A Social, Scientific,
and Technological Documentaty. Vol. 2: Thunderstorm Morphol-
ogy and Dynamics, 2nd edn., revised and enlarged. Norman,
OK and London: University of Oklahoma Press. Adapted with
permission from Lemon LR and Doswell 111 CA (1979) Severe
Figure 4 (Above) Vertical cross-section as might be observed on thunderstorm evolution and mesocyclone structure as
a radar scope during the mature phase of an intense supercell related to tornadogenesis. Monthly Weather Reviews 107:
storm. Low-level inflow, updraft, and outflow aloft (solid lines) are 1184-1 197.)
superimposed on the radar reflectivity (dashed lines). (Below)
Composite tilt sequence. Solid lines are the low-level reflectivity
contours, dashed lines outline the echo greater than 20dBZ
derived from the middle-level elevation scan, and the black dot is much stronger vertical accelerations and resul-
the location of the maximum top from the high-level scan. (Adapted
with permission from Lemon LR (1980) Severe Thunderstorm
ting vertical motions (both upward and downward)
Radar Identification Techniques and Warning Criteria. NOAA that are produced. Thus, the most fundamental
Technical Memorandum, NWS NSSFC-3, Kansas City, MO (NTIS equation relevant to convective storm dynamics is
PB81-234809). the non-hydrostatic vertical momentum equation:
multicells, or supercells. In the following, we review PI

the basic physical processes that contribute to the wide
spectrum of observed convective storm properties. where n is a nondimensional form of the pressure,
referred to as the Exner function,
Buoyancy Effects
Convective storms differ dynamically from larger-

scale atmospheric phenomena primarily due to the
Rd is gas constant for dry air, 0 represents the potential

temperature, qv represents the water vapor mixing
ratio, and qc, qr, and qi represent cloud water,
rainwater, and ice mixing ratios, respectively.
For an undisturbed environment (e.g., characterized
by no variation of wind with height), the pressure
contributions to vertical accelerations are usually very
small relative to the buoyancy contributions, and are
neglected. Under this assumption, an estimate of
potential updraft (and downdraft) strength in a
convective storm is often made by integrating the
potential temperature contributions from buoyancy
along a representative parcel path. For an updraft
parcel, this quantity is referred to as the convective
available potential energy (CAPE):
where Q'(z)defines the potential temperature of a

representative adiabatically ascending surface parcel,
e ( z ) defines the environmental potential temperature
profile, and the integral is taken over the vertical
interval where the lifted parcel is warmer than its
environment (usually from the level of free convection,
180" LFC, to the equilibrium level, EL). This calculation is
30
equivalent to evaluating the positive area represented
on a skew-T diagram. Maximum potential tempera-
I LM
11
t2 ture excesses in convective updrafts can be greater
than 10K, with magnitudes of CAPE larger than
6000 m2 s - 2 , but generally potential temperature
km
excesses range between 3 and 6 K , with CAPES of
1500 and 2500 m2 s - 2 for moderately unstable con-
vective days. A similar quantity can be calculated for
downdraft parcels, and is referred to as DCAPE
RM , I I I 270"
0 10 20 30 40 50 mls (downdraft CAPE).
By equating this CAPE (DCAPE) to vertical kinetic
Figure 6 (Top) WSR-57radar history of a splitting storm energy, one can then estimate the maximum updraft
observed in south central Oklahoma. The solid contours indicate
(downdraft) that would be expected from a given
return greater than lOdBZ, and the stippled regions indicate a
return greater than 40dBZ. Times adjacent to each outline are environment:
CST. (Bottom) A hodograph representative of the storm's environ-
ment. RM and LM indicate the observed motion of the right-moving W,, = (2 x 151
and left-moving cells. (Adapted with permission from Burgess DW
(1974) Studyofa Right-Moving ThunderstormUtilizing New Single Using this relationship, a CAPE of 2500 m2 s V 2would
Doppler Radar Evidence. Master's thesis, Department of Meteo-
translate to a maximum possible updraft strength of
rology, University of Oklahoma.
70 m s - '. However, water loading, perturbed vertical
pressure gradients, and mixing effects reduce these
and B is the buoyancy, including the effects of water estimates by roughly 50%. Vertical motions of
vapor and precipitation loading: 60 m s - or greater have been observed in the strong-
est storm updrafts, but maximum downdrafts rarely
exceed 3 0 m s - l .
Vertical Wind Shear Effects
In these equations, C, represents the specific heat at While the thermodynamic structure influences
constant pressure, po is a reference surface pressure, strongly the vertical accelerations in a convective
storm, vertical wind shear has a strong influence on 180"

what form convection might take. In particular, short-
lived ordinary cells tend to be the preferred mode of
organization in weak wind shear regimes, while
multicells and supercells become the respective pre- ~ I 270" Ordinarycells
ferred mode of organization for increasing magnitude
of vertical wind shear. -
The characteristics of the wind profile in this regard 10 20 m s-l
are best represented in the from of a hodograph, where Storm
motion
the wind vectors at each height are plotted from
the origin, and then the tips of the vectors are
connected to produce a hodograph trace (Figure 7 ) .
Vertical wind shear vectors are everywhere tangent
to this hodograph trace, with the length of
the hodograph curve over a given depth being a direct 180" 12 km
I
measure of the magnitude of the wind shear over that
Multicells
depth.
The relationship between wind shear and basic
storm type is demonstrated in Figure 8, which depicts 1
composite hodographs from a study of hailstorms in 270"
I
/ ' I ' I ' I ' I
Alberta, showing an increasing length of the hodo-
20 30 40ms-'
graph (especially over the lowest 6 km agl) as the type 'Storm 10
of convection progresses from short-lived storms to
supercells. Generally, multicell storms become more
prevalent when the length of the hodograph over the 180"
lowest 4-6 km agl is greater than 10-15 m s - ', with I Supercells
supercells becoming more prevalent when the length
of the hodograph is greater than 20-25 m s - over the
lowest 4-6 km agl. Also included on the hodograph
plots are observed cell motions. For ordinary cells and
multiple-cell systems, cell motion tends to be with the
mean wind over the lower 6-8km of the profile, storm 40 m s-l
appearing on or near the hodograph trace. For the 7(C)
-+--0° motion
supercell, however, cell motion is well off the hodo-
Figure 8 Typical wind hodographs for (A) ordinary cell,

(B) multicell, and (C) supercell storms observed during the Alberta
20 Hail Studies project. (Adapted with permission from Chisholm AJ
and Renick JH (1972) The Kinematics of multicell and supercell
Alben'a hailstorms. Alberta Hail Studies, Research Council of
Alberta Hail Studies, Rep. 72-2, Edmonton, Canada, pp. 24-31 .)
graph, well to the right of a calculated mean wind from

the profile. Similar off-hodograph propagation is
20 40 evident for the 19 April splitting storm (Figure 6),
U and reflects the unique dynamical character of super-
Figure 7 Composite hodograph (ms-') for tornadic supercell cell storms, as will be described below.
storms. Light arrows represent the wind vectors at each level, and Two physical mechanisms help explain the organ-
heavy arrows indicate the direction of the shear vector at each level izational capacity of vertical wind shear. The first is
(labeled in mbar). The estimated mean storm motion is denoted by related to the ability of a cold pool to trigger new
an encircled X. (Reproduced with permission from Klemp JB convective cells. The second is related to the interac-
(1987) Dynamics of tornadic thunderstorms. Annual Reviews of
Fluid Mechanics 19: 369-402. Adapted with permission from tion of an updraft with the environmental vertical
Maddox RA (1976) An evaluation of tornado proximity wind and wind shear to produce an enhanced, quasi-steady
stability data. Monthly Weather Reviews 104: 133-1 42.) storm structure.
CONVECTIVE STORMS I Overview 555
(A) (B)
Figure 9 Schematics of cold pool-shear interactions. (A) A convective cell in a zero-shear environment produces a cold pool that
propagatesawayfrom the cell. Without the presence of low-levelshear, the circulation of the spreadingcold pool inhibits deep lifting, and is
less apt to trigger a new convective cell. (B) The presence of low-level shear counteracts the circulation of the cold pool on the downshear
side, promoting deeper lifting and an enhanced potentialto trigger new convective cells. (Adaptedwith permissionfrom Rotunno R, Klernp
JB, and Weisman ML (1 988) A theory for strong, long-lived squall lines. Journal of the Atmospheric Sciences 45: 463-485.)
(i)Cold Pool-Shear Interactions: Cold pools are one due primarily to the enhanced ability of the cold pool
of the most prominent features of convective storms, to trigger new cells.
and have a critical role in determining whether a storm (ii) Updraft-Shear Interactions: Vertical wind shear
system can be maintained over a long period of time. can further contribute to convective storm strength,
This is due primarily to the ability of the cold pool to organization, and sustenance through the interaction
lift the surrounding air mass, thus serving potentially of the sheared flow with the convective updrafts. These
as a trigger for new convective cells. For a zero wind effects can be both positive and negative. The negative
shear environment, the lifting along the leading edge of effects are most clearly evident during the early stages
a cold pool is generally restricted to the depth of the of a storm's life, as clouds are observed to lean over in
cold pool's nose, as the circulation generated by the the direction of the mean tropospheric shear vector.
cold pool rapidly drags the lifted air rearwards. If the This process takes vertical kinetic energy out of
LFC is significantly higher than the nose of the cold the accelerating buoyant plume, converting it to
pool, then it is unlikely that new cells can be triggered horizontal kinetic energy. If the shear is too strong
as the cold pool propagates away from a given cell relative to the buoyancy, a cloud can be literally torn
(e.g., Figure 9A). apart.
This picture changes significantly with the addition The positive attributes of the shear are most clearly
of environmental low-level vertical wind shear. Asso- associated with the development of rotation about a
ciated with this vertical wind shear is an opposing vertical axis within the storm. This rotation originates
circulation that can balance the circulation of the cold through the tilting of horizontal vorticity inherent in
pool somewhat on its downshear side, producing a the vertically sheared flow, as can be shown from the
more vertically oriented jet of air and deeper lifting at vertical vorticity equation:
its leading edge (e.g., Figure 9B).An 'optimal' state for
lifting along the cold pool can be envisioned when the
circulation generated by the cold pool matches the
opposite circulation associated with the environmen-
tal vertical wind shear. The deptldlayer of vertical where W H and [ represent the horizontal and vertical
wind shear most important to this process is equiva- components of vorticity, respectively. This process
lent to the depth of the cold pool itself, but deeper is visualized in Figure 10A, for an isolated up-
shear layers will also contribute somewhat to this draft developing in a unidirectionally sheared flow.
process. Ambient shear can further enhance the The updraft initially deforms the ambient vortex
strength of new cells by virtue of the fact that such lines upwards, leading to the development of a
cells will move downshear along with the cold pool, vortex couplet at midlevels, centered on the up-
increasing both the relative inflow into these cells and draft. Cyclonic vorticity is generated on the right flank
the time over which the cells maintain their low-level of the updraft (relative to the direction of the shear
convergence and feed on the warm air ahead of the vector), with anticyclonic vertical vorticity on the left
gust front. All in all, the strength and longevity of flank.
multiple-cell convective systems is enhanced for The main impact of this rotation on storm structure
increasing magnitudes of ambient vertical wind shear, occurs through the relationship between the velocity
updraft to propagate to both flanks of the original

cell.
Once the updrafts propagate to the flanks (Figure
lOB), they become more colocated with the midlevel
rotation centers, which are then further enhanced by
vortex stretching. The vortex tilting process continues
to generate new rotation on the flanks of the storm,
and the updrafts will continue to propagate towards
these midlevel rotational centers. Thus, the original
cell splits into mirror image cyclonic and anticyclonic
storms that propagate to the right and left of the shear
vector, respectively. This is the most basic process
by which supercell storms may be generated and
sustained.
The relationship between the velocity and pressure
fields in a convective storm can be derived by taking
the divergence of the momentum equations, assuming
incompressibility, which leads to the follow-
ing Poisson equation for the nondimensional
pressure, n:
V . (Cp$vV~)= - V . ( P v . VV) - + aB
az [7]
This equation can be solved for the contributions

to the perturbation pressure field from the velocity
and buoyancy terms on the right-hand side of
Figure 10 Schematic depicting how a typical vortex tube
eqn [ 71 individually, allowing the vertical momentum
contained within (westerly) environmental shear is deformed as it equation [l]to be rewritten to reflect the contribu-
interacts with a convective cell (viewed from the south-east). tions from velocity-related pressure perturbations
Cylindrical arrows show the direction of cloud-relative airflow, and and buoyancy-related processes individually
heavy solid lines represent vortex lines with the sense of rotation as well:
indicated by circular arrows. Shaded arrows represent the forcing
influencesthat promote new updraft and downdraft growth. Vertical
dashed lines denote regions of precipitation. (A) Initial stage: vortex
tube loops into the vertical as it is swept into the updraft. (6)
Splitting
stage: downdraft forming between the splitting updraft cells tilts
vortex tubes downward, producing two vortex pairs. A new updraft The first term on the right-hand side of eqn [SI is
is forced on the flanks of the splitting cell in response to upward- referred to as the dynamic contribution to vertical
directed vertical pressure gradient forcing associated with the acceleration, and includes all the effects of shear on an
midlevel rotation. The barbed line at the surface marks the
boundary of the cold air spreading out beneath the storm.
updraft, such as the initial tendencies for a cell to lean
(Reproduced with permission from Klemp JB (1987) Dynamics of in the direction of the shear as well as the positive
tornadic thunderstorms. Annual Reviews of Fluid Mechanics 19: influences due to the development of rotation. The
369-402.) second term on the right-hand side of eqn [8] includes
the usual effects of buoyancy as well as the compen-
sating effects due to the buoyant contributions to
field and the pressure field. In particular, the localized the pressure field. For ordinary convective cells,
development of rotation in a fluid is associated which develop in weakly sheared environments, the
with lowered pressures (e.g., consider what happens buoyancy terms generally contribute 60-70% of the
when you stir a cup of coffee). For convective scales maximum updraft strength in a storm. However, the
of motion, this lowering of pressure occurs whether supercell storms, which develop in strongly sheared
the rotation is cyclonic or anticyclonic. If the resulting environments, 60-70 YO of the maximum updraft
rotation at midlevels in a storm is sufficiently strong strength can come from the dynamic contributions,
(e.g., if the storm is developing in a sufficiently with most of this contribution coming in the lowest
sheared environment), the induced pressure deficits several kilometers of the storm. This explains why
at midlevels will produce a significant upward-direct- supercell storms can be unusually strong, and can
ed vertical pressure gradient force that will force the persist, sometimes even in the presence of significant
CONVECTIVE STORMS / Overview 557
~~~~~~~
low-level capping inversions, as are generally observed Summary

at night.
For convective storms, cold pool generated lifting and
The updraft-shear interaction processes described
dynamic pressure forcing work together to produce
above are symmetric about the ambient shear vector
the observed storm characteristics. The relative
for unidirectionally sheared environments (e.g., shear
importance of each mechanism is dependent on the
environments characterized by a straight line on a
characteristics of the thermodynamic profile as well as
hodograph). In such cases, mirror image supercells
the vertical wind shear profile of the environment in
propagating off the hodograph to the right and left of
which the storm grows. A convective system may be
the shear vector can be produced, as demonstrated in
composed of both ordinary cells and supercells simul-
idealized cloud model simulations presented in Figure
taneously, while maintaining a general multicell char-
11A. This symmetry is modified, however, by the
acter. Storm types also have a tendency to change
addition of directional shear to the environment. If the
during the lifetime of an event. For instance, an
environmental vertical wind shear vector turns clock-
isolated supercell will often evolve into a more
wise with height over the lowest few kilometers agl
multicellular line of ordinary cells over time as the
(referred to as a clockwise curved hodograph), as
storm generated cold pool and associated lifting
presented in Figure l l B , the pressure forcing is
becomes stronger and begins to dominate over the
enhanced on the cyclonic flank of the original cell,
dynamic lifting effects associated with the rotating
and a dominant cyclonically rotating supercell results
updraft. In such cases, a supercell is said to have
from the original splitting process. However, if the
‘gusted out’ or ‘lined out’. Convective storms also
environmental vertical wind shear turns counterclock-
change character as they move into a different meso-
wise with height (not shown), the anticyclonic mem-
scale environment, or when they interact with each
ber of the original split would have been favored
other, as within a squall line.
instead. Climatologically, environmental hodographs
While convective updraft characteristics can gener-
in the vicinity of supercell storms exhibit cyclonic
ally be anticipated quite well from environmental
turning of the shear vector at low levels (e.g., consider
thermodynamic and shear profiles, potential down-
the hodographs in Figures 7 and 8C), and thus
draft and resulting cold pool characteristics are much
cyclonically rotating supercells tend to be more
more difficult to gauge from environmental condi-
common and dominant than anticyclonically rotating
tions. The storm-generated downdraft and cold pool is
supercells.
certainly sensitive to the amount of thermodynamic
Figure 12 presents the overall flow structure
instability and the distribution of moisture in the
for a mature, cyclonically rotating supercell storm.
environment, but it is also sensitive to the character-
An anticyclonically rotating supercell is the mirror
istics of the precipitation that is produced within the
image of this. The flow vectors depict the main
storm. For instance, a convective cloud that predom-
interwoven airstreams, with the low-level flow con-
inantly produces a few large raindrops or hailstones
verging from both ahead and behind of the surface gust
will tend to have a weaker downdraft and cold
front and rising into a deep, rotating updraft, and
pool than a cloud that produces a large quantity of
the midlevel flow passing in front of and then
smaller drops, due to decreased evaporation rates.
descending behind the updraft. The updraft reaches
Along these lines, supercell storms have been subclas-
the top of the storm, where it then diverges within
sified into high-precipitation (HP), classic, and low-
the anvil, primarily in the downshear direction. While
precipitation (LP) varieties, based on intensity and
the midlevel rotation in the storm is generated via the
distribution of the precipitation and the resulting
tilting of horizontal vorticity associated with
strength of the system-generated cold pool. Many of
the warm, ambient environment (e.g., Figure lOA),
these factors are discussed in companion chapters
the air that feeds the low-level rotation originates
within this volume.
largely from the cold side of the surface cold
pool boundary. Horizontal vorticity is generated
in response to the buoyancy gradients across Climatology of Convective Storm
this boundary, as depicted by the low-level vortex
lines turning towards the storm on the cold side of Types
the forward flank gust front, and this horizontal Ordinary cell and multicell storm systems are com-
vorticity feeds into the low-level updraft in a stream- monly observed from the tropics through midlati-
wise sense, leading to the low-level updraft rotation. tudes, whenever thermodynamic instability exists and
It is this low-level, rotating updraft that leads to the there is a sufficient triggering mechanism for the
development of significant tornadoes within supercell convection. Supercell storms, however, tend to be
storms. more limited to midlatitude, continental regions,
Figure 11 Plan views of numerically simulated convective storms at 40,80, and 120 min for two environmental wind profiles (displayed
at the upper left) having wind shear between the surface and 7.5 km agl. The storm system in the lower portion of the figure evolves in
response tothewind profile forwhich theverticalwind shearvectorturnsclockwisewithheight between the groundand 2.5 km (heavy solid
line in the hodograph), while the upper system develops when the shear is unidirectional (same wind profile except follow the heavy
dashed line below 2.5 km). The plan view depicts the low-level (1.8 km) rainwater field (similar to radar reflectivity) contoured at 2 g kg-'
intervals, the midlevel (4.6 km) updraft (shaded regions), and the location of the surface cold pool boundary (barbed lines). The maximum
updraft velocity is labeled (in m s- ') within each updraft at each time. The dashed lines track the path of each updraft center. Arrows on the
hodograph indicate the supercell propagation velocities for the unidirectional (dashed) and turning (solid) wind shear profiles.
(Reproduced with permission from Klemp JB (1987) Dynamics of tornadic thunderstorms. Annual Reviews of Fluid Mechanics 19:
369-402.)
where sufficient vertical wind shear can exist in source of low-level moisture to enhance thermody-
association with thermodynamic instability. Supercell namic instability, and the frequent passage of synop-
storms are especially prevalent in the spring and early tic-scale waves offers a source for the vertical wind
summer in the plains and mid-western regions of the shear. The frequency of supercell storms and the
United States, where the Gulf of Mexico supplies a associated tornadoes in this part of the United States
Next Page
continental convection, due to less thermodynamic

instability over oceanic regions. Convective storms
can occur at any time of the day or night, when
thermodynamic instability and a trigger is available,
but severe convection tends to maximize in the late
afternoon and evening hours, in response to the
enhanced thermodynamic instability associated with
diurnal heating.
See also
Bow Echos and Derecho. Convective Storms: Con-
Figure 12 Three-dimensional schematic view of a mature vective Initiation. Density Currents. Gust Fronts. Hail
supercell thunderstorm at a stage when low-level rotation is and Hailstorms. Mesoscale Meteorology: Cloud and
intensifying. The storm, viewed from the south-east, is evolving in Precipitation Bands. Tornados. Waterspouts. Weather
westerly environmental wind shear. The cylindrical arrows depict Prediction: Severe Weather Forecasting.
the flow in and around the storm. The thick lines show the low-level
vortex lines, with the sense of rotation indicated by the circular-
ribbon arrows. The heavy barbed line marks the boundary of the
cold air beneath the storm. (Reproduced with permission from Further Reading
Klemp JB (1987) Dynamics of tornadic thunderstorms. Annual
Reviews of Fluid Mechanics 19: 369-402.) Browning KA (1977) The structure and mechanism
of hailstorms. Hail: A Review of Hail Science and
Hail Suppression, Meteorological, Monographs, vol. 16,
pp. 1-43. Boston: American Meteorological Society.
has led to this region being referred to as 'Tornado Byers HR and Braham RR Jr (1949) The Thunderstorm.
Alley.' Supt of Documents, Washington, DC: US Government
Supercell storms can also be embedded within the Printing Office.
rain bands of landfalling tropical storms and hurri- Church C, Burgess D, Doswell C and Davies-Jones R (1993)
canes. These supercells tend to be shallower than more The tornado: its structure, dynamics, prediction, and
classic supercells, as minimal instability (CAPE less hazards. Geophysical Monographs 79: 637.
than 1000 Jkg-') is available within the associated Doswell I11 CA (1985) The Operational Meteorology
of Convective Weather. Volume II: Storm Scale Analysis.
tropical air mass. However, very strong low-level
NOAA Technical Memorandum ERL ESG-15.
vertical wind shears are generated as the rain bands Kessler E (1986) Thunderstorms: A Social, Scientific, and
come ashore, and very intense (of order 1Oms-') TechnologicalDocumentary. Vol. 2: Thunderstorm Mor-
updrafts can be generated in the lowest 1-2 km agl in phology and Dynamics, 2nd edn., revised and enlarged.
such cells, due to the dynamic vertical pressure Norman, OK and London: University of Oklahoma
gradients associated with the rotating updrafts. Such Press.
shallow supercells are hypothesized to be the source of Klemp JB (1987) Dynamics of tornadic thunderstorms.
tornado outbreaks within landfalling tropical storms Annual Reviews of Fluid Mechanics 19: 369-402.
and hurricanes. Lemon LR (1980) Severe Thunderstorm Radar Identifica-
In the midlatitudes, convection often occurs in the tion Techniques and Warning Criteria. NOAA Technical
warm sectors of synoptic-scale waves, in association Memorandum, N W S NSSFC-3, Kansas City, MO (NTIS
with cold fronts and warm fronts. In the tropics, PB81-234809).
Rotunno R, Klemp JB and Weisman ML (1988) A theory
convective activity is commonly located along the
for strong, long-lived squall lines. Journal of the
intertropical convergence zone (ITCZ). In both the Atmospheric Sciences 45: 463-485.
tropics and midlatitudes, convection also commonly Weisman ML and Klemp JB (1986) Characteristics of
occurs along sea breeze fronts or in association with isolated convective storms. Mesoscale Meteorology and
topographic features such as mountain ridges. Gener- Forecasting, pp. 331-358. Boston: American Meteoro-
ally, oceanic convection tends to be weaker than logical Society.
570 CORIOLIS FORCE
Kessler E (1986) Thunderstorms: A Social, Scientific, and Schaefer JT (1986) The dryline. In: Ray PS (ed.) Mesoscale
Technological Documentary, vol. 2, Thunderstorm Meteorology and Forecasting, pp. 549-572. Boston:
Morphology and Dynamics. Norman, OK: University American Meteorological Society.
of Oklahoma Press. Simpson JE (1997) Gravity Currents in the Environment
Pielke RA and Segal M (1986)Mesoscale circulations forced and the Laboratory. Cambridge: Cambridge University
by differential terrain heating. In: Ray PS (ed.)Mesoscale Press.
Meteorology and Forecasting, pp. 516-548. Boston: Stull RB (1988) An Introduction to Boundary Layer
American Meteorological Society. Meteorology. Dordrecht: Kluwer Academic.
D W Moore, Pacific Marine Environmental Laboratory, the Earth is rotating, it bulges at the Equator and is
Seattle, WA, USA flattened at the poles. The reference geopotential
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. includes both the gravitational and centrifugal poten-
tial, so the reference surface is not quite spherical. The
Earth’s equatorial radius exceeds the polar radius by
Introduction about 21 kilometers. The shape of the geopotential
The Coriolis force arises when Newton’s equations and the Coriolis force both depend on the Earth’s
of motion are written in a rotating coordinate rotation rate.
system. It is named for Gaspard Gustave de
Coriolis (1792-1843). His studies of dynamical
problems associated with rotating machinery were
Derivation of the Coriolis Force
presented to the Acadtmie des Sciences in Paris in Consider two coordinate systems, one of which is
1831. inertial, fixed with respect to the fix_ed stars, and the
All of the terms in the equations of motion related to other rotating with angular velocity SZ We use a prime
the Coriolis force were actually included in the tidal to denote a quantity as viewed in the rotating frame
equation of P. S. Laplace, published in 1775and 1776, and use unprimed quantities to denote the inertial
and repeated in Book IV of his Treatise on Celestial frame. Let 4‘ be a fixed vector in the inertial frame, so
Mechanics, published in 1799. that
Newton’s equation of motion for a particle is
dq‘
-
=o
dij +
dt
m-=F
dt
Viewed in the rotatizg frame, $ appears to rotate with
where m is the mass of the particle, ij = d7/dt is the angular velocity -zZ in the direction opposite to the
velocity of the particle, 7 is the position vector, and rotation of the coordinate systeq So its apparent
d/dt is the time derivative following the motion of the velocit_yin the rotating system is -zZ x $, and we must
particle. The quantity a‘ = dri/dt is the acceleration. add + a x to this apparent velocity to bring it to zero.
This equation holds in an inertial reference frame, We are thus led to the rule
fixed with respect to the ‘fixed’stars.
But the Earth on which we live rotates at a nearly
constant rate about an axis whose direction is more or
d,
-q=-q
dt
d,,
dt’
+Rxq“
- [31
less fixed in inerti:l space. y e denote the Earth’s
rotation vector by a , so C2 = is the rotation rate. where the terms on the left are viewed in the inertial
The direction of S2 points from the South Pole to the frame and those on the right in the rotating frame.
North. The Earth rotates once each sidereal day, which Furthermore, this same result, eqn [3], holds even if the
is about 8.62 x lo4 solar seconds. So the Earth’s left-hand side is not zero.
rotation rate is R = 7.29 x lop5s-l. Consider a particle with position vector 7 in the
The motions of the atmosphere and the oceans are inertial frame and 7’ in the rotating frame. The velocity
generally described in a coordinate system which is given by
rotates with the Earth. For example, longitude, lati-
tude, and height above a reference geopotential
surface form a convenient rotating system. Because
CORIOLIS FORCE 571
and the acceleration is The Coriolis force & = -26 x u’ is given by

a =-
dG = ($+ a x ) (5’ +a x i’)
- eOw + sin eou)l-
ZC = 2 ~ [ ( COS sin eou3
dt
+ cos eoui] 191
This force is perpendicular to u’ (and to 6) and

The first term on the right is the acceleration as seen in therefore does no work on the moving particle or fluid.
the rotating system. The second term is the Coriolis In the equation for the time rate of change of kinetic
acceleration. The third term is the centripetal accele- energy the Coriolis terms vanish identically. This is
ration arising from the rotating coordinate frame. important to note because in most problems the part of
the Coriolis force proportional to the locally horizon-
tal component of R can be shown by a scaling
Modification of the Gravitational argument are to be negligible. These are the terms
Potential proportional to 2R cos 80. When they neglected the
Let -2R cos 80w; term and the +2R cos 8ouk term must
both be omitted to avoid introducing a spurious source
of kinetic energy. Likewise if one of these terms is
important then the other should be included as well. If
these terms are omitted then the only quantity involv-
Then R is the distance from the position J to the axis of
ing the rotation is fo = 2R sin 80. This is called the
rotation, and it is easy to check that
Coriolis parameter. More generally,
[71 f = 2R sin 8 1101

Since this is the gradient of a scalar, it can be combined is the Coriolis parameter at latitude 8.
with the gravitation potential to define an effective
gravitational potential. This then accounts for the
nonspherical shape of the rotating Earth and the The f-Plane and the Beta Plane
reduction of the strength of the apparent gravitational If the problem under consideration involves a physical
force due to the centripetal acceleration. domain that includes the whole globe or a large
portion of it, then spherical coordinates are a natural
choice and the full dependence of the Coriolis force on
Motions on a Rotating Earth
latitude is retained. If the latitudinal extent of the
We now consider motions as observed on a rotating physical domain is small enough that f = 2R sin 8 is
Earth, and drop the primes used earlier to denote nearly constant, then f is approximated by the
the rotating coordinates. Consider a local Cartesian constant fo and we speak of studying the problem
coordinate system fixed with respect to the rotating ‘on an f-plane’.
Earth. Let x be positive toward the east, y positive A more common situation which arises in both
toward the north, and z positive upward, opposite meteorology and oceanography is one in which the
to the direction of the apparent gravitational force. latitudinal extent of the domain is large enough so that
Let 1, j , and k be unit vectors in the directions of the spatial variation of f = 2Q sin 8 needs to be
x , y , and z, and let u, 0 , and w be the components considered, but the full spherical geometry does not.
of the velocity in those directions. If we are considering Then the latitude is written as 8 N 8 0 + y/R, where R
particle motion then u , u, and w are the components is the radius of the Earth, and sin 8 is approximated by
of the particle velocity. If we are considering the first two terms in its Taylor expansion. We
fluid motion then they are the components of write
the velocity of the fluid element at the position
252
x , y , and z. Let the origin of this system be at latitude f =2R sin 8 N 2R sin 80 +-R
cos Boy
80, which is the angle between the Earth’s equatorial
plane, and the local vertical direction. Then the Earth’s =fo + PY [I11
rotation vector fi written in the local coordinate
system is The approximation f = fo + Py is called the beta plane
approximation. The parameter P = (2R/R) cos 8 0
6 = R cos eo?+ R sin e o i PI was first introduced by Carl Gustav Rossby in 1939.
572 CORIOLIS FORCE
Examples: The Foucault Pendulum rotation of the plane of the orbit; the nonzero initial
value of y gives the pendulum just the correct initial
We now turn to some simple examples to illustrate
velocity to produce this solution.
the effect of the Coriolis force. The examples are
In the Foucault pendulum problem the Coriolis
the Foucault pendulum, inertial oscillations, Ekman
force is a relatively small perturbation on the orbit
layers, and geostrophic balance. For the Foucault
of the pendulum as it goes back and forth once,
pendulum, let a particle of mass m be suspended by
but over the course of a day it has a substantial
a string of length e. Take the origin at the equilibrium
cumulative effect. In most atmospheric and oceanic
position of the mass, directly below the point of
problems we are interested in time scales of a day
suspension. If the horizontal displacements are small
or longer. The Coriolis force then becomes one
compared with e, then the x and y components of the
of the dominant terms in the equations of motion.
tension in the string are approximately -mg(x/k)i
In the remaining illustrations the Coriolis term
and -mg(y/e)?. In the absence of rotation the equa-
is balanced by various other possible important
tions of motion would be mx = -m(g/k)x and
terms.
my = -m(g/k)y, where the dots denote a time deri-
vative. If we add the Coriolis force term and divide by
m we obtain Geostrophic Balance
X-fy+W2X=O [I21 If the motions are varying slowly enough with time to
be regarded as nearly steady, and turbulent dissipation
and is small, then the approximate balance in the horizon-
tal momentum equation is
y + f k + o2y =0 1131
where co2 = g / l is the pendulum frequency in the
nonrotating system. For positive f we see that the
effect of the Coriolis force is to accelerate the particle and
to the right of its motion. A solution of eqns [12]
and [13] with initial conditions x = xg, y = 0 , i = 0,
and y = 0 at t = 0 is given by
where p is pressure, p density, and u, u , and f the
x + iy = xg exp{ -ig} [cos w’t +-
201
if
sin o’t
1 velocity components and the Coriolis parameter.
Equations [16] and [17] show that pressure gradient
1141 on the left balances the Coriolis force on the right. This
is called geostrophic balance and is illustrated else-
where we have used complex notation and
where in this encyclopedia (see Dynamic Meteoro-
w’2 = w2 + f2/4.
logy: Overview, Figure 1).So if the pressure field is
The term in the square bracket describes an elliptical
known then the horizontal velocities can be estimated
orbit, almost rectilinear iff’ << o’, with the pendulum
from eqns [ 161 and [ 171.
oscillating at frequency 0’.The exponential term
indicates that this elliptical orbit slowly rotates clock-
wise in the Northern Hemisphere, at frequency f / 2 .
This rotation is due to the deflection of the orbit by the Inertial Oscillations
Coriolis force. The frequency f / 2 = R sin 8. At the Various departures from geostrophic balance are
North Pole the period of the exponential terms is 24 h, possible, and for many of these the Coriolis force still
and can be thought of as the Earth rotating under the plays a dominant role. Inertial oscillations provide one
swinging pendulum. such example. In the simplest case, we ignore hori-
Note that the solution given by eqn [14]never passes zontal pressure gradients and balance horizontal
through the origin. A slight change in initial condi- accelerations by the Coriolis force. The governing
tions, with x = X O ,y = 0, k = 0, and i, = -f /2 xg at equations are
t = 0, produces the solution
du
- = fu
dt
and
The [cos w’t] represents a true rectilinear orbit through du
- = -f.
dt 1191
the origin and the exponential factor describes the
CORIOLIS FORCE 573
One such observation lasting nearly a week
t
24AUG. Oh 0 1 2 3 4 5km
was made in the Baltic Sea by Gustafson and
Kullenberg. It is described in some detail by Sverdrup,
Johnson, and Fleming in their book The Oceans
(see Figure 1).
Ekman Layers
Another important example is the Ekman layer,
b where the Coriolis force is balanced by vertical

mixing. The original work, published by V. Walfrid
Ekman in 1905, is called On the Influence of the
Earth's Rotation on Ocean Currents. Ekman consid-
ered the problem of determining the motion of
the upper layer of the ocean due to a wind stress
21 AUG. acting on the sea surface. Let z measure distance
positive upward from the sea surface, so the domain of
interest is z < 0. In this model the only horizontal forces
acting on the fluid are the Coriolis force and the
divergence of the stress tensor describing the turbulent
mixing. The stress is modeled in terms of an eddy
viscosity as
du
T ; = p- 1231
dz
and
du
7; = p-
dz
where 7: means the stress in the x-direction acting
Figure 1 Rotating currents of period one-half pendulum day
observed in the Baltic Sea and represented by a progressive vector
on a surface whose normal is in the z-direction. The
diagram for the period, 17-24 August 1933, and by a central vector quantity p is the coefficient of eddy viscosity, and v =
diagram between 06.00 and 20.00 on 21 August (according to p / p is taken as a constant. The equations of motion in
Gustafson and Kullenberg). (Adapted with permission from the x- and y-directions are
Sverdrup HU, Johnson MW, and Fleming RH (1942) The Oceans:
Their Physics, Chemistry and General Biology. New York:
Prentice-Hall.) 1251
and
If the initial condition is u = U O ,u = 0, then the fu =

; (v;)
solution is
If a stress of magnitude 7: acts in the x-direction on the
u = 210 cos ( f t ) , [201 sea surface, the boundary conditions are
u = -uo sin ( f t ) , P
I au
v-=-
7$
onz = 0
az P
or in complex form
and
u + iu = uo exp { -ift} 1221 au
v-=O onz=O ~ 7 1
Equation [22] clearly shows that the velocity az
vector rotates at frequency f , clockwise in the Furthermore, the stresses are assumed to vanish as
Northern Hemisphere and counterclockwise in z4 -m. If we write
the Southern. Such oscillations have been observed
in the sea. q=u+iu PI
574 COUPLED OCEAN-ATMOSPHERE MODELS
then eqns [25]and [26]can be combined in the form Hemisphere. In the Southern Hemisphere the trans-
ports are 90" to the left of the wind stress, and surface
v9zz = if9 ~ 9 1 currents 45"to the left.
and for f > O (Northern Hemisphere), a solution Since f = 2 R sin 8 vanishes at the Equator, there is
which vanishes as z +. -cc is an equatorial transition from Northern to Southern
Hemisphere behavior. All the examples discussed here
q=Aexp - (lAi){ gz} apply to regions away from the Equator. In the tropics
[30] other terms become important in the equations
describing the motion.
Applying the boundary condition given in eqn [27]
determines A, and the solution is
See also
q=u+iv
Boundary Layers: Neutrally Stratified Boundary Layer.
=c P & {
exp (1 + i ) & z } [31]
Dynamic Meteorology: Overview.
Equation [31]describes a flow which at the sea surface Further Reading

z = 0 is directed at 45"to the right of the surface stress.
Chandrasekhar S (1963)Ellipsoidal Figures of Equilibrium.
The direction of the flow spirals to the right as z
New Haven: Yale University Press. (Reprinted with
becomes more negative. The net vertically integrated minor changes, New York: Dover, 1987.)
transport can be determined by integrating eqns [25] Ekman VW (1905) On the influence of the Earth's rotation
and [26] and applying the boundary conditions eqn on the ocean currents. Arkiv for Mathematik, Astronomi,
[27].The result is och Fysik (Stockholm) 2: 1-53.
Rossby CG and collaborators (1939) Relation between
variations in the intensity of the zonal circulation of
the atmosphere and the displacements of the semi-
If a nonzero stress in the y-direction, zi, had been permanent centers of action. Journal o f Marine Research
specified as well, we would have 2: 38-55.
Stommel HS and Moore DW (1989) An Introduction to
U = TO1 V = - - 7: the Coriolis Fovce. New York: Columbia University
[331 Press.
f f Sverdrup HU, Johnson M W and Fleming RH (1942) The
So the net transport in the frictional boundary layer Oceans: Their Physics, Chemistry and General Biology.
is 90" to the right of the wind stress in the Northern New York: Prentice-Hall.
D Anderson, ECMWF, Reading, UK To address such an issue requires the most sophis-
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. ticated and advanced models of the Earth's environ-
ment. At present these models capture only part of
what is thought to be involved in representing the
many complex interactions in the climate system. The
Introduction models have deficiencies, both in their ability to solve
Coupled models represent the best available tools the equations correctly and in that they do not take
for addressing some of the most pressing into account all known interactions, not to mention
current scientific questions. For example, it has possible interactions that have not yet even been
been predicted that global mean sea level will thought of. Nonetheless, they represent the main tool
rise by between about 10 and 90cm in the for climate prediction, interpretation of climate var-
twenty-first century, compared with the 10-20 cm iability, and understanding of climate feedbacks and
rise observed in the twentieth century. How can interactions. In the last few years, considerable
these predictions be made, and with what con- progress has been made in the development and use
fidence? of such models.
then eqns [25]and [26]can be combined in the form Hemisphere. In the Southern Hemisphere the trans-
ports are 90" to the left of the wind stress, and surface
v9zz = if9 ~ 9 1 currents 45"to the left.
and for f > O (Northern Hemisphere), a solution Since f = 2 R sin 8 vanishes at the Equator, there is
which vanishes as z +. -cc is an equatorial transition from Northern to Southern
Hemisphere behavior. All the examples discussed here
q=Aexp - (lAi){ gz} apply to regions away from the Equator. In the tropics
[30] other terms become important in the equations
describing the motion.
Applying the boundary condition given in eqn [27]
determines A, and the solution is
See also
q=u+iv
=c P & {
exp (1 + i ) & z } [31]
Dynamic Meteorology: Overview.
Equation [31]describes a flow which at the sea surface Further Reading

z = 0 is directed at 45"to the right of the surface stress.
Chandrasekhar S (1963)Ellipsoidal Figures of Equilibrium.
The direction of the flow spirals to the right as z
New Haven: Yale University Press. (Reprinted with
becomes more negative. The net vertically integrated minor changes, New York: Dover, 1987.)
transport can be determined by integrating eqns [25] Ekman VW (1905) On the influence of the Earth's rotation
and [26] and applying the boundary conditions eqn on the ocean currents. Arkiv for Mathematik, Astronomi,
[27].The result is och Fysik (Stockholm) 2: 1-53.
Rossby CG and collaborators (1939) Relation between
variations in the intensity of the zonal circulation of
the atmosphere and the displacements of the semi-
If a nonzero stress in the y-direction, zi, had been permanent centers of action. Journal o f Marine Research
specified as well, we would have 2: 38-55.
Stommel HS and Moore DW (1989) An Introduction to
U = TO1 V = - - 7: the Coriolis Fovce. New York: Columbia University
[331 Press.
f f Sverdrup HU, Johnson M W and Fleming RH (1942) The
So the net transport in the frictional boundary layer Oceans: Their Physics, Chemistry and General Biology.
is 90" to the right of the wind stress in the Northern New York: Prentice-Hall.
D Anderson, ECMWF, Reading, UK To address such an issue requires the most sophis-
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. ticated and advanced models of the Earth's environ-
ment. At present these models capture only part of
what is thought to be involved in representing the
many complex interactions in the climate system. The
Introduction models have deficiencies, both in their ability to solve
Coupled models represent the best available tools the equations correctly and in that they do not take
for addressing some of the most pressing into account all known interactions, not to mention
current scientific questions. For example, it has possible interactions that have not yet even been
been predicted that global mean sea level will thought of. Nonetheless, they represent the main tool
rise by between about 10 and 90cm in the for climate prediction, interpretation of climate var-
twenty-first century, compared with the 10-20 cm iability, and understanding of climate feedbacks and
rise observed in the twentieth century. How can interactions. In the last few years, considerable
these predictions be made, and with what con- progress has been made in the development and use
fidence? of such models.
COUPLED OCEAN-ATMOSPHERE MODELS 575
In this article, the physics of coupled models will other components of the coupled system. As there is no
first be considered and then some examples of appli- unique way of making these parameterizations, dif-
cations will be given. Although for certain applica- ferent models will have different parameterizations.
tions gross simplifications of some components can be By comparing and combining the results of many
made such, models are highly specific to a given models, it is possible to put some measure on the
application and will not be considered here: rather we uncertainty arising from sub-grid-scale processes.
will discuss only coupled general circulation models There are over 30 different coupled models extant
(CGCMs). today. The atmosphere is quite complicated and differs
Any CGCM must contain a comprehensive model substantially from model to model. It is not possible to
of the atmosphere, together with a comprehensive describe these individually, so instead we will discuss
model of the ocean. Usually both will cover the globe, primarily one atmospheric model as a typical example.
and a model of land conditions covering soil moisture A typical ocean model will also be described. Appli-
and snow cover is generally included as part of the cations of coupled models will then be given.
atmospheric model. Frequently there is a model of the
sea ice. Interactions with chemistry, biology, and
geochemistry are under development but not yet The Atmospheric Model Formulation
generally included. Not all coupled models will have
The Explicit Equations
all these components. For example, models used to
study seasonal climate variability -why one season is The model formulation can be summarized by seven
not like its counterpart a year later - do not yet have basic physical equations, the resolution in time and
sea ice models, whereas models to study global space, and the way the numerical computations are
change - how climate will respond to changing carried out. Of the equations governing the ECMWF
levels of ‘greenhouse’ gases - should have an accurate (European Centre for Medium-Range Weather Fore-
sea ice component. casts) primitive equation atmospheric model, two are
In its simplest form, modeling the atmosphere is a diagnostic and tell us about the static relation between
case of solving the equations governing both the different parameters: namely the gas law, which gives
atmospheric flow and thermodynamics. Unfortunate- the relation between pressure, density, and tempera-
ly, the equations are horribly complex and can only be ture, and the hydrostatic equation, which shows the
solved approximately by numerical means. Concep- relationship between the density of the air and the
tually imagine the atmosphere of the Earth as split up change of pressure with height. Four prognostic
into boxes. Box-average values are defined and the equations describe the changes with time of the
equations modified to solve for these average quanti- horizontal wind components, temperature and water
ties. Typically in climate models the size of the boxes is vapor content of an air parcel. The equation of
a few hundred kilometers on a side by some ten to continuity expresses the mass conservation and makes
hundreds of meters deep, but look out the window and it possible to determine the vertical velocity and
you will see a lot can happen on scales much smaller change in the surface pressure.
than this. Representing the atmosphere by box aver- Newton’s equations of motion describes how the
ages could therefore leave many processes unaccount- momentum of an air parcel changes due to the pressure
ed for. We can take account of how these boxes interact gradient and the Coriolis force. The thermodynamic
in a mean sense explicitly, but the smaller-scale equation expresses how a change in an air parcel
processes have to be parameterized. Thus how many temperature is brought about by adiabatic cooling or
clouds in a box, what type, what height, what effect warming due to vertical displacements. The continuity
they will have on the radiation - if you are under a equation for moisture assumes that the moisture
cloud on a sunny day it is a lot cooler than if you are in content of an air parcel is constant, except for losses
full sun - all have to be represented in some form due to precipitation and condensation or gains by
though parameterization. Unfortunately, the smaller evaporation from clouds and rain or from the oceans
scale processes have an important influence on the and continents. Added to this, there are specific
box-average values and so should be represented prognostic equations for the cloud fraction, water,
accurately, albeit only statistically. These aspects will ice content, and ozone. Latent heat release, radiation
be discussed later: for now, simply note that any model from the Sun and the Earth’s surface, and frictional or
of the atmosphere must include a mechanism for turbulent processes (diffusion) are described in a
handling the explicit part (the averages) and the statistical way as parameterized processes.
smaller-scale processes - the physical parameteriza- The vertical resolution (measured in geometric
tion. A similar scale separation into explicitly resolved height) is highest in the planetary boundary layer
and sub-grid-scale processes applies to the ocean and (PBL) and lowest in the stratosphere. The atmosphere
is divided into 40 levels (Figure 1). These follow the inhomogeneously, so that it is difficult to maintain
Earth’s surface in the boundary layer but are surfaces uniform resolution over the forecast region. A semi-
of constant pressure in the upper stratosphere with a Lagrangian scheme is used to overcome this difficulty:
smooth transition between these types of levels. For its the grid points are stationary and at each time step the
horizontal resolution the ECMWF coupled model uses scheme computes a backward trajectory from every
two different numerical representations. A spectral grid point. The point reached defines where the air
method, based on a spherical harmonic expansion, parcel was at the beginning of the time step. The
truncated at total wavenumber 95, is used for the interpolated value of the variable at that point is then
representation of upper air fields, and the computation carried forward to the grid point, applying the various
of the horizontal derivatives. This spectral technique physical processes. Whereas all Eulerian schemes
was introduced to improve the accuracy in computing require small time steps to avoid numerical instability,
terms involving spatial derivatives, which is crucial for the semi-Lagrangian scheme allows longer time steps.
good phase speed estimation. With higher resolution (The limitation for stability is that the trajectories do
with both grid point and spectral models there is less not cross, i.e. a parcel cannot ‘overtake’ another one.)
significant difference in accuracy, though there are still Tests have shown that a semi-Lagrangian timestep can
some advantages in using a spectral representation. be at least 15 times longer than the Eulerian without
The momentum, temperature, and humidity equa- becoming unstable, though such an increase in time
tions have to be stepped forward in time. An innova- step may be too long to deal adequately with physical
tive feature of the ECMWF model is the use of a semi- processes within the model. For a spectral resolution
Lagrangian numerical scheme instead of an Eulerian of T95, the time step is 1h. (If the semi-Lagrangian
one to save computer time and speed up the forecast. In approach was not used, the time step might be -5 min,
a pure Lagrangian framework one seeks to follow a set showing the great advantage of the technique. Most
of marked fluid parcels. A difficulty is that shear and coupled models, however, do not yet use this tech-
stretching deformations tend to concentrate parcels nique.) A further advantage of the semi-Lagrangian
technique is that it allows a better representation of the
moisture equation, where the large gradients in
moisture often give rise to negative values in an
Eulerian formulation if special steps are not taken.
In addition to the spectral representation, there is a
grid point representation used for computing dynamic
tendencies and the diabatic physical parameterization.
This so-called Gaussian grid is regular in longitude and
almost regular in latitude. Due to the convergence of
the longitudes toward the poles, the east-west distance
between the grid points decreases poleward. To avoid
some numerical problems around the poles, but most
importantly to save computing time, a reduced
Gaussian grid is introduced by reducing the number
of grid points along the shorter latitude lines near the
poles, so as to keep the east-west separation between
points on different latitudes almost constant. The
model surface is logically divided into sea and land
points, by using a land-sea mask. A grid point is
defined as a land point if more than 50% of the actual
surface of the grid box is land. A typical resolution of
approximately 200 km.
Parameterization of Physical Processes
The physical processes associated with radiative

transfer, turbulent mixing, sub-grid-scale orographic
drag, moist convection, clouds, and surfacehoil pro-
cesses have a strong impact on the large-scale flow of
Figure 1 Vertical distribution of levels in the atmosphere. the atmosphere. However, these mechanisms are often
(Courtesy of the ECMWF.) active at scales smaller than the horizontal grid size.
Parameterization schemes are then necessary in order uses the larger-scale variables such as wind, temper-
to properly describe the impact of these sub-grid-scale ature, and specific humidity, with the assumption that
mechanisms on the large-scale flow of the atmosphere. the fluxes are proportional to the vertical gradients. At
In other words, the ensemble effect of the sub-grid- the Earth’s surface, the turbulent fluxes of momentum,
scale processes has to be formulated in terms of the heat, and moisture are computed as a function of air-
resolved grid-scale variables. Processes of special surface differences and surface characteristics. Over
interest are those happening near the ground and the land areas, snow depth, soil temperature, and wetness
interchange with the ocean; the effects of radiation, are forecast variables, calculated by a model of the soil
clouds, and precipitation; and the mixing and stirring with four layers within the top 2 m.
by small-scale processes. Figure 2 represents schemat- For the albedo, a background monthly climate field
ically the different physical processes considered in the is used over land. Two spectral bands are used in the
ECMWF atmospheric model. solar radiation calculation, corresponding to direct
The treatment of the planetary boundary layer input from the Sun and diffuse radiation. Over open
(PBL) plays a fundamental role for the whole atmos- water the albedo for these varies differently. Over land
phere-Earth system. It is through the surface exchang- the forecast albedo depends on the background albedo
es of momentum, heat, and moisture that the and the snow depth. It has a minimum of 0.07 and can
atmosphere ‘feels’ that it moves over a rough land go up to 0.70 for exposed snow but only to 0.20 for
surface or a wet smooth sea. The lowest model level is snow in forest. The thermal properties of snow-
close to IO m above the surface and several more levels covered ground depend only on the depth of the
are within a few hundred meters of the surface, where snow mass per unit area. The snow depth evolves
velocity humidity and temperature can change rapidly. through the combined effect of snowfall, evaporation,
Even with this fairly high resolution, the vertical and melting.
gradients of temperature, wind, moisture, etc., in the The soil moisture is divided into skin and soil
PBL cannot be described very accurately, let alone the reservoirs. The skin reservoir (which is mainly mois-
turbulent transports of momentum, heat, and moisture on vegetation) evolves under the action of its own
ture. For the estimation of these parameters the model evaporation and its ability to collect dew and intercept
Figure 2 Schematic diagram of the different physical processes represented in the ECMWF atmospheric model. (Courtesy of the
ECMWF.)
precipitation. The soil reservoir takes into account Sub-grid vertical fluxes of mass, heat, water vapor,
precipitation and snow melt, as well as vertical and momentum are computed at each model level with
transfer of water due to drainage and capillarity, the help of a simple mass flux model interacting with
evaporation over bare ground, and root uptake by its environment. The scheme is applied to penetrative
vegetation. The vegetation ratio is specified in each convection, shallow convection and mid-level convec-
grid point and used by the model to estimate the tion. They are mutually exclusive, so only when the
roughness and the evaporation. scheme fails to create cloud of one type, does it try the
In view of the importance of cloud-radiation next. Deep convection predominantly occurs in dis-
interaction in both long-term (climate) and short- turbed situations with a deep layer of conditional
term (weather) processes, high emphasis is placed on instability and large-scale moisture convergence. The
the treatment of the absorption and scattering by downdraught mass flux is assumed to be proportional
clouds of solar and terrestrial radiation. A significant to the updraught mass flux. Shallow convection
fraction of the overall computational time is devoted predominantly occurs in undisturbed flow, in the
to the radiation scheme, even though some simplifica- absence of large-scale convergent flow. The moisture
tions are made to reduce the cost such as calculating supply is from surface evaporation. It does not
radiation only every 3 h rather than every time step normally produce precipitation. Mid-level convection
only over a subset of the grid points. describes convective cells that originate at levels above
The radiation spectrum is divided into eight fre- the boundary layer. Stratocumulus clouds are linked to
quency bands: two in the short-wave spectrum (direct the boundary layer moisture flux produced by the
from the Sun and diffuse radiation), and six in the vertical diffusion scheme. Stratiform clouds (e.g., low-
long-wave spectrum (from the Earth and within the level stratus and medium-level nimbostratus types) are
atmosphere). The upward and downward diffused determined by the rate at which the saturation specific
radiation is computed for each of the eight spectral humidity decreases due to upward vertical motion and
bands. The parameters influencing the emission and radiative cooling. Evaporation processes in connec-
absorption are pressure, temperature, moisture, cloud tion with clouds are accounted for in several ways:
cover, and cloud water content. In addition, emission large-scale and cumulus-induced subsidence and ra-
and absorption are affected by the amount of carbon diative heating, evaporation at the cloud sides due
dioxide, ozone, methane, nitrous oxide, CFC-11, and to turbulent processes, and turbulent motion at the
CFC-12 present. Assumed parameters are the concen- cloud tops.
tration of these trace gases, the solar constant, the Precipitation processes not only take into account
distribution and optical properties of aerosols, and the local waterhce content, but also different precip-
ground albedo, which is modified according to the itation enhancement processes as well as evaporation
snow cover. of falling precipitation. Two mechanisms to generate
The radiation scheme is designed to take the cloud- precipitation are included - for convective and for
radiation interactions into account in considerable stratiform (frontal or dynamical) precipitation. In the
detail. It allows partial cloud cover in any layer of the case of convective precipitation the condensate
model. For cloudy grid points, computations are made formed in the updrafts of the convection is water
both for clear and overcast conditions, and the total above O'C, ice below - 23"C, and a mixture of the two
amount weighted according to the forecast cloud in between. If the amount of condensate formed
amount. The radiative effects of various types of exceeds the value that can be sustained by the vertical
aerosols (oceanic, continental, desert, urban, and velocity, precipitation is formed in the form of snow or
stratospheric background) are taken into account. water. When stratiform precipitation occurs cloud
The main purpose of the cloud scheme is to provide water and ice from the cloud scheme are converted
input to the radiation computations and to calculate into precipitation dependent on the waterhce content.
precipitation. The clouds are generated by large-scale It is assumed that falling precipitation evaporates in
ascent, cumulus convection, boundary layer turbu- nonsaturated layers before reaching the ground, a
lence, and radiative cooling. They are dissipated process that can substantially reduce the precipitation
through evaporation due to large-scale descent, cu- reaching the surface.
mulus-induced subsidence, radiative heating, and The circulation of the atmosphere is shaped by the
turbulence at both cloud tops and sides. The cloud contrasts between land and sea and the presence of the
scheme treats the main cloud-related processes in a major mountain ranges of the world, through the
consistent way by forecasting both cloud fraction and blocking effects on the flow and through the surface
cloud waterhce content with their own prognostic drag. Mountains have many scales associated with
equations. The cloud processes are strongly coupled to valleys and peaks and it is not sufficient just to
other parameterized processes. represent them by their average height. In fact, the
representation of the orography uses the mean oro- mixing processes are represented by a diffusion
graphy and four additional fields describing the operator with coefficients dependent on the local
standard deviation, orientation, anisotropy, and slope Richardson number. In addition, some algorithm is
of the sub-grid orography. This takes some account of used to estimate the depth of the mixed layer, within
the orographic variability, but even so, the model which there is enhanced mixing (the upper ocean
orography is still significantly smoother than reality. mixed layer is the equivalent of the atmospheric
However, the parametrization allows a realistic rep- boundary layer). Finally, models incorporate a
resentation of the mountain drag, which is important parametrization of convection that ensures stable
for the creation of large-scale atmospheric eddies. An conditions in a given ocean column.
important part of the scheme is that, depending on Early versions of ocean models obtained efficiency
dynamical criteria, it can block the low-level flow by assuming the ocean had a rigid lid, so excluding fast
rather than make the air go over the orography. When external gravity waves and allowing longer time steps.
stably stratified air crosses a mountain ridge, gravity Indeed, some models still use this approximation,
waves are excited. Depending on the static stability though free surface conditions whereby the top surface
and vertical wind shear, these gravity waves can can rise and fall as a result of replacing cold water by
propagate vertically until they have sufficiently large warm or salty by fresh are now more common.
amplitude to break. A gravity wave drag scheme is The ocean is forced by the transfer of momentum
included to represent the momentum transport due to between the atmosphere and the ocean. Strictly this
sub-grid gravity waves, contributing to the formation should be through the generation of surface waves, but
of blocking highs and cut-off lows. in fact a much simpler and local formulation is used
whereby the near-surface wind generates a momentum
flux that is passed to the upper ocean locally and
directly. The heat fluxes from the atmospheric model
Ocean Components (latent heating, sensible and long-wave) are applied at
An ocean GCM typically solves the equations of the surface. Solar heating can penetrate into the ocean
momentum, thermal energy, the continuity equation, a few tens of meters. Usually the transparency of the
the hydrostatic equation, and the state equation. An water is fixed, though some models are capable of
equation for salinity is also included as salt has an including a biological component and predicting the
important effect on density. The ocean is usually turbidity.
assumed to be incompressible, which allows some One problem that is more acute in the modeling of
simplifications to be made. A model typically has a the ocean than the atmosphere is the range of energetic
relatively high spatial resolution (-30 km) near the scales that are present and have to be accounted for.
Equator so that it can resolve and give a good The radius of deformation is much smaller in the ocean
representation of the equatorial waves which are so than in the atmosphere, meaning that energetic proc-
important in processes such as El Niiio. The zonal esses may take place in the ocean at scales of only a few
resolution is in the range 1-2". The vertical resolution kilometers, such as mesoscales eddies (the equivalent
is typically 30 vertical levels with more than 10 levels of atmospheric weather systems) and boundary cur-
within the upper 200 m. A realistic bottom topogra- rents. These eddies can be simulated explicitly in
phy is used. Unlike in the atmospheric case, depth- purpose-built models with specific applications in
following coordinates are not generally used. A depth mind, but not in the models used for climate processes.
( z ) coordinate system is by far the most common, Further, boundary currents such as the Kuroshio and
though a hybrid system using isopycnic coordinates in the Gulf Stream are poorly resolved in climate models
the deep ocean and z coordinates near the surface is leading to erroneous separation from the coast,
sometimes used. frequently resulting in errors in the interior circula-
Just as in the atmospheric case, ocean models must tion. To properly resolve these currents and eddies
parametrize sub-grid-scale processes. In some respects requires resolution of -5 km, which is far beyond
the ocean is a simpler system than the atmosphere, what can be achieved in climate models in the near
with fewer processes to parametrize. Mixing is a key future. The effects of eddies must therefore be
process. Horizontal mixing is usually done by some parameterized.
combination of straight Laplacian mixing or higher- Another difficulty is that it is thought that the world
order biharmonic mixing. The physical justification ocean circulation has a few important choke points.
for the latter is weak, but it has the advantage of One of these is the throughflow between the Pacific
reducing mixing except at larger scales. Variants on and Indian oceans. In reality the channels are narrow,
this include flow-dependent mixing such that it is much narrower than the model resolution. To over-
small, except where shear or strain are large. Vertical come this, the models must either use an artificially
wide channel, which may then allow incorrect trans- scales is that small uncertainties in the initial condi-
fers of heat and salt between the two oceans or increase tions will always give a large uncertainty in the state
the resolution in the neighborhood of the channels by a of the atmosphere after a few days. Because the
grid mapping. The overflow from the Greenland/ atmosphere and climate system are so chaotic, ensem-
Norwegian seas into the Atlantic is another region in ble techniques are needed whereby the forward
which the transfers are in error. Many models do not calculation of the ocean-atmosphere system is repeat-
take any special measures to try to limit the impact of ed many times, with small differences in the initial
such deficiencies on the circulation; others have tried conditions. The overall statistics of the predictions
specific local fixes, though these have not yet been used are then considered. The size of the ensemble
in extended climate integrations. is typically about 30, which is enough to give a
A potentially serious problem with using coupled reasonable estimate of seasonal changes in many areas
models to calculate the behavior of the ocean and of the tropics. A larger ensemble is needed to better
atmosphere is that errors in the models can accumu- resolve the weaker signals that are often present in
late, giving rise to what is called model drift, i.e., a midlatitudes.
tendency for the model to evolve towards an unreal- As mentioned earlier, fully coupled models drift, as a
istic state. In the ECMWF system many parts of the result of errors in one or both of the components. A
ocean surface cool down, while in a few places the strategy for dealing with model drift is necessary. In
ocean surface systematically warms up. Other models seasonal forecasting, one strategy is to initialize the
have different systematic errors. Since the size of the ocean, atmosphere, and land surface to be as close to
drift is comparable to that of the signal model drift reality as possible, and calculate the forward evolution
would be a serious problem if it were not dealt with in of the system as best we can using numerical approx-
some way. Two ways will be mentioned as examples: imations of the laws of physics. No ‘artificial’terms are
a posteriori correction as sometimes used in seasonal introduced to try to reduce the drift of the model, and
forecasting, and a priori correction in the form of flux no steps are taken to remove or reduce any imbalances
correction, as sometimes used in climate change in the coupled model initial state: we simply couple the
scenarios. models together and start to integrate forward. The
effect of the drift on the model calculations is
estimated from previous integrations of the model in
Applications of Coupled Models previous years. This mean drift is then removed from
The uses of coupled models are many but here we the model solution during the post-processing and
will consider only two: seasonal forecasting and before a forecast is issued.
climate change. The former is associated with trying The behavior of the atmosphere is influenced by a
to predict what will happen in the next few months, number of factors that themselves show variations
the latter in trying to predict what will happen in the from year to year. The most important of these is the
next hundred years. Curiously, there has been less ocean, whose surface temperature can have a substan-
effort in trying to predict what will happen in the next tial impact on the atmosphere, especially in the
10 years, though this application is likely to increase in tropics, where relatively small temperature changes
the future. can cause large shifts in the positions of tropical rain
It is not possible to predict the details of the weather systems. Although the ocean is the most important
beyond a few days, nor is it possible to predict exactly factor in seasonal forecasts, it is not the only one:
the average weather a season ahead, but it is possible changes in soil moisture and snow cover of the land
to predict the range of values that are likely to occur. In surface also play a role. It is also possible that some
some parts of the world, and in some circumstances, it aspects of the initial state of the atmosphere have a
may be possible to give a relatively narrow range part to play, although this is not certain. The impact of
within which weather values are expected to occur and all these factors on the atmosphere can be described
how this changes from year to year. Some of the technically as a shift in the (chaotic) attractor of the
forecasts associated with strong El Nifio events fall atmospheric system, meaning that these factors influ-
into this category. More typically, the probable ranges ence the statistics of the atmospheric behavior. They
of the weather are shifted only slightly from year to do not determine exactly what will happen, but they
year. Some tropical areas have a moderate amount of do influence, for example, the amount of storm
predictable signal, whereas in the midlatitudes ran- activity that might be expected at a certain place
dom weather fluctuations are usually larger than the during a certain period of time.
predictable component of the weather. An application of coupled models to seasonal
The reason the behavior of the atmosphere is not forecasting of El Niiio is shown in Figure 3. The value
predictable in a deterministic sense on seasonal time of the SST (sea surface temperature) in the central
Figure 3 Seasonal predictions of EL NiAo, 1997-98. (Courtesy of the Max Planck Institute, Hamburg, Germany and the ECMWF.)
eastern equatorial Pacific (Niiio-3)is shown, together Applications to Global Change

with various forecasts from different starting times
Use of coupled models to quantify the future climate
(indicated by color). The skill of the predictions is
given prescribed changes in radiatively active gases is a
generally quite good, though one can see cases when
major activity at a number of climate modeling centers
every ensemble member underpredicted the rising SST.
around the world. Just as ensembles of forecasts are
Coupled models can themselves be used to estimate the
needed for seasonal and decadal prediction, so, too,
accuracy with which El Niiio events could be predicted
are they needed for global change forecasts. Unfortu-
if the model were accurately specified and the initial
nately, each modeling group performs only a few
conditions were accurately known. On the basis of
integrations. The combination of forecasts by many
present estimates, there are grounds for optimism.
models should give a measure of the probability
One major uncertainty in the evolution of the El Niiio
distribution function spanning uncertainty in model
Southern Oscillation (ENSO) is the role played by
formulation (both parameterization and dynamical
the intraseasonal oscillation, poorly represented in
methods). Each model should be run with a set of
current models.
different initial conditions spanning the uncertainty in
knowledge of the current ocean state and to represent
Applications to Decadal Forecasting
the uncertainty resulting from chaotic processes. This
There are currently no applications of a regular is not yet being done in an organized way, as it is highly
and routine nature for this type of problem. How- computer intensive and sufficient resources are not yet
ever, the same strategy as used for seasonal forecasting available. At longer forecast leads the importance of
could be used for this longer time scale problem. the initial conditions is expected to decrease, unless a
It would be desirable to have a sea ice model, though major state change should result, such as the shutting
even for this time scale it may not be essential. down of the thermohaline circulation.
The ocean initial conditions would be obtained Despite the limitations of coupled models, they
as for seasonal forecasting using a combination remain the only way of making quantitative predic-
of past history of the atmospheric forcing fields and tions of the consequences of projected changes in
assimilation of ocean data. An assessment of greenhouse gases. Over the last few years, some
the scientific case and justification for such an activity models have reached a stage of development where a
is in progress, however. Whereas for seasonal fore- flux correction to offset climate drift is not required.
casting the emphasis has been on the Pacific, for Systematic errors still remain, however. For example,
decadal variability greater attention has been given to the distribution of stratus clouds over the colder ocean
the Atlantic. towards the eastern side of the subtropical oceans is
582 CRITICAL LAYERS
not well represented in models. Models do not in climate-related issues, there is no reason to believe that
general simulate the correct timing for the onset, rapid improvement in climate forecasting will not
decay, or intensity of the Asian monsoon. continue.
Until recently, it was thought that the stratosphere
played only a limited role in climate change but that
view is changing. The idea that climate change may See also
influence the frequency of occurrence of basic modes Boundary Layers: Modeling and Parameterization. Cli-
of variability has given greater impulse to the verifi- mate Variability: Decadal to Centennial Variability. Dy-
cation of a model’s ability to reproduce the preferred namic Meteorology: Primitive Equations. El Niiio and
patterns of variability. Three of these are the El Niiiol the Southern Oscillation: Theory. General Circula-
La Nifia or ENS0 cycle, the Pacific North American tion: Models. Global Change: Biospheric Impacts and
pattern and the North Atlantic Oscillation. The better Feedbacks; Surface Temperature Trends; Upper Atmos-
models do represent ocean-atmosphere interaction in pheric Change. Mesoscale Meteorology: Models. Nu-
the tropics with roughly the correct time scales. There merical Models: Chemistry Models; Methods. Spectral
is no consensus on how El Niiio events will change - Models. Weather Prediction: Regional Prediction
Models.
whether they become more frequent, or more intense.
A low-probability but high-consequence outcome
of global change is that the Atlantic thermohaline
circulation might shut down. Several coupled models Further Reading
suggest that the thermohaline circulation might be Anderson DLT, Sarachik ES, Webster PJ and Rothstein LM
reduced in the next hundred years, though none (eds)(1998)The TOGA Decade. Reviewing the progress
predict its collapse. However, the thermohaline circu- of El Niiio research and prediction. Journal of Geophysi-
lation is sensitive to processes not well represented in cal Research.
models. These include in particular the outflow from Davey MK et al. (2002) STOIC - A study of coupled model
the Greenland Norwegian Sea, and the fresh water climatology and variability in tropical ocean regions.
input from the atmosphere. Climate Dynamics 18: 4 0 3 4 2 0 .
Although coupled modeling has advanced consid- Gordon C et al. (2000)The simulation of SST, sea ice extents,
erably in the past decade, we are far from having well- and ocean heat transports in a version of the Hadley
validated models that can be used with confidence in Centre Coupled model. Climate Dynamics 16: 147-168.
climate forecasting. The different time scales and Hartman DL (1994) Global Physical Climatology.
Academic Press.
feedback mechanisms between the various processes
Houghton JT (eds) (2001) Climate Change 2001: The
makes the computations extremely complex and
Scientific Basis. IPCC report. Cambridge University
expensive. The mechanisms for these processes are Press.
mainly related to disturbances smaller than the scales Latif M et al. (2001) ENSIP - The El Nino simulation
explicitly resolved by the model, from convective intercomparison project. Climate Dynamics 18:
clouds down to molecular processes. The effect that 255-276.
these sub-grid-scale processes have on the larger scales Trenberth KE (ed.) (1992) Climate System Modeling.
can be computed only by parametrization, i.e., Cambridge University Press.
formulating indirectly their overall effect in terms of Washington W M and Parkinson CL (1987)An Introduction
known grid-scale variables. However, with increasing to Three Dimesional Climate Modeling. University
computer power likely and a growing interest in Science Books.
P H Haynes, University of Cambridge, Cambridge, UK background state that is a shear flow. One example is
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. that of Rossby waves (or planetary waves) propagat-
ing from the extratropical troposphere into the strat-
osphere, The background state here is the
Introduction longitudinally averaged flow, which may include
Theoretical models of waves in the atmosphere nat- westerly winds increasing in strength with height
urally require consideration of propagation on a (e.g., in the winter) or westerly winds at lower levels
582 CRITICAL LAYERS
not well represented in models. Models do not in climate-related issues, there is no reason to believe that
general simulate the correct timing for the onset, rapid improvement in climate forecasting will not
decay, or intensity of the Asian monsoon. continue.
Until recently, it was thought that the stratosphere
played only a limited role in climate change but that
view is changing. The idea that climate change may See also
influence the frequency of occurrence of basic modes Boundary Layers: Modeling and Parameterization. Cli-
of variability has given greater impulse to the verifi- mate Variability: Decadal to Centennial Variability. Dy-
cation of a model’s ability to reproduce the preferred namic Meteorology: Primitive Equations. El Niiio and
patterns of variability. Three of these are the El Niiiol the Southern Oscillation: Theory. General Circula-
La Nifia or ENS0 cycle, the Pacific North American tion: Models. Global Change: Biospheric Impacts and
pattern and the North Atlantic Oscillation. The better Feedbacks; Surface Temperature Trends; Upper Atmos-
models do represent ocean-atmosphere interaction in pheric Change. Mesoscale Meteorology: Models. Nu-
the tropics with roughly the correct time scales. There merical Models: Chemistry Models; Methods. Spectral
is no consensus on how El Niiio events will change - Models. Weather Prediction: Regional Prediction
Models.
whether they become more frequent, or more intense.
A low-probability but high-consequence outcome
of global change is that the Atlantic thermohaline
circulation might shut down. Several coupled models Further Reading
suggest that the thermohaline circulation might be Anderson DLT, Sarachik ES, Webster PJ and Rothstein LM
reduced in the next hundred years, though none (eds)(1998)The TOGA Decade. Reviewing the progress
predict its collapse. However, the thermohaline circu- of El Niiio research and prediction. Journal of Geophysi-
lation is sensitive to processes not well represented in cal Research.
models. These include in particular the outflow from Davey MK et al. (2002) STOIC - A study of coupled model
the Greenland Norwegian Sea, and the fresh water climatology and variability in tropical ocean regions.
input from the atmosphere. Climate Dynamics 18: 4 0 3 4 2 0 .
Although coupled modeling has advanced consid- Gordon C et al. (2000)The simulation of SST, sea ice extents,
erably in the past decade, we are far from having well- and ocean heat transports in a version of the Hadley
validated models that can be used with confidence in Centre Coupled model. Climate Dynamics 16: 147-168.
climate forecasting. The different time scales and Hartman DL (1994) Global Physical Climatology.
Academic Press.
feedback mechanisms between the various processes
Houghton JT (eds) (2001) Climate Change 2001: The
makes the computations extremely complex and
Scientific Basis. IPCC report. Cambridge University
expensive. The mechanisms for these processes are Press.
mainly related to disturbances smaller than the scales Latif M et al. (2001) ENSIP - The El Nino simulation
explicitly resolved by the model, from convective intercomparison project. Climate Dynamics 18:
clouds down to molecular processes. The effect that 255-276.
these sub-grid-scale processes have on the larger scales Trenberth KE (ed.) (1992) Climate System Modeling.
can be computed only by parametrization, i.e., Cambridge University Press.
formulating indirectly their overall effect in terms of Washington W M and Parkinson CL (1987)An Introduction
known grid-scale variables. However, with increasing to Three Dimesional Climate Modeling. University
computer power likely and a growing interest in Science Books.
P H Haynes, University of Cambridge, Cambridge, UK background state that is a shear flow. One example is
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. that of Rossby waves (or planetary waves) propagat-
ing from the extratropical troposphere into the strat-
osphere, The background state here is the
Introduction longitudinally averaged flow, which may include
Theoretical models of waves in the atmosphere nat- westerly winds increasing in strength with height
urally require consideration of propagation on a (e.g., in the winter) or westerly winds at lower levels
CRITICAL LAYERS 583
changing to easterly winds at upper levels (e.g., in eastward direction and y measured in the northward
the summer). Another example is that of small-scale direction. The corresponding velocity components are
internal gravity waves excited by a mountain and taken to be ( u ,v ) .The assumption of incompressibility
propagating upward through a large-scale flow implies that the velocity components may be expressed
that changes strength (and perhaps direction) with in terms of a streamfunction $ ( x , y, t ) (with t time)
height. where u = -a$/ay and v = a$/ax.
Suppose that the background flow (i.e., the flow in The governing equation is based on the fact that, in
the absence of the waves) is in the x-direction with the absence of dissipation, the absolute vorticity,
speed U that is a function of a second space coordinate which is the sum of the relative vorticity [ = a v / a x -
y and that the waves have a well-defined phase speed c au/ay = V2$ and the planetary vorticity by, is con-
in the x-direction. Then a location where U(y) = c served following the fluid motion. 8 is a constant and
(i.e., where the flow speed matches the phase speed) is in an Earth-like atmosphere is positive. It is convenient
a line parallel to the x-axis and at a fixed value of y, to include linear damping of vorticity in the model as a
called a critical line. Where the second space coordi- simple representation of a dissipative processes. The
nate is height, the equivalent term critical level is often governing equation then becomes eqn [l], where t( is
used. If the speed U were a function of two space the damping rate.
coordinates y and z, then the location U(y,z) = 0
would define a critical surface.
Simple theories for the structure of waves are often
based on the assumptions that the waves are propa-
gating steadily, that dissipative or diabatic processes (Another possibility for a dissipative process would be
such as friction or radiative transfer may be neglected, diffusion of vorticity. Neither linear damping nor
and that the waves are small in amplitude, so that diffusion is likely to be a realistic representation of
terms in the equations of motion that are nonlinear in dissipative processes in the real atmosphere, but either
wave quantities may be neglected. These theories lead serves as a convenient example that captures the basic
to a straightforward differential or partial differential effect of dissipation in the critical layer.)
equation that describes the spatial structure of the It is assumed that in the absence of waves the flow is
waves. The importance of the critical line is that it is a in the x-direction with speed U(y). Waves are super-
location where these differential equations are singu- imposed on this flow giving a contribution $ ’ ( x , y, t )to
lar; in other words, the solutions imply that some the streamfunction. Then the eqn [ 11may be written in
physical quantity becomes infinite. terms of’)I as eqn [2], where u’ = -a$’/ay and v’ =
As in many physical contexts, the appearance of a$’/ax are the wave velocity components
singular behavior in a mathematical model implies
that the simplifications that lead to that model cannot
be justified and that some physical process that was
neglected must be retained. To remove the critical-line
singularity, one of the neglected processes mentioned
above must therefore be included (however weak such
processes might have been estimated to be). The
neglected process will be essential only in a small, but If the damping is weak, it is reasonable to neglect the
finite, region around the critical line and may still be first term on the right-hand side. If the waves are small-
negligible elsewhere. This small but finite region is amplitude, it is reasonable to neglect the second and
named the critical layer. third terms on the right-hand side, which are quadratic
in wave quantities. Since the resulting equations are
linear and contain no explicit x-dependence, it is
The Rossby Wave Critical Layer possible to consider waves with different wavelengths
in the x-direction as independent.
A Simple Model
Consider one such wave, with x-wavenumber k
One of the simplest examples of the critical-line (i.e., wavelength 2 z / k in the x-direction), assumed
singularity and its resolution in a finite critical layer to be steadily propagating in the x-direction with
arises in a two-dimensional model of Rossby wave phase speed c. It follows that the streamfunc-
propagation on a 8-plane (a mathematical device to tion for this wave may be written in the form
include the effect of the variation of the Coriolis $ ’ ( x , y, t ) = Re[$(y)eik(x-ct)1, where $(y) is a (com-
parameter with latitude). Two-dimensional Cartesian plex) function of y. Substituting into eqn [2] and
coordinates ( x , y) may be used, with x measured in the neglecting terms on the right-hand side gives the
584 CRITICAL LAYERS
ordinary differential equation [3]. K

.-0 Wave source A
Equation [3] is known as the Rayleigh-Kuo equation

and determines the stability of the shear flow U(y), as
well as describing (through the function $(y)) the
structure of propagating waves.
The appearance of the factor (U(y) - c)-l in part of
the coefficient of $ indicates that the equation has a
singular point at values of y such that U(y) = cy i.e.,
where the phase speed matches the flow speed. These
locations are the critical lines. If U(y) is an increasing
or decreasing function of y, there is at most one critical
Figure 1 Schematic diagram of Rossby wave propagation on
line. If U(y) has a turning point (as would be the case a shear flow U(y) with a critical line. The flow is positive
for a jetlike flow, for example), there may be more than (Le., eastward) in y > 0 (upper portion of the diagram) and negative
one critical line. (Le., westward) in y<O (lower portion of the diagram). The waves
Consider the solution near a critical line at y = yc. are forced, with zero phase speed in the x-direction, in y > 0 and
The nature of eqn [3] is such that while $(y) propagate toward y = 0. In y<O the waves are evanescent (Le.,
non-propagating and decaying as ybecomes more negative). The
(proportional to the velocity in the y-direction) is critical line is at y = 0, where U(y) = 0. In the neighborhood of
finite and continuous in the neighborhood of the y = 0 the streamlines are closed and form a Kelvin's cat's eye
critical line, d2$/dy2 (representing part of the vorti- pattern. The width of the closed streamline region, which increases
city) is proportional to (y - yC)-' and d$/dy (repre- as the wave amplitude increases, defines the width of the nonlinear
senting the velocity in the x direction) is proportional critical layer. If dissipation were strong enough, dissipative effects
would dominate over a relatively broad region near y = 0 and the
to log J y- ycl; i.e., both are singular. This is clearly closed streamlines would essentially be irrelevant to the dynamics.
unphysical, but what is more problematic, in a way, is (The critical layer would then be linear and dissipative, rather than
that the singularity in d$/dy implies that there is no nonlinear.) There may be some reflected wave in y > 0, but the
unique way to match solutionsAofeqn [3] across amount of reflection can be determined only by considering the
y = yc. In particular, the jump in d$/dy corresponding detailed dynamics of the critical layer.
to the jump in u', across y = yc remains unknown. It
follows there is no unique solution in the whole flow
domain for the function $(y). The critical-line singu- waves. One feature of the solution that is naturally
larity must therefore be resolved not only to remove of interest is the relative amount of northward- and
the local singular behavior in certain physical quan- southward-propagating waves in the region between
tities but also to determine the structure of the waves the wave forcing and the critical layer. This measures
over the whole flow. the absorption-reflection behavior of the critical layer.
If the critical layer acted as an absorber of waves, the
Absorption-Reflection
region between the forcing and the critical line would
To note the implications of the critical layer for the contain only waves propagating southward. If it acted
waves elsewhere in the flow, it is useful to focus on the as a reflector of waves, there would be some contri-
following geometry, shown in Figure 1. Assume that bution to the solution in this region from waves
the waves are forced at some large positive value of y, propagating northward. The reflection could be par-
with phase speed c = 0; i.e., the waves are stationary. tial or perfect. Indeed, there could in principle be
Assume also that the flow speed U ( y ) is positive in overreflection, which would be associated with a
y > 0 and negative in y < 0, so that the waves have a greater proportion of northward-propagating than
critical line at y = 0, and that the curvature term southward-propagating waves, implying that the crit-
d2U/dy2 is not too large (so that /3 - d2U/dy2 is ical layer was actively emitting waves.
positive). Equation [3] predicts that the function $(y) A useful quantitative measure of the wave propa-
is oscillatory in y > 0, implying that there are prop- gation is the momentum flux, u", where indicates
agating waves (as expected from the basic properties an average in the x-direction (keeping y and t
of Rossby waves). In y<O, on the other hand, the constant). u " indicates the correlation between the
function $(y) is exponentially increasing or decreasing two velocity components and the basic properties of
with y and physical considerations require that $(y) Rossby waves imply 27 > 0 for southward-propa-
decreases as y decreases, representing evanescent gating waves and 27<0 for northward-propagating
CRITICAL LAYERS 585
waves. It follows from eqn [3] that UIV‘ is constant in are predicted to increase without bound. The time-
y > 0, except at the critical line at y = 0. The evanes- dependent analysis does not resolve the problem of the
cence of the waves in y < 0 implies that u” = 0 there. singularity, but it shows that the terms neglected in
However, the value of u” cannot be determined from going from eqn [2] to eqn [3] inevitably become
eqn [3] alone. Instead, the critical line must be resolved important at large times, however small they might
into a finite critical layer to allow the jump in u” have first appeared.
,:I‘&[
across the layer, denoted by and hence the value Second, consider the dissipative term -aV2$’ in eqn
of u“in y > 0, to be evaluated. [2]. This may be compared with the advection term
The continuity of $(y) across the critical-line U(y) aV2$’/ax.The relative sizes of these terms, near
singularity suggests that, when the critical line is to y = 0, may be estimated as a/(kU’(O)y), where
resolved into a thin critical layer, $’ and hence v’ = U’(0) represents dU/dy at y = 0, and it follows that
a$’/ax will vary only weakly across the critical layer. the dissipative term cannot be neglected in a region of
In addition, y-derivatives within the critical layer will size 6, = a/(kU’(O)). This is (potentially) the thick-
generally be much larger than x-derivatives (because ness of the dissipative critical layer.
the layer is thin), implying that [’ N -au’/ay. Putting Finally, consider the nonlinear term. It turns out that
these pieces of information together, it follows that eqn the most important part is v’ aV2$’/i3y. If this is to
[4] holds, where [u’]: denotes the jump in u’ across the balance U(y) aV2$’/i3x N U’(0)yaV2$’/ax in a thin
critical layer, the integrals are taken across the criticalregion of thickness ~ N Lthen , U ’ / ~ N L ~U’(O)GNL,
N Le.,
layer, and v’ may be taken as constant within the ~ N L ( ~ ’ / k U ’ ( o ) ) lThis
N ’ ~ . is (potentially) the thick-
second integral. ness of the nonlinear critical layer.
1
[u’]$.= [‘dy
Whether nonlinearity or dissipation is dominant in
the critical layer de ends on the relative size of
6 ~ ~ =/ (kv’U’(0))
6 ~ 1pz / a . If 8 ~ ~ / 6 ~ < <then
1 , the
and critical-layer dynamics are dominated by dissipation
/
u ” ( i n y > 0 ) = [u”]!= m d y
and the critical layer thickness is 6,. If ~ N L / & >> 1,
141 then the critical-layer dynamics are dominated by
nonlinear processes and the critical layer thickness is
The first equality is the missing matching condition ~ N L In . the case of Rossby waves in the real atmos-
across the critical layer. The second shows that the phere, wave amplitudes are relatively large and
critical layer acts as a net absorber of waves when there dissipation is relatively weak, so that the nonlinear
is (in a y-integrated sense) negative correlation dynamics are the most relevant.
between 5’ and v’ in the critical layer, as a perfect The fully nonlinear equations state that + Py is
reflector when there is zero correlation, and as a conserved following the flow (which is in turn deter-
net emitter (i.e., an overreflector) when there is mined by the [ field). If the critical layer is thin, i.e., ~ N L
positive correlation. is small, there is a simplification because the flow may
To summarize, the nonuniqueness in the solution of be approximated by the superposition of the basic flow
eqn [3] left by the critical-line singularity leaves U(y) 11 U’(0)y and the y-component of the distur-
the absorption-reflection behavior of the critical layer bance velocity field, which is simply a function of x,
uncertain. Only by determining the correlation because of the continuity of $’ across the critical layer.
between v’ and [’ in the critical layer is it possible This superposition gives a flow whose streamlines
to determine the precise absorption-reflection form a pattern known as Kelvin’s cat’s eyes, with
properties. closed streamlines near y = 0 (see Figure 1). The
dynamics of the nonlinear critical layer is therefore
The Dynamics of the Critical Layer
that fluid particles are advected around these stream-
+
The dynamical balance in the critical layer depends on lines, conserving their values of [ by. The rearrange-
the parameters of the problem. Consider in turn the ment of the [ + Py field changes the [’ field, thereby
processes that have been neglected in arriving at eqn changing [u’]: and hence the structure of the waves
[3]. First, it has been assumed that the waves are outside the critical layer. Furthermore, there is a
steadily propagating, i.e., that their amplitude is not corresponding change in the correlation between v’
changing with time. It is possible to analyze the non- and c’,which determines the absorption-reflection
dissipative, linearized equations (eqn [2] with the properties.
right-hand side set to zero) without making this A schematic diagram of the evolution of the
assumption and to show that the singular behavior vorticity field in a simple nonlinear critical layer
predicted by eqn [3] develops with time. For example, is presented in Figure 2 and the absorption-reflection
c’
both the vorticity and the x-component of velocity u’ properties are deduced. At early times v‘ and are c’
586 CRITICAL LAYERS
Figure 2 Evolution of thevorticity field in the nonlinear critical layer. The panels show an expandedview of the vorticityfield in the closed
streamline region indicated in Figure 1. Note that this region may be vety thin in the y-direction if the wave amplitude is small. Heavy
dashed curves are the bounding closed streamlines. Heavy solid curves are contours of absolute vorticity [’ + f l y . Light curves are
contours of wave relative vorticity [‘, with solid curves indicating positive values and dashed curves indicating negative values. The three
panels show (A) absorbing stage, where y-velocity v’ is negativelycorrelated with 5‘; (B) reflecting stage, where correlation between v’ and
5’ is close to zero; and (C) overreflecting stage, where v’ is positively correlated with i‘.
anticorrelated and the critical layer acts as an absorber. rearrangement continues, to give a positive correla-
If there is strong dissipation, the vorticity in the tion between u‘ and 5’ and hence overreflection, and
center of the critical layer is essentially frozen in the critical layer subsequently oscillates between
its early-time configuration and the critical layer a weakly absorbing and weakly overreflecting state,
continues to act as an absorber at later times. (Detailed converging to a state of perfect reflection at large
calculation shows that in this early-time/dissipative times.
regime, the absorption is effectively perfect.) How- The precise details of the evolution depend on the
ever, if dissipation is weak, the advective rearrange- particular flow configuration. However, a general
ment continues and, after about half a turn-round description of absorption-reflection behavior can be
time for the closed-streamline flow, the 5’ field in formulated by considering -22 as the flux (in the
the center of the critical layer (which gives the major y-direction) of wave activity (i.e., a quantity that is
contribution to the integral) is such that there is positive when waves are present and zero when they
no y-integrated correlation with the u’ field; are not). In the early-time absorbing stage, wave
that is, the critical layer acts as a perfect reflector. activity builds up in the critical layer. As the reflecting
According to this particular model, the advective stage approaches, the rate of build-up decreases to
CRITICAL LAYERS 587
zero and in the overreflecting stage the critical layer Wave Breaking
reemits some of its wave activity. If there is dissipation,
The behavior seen in the nonlinear critical layer for
then the flux of wave activity into the critical layer may
~ N Lsmall may be interpreted as an example of the
be balanced by local dissipation of wave activity and
breaking of Rossby waves. By ‘breaking’ it is meant
an absorbing state may persist.
that the material contours or surfaces that would, in
However, for the critical layer to continue to
wave propagation, be reversibly undulated are strong-
act as an absorber without dissipation, the amount
ly and irreversibly deformed. The most familiar
of wave activity in the critical layer must continue
example of wave breaking occurs for surface waves.
to increase. The total amount of wave activity in
There the wave dynamics is associated with the
the critical layer may be shown to depend on the
undulation of the ocean surface. Waves might be
thickness of the region over which the vorticity field
forced in one region (e.g., by a storm) and propagate
has been rearranged, i.e., the thickness of the critical
through large distances. The presence of the waves in
layer. If this thickness is finite, there is an upper bound
this propagation stage is associated with distortion of
to the total amount of wave activity that can be stored
the ocean surface, but the distortion is weak and
there, and it is therefore not possible to sustain
reversible. As the waves enter shallow water in a
absorption. In such a case, the long-time average of
coastal region, the distortion of the ocean surface
the flux of wave activity must approach zero and
becomes stronger and, ultimately, complex and irre-
one can therefore say that the long-time average
versible and the flow will become three-dimensionally
behavior is perfect reflection. The only way that
turbulent.
absorption could be sustained in the long-term would
Rossby wave propagation involves the reversible
be if the thickness of the critical layer systematically
distortion of contours of potential vorticity (absolute
increased in size.
vorticity in the simple two-dimensional context dis-
A complementary viewpoint comes from consider-
cussed above). In the critical-layer region, the distor-
ing u” as a momentum flux. In the absorbing stage
tion of these contours is strong and irreversible and the
[u”]? is positive, implying that there is a negative
waves may be said to be breaking. Indeed, in many
force exerted on the x-averaged flow in the critical
cases the flow in the critical layer may be shown to
layer. The time-averaged perfect reflection in the case
involve a sort of turbulence (quasi-geostrophic or two-
where dissipation is zero translates into no time-
dimensional), but this is not essential for the behavior
averaged x-average force exerted on the flow in the
to be described as breaking. As in the surface-wave
critical layer. (If there were such a force then the
case, where the breaking may allow the waves to drive
critical line, and hence the critical layer, would move
systematic long-shore currents, the breaking of Rossby
closer and closer to the wave source.) Sustained
waves allows a systematic force to be exerted by the
absorption where there is dissipation translates into waves.
a nonzero time-averaged x-averaged force exerted on
the flow in the critical layer, with this force being
Implications for the Atmosphere
balanced by forces provided by dissipative processes
(Le., by the linear vorticity damping in the model In the nonlinear Rossby wave critical layer described
described above). above there is a clear division (described by simplified
The critical layer theory makes clear the nature of mathematical equations that may be derived by a
the two-way interaction between the wave propaga- formal approximation procedure) between the broad
tion region outside the critical layer and the flow in the region outside the critical layer where the dynamics is
critical layer itself. The waves outside the critical layer wavelike and the thin critical layer itself where the
directly determine the flow pattern inside it (because of dynamics is strong advective rearrangement of the
the continuity oft.” across the critical layer). However, potential vorticity or absolute vorticity field, which
inside the critical layer, the flow changes the vorticity might be called wave breaking. In the real atmosphere,
field and hence the jump in u’ across the critical layer, the wave amplitudes are very large and the formal
thereby changing the waves outside it. It is important estimate ~ N of
L the nonlinear critical layer thickness is
to note that there is no wave propagation within generally as large as the other length scales in the
the critical layer itself. The dynamics is simply that problem. The same formal mathematical equations do
of vorticity advection by a simple cat’s eye flow not hold. Nonetheless, observations and numerical
whose structure is determined by the waves outside models show clearly that there are regions of wave
the critical layer. It is not the case that waves can be propagation and regions of wave breaking and that
said to propagate into the critical layer and be reflected these exist side by side. There are at least two
by the structure of the flow profile that they encounter important examples. One is in the winter stratosphere,
there. where planetary-scale Rossby waves propagate up
588 CRITICAL LAYERS
from the troposphere, distort and shift the polar-night This equation is known as the Taylor-Goldstein
vortex, and break in what is now called the strato- equation and determines the stability of density-
spheric ‘surf zone’ (which covers a large region of the stratified shear flows, as well as governing the struc-
mid-latitudes and subtropics). A second example is in ture of waves propagating on such flows. Again the
the upper troposphere and lower stratosphere, where critical-line singularity at U ( z ) = c is manifested by
synoptic-scale baroclinic eddies lead to a wavelike the inverse powers of ,U(z)-c appearing in the
distortion of the subtropical jet and to a wave- expression multiplying $. There is an important
breaking region on the equatorward side of the jet. difference from eqn [3] in that the one of expressions
Nonlinear critical layer theory provides quantitive contains the factor of ( U ( z )- c)~.This means that the
guidance as to how the different regions interact. For behavior of solutions near the critical-line singularity
example, it indicates that the wave-breaking regions is quite different from the Rossby wave case. In fact,
may be considered to absorb, reflect, or emit wave provided that N(z)2/(dU/dz)2 > (which is precisely
activity and that the waves may have a systematic the condition required for the flow to be stable) the
effect on the flow in the wave-breaking region. function $(z) oscillates rapidly in z near to the critical
There have been attempts, based in part on Rossby line and the oscillations become infinitely rapid as the
wave critical-layer theory to parameterize the effects critical line is approached. Indeed there are infinitely
of Rossby wave propagation and breaking on the many oscillations before the critical line is reached.
average flow and on the average distribution of These oscillations are a manifestation of the rapid
advected chemical species. These parameterizations shrinking of the vertical wavelength of the wave as the
have had most impact in modeling of stratospheric critical line is approached, due to the tilting of the
chemistry, where two-dimensional height-latitude wave by the shear. An analogous shrinking of the
models remain an important tool for assessment of wavelength occurs in the Rossby wave case, but there
chemical change over time scales of many decades. are only a finite number of oscillations before the
The trend in chemical modeling is to move to models critical line is reached - a subtle and important
that are three-dimensional and explicitly resolve the difference between this and the internal gravity wave
Rossby wave dynamics. It is therefore likely that future case. The reason for the difference is that in the
development of Rossby wave parameterizations will internal gravity wave case decrease in wavelength
not proceed much further. In the Rossby wave case the gives a stronger decrease in the group velocity (i.e., the
greatest importance of critical layer theory is as a propagation velocity for wave packets). Indeed in the
paradigm on which to base understanding of complex Rossby wave case the idea of group velocity is simply
flows to which the theory is not formally applicable, not at all useful in the neighborhood of the critical line,
but for which the only possibility of accurate quanti- whereas in the internal gravity wave case it is.
tative calculation is full numerical simulation. Again the critical-line singularity can be resolved
either by dissipation or by nonlinearity, depending on
Internal Gravity Wave Critical Layers the relative strengths of the two. One possibility is that
the wave will eventually dissipate. This is possible
Description however weak dissipative processes might seem, since
Critical lines and critical layers arise generically in any the decrease in group velocity as the critical line is
problem of wave propagation in a fluid. Another approached means that there is infinite time for the
example that is particularly important for the atmos- dissipation to act. Indeed, the wave will dissipate
phere is that of internal gravity waves. This has some before the critical line is reached. In this case the
important differences from the Rossby wave case. thickness of the dissipative critical layer may be
Consider the propagation of internal gravity waves defined as the distance to the critical line at which
on a background state that has stable stratification the dissipation occurs, and the critical layer may be
with buoyancy frequency N ( z ) and flow in the regarded as a wave absorber. If wave amplitudes are
x-direction with speed U ( z ) , where z is height. sufficiently large compared to dissipative processes,
Assuming that the flow is incompressible (which is however, then nonlinear terms in the equations may
not necessarily defensible for many atmospheric grav- become important before dissipation occurs and,
ity waves, but the model serves to illustrate important again, before the critical line is reached. The distance
points that continue over to the compressible case), it to the critical line defines the thickness of the nonlinear
may be shown that the analogue of [3] is eqn [ 5 ] . critical layer. Here the situation is much more
complicated than in the Rossby wave case. For
example, it is not possible to argue that the velocity
component in the z-direction (analogous to v’ in the
Rossby wave case) is continuous across the nonlinear
CYCLOGENESIS 589
critical layer (and therefore independent of z within essential for useful simulation of the stratosphere
the critical layer). If governing equations for the and mesosphere. One very simple parameterization
nonlinear critical layer are derived, they are essentially would be that, for a spectrum of upward-propagating
the full nonlinear governing dynamical equations, gravity waves, each component of the spectrum
with a slight simplification because the structure is dissipates at its critical line and therefore gives rise to
very thin in the z-direction. The critical-layer dynam- a force at that location. In practice, some kind of
ics is therefore a complex juxtaposition of wave breaking criterion is applied so that waves break
propagation and nonlinearity. Furthermore, if nonlin- before the critical line is reached. Almost all current
earity is important it is also almost inevitable that there parameterizations assume the equivalent of critical-
will be the potential for gravitational instability and layer absorption. If critical-layer reflection had to be
therefore, in reality, breakdown of the flow into taken into account then it would greatly increase the
complex three-dimensional turbulence. For this rea- complexity of the parameterization problem.
son there are few clear results on the evolution of the
nonlinear internal gravity wave critical layer, although
it is believed that some nonlinear reflection effect is See also
possible. Dynamic Meteorology: Overview; Waves. Middle At-
mosphere: Gravity Waves; Planetary Waves; Quasi-
Implications for the Atmosphere Biennial Oscillation. Rossby Waves. Wave Mean-Flow
Dissipation and breaking of internal gravity waves as Interaction.
they approach critical lines is potentially an important
process in the atmosphere, since it implies the possi-
bility of wave-induced forces. Breaking may also be Further Reading
caused by the decrease of density with height, which Andrews DG, Holton JR and Leovy CB (1987) Middle
leads t o a corresponding increase in wave amplitudes. Atmosphere Dynamics. London: Academic Press.
However, there is little doubt that breaking at (or more Brunet G and Haynes PH (1996)Low-latitude reflection of
strictly near) critical lines also plays a major role. For Rossby wavetrains. Journal of the Atmospheric Sciences
example, the mechanism for the equatorial quasi- 53: 482-496.
biennial oscillation in the stratosphere requires selec- Dornbrack A (1998) Turbulent mixing by breaking gravity
tive filtering, breaking, and dissipation of waves waves. Journal of Fluid Mechanics 375: 113-141.
(depending on their horizontal phase speed) by the Maslowe SA (1986) Critical layers in shear flows. Annual
background flow. Such waves are believed to arise Review of Fluid Mechanics 18: 405-432.
McIntyre ME and Norton WA (1990) Dissipative wave-
primarily from convection in the tropical troposphere
mean interactions and the transport of vorticity or
(on a whole range of different scales). Observations
potential vorticity. Journal of Fluid Mechanics 212:
confirm the expected relation between the phase speed
403-435. [Corrigendum 220: 693.1
of the waves observed at a particular height and the McIntyre ME (2000) On global-scale atmospheric circula-
background flow at lower levels, through which they tions. In: Batchelor GK, Moffat HK and Worster MG
would have propagated. (eds) Perspectives in Fluid Dynamics: A Collective
Critical-line/critical-layerbehavior is an important Introduction to Current Research, pp. 557-624.
ingredient of gravity-wave parameterizations that Cambridge: Cambridge University Press.
seek to represent the effects (primarily the wave- Staquet C and Sommeria J (2002) Internal Gravity Waves:
induced forces) of small-scale gravity waves in global- From Instabilities to Turbulence.Annual Review of Fluid
scale numerical models. Such parameterization is Mechanics 34: 559-594.
G J Hakim, University of Washington, Seattle, WA, USA of planetary rotation; this circulation is clockwise in
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. the Southern Hemisphere and counterclockwise in the
Northern. Cyclogenesis is defined as the formation
and amplification of cyclones. The emphasis here will
Introduction
be on cyclogenesis in the extratropical latitudes, near
Cyclones are associated with horizontal winds circu- the westerly jet streams; i.e., poleward of about 30"
lating in the same sense as the local vertical component latitude. Extratropical cyclones (hereafter, cyclones)
CYCLOGENESIS 589
critical layer (and therefore independent of z within essential for useful simulation of the stratosphere
the critical layer). If governing equations for the and mesosphere. One very simple parameterization
nonlinear critical layer are derived, they are essentially would be that, for a spectrum of upward-propagating
the full nonlinear governing dynamical equations, gravity waves, each component of the spectrum
with a slight simplification because the structure is dissipates at its critical line and therefore gives rise to
very thin in the z-direction. The critical-layer dynam- a force at that location. In practice, some kind of
ics is therefore a complex juxtaposition of wave breaking criterion is applied so that waves break
propagation and nonlinearity. Furthermore, if nonlin- before the critical line is reached. Almost all current
earity is important it is also almost inevitable that there parameterizations assume the equivalent of critical-
will be the potential for gravitational instability and layer absorption. If critical-layer reflection had to be
therefore, in reality, breakdown of the flow into taken into account then it would greatly increase the
complex three-dimensional turbulence. For this rea- complexity of the parameterization problem.
son there are few clear results on the evolution of the
nonlinear internal gravity wave critical layer, although
it is believed that some nonlinear reflection effect is See also
possible. Dynamic Meteorology: Overview; Waves. Middle At-
mosphere: Gravity Waves; Planetary Waves; Quasi-
Implications for the Atmosphere Biennial Oscillation. Rossby Waves. Wave Mean-Flow
Dissipation and breaking of internal gravity waves as Interaction.
they approach critical lines is potentially an important
process in the atmosphere, since it implies the possi-
bility of wave-induced forces. Breaking may also be Further Reading
caused by the decrease of density with height, which Andrews DG, Holton JR and Leovy CB (1987) Middle
leads t o a corresponding increase in wave amplitudes. Atmosphere Dynamics. London: Academic Press.
However, there is little doubt that breaking at (or more Brunet G and Haynes PH (1996)Low-latitude reflection of
strictly near) critical lines also plays a major role. For Rossby wavetrains. Journal of the Atmospheric Sciences
example, the mechanism for the equatorial quasi- 53: 482-496.
biennial oscillation in the stratosphere requires selec- Dornbrack A (1998) Turbulent mixing by breaking gravity
tive filtering, breaking, and dissipation of waves waves. Journal of Fluid Mechanics 375: 113-141.
(depending on their horizontal phase speed) by the Maslowe SA (1986) Critical layers in shear flows. Annual
background flow. Such waves are believed to arise Review of Fluid Mechanics 18: 405-432.
McIntyre ME and Norton WA (1990) Dissipative wave-
primarily from convection in the tropical troposphere
mean interactions and the transport of vorticity or
(on a whole range of different scales). Observations
potential vorticity. Journal of Fluid Mechanics 212:
confirm the expected relation between the phase speed
403-435. [Corrigendum 220: 693.1
of the waves observed at a particular height and the McIntyre ME (2000) On global-scale atmospheric circula-
background flow at lower levels, through which they tions. In: Batchelor GK, Moffat HK and Worster MG
would have propagated. (eds) Perspectives in Fluid Dynamics: A Collective
Critical-line/critical-layerbehavior is an important Introduction to Current Research, pp. 557-624.
ingredient of gravity-wave parameterizations that Cambridge: Cambridge University Press.
seek to represent the effects (primarily the wave- Staquet C and Sommeria J (2002) Internal Gravity Waves:
induced forces) of small-scale gravity waves in global- From Instabilities to Turbulence.Annual Review of Fluid
scale numerical models. Such parameterization is Mechanics 34: 559-594.
G J Hakim, University of Washington, Seattle, WA, USA of planetary rotation; this circulation is clockwise in
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. the Southern Hemisphere and counterclockwise in the
Northern. Cyclogenesis is defined as the formation
and amplification of cyclones. The emphasis here will
Introduction
be on cyclogenesis in the extratropical latitudes, near
Cyclones are associated with horizontal winds circu- the westerly jet streams; i.e., poleward of about 30"
lating in the same sense as the local vertical component latitude. Extratropical cyclones (hereafter, cyclones)
590 CYCLOGENESIS
are responsible for organizing significant short-term surface pressure, and is associated with net divergence
weather, such as cloud and precipitation patterns. of air from the air column. Frictional drag at the
Cyclones are also crucial components of the Earth’s surface disrupts geostrophic balance, so that the near-
climate system because they regulate the pole-Equator surface winds converge toward low pressure. There-
temperature contrast, stabilize the density stratifica- fore we can conclude that upper-level divergence must
tion, and maintain the westerly winds in midlatitudes be important for cyclogenesis. Moreover, since the
against frictional dissipation. atmosphere is nearly incompressible on horizontal
Cyclones have horizontal length scales on the order length scales typical of cyclones, lower-level conver-
of about 500-2500 km and may span the depth of the gence and upper-level divergence are linked by mass
troposphere (- 10 km). When compared with an continuity to upward air motion in the troposphere.
undisturbed environment, cyclones are associated It is reasonable to assume the existence of a
with relatively lower pressure (- - 10 hPa), circulat- constant-pressure surface, Po, in the stratosphere
ing winds (- 10 m s - ’), relatively warm air (- 10 K), that is undisturbed by the cyclone. Decreasing surface
and rising air (- 10 cm s - I ) . The wind field is nearly in pressure during cyclogenesis implies that the thickness
geostrophic balance, so that the wind flows parallel to between a constant-pressure surface near the ground
lines of constant pressure and the wind speed is and Po increases; i.e., the layer is warming with time.
proportional to the magnitude of the horizontal Falling pressure in the cyclone center also implies that
pressure gradient (the spacing between the pressure the magnitude of the pressure difference between the
contours). Cyclones also represent locally large values cyclone center and the surrounding environment, and
of other quantities that are derived from the wind, therefore the geostrophic wind speed, also increases
temperature, and pressure fields, such as: kinetic with time.
energy, vorticity, and potential vorticity. Vorticity pro- These deductions on cyclogenesis, based on geo-
vides a local measure for the rotation rate of the wind. strophic and hydrostatic balance, are self-consistent;
Potential vorticity is approximately the product of however, they do not reveal why or how these changes
vorticity and a measure of density stratification; it gives come about. Further analysis of the time variation of
the vorticity that a sample of air would have (i.e., individual quantities, such as surface pressure, are
potentially) if taken to a reference latitude and rear- incomplete, since, importantly, all quantities are
ranged adiabatically to a reference density stratification. dynamically related. An analogy for this difficulty
Potential vorticity (PV) plays a central role in the considers analyzing a moving automobile to discover
modern understanding of cyclogenesis, and is reviewed the process responsible for locomotion. One might
below under ‘Dynamics of Cyclone Development’. consider the decreasing mass of gasoline in the fuel
tank as a crucial aspect for locomotion, rather than as
a diagnostic indicator for the action of the internal
combustion engine. Fortunately, the dynamics of
Basic Facts and Definitions cyclogenesis can be concisely described in terms of a
Cyclones can be categorized with regard to the thermal single quantity that implicitly incorporates all others:
structure of the atmosphere near their centers. Warm- the potential vorticity. Before considering the dynam-
core cyclones are strongest near the Earth’s surface ics of cyclones, it will prove useful to document the
(hereafter, surface) and weaken with height. Cold-core evolving structure of a typical cyclone.
cyclones are strongest near the tropopause and weak-
en toward lower height. Extratropical cyclones can be
viewed as an amalgam of warm- and cold-core Structure of Developing Extratropical
cyclones, with a warm cyclonic circulation at the
surface and a cold cyclonic circulation near the
Cyclones
tropopause. These circulations are not aligned verti- Cyclones originate on zones of horizontal temperature
cally, but are displaced laterally by hundreds of contrast that are located in the vicinity of the extra-
kilometers in developing cyclones. This configuration tropical jet streams (Figure 1A); occasionally, the
allows the disturbance to extract energy from the jet temperature contrast is concentrated in a narrow
stream. frontal zone. The nascent cyclone appears as a region
Since the atmosphere near cyclones is in, to very of low pressure downwind (following upper-level
good approximation, hydrostatic balance (vertical winds) from a preexisting upper-level disturbance
pressure-gradient forces balance gravitational forces), called a short-wave trough. This horizontal displace-
cyclogenesis implies that there is a net loss of mass ment between the surface cyclone and the short-wave
from an imaginary column of air over the surface trough is required for the disturbance to extract energy
cyclone. This loss of mass accounts for the drop in from the jet stream. Viewed on a level surface, the
CYCLOGENESIS 591
marks the location of surface warm and cold fronts,

respectively (Figure 1B). Warm, moist air flowing
poleward rises at the warm front up and over the
colder, denser, air at low levels, producing a wide-
spread region of precipitation. Cold, dense air associ-
ated with the cold front displaces warmer air at the
cold front, producing a narrow band of precipitation.
Eventually, the cyclone separates from the surface
warm front and migrates toward the cold air forming
an occluded front. The deepening rate of the cyclone
slows during this time, and eventually ceases when the
surface cyclone and upper-level trough are nearly
coincident in the vertical (Figure 1C).Viewed from an
energy perspective, cyclogenesis represents a conver-
sion of potential energy associated with horizontal
temperature contrasts into kinetic energy. Cyclogen-
esis is a thermally direct circulation with warm air
rising as it flows poleward, and cold air sinking as it
flows Equatorward. A net result of this process is a
reduction in the pole-Equator temperature contrast
and more-stable density stratification (dense air un-
derlying less-dense air). A typical duration for the
cyclogenesis process depicted in Figure 1 ranges from
about 12 to about 60 h.
Figure 1 Schematic illustration of a developing extratropical
cyclone. Heavy solid lines give contours of 500 hPa geopotential
height; thin solid lines give contours of lOOOhPa geopotential Dynamics of Cyclone Development
height; dashed lines give contours of 500-1000 hPa thickness,
Cyclogenesis is associated with the development of
which is proportionalto the mean temperature in the500-1000 hPa
layer. Fronts are denoted by heavy solid lines with filled half-circles disturbances in the fields of wind, temperature, pres-
and triangles. (A) shows the early stages of development, sure, and density. Deriving a comprehensive dynam-
(B)shows the mature stage, and (C) shows the occluded stage. ical understanding of cyclogenesis in terms of the
(Reproduced with permission from Palmen E and Newton CW temporal variation of each of these quantities is
(1969) Atmospheric Circulation Systems: Their Structure and
laborious and redundant, since these quantities are
Physical Interpretation. New York: Academic Press.)
dynamically related. This difficulty is overcome
through two simplifications that engender concise
and deep insight: approximating the dynamical equa-
upper-level disturbance appears as a trough of low tions (using the quasigeostrophic approximations)
pressure or, equivalently (as shown in Figure l),as a and casting the resulting equations in terms of a single
trough of low geopotential height (or, simply, height) variable, the potential vorticity (PV).
on a constant pressure surface. These features
are referred to as short-wave troughs because their Quasigeostrophic Dynamics
1000-2000 km horizontal length scale compares with The observations that cyclones are nearly in geo-
greater than 5000 km for so-called long-wave troughs. strophic and hydrostatic balance motivates adopting
The short-wave trough is actually a manifestation of a quasi-geostrophic (QG) dynamics, which represent a
downward displacement of the tropopause, with large simplification of the actual governing equations. The
values of stratospheric PVextending to lower altitude. important result for our purposes concerns the fact
(PVis typically much smaller in the troposphere due to that the entire QG system can be formulated in terms
less-stable density stratification.) A region of large of PV, assuming adiabatic and frictionless conditions:
wind speed often can be found near the trough when
the flow is not strongly curved - this feature is called a DgQ= 0
jet streak, since it represents a local region of strong
winds within the jet stream. where D , is the time rate of change following the
As the cyclone continues to deepen, warm air geostrophic motion. Q is the PV, which in the QG case
advances poleward and cold air advances Equator- is the sum of geostrophic vorticity and a measure of
ward - the leading edge of these advancing air masses static stability, except at the surface where Q is
592 CYCLOGENESIS
proportional to temperature. Cyclonic Q has the same A Potential Vorticity Schematic of Cyclogenesis
sign as the local vertical component of planetary
rotation - positive in the Northern Hemisphere and Figure 2 gives idealized schematics of the dynamics of
cyclogenesis in vertical cross-section and plan view. In
negative in the Southern. At the surface, cyclonic Q is
associated with relatively warm air. the vertical cross-section view, a depression of the
tropopause gives a local blob of cyclonic PV on level
Note that eqn [l]is a simplified version of a more
surfaces (Figure 2A; note the plus sign). This PV
general relation that applies to the unapproximated
disturbance is associated with the 500 hPa short-wave
governing equations. Equation [l]indicates that PV is
trough noted in Figure 1. There is low pressure near the
conserved (i.e., unchanging) following the geostrophic
PV disturbance and a region of cold air below it in
motion, and that QG dynamics are simply described
accordance with the simple example given previously.
by the rearrangement of PV. From this perspective, all
(Note that the upward bulge in the isentropic surface is
other quantities (pressure, temperature, etc.) can be
due to the fact that potential temperature increases
viewed as subservient to PV in that, it should be
with height in a stably stratified atmosphere.) Low
emphasized, they evolve so as to conserve PV. Since Q
pressure at the surface is located near relatively warm
cannot be measured directly, it may seem less gratify-
air downwind of the upper-level low pressure
ing than quantities such as pressure and temperature.
(cf. Figure 1).In contrast, the temperature disturbanc-
However, one may imagine that a discovery of ‘PV
es tilt downwind with height; this configuration is
meters’ prior to barometers and thermometers would
necessary to achieve a conversion of potential to
perhaps have rendered pressure and temperature as
kinetic energy.
foreign and contrived quantities!
A second important fact about PV is that all other
+
variables can be recovered from it through a process
called inversion:
rE.f W
Stratosphere
where cg is the geostrophic wind, w the vertical

component of the wind, P pressure, and T tempera-
H L W
ture. Inversion I can be performed analytically for
simple distributions of Q, and otherwise can be
performed numerically; the details are unimportant
here. A useful illustration of inversion applies to a
small sphere of cyclonic &, which can be viewed as an
elemental building block from which arbitrarily com-
plex flow patterns can be constructed. Application of I
to this small sphere shows that relatively low pressure
is found in and around the sphere, a cyclonic
geostrophic wind field is found around the sphere,
and relatively warm (cold) air is found above (below) Pole
the sphere. These patterns have largest magnitude near
L E
the sphere, and weaken with distance in all directions.
The fact that the fields of wind, temperature, etc.,
extend far from the small sphere is fundamental (6)
to understanding how atmospheric disturbances Figure 2 Schematic illustration of potential vorticity concepts
interact. applied to cyclogenesis. (A) East-west vertical section showing the
More complicated flow patterns can be recovered tropopause (thick solid line), the surface thick cross-hatched line),
a lower-tropospheric isentropic surface (dashed line), and the air
from this simple example by summing the contribu-
circulations required to conserve potential vorticity for the typical
tions from PV in all locations. This ‘superposition case where westerly winds are increasing from the surface to the
principle’ allows one to rigorously decompose the tropopause. Regions of relatively warm and cold air are given by W
atmosphere into essential components, and to diag- and C, respectively, and regions of relatively low and high pressure
nose dynamical interactions among the components. are given by L and H, respectively. (E) Horizontal plan view
For example, the following section shows how the showing the surface isotherms (solid lines), surface wind associ-
ated with the upper-level PV disturbance (dashed lines with
dynamics of cyclogenesis are given by an interaction arrows), and surface wind associated with the surface cyclone
between the tropopause PV disturbance and the (solid lines with arrows). In both (A) and (E), the plus sign shows the
surface cyclone. location of the PV disturbance due to the lowered tropopause.
CYCLOGENESIS 593
The arrows in Figure 2 show the air circulations that grow exponentially in time. The most rapidly
necessary to conserve PV for the typical case where growing disturbances in this theory are waves in the
westerly winds are increasing from the surface to the horizontal direction, and they have maximum ampli-
tropopause. Air converges into the cyclone near the tude at the tropopause and surface. For typical
surface and then rises toward the tropopause where it extratropical conditions, these disturbances have
diverges. Conversely, on the upwind side of the PV horizontal wavelengths around 4000 km, propagate
disturbance, air converges near the tropopause and at about l S m s - l , and double their amplitude
then sinks toward the surface where it diverges near over slightly less than one day. These characteristics
the surface anticyclone. Note that a reverse circulation are qualitatively in accord with observations,
exists in the lower stratosphere, and is only hinted at in although the growth rate may appear to be too slow
Figure 2 . Downwind of the PV disturbance, these to account for the development of observed cyclones
circulations raise (lower)the isentropic surfaces below from infinitesimal perturbations over a 4 8 h time
(above) the level of the disturbance in order to period.
conserve PV as the westerly winds transport the PV
blob eastward. Objections to Baroclinic Instability Theory
In plan view, the horizontal wind field due to the
When surface friction is introduced to the baroclinic
upper-level PV disturbance extends downward to the
instability theory, the already small disturbance
surface and contributes to the poleward (Equator-
growth rates become even smaller. In fact, some
ward) transport of warm (cold) air near the surface
research suggests that the jet streams are stable
cyclone (anticyclone) (Figure 2B; note that this illus-
to baroclinic instability; in other words, the growing
tration applies to the Northern Hemisphere). These
disturbances do no exist. Another objection points
transport patterns amplify the surface cyclone by
to the fact that observed cyclogenesis events proceed
warming the center of the cyclone (or, an increase in
from preexisting large-amplitude disturbances
surface PV, since relatively warm air at the surface
on the tropopause, not infinitesimal noise. More-
represents cyclonic PV). In contrast, the horizontal
over, cyclogenesis events tend to be highly localized in
wind field due to the surface cyclone produces
the horizontal, not plane waves as in baroclinic
temperature transport patterns that are out of phase
instability theory, and the vertical structure of
with respect to the surface cyclone. These transport
observed cyclones tends to change with time as
patterns are responsible for propagating the distur-
compared with the fixed structure of an unstable
bance toward the east. Similar interpretations apply
disturbance.
near the tropopause, assuming that the PV increases
A second theory for cyclogenesis attempts to
horizontally toward the pole, as is typical. Near the
account for the noted deficiencies by employing
tropopause, winds attributable to the surface cyclone
disturbances having structures and growth rates that
amplify the upper-level PV disturbance, and winds
change in time. Although these disturbances decay
attributable to the upper-level PV disturbance propa-
over long time intervals, they can produce very large
gate the disturbance toward the west, relative to the
growth over short time intervals. The theory for
westerly winds.
optimal perturbations formalizes the search for dis-
Finally, we note that diabatic heating can be
turbances that have the largest growth over short time
accounted for within the PV framework. Diabatic
intervals, such as 48 h. An amazing result of this theory
heating contributes to the generation of the PV
is that even in cases where friction is sufficiently strong
disturbances (non-conservation of PV, or a source
to stabilize a particular jet stream to baroclinic
term on the right-hand side of eqn [l]), with a cyclonic
instability, optimal perturbations may still exhibit
PV disturbance below, and an anticyclonic PV distur-
large growth over the time interval characteristic of
bance above, a region of maximum warming. Since the
observed cyclogenesis. Although the rapid growth and
cyclonic PV disturbance below the heating is in close
time-varying structure of optimal perturbations make
proximity to the surface, it contributes to a reduction
them appealing candidates to explain observed cyclo-
in the cyclone surface pressure and enhances the winds
genesis, there is at present no evidence to support such
at the cyclone center.
claims.
Reconciliation of Baroclinic Instability Theory with

Theory of Cyclogenesis Observations
The leading theory of cyclogenesis, baroclinic insta- Careful examination of observations and the stability
bility, posits that the westerly jet streams in the properties of observed flows together indicate that
extratropics are unstable to infinitesimal disturbances baroclinic instability theory can in fact account for
594 CYCLONES, EXTRA TROPICAL
observed cyclogenesis, once two factors are taken into Another energy source contributing to cyclogenesis
consideration. First, observed precursor disturbances involves a transfer of existing energy from upwind
tend to be localized in the horizontal direction, which disturbances. Since the disturbances propagating
means that a sum of disturbances, rather than a single along the jet stream are strongly dispersive Rossby
disturbance, should be considered; the sum allows for waves (phase speeds vary with wavelength, so that a
a localized disturbance through cancelation of the group of waves with similar wavelength does not move
components away from the local maximum. Second, at the phase speed), the energy in these disturbances
these localized disturbances have large, not infinites- can radiate away from existing disturbances. This
imal, amplitude (cf. Figures 1 and 2).An initial distur- energy radiation can contribute to a sequence of
bance of modest initial amplitude comprising an cyclogenesis events through a process known as
unstable component can grow over a few doubling downstream development.
times to account for the amplitude of observed
disturbances.
It is clear, however, that other factors not empha- See also
sized here play a contributing role in cyclogenesis; for
Baroclinic Instability. Cyclones, Extra Tropical. Dy-
some cyclones this role may be essential. For example, namic Meteorology: Balanced Flows; Overview. Jet
diabatic heating appears to contribute positively to Streaks. Quasi-geostrophicTheory. Synoptic Mete-
cyclogenesis in many cases. Another potentially im- orology: Forecasting; Weather Maps. Tropopause.
portant process involves the transfer of heat and Vorticity.
moisture from the sea surface during oceanic cyclo-
genesis. The most intense and most rapidly developing
cyclones are found over the western North Atlantic Further Reading
and North Pacific oceans, where strong ocean currents
Bluestein HB (1992) Synoptic-Dynamic Meteorology
maintain regions of large horizontal temperature in Midlatitudes, Vol. II: Observations and Theory
contrast in the lower troposphere. Another important of Weather Systems. New York: Oxford University
effect in these locations is due to cold continental air Press.
flowing over relatively warm water, which results in Carlson TN (1991) Mid-Latitude Weather Systems. New
weak static stability. Weak static stability reduces the York: HarperCollins Academic.
doubling time for baroclinic instability and may allow Holton JR (1992) An Introduction to Dynamic Meteorol-
greater latent heat release through stronger upward air ogy. New York: Academic Press.
motion and precipitation. Orographic barriers are also Newton CW and Holopainen EO (eds) (1990) Extra
important in initiating or enhancing cyclogenesis, Tropical Cyclones: The Erik Palme'n Memorial Volume.
owing to the strong adiabatic warming that occurs in Boston, MA: American Meteorological Society.
the lee when air flows over the mountains. Cyclones Shapiro MA and Grgnds S (eds) (1999) T h e Life Cycles of
often form in the lee of major mountain barriers and Extratropical Cyclones. Boston, MA: American Meteor-
then move downwind. ological Society.
See HURRICANES; POLAR LOWS; THERMAL LOW
A Joly, F Ayrault and S Malardel, Meteo-France, Introduction

Centre National de Recherches Meteorologiques,
Toulouse cedex, France One or two days of mild air temperature and the next
day the air feels cold, even chilly; dark, low clouds
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. persist for an entire day together with continuous
disturbances.

CYCLONES, EXTRA TROPICAL 595
~~~ ~
rainfall or snowfall and then a spell of interwoven Several other aspects of that same extratropical
showery and sunny hours follows; a day of calm or cyclone will be presented. Since extratropical cyclones
breezy air, and the next day long hours of howling are ubiquitous in mid-latitude meteorology, they have
winds; storms: mid- and high-latitude weather is been studied for a long time. Several models and
characterized by these contrasts that take place on paradigms, some of them very well known, have been
a time scale of the order of one day or few days. Most proposed to describe or explain these systems; the
of these changes in the temperature, wind, and sky most recent ones will be reviewed. An up-to-date
are the effects of extratropical cyclones. They occur at climatological description of North Atlantic cyclones
all seasons, but are most frequent in autumn will provide the current observational picture, com-
and winter. They are the main purveyors of water, so pleting the single example description and under-
that agriculture and many human activities would not pining the strengths but also the shortcomings of
be possible without them. At the same time, some of existing conceptualizations. A summary of the main
them are deadly, destroying trees and damaging physical mechanism of cyclone development will be
buildings. given. All these descriptions rest on looking at
Extratropical cyclones look like large cloud-covered characteristic, coherent shapes in fields such as those
vortices when seen from space. Figure 1 is an example of temperature or wind. It is also possible to picture
of extratropical cyclone followed over a period of extratropical cyclones in terms of air streams flowing
two days. A series of surface pressure maps overlaid into or out of them. This overview will be concluded
on infrared composite satellite images introduce by looking at the larger-scale picture and discussing
the most basic features of such a weather system. facts and ideas relating to the role of extratropical
The characteristic horizontal scale is of the order cyclones in the general circulation.
of 2000 km and the system travels across the North
Atlantic ocean. While the pressure at the core deepens,
the clouds expand, until the system appears no longer
Example of Cyclone Life Cycle
to change much and spins: extratropical cyclones go The example introduced with Figure 1 is one well-
through a life cycle, appearing, growing, and later observed case. After briefly indicating how this has
dying. been done, further aspects of its features are presented.
Figure 1 Example of mid-latitude cyclone over the North-Atlantic ocean, in February 1997, shown every 24h. The images are
composite infrared pictures from NOAA GOES and EUMETSAT METEOSAT geostationary satellites. The fields are mean-sea-level
pressure from a 4D-VAR reanalysis of all the available data, drawn every 5 hpa. Location marks of the trajectory are placed every 12 h.
Observing Extratropical Cyclones humidity during their fall. Ships sent similar sondes,
but upward, tied to balloons. The mature cloud system
One way to properly describe extratropical cyclones is
of the cyclones was also sampled with sophisticated
to gather scientists, instruments, and platforms within
airborne Doppler radars on the eastern side of the
an international project. There were several such
ocean.
initiatives in the 1980s and 1990s. One of them is
The data, in particular all the profiles (about 9000
called the Fronts and Atlantic Storm-Track Experi-
for the two months), have been checked, sometimes
ment. All countries around the North Atlantic basin
corrected, and put into an Internet-accessible data
have been involved, some of them quite heavily:
base open to all. In order to combine together these
Canada, France, Iceland, Ireland, the United King-
profiles and other sources into coherent fields, these
dom, the United States, and others. Its observational
data have been assimilated, the last and best reanalysis
phase took place in January and February 1997. The
having been performed at Mttto-France using the
unique feature of this particular example was that its
sophisticated four-dimensional variational scheme it
observational grounds were the whole of the North
has developed with ECMWF. This scheme associates
Atlantic ocean because its main experimental objec-
data with various realistic dynamical constraints,
tive was to track a number of cyclones throughout
taking into account uncertainties associated to each
their life cycles.
source.
Radiosoundings were reinforced on the lands
around the oceanic basin. Ships held positions along
Low 30 from FASTEX
the 30"W meridian, moving north and south to
remain in the vicinity of the Atlantic jet stream Figure 1 shows that, as the pressure minimum deepens,
(otherwise, it can be shown that their data are somewhat unorganized cloud bands off the east coast
practically useless). Aircraft were based on both sides are shaped into a large and deep cloudy area on
of the ocean. About 10 cases have been successfully the fore side, called the cloud head. The low is trailed
tracked at several key times of their evolution. by a more bandlike cloud feature that is read-
The example already glimpsed in Figure 1 is Low 30, ily identified as a cold front. The characteristic scale
a cyclone that formed on 4 February off of the cloud head is about 1500-2000km. This
New York and retired three days later off the North particular system becomes quite meridionally
Cape, Norway, 5000km away (Figure 2). The elongated as it moves north-eastward across the
observational strategy of FASTEX was to operate ocean. An interesting feature is the late development
observing platforms in sequence, along the track of the of a secondary spiral of clouds just south of the main
cyclone. The most frequently used instruments were system.
dropsondes thrown out of aircraft and measuring The low level thermal field shows a well-defined
horizontal wind, pressure, temperature, and relative preexisting area of transition from warm to cold
Figure 2 A schematic of the deployment of observational platforms to follow the life cycle of Low 30, shown in Figure 1, during the
FASTEX project. The balloons symbols mark the location of radiosounding stations. The darker they are colored, the most frequent were
the soundings, from 12-hourly to 3-hourly. The ships are also intensive radiosounding locations. In addition to the cyclone tracks, some of
the flight tracks are also shown.
Figure 3 The low-level (about 1.5 km above the surface) structure and environment of the mid-latitude cyclone shown in Figure 1. The
fields are moist potential temperature at 850 hpa (thin brick red lines, contour interval 2 K) and relativevorticity (heavy purple solid lines,
contour interval 5 x 10-5s-’ above 5 x 10-5s-’).The system of interest is highlighted by shading of the corresponding vorticity
maximum.
already orientated along a north-east axis (Figure 3). It is important to pay attention to some upper-
This is the signature of what is here called a baroclinic level fields as well (Figure 5 ) . The wind has been
zone. This transition is clearly reinforced in the plotted only above a threshold value of 40 m s - ’;
vicinity of the cyclone center during roughly the first this provides a good visualization of the jet stream.
half of the evolution. Fronts correspond to these Its close association with the background thermal
limited zones of stronger changes of temperature. change shows that the jet stream is another key feature
Significant wind changes also take place in those same
areas. The vorticity field is one way to bring out these
changes. Vorticity is a measure of the rotation or spin
due to the wind.
The pressure field is the oldest way to spot a cyclone
- indeed it is sometimes called a depression - but,
as can be seen in Figure 1, the pressure field also shows
a large quasi-permanent south-north change.
This strong background organization masks the
cyclone’s own signature. The vorticity field, on
the other hand, allows a measure of the intensity
of a cyclone that does not have this problem. Vorticity
is a property of the wind that measures the local spin
it induces: cyclones are well-defined maxima of
relative vorticity. Plotting the time-series of low-
level vorticity in the core of Low 30, it appears
that it has been through two well-separated growth Figure 4 Evolution of the maximum vorticity [ at 850 hpa at the
phases (Figure 4). This behavior has been observed core of the extratropical cyclone shown in Figures 1, 3, and 5. It
provides a better measure of the cyclone’s own amplitude than the
in about half the cases observed in FASTEX, which minimum central pressure. (Adapted from Baehr Ch, Pouponneau
did not use a random selection but chose active B, Ayrault F, Joly A (1999)Quarterly Journal of the Royal
cases. Meteorological Society 125:3469-3494.
Figure 5 The tropopause level structure and environment of the mid-latitude cyclone shown in Figures 1 and 3. The fields are the
geopotential height of the constant potential vorticity surface P = 1.5 pvu (solid lines, contour interval 1000 mgp) and the wind vector
’,
plotted with the standard meteorological conventionswhen thevelocity is largerthan 40 m s- The system of interest is highlighted by the
shaded low level vorticity maximum of Figure 3.
of a baroclinic zone. The wind field is shown at a framework moving with the low-level cyclone, as
the tropopause defined with potential vorticity; shown by Figure 7.
since the latter is a conserved -property,
- . it is a kind
of material surface. But this surface has a complex
shape, as shown by its geopotential height. An upper- Recent Past Conceptualizations
level signature accompanying the cyclone in the of Extratropical Cyclones
first half of its life cycle is easy to see. It is a The evolution of the understanding of extratropical
well-defined upper-level trough that also amplifies cyclones is tied to the progress of two areas: better
but moves at its own phase speed, slower than observations on the one hand, and on the other
that of the cyclone. A key thing to note on this the introduction into meteorology of new mathe-
case is that this upper-level feature definitely pre- matical tools for physics, allowing assumptions
exists. and their consequences to be quantified and checked.
Indeed, on 3 February at 1200 UTC, it is above the From that point of view, early attempts using concepts
Great-Lakes, but without any significant signature like the thermal cyclone or the polar front cyclone,
either in the satellite image or the surface pressure field although seminal in some respects, have not been very
(Figure 6 ) . It is when this feature closes on the successful.
acceleration area of the jet stream that Low 30 takes
shape. The interaction between the upper-level trough,
Upper-Level Induced Development Model
the baroclinic zone, and the developing cyclone fits
well with the first growth phase. The cyclone then The emergence of operational meteorology accompa-
seems to escape away from the upper-level trough nying the development of aviation led to better-
during the decreasing phase. The final growth stage coordinated observational networks. The surface
coincides with the cyclone crossing the baroclinic zone measurements benefited from the first ships able to
from the warm to the cold side, without a clear hold a fixed location in the ocean. Most importantly,
independent upper-level vortex-like feature being upper-air observations over continents became more
involved. The various phase relations are best seen in frequent; they used sondes attached to balloons.
Figure 6 The situation preceding the development of the extratropical cyclone shown in Figures 1,3 and 4, showing the upper-level
precursor and its environment. Although it is well organized at the tropopause, there are few associated features elsewhere. The fields in
the top panel are as in Figure 1, in the middle panel as in Figure 3, and in the bottom panel as in Figure 5.
Figure 7 Several fields showing the extratropical cyclone of Figures 1 , 3 ,and 4 and its environment in aframe centered on the low-level
core of the cyclone and rotating so that the horizontal axis remains roughly parallel to the mean isotherms. The fields are the 850 hpa
relative vorticity (heavy lines, same contouring as in Figure3, system of interest shaded); the tropopause level vorticity (light lines, same
contouring); and the tropopause wind field above 40ms-’.
Radiosondes transmitting their data while ascending minority among those attempting to understand
were introduced, removing the need to recover the cyclones.
recorded measurements after the balloon burst and
the sonde fell in some unlikely place. In 1935, a Upper-Level Jet, Baroclinic Zone, and Baroclinic
‘coordinated ascents’ experiment on the scale of Instability
western and northern Europe took place. For the first The school of Sutcliffe, Kleinschmidt, Petterssen, and
time, a few mature cyclones were sampled in three others was in a minority because an elegant mathe-
dimensions.
matical framework was set up in parallel and attracted
The upper-level structure of the flow about extra-
most of the attention. Discarding the unrealistic
tropical cyclones was then quite clearly obtained. The
extreme front model, Charney in California and
wind increased with height much more rapidly near
Eady in England applied, in the early 1940s, the linear
the cyclone center than away from it. A very clear
stability technique to the study of the properties of a
vertical tilt with height of the minimum pressure
simple broad baroclinic zone. This is an area of
location was definitely shown to exist. It took some
continuous and moderate transition from warm to
time to make something of the wind distribution:
cold air, which, as a consequence of rotation, is
ultimately, and with much further observations, it was
accompanied by a zonal wind increasing with
recognized to be organized in meridionally localized
height - an idealized form of the upper-level jet
streams of air, called jet streams. But the importance of
stream. This basic or initial flow indeed fits the
the vertical tilt and the related upper-level mass
observed average flow at mid-latitudes much more
distribution was grasped very quickly, partly because
reasonably than the extreme front model. This
it had been anticipated both with observations and
problem can be idealized to a point that allows
theoretically on several occasions before.
analytical solutions to be derived that still share
Several approaches have since been developed. One
some properties with real cyclones. Most notably,
is called ‘development’. This had been outlined shortly
the vertical tilt with height is part of the most unstable
before the 1940s by the British scientist Sutcliffe, who
solution, together with reasonable spatial scales. This
proposed that the emergence and amplification of a
is the baroclinic instability theory of cyclogenesis.
surface cyclone resulted from the influence of a
These similarities made linear studies the preferred
preexisting upper-level trough in a broad baroclinic
path to explaining natural phenomenon like the
zone (rather than an extreme front). This view is
growth of cyclones. Most textbooks on dynamical
supported by simplifying fundamental equations,
meteorology depend heavily on linearized equations
rather than linearizing them, and concentrating, in
and stability problems. Very elegant general existence
particular, on vorticity dynamics. In problems where
theorems have been derived.
rotation is important, looking at vorticity is a good
Because it is so elegant and appealing to mathema-
tactic.
tically oriented minds, the instability model domi-
Along this line of thought, the German meteorolo-
nated the theoretical scene throughout the 1950s
gist Kleinschmidt put some flesh on the concept of the
and 1960s. But because part of its elegance comes
upper-level component of an extratropical cyclone
from mathematical tractability, only extremely idea-
and its relation to vorticity. Using the homogeneous
lized problems can be studied and the relationship
and extended network of radiosoundings available
with observations is the weak spot of all of this work.
over Europe in 1943, he built a striking three-
As a consequence (or measure) of this, the results have
dimensional picture of this structure. Furthermore,
had little impact on practical forecasting aside from
he gave the description in terms of a quantity akin to
some guidelines for interpreting numerical model
vorticity and called potential vorticity. He particularly
output.
undertook to show that, under certain assumptions, it
is possible t o derive the wind and temperature
Quasi-Geostrophic Model and Theoretical Results
distributions from that of potential vorticity, a process
called inversion. In the early version of the instability approach, subtle
Sutcliffe and Kleinschmidt are, together with the simplifications were introduced at key points in the
Norwegian Petterssen and a few others, representa- derivation, limiting the number of solutions. This
tives of a school of thought that made the most practice is in fact very powerful as far as gaining an
of available observations to conclude that cyclones understanding is concerned. It was soon reconsidered
are largely caused by finite-amplitude interactions of more formally by Charney in terms of a scale analysis
some pre-existing ‘structures’: vertically confined of the basic equations; this is called filtering on the
vortices, baroclinic zones or jet streams, fronts, and basis of a balance assumption. A simplified set of
the like. However, until recently, they represented a equations specifically oriented toward understanding
extratropical cyclones and their large-scale environ- described entirely numerically and of known
ment can be derived at the outset. This is the quasi- representativeness.
geostrophic model. These equations represent
cyclones and some of their behavior, but they exclude, Data and Processing, Definition of Life Cycle Phases
for example, gravity waves. They include well-defined
cause-and-effect relationships. In this framework, for The data cover the period 1979-1993 and are an
example, vertical motion is the computable effect of extraction of the winter season (15 October-15 April)
the interaction between the thermal and horizontal over a domain centered over the North Atantic, with a
wind fields. resolution of 1.5” and over 1 3 pressure levels.
It is important to realize that this model is the Events of interest are defined using vorticity
common ground of both the intrinsic instability rather than pressure, for reasons already given. The
approach and the finite-amplitude interactions trajectories of 850 hpa vorticity maxima have
approach. With hindsight, it is easy to show that been constructed using an automatic tracking
Sutcliffe’s calculations are based on an early form of algorithm. In order to distribute events into classes,
the quasi-geostrophic filtering and so are Eady’s. the large state vectors that result from extracting fields
Potential vorticity inversion, as outlined by Kleinsch- in boxes of size of about 2500 km centered on the low-
midt, is a well-posed classical Poisson problem in this level vorticity maxima along the trajectories are
framework. The current theoretical understanding of reduced by a principal component analysis. The
cyclones undoubtedly rests on the quasi-geostrophic remaining components are distributed into classes
framework. using ascending hierarchical classification. Paradigms
of situations conducive to cyclogenesis are then
Recent Observational Description of Cyclones
obtained by averaging the original fields within each
class.
The explosion of observations that characterized Study of the distributions of amplitudes, of dura-
the last quarter of the twentieth century quickly tions, and of the various growth phases discourages
showed that the idea of a single type of cyclone was the automatic classification of full life cycles. Two
an idealization. Perhaps it was necessary in a period particular stages have been isolated. One is the 12h
when measurements were scarce. Although most period centered on time 0 h of each case, extrapolating
cyclones form and evolve in baroclinic regions a location for the field extraction at - 6 h: this phase is
accompanying the jet stream, some do not. Even in called the initiation of a new cyclone. The other is the
the baroclinic family, there is an extraordinary 24 h period preceding the time of maximum ampli-
diversity of shapes and histories, and therefore tude: this is called the maturation phase and generally
perhaps of mechanisms. includes the period of maximum vorticity growth
At the moment, knowledge of cyclones results from undergone by each case.
two main sources, both combining numerous observa-
tions in all dimensions. A number of dedicated projects Summary of Composite Genesis Situations
such as FASTEX have been set up to concentrate The classification of initiation has 1 2 classes.
observational platforms so as to document specific This classification uses nearly 6000 cases. As it is
cases in a very detailed way, such as the case presented impossible to fully account for all the classes here,
above. Beside these cases, the work on the possible composites are gathered into ‘families’, of which there
evolution of climate has led to reassimilation of rather are five. The best previous studies dealt with two dozen
long series of global data, providing homogeneous, cases.
consistent datasets from which thousands of cases can The most frequent family corresponds to the genesis
be extracted and studied. of a cyclone on the northern side of the baroclinic
zone, in the cold air (28% of cases). Nearly as
Types of North Atlantic Extratropical frequent, there is a family that comes close to the
extensively studied frontal instability problem (24%,
Cyclones
Figure 8).It departs from the idealized situation on one
Looking for a more generic description of ‘typical’ critical aspect: The flow is three-dimensional and the
extratropical cyclones, one is led to take advantage of new cyclone forms at the end of the frontal zone,
the former set of fields obtained by reprocessing years below a jet entrance. Neither of these families seems to
of archived data into a given version of the data feature a precursor of the cyclogenesis. The next two,
assimilation scheme. One such project is the ECMWF on the other hand, do have such a preexisting structure
Re-Analysis (ERA).It is used here to build composites of cyclone scale: the class that corresponds to the
representative of hundreds of individual cases, well-known (but clearly overestimated) development
Figure 8 One of the initiation class W composite at times (from left to right) - 6 h, 0 h (first detection time of an individual vorticity
maximum) and +6 h, at 300 hpa (top) and at 850 hpa (bottom). The fields on the top row are wind vectors (knots) coloured according to
speed so as to highlight the jet-flow structure (darkest blue above 40 m s - ’ ) and relative vorticity i,(light purple lines, contour interval
2 x 1 0 - 5 s - ’ , negative contours dashed). The green-shaded background field is the 600 hpa relative humidity (contour interval lo%,
white for values largerthan 90%, colordarkerfor smallervalues). Bottom fields are relativevorticity (as at the top, with darker contours) and
equivalent potential temperature 0, (brick-red thin contours and shading, interval 3 K, limit between blue-green shade and light orange
300K). The 850 hpa wind is also shown as above, the darkest blue threshold is changed to 10ms-’. The boxed arrow is a compass
indicating the average direction of the geographical north, since the area shown is rotated. (Adapted from Ayrault F, Joly A (2000),
Comptes-rendus de I’ Academie des Sciences, Paris, Sciences de la Terre et des planetes 330: 173-1 78.)
envisioned by Sutcliffe and Petterssen, and a new Summary of Maturation Situations

family in which the precursor is a full-scale cyclone
that splits into the old and a new system. A single The maturation phase composites show seven classes.
picture summary of each family is shown in Figure 9. Figure 10 shows some of the morphological character-
Figure 9 A single-time summary of three of the other families of initiation composites. The structures are shown at the time of the first
detection of an individual vorticity maximum, that is, the appearance of a new cyclone. From left to right: a class of cold air cyclogenesis;
cyclogenesis induced by a preexisting upper-level precursor; and a class of cyclogenesis resulting from the apparent aft-splitting of an
older cyclone. The fields are displayed as in Figure 8. (Adapted from Ayrault F, Joly A (2000) Comptes-rendus de I’ Academie des
Sciences, Paris, Sciences de la Terre et des planetes 330: 173-1 78.)
Next Page CYCLONES, EXTRA TROPICAL 603
Figure 10 Example of a class of mature extratropical cyclone, the most intense one, the meteorological ‘bomb’ (Class A). The
composite structure is shown at times - 24 h, - 12 h, and at the time of maximum amplitude (from left to right), at 300 hpa (top) and at
850 hpa (bottom).Thefieldsareas in Figure8exceptforthewindshownabove50 knots. (From Ayrault F, JolyA(2000), Comptes-rendus
de I’ Academie des Sciences, Paris, Sciences de la Terre et des planetes, 330: 167-1 72.)
istics of Type A, the one that undergoes the largest The method employed so far yields two sets of
deepening (-25 hpa/24 h) and happens in 12% of composites. A fundamental result comes from the
cases: It is what some authors have termed the clear separation between these two sets: The subse-
meteorological ‘bomb’. Centered on the low-level quent amplification of a cyclone does not necessarily
component of the system, the figure clearly depicts the involve the same processes as its genesis. Indeed, while
phase change between this and the upper-level
component: As there are independent studies of these
strong cases, this class validates the approach.
Class Byalthough not deepening, remains in warm
air and yields significant precipitations. An interesting
characteristic of this type is that it occurs in rather
strong baroclinic zones, like Class A and with the same
frequency. Yet it does not develop strongly (Figure 11).
The magnitude of the baroclinicity does not explain
the two different behaviors, and this calls into question
explanations in terms of simple instabilities. Class B
does not have the upstream upper-level signature that
Class A shows well.
The other types correspond to cyclones deepening
between 9 and 15 hpa in 24 h. They share with Type A
the presence of a well-defined upper-level signature.
One feature that discriminates these types is the shape
of the upper-level jet and the position of the cyclone
relative to it (Figure 12). Using time filtering to
separate the lows and their environment, it is possible
to estimate the energy transfers. The preferred energy
conversions are similar for all the developing classes
and correspond to the early twentieth century idea that Figure 1 1 The mean 24 h pressure deepening A P P 4 h of the
seven classes of maturation phases is plotted against the low-pass
cyclones develop by converting the energy contained
average baroclinity, the temperature gradient of their environment.
in gradients Of temperature into wind. In (From Ayrault F and JolyA (2000) Comptes-rendus de/’Acad&mie
that sense, the core physical mechanism for growth is des Sciences, Paris, Sciences de la Terre et des planetes 330:
the same for all these types. 167-1 72.)
disturbances.

~~~ ~
rainfall or snowfall and then a spell of interwoven Several other aspects of that same extratropical
showery and sunny hours follows; a day of calm or cyclone will be presented. Since extratropical cyclones
breezy air, and the next day long hours of howling are ubiquitous in mid-latitude meteorology, they have
winds; storms: mid- and high-latitude weather is been studied for a long time. Several models and
characterized by these contrasts that take place on paradigms, some of them very well known, have been
a time scale of the order of one day or few days. Most proposed to describe or explain these systems; the
of these changes in the temperature, wind, and sky most recent ones will be reviewed. An up-to-date
are the effects of extratropical cyclones. They occur at climatological description of North Atlantic cyclones
all seasons, but are most frequent in autumn will provide the current observational picture, com-
and winter. They are the main purveyors of water, so pleting the single example description and under-
that agriculture and many human activities would not pining the strengths but also the shortcomings of
be possible without them. At the same time, some of existing conceptualizations. A summary of the main
them are deadly, destroying trees and damaging physical mechanism of cyclone development will be
buildings. given. All these descriptions rest on looking at
Extratropical cyclones look like large cloud-covered characteristic, coherent shapes in fields such as those
vortices when seen from space. Figure 1 is an example of temperature or wind. It is also possible to picture
of extratropical cyclone followed over a period of extratropical cyclones in terms of air streams flowing
two days. A series of surface pressure maps overlaid into or out of them. This overview will be concluded
on infrared composite satellite images introduce by looking at the larger-scale picture and discussing
the most basic features of such a weather system. facts and ideas relating to the role of extratropical
The characteristic horizontal scale is of the order cyclones in the general circulation.
of 2000 km and the system travels across the North
Atlantic ocean. While the pressure at the core deepens,
the clouds expand, until the system appears no longer
Example of Cyclone Life Cycle
to change much and spins: extratropical cyclones go The example introduced with Figure 1 is one well-
through a life cycle, appearing, growing, and later observed case. After briefly indicating how this has
dying. been done, further aspects of its features are presented.
Figure 1 Example of mid-latitude cyclone over the North-Atlantic ocean, in February 1997, shown every 24h. The images are
composite infrared pictures from NOAA GOES and EUMETSAT METEOSAT geostationary satellites. The fields are mean-sea-level
pressure from a 4D-VAR reanalysis of all the available data, drawn every 5 hpa. Location marks of the trajectory are placed every 12 h.
Observing Extratropical Cyclones humidity during their fall. Ships sent similar sondes,
but upward, tied to balloons. The mature cloud system
One way to properly describe extratropical cyclones is
of the cyclones was also sampled with sophisticated
to gather scientists, instruments, and platforms within
airborne Doppler radars on the eastern side of the
an international project. There were several such
ocean.
initiatives in the 1980s and 1990s. One of them is
The data, in particular all the profiles (about 9000
called the Fronts and Atlantic Storm-Track Experi-
for the two months), have been checked, sometimes
ment. All countries around the North Atlantic basin
corrected, and put into an Internet-accessible data
have been involved, some of them quite heavily:
base open to all. In order to combine together these
Canada, France, Iceland, Ireland, the United King-
profiles and other sources into coherent fields, these
dom, the United States, and others. Its observational
data have been assimilated, the last and best reanalysis
phase took place in January and February 1997. The
having been performed at Mttto-France using the
unique feature of this particular example was that its
sophisticated four-dimensional variational scheme it
observational grounds were the whole of the North
has developed with ECMWF. This scheme associates
Atlantic ocean because its main experimental objec-
data with various realistic dynamical constraints,
tive was to track a number of cyclones throughout
taking into account uncertainties associated to each
their life cycles.
source.
Radiosoundings were reinforced on the lands
around the oceanic basin. Ships held positions along
Low 30 from FASTEX
the 30"W meridian, moving north and south to
remain in the vicinity of the Atlantic jet stream Figure 1 shows that, as the pressure minimum deepens,
(otherwise, it can be shown that their data are somewhat unorganized cloud bands off the east coast
practically useless). Aircraft were based on both sides are shaped into a large and deep cloudy area on
of the ocean. About 10 cases have been successfully the fore side, called the cloud head. The low is trailed
tracked at several key times of their evolution. by a more bandlike cloud feature that is read-
The example already glimpsed in Figure 1 is Low 30, ily identified as a cold front. The characteristic scale
a cyclone that formed on 4 February off of the cloud head is about 1500-2000km. This
New York and retired three days later off the North particular system becomes quite meridionally
Cape, Norway, 5000km away (Figure 2). The elongated as it moves north-eastward across the
observational strategy of FASTEX was to operate ocean. An interesting feature is the late development
observing platforms in sequence, along the track of the of a secondary spiral of clouds just south of the main
cyclone. The most frequently used instruments were system.
dropsondes thrown out of aircraft and measuring The low level thermal field shows a well-defined
horizontal wind, pressure, temperature, and relative preexisting area of transition from warm to cold
Figure 2 A schematic of the deployment of observational platforms to follow the life cycle of Low 30, shown in Figure 1, during the
FASTEX project. The balloons symbols mark the location of radiosounding stations. The darker they are colored, the most frequent were
the soundings, from 12-hourly to 3-hourly. The ships are also intensive radiosounding locations. In addition to the cyclone tracks, some of
the flight tracks are also shown.
Figure 3 The low-level (about 1.5 km above the surface) structure and environment of the mid-latitude cyclone shown in Figure 1. The
fields are moist potential temperature at 850 hpa (thin brick red lines, contour interval 2 K) and relativevorticity (heavy purple solid lines,
contour interval 5 x 10-5s-’ above 5 x 10-5s-’).The system of interest is highlighted by shading of the corresponding vorticity
maximum.
already orientated along a north-east axis (Figure 3). It is important to pay attention to some upper-
This is the signature of what is here called a baroclinic level fields as well (Figure 5 ) . The wind has been
zone. This transition is clearly reinforced in the plotted only above a threshold value of 40 m s - ’;
vicinity of the cyclone center during roughly the first this provides a good visualization of the jet stream.
half of the evolution. Fronts correspond to these Its close association with the background thermal
limited zones of stronger changes of temperature. change shows that the jet stream is another key feature
Significant wind changes also take place in those same
areas. The vorticity field is one way to bring out these
changes. Vorticity is a measure of the rotation or spin
due to the wind.
The pressure field is the oldest way to spot a cyclone
- indeed it is sometimes called a depression - but,
as can be seen in Figure 1, the pressure field also shows
a large quasi-permanent south-north change.
This strong background organization masks the
cyclone’s own signature. The vorticity field, on
the other hand, allows a measure of the intensity
of a cyclone that does not have this problem. Vorticity
is a property of the wind that measures the local spin
it induces: cyclones are well-defined maxima of
relative vorticity. Plotting the time-series of low-
level vorticity in the core of Low 30, it appears
that it has been through two well-separated growth Figure 4 Evolution of the maximum vorticity [ at 850 hpa at the
phases (Figure 4). This behavior has been observed core of the extratropical cyclone shown in Figures 1, 3, and 5. It
provides a better measure of the cyclone’s own amplitude than the
in about half the cases observed in FASTEX, which minimum central pressure. (Adapted from Baehr Ch, Pouponneau
did not use a random selection but chose active B, Ayrault F, Joly A (1999)Quarterly Journal of the Royal
cases. Meteorological Society 125:3469-3494.
Figure 5 The tropopause level structure and environment of the mid-latitude cyclone shown in Figures 1 and 3. The fields are the
geopotential height of the constant potential vorticity surface P = 1.5 pvu (solid lines, contour interval 1000 mgp) and the wind vector
’,
plotted with the standard meteorological conventionswhen thevelocity is largerthan 40 m s- The system of interest is highlighted by the
shaded low level vorticity maximum of Figure 3.
of a baroclinic zone. The wind field is shown at a framework moving with the low-level cyclone, as
the tropopause defined with potential vorticity; shown by Figure 7.
since the latter is a conserved -property,
- . it is a kind
of material surface. But this surface has a complex
shape, as shown by its geopotential height. An upper- Recent Past Conceptualizations
level signature accompanying the cyclone in the of Extratropical Cyclones
first half of its life cycle is easy to see. It is a The evolution of the understanding of extratropical
well-defined upper-level trough that also amplifies cyclones is tied to the progress of two areas: better
but moves at its own phase speed, slower than observations on the one hand, and on the other
that of the cyclone. A key thing to note on this the introduction into meteorology of new mathe-
case is that this upper-level feature definitely pre- matical tools for physics, allowing assumptions
exists. and their consequences to be quantified and checked.
Indeed, on 3 February at 1200 UTC, it is above the From that point of view, early attempts using concepts
Great-Lakes, but without any significant signature like the thermal cyclone or the polar front cyclone,
either in the satellite image or the surface pressure field although seminal in some respects, have not been very
(Figure 6 ) . It is when this feature closes on the successful.
acceleration area of the jet stream that Low 30 takes
shape. The interaction between the upper-level trough,
Upper-Level Induced Development Model
the baroclinic zone, and the developing cyclone fits
well with the first growth phase. The cyclone then The emergence of operational meteorology accompa-
seems to escape away from the upper-level trough nying the development of aviation led to better-
during the decreasing phase. The final growth stage coordinated observational networks. The surface
coincides with the cyclone crossing the baroclinic zone measurements benefited from the first ships able to
from the warm to the cold side, without a clear hold a fixed location in the ocean. Most importantly,
independent upper-level vortex-like feature being upper-air observations over continents became more
involved. The various phase relations are best seen in frequent; they used sondes attached to balloons.
Figure 6 The situation preceding the development of the extratropical cyclone shown in Figures 1,3 and 4, showing the upper-level
precursor and its environment. Although it is well organized at the tropopause, there are few associated features elsewhere. The fields in
the top panel are as in Figure 1, in the middle panel as in Figure 3, and in the bottom panel as in Figure 5.
Figure 7 Several fields showing the extratropical cyclone of Figures 1 , 3 ,and 4 and its environment in aframe centered on the low-level
core of the cyclone and rotating so that the horizontal axis remains roughly parallel to the mean isotherms. The fields are the 850 hpa
relative vorticity (heavy lines, same contouring as in Figure3, system of interest shaded); the tropopause level vorticity (light lines, same
contouring); and the tropopause wind field above 40ms-’.
Radiosondes transmitting their data while ascending minority among those attempting to understand
were introduced, removing the need to recover the cyclones.
recorded measurements after the balloon burst and
the sonde fell in some unlikely place. In 1935, a Upper-Level Jet, Baroclinic Zone, and Baroclinic
‘coordinated ascents’ experiment on the scale of Instability
western and northern Europe took place. For the first The school of Sutcliffe, Kleinschmidt, Petterssen, and
time, a few mature cyclones were sampled in three others was in a minority because an elegant mathe-
dimensions.
matical framework was set up in parallel and attracted
The upper-level structure of the flow about extra-
most of the attention. Discarding the unrealistic
tropical cyclones was then quite clearly obtained. The
extreme front model, Charney in California and
wind increased with height much more rapidly near
Eady in England applied, in the early 1940s, the linear
the cyclone center than away from it. A very clear
stability technique to the study of the properties of a
vertical tilt with height of the minimum pressure
simple broad baroclinic zone. This is an area of
location was definitely shown to exist. It took some
continuous and moderate transition from warm to
time to make something of the wind distribution:
cold air, which, as a consequence of rotation, is
ultimately, and with much further observations, it was
accompanied by a zonal wind increasing with
recognized to be organized in meridionally localized
height - an idealized form of the upper-level jet
streams of air, called jet streams. But the importance of
stream. This basic or initial flow indeed fits the
the vertical tilt and the related upper-level mass
observed average flow at mid-latitudes much more
distribution was grasped very quickly, partly because
reasonably than the extreme front model. This
it had been anticipated both with observations and
problem can be idealized to a point that allows
theoretically on several occasions before.
analytical solutions to be derived that still share
Several approaches have since been developed. One
some properties with real cyclones. Most notably,
is called ‘development’. This had been outlined shortly
the vertical tilt with height is part of the most unstable
before the 1940s by the British scientist Sutcliffe, who
solution, together with reasonable spatial scales. This
proposed that the emergence and amplification of a
is the baroclinic instability theory of cyclogenesis.
surface cyclone resulted from the influence of a
These similarities made linear studies the preferred
preexisting upper-level trough in a broad baroclinic
path to explaining natural phenomenon like the
zone (rather than an extreme front). This view is
growth of cyclones. Most textbooks on dynamical
supported by simplifying fundamental equations,
meteorology depend heavily on linearized equations
rather than linearizing them, and concentrating, in
and stability problems. Very elegant general existence
particular, on vorticity dynamics. In problems where
theorems have been derived.
rotation is important, looking at vorticity is a good
Because it is so elegant and appealing to mathema-
tactic.
tically oriented minds, the instability model domi-
Along this line of thought, the German meteorolo-
nated the theoretical scene throughout the 1950s
gist Kleinschmidt put some flesh on the concept of the
and 1960s. But because part of its elegance comes
upper-level component of an extratropical cyclone
from mathematical tractability, only extremely idea-
and its relation to vorticity. Using the homogeneous
lized problems can be studied and the relationship
and extended network of radiosoundings available
with observations is the weak spot of all of this work.
over Europe in 1943, he built a striking three-
As a consequence (or measure) of this, the results have
dimensional picture of this structure. Furthermore,
had little impact on practical forecasting aside from
he gave the description in terms of a quantity akin to
some guidelines for interpreting numerical model
vorticity and called potential vorticity. He particularly
output.
undertook to show that, under certain assumptions, it
is possible t o derive the wind and temperature
Quasi-Geostrophic Model and Theoretical Results
distributions from that of potential vorticity, a process
called inversion. In the early version of the instability approach, subtle
Sutcliffe and Kleinschmidt are, together with the simplifications were introduced at key points in the
Norwegian Petterssen and a few others, representa- derivation, limiting the number of solutions. This
tives of a school of thought that made the most practice is in fact very powerful as far as gaining an
of available observations to conclude that cyclones understanding is concerned. It was soon reconsidered
are largely caused by finite-amplitude interactions of more formally by Charney in terms of a scale analysis
some pre-existing ‘structures’: vertically confined of the basic equations; this is called filtering on the
vortices, baroclinic zones or jet streams, fronts, and basis of a balance assumption. A simplified set of
the like. However, until recently, they represented a equations specifically oriented toward understanding
extratropical cyclones and their large-scale environ- described entirely numerically and of known
ment can be derived at the outset. This is the quasi- representativeness.
geostrophic model. These equations represent
cyclones and some of their behavior, but they exclude, Data and Processing, Definition of Life Cycle Phases
for example, gravity waves. They include well-defined
cause-and-effect relationships. In this framework, for The data cover the period 1979-1993 and are an
example, vertical motion is the computable effect of extraction of the winter season (15 October-15 April)
the interaction between the thermal and horizontal over a domain centered over the North Atantic, with a
wind fields. resolution of 1.5” and over 1 3 pressure levels.
It is important to realize that this model is the Events of interest are defined using vorticity
common ground of both the intrinsic instability rather than pressure, for reasons already given. The
approach and the finite-amplitude interactions trajectories of 850 hpa vorticity maxima have
approach. With hindsight, it is easy to show that been constructed using an automatic tracking
Sutcliffe’s calculations are based on an early form of algorithm. In order to distribute events into classes,
the quasi-geostrophic filtering and so are Eady’s. the large state vectors that result from extracting fields
Potential vorticity inversion, as outlined by Kleinsch- in boxes of size of about 2500 km centered on the low-
midt, is a well-posed classical Poisson problem in this level vorticity maxima along the trajectories are
framework. The current theoretical understanding of reduced by a principal component analysis. The
cyclones undoubtedly rests on the quasi-geostrophic remaining components are distributed into classes
framework. using ascending hierarchical classification. Paradigms
of situations conducive to cyclogenesis are then
Recent Observational Description of Cyclones
obtained by averaging the original fields within each
class.
The explosion of observations that characterized Study of the distributions of amplitudes, of dura-
the last quarter of the twentieth century quickly tions, and of the various growth phases discourages
showed that the idea of a single type of cyclone was the automatic classification of full life cycles. Two
an idealization. Perhaps it was necessary in a period particular stages have been isolated. One is the 12h
when measurements were scarce. Although most period centered on time 0 h of each case, extrapolating
cyclones form and evolve in baroclinic regions a location for the field extraction at - 6 h: this phase is
accompanying the jet stream, some do not. Even in called the initiation of a new cyclone. The other is the
the baroclinic family, there is an extraordinary 24 h period preceding the time of maximum ampli-
diversity of shapes and histories, and therefore tude: this is called the maturation phase and generally
perhaps of mechanisms. includes the period of maximum vorticity growth
At the moment, knowledge of cyclones results from undergone by each case.
two main sources, both combining numerous observa-
tions in all dimensions. A number of dedicated projects Summary of Composite Genesis Situations
such as FASTEX have been set up to concentrate The classification of initiation has 1 2 classes.
observational platforms so as to document specific This classification uses nearly 6000 cases. As it is
cases in a very detailed way, such as the case presented impossible to fully account for all the classes here,
above. Beside these cases, the work on the possible composites are gathered into ‘families’, of which there
evolution of climate has led to reassimilation of rather are five. The best previous studies dealt with two dozen
long series of global data, providing homogeneous, cases.
consistent datasets from which thousands of cases can The most frequent family corresponds to the genesis
be extracted and studied. of a cyclone on the northern side of the baroclinic
zone, in the cold air (28% of cases). Nearly as
Types of North Atlantic Extratropical frequent, there is a family that comes close to the
extensively studied frontal instability problem (24%,
Cyclones
Figure 8).It departs from the idealized situation on one
Looking for a more generic description of ‘typical’ critical aspect: The flow is three-dimensional and the
extratropical cyclones, one is led to take advantage of new cyclone forms at the end of the frontal zone,
the former set of fields obtained by reprocessing years below a jet entrance. Neither of these families seems to
of archived data into a given version of the data feature a precursor of the cyclogenesis. The next two,
assimilation scheme. One such project is the ECMWF on the other hand, do have such a preexisting structure
Re-Analysis (ERA).It is used here to build composites of cyclone scale: the class that corresponds to the
representative of hundreds of individual cases, well-known (but clearly overestimated) development
Figure 8 One of the initiation class W composite at times (from left to right) - 6 h, 0 h (first detection time of an individual vorticity
maximum) and +6 h, at 300 hpa (top) and at 850 hpa (bottom). The fields on the top row are wind vectors (knots) coloured according to
speed so as to highlight the jet-flow structure (darkest blue above 40 m s - ’ ) and relative vorticity i,(light purple lines, contour interval
2 x 1 0 - 5 s - ’ , negative contours dashed). The green-shaded background field is the 600 hpa relative humidity (contour interval lo%,
white for values largerthan 90%, colordarkerfor smallervalues). Bottom fields are relativevorticity (as at the top, with darker contours) and
equivalent potential temperature 0, (brick-red thin contours and shading, interval 3 K, limit between blue-green shade and light orange
300K). The 850 hpa wind is also shown as above, the darkest blue threshold is changed to 10ms-’. The boxed arrow is a compass
indicating the average direction of the geographical north, since the area shown is rotated. (Adapted from Ayrault F, Joly A (2000),
Comptes-rendus de I’ Academie des Sciences, Paris, Sciences de la Terre et des planetes 330: 173-1 78.)
envisioned by Sutcliffe and Petterssen, and a new Summary of Maturation Situations

family in which the precursor is a full-scale cyclone
that splits into the old and a new system. A single The maturation phase composites show seven classes.
picture summary of each family is shown in Figure 9. Figure 10 shows some of the morphological character-
Figure 9 A single-time summary of three of the other families of initiation composites. The structures are shown at the time of the first
detection of an individual vorticity maximum, that is, the appearance of a new cyclone. From left to right: a class of cold air cyclogenesis;
cyclogenesis induced by a preexisting upper-level precursor; and a class of cyclogenesis resulting from the apparent aft-splitting of an
older cyclone. The fields are displayed as in Figure 8. (Adapted from Ayrault F, Joly A (2000) Comptes-rendus de I’ Academie des
Sciences, Paris, Sciences de la Terre et des planetes 330: 173-1 78.)
Next Page CYCLONES, EXTRA TROPICAL 603
Figure 10 Example of a class of mature extratropical cyclone, the most intense one, the meteorological ‘bomb’ (Class A). The
composite structure is shown at times - 24 h, - 12 h, and at the time of maximum amplitude (from left to right), at 300 hpa (top) and at
850 hpa (bottom).Thefieldsareas in Figure8exceptforthewindshownabove50 knots. (From Ayrault F, JolyA(2000), Comptes-rendus
de I’ Academie des Sciences, Paris, Sciences de la Terre et des planetes, 330: 167-1 72.)
istics of Type A, the one that undergoes the largest The method employed so far yields two sets of
deepening (-25 hpa/24 h) and happens in 12% of composites. A fundamental result comes from the
cases: It is what some authors have termed the clear separation between these two sets: The subse-
meteorological ‘bomb’. Centered on the low-level quent amplification of a cyclone does not necessarily
component of the system, the figure clearly depicts the involve the same processes as its genesis. Indeed, while
phase change between this and the upper-level
component: As there are independent studies of these
strong cases, this class validates the approach.
Class Byalthough not deepening, remains in warm
air and yields significant precipitations. An interesting
characteristic of this type is that it occurs in rather
strong baroclinic zones, like Class A and with the same
frequency. Yet it does not develop strongly (Figure 11).
The magnitude of the baroclinicity does not explain
the two different behaviors, and this calls into question
explanations in terms of simple instabilities. Class B
does not have the upstream upper-level signature that
Class A shows well.
The other types correspond to cyclones deepening
between 9 and 15 hpa in 24 h. They share with Type A
the presence of a well-defined upper-level signature.
One feature that discriminates these types is the shape
of the upper-level jet and the position of the cyclone
relative to it (Figure 12). Using time filtering to
separate the lows and their environment, it is possible
to estimate the energy transfers. The preferred energy
conversions are similar for all the developing classes
and correspond to the early twentieth century idea that Figure 1 1 The mean 24 h pressure deepening A P P 4 h of the
seven classes of maturation phases is plotted against the low-pass
cyclones develop by converting the energy contained
average baroclinity, the temperature gradient of their environment.
in gradients Of temperature into wind. In (From Ayrault F and JolyA (2000) Comptes-rendus de/’Acad&mie
that sense, the core physical mechanism for growth is des Sciences, Paris, Sciences de la Terre et des planetes 330:
the same for all these types. 167-1 72.)
DATA ANALYSIS / Empirical Orthogonal Functions and Singular Vectors 617
Contents
Empirical Orthogonal Functions and Singular Vectors

Time Series Analysis
than this specific technique of spatial data analysis);

Empirtcal Orthogonal canonical correlation analysis (CCA); redundancy
analysis; and variations. We discuss space-time data
Functions and Singular fitting with a linear autoregressive model, called
vectors principal oscillation pattern (POP) analysis. All of
the above methods are thoroughly discussed and
C S Bretherton, University of Washington, Seattle, WA, referenced by von Storch and Zwiers (see Further
USA Reading). Lastly, we briefly mention some nonlinear
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. techniques for space-time analysis.
Principal Component Analysis

Introduction PCA was pioneered in the social/biological sciences by
Hoteling and Pearson in the 1930s, and first applied to
One frequently needs to analyze atmospheric obser-
atmospheric data by Lorenz in 1956. Consider a
vations or model outputs that are time-varying fields
multivariate data set yi(t,) consisting of several
of data at many points in space. Typically, one desires
samples (indexed by n ) of a set of variables (indexed
to isolate a few important and physically interpretable
by i). In a biological science application, the vari-
modes of variability in the data. Until the 1960s, this
ables might be the sizes of different body parts
was usually done subjectively, by constructing an (forearm length, waistline, height, foot size, etc.),
index (such as the Southern Oscillation Index of sea-
sampled across many people. In an atmospheric con-
level pressure difference between Tahiti and Darwin) text, the variables might be annually-averaged sea-level
that appears to correlate strongly with the variability pressure anomaly yi(t,) at I observing stations
in one or more data fields (sea-level pressure, sea
i = 1 , .. . , I , sampled for N = 50 years t,,
surface temperature, winds, precipitation) at many n = 1,.. . , N. PCA seeks a single linear combination
locations. Regressing this index onto the data
of the variables that explains the maximal fraction of the
or binning the data corresponding to different ranges
overall sample variance.
of the index yields composite spatial patterns of
variability.
Mathematical Derivation
To avoid the trial and error of concocting such
indices, it has become common to use techniques that This can be phrased as the following mathematical
directly find dominant modes of variability in the data. optimization problem. Let y(t,) be the vector of
This article describes some popular techniques for observations at time t,. Any linear combination of
doing this. these observations can be written a(t,) = u y ( t , ) , e
We first discuss principal component analysis (PCA) where u is a vector of station weights ui. It suffices to
and some variants. PCA seeks spatial patterns that take u to be a vector of unit norm. Then a(t,)he
explain the maximal variance in a single data field. We projection of the data onto the direction u, and u2 ( t )is
then examine popular methods for isolating important the explained variance (eqn [l]).
spatial patterns of coupled variability between several -
data fields. Such methods include maximum covari- $ ( t ) = uT y y Tu = uT c,,u [I1
ance analysis or MCA (also widely called singular Here, an overbar represents an average over the N
value decomposition or SVD - we prefer the name sampling times, and the (i, j)th element of the matrix
MCA, suggested by von Storch and Zwiers in 1999, Cy, is the covariance of yi with yi. In this language,
since SVD is the name of a fundamental matrix PCA seeks a u that maximizes eqn [l] subject to
decomposition whose applications are far broader IUI = 1.
618 DATA ANALYSIS / Empirical Orthogonal Functions and Singular Vectors
Since Cy, is a symmetric positive definite matrix, Because higher EOFs are forced to be orthogonal to
standard linear algebra tells us that the desired u = el, their predecessors, they may not be amenable to simple
the eigenvector of Cy, corresponding to its largest physical interpretation. Furthermore, since the covar-
eigenvalue 111. If the other eigenvalues are ordered iance matrix was estimated on the basis of a finite
from largest to smallest, the kth eigenvector is the number of physical realizations, its EOFs may contain
direction that explains the maximal variance in the sampling biases. This problem becomes more acute for
data and is orthogonal to all previous eigenvectors. higher-index EOFs, and can also mix EOFs with
This eigenvector is called the kth empirical orthogonal closely spaced eigenvalues. North’s criterion states
function (EOF)of y and the corresponding time series that if the covariance matrix is constructed on the basis
ak(t) of the projection of the data onto this EOF is of N independent samples, its eigenvalues have sam-
called the kth principal component (PC).The data field pling uncertainties 6& FZ 2&/N, and if the spacing
can be exactly rewritten in ‘canonical form’ in terms of between successive eigenvalues is comparable to or
its EOFs and PCs (eqn [ 2 ] ) . less than this sampling uncertainty, their EOFs will be
heavily contaminated by sampling uncertainty. An-
PI other simple check on the robustness of the PCA is to
divide the data set into randomly chosen halves, then
k=l
examine the similarity of the EOFs given by the two
Furthermore, the ak (tn)are temporally uncorrelated, halves.
so the total variance in y summed over all spatial
locations is just the sum of the eigenvalues &, and the
fraction of the variance explained by the kth PC/EOF Rotated PCA
pair is given by eqn [ 3 ] . The spatial patterns produced by PCA can sometimes
f k =- Ak
[31 be difficult to interpret. For instance, two ocean basins
1=1 1 might have characteristic patterns of sea surface
temperature anomalies that explain a lot of the
Owing to these orthogonality properties, the EOFs can intrabasin variability, but the variability in the two
be interpreted as spatial patterns of variability as well basins might only be weakly coupled. To maximize the
as just weight functions. In particular, if y i ( t ) is leading mode variance and maintain an orthogonal
regressed on the kth PC, the regression slope is just second mode, PCA would yield a leading mode in
eik, the ith element of the kth EOF. Thus 11:”ek can be which the spatial patterns in the two basins are in-
interpreted as the characteristic change of y ( t ) corre- phase and a second mode of slightly smaller explained
sponding to a one standard deviation change in variance in which they are out of phase. This might
the kth PC. direct attention to the weak coupling between the
basins rather than the much more robust intrabasin
Practical Interpretation of PCA
variability.
In practice, only a few leading EOFs explain most of Rotated PCA (in statistical parlance, a type of factor
the variance of the field. Since most atmospheric fields analysis) can be used to obtain localized modes that
have long-range spatial correlations, these EOFs might have a simpler physical interpretation. It con-
typically correspond to patterns of broad spatial scale, sists of choosing a set of spatial weights or directions
and are relatively robust even when the number of that optimize some measure of localization. The data
samples is low or the data are noisy. Thus, the leading are typically filtered onto the leading EOFs before this
PCs are often useful index time-series for compositing procedure is carried out. Varimax rotation, which is
with respect to the leading modes of variability of a most commonly used, maximizes a function of the
given field. Often, 1-5 PCs explain as much as 90% of weights that tends to force them to be closer to either
the overall variance in the field, and these PCEOF pairs zero or one than regular PCA. This favors modes that
form an efficient low-dimensional description of the are weighted mainly to a few spatial points, inhibiting
entire field. In this case, projection onto these PCs is a mode mixing of the kind discussed above. However,
useful spacehime filter before other statistical methods the modes are no longer temporally uncorrelated
such as canonical correlation analysis are employed. (hence they do not form a partitioning of the variance)
If the variability of the field is dominated by and may not be spatially orthogonal for some rotation
propagating disturbances (e.g., mid-latitude sea-level strategies. Furthermore, the leading mode will not
pressure variations on synoptic time scales), the explain so much of the variance as with regular PCA,
corresponding eigenvalues come in a nearly equal and many spatial patterns do not have the localized
pair corresponding to two phases of the disturbances character that rotated PCA selects for, so this method
that are in quadrature. should be applied thoughtfully.
DATA ANALYSIS / Empirical Orthogonal Functionsand Singular Vectors 619
PCA for Propagating Disturbances direction has maximal possible covariance with the
Several methods for extending PCA to better analyze
-
corresponding projection bl ( t )= v z ( t ) of the right-
field values in the v direction. The solution is obtained
spatially propagating disturbances have been pro-
by singular-value decomposition of the cross-covari-
posed. One such technique is complex PCA (CPCA),in
ance matrix Cy, between the gridpoint values of the
which a discrete Hilbert transform in time is applied to
left field and those of the right field. If its left and right
the field of data. A new field is created whose real and
singular vectors are lk and rk, ordered from largest to
imaginary parts are the original field and its Hilbert
smallest singular value ok, then the optimal u and v are
transform. A PCA of the (now complex) covariance
the leading left and right singular vectors 11 and r l .
matrix of this field yields complex eigenvectors whose
Succeeding singular vectors define coupled spatial
real and imaginary parts give the characteristic spatial
modes that maximize the covariance subject to
patterns in both phases of the wave.
orthogonality with the previous modes. Unlike with
Singular Spectral Analysis (SSA) PCA, the corresponding modal amplitudes may be
temporally correlated.
Many modes of spatial variability have characteristic
MCA is efficient to numerically implement, requires
time evolutions, for instance, the life cycle of a no prefiltering of the fields, and produces spatial
baroclinic cycle. Singular spectral analysis is a method patterns that exhibit broad coupling over as much as
popular in some circles for characterizing the domi- possible of the spatial domains being analyzed. Its
nant patterns of space-time variability, and has been optimality characteristics usually provide excellent
used as an alternative to power spectral analysis. results, but there are caveats. One is that even if two
Conventional PCA is based on the matrix of simulta- fields are perfectly linearly coupled (e.g. a two-
neous (unlagged) covariances. SSA is based on PCA of dimensional streamfunction and corresponding vorti-
an extended covariance matrix that also includes city field), MCA may only approximately identify this
lagged covariances with a sequence of time lags. The coupling. A second caveat is that as with PCA, the
EOFs are parsed into characteristic spatial patterns spatial orthogonality constraints on succeeding spatial
evolving through the sequence of lag times. patterns may mix physically distinct but spatially
nonorthogonal modes of variability.
Methods for Isolating Patterns of
Coupled Variability Canonical Correlation Analysis
Two linear methods, MCA and CCA, are widely used CCA, first developed by Hoteling in 1936, seeks a
for isolating patterns of coupled variability in two linear combination a ( t ) = u . y ( t ) of the left-field
fields. A third method, called ‘redundancy analysis’ values and a linear combination b ( t ) = v z ( t ) of the
e
(RA) by von Storch and Zwiers, is also useful for right-field values that have maximal temporal corre-
predicting the time-space structure of one field based lation. This optimality criterion is similar to MCA
on observations of another, correlated field. (which maximizes covariance). In practice, the differ-
Mathematically, these methods can be phrased as ence is that CCA can choose linear combinations that
follows. We call the two fields the ‘left’ field yi(tn) and explain little of the variance in their field but are highly
the ‘right’ field zj(tn).For example, we might look for correlated with the other field. MCA will select linear
coupling between monthly anomalies of tropical combinations that explain significant variance in their
Pacific sea surface temperature at an array of buoys fields but may be slightly less well correlated with the
(left field) with monthly gridded Northern Hemi- other field.
sphere 500 hPa height anomalies (right field). The two The optimal solution can be found by first perform-
fields must be given at the same time, but not ing PCA on each field. This yields standardized left and
necessarily at the same spatial locations. right principal components a k ( t ) and P l ( t ) . Then a
singular value decomposition is performed on the
Maximum Covariance Analysis (MCA)
cross-covariance matrix between a k ( t ) and P l ( t ) . The
MCA (also commonly known as SVD) was first lefthight singular vectors are transformed back into
applied to an atmospheric problem in 1976 by gridpoint coordinates to get the weight vectors uk/vk.
Prohaska, and was comprehensively compared with The singular values are the correlations between the
other methods of space-time analysis by Bretherton corresponding time-series a ( t ) and b ( t ) .
and colleagues. MCA finds an optimally coupled left To obtain robust results with CCA if a limited
spatial pattern u and a right spatial pattern v, both unit number of times have been sampled, one should first
vectors. They are chosen such that the projection prefilter the data onto a small set of leading principal
a l ( t ) = u - y ( t ) of the left-field values onto the u components of each field (that collectively explain
620 DATA ANALYSIS I Empirical Orthogonal Functions and Singular Vectors
70-90% of the variance) to avoid spurious large The corresponding eigenvectors of A (the POPS)
correlations associated with overfitting. In this case, describe the characteristic unforced patterns of spatial
CCA tends to give quite similar results to MCA for the variability of the system; those with eigenvalues near
leading pattern. Successive patterns are temporally to one in magnitude are slowly damped and will tend
uncorrelated but may be spatially nonorthogonal. to dominate the observed response. Complex eigen-
CCA is slightly more numerically intensive than MCA, values correspond to damped oscillations; the corre-
and the prefiltering step slightly complicates the sponding POPs are also complex, and the real and
analysis procedure. However, it can sometimes be imaginary parts of the POP describe the oscillations in
preferable if strong coupling between the fields is the spatial structure of the response.
geographically localized. To estimate A from the data, consider the expected
value E [ . ] of an outer product of eqn [4] with y,
Redundancy Analysis
(eqn [si).
In redundancy analysis, an application of multiple
regression first developed in the 1970s, one tries to 151
explain the maximal fraction of the variance of the From eqn [.SI, A can be estimated based on the
right (predictand) field using the left (predictor)field. empirical lag-0 and lag- 1 covariance matrices.
This can be viewed as a collection of least-squares
multiple-regression problems for the right-field values
at each location. As with CCA, prefiltering is recom- Nonlinear Methods
mended to avoid overfitting of the data. The above ‘linear’ analysis techniques decompose an
Aggregated PCA observed field into the sum of modes, each of which is
the product of a spatial pattern and a time-varying
The above methods look for coupling between two amplitude. Some phenomena, such as ENS0 sea
scalar fields. Coupling between more than two fields, surface temperature variability, show a somewhat
or a vector field and a scalar field, can be handled by different spatial anomaly pattern in the positive phase
aggregating some of the fields after standardizing them from that in the negative phase, and may be more
to comparable variance, and analyzing coupling efficiently described as nonlinear modes whose pattern
between two aggregated fields. Another approach in covaries with their amplitude. One standard check for
this case is to aggregate all of the fields into a single this behavior is to composite the spatial anomalies
field, whose modes of variability are found using PCA. corresponding to positive and negative values of an
The resulting modes optimize a combination of the index time-series and see whether they are substan-
explained variance within individual fields and the tially different in structure. Neural net techniques have
covariance between fields, and may not pick out the recently been developed that may prove to be useful
patterns of coupling between the fields as effectively as for isolating such nonlinear modes of variability.
the previous methods.
See also
Principal Oscillation Pattern (POP)
Climate Variability: North Atlantic and Arctic Oscillation.
Analysis
El Nifio and the Southern Oscillation: Theory.
Some physical systems can profitably be idealized as a
damped linear response to stochastical forcing. POP
analysis fits such a model to the space-time variability Further Reading
of a field or coupled fields. Given a vector yn of Bretherton CS, Smith C and Wallace JM (1992) An
observations at times t, of one or more fields at one intercomparison of methods for finding coupled patterns
or more locations, the fitting model has the form of in climate data. Journal of Climate 5: 541-560.
eqn [41. North GR, Bell TL, Cahalan RF and Moeng FJ (1982)
Yn+l = A y n + fn 141 Sampling errors in the estimation of empirical orthogonal
functions. Monthly Weather Review 110: 699-706.
The stochastic forcing fn is a noise vector uncorrelated Strang G (1988) Linear Algebra and Its Applications, 3rd
between observation times, and A describes the edn. New York: Harcourt-Brace-Jovanovich.
evolution of the unforced system. The eigenvalues of von Storch H and Zwiers FW (1999)Statistical Analysis in
A should all have magnitudes less than one, so that this Climate Research. Cambridge: Cambridge University
model can describe a statistically stationary process. Press.
DATA ANALYSIS / Time Series Analysis 621
Time Series Analysis

G R North, Texas A&M University, TX, USA to generating realizations of the height random varia-
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. ble. Most realizations of such a process yield values near
the center of the bell-shaped curve, with large excur-
sions from the population average being very rare.
The mean of a random variable is the arithmetic
Introduction average of its values taken over many realizations
Study of the records of weather and climate over time (actually the limiting value derived from infinitely
is extremely valuable for practical as well as theoret- many). The variance is the average of the squares of the
ical purposes. For example, the instantaneous temper- deviations from the mean. The deviation from the
ature or humidity taken at hourly intervals constitutes mean for two variables can be multiplied together and
a time series of measurements. Such data might be used averaged over many realizations to form the covari-
for many purposes such as assembling a climatological ance of the two random variables. As the name
summary for the site or for performing an analysis to implies, the covariance is a measure of how two
identify the underlying physical processes responsible random variables covary. If the two variables are
for certain interesting features of the data. In such identical then their covariance is simply the variance.
an example, we might not be surprised to find a daily If the two variables are not statistically related to one
swing of the temperature and humidity as well as another then the covariance is zero. The correlation
an annual oscillation. Were we to plot these hourly between two random variables is the covariance
temperatures, a nearly repeating graph of period 24 divided by the square root of the product of the
hours would result (the diurnal cycle). Examination variances. Its magnitude is always less than or equal to
over longer time spans leads to the identification of an unity. Anticorrelated variables have negative correla-
annual cycle in the data. By simple inspection, we tion (and covariance). In a time series we are often
would rather quickly decide that there is some interested in how an entry covaries with or is related to
underlying physical agent responsible for these near past (or future) entries. In particular, we would like to
repetitions or periodic statistics in the data stream. know if it is correlated with immediately past entries, a
This is an example of the most primitive form of time property known as serial correlation.
series analysis.
Many decades of experience have led environmental
and statistical scientists to devise very sophisticated
White Noise
methods of studying records of time series. The most
powerful innovation is the idea of a mathematical The simplest model of a randomly based time series is
model of the time series. A mathematical model one in which each new time step leads to a statistically
involving random components is a very convenient independent (no serial correlation) drawing (like the
way of representing a time series of data. Such models heights of individuals above); but each new entry is
always employ simplifying assumptions, but such drawn from an identical distribution. Each data entry
techniques work in a surprisingly large number of is statistically independent of the previous one. This
applications, especially in the geosciences. model is called the statistically independent, identi-
Before describing how such a time series can be cally distributed white-noise model. It is the most
constructed, we must introduce the idea of a random important model in time series analysis since almost all
variable. A random variable can take on values drawn other statistical models are derived from it.
from a certain probability distribution. For example, A few properties of the white noise model (we
the outcomes of flipping a coin are a random variable. hereafter assume that the identically distributed
Each flip constitutes a realization of the random property is included) are worth mentioning. First,
variable -either heads (H)or tails (T).The probability the mean of the entries of such a time series can be
distribution is 50-50 (probability 0.5 for H, 0.5 for T). estimated by adding the values of a long series of
More commonly encountered in geophysical and entries and dividing by the number of entries used. In
behavioral sciences is the case where the variable can doing this we are actually finding an estimate of the
take on a range of discrete or continuous values and its underlying mean of the probability distribution of the
frequency distribution is the normal or bell-shaped individual entries. Another property of time series that
curve distribution. An example is the distribution of is easily demonstrated with the white-noise model
heights of individuals. Drawing names from a hat and time series is that of an independent realization of a
announcing the height of the individuals is equivalent time series. Above we introduced the random variable.
622 DATA ANALYSIS I Time Series Analysis
n
20 40 60 80 100
Figure 1 Anexampleof a realizationofawhite-noisetimeserieswith100entries.Themeanof thistimeseriesis2.00andtheentriesare

drawn from a normal distribution with unit variance. In this plot as in subsequent ones the entries are joined by a continuous line.
In time series we actually have a random function or and these are common in nature. (Strictly speaking this
string of random variables, with individual realiza- only defines a second moment stationary time series,
tions being individual graphs of the function. The since nonstationary properties of the probability
property that is so evident is that averaging along the distribution may still be present. If the random
time series is equivalent to fixing the time and variables are normally distributed, these mean and
averaging across realizations to find the mean. This covariance stationary properties suffice to determine
property of ensemble averaging is a very powerful one, the strong forms of stationarity.) Perhaps some exam-
which enables many of the proofs and analyses of time ples of nonstationary time series will help clarify the
series analysis. An example of a single realization of a concept. The diurnal or seasonal data mentioned
white-noise process is shown in Figure 1. above are examples of nonstationary time series since
An example of a white-noise time series is the their means depend on local time of day or time of year.
heights of students standing in a cafeteria queue. On In addition, their variances will also have such a phase
the other hand, consider the heights of succeeding first dependence; even their serial correlation structure
sons in an ancestral sequence. There is a genetically may have a phase dependence -for example, the serial
determined correlation between the height of a father correlation between entries may by greater in winter
and that of son. Such a time series exhibits a positive than in summer. At first glance the sequence of heights
serial correlation which diminishes to zero after a few of first sons across generations may seem like a
generations - a phenomenon known as regression to stationary time series, but there is known to be a
the mean. secular trend of increasing heights over generations,
probably because of better nutrition.
Despite our ability to enumerate many time series
that are nonstationary, the model of a stationary time
Stationary Time Series series is very valuable in the geosciences. For example,
Consider next a time series which is not necessarily annual averages of temperature at a location are likely
white-noise. That is to say, an entry may not neces- to form a stationary time series at least to a good
sarily be statistically independent of its predecessor - approximation. The statistics of such a time series
there may be serial correlation. Nevertheless, the time (mean, variance, serial correlation properties) make a
series may not have any preferred origin. That is, as we good summary of the sequence and for many purposes
look along the time series, statistically speaking, each may form an adequate substitute in practical applica-
time is equivalent to each other time. In such a time tions. For example, an insurance company may want
series the mean is independent of time and so is the to know the likelihood of the temperature (or flood
variance. The covariance between an entry and that a water level) exceeding a given threshold. The serial
certain number of steps, say n, earlier depends only on correlation structure is particularly important in
the temporal separation or lag, n. A time series having drought, where sequences of dry years can be the
the above properties is called a ‘stationary’ time series most important indicator of consequences.
DATA ANALYSIS I Time Series Analysis 623
Autoregressive Processes A
The most common type of time series encountered in

the geosciences is the first-order autoregressive process
(known as the AR1 process). In this process each new
entry can be written mathematically as the sum of two
terms, the first proportional to the previous entry, the
second an additive white-noise term. Higher-order
autoregressive processes (ARn) model the next entry
as a sum of n + 1 terms, the first n of which are
proportional to the previous n entries along with the
additive white-noise term. We concentrate here on the
AR1 process because of its central importance. The
parameters which describe the time series are its mean, Figure 3 The autocorrelationfunction corresponding to the AR1
its variance, and its so-called lag-one serial correla- process depicted in Figure 2. The lag is treated as a continuous
tion. It is the job of the analyst to take the given data variable for this plot for clarity of the display. The autocorrelation
time for this process is about 4.0.
series and determine or fit the parameters to the data.
That is, one wants to know the mean, variance, and
lag-one serial correlation in the data. If the lag-one
applications. One begins the analysis by taking the
serial correlation turns out to vanish, then we infer
finite-length segment of data in the sequence and
that the series can be modeled by a white-noise time
estimating the Fourier coefficients for representing the
series (ARO). If the lag-one serial correlation is Y then
data as a Fourier series on the segment. In this process
the lag-two is r2, and so on. In the limit of very small
one is representing the data in terms of the Fourier
time steps in the series this tends to an exponential
coefficients instead of the temporal entries. The two
falloff of serial correlation. The value of n for which
are equivalent ways of expressing the content of the
the serial correlation falls to l / e ( = 0.3678.. .) is
data. Each Fourier coefficient is a component or
known as the autocorrelation time. The autocorrela-
amplitude of a certain sinusoidal waveform in the data
tion time is a measure of the memory of the system. It is
stream. From the point of view of time series modeling,
often said that the system ‘forgets’ its past values after
the Fourier coefficients are random variables, since
a few autocorrelation times (Figures 2 and 3).
from one realization of the process on the same
segment to another the coefficients will differ. How-
ever, they will have certain statistical properties
Fourier Analysis of Time Series common across the ensemble of realizations. If the
If there are physical reasons to think that a time series segment is sufficiently long and the series is stationary,
of data is stationary, then Fourier analysis of the data it can be proven that the Fourier coefficients corre-
can lead to a number of powerful techniques useful in sponding to different frequencies are uncorrelated.
This permits us to perform an analysis of variance over
the different frequency bands to examine how vari-
ance is distributed over frequencies. It is routine to plot
a graph of the variance or sometimes known as power
as a function of frequency. This is known as spectral
analysis.
The most common example is the white-noise time
series. The white-noise spectrum is flat; that is, every
frequency band has allotted the same variance. Hence,
one way of determining whether a certain time series is
white-noise is to perform the Fourier analysis and plot
the spectrum (variance or power versus frequency). If
the spectrum is flat, we can infer that the time series is
white-noise. Of course, if the time series segment is
20 40 60 80 100 short, there will be problems in estimating the spec-
Figure 2 An example of a realization of an autoregressive
trum of the underlying process because of sampling
processof orderone. In this examplethe present entryis0.75 times error, Analysts have devised many useful techniques
the previous entry with an added normally distributed variable of for statistical testing of the white-noise hypothesis.
variance 0.25. The mean of the time series is 3.00. The term ‘white-noise spectrum’ derives from optics,
624 DATA ANALYSIS I Time Series Analysis
0.15
0.2
r =
S The estimate will not only provide a most probable
value of the future entry, but some assessment of the
uncertainty in the forecast - perhaps even a theoretical
frequency distribution of values that can be expected.
\
Interpolation is a second application of time series
modeling. Suppose there are missing values in an
0.02 empirical time series and for some reason one wishes to
o0.01
oo l insert values that are statistically consistent with the rest
of the entries. First one finds a model of the time series and
I I I I 1 1 , # # 1 I I cllllll I I # I h < # hFrequency
0.01 0.05 0.1 0.5 1 5 10
then one can find the most probable entry with an
associated theoretical frequency distribution. Depending
Figure 4 A log-log plot of the spectrum associatedwith the AR1 on the application, one may wish to insert the most
processdepictedin the previoustwo figures. Note that the log of the probable value or add to it a random number which is
spectrum turns over at about a frequency of 2n divided by the
autocorrelationtime, which in this case is taken to be 4.00.
consistent in a statistical sense with the nearest neighbors.
A third application is in the area of signal process-
ing. One often finds a deterministic signal embedded in
where it refers to white light, which has an electro- some kind of noise (white or colored). The object
magnetic energy intensity which is somewhat uni- usually is to separate the noise from the signal and to
formly distributed across the visible part of the estimate the amplitude of the signal. This is the
spectrum. By the same analogy, a red noise spectrum problem in radio reception. We refer to the process as
is one which has its energy more concentrated in the detection. By time series modeling and knowing some
lower frequencies as opposed to being uniform. characteristics of the signal wave form one can find an
The AR1 spectrum is of the red noise type. The optimal estimate of the signal strength. In electronics
characteristic frequency where the spectrum begins one might want to clean the noise away from the signal
to turn downwards is at about the inverse of the and amplify the residual, while in other applications
autocorrelation time (Figure 4). one might want to find the amplitude of a periodic
signal such as the diurnal or seasonal cycle. One of the
Applications of Time Series Analysis most famous applications of this type in the geosci-
ences is the detection of periodic signals in the record
Not only do time series analyses provide new insights of changes in continental ice sheet volume. These
into the underlying physical processes generating an excursions have been found to contain significant
empirical time series, but they are useful in a variety of variance peaked at periods of 100 000, 43 000 and
practical applications. 20 000 years, and these just happen to coincide with
A very common use of time series analysis is data the periods of the changes in the elliptical orbital
smoothing. Often one wishes to smooth out the highly parameters of planet Earth (eccentricity, obliquity, and
irregular short-term fluctuations in a time series to get a precession of the equinoxes). Thus time series analysis
better view of longer-term trends or undulations. This was able to show conclusively that the ice ages are
can be accomplished by running a smoother over the linked to the changes in the Earth’s orbital elements.
time series. For example, one might take as the value at a
certain time the arithmetic average over several future See also
and past values. These past and future entries can be
Data Analysis: Empirical Orthogonal Functions and
weighted in various ways to make the smoothing
Singular Vectors.
optimal for the particular application. This class of
operations is known as moving-average smoothing. Further Reading
Perhaps the most important application is forecast-
ing. One might ask, given a segment of a data-derived Bendat JS and Piersol AG (1986) Random Data: Analysis
time series, if it possible to use this information to and Measurement Procedures. New York: Wiley.
forecast future entries. The answer in principle is Bloomfield P (1976) Fourier Analysis of Time Series: An
Introduction. New York: Wiley.
simple. In the case of a white-noise time series (or an
Chatfield C (1992) The Analysis of Time Series - An
empirical one which is indistinguishable from white Introduction. New York: Chapman & Hall.
noise) there can be no forecast skill, since each entry is Percival DB and Walden AT (1993) Spectral Analysis for
statistically independent of the past entries. But in the Physical Applications. Cambridge: Cambridge Universi-
case of an AR1 process there is correlation with past ty Press.
entries and this will permit some statistical estimate of Wei WWS (1990) Time Series Analysis. Redwood City CA:
future entries out to roughly one autocorrelation time. Addison-Wesley.
DENSITY CURRENTS 625
P G Baines, CSIRO, Aspendale, Australia cold fronts. Such flows are sometimes made visible by
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. suspended dust, which mark out the features of the
cold density current. Noted examples of these are seen
in the Sudan (where they are known as haboobs), and
Density currents (also known as ‘gravity currents’) in in India, Australia, and Arizona during dry summers.
the atmosphere are air flows that are principally due to In these flows the dust usually makes a negligible
differences in density (i.e., temperature) between contribution to the density difference, but this is not
neighboring bodies of air. Descending cold dense the case in another example of density currents -
bodies of air down sloping terrain are sometimes powder snow avalanches. Here turbulence causes the
termed ‘katabatic flows’ or ‘katabatic winds’, and snow to be suspended in the air, producing the density
ascending heated air constitutes ‘anabatic winds’. difference that causes the downflow, which in turn
Such density differences cause lateral pressure gradi- produces the turbulence.
ents which produce horizontal motion, with cold air
moving over the ground into a warmer air mass.
Strong vertical motion occurs on length scales of Unidirectional Density Currents
several hundred meters but the currents can travel
hundreds of kilometers. The Earth’s rotation (via the The dynamical essentials of the above phenomena can
Coriolis force) is normally not important for these be encapsulated by looking at simple idealized flows
flows, unless they last for many hours. In this article I that represent them. These can best be seen from
give some examples of these flows, and then describe simple laboratory experiments. Here water is used to
their basic dynamical properties for flows over hori- represent air, because both are effectively incompress-
zontal ground that move in one direction, and that ible for motions on these length- and timescales. A
move radially outward in two dimensions. I then density current can be created in the laboratory by
proceed to discuss density currents that flow down releasing cold water into relatively warmer water, but
slopes, taking into account the effects of radiative it is usually simpler to work with isothermal fresh
forcing and environmental stratification. water and use dissolved salt to make the water denser.
In Figure 1 a body of dense salty water has been
released into a tank of fresh water, producing a density
Examples current moving from right to left over a horizontal
surface, viewed from the side. The leading part of the
Atmospheric examples of such flows are common.
current consists of a ‘head’ with an overhanging
They include sea breezes, in which heating of the air
leading nose. This head may be regarded as a limiting
over the land by convection from the solar heated
form of hydraulic jump in a cold layer of dense fluid, in
surface below causes a lateral density difference
which the depth upstream of the jump is zero (see
between that and the relatively cooler air over the
Hydraulic Flow). Behind the head the fluid has a three-
ocean. The onshore flowing sea breeze is the result.
Similarly, the weaker offshore nocturnal land breeze
occurs due to radiative nocturnal cooling over the
land. In fact, nocturnal cooling is a common cause of
density currents over uneven topography, which tends
to cause larger lateral gradients and currents than over
horizontal terrain: radiative cooling of the air causes
flow down sloping topography, forming drainage
flows, and valley winds. Another source is thunder-
storms, which contain downdraughts of cold air due to
the drag of falling raindrops, and their evaporation.
On reaching the ground, these downdraughts spread
as gust fronts (see Gust Fronts) and form density
currents. Cold fronts usually contain one or more
squall lines, which consist (by definition) of a line of
thunderstorms, so that the associated downdraught- Figure 1 Shadowgraph showing the structure of the flow in and
produced density currents are conspicuous features of behind a density current head.
626 DENSITY CURRENTS
layer structure. The layer of dense fluid constituting We may consider an idealized two-dimensional
the main part of the current is at the bottom. This fluid model of a density current, in which the volume flux Q
moves faster than the head, and catches up with it. and density p1 of the dense fluid are specified at a given
There it rises and is mixed with the surrounding lighter source location. If the density of the environmental
environmental fluid, and forms a density-stratified fluid is po, the buoyancy of the dense fluid is then
layer, spread out above the bottom current. Vorticity is g‘ = ( p l - po)g/po. Behind the head, the velocity zll
produced in the upper part of the head by shear and thickness dl of the dense layer are approximately
between the head and the ambient fluid, and this is also constant. We then have Q = q d l , and from dimen-
deposited in the stratified layer (see Vorticity). This sional analysis alone we have
mixed stratified layer moves slowly in the direction of
the current. It is highly turbulent immediately behind
the head, where most of the mixing takes place, and
The speed V H of the head also scales with V I , with
this turbulence decays with distance from the head.
z / H < z / ~ . Laboratory observations of flow into a
Here the interfaces between the mixed layer and the
deep homogeneous environment at large Reynolds
dense fluid below and the environmental fluid above
numbers give
are generally stable, so that apart from the decaying
turbulence little further significant mixing occurs.
Figure 1 shows a two-dimensionally averaged
picture. In fact, density current heads have three- which are approximately constant with time for a
dimensional structure, as shown in Figure 2 . The sustained source of fluid. The same observations also
leading edge contains many lumps and bumps, and show that the total height of the head is typically 3dl,
these shapes change continually as the head propa- and the height of the nose is about 0.15dl. These
gates. Some small parts of the head are more advanced expressions apply to density current heads in general,
than others, but these then disappear and are over- when the buoyancy and thickness of the dense fluid
taken by others. This lobe and cleft structure is due to approaching the head are given by g‘ and d l . This
the drag on the current by the rigid lower surface over steady pattern is maintained because the driving
which it propagates. This retards the lowest levels of pressure gradient is retarded by loss of momentum
dense fluid, causing the overhanging noses. The lighter through mixing in the head (see below), and the
fluid beneath these noses then rises through the dense following current is restrained by friction with the
fluid driven by its buoyancy, and this causes the bottom surface and the fluid above. When the current
unstable three-dimensional lobe and cleft structure becomes long, the thickness of the dense layer
that is seen in Figure 2. This structure can sometimes decreases gradually away from the source, providing
be seen in the atmospheric examples cited above, a small pressure gradient to overcome friction.
where dust or moisture make the cold air visible.
Unidirectional Density Currents due to Collapse
of a Dense Body of Fluid
Density currents may also be created by suddenly
releasing a large body of dense fluid within a stationary
environment of lighter fluid. This models a finite-sized
source, and is readily simulated in a laboratory tank by
rapidly raising a vertical barrier that separates the
dense and lighter fluid. In a two-dimensional situation
where the fluid spreads in the x-direction only, the flow
passes successively through three stages. In the first
stage of collapse of the dense fluid (of initial depth do,
length LO,and area A = A0 = doLo in fluid of total
depth D O )termed
, the slumping stage, the front travels
as a density current of constant speed and depth.
However, the surface of the dense fluid behind this
front is not horizontal as large-amplitude internal
waves propagate on it (see Figure 3). These waves
Figure2 A laboratorysimulation of adensitycurrent, in which the
reflect from the left-hand end (or center, for a
dense fluid has been made visible by milk. Note the lobe and cleft symmetric collapsing body), and the character of the
structure of the dense fluid interface, and the low-level transient flow changes when they catch up to the front, or
overhanging noses. density current head.
stage is reached, where the motion becomes sufficient-

ly thin and slow for viscous effects to become
important, a regime that is not relevant here,
In the second stage, dense fluid enters the head from
behind where it is mixed with environmental fluid. The
result is spread out behind, in a layer over the dense
layer. The rate of mixing within the head is roughly
proportional to its mean height, and hence decreases
with it. At early times in this stage ( x > x , ) , the
Figure 3 Schematic diagrams showing various stages at suc- detrained mixed fluid consists mostly of the dense fluid
cessive times in the initial slumping stage of the collapse of a
rectangular volume (shown dashed) of dense fluid with d o / D << 1.
with a small part of environmental fluid. However, as
the current proceeds, this proportion reverses and near
the end ( x >>x,), nearly all the mixed fluid is
Upper-level stratification often has a significant environmental. When the end of the second stage has
effect on density current propagation in the atmos- been reached (at x = x s + 29At'2), and mixing has
phere. For example, it is common for sea breeze fronts effectively ceased, the total volume of mixed fluid that
to propagate inland under an atmospheric inversion, has been produced is slightly more than twice the
which acts as an effective lid. This may be represented initial volume of dense fluid. The volume of the
in experiments by a finite total depth. If this is present, remaining unmixed dense fluid is quite small, so that
the resulting motion of the ambient fluid may affect the overall, the dense fluid and environmental fluid have
waves on the interface with the dense fluid, and hence been mixed in approximately equal proportions.
affect the details of the slumping behavior. In partic- The vertical shear of the horizontal wind also affects
ular, the presence of a reversed flow in the upper layer the behavior of density currents. For small positive
causes a corresponding reduction in the speed over the shear (i.e. stronger low-level flow opposing the
ground of the collapsing front, relative to eqn [2]. The current, weaker aloft), the forward face of the density
distance x , at which the slumping phase ends and the current steepens and the current thickness increases,
waves reach the head, measured from the end of the with the opposite effects for negative shear. This has
tank, is then observed to be important implications for convection initiation by
downdraught outflows and sea breezes over islands.
X
2 = 3 7.4- + do
131
LO DO
After the reflected waves have caught up with the
Radial Density Currents
density current head the flow enters the second self-
similar stage, which is dominated by inertia, buoyan- In the atmosphere, collapsing localized bodies of dense
cy, and mixing. Here the dense fluid collapses in the fluid may be constrained to spread in one direction if
form of a rectangle of approximately uniform area, they are spatially confined such as in a valley. How-
with increasing length and uniform decreasing height ever, this is usually not the case for thunderstorm
d l . This rectangular uniformity is maintained by downdraughts, for example, which are often free to
internal waves propagating outwards on the dense spread horizontally without confinement. These and
fluid interface. Mixing in the main body of the other localized sources of cold air in flat terrain, or in a
collapsing rectangular current is generally very small. valley that widens at a constant rate, can cause density
This is because the mean gradient Richardson number currents that spread in the radial direction. These have
at its upper boundary (on which mixing depends) is a curved front or head, expanding radially away from
generally greater than 0.25, implying that the mean a nominal central source. Provided the curvature of the
flow is stable and does not generate local mixing (see front is not too large, the speed of the front and the
Kelvin-Helmholtz Instability). Hence the density fluid behind it are given approximately by eqn [2].
within the collapsing rectangular body of dense fluid Such flows may be modeled by axisymmetric
remains largely unaltered until the fluid enters the collapsing bodies of dense fluid, which have some
head, where most of the mixing takes place. The total properties that resemble those of one-dimensional
rectangular area A ( t ) of unmixed dense fluid contin- spreading. Again, there are three stages of spreading:
ually decreases because V I < U H in eqn [2].During this the slumping, inertial-buoyancy balance, and viscous
stage, the length of the dense fluid increases as stages. For a body of dense fluid of initial height do,
(Ag't2)1'3,so that the velocity of the density current radius Ro, and volume Vo, in fluid of overall depth D ,
head decreases with time as (Ag'/t)lI3because of the the initial collapse occurs in the 'slumping' stage,
steady decrease of dl. This continues until the third governed by wave propagation and adverse flow of the
628 DENSITY CURRENTS
radial outflow are not uniform and decrease with

radial distance r. The only significant mixing occurs
behind the head, and for steady flow, dimensional
analysis gives
In the inertio-gravity range, the radial speed of the

head is then given by
Figure 4 Schematic diagram of the ring vortex shortly after a
downdraught reaches the ground. [61
ambient fluid. Here the radial position R of the front
increases at a constant speed in the range Ro < R < R,. Density Currents down Slopes
The shear between the front of the dense fluid and the
When the terrain is not horizontal but slopes down-
ambient fluid produces vorticity that is initially con-
wards, buoyancy acts directly to drive the current
tained within the mixing head of the current, and as
downward, and this is stronger than the indirect effect
the front expands this vorticity increases by vortex
of establishing a horizontal pressure gradient as
stretching (Figure 4). Downbursts that cause this
described above. The downslope buoyancy force is
process may occur on a range of scales, from several
hundred meters or more for thunderstorm down- now g' sin 6 per unit mass of dense fluid, where e is the
angle between the slope and the horizontal. The onset
bursts, and somewhat less than this for microbursts
of a steady source of dense fluid at the top of the slope
(see Microbursts). The latter constitute particular
leads to the formation of a density current head similar
hazards for light aircraft on take-off and landing,
to that described above (see below). In the following
because they may create strong localized circular
current the physics is different, because the flow is now
vortices with radii of 100 m or more.
unstable, unless the slope angle is very small. Mixing
The vortex intensifies as it expands, and initially
now occurs along the whole length of the density
may be much deeper than the following fluid. For a
small initial volume of dense fluid, this expanding current, and not just behind the head. After the passage
vortex ring of dense mixed fluid is all that is produced. of the head an approximately steady flow is estab-
But, for a larger initial volume, as the dense fluid lished, with the buoyancy force being balanced by the
expands and the head moves forward, the initial drag on the bottom surface, and by the entrainment of
vortex progressively breaks up and is subsumed into environmental fluid into the current from above. For a
homogeneous environment, both the size of the head
the mixed layer. New vorticity is continually created at
and the thickness of the following current increase
the head, as for the unidirectional currents, but this is
with downslope distance.
associated with newly mixed fluid and the stretching is
much less than that in the initial vortex. In this Behavior of the Head
second inertial-buoyancy phase (where R > R,), grav-
ity waves maintain approximately uniform depth of For a constant supply of dense fluid of buoyancy gb
the body of dense fluid, and it collapses in the form of and flow rate Qo at the top of the slope, the speed of the
an axisymmetric pillbox, with the volume slowly head V H is almost independent of slope angle and is
decreasing due to mixing and detrainment behind the given approximately by
circular head. If the volume of dense fluid at any given
time is V = zR2dl, and the properties of the front are VH = ~ 15~e, 5 90"
(1.5 0 . 2 ) ( g b ~ ~ )50 [7]
given by eqn [2], the radius R increases as However, in spite of this uniform speed, the size of the
head increases as it moves downslope, at the rate
[41
dH-
_ 0.728
V slowly decreases as dense fluid entering the head is -
181
dx TC
mixed, and this dynamical regime continues until (in
the laboratory) the viscous regime is reached. where H i s the height (or thickness) of the head, x is the
If the source of dense fluid is maintained with downslope distance, and 6 is in radians. The increased
constant volume flux Q1, the initial vortex and head buoyancy at steeper slopes is balanced by increased
form as above but the flow behind it evolves differ- entrainment that acts to keep the head speed constant,
ently, as the depth dl and velocity V I of the following regardless of downslope distance and slope angle.
Entrainment in Downslope Flows

After the passage of the head the following flow is
unstable and mixing with the overlying fluid results.
This process becomes stronger with increasing slope
angle 0. The mixed fluid above the dense layer is also
denser than the environment, and hence it also moves
downslope under gravity but at a reduced speed. The
thickness and volume of this mixed layer both increase
with downslope distance. This combined flow may
now be regarded as a single entity, and the net
downslope buoyancy flux is constant with x and t,
and equal to g’oQo. The net entrainment of environ-
mental fluid into this overall downslope flow may be
Figure 5 The turbulent, mixing interface in the body of a dense
described by an entrainment coefficient E which is a layer flowing down a slope at 6” to the horizontal, in a density-
function of the bulk Richardson number Ri, defined by stratified environment.The dense fluid is the lightlydyed layerclose
to the boundary. Note the filaments extending from it, and the
g‘d cos 6 partially mixed fluid above.
R; = [91
U2
For slope angles less than about 20°, it is appropriate
where U is the mean velocity and d the mean thickness
here to regard the main dense current and the partially
of the total flow (see Kelvin-Helmholtz Instability.
mixed layer above it as separate entities. In laboratory
Turbulence, Two Dimensional). At each level, the
experiments a clear interface is visible between them,
velocity of inflow we of the environmental fluid into
although there are turbulent fluxes across it, as shown
the downflow is given by
schematically in Figure 5 . These flows are governed by
two dimensionless parameters - a bulk Richardson
we = E(Ri)U [lo1 number Ri defined as in eqn [9] but based on the dense
where U is the mean velocity of the downflow. E layer only, and a parameter n/i, defined by
decreases monotonically with increasing Ri, from E =
0.075 at Ri = 0 to very small values for Ri > 0.8. In
these flows, Ri is approximately constant with down- which is a measure of the effect of the ambient
slope distance, and decreases with increasing slope stratification. The dense layer is observed to have
angle. Experiments show that U , and hence w e , are approximately uniform thickness over most of its
also constant, so that the downslope flux Q and the length, but its velocity mostly decreases with down-
lateral spreading increase linearly with distance. slope distance. It loses fluid to the layer above, but also
entrains fluid from it so that its density progressively
Downslope Flows into Stratified Environments
decreases. The fluid that remains reaches its ambient
If the environmental fluid is density stratified, its effect level where it spreads out, at a distance somewhere less
on density currents flowing over a horizontal surface than D below the source. Entrainment into this dense
are mostly limited to its effects on the surrounding layer may be expressed in terms of an entrainment
flow, and the generation of internal waves (see coefficient that depends on Ri in a manner similar to
Buoyancy and Buoyancy Waves: Theory). However, eqn [lo]. There is also a loss of dense fluid to the
for flows down slopes, if ambient stratification is mixing layer, and an exchange between the mixing
present it is a major parameter. If N is the buoyancy layer and the environment. Fluid may leave the mixing
frequency (see Buoyancy and Buoyancy Waves: layer to find its own neutral level, and this occurs
Theory) of the ambient stratification, one may identify continuously along the path of the current. Dense fluid
a depth D below the source, defined by N 2 = g’o/D, from the source may therefore be distributed over a
where the ambient density equals the inflowing den- range of depths, and not all at one level. This
sity. If there were no mixing, all of the inflowing fluid detrainment into the environment depends on both
would be expected to reach and spread horizontally at Ri and M - larger M implies larger detrainment, to the
this level. The downslope speed of the head is again extent that all the dense fluid may be detrained before
approximately constant over most of its distance it reaches its ambient level.
travelled, and scales with eqn [7] above. The main In these flows, detrainment exceeds entrainment.
differences from the homogeneous case concern the The buoyancy force is mostly balanced by bottom
following current, which we next discuss. drag. If this balance is disturbed by, for example, the
630 DEPOSITION
flow encountering a sudden increase in bottom slope, (supercritical) air stream may undergo an internal
the balance is temporally disturbed, and the current hydraulic jump (see Hydraulic Flow) and adjust to a
may vigorously entrain over a short distance in broader, more slowly moving (subcritical) air stream.
reaching a new balanced state, beyond which detrain- These cold offshore katabatic flows may push the ice
ment continues. away from the coast, causing rapid new ice formation
in the open water that takes its place, particularly in
Katabatic Flows Forced by Radiative Cooling
winter.
A prominent source of density currents is cooling of air
near the ground through radiation. This is a common Anabatic Flows Forced by Radiative Heating
occurrence at night, particularly under clear skies, and
causes drainage flows over uneven terrain. In complex Daytime solar heating of sloping terrain causes heat-
terrain, this can occur at a number of source regions, at ing of the adjacent air by convection and radiation.
different altitudes. Depending on the circumstances, This tends to cause flow upslope, although these are
these flows may be quite strong and have depths of not as strong as the drainage flows because they
several hundred meters, or only have speeds of several depend on lateral gradients and involve mixing with
m s - and depths less than 50 m. Nocturnal cooling the overlying fluid for their existence. Nonetheless,
also tends to re-establish the ambient stratification at they are an important part of the diurnal cycle in valley
low levels that is destroyed by convection during the flows influenced by radiative heating and cooling.
day. The processes described above can then cause a They may also be important in promoting wild (bush)-
complex interleaving of stratified layers of air that flow fire propagation up hillsides.
progressively down several slopes, or find their own
environmental level, giving flow in various different
See also
directions at different heights. Since this occurs at
night, good observations of these complex flow Antarctic Climate. Buoyancy and Buoyancy Waves:
patterns are rare. Theory. Gust Fronts. Hydraulic Flow. Kelvin-Helm-
One region where katabatic flows are common and holtz Instability. Microbursts. Turbulence, Two
Dimensional. Vorticity.
are reasonably well observed and understood is
Antarctica (and to a lesser extent, Greenland), where
radiative cooling over the ice sheets sets up a perpetual
drainage regime over the whole continent (see Ant- Further Reading
arctic Climate). Cold air produced over the central
plateau drains off it towards the coast in a layer that is Baines PG (1995) Topographic Effects in Stratified Flows.
several hundred meters thick, and several degrees New York: Cambridge University Press.
Blumen W (ed.) (1990) Atmospheric Processes over Com-
colder than the air above. This effect exists throughout
plex Terrain. Meteorological Monographs No. 45,
the year, and is particularly strong in winter. Speeds American Meteorological Society.
increase as the topographic gradient increases toward Simpson JE (1997) Gravity Currents. Cambridge: Cam-
the coast. Here the intensity of the katabatic flows bridge University Press.
varies according to the local synoptic situation, but Turner JS (1986)Turbulent entrainment: the development of
very strong winds of 40 m s - or more may last for the entrainment assumption, and its application to
days, or even weeks in certain locations. Near the geophysical flows. Journal of Fluid Mechanics 173:
coast, or up to 50 km out to sea, this shallow, intense 431-471.
R J Vong, Oregon State University, Corvallis, OR, turbulence brings gas molecules close to surfaces,
USA whereas aerosol particles are transported both by
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. turbulence and gravitational settling. Deposition oc-
curs if the transported material is then removed from
the atmosphere when it contacts a water, soil, or
Deposition is the process in which mass is transferred vegetated surface. The transfer of gases and particles
from the atmosphere to the Earth’s surface. Trans- to the Earth’s surface is referred to as dry deposition. In
ferred mass can be in the gaseous, particulate, or contrast, wet deposition involves the transfer of
aqueous phase while in the atmosphere. Atmospheric atmospheric mass to these surfaces through the
630 DEPOSITION
flow encountering a sudden increase in bottom slope, (supercritical) air stream may undergo an internal
the balance is temporally disturbed, and the current hydraulic jump (see Hydraulic Flow) and adjust to a
may vigorously entrain over a short distance in broader, more slowly moving (subcritical) air stream.
reaching a new balanced state, beyond which detrain- These cold offshore katabatic flows may push the ice
ment continues. away from the coast, causing rapid new ice formation
in the open water that takes its place, particularly in
Katabatic Flows Forced by Radiative Cooling
winter.
A prominent source of density currents is cooling of air
near the ground through radiation. This is a common Anabatic Flows Forced by Radiative Heating
occurrence at night, particularly under clear skies, and
causes drainage flows over uneven terrain. In complex Daytime solar heating of sloping terrain causes heat-
terrain, this can occur at a number of source regions, at ing of the adjacent air by convection and radiation.
different altitudes. Depending on the circumstances, This tends to cause flow upslope, although these are
these flows may be quite strong and have depths of not as strong as the drainage flows because they
several hundred meters, or only have speeds of several depend on lateral gradients and involve mixing with
m s - and depths less than 50 m. Nocturnal cooling the overlying fluid for their existence. Nonetheless,
also tends to re-establish the ambient stratification at they are an important part of the diurnal cycle in valley
low levels that is destroyed by convection during the flows influenced by radiative heating and cooling.
day. The processes described above can then cause a They may also be important in promoting wild (bush)-
complex interleaving of stratified layers of air that flow fire propagation up hillsides.
progressively down several slopes, or find their own
environmental level, giving flow in various different
See also
directions at different heights. Since this occurs at
night, good observations of these complex flow Antarctic Climate. Buoyancy and Buoyancy Waves:
patterns are rare. Theory. Gust Fronts. Hydraulic Flow. Kelvin-Helm-
One region where katabatic flows are common and holtz Instability. Microbursts. Turbulence, Two
Dimensional. Vorticity.
are reasonably well observed and understood is
Antarctica (and to a lesser extent, Greenland), where
radiative cooling over the ice sheets sets up a perpetual
drainage regime over the whole continent (see Ant- Further Reading
arctic Climate). Cold air produced over the central
plateau drains off it towards the coast in a layer that is Baines PG (1995) Topographic Effects in Stratified Flows.
several hundred meters thick, and several degrees New York: Cambridge University Press.
Blumen W (ed.) (1990) Atmospheric Processes over Com-
colder than the air above. This effect exists throughout
plex Terrain. Meteorological Monographs No. 45,
the year, and is particularly strong in winter. Speeds American Meteorological Society.
increase as the topographic gradient increases toward Simpson JE (1997) Gravity Currents. Cambridge: Cam-
the coast. Here the intensity of the katabatic flows bridge University Press.
varies according to the local synoptic situation, but Turner JS (1986)Turbulent entrainment: the development of
very strong winds of 40 m s - or more may last for the entrainment assumption, and its application to
days, or even weeks in certain locations. Near the geophysical flows. Journal of Fluid Mechanics 173:
coast, or up to 50 km out to sea, this shallow, intense 431-471.
R J Vong, Oregon State University, Corvallis, OR, turbulence brings gas molecules close to surfaces,
USA whereas aerosol particles are transported both by
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. turbulence and gravitational settling. Deposition oc-
curs if the transported material is then removed from
the atmosphere when it contacts a water, soil, or
Deposition is the process in which mass is transferred vegetated surface. The transfer of gases and particles
from the atmosphere to the Earth’s surface. Trans- to the Earth’s surface is referred to as dry deposition. In
ferred mass can be in the gaseous, particulate, or contrast, wet deposition involves the transfer of
aqueous phase while in the atmosphere. Atmospheric atmospheric mass to these surfaces through the
DEPOSITION 631
occurrence of precipitation. Thus, wet deposition the atmospheric concentrations of the gases and
occurs only when precipitation delivers its chemical aerosol that can be subsequently deposited. Air
constituents to the Earth’s surface, but dry deposition concentrations and deposition to the Earth’s surface
occurs continuously, in all types of weather. Cloud are related but are not directly proportional; precip-
water deposition involves the interception of cloud itation scavenging, turbulent transfer, and gravita-
droplets at the surface, primarily by vegetation. tional processes also govern wet, cloud, and dry
Depending on location, surface type, and meteoro- deposition fluxes. There is great variation in the
logy, all three of these deposition pathways can be effectiveness of the many steps that convert air
significant contributors to the transfer of atmospheric concentrations to chemical deposition.
gases and aerosol to Earth’s surface. Deposition is Aqueous phase solute concentrations in precipita-
expressed as a flux density, with typical units of mass tion reflect the air concentrations of the chemical
per surface area over a given time period (e.g., kg- species of interest as well as precipitation formation
SO:- ha-’ y-’). and scavenging processes within and below clouds.
Deposition processes result in a cleansing of the These processes include nucleation of hygroscopic
atmosphere, but this removal of gases and particles aerosol to form cloud droplets, cloud droplet conden-
can result in significant chemical inputs to terrestrial sational growth, mass transfer and solubility of gases
and aquatic ecosystems. The chemical inputs associ- in the drop, below-cloud scavenging of gases and
ated with wet deposition are typically referred to as aerosol, and evaporation of falling precipitation. The
acidic deposition or acid rain. Dry deposition adds to effectiveness of scavenging processes differs for ice and
the sulfur, nitrogen, trace elements, and acidity that liquid-phase precipitation. The ratio of wet deposition
are deposited by precipitation. Cloud water deposi- to precipitation solute concentration is the precipita-
tion, whether viewed as a form of dry deposition (with tion amount (my-’ of rain). Wet chemical deposition
a similar transport mechanisms) or as a form of acid typically is larger at high elevations than in nearby
rain (to which it is similar in that gases and aerosol are low-elevation areas with similar air concentrations
dissolved in hydrometers), represents a third pathway because there is more precipitation. Wet deposition
for chemical deposition to Earth’s surface. networks in many regions of the world are extensive
The chemical inputs that deposition provides to and, thus, may be representative enough to character-
terrestrial and aquatic ecosystems can have impacts on ize wet chemical fluxes from the spatial interpolation
the environment. The potential negative impacts of local measurements. Figures 1and 2 present the wet
include changes in lake chemistry, fish mortality, deposition of SO:- and NOT in the United States for
crop yields, forest health, leaching of soil cations, the year 2000.
weathering of buildings and statues, and corrosion of The relationship between air concentrations and
surfaces. Any positive impacts of chemical deposition dry deposition rates is governed by atmospheric
are due to the potential for deposited nitrogen com- boundary layer turbulence and the physical structure
pounds to fertilize nutrient poor ecosystems. of the receiving surface (see Land-Atmosphere Inter-
Natural and anthropogenic emissions, atmospheric actions: Trace Gas Exchange). Dry deposition of
transport and mixing, and chemical reactions regulate aerosol also depends on gravitational settling. Dry
Figure 1 Wet deposition of sulfate for the year 2000. (National Atmospheric Depotion Program/National Trends Network.)
632 DEPOSITION
Figure 2 Wet deposition of nitrate for the year 2000. (National At1mospheric Depotion Program/National Trends Network.)
deposition measurements are performed at fewer loca- See also

tions than wet deposition measurements and with larger Biogeochemical Cycles: Nitrogen Cycle; Sulfur Cycle.
uncertainties. Dry deposition has typically been deter- Global Change: Biospheric Impacts and Feedbacks.
mined for flat, uniform terrain. Dry deposition fluxes to Land-Atmosphere Interactions: Overview: Trace Gas
forests and tall crops are larger than to nearby soil or Exchange.
water surfaces owing to increased aerodynamic rough-
ness. Any ‘edge effects’ on dry deposition fluxes at
transitions between different surface types are difficult Further Reading
to quantify but may be important.
Cloud water deposition sometimes is referred to as Co-operative Programme for Monitoring and Evaluation of
the Long-range Transmission of Air Pollutants in Europe
‘occult deposition’ because it can occur from a low- (EMEP) web sites. Deposition modeling: http://www.
altitude, nonprecipitating cloud (ie., fog) while a emep.int and Monitoring data: http://www.nilu.no/pro-
nearby rain gauge indicates that no precipitation has jects/ccc/emepdata.html
occurred (it can be viewed as hidden, and thus Dabberdt WF, Lenschow DH, Horst TW,et al. (1993)
‘occult’). Substantial cloud water deposition occurs Atmosphere-surface exchange measurements. Science
almost exclusively to windy, hilly terrain where lifting 260: 1472-1781.
produces condensation, droplets, and frequent cloud- Grennfelt P, Rodhe H, Thornelof E and Wisniewski J. (eds)
iness ( >15% of hours) at the surface. Cloud water (1995) Acid Ruin ’95. Proceedings of the 5th Interna-
deposition depends on turbulence and gravitational tional Conference on Acidic Deposition: Science and
settling to transport mass close to the surface. The Policy. Water Air and Soil Pollution 85(4):1921-2582.
Kaimal JC and Finnigan JJ (1994) Atmospheric Boundary
cloud droplets are intercepted or impact onto vegeta-
Layer Flows. New York: Oxford University Press.
tion and, thus, are removed from the atmosphere Kowalski AS and Vong RJ (2000) Near-surface fluxes of
along with their chemical constituents. Cloud water cloud water evolve vertically. Quarterly Journal of the
deposition occurs simultaneously with wet and dry Royal Meteorological Society 125: 2663-2684.
deposition when precipitating cloud occurs at ground Mohnen VA (1988) The challenge of acid rain. Scientific
level. This occult deposition process has large uncer- American 259(2): 30-38.
tainties as to its magnitude and spatial extent. Sisterson DL, Bowersox VC, Meyers TP, et ul. (1990)
Globally, dry deposition is typically smaller in Deposition monitoring: methods and results (NAPAP
magnitude than wet deposition except in arid regions. Report #6): Acidic Deposition: State of Science and
In the eastern United States, wet and dry deposition are Technology, Vol. I, Final Assessment Reports of the
similar in magnitude for the sulfur compounds that are National Acidic Precipitation Assessment Program.
Washington, DC: US Government Printing Office.
often considered to be environmentally important. In
US Environmental Protection Agency web site. Acid depo-
some coastal and mountainous regions, the magnitude sition progress report http://www.epa,gov/airmarkets/
of cloud water deposition can approach that for wet progress/arpreport/acidrainprogress.pdf
deposition. In the future, it will be important to US National Atmospheric Deposition ProgramNational
improve the understanding of deposition to oceans, Trends Network (NADPNTN) web site. Deposition
heterogeneous landscapes, and complex terrain. monitoring data: http://nadp.sws.uiuc.edu
DESERTSAND DESERTIFICATION 633
Vong RJ and Charlson RJ (1985) The equilibrium pH of a Vong RJ, Sigmon JT and Mueller SF (1991) Cloud water
rain or cloud droplet: a computer based solution for a six- deposition to Appalachian forests. ES&T Critical
component system. Journal of Chemical Education 62: Review. Environmental Science Technology 25( 6):
141-143. 1014-1 021.
See BOW ECHOS AND DERECHO
G Wang, University of Connecticut, Storrs, CT, USA ical productivity is extremely low, and very little
G S Jenkins, Pennsylvania State University, University organic material exists in the soil. Only a sparse
Park, PA, USA vegetation cover, if any, can develop over desert.
Except for rivers that are fed by water from outside the
desert, no stream can survive the rapid evaporation
and soil infiltration in deserts. Owing to the lack of
Introduction leaching by streamflows, the soluble minerals released
from the soil accumulate within the desert basin,
Approximately one-third of the Earth's land surface is leading to highly saline soils as well as saline water in
desert, where rainfall is so scarce that the potential for both the groundwater reservoir and dry lakes.
biological productivity is nearly zero. The inability to About one-third of the Earth's land surface can be
produce food and the harsh climate make the desert a described as desert, with desert-covered lands even
hostile environment for human life. Unfortunately, located in the Dry Valleys of Antarctica. Nonpolar
such unfavorable environments have been expanding deserts are found primarily in Africa, the Americas,
in recent times. As a result of anthropogenic distur- Asia, and Australia. The deserts of Africa include the
bances, a vast area of previously productive land has Sahara in North Africa, and the Kalahari and Namib
been severely degraded, and some places are being deserts in South Africa. The Atacama and Peruvian
converted into desert-like conditions. This process is
deserts are found in the western part of South America,
referred to as desertification. Both desert as a land type and the Great Basin, Mojave, Sonoran, and Chihua-
and desertification as a human-induced process are huan deserts exist in the western part of North
closely related to a region's atmospheric conditions.
America. Asia has the largest number of deserts,
Natural deserts were formed by terrestrial surficial among which the major ones include the Arabian
processes and atmospheric processes interacting over
desert, the Thar, Taklimakan, and Gobi. The Great
long periods of time. Desertification, more than just
Sandy and Simpson deserts cover a considerable part
reducing the land productivity, changes the surface of Australia.
energy balance and hydrological cycle, and leads the The global atmospheric circulation pattern is an
regional climate toward greater aridity. Interplaying important factor influencing the formation of most
with atmospheric processes, desertification may cause deserts. The majority of the world's deserts are located
the formation of man-made desert within a relatively in the vicinity of the tropics of Cancer (23.5"N ) and
short time. Our best understanding of the problem is Capricorn (23.5"S), where strong subsidence of dry
required in order to effectively prevent or remedy such air associated with the descending branch of the
environmental deteriorations. Hadley circulation dissipates cloud cover and sup-
presses rainfall. The Sahara in Africa is an example of
Desert this desert type. Another type of desert is formed
where water vapor in the air is precipitated out before
Deserts are extremely arid areas that generally receive arrival, either because of the long distance from oceans
less than 1mm day-' rainfall on an annual basis or owing to special topographic configurations. Such
(Figure 1)and have very high potential evaporation deserts (e.g., the Sonoran and the Taklimakan) typi-
rates. Because of this aridity, the potential for biolog- cally exist in the middle latitudes. Deserts are also
141-143. 1014-1 021.
634 DESERTS AND DESERTIFICATION
Figure 1 Annual precipitation rates for areas that receive less than 2 mm day-’ of rain. The darkest shaded regions receive less than
0.5 mm day-’ of rain.
formed on the western coastline areas of the conti- the dry, bare, and bright soils. Compared to other land
nents, where the local weather systems are strongly types at the same latitude, the amount of incident solar
affected by the cold ocean currents along the coast radiation reaching the desert surface is much higher
such as the deserts at the western edges of South because of the low cloud fractions, but the amount of
America and South Africa. solar radiation absorbed by the desert is often smaller
because of the higher surface albedo. Consequently,
deserts have a net annual radiative balance (absorbed
Atmospheric Conditions over Deserts solar minus outgoing long-wave radiation (OLR))at
From a meteorological point of view, in addition to the the top of the atmosphere that is approximately zero or
low precipitation rates (Figure 1)and high potential even negative, while other tropical and subtropical
evaporation rates, deserts have a higher surface albedo land types have large positive values of net radiation
than other snow-free or ice-free land types because of (Figure2).The largest negative net annual radiation is
Figure 2 Mean annual net radiative balance at the top of atmosphere. Regions with net radiation less than 20 W m-* are shaded.
The darkest shaded regions have net radiation smaller than -20 W m-2,
observed over the Arabian and Sahara deserts as their lower than -2O"C, in part because of its high elevation
clear-sky albedo is the highest among all deserts. In and the midlatitude location. Regardless of the season,
general, there is a net loss of energy from the autumn surface temperatures of all deserts have extremely
through spring seasons and a net surplus of energy large variations on the diurnal time scale. At the desert
during the spring through autumn seasons over surface, owing to extreme aridity, latent heat flux is
deserts. negligible (Table 1). During the daytime, solar radi-
Subtropical to low-latitude deserts experience hot ation quickly heats up the desert surface because of the
summers and cool winters. For example, in the Sahara lack of latent heat cooling, causing high sensible heat
and Arabian and Australian deserts, the midwinter and ground heat fluxes; at night, when terrestrial
monthly average temperature is around 1O-2O0C, but radiation dominates the surface energy budget, the
the midsummer monthly temperature is in excess of desert surface cools quickly because the dry and
30°C. Midlatitude deserts have relatively hot summers cloudless atmosphere cannot effectively trap the
but cold winters. In Central Asia, the Gobi desert has outgoing terrestrial radiation and reemit back to the
midwinter monthly temperatures lower than -1O"C, land surface. As a result, temperature during the day is
and midsummer daily temperatures greater than 25°C much higher than at night, and such temperature
(Figure 3). For some desert areas in South America differences can easily exceed 20°C on a normal day.
(Patagonia),midwinter monthly temperatures average The greatest diurnal fluctuations occur in deserts near
Figure 3 (A) January and (B) July long-term observed temperature. Temperatures greater than 25°Care shaded. The darkest shading
represents temperatures greater than 30°C.
Table 1 An example of surface energy budget over the deserta period known as the wet season, ranging from several
Net short-wave radiation 600
weeks in arid areas to several months in semi-arid
Net long-wave radiation - 300 areas; dry conditions similar to desert climate persist
Sensible heat flux - 200 during the rest of the year. Owing to the low level of
Latent heat flux Near zero rainfall and the shortness of a favorable growing
Ground heat flux - 100 season, the ecosystem of dry lands feature low biomass
aAverageheat fluxes in W m - 2 at 1300 Local Solar Time observed and poor resilience; the carrying capacity of the land,
over the Empty Quarter of Saudi Arabia (from Figure 11-16 in which is the number of people and animals the land
Pielke(1984)).Apositivevalue representsagaintothesurfaceand can normally support, is very small. These factors
a negative value is a loss from the surface to the atmosphere or to make dry lands subject to anthropogenic desertifica-
the soil depth.
tion. Desertification occurs when the anthropogenic-
ally induced degradation exceeds the ecosystem’s
the Equator. For example, the temperature range restorative capacity.
within one day can be larger than 50°C in the Sahara The major causes of desertification include over-
desert. grazing, overcultivation of marginal land, unsustain-
Deserts around 20-30” latitude in both hemispheres able ‘slash-and-burn’ agricultural practice, logging for
are under the influence of strong subsidence associated fuelwood, and poor irrigation techniques. The dom-
with the descending branch of the Hadley circulation. inant cause varies on the land type. For pasture lands,
In the midtroposphere (500 hPa) sinking air can be livestock overgrazing is the main cause of desertifica-
found over most of the desert regions. During midday tion; removal of vegetation directly by humans for
hours, especially in the summer, intense vertical agriculture and other purposes has caused most of the
turbulent mixing can be found throughout most of desertification in rain-fed lands; incorrect irrigation
the lower troposphere. Consequently, the planetary practice is responsible for the degradation of most
boundary layer (PBL)can extend up to pressure levels irrigated lands. There are several different processes
of 600hPa. There is no rain associated with this associated with the desertification phenomenon: deg-
vertical mixing in the PBL over deserts owing to the radation of vegetation cover; soil erosion; soil com-
lack of moisture in the atmosphere. The intense paction; and waterlogging and salinization in irrigated
turbulence during the daytime is driven by the high lands. One process or a combination of several
sensible heat flux at the surface, which balances the net processes may be involved in the desertification.
radiation absorbed by the desert (Table 1). With the exception of irrigation-induced land deg-
radation, desertification usually starts with the
removal of vegetation cover by humans or livestock.
Overgrazing and excessive fuelwood collection de-
Desertification stroy the protective layer of plants, exposing the top
Desertification refers to land degradation in arid, layer soil to wind and water erosion. As the vegetation
semi-arid, and dry semi-humid areas resulting from cover decreases, soil compaction occurs as a result of
various human impacts. It reduces the land’s potential the livestock trampling and rain drop tapping, which
for biological productivity, and causes conversion of increases the proportion of fine materials in the top
productive lands used for pasture and agriculture into soil, thus accelerating soil erosion. Gradually, the top
desert-like conditions. Desertification can be meas- soil loses its fertility. In addition, compacted soil
ured by the loss of ecosystem productivity it causes, allows less water to infiltrate, limiting the water
and ranges from slight to severe. For example, resources for plant uptake. The reduced fertility and
moderate desertification causes a 10-25% drop in the increased water stress cause the land’s productivity
agricultural productivity, while severe desertification to drop, which then further degrades the vegetation
can result in a productivity loss of 50% or more. cover.
Desertification is not directly related to desert. In In the case of overcultivation of marginal lands, land
most cases, the desert is not expanding continuously is cleared of natural vegetation for agriculture, but is
on the human time scale, although its boundary may then abandoned after crop failure. Without the pro-
advance and retreat in response to interannual climate tection of plants, the loose well-plowed soil easily
variability. The desert has been unjustly thought of by develops into sand dunes under the strong wind
some as the cause of desertification because semi-arid impact. Another destructive agricultural practice is
areas surround desert. Arid and semi-arid areas (or the ‘slashing and burning’ of natural perennial vege-
‘dry lands’) generally receive an annual rainfall less tation, including trees and shrubs, to clear land for
than 600 mm, with high interannual variability. This annual agriculture. As annual agriculture expands, the
small amount of rainfall is concentrated within a short land loses the protection of vegetation, especially
during the dry season, and suffers from severe wind same texture, albedo increases as vegetation cover
erosion. Consequently, biological productivity of the degrades and soil gets drier; degraded vegetation is
land drops and sand dunes start to form. often shorter, which reduces surface roughness.
Salinization and waterlogging lead to the destruc- One of the most direct impacts of desertification on
tion of vegetation cover in irrigated lands. The salinity the hydrological cycle is the decrease of evaporation
is high in both the soil and the stored water in dry and transpiration. Owing to the removal of vegetation
lands, and excessive evaporation and transpiration cover, plant transpiration drops; and the reduced
(collectively referred to as ‘evapotranspiration’) tend surface roughness causes a reduction in turbulent
to accumulate soluble minerals in the upper soil. As a transport, which also limits the rate of evapotranspi-
result, salts in the irrigation water are deposited in the ration. Following desertification, evapotranspiration
root zone. Such soil salinization increases the osmotic (therefore the surface latent heat flux) significantly
pressure of the root medium, leading to a reduction of decreases and the surface Bowen ratio (i.e., the ratio of
the vegetation’s ability to tolerate water stress. Water- sensible heat flux to latent heat flux) increases.
logging occurs when water table rises to the root zone Precipitation, one of the most important climate
as a result of overirrigation - repeated incorrect indicators of dry lands, decreases following desertifi-
irrigation causes the formation of a shallow imperme- cation. The increase of albedo due to desertification
able layer that prevents water from infiltrating down. reduces the amount of net radiation absorbed by the
Under this condition, irrigation water fills all the soil land surface (and therefore the amount of total heat
pores in the root zone and obstructs the gas exchange flux from the surface to the overlying atmosphere).
between soil and air. This causes the buildup of This causes a radiative cooling of the atmosphere and
chemicals harmful to plant growth. Severe salinization induces an additional sinking motion to maintain the
and waterlogging may lead to complete crop failure. atmospheric thermal equilibrium, thus suppressing
In the past several decades, desertification has precipitation. In addition, the increase of Bowen ratio
become a devastating problem of global concern. At reduces the atmospheric humidity, which limits the
least 25% of the world’s land area is affected by moisture source of precipitation.
desertification to some degree. It is especially a The impact of desertification on surface tempera-
problem in Africa, Asia, and South America. Desert- ture involves several competing factors, of which the
ification also affects North America, the Middle East, most important are the increase of surface albedo and
Australia, and some European countries in the Med- the decrease of evapotranspiration. On one hand, the
iterranean area. In terms of severity, the Sudano- decrease of surface net radiation caused by the higher
Sahelian region in Africa is the most affected region of albedo tends to cool the land surface; on the other
the world. More than 70% of the Sudano-Sahelian dry hand, the decrease of latent heat cooling induces a
lands are degraded; of the world’s population that are warming impact on the land surface. During the dry
moderately to severely affected by desertification, season in arid and semi-arid areas, evaporation and
more than 80% reside in this region. In terms of the transpiration are minimal in any case, so their reduc-
affected land area, Asia suffers the most, with more tion is very small. As a result, the albedo effect of
than 1.3 billion hectares of dry lands degraded. desertification is dominant, which cools the land
Globally, the immediate consequence of desertificat- surface. However, during the wet season, latent heat
ion is the reduction of agricultural productivity and becomes an important component of the surface
the resulting threat of famine, which has serious social energy budget, so much so that the effect of an
and economic impact. evapotranspiration reduction outcompetes the albedo
effect and warms up the land surface. Consequently,
the response of surface temperature to desertification
is highly seasonal and is closely related to the hydro-
Climatic Impact of Desertification logical cycle.
Desertification is a detrimental phenomenon not only Desertification can also modify the regional and
because of the resulting loss of land productivity but global climate by changing the atmospheric composi-
also because of its significant impact on regional tion. This takes place mainly as a result of the
climate. Among the geophysical changes accompany- increased dust production from the denuded dry soil
ing desertification, three are considered to directly and in some regions from biomass burning. Desertifi-
affect the water and energy balance at the land surface cation exposes the dry, bare soil to strong wind
and in the overlying atmosphere: increase of the erosion, which significantly increases the atmospheric
surface albedo, decrease of soil moisture, and decrease dust loading in both the desertification zone and the
of the surface roughness. These changes are closely regions downwind. These mineral dusts influence the
related to each other. For example, with soils of the atmospheric radiative transfer directly by scattering
and absorbing solar radiation, and indirectly by lation and thus shows little discernible response to
modifying the optical property and lifetime of clouds. land cover changes, the feedback between the bio-
During ‘slash-and-burn’ agricultural practice, biosphere and the atmosphere may never become posi-
mass burning releases significant amount of ‘green- tive, regardless of the magnitude of desertification. On
house’ gases (e.g., carbon dioxide, methane, nitrous one hand, desertification reduces rainfall (e.g., from PO
oxide) into the atmosphere, and also causes the atmos- to PI = cxP0 in Figure 4, where a< 1);on the other
pheric buildup of reactive trace gases (e.g., carbon hand, owing to the vegetation degradation, water
monoxide and nitric oxide) that play important roles consumption by the plant community also decreases
in controlling atmospheric chemistry. Another climat- (e.g., from DOto D1 = /?Do in Figure 4, where /?<1).
ically important product of biomass burning is black With low sensitivity of climate to land cover changes
carbon, a carbonaceous aerosol that is a strong (i.e., a in Figure 4 close to l . O ) , the water availability
absorber of solar radiation. However, owing to the after desertification, although reduced, may always
complexity of the processes involved, the overall exceed the reduced plant water demand (i.e.,
impact of the atmospheric composition changes due PI > 01). Therefore, the degraded vegetation will
to desertification is not completely understood. tend to grow back, and the biosphere-atmosphere
feedback works to moderate, instead of reinforcing,
the anthropogenic desertification.
Enhancement of Desertificationdue to In regions where the atmospheric condition is highly
sensitive to land cover changes, however, large-scale
Biosphere-Atmosphere Interactions desertification is very likely to trigger a positive
Water plays a determining role in the ecosystem of dry biosphere-atmosphere feedback and become self-per-
lands where most desertification occurs. The desert- petuating. In such regions, when the magnitude of
ification-induced aridity, reflected by a decrease in man-made desertification reaches a certain level, the
both rainfall and specific humidity, may suppress resultant changes in regional climate (rainfall reduction
vegetation growth and further degrade the already in particular) become so dramatic that they start to
disturbed vegetation cover. Thus, enhancement of cause the deterioration of the biota in both the desert-
anthropogenic desertification may result from the ification zone and its neighboring areas; this drought-
biosphere-atmosphere interactions. Whether such a induced vegetation degradation will then reinforce the
positive feedback does take place depends on the initial man-made desertification in changing the region-
regional atmospheric conditions, the magnitude of the al climate, leading to a self-degradation of the land
initial desertification, and the ecological property of surface as well as a persistent drought (Figure 5 ) .
the natural vegetation. Through this mechanism, a new desert may form
The sensitivity of regional climates to desertification following intensive anthropogenic desertification.
varies significantly from place to place. In regions The growth form of the natural vegetation is of the
where precipitation is controlled by large-scale circu- utmost importance in forming the mechanism illus-
Figure 4 Negative and positive biosphere-atmosphere feedbacks associated with desertification. Po and Pi represent the water
availability provided by precipitation before and after man-made land cover changes; Do and D, represent the water demand for
maintenance of the vegetation before and after man-made land cover changes; c( and p are fractions between 0 and 1. When Pi > D1 ,
vegetation tends to recover; when PI < D1,land cover develops toward greater degradation.
Man-made desertification
rainfall
Desert
Rainfall
reduction 1 Persistent
drought 1
Figure 5 Schematic mechanism for the enhancement of desertification through biosphere-atmosphere interactions.
trated by Figure 5. For the biospheric feedback to Desertification and the Sahel Drought:
enhance the initial desertification and cause a persist- A Case Analysis
ent drought, vegetation must be able to carry the
environmental stress from one year to the next. This is Because of biosphere-atmosphere interactions, desert-
obvious for all woody plants such as trees and shrubs, ification may enhance an existing climate trend toward
since their permanent above-ground structures result desiccation, and it may also initiate a change of regional
from multiple-year carbon accumulation. For herba- climate toward greater aridity. Therefore, desertificat-
ceous plants that dominate the biota of dry lands, two ion and drought are often closely related to each other. A
groups of different growth form exist: perennial good example occurs in the Sahel of West Africa, which
grasses have perennating root structures, although is the arid and semi-arid region south of the Sahara
their aerial parts do not survive from one annual cycle desert and spans approximately from 12"N to 18" N.
to another; annual grasses lack any perennating In most parts of the Sahel, natural vegetation is
vegetative structure, of which only the seed bank dominated by savanna, which features a continuous
persists during the dry season. In the beginning of the grass stratum with shrubs, trees, or palms present.
growing season, perennials sprout from their under- During the past several decades, increasing human
ground root structures, while annuals start from seeds. population in this region has led to the overexploita-
The change of the live biomass in the roots of perennial tion of land resources. The 'slash-and-burn' agricul-
plants depends on both the net primary productivity tural practice became one of the most important
and the root turnover rate. Under the condition of causes of desertification, leading to impoverished soil
severe water stress, the low productivity cannot and sand dunes in many places of the Sahel. For
provide sufficient carbon to compensate the root example, Khuwei village in the Sahelian region of
turnover, which causes the root amount to decrease. western Sudan was once well protected by trees, but
Therefore, the perennial grass will have less resource today sand dunes rise against some of the structures to
to start with in the subsequent growing season, thus roof level. Instead of being blown in from the Sahara
carrying information about environmental stress from desert, these dunes have actually formed locally from
one year to the next. The natural enhancement of eroded soil after trees were cut down. Overgrazing and
desertification through the biosphere-atmosphere in- overcultivation are also widespread in the Sahel. As a
teractions can only take effect in regions where the result, the West African Sahel became the region on the
dominant natural vegetation is perennial, whether Earth most severely affected by desertification. Given
herbaceous or woody. It is worth mentioning that this the high sensitivity of the West African climate to land
condition is often satisfied as observations show that cover changes as documented by numerous modeling
perennial grass and shrubs are the dominant plant studies, the intense and extensive desertification in the
species for most dry lands in the tropics. past several decades should leave a fingerprint in the
640 DIURNALCYCLE
regional climate. Specifically, rainfall was expected to See also

decrease, and in fact it did.
Climate: Overview. Drought. Global Change: Surface
After a period of rainfall abundance in the 1950s Temperature Trends. Hydrology: Soil Moisture. Ozone:
and early 1960s, a severe decades-long drought Surface Ozone Effects on Vegetation. Palmer Drought
commenced around 1968. Averaged over the whole Index.
Sahel, the mean rainfall in the last three decades of the
twentieth century is 40-50% less than in the favorable
1950s. During this drought, water of Lake Chad, Further Reading
located in the middle of Sahel, reached its lowest level
on record; some areas near the desert border even Dregne HE (1983)Desertification of Arid Lands. Harwood
experienced a complete loss of rainfall. Proxy data Academic Publishers.
show that West Africa experienced several major Ellis WS (1987)Africa’s Sahel -The stricken land. National
Geographic Magazine 172: 140-179.
droughts in recent history, but none of those was
Glantz MH (1994) Drought follows the Plow. Cambridge:
comparable in severity with the one in the late Cambridge University Press.
twentieth century. The drought-induced famine Mainguet M (1994) Desertification: Natural Background
claimed about 100000 lives in the early 1970s, and and Human Mismanagement. Berlin: Springer-Verlag.
many people have been displaced in order to survive. National Research Council (1992) Grasslands and Grass-
Such a severe drought as observed in West Africa land Sciences in Northern China. Washington DC:
unavoidably causes a certain degree of land degrada- National Academy Press.
tion, which then mingles together with the concurrent Peixoto JP and Oort AH (1992) Physics of Climate. New
anthropogenic desertification. It is almost impossible to York: Springer-Verlag.
distinguish these two types of land degradation without Petrov MP (1976) Deserts of the World. Jerusalem: Keter
detailed observation of vegetation conditions. Owing to Publishing House.
Pielke RA (1984) Mesoscale Meterological Modeling. New
the lack of such observational data before and right after York: Academic Press.
the drought onset, there has been no consensus among Sarmiento G and Monasterio M (1983) Life forms and
scientists whether human activity was the original phenology. In: Bourliere F (ed.)Ecosystem of the World-
trigger of the twentieth century Sahel drought. Howev- Tropical Savannas, pp. 79-108. New York: Elsevier
er, it is generally believed that desertification at least Science.
enhanced this desiccation trend through biosphere- Secretariat of the United Nations Conference on Desertifi-
atmosphere interactions. Using the most up-to-date cation (1977) Desertification: Its causes and conse-
technique in modeling biosphere-atmosphere interac- quences. New York: Pergamon.
tions and considering realistic human impacts, recent UNEP (1992) World Atlas of Desertification. London:
studies have shown that it is conceivable for the Edward Arnold.
anthropogenic desertification in West Africa to cause a Walker AS (1996) Desert: Geology and Resources.
Monograph of US Geological Survey. Denver, CO:
persistent drought similar to what has been observed in USGS.
Sahel. In addition, it was also shown that the interplay Wang G and Eltahir EAB (2000) Ecosystem dynamics
between the drought and desertification may cause the and the Sahel drought. Geophysical Research Letters
northern part of Sahel t o become desert, which appears 27: 795-798 (published by American Geophysical
as a southward expansion of the Sahara. Union).
A Betts, Atmospheric Research, Pittsford, VT, USA wave radiation from the Sun and relatively opaque
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. to the thermal radiation from the Earth. As a result
the surface is warmed by a positive net radiation
balance in the daytime, and cooled by a negative
Introduction
radiation balance at night. The surface temperature
Near the Earth’s surface, many variables have a oscillates almost sinusoidally between a minimum at
characteristic diurnal or daily cycle, driven by the sunrise and a maximum in the afternoon. This is
diurnal cycle of the incoming solar radiation, which referred to as the diurnal cycle of temperature. In
is zero at night and peaks at local noon. The warm seasons, the daily net radiation balance
atmosphere is relatively transparent to the short- is positive, and the daily mean temperature is
141-143. 1014-1 021.
Figure 1 Annual precipitation rates for areas that receive less than 2 mm day-’ of rain. The darkest shaded regions receive less than
0.5 mm day-’ of rain.
formed on the western coastline areas of the conti- the dry, bare, and bright soils. Compared to other land
nents, where the local weather systems are strongly types at the same latitude, the amount of incident solar
affected by the cold ocean currents along the coast radiation reaching the desert surface is much higher
such as the deserts at the western edges of South because of the low cloud fractions, but the amount of
America and South Africa. solar radiation absorbed by the desert is often smaller
because of the higher surface albedo. Consequently,
deserts have a net annual radiative balance (absorbed
Atmospheric Conditions over Deserts solar minus outgoing long-wave radiation (OLR))at
From a meteorological point of view, in addition to the the top of the atmosphere that is approximately zero or
low precipitation rates (Figure 1)and high potential even negative, while other tropical and subtropical
evaporation rates, deserts have a higher surface albedo land types have large positive values of net radiation
than other snow-free or ice-free land types because of (Figure2).The largest negative net annual radiation is
Figure 2 Mean annual net radiative balance at the top of atmosphere. Regions with net radiation less than 20 W m-* are shaded.
The darkest shaded regions have net radiation smaller than -20 W m-2,
observed over the Arabian and Sahara deserts as their lower than -2O"C, in part because of its high elevation
clear-sky albedo is the highest among all deserts. In and the midlatitude location. Regardless of the season,
general, there is a net loss of energy from the autumn surface temperatures of all deserts have extremely
through spring seasons and a net surplus of energy large variations on the diurnal time scale. At the desert
during the spring through autumn seasons over surface, owing to extreme aridity, latent heat flux is
deserts. negligible (Table 1). During the daytime, solar radi-
Subtropical to low-latitude deserts experience hot ation quickly heats up the desert surface because of the
summers and cool winters. For example, in the Sahara lack of latent heat cooling, causing high sensible heat
and Arabian and Australian deserts, the midwinter and ground heat fluxes; at night, when terrestrial
monthly average temperature is around 1O-2O0C, but radiation dominates the surface energy budget, the
the midsummer monthly temperature is in excess of desert surface cools quickly because the dry and
30°C. Midlatitude deserts have relatively hot summers cloudless atmosphere cannot effectively trap the
but cold winters. In Central Asia, the Gobi desert has outgoing terrestrial radiation and reemit back to the
midwinter monthly temperatures lower than -1O"C, land surface. As a result, temperature during the day is
and midsummer daily temperatures greater than 25°C much higher than at night, and such temperature
(Figure 3). For some desert areas in South America differences can easily exceed 20°C on a normal day.
(Patagonia),midwinter monthly temperatures average The greatest diurnal fluctuations occur in deserts near
Figure 3 (A) January and (B) July long-term observed temperature. Temperatures greater than 25°Care shaded. The darkest shading
represents temperatures greater than 30°C.
Table 1 An example of surface energy budget over the deserta period known as the wet season, ranging from several
Net short-wave radiation 600
weeks in arid areas to several months in semi-arid
Net long-wave radiation - 300 areas; dry conditions similar to desert climate persist
Sensible heat flux - 200 during the rest of the year. Owing to the low level of
Latent heat flux Near zero rainfall and the shortness of a favorable growing
Ground heat flux - 100 season, the ecosystem of dry lands feature low biomass
aAverageheat fluxes in W m - 2 at 1300 Local Solar Time observed and poor resilience; the carrying capacity of the land,
over the Empty Quarter of Saudi Arabia (from Figure 11-16 in which is the number of people and animals the land
Pielke(1984)).Apositivevalue representsagaintothesurfaceand can normally support, is very small. These factors
a negative value is a loss from the surface to the atmosphere or to make dry lands subject to anthropogenic desertifica-
the soil depth.
tion. Desertification occurs when the anthropogenic-
ally induced degradation exceeds the ecosystem’s
the Equator. For example, the temperature range restorative capacity.
within one day can be larger than 50°C in the Sahara The major causes of desertification include over-
desert. grazing, overcultivation of marginal land, unsustain-
Deserts around 20-30” latitude in both hemispheres able ‘slash-and-burn’ agricultural practice, logging for
are under the influence of strong subsidence associated fuelwood, and poor irrigation techniques. The dom-
with the descending branch of the Hadley circulation. inant cause varies on the land type. For pasture lands,
In the midtroposphere (500 hPa) sinking air can be livestock overgrazing is the main cause of desertifica-
found over most of the desert regions. During midday tion; removal of vegetation directly by humans for
hours, especially in the summer, intense vertical agriculture and other purposes has caused most of the
turbulent mixing can be found throughout most of desertification in rain-fed lands; incorrect irrigation
the lower troposphere. Consequently, the planetary practice is responsible for the degradation of most
boundary layer (PBL)can extend up to pressure levels irrigated lands. There are several different processes
of 600hPa. There is no rain associated with this associated with the desertification phenomenon: deg-
vertical mixing in the PBL over deserts owing to the radation of vegetation cover; soil erosion; soil com-
lack of moisture in the atmosphere. The intense paction; and waterlogging and salinization in irrigated
turbulence during the daytime is driven by the high lands. One process or a combination of several
sensible heat flux at the surface, which balances the net processes may be involved in the desertification.
radiation absorbed by the desert (Table 1). With the exception of irrigation-induced land deg-
radation, desertification usually starts with the
removal of vegetation cover by humans or livestock.
Overgrazing and excessive fuelwood collection de-
Desertification stroy the protective layer of plants, exposing the top
Desertification refers to land degradation in arid, layer soil to wind and water erosion. As the vegetation
semi-arid, and dry semi-humid areas resulting from cover decreases, soil compaction occurs as a result of
various human impacts. It reduces the land’s potential the livestock trampling and rain drop tapping, which
for biological productivity, and causes conversion of increases the proportion of fine materials in the top
productive lands used for pasture and agriculture into soil, thus accelerating soil erosion. Gradually, the top
desert-like conditions. Desertification can be meas- soil loses its fertility. In addition, compacted soil
ured by the loss of ecosystem productivity it causes, allows less water to infiltrate, limiting the water
and ranges from slight to severe. For example, resources for plant uptake. The reduced fertility and
moderate desertification causes a 10-25% drop in the increased water stress cause the land’s productivity
agricultural productivity, while severe desertification to drop, which then further degrades the vegetation
can result in a productivity loss of 50% or more. cover.
Desertification is not directly related to desert. In In the case of overcultivation of marginal lands, land
most cases, the desert is not expanding continuously is cleared of natural vegetation for agriculture, but is
on the human time scale, although its boundary may then abandoned after crop failure. Without the pro-
advance and retreat in response to interannual climate tection of plants, the loose well-plowed soil easily
variability. The desert has been unjustly thought of by develops into sand dunes under the strong wind
some as the cause of desertification because semi-arid impact. Another destructive agricultural practice is
areas surround desert. Arid and semi-arid areas (or the ‘slashing and burning’ of natural perennial vege-
‘dry lands’) generally receive an annual rainfall less tation, including trees and shrubs, to clear land for
than 600 mm, with high interannual variability. This annual agriculture. As annual agriculture expands, the
small amount of rainfall is concentrated within a short land loses the protection of vegetation, especially
during the dry season, and suffers from severe wind same texture, albedo increases as vegetation cover
erosion. Consequently, biological productivity of the degrades and soil gets drier; degraded vegetation is
land drops and sand dunes start to form. often shorter, which reduces surface roughness.
Salinization and waterlogging lead to the destruc- One of the most direct impacts of desertification on
tion of vegetation cover in irrigated lands. The salinity the hydrological cycle is the decrease of evaporation
is high in both the soil and the stored water in dry and transpiration. Owing to the removal of vegetation
lands, and excessive evaporation and transpiration cover, plant transpiration drops; and the reduced
(collectively referred to as ‘evapotranspiration’) tend surface roughness causes a reduction in turbulent
to accumulate soluble minerals in the upper soil. As a transport, which also limits the rate of evapotranspi-
result, salts in the irrigation water are deposited in the ration. Following desertification, evapotranspiration
root zone. Such soil salinization increases the osmotic (therefore the surface latent heat flux) significantly
pressure of the root medium, leading to a reduction of decreases and the surface Bowen ratio (i.e., the ratio of
the vegetation’s ability to tolerate water stress. Water- sensible heat flux to latent heat flux) increases.
logging occurs when water table rises to the root zone Precipitation, one of the most important climate
as a result of overirrigation - repeated incorrect indicators of dry lands, decreases following desertifi-
irrigation causes the formation of a shallow imperme- cation. The increase of albedo due to desertification
able layer that prevents water from infiltrating down. reduces the amount of net radiation absorbed by the
Under this condition, irrigation water fills all the soil land surface (and therefore the amount of total heat
pores in the root zone and obstructs the gas exchange flux from the surface to the overlying atmosphere).
between soil and air. This causes the buildup of This causes a radiative cooling of the atmosphere and
chemicals harmful to plant growth. Severe salinization induces an additional sinking motion to maintain the
and waterlogging may lead to complete crop failure. atmospheric thermal equilibrium, thus suppressing
In the past several decades, desertification has precipitation. In addition, the increase of Bowen ratio
become a devastating problem of global concern. At reduces the atmospheric humidity, which limits the
least 25% of the world’s land area is affected by moisture source of precipitation.
desertification to some degree. It is especially a The impact of desertification on surface tempera-
problem in Africa, Asia, and South America. Desert- ture involves several competing factors, of which the
ification also affects North America, the Middle East, most important are the increase of surface albedo and
Australia, and some European countries in the Med- the decrease of evapotranspiration. On one hand, the
iterranean area. In terms of severity, the Sudano- decrease of surface net radiation caused by the higher
Sahelian region in Africa is the most affected region of albedo tends to cool the land surface; on the other
the world. More than 70% of the Sudano-Sahelian dry hand, the decrease of latent heat cooling induces a
lands are degraded; of the world’s population that are warming impact on the land surface. During the dry
moderately to severely affected by desertification, season in arid and semi-arid areas, evaporation and
more than 80% reside in this region. In terms of the transpiration are minimal in any case, so their reduc-
affected land area, Asia suffers the most, with more tion is very small. As a result, the albedo effect of
than 1.3 billion hectares of dry lands degraded. desertification is dominant, which cools the land
Globally, the immediate consequence of desertificat- surface. However, during the wet season, latent heat
ion is the reduction of agricultural productivity and becomes an important component of the surface
the resulting threat of famine, which has serious social energy budget, so much so that the effect of an
and economic impact. evapotranspiration reduction outcompetes the albedo
effect and warms up the land surface. Consequently,
the response of surface temperature to desertification
is highly seasonal and is closely related to the hydro-
Climatic Impact of Desertification logical cycle.
Desertification is a detrimental phenomenon not only Desertification can also modify the regional and
because of the resulting loss of land productivity but global climate by changing the atmospheric composi-
also because of its significant impact on regional tion. This takes place mainly as a result of the
climate. Among the geophysical changes accompany- increased dust production from the denuded dry soil
ing desertification, three are considered to directly and in some regions from biomass burning. Desertifi-
affect the water and energy balance at the land surface cation exposes the dry, bare soil to strong wind
and in the overlying atmosphere: increase of the erosion, which significantly increases the atmospheric
surface albedo, decrease of soil moisture, and decrease dust loading in both the desertification zone and the
of the surface roughness. These changes are closely regions downwind. These mineral dusts influence the
related to each other. For example, with soils of the atmospheric radiative transfer directly by scattering
and absorbing solar radiation, and indirectly by lation and thus shows little discernible response to
modifying the optical property and lifetime of clouds. land cover changes, the feedback between the bio-
During ‘slash-and-burn’ agricultural practice, biosphere and the atmosphere may never become posi-
mass burning releases significant amount of ‘green- tive, regardless of the magnitude of desertification. On
house’ gases (e.g., carbon dioxide, methane, nitrous one hand, desertification reduces rainfall (e.g., from PO
oxide) into the atmosphere, and also causes the atmos- to PI = cxP0 in Figure 4, where a< 1);on the other
pheric buildup of reactive trace gases (e.g., carbon hand, owing to the vegetation degradation, water
monoxide and nitric oxide) that play important roles consumption by the plant community also decreases
in controlling atmospheric chemistry. Another climat- (e.g., from DOto D1 = /?Do in Figure 4, where /?<1).
ically important product of biomass burning is black With low sensitivity of climate to land cover changes
carbon, a carbonaceous aerosol that is a strong (i.e., a in Figure 4 close to l . O ) , the water availability
absorber of solar radiation. However, owing to the after desertification, although reduced, may always
complexity of the processes involved, the overall exceed the reduced plant water demand (i.e.,
impact of the atmospheric composition changes due PI > 01). Therefore, the degraded vegetation will
to desertification is not completely understood. tend to grow back, and the biosphere-atmosphere
feedback works to moderate, instead of reinforcing,
the anthropogenic desertification.
Enhancement of Desertificationdue to In regions where the atmospheric condition is highly
sensitive to land cover changes, however, large-scale
Biosphere-Atmosphere Interactions desertification is very likely to trigger a positive
Water plays a determining role in the ecosystem of dry biosphere-atmosphere feedback and become self-per-
lands where most desertification occurs. The desert- petuating. In such regions, when the magnitude of
ification-induced aridity, reflected by a decrease in man-made desertification reaches a certain level, the
both rainfall and specific humidity, may suppress resultant changes in regional climate (rainfall reduction
vegetation growth and further degrade the already in particular) become so dramatic that they start to
disturbed vegetation cover. Thus, enhancement of cause the deterioration of the biota in both the desert-
anthropogenic desertification may result from the ification zone and its neighboring areas; this drought-
biosphere-atmosphere interactions. Whether such a induced vegetation degradation will then reinforce the
positive feedback does take place depends on the initial man-made desertification in changing the region-
regional atmospheric conditions, the magnitude of the al climate, leading to a self-degradation of the land
initial desertification, and the ecological property of surface as well as a persistent drought (Figure 5 ) .
the natural vegetation. Through this mechanism, a new desert may form
The sensitivity of regional climates to desertification following intensive anthropogenic desertification.
varies significantly from place to place. In regions The growth form of the natural vegetation is of the
where precipitation is controlled by large-scale circu- utmost importance in forming the mechanism illus-
Figure 4 Negative and positive biosphere-atmosphere feedbacks associated with desertification. Po and Pi represent the water
availability provided by precipitation before and after man-made land cover changes; Do and D, represent the water demand for
maintenance of the vegetation before and after man-made land cover changes; c( and p are fractions between 0 and 1. When Pi > D1 ,
vegetation tends to recover; when PI < D1,land cover develops toward greater degradation.
Man-made desertification
rainfall
Desert
Rainfall
reduction 1 Persistent
drought 1
Figure 5 Schematic mechanism for the enhancement of desertification through biosphere-atmosphere interactions.
trated by Figure 5. For the biospheric feedback to Desertification and the Sahel Drought:
enhance the initial desertification and cause a persist- A Case Analysis
ent drought, vegetation must be able to carry the
environmental stress from one year to the next. This is Because of biosphere-atmosphere interactions, desert-
obvious for all woody plants such as trees and shrubs, ification may enhance an existing climate trend toward
since their permanent above-ground structures result desiccation, and it may also initiate a change of regional
from multiple-year carbon accumulation. For herba- climate toward greater aridity. Therefore, desertificat-
ceous plants that dominate the biota of dry lands, two ion and drought are often closely related to each other. A
groups of different growth form exist: perennial good example occurs in the Sahel of West Africa, which
grasses have perennating root structures, although is the arid and semi-arid region south of the Sahara
their aerial parts do not survive from one annual cycle desert and spans approximately from 12"N to 18" N.
to another; annual grasses lack any perennating In most parts of the Sahel, natural vegetation is
vegetative structure, of which only the seed bank dominated by savanna, which features a continuous
persists during the dry season. In the beginning of the grass stratum with shrubs, trees, or palms present.
growing season, perennials sprout from their under- During the past several decades, increasing human
ground root structures, while annuals start from seeds. population in this region has led to the overexploita-
The change of the live biomass in the roots of perennial tion of land resources. The 'slash-and-burn' agricul-
plants depends on both the net primary productivity tural practice became one of the most important
and the root turnover rate. Under the condition of causes of desertification, leading to impoverished soil
severe water stress, the low productivity cannot and sand dunes in many places of the Sahel. For
provide sufficient carbon to compensate the root example, Khuwei village in the Sahelian region of
turnover, which causes the root amount to decrease. western Sudan was once well protected by trees, but
Therefore, the perennial grass will have less resource today sand dunes rise against some of the structures to
to start with in the subsequent growing season, thus roof level. Instead of being blown in from the Sahara
carrying information about environmental stress from desert, these dunes have actually formed locally from
one year to the next. The natural enhancement of eroded soil after trees were cut down. Overgrazing and
desertification through the biosphere-atmosphere in- overcultivation are also widespread in the Sahel. As a
teractions can only take effect in regions where the result, the West African Sahel became the region on the
dominant natural vegetation is perennial, whether Earth most severely affected by desertification. Given
herbaceous or woody. It is worth mentioning that this the high sensitivity of the West African climate to land
condition is often satisfied as observations show that cover changes as documented by numerous modeling
perennial grass and shrubs are the dominant plant studies, the intense and extensive desertification in the
species for most dry lands in the tropics. past several decades should leave a fingerprint in the
640 DIURNALCYCLE

Introduction
640 DIURNALCYCLE

Introduction
DIURNAL CYCLE 641
determined by the daily mean surface energy balance, 700 45 4 FIFE1987 A I

which involves not only the short- and long-wave 600
500
radiation components, but also heat transfers to the
atmosphere.
E 400
5 300
s
v
The magnitude of this diurnal range of temperature is 6 200 2

determined by many factors, which we will discuss. The 100
0
most important are the nature of the underlying surface,
0 3 6 91215182124 0 3 6 9 12151821 24
whether land or water, and the coupling to the atmos- (A) LST (D) LST
phere above. The phase change of water, particularly
evaporation and condensation, plays an important role
in moderating the diurnal range of temperature, because ...
of the large latent heat of vaporization. (In cold climates

the freezing and thawing of the soil is also important on 2 150
the seasonal time scale.)
50
Over the ocean (and large lakes), the diurnal 0 .
temperature range is small, because the incoming -50
0 3 6 9 1215182124
20
0 3 6 9 1215182124
solar energy is mixed downward into an ocean ‘mixed
layer’, which is usually tens of meters deep. One day of
solar heating will warm a layer of water 50 m deep by 100 , I
90
less than 0.1 K, because of its large thermal capacity.
-8
,,---\
80
Only in light winds, when the downward mixing is 70
small, does the diurnal range of sea surface tempera-
v
60
I
a: 50
ture reach 1K. On time scales longer than the diurnal, 40
evaporation of water primarily balances the surface 30
net radiation budget. 20
0 3 6 9 12151821 24
Over land, only a small fraction (<20%)of the net (C) LST (F) LST
radiation at the surface is conducted downward in
daytime, or stored by warming trees on the surface for Figure 1 Diurnal cycle of net radiation (&), sensible heat flux ( H )
and latent heat flux (LE) (A-C); surface temperature, air temper-
example. As a result, the surface temperature rises ature and relative humidity (D-F), stratified by percent volumetric
rapidly after sunrise, until near balance is achieved soil moisture in first lOcm of the soil.
between the net radiation and the direct transport of
heat to the atmosphere (referred to as the sensible heat
flux) and evaporation of water (or transpiration from balance can be written as
plants), referred to as the latent heat flux. If the surface
is a desert then the daytime temperature rise is large, R, =H + LE + G 111
but if water is readily available for transpiration then
the daytime rise of temperature is greatly reduced, where G is the storage in the ground and vegetation,
because most of the net radiation goes into the latent which we do not show. In addition a small amount of
heat of vaporization. The surface sensible and latent energy goes into photosynthesis, which again we do
heat fluxes have a large diurnal cycle, with a peak nearnot show. The time access is local solar time, which is
local noon, as they are driven primarily by the U T C - 6 hours. The data have been grouped and
incoming solar radiation. The surface temperature averaged based on the percentage of soil moisture
peaks a little later in the afternoon, when the surface (SM) in the first 10 cm of soil, so that there are three
sensible heat flux goes negative as the surface cools. curves (each an average of about 10 days), represent-
ing dry, medium, and wet soils. Panel A shows that the
mean net radiation on these sunny days is very similar.
Dependence of Diurnal Cycle on However, because soil moisture is a major control on
evaporation, the partition of the net radiation into
Evaporation
sensible and latent heat is very different. When the soil
Figure 1 illustrates this diurnal variation using data is wet, the latent heat flux (or ‘evaporative energy’
from sunny days in midsummer during a 1987 field flux) is about three times the sensible heat flux,
experiment (with the acronym FIFE) conducted over whereas when the soil is dry these two fluxes are
grassland near Manhattan, Kansas. Panels on the left nearly equal. Panels on the right side show the
show net radiation, R , (A), sensible heat flux, H (B), response to the different surface forcing. Panel D
and latent heat flux, LE (C). The surface energy shows the surface temperature (measured by an
642 DIURNAL CYCLE
infrared radiation thermometer, mounted on a tower 24 BOREAS 1994 302

and pointed downward at the grass). Although R , is 22 300
the same, on days when the soil is dry and water is not 20 298
readily available for evaporation, the surface gets very 18 296
hot, as warm as 44°C near noon. This warm surface 16 294
temperature drives the large sensible heat flux H and 14 292
heats the air above the surface. The diurnal range of E 12
290 Q
Y
288 m
the surface temperature is more than 20°C on these 10
286
days, while for the air at 2 m above the surface in panel 8
284
E, the diurnal range is only 12°C. As soil moisture 6
282
increases, the daily maximum surface and air temper- 4
280
ature decrease. Panels D and E are similar, except that 2- ........
278
the amplitude of the surface temperature is larger than 0 I " I " I " I " I " I " I "
that of the air temperature. Both are related to the
sensible heat flux H . Note that the air temperature has a (A) LST (UTC-7)
broad afternoon maximum, because H is upward as
long as the surface is warmer than the air. The surface 100
-26 May
temperature falls below the air temperature only in late -----_31 May
afternoon, H then changes sign, and at night the surface
is cooler than the air. Panel F shows the diurnal cycle of
relative humidity (RH) as a percent. Over the wetter
70
soils, the RH of the air at 2 m reaches 85% before h
sunrise, and falls in the daytime as the surface and air 6 60

I
warms. The fall of RH is smallest on the days with the [r
50
greatest evaporation, LE. When evaporation is reduced
because the soil is dry, daytime RH falls as low as 30%, 40
and even at night only reaches 72% at sunrise.
30
20
Coupling Between the Surface Diurnal 0 3 6 9 12 15 18 21 24
Cycle and the Atmospheric Mixed (8) LST (UTC-7)
Layer Figure2 Diurnal cycle of temperature, above and below a boreal

forest canopy (A), and of relative humidity above the canopy (6)
for
As the land surface is heated during the daytime, a dry two days in May 1994.
convective boundary layer grows in depth. This is
called the mixed layer, because the turbulent dry the atmosphere above is more stable (see Figure 3 ) .
convection rapidly stirs the layer t o near-neutral The air under the canopy becomes effectively de-
buoyancy and near-constant water vapor mixing ratio. coupled from the atmosphere above and the stable
The diurnal cycle of the surface and the mixed layer are temperature gradient across the canopy at night reaches
tightly coupled. As a result the preexisting atmospher- 7 K . There is very little evaporation from either the
ic structure above the surface at sunrise has a consid- forest or the cold lakes at this time in spring. The lower
erable impact on the daytime diurnal cycle, as panel shows RH measurements above the canopy. In the
illustrated in the following figures using surface and late afternoon, R H falls as low as 20% on 31 May.
sounding data collected over the boreal forest in Before sunrise on this day, RH above the canopy reaches
Saskatchewan during the Boreal Ecosystem-Atmos- 90% as Tu falls to a minimum of 4°C. RH was not
phere Study (BOREAS) in 1994. Figure 2 shows the measured below the canopy, but the temperatures there
surface diurnal cycle for two days in spring. The upper are cold enough to saturate the air in the hours before
panel shows for each day the temperature at two sunrise. The dew point is often used to estimate
levels, an upper level Tu which is at 21 m, about 5 m minimum nighttime temperatures at the surface.
above the canopy of a jack pine forest, and a lower The right-hand scale of Figure 2A shows the corre-
level about 5 m above the forest floor. O n both days the sponding dry potential temperature, which is defined as
surface cools strongly at night and rises steeply after
sunrise with a greater diurnal range than in Figure 1.
The diurnal range under the canopy is larger than
above it. At night on 26 May the winds are lighter, and
DIURNAL CYCLE 643
where p is the surface pressure (here about 950 hPa, while on 31 May it occurs at 7.8 h, when 8 = 289K,
since the observation site is about 500m above sea and on this day the change is smaller.
level). The potential temperature, 8, is useful as a Figure 3 shows sequences of seven profiles of
variable because it allows us to compare surface and potential temperature in the lower troposphere, meas-
atmosphere above. During the daytime the boundary ured by rawinsonde ascents, nominally every 2 hours
layer above the surface is mixed to almost constant from sunrise to late afternoon on the two days. Figure 3A
potential temperature (see Figure 3 ) . The strong shows at sunrise (04.17LST, solid) a cold (stable)
radiative cooling of the surface at night generates a surface layer only about 25 hPa deep (200 m), with a
stable layer close to the ground, typically only a few deep layer above of constant 8, which is the residual or
hundred meters deep. About 3 to 4 hours after sunrise, 'fossil' mixed layer from the previous day. At the
the surface has warmed enough to remove this stable surface the temperature warms rapidly, as the surface
surface layer and reconnect to a deeper layer. When sensible heat flux is trapped in this shallow surface layer.
this happens, the rate of rise of temperature and fall of The profile at 08.24 LST shows a mixed layer with 8 =
RH decrease sharply. In Figure 2, this occurs on 26 294.5K to 890hPa. Shortly afterwards, when the
May at a local time of 8.8 h, when 8 reaches 296K; surface potential temperature reaches 8 = 296 K, the
new growing boundary layer merges with the deep
residual mixed layer. From then on, the surface and
mixed layer warm much more slowly, as seen in Figure 2.
Even though H exceeds 300 W m - 2 at all the forest
sites for several hours around local noon (not shown),
this large heat flux is distributed through a deep layer.
Figure 3B shows the time sequence on 31 May. Note
that at sunrise (solid),the profile is quite different from
750
m that on 26 May. Instead of a deep layer of constant 8,
produced by dry convection the previous day (a so-
p 800
-04.17LST called dry adiabatic structure), there is a layer from
920 to 6SOhPa in which 8 increases steadily with
height. In fact this layer was produced by showers the
previous evening (and it has a so-called wet adiabatic
structure). The change in slope of the early morning
profile at 920 hPa is at 8 = 289 K, and hence we see on
Figure 2 a change in the rate of warming, once the
280 290 300 310 320 surface reaches this potential temperature. This
(4 0 (K) change of slope is more dramatic on 26 May because
the change in the vertical profile is also greater. On 31
May, the mixed layer grows steadily all day until it is
300 hPa deep (about 3000 m) in the late afternoon. On
both these days, there is some broken cumulus cover in
600 the afternoon at the top of the mixed layer. The rapid
warming on 31 May between 500 and 600 hPa is
650
related to the lowering and change in structure of a
700 powerful jet stream above, not by surface processes.
a"
5. 750
P
-04.18LST
Diurnal Cycle of COP
800
The diurnal cycle of the solar radiation drives a diurnal
850 cycle in COZ through photosynthesis and respiration
in plants. Figure 4 shows the mean diurnal cycle over a
900
young jack pine canopy (about 5-6m tall) near
950 Thompson, Manitoba, from the 1996 BOREAS
experiment for the three months June, August, and
280 290 300 310 320
October. During the summer months, COZ decreases
(B) 6 (K) during the daylight hours as it is taken up in photo-
Figure 3 Profiles of potential temperature in the lower tropo- synthesis, and increases at night as it is released by
sphere on 26 and 31 May. respiration from both plants and soil. The amplitude
644 DOWNSLOPEWINDS
.- Above jack pine forest See also

Manitoba 1996
Air-Sea Interaction: Momentum, Heat and Vapor Flux-
-June es; Sea Surface Temperature. Boundary Layers: Ocean
_----August
October
Mixed Layer: Overview; Surface Layer. Carbon Dioxide.
Climate: Overview. Climate Variability: Seasonal to
370
lnterannual Variability. Energy Balance Model, Sur-
face. Hydrology: Ground and Surface Water: Soil Mois-
ture. Land-Sea Breeze.
350 4 Further Reading

Arya SP (1988) Introduction to Micrometeorology. New
York: Academic Press.
Betts AK and Ball JH (1995)The FIFE diurnal cycle climate.
Journal of Geophysical Research 100: 25679-25693.
Budyko MI (1974) Climate and Life. New York: Academic
0 3 6 9 12 15 18 21 24 Press.
24-hour day Curry JA and Webster PJ (1999) Thermodynamics of
Atmospheres and Oceans. New York: Academic Press.
Figure 4 Monthly mean diurnal cycle of COP for June, August, Garratt JR (1992) The Atmospheric Boundary Layer.
and October for a boreal jack pine site. (Data from Dr JH Cambridge: Cambridge University Press.
McCaughey, Queens University, Kingston, Ontario.)
Geiger R (1965)The Climate Near the Ground. Cambridge:
Harvard University Press.
of the diurnal cycle increases from June t o August, but Hartman DL (1994) Global Physical Climatology. New
the monthly mean decreases as there is a net COZ York: Academic Press.
uptake by the entire Northern Hemisphere. By Octo- Simpson JE (1994)Sea Breeze and Local Wind. Cambridge:
ber of this year, however, the diurnal cycle is very small, Cambridge University Press.
because temperatures have dropped so low that both Stull RB (1988) A n Introduction to Boundary Layer
photosynthesis and respiration have almost ceased. Meteorology. Dordrecht: Kluwer.
Dale R Durran, University of Washington, Seattle, WA, ( 2 1 6 k m h - I ) . An anemometer trace recorded at the
USA National Center for Atmospheric Research in Boulder,
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. CO, during a strong chinook is shown in Figure 1.
240
Introduction
Very strong surface winds sometimes develop when air
flows over a high mountain ridge with a steep lee slope.
Such winds are known t o occur at many locations
.
E 180
E
?5
throughout the middle latitudes. Local names for these U
g 120
winds include the Alpine foehn, the Rocky Mountain (I
v)
chinook, the Croatian bora, the Santa Ana in southern U

K
California, and the Argentine zonda. These winds are 3 60

collectively referred to as downslope winds. Down-
slope winds in the lee of major mountain barriers can 0
approach hurricane strength. (By definition, hurri- 02:OO 0130 01:OO 00:30
canes are storms with sustained winds of at least Time (MST)
32 m s - (115 km h - ').) Every few years, for exam- Figure 1 Anemometer trace recorded at the National Center for
ple, the eastern slope of the Colorado Front Range Atmospheric Research during the onset of the 17 January 1982
(part of the Rocky Mountains) experiences a damag- Boulder windstorm. Time reads right to left. (Reproduced with
ing windstorm with peak gusts as high as 60 m s - permission from Durran (1990, Figure 4.1 l).)
644 DOWNSLOPEWINDS
.- Above jack pine forest See also

Manitoba 1996
Air-Sea Interaction: Momentum, Heat and Vapor Flux-
-June es; Sea Surface Temperature. Boundary Layers: Ocean
_----August
October
Mixed Layer: Overview; Surface Layer. Carbon Dioxide.
Climate: Overview. Climate Variability: Seasonal to
370
lnterannual Variability. Energy Balance Model, Sur-
face. Hydrology: Ground and Surface Water: Soil Mois-
ture. Land-Sea Breeze.
350 4 Further Reading

Arya SP (1988) Introduction to Micrometeorology. New
York: Academic Press.
Betts AK and Ball JH (1995)The FIFE diurnal cycle climate.
Journal of Geophysical Research 100: 25679-25693.
Budyko MI (1974) Climate and Life. New York: Academic
0 3 6 9 12 15 18 21 24 Press.
24-hour day Curry JA and Webster PJ (1999) Thermodynamics of
Atmospheres and Oceans. New York: Academic Press.
Figure 4 Monthly mean diurnal cycle of COP for June, August, Garratt JR (1992) The Atmospheric Boundary Layer.
and October for a boreal jack pine site. (Data from Dr JH Cambridge: Cambridge University Press.
McCaughey, Queens University, Kingston, Ontario.)
Geiger R (1965)The Climate Near the Ground. Cambridge:
Harvard University Press.
of the diurnal cycle increases from June t o August, but Hartman DL (1994) Global Physical Climatology. New
the monthly mean decreases as there is a net COZ York: Academic Press.
uptake by the entire Northern Hemisphere. By Octo- Simpson JE (1994)Sea Breeze and Local Wind. Cambridge:
ber of this year, however, the diurnal cycle is very small, Cambridge University Press.
because temperatures have dropped so low that both Stull RB (1988) A n Introduction to Boundary Layer
photosynthesis and respiration have almost ceased. Meteorology. Dordrecht: Kluwer.
Dale R Durran, University of Washington, Seattle, WA, ( 2 1 6 k m h - I ) . An anemometer trace recorded at the
USA National Center for Atmospheric Research in Boulder,
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. CO, during a strong chinook is shown in Figure 1.
240
Introduction
Very strong surface winds sometimes develop when air
flows over a high mountain ridge with a steep lee slope.
Such winds are known t o occur at many locations
.
E 180
E
?5
throughout the middle latitudes. Local names for these U
g 120
winds include the Alpine foehn, the Rocky Mountain (I
v)
chinook, the Croatian bora, the Santa Ana in southern U

K
California, and the Argentine zonda. These winds are 3 60

collectively referred to as downslope winds. Down-
slope winds in the lee of major mountain barriers can 0
approach hurricane strength. (By definition, hurri- 02:OO 0130 01:OO 00:30
canes are storms with sustained winds of at least Time (MST)
32 m s - (115 km h - ').) Every few years, for exam- Figure 1 Anemometer trace recorded at the National Center for
ple, the eastern slope of the Colorado Front Range Atmospheric Research during the onset of the 17 January 1982
(part of the Rocky Mountains) experiences a damag- Boulder windstorm. Time reads right to left. (Reproduced with
ing windstorm with peak gusts as high as 60 m s - permission from Durran (1990, Figure 4.1 l).)
DOWNSLOPE WINDS 645
In modern meteorological usage, downslope winds flow, which is moving from left to right. (The isen-
are distinguished from katabatic winds by the dynam- tropes would be exactly identical to streamlines if the
ical processes driving each flow. Katabatic winds flow were steady, inviscid, adiabatic, and two dimen-
usually refer to shallow gravity currents generated by sional.) A large-amplitude mountain wave is clearly
the cooling of surface air over sloping terrain. Down- visible in the potential temperature field just to the lee
slope winds usually refer to winds generated as a of the continental divide. The apparent horizontal
deeper layer of air is forced over topography. In displacement of the wave trough at upper levels from
contrast to katabatic winds, the diabatic cooling of air its position at low levels is due to a 2 h difference
in contact with a cold surface plays no essential role in between the times at which observations were collect-
the dynamics of downslope winds. ed in the upper and lower flight levels. Also apparent in
In most downslope wind events (including the Figure 2 is a layer of enhanced static stability around
typical foehn and chinook), the onset of the downslope the 550 mb level in the upstream flow. When intense
wind is accompanied by an increase in the surface downslope winds develop in a deep cross-mountain
temperature and a drop in the dew point. Whereas the flow, strong mountain waves and low-level stable
area of violent wind is limited to a relatively narrow layers similar to those shown in Figure 2 are usually
swath along and adjacent to the lee slope, the warmer present.
drier air mass can extend much further downstream. The connection between mountain waves and
Nevertheless, in some cases the upstream conditions strong downslope winds is less apparent in situations
may be so cold, and the initial downstream conditions where the cross-mountain wind component reverses
sufficiently warm, that the onset of a downslope wind with height at some level in the middle or lower
brings a drop in temperature. The most well-known troposphere, as is often the case in the Croatian bora or
example of this type of cold downslope wind is the when strong winds blow from the east down the
Croatian bora. Despite the difference in the evolution western slopes of the Wasatch mountains in Utah.
of the surface temperature, there does not appear to be Contours of the potential temperature observed dur-
any significant dynamical distinction between the ing a moderate bora along a cross-section through
processes responsible for the development of high Senj, Croatia, are shown in Figure 3. The flow in this
downslope winds in cold and warm events. example is from right to left. A low-level inversion is
Contours of the potential temperature observed on once again apparent upstream of the mountains;
11 January 1972 during an intense downslope wind- however, no significant wave activity is present above
storm are plotted in the vertical cross-section through the 3 km level. In this case, the upstream inversion is
Boulder, COY shown in Figure 2 . These contours coincident with a region of strong vertical wind shear
provide a rough indication of the streamlines in the in which the cross-mountain wind component reverses
Sabreliner. Queen air 11 January 1972
mb I +++++Tu;buience encountered
OK
.......................... L,..,
....
-110 -lW -90 -80 -70 -60 -50 -40 -30 -20 -10 0 10 20 30 40 50 6
.j5
0 0
Figure 2 Cross-section of potential temperature along an east-west line through Boulder, GO, from aircraft observations collected on
11 January 1972. The heavy dashed line separates data collected by two aircraft at different times. Flight tracks are indicated by the dotted
lines: those segments along which significant turbulence was encountered are denoted by pluses. (Reproduced with permission from Lilly
(1978, Figure 7).)
646 DOWNSLOPE WINDS
=Potential temperature (“K) 1 4
-
Zagreb
I
-1 00 -50
I
0
I I
50
v
I
100 150
SL
Distance (km)
Figure 3 Cross-section of potential temperature along a north-east-south-west line through Zagreb and Senj, Croatia. (Reproduced
with permission from Smith (1987, Figure 9b).)
direction. The level at which the cross-mountain wind by the horizontal momentum equation
component drops to zero is a critical level for steady
two-dimensional mountain waves, and any gravity
waves triggered by the mountain break down
and dissipate as they approach this critical level. (A and the mass continuity equation
critical level is a level at which the phase speed of a
wave equals the speed and direction of the basic -auD
=o PI
state flow.) ax
where x is the horizontal coordinate directed perpen-
dicular to the ridgeline, u is the velocity in the x
The Hydraulic Analogue direction, D is the thickness of the fluid, and h is the
local height of the obstacle. Using eqn [2] to substitute
The dynamics governing the development of strong for au/ax into eqn [l]yields
downslope winds in the atmosphere are analogous to
those governing the rapid increase in speed that occurs ah
ax ( D + h ) = -
2 a
when water flowing over a rock in a river undergoes a (1 - F- )- ax 131
transition from a relatively slow velocity upstream to a
thin layer of high-velocity fluid over the downstream Here
face. In such circumstances, a turbulent hydraulic U
F=-
jump often develops downstream of the rock at the
point where the high-speed flow decelerates back to
vm [41
the ambient velocity of the river. Since the fundamen- is the Froude number, which is the ratio of the local
tal processes responsible for the rapid acceleration of flow speed to the local phase speed of a linear shallow
water flowing over a rock can be explained more water gravity wave.
simply than those which govern downslope winds in According to eqn [3], the magnitude of the Froude
the atmosphere, let us begin by considering the number determines whether the free surface rises or
hydraulic model for a shallow layer of water flowing falls as the fluid ascends the upstream slope of the
over an obstacle in an open channel. obstacle. The case F > 1,known as supercritical flow,
Suppose a homogeneous fluid, such as fresh water, is is shown in Figure 4A; the fluid thickens and slows as it
flowing over a ridge-like obstacle. Assuming the flow is passes over the top of the obstacle, and it reaches its
steady and that there are no variations in the coordi- minimum speed at the crest. The accelerations expe-
nate direction parallel to the ridge axis, and making rienced by the fluid are qualitatively similar to those
the hydrostatic approximation, the flow is governed experienced by a hockey puck traversing a frictionless
DOWNSLOPE WINDS 647
Thus in steady open-channel hydraulic flow, acceler-

ation always opposes the pressure gradient force due
to changes in fluid depth. Furthermore, F2 may be
interpreted as the ratio of the magnitude of the
acceleration to the magnitude of the pressure gradient
force generated by changes in the fluid depth. In
supercritical flow ( F > 1) acceleration dominates the
pressure gradient force and the three-way balance in
eqn [l]is satisfied such that fluid parcels ascending the
upstream slope decelerate as they do work against
gravity.
Before discussing the subcritical case, it is helpful to
recast the discussion in terms of the conversions
between kinetic energy (KE)and potential energy (PE).
+ +
Equation [l]implies that u 2 / 2 g ( D h ) is constant
along a streamline. This is just Bernoulli’s theorem for
steady incompressible hydrostatic flow since the con-
tribution of w 2 / 2 to the total KE is neglected in the
hydrostatic approximation. The term g ( D + h ) repre-
sents the combined PES associated with the gravita-
tional and pressure fields, as may be verified by taking
the hydrostatic pressure to be zero at the top of the
water and choosing the z = 0 level to coincide with the
bottom of the channel away from the obstacle; then at
an arbitrary level z,
Figure 4 Behavior of shallow water flowing over an obstacle:

(A) everywhere supercritical flow; (B) everywhere subcritical flow;
(C) hydraulic jump after a transition from subcritical to supercritical
flow over the crest. (Reproduced with permission from Duran
According to this generalized interpretation of PE,
(1990, Figure 4.5)) fluid parcels ascending the obstacle in a supercritical
flow slow down as they convert KE to PE, and after
ridge of ice. The case F < 1,known as subcritical flow,
passing the crest they reaccelerate as PE is converted
is shown in Figure 4B. The fluid parcel accelerations in
back to KE (Figure 4A).
the subcritical flow seem counterintuitive in that the
On the other hand, in subcritical flow ( F < 1) the
fluid thins and accelerates as it crosses the top of the
pressure gradient force dominates acceleration and the
obstacle, reaching its maximum speed at the crest.
three-way balance in eqn [ l ] requires that fluid parcels
Why does a subcritical flow accelerate as it encoun-
accelerate in the direction opposite to the component
ters rising bottom topography? In contrast to a
of gravity parallel to the topography. As shown in
frictionless hockey puck, the acceleration of a fluid
Figure 4B, fluid parcels ascending the obstacle accel-
parcel is determined not only by gravity and by the
erate as the free surface drops and PE is converted to
angle of the slope but also by pressure gradient forces.
KE. After passing the crest, the parcels decelerate as
The steady-state momentum equation [l]requires a
KE is converted back to PE. The disturbance centered
three-way balance between acceleration (the first
over the obstacle in Figure 4B is a steady surface
term), pressure gradient forces arising from changes
gravity wave.
in the fluid depth (the second term), and the work per
The flow regime that serves as an analogue for
unit mass done against gravity while ascending the
downslope windstorms is shown in Figure 4C. If the
sloping topography (the third term). The value of the
flow is subcritical upstream and if a column of fluid
Froude number determines whether the work done
undergoes a sufficient acceleration and experiences a
against gravity is predominantly balanced by acceler-
ations or by the pressure gradient force. From eqn [2] sufficient decrease in thickness as it ascends toward the
crest, a transition from subcritical to supercritical flow
occurs at the top of the obstacle. Since the lee-slope
(uE)/(gg) =(uE)/(-$E) = -F2 flow is now supercritical, fluid parcels continue to
accelerate as they descend, and very high velocities can
be produced because PE is converted to KE during the
648 DOWNSLOPE WINDS
entire time over which a fluid parcel traverses the U / f l in eqn [3], will not reliably yield reasonable
obstacle. The deceleration that would otherwise occur approximations to the speed and depth of the down-
in the lee-side portion of the standing gravity wave is slope flow in actual windstorms.
disrupted when the flow becomes supercritical. In this Significant downslope winds have been observed to
case fluid parcels eventually decelerate when they pass develop in three basic situations: (1)when a standing
through a turbulent hydraulic jump at some point mountain wave in a deep cross-mountain flow
downstream from the crest. achieves sufficient amplitude to overturn and break
down at some level in the troposphere; (2) when
standing mountain waves break and dissipate at a
Application of the Hydraulic Analogue critical level in a shallow cross-mountain flow; and ( 3 )
when there is sufficient static stability near mountain-
to the Atmosphere top level in the cross-mountain flow to create high
The hydraulic analogue is best applied to the atmos- downslope winds even without wave breaking. The
phere in a qualitative rather than a quantitative, qualitative application of hydraulic theory to the
manner. Quantitative application is hindered by the dynamics of downslope winds centers on the idea that
difficulty of defining a dynamically meaningful Froude in all three of these cases there is a transition from
number in vertically unbounded continuously strati- wavelike behavior over the upstream slopes of the
fied flow. A variety of expressions have been described topography to a non-wavelike regime in the lee.
as Froude numbers in the literature, but all of the First, consider the case of breaking waves in a deep
simple expressions have serious deficiencies. cross-mountain flow. The structures of the low-level
The parameter U/(Nho, where N is the Brunt- horizontal velocity perturbations in a stationary 2D
Vaisala frequency, U the wind speed, and ho the internal gravity wave forced by an isolated ridge are
maximum mountain height, is sometimes referred to shown in Figure SA. In this case, the upstream wind
as the Froude number in idealized cases in which N and static stability are constant with height such that
and U are constant throughout the upstream flow. N = 0 . 0 1 0 4 7 ~ - ~U, = lOms-l, and Nho/U = 0.6.
Unlike the denominator in the conventional shallow Streamlines for this same stationary internal gravity
water Froude number, Nho is not the horizontal phase wave are plotted in Figure 3A of the entry Lee Waves
speed of any particularly significant wave. (Nho is the and Mountain Waves. As is apparent in Figure SA, the
phase speed of a hydrostatic internal gravity wave with detailed structures of the velocity perturbations in the
vertical wavelength 271h0, but there is nothing parti- internal gravity wave are somewhat different from
cularly significant about this wavelength in contrast to those in the surface gravity wave schematically illus-
other similar waves with wavelengths such as 5ho or trated in Figure 4B. In particular the maximum
6ho.) On the other hand, the maximum perturbation perturbation surface wind speed occurs halfway
horizontal wind speed u’ in linear flow over an down the lee slope in the internal gravity wave,
obstacle with constant N and U scales like Nho, so whereas it occurs at the crest in the surface gravity
that U / ( N h o )x U / u ’ might be better described as a wave. Nevertheless, both types of waves allow a fluid
nonlinearity parameter. parcel to arrive at the ridge crest with a positive
When there is a strong well-defined inversion at perturbation velocity (i.e., to undergo a net accelera-
some elevation H in the upstream flow, many authors tion while ascending to the crest), and in both cases the
have attempted to define a Froude number as wind speed eventually returns to its ambient value well
U/&E, where g’ = gA0/80 is the ‘reduced gravity’, downstream of the crest as KE is converted back to PE
A0 is the increase in potential temperature across the in the lee-side portion of the stationary gravity wave.
inversion, and 00 is the mean potential temperature The enhancement of the perturbation horizontal
below the inversion. The difficulty with this approach winds along the lee slope in Figure SA is too weak to
is that it implies that the pressure gradient force is create significant downslope winds. (The total wind
entirely determined by the vertical displacements of speed increases from 10 m s - far upstream to ap-
the inversion layer and thereby neglects the influence, proximately 1 5 m s - 1 in the lee.) Much stronger
on the surface pressure gradient, of vertical displace- downslope winds occur in the case shown in Figure
ments in the stably stratified fluid above and below the SB, which is a vertical cross-section of the perturba-
inversion. Moreover, it is also very difficult to deter- tion horizontal velocity in a simulation identical to
mine a precise quantitative value for u/&E in more that shown in Figure SA, except that the height of the
general applications in which the wind speed is not mountain has been doubled so that Nho/U = 1.2. The
constant below the inversion and the inversion itself higher topography in this case forces the internal
may be indistinct. As a consequence, the reduced- gravity wave to overturn and produces a well-mixed
gravity shallow water model, in which F is replaced by region of weakly reversed flow at elevations around
DOWNSLOPE WINDS 649
-20 0 20 40 -20 0 20 40
Cross-ridge distance (km) Cross-ridge distance (km)
Figure 5 Perturbation horizontal velocity in flow over an isolated mountain when (A) Nho/U = 0.6, contour interval 1 m s - ’ and
(B) Nho/U= 1.2,contourinterval2ms-’
3 km over the lee slope. (The region of reversed flow is undergo a net acceleration in the wavelike upstream
that in which the horizontal perturbation velocity is flow as they ascend the mountain crest and then
less than - 10 m s - I.) Streamlines for this same wave- continue to accelerate as they convert PE to KE while
breaking case are shown in Figure 4A of the entry Lee descending the entire lee slope. The vertical displace-
Waves and Mountain Waves. Although the lee-side ment of a streamline about its initial undisturbed level
flow is dramatically different when the wave is 6 ( x ,z ) can be modeled with reasonable fidelity in the
breaking, the flow upstream of the crest remains flow beneath the critical layer by solving the hydro-
consistent with that in a stationary internal gravity static Long’s equation
wave. Linear theory for stationary internal gravity
waves predicts that doubling the mountain height
should double the amplitude of the perturbation 171
horizontal velocities without changing the spatial
subject to the lower boundary condition that the
distribution of the perturbations relative to the
streamline follow the topography
mountain, and this is essentially the case in the region
upstream of the crest. Note, for example, the similarity 6 [ x z, = h ( x ) ]= h ( x )
’
of the 3 m s - contour in Figure SA to the 6 m s -
181
contour in Figure 5B. and an upper boundary condition in which the
Since the wave in Figure 5B has become unstable horizontal wind speed is held constant along a
and overturned above the lee slope, there is no ‘dividing streamline’ separating the well-mixed turbu-
standing gravity wave to decelerate the fluid parcels lent region from the underlying high-speed flow. In the
as they descend. Instead these parcels continue to case shown in Figure 3, the 294 K isentrope approx-
accelerate as PE is converted to KE along the entire lee imates a dividing streamline while the 296 K isentrope
slope, generating strong downslope winds in which roughly coincides with the top of the wedge of well-
the maximum surface wind speeds ( > 28 m s- are ’) mixed air downwind of the crest. Very close mathe-
approximately three times stronger than the 10 m s - matical analogies exist between conventional shallow
flow far upstream. Wave breaking in a deep cross- water hydraulic theory and the mathematical solu-
mountain flow appears to have played an important role tions to eqns [7] and [8], although there is no simple
in the generation of the 11January 1972 Boulder, CO, parameter that plays the role of the Froude number in
windstorm. The presence of breaking waves is suggested this analogy.
by the almost vertical orientation of the isentropes on The third situation that produces strong downslope
the lee side of the trough in the upper-level wave in winds may occur when there is high static stability at
Figure 2 and by the turbulence encountered along the low levels in the cross-mountain flow and lower
flight legs through this region. stability aloft. A prototypical example of this type is
The second type of situation conducive to the presented in Figure 6, which shows contours of the
development of strong downslope winds is illustrated perturbation horizontal velocity field and streamlines
in Figure 3. In this bora event a critical level at an from a numerical simulation identical to that de-
elevation of about 2 km disrupts the lee-side gravity scribed in Figure SA, except that the Brunt-Vaisala
wave so that, once again, fluid parcels near the surface frequency above 3 km in the upstream flow is reduced
650 DROUGHT
-20 0 20 40 -20 0 20 40
Figure 6 Two-layer flow over an isolated mountain in which the upstream value of Nho/U is 0.6 in the lower layer and 0.24 above:
(A) perturbation horizontal velocity contour interval 2 m s - ’ ; (B)streamlines within the lower layer.
by a factor of 0.4. Comparison of the horizontal wind with wave breaking to generate very high winds.
speed perturbations between Figures SA and 6Ashows Indeed climatological data and numerical experiments
that the perturbation horizontal winds are twice as suggest this is often the case in Boulder windstorms. In
strong and that the maximum winds have shifted to the particular, nonlinear wave amplification due to the
surface along the lee slope in the two-layer flow. The presence of a low-level stable layer appears to have
amplification of the surface winds in the two-layer served as a necessary precursor to wave breaking
simulation is produced without wave breaking; in fact, during the 11January 1972 event.
the flow does not come close to stagnation.
The streamlines within the lower layer shown in
Figure 6B appear similar to those in water undergoing See also
a transition from subcritical to supercritical flow over Buoyancy and BuoyancyWaves: Optical Observations;
the crest of an obstacle. Near the base of the lee slope in Theory. Katabatic Winds. Lee Vortices. Lee Waves
Figure 6, the flow recovers toward ambient conditions and Mountain Waves.
by radiating energy downstream in a series of verti-
cally trapped gravity waves. The removal of energy by
these trapped waves is analogous to the dissipation of Further Reading
energy at the point where the flow recovers toward Durran DR (1990) Mountain waves and downslope winds.
ambient downstream conditions in a hydraulic jump In: Blumen W (ed.) Atmospheric Process over Complex
in the standard shallow water model (Figure 4C). Terrain, pp. 59-8 1. Boston: American Meteorological
Additional sensitivity studies have demonstrated that Society.
the changes in the depth of the lower layer and the Lilly DK (1978)A severe downslope windstorm and aircraft
maximum height of the mountain modify the two- turbulence event induced by a mountain wave. Jouvnal of
the Atmospheric Sciences 35: 59-77.
layer flow in a manner one would expect on the basis of Smith RB (1987) Aerial observations of the Yugo-
hydraulic theory. In particular, making the lower layer slavian bora. Journal of the Atmospheric Sciences 44:
too deep or the mountain too small eliminates the 2 69-2 9 7.
transition to a high wind regime. Smith RB (1989) Hydrostatic air flow over mountains. In:
In actual downslope wind events the dynamical Saltzman B (ed.) Advances in Geophysics, vol. 31, pp.
influence of a low-level stable layer may act in concert 1-41. New York: Academic Press.
D A Wilhite, National Drought Mitigation

Center, University of Nebraska-Lincoln, Lincoln, Introduction
NE, USA
Drought results from a deficiency of precipitation
Copyright 2003 Elsevier Science Ltd. All Rights Resewed. from expected or ‘normal’ that, when extended over a
650 DROUGHT
-20 0 20 40 -20 0 20 40
Figure 6 Two-layer flow over an isolated mountain in which the upstream value of Nho/U is 0.6 in the lower layer and 0.24 above:
(A) perturbation horizontal velocity contour interval 2 m s - ’ ; (B)streamlines within the lower layer.
by a factor of 0.4. Comparison of the horizontal wind with wave breaking to generate very high winds.
speed perturbations between Figures SA and 6Ashows Indeed climatological data and numerical experiments
that the perturbation horizontal winds are twice as suggest this is often the case in Boulder windstorms. In
strong and that the maximum winds have shifted to the particular, nonlinear wave amplification due to the
surface along the lee slope in the two-layer flow. The presence of a low-level stable layer appears to have
amplification of the surface winds in the two-layer served as a necessary precursor to wave breaking
simulation is produced without wave breaking; in fact, during the 11January 1972 event.
the flow does not come close to stagnation.
The streamlines within the lower layer shown in
Figure 6B appear similar to those in water undergoing See also
a transition from subcritical to supercritical flow over Buoyancy and BuoyancyWaves: Optical Observations;
the crest of an obstacle. Near the base of the lee slope in Theory. Katabatic Winds. Lee Vortices. Lee Waves
Figure 6, the flow recovers toward ambient conditions and Mountain Waves.
by radiating energy downstream in a series of verti-
cally trapped gravity waves. The removal of energy by
these trapped waves is analogous to the dissipation of Further Reading
energy at the point where the flow recovers toward Durran DR (1990) Mountain waves and downslope winds.
ambient downstream conditions in a hydraulic jump In: Blumen W (ed.) Atmospheric Process over Complex
in the standard shallow water model (Figure 4C). Terrain, pp. 59-8 1. Boston: American Meteorological
Additional sensitivity studies have demonstrated that Society.
the changes in the depth of the lower layer and the Lilly DK (1978)A severe downslope windstorm and aircraft
maximum height of the mountain modify the two- turbulence event induced by a mountain wave. Jouvnal of
the Atmospheric Sciences 35: 59-77.
layer flow in a manner one would expect on the basis of Smith RB (1987) Aerial observations of the Yugo-
hydraulic theory. In particular, making the lower layer slavian bora. Journal of the Atmospheric Sciences 44:
too deep or the mountain too small eliminates the 2 69-2 9 7.
transition to a high wind regime. Smith RB (1989) Hydrostatic air flow over mountains. In:
In actual downslope wind events the dynamical Saltzman B (ed.) Advances in Geophysics, vol. 31, pp.
influence of a low-level stable layer may act in concert 1-41. New York: Academic Press.
D A Wilhite, National Drought Mitigation

Center, University of Nebraska-Lincoln, Lincoln, Introduction
NE, USA
Drought results from a deficiency of precipitation
Copyright 2003 Elsevier Science Ltd. All Rights Resewed. from expected or ‘normal’ that, when extended over a
DROUGHT 651
season or longer period of time, is insufficient to meet must be region-specific and application- (or impact-)
the demands of human activities. Scores of definitions specific. This is one explanation for the scores of
of drought exist, reflecting different climatic charac- definitions that have been developed. Many defini-
teristics from region to region and sector-specific tions do not adequately define drought in meaningful
impacts. Although droughts are usually classified as terms for scientists and policy makers. In many cases,
meteorological, agricultural, hydrological, or socio- the thresholds for declaring drought are arbitrary (i.e.,
economic, all types of drought originate from a they are not linked to specific impacts in key economic
deficiency of precipitation that results in water short- sectors).
age for some activity or some group. Drought must be Third, drought impacts are nonstructural and
considered a relative, rather than absolute, condition. spread over a larger geographical area than damages
The ultimate results of these precipitation deficiencies that result from other natural hazards. The (US)
are, at times, enormous economic and environmental National Drought Mitigation Center recently deter-
impacts as well as personal hardship. Impacts of mined that for the 48 contiguous states of the United
drought appear to be increasing in both developing States, severe and extreme drought affected more than
and developed countries, a clear indication of nonsus- 25% of the country in 2 7 of the past one hundred
tainable development in many cases. Lessening the years. This represents an area of 1 9 4 2 500 km2 or
impacts of future drought events will require nations more. During the drought of 1934, more than 60% of
to pursue development of drought policies that the 48 contiguous states experienced severe or extreme
emphasize a wide range of risk management tech- drought conditions. Quantifying the impacts and
niques, including improved monitoring and early providing disaster relief are far more difficult tasks
warning systems, preparedness plans, and appropriate for drought than they are for other natural hazards.
mitigating actions and programs. Emergency managers, for example, are more accus-
Drought is considered by many to be the most tomed to dealing with impacts that are structural and
complex but least understood of all natural hazards, localized, responding to these events by restoring
affecting more people than any other hazard. It is a communication and transportation channels, provid-
normal feature of climate and its recurrence is inev- ing emergency medical supplies, ensuring safe drink-
itable. However, there remains much confusion about ing water, and so forth. These characteristics of
its characteristics. It is precisely this confusion that drought have hindered the development of accurate,
explains, to some extent, the lack of emphasis on reliable, and timely estimates of severity and impacts
proactive drought management efforts in most parts of and, ultimately, the formulation of drought prepared-
the world. Through an improved understanding and ness plans.
awareness of the concept and characteristics of Many people consider drought to be largely a
drought and its differences from other natural haz- natural or physical event. Figure 1 illustrates that, in
ards, scientists, policy makers, and the public will be reality, drought has both natural and social compo-
better equipped to establish much-needed policies and nents. The risk associated with drought for any region
plans whereby vulnerability can be reduced for future is a product of both the region’s exposure to the event
generations. (i.e., probability of occurrence at various severity
levels) and the vulnerability of society to the event. The
natural event (i.e., meteorological drought) is a result
The Concept of Drought
of the occurrence of persistent large-scale disruptions
Drought differs from other natural hazards (such as in the global circulation pattern of the atmosphere.
floods, tropical cyclones, and earthquakes) in several Exposure to drought varies spatially and there is little,
ways. First, drought is a slow-onset, creeping natural
hazard. Its effects often accumulate slowly over a
-
considerable period of time and may linger for years Components of
after the termination of the event. Therefore, the onset
and end of drought are difficult to determine. Because
of this slow-onset characteristic, it is difficult to
recognize the onset of drought, and scientists and
F
HAZARD
(natural event)
Drought \ Vulnerability
(social factors)
policy makers continue to debate the basis (i.e., the
*Hazard +Vulnerability = RISK *Prediction
criteria) for declaring an end to a drought. Monitoring/Early warning
Second, the absence of a precise and universally *Mitigation
accepted definition of drought adds to the confusion Preparedness
about whether or not a drought exists and, if it does, its Figure 1 Components of drought. (Source: National Drought
degree of severity. Realistically, definitions of drought Mitigation Center.)
652 DROUGHT
if anything, that we can do to alter drought occur- drought is rarely static during the course of the event.
rence. Vulnerability, on the other hand, is determined As drought emerges and intensifies, its core area or
by social factors such as population changes, popula- epicenter shifts and its spatial extent expands and
tion shifts (regional and rural to urban), demographic contracts throughout the duration of the event.
characteristics, technology, policy, and social be- Because drought affects so many economic and
havior. These factors change over time and thus social sectors, scores of definitions have been devel-
vulnerability is likely to increase or decrease in oped by a variety of disciplines. In addition, because
response to these changes. Subsequent droughts in drought occurs with varying frequency in nearly all
the same region will have different effects, even if they regions of the globe, in all types of economic systems,
are identical in intensity, duration, and spatial char- and in developed and developing countries alike, the
acteristics, because societal characteristics will have approaches taken to define it also reflect regional and
changed. However, much can be done to lessen ideological differences. Impacts also differ spatially
societal vulnerability to drought. and temporally, depending on the societal context of
drought. A universal definition of drought is an
unrealistic expectation.
Defining Drought Many disciplinary perspectives of drought exist,
Drought is the consequence of a natural reduction in often causing considerable confusion about what
the amount of precipitation received over an extended constitutes a drought. Research has shown that the
period of time, usually a season or more in length, lack of a precise and objective definition in specific
although other climatic factors (such as high temper- situations has been an obstacle to understanding
atures, high winds, and low relative humidity) are drought, which has led to indecision and/or inaction
often associated with it in many regions of the world on the part of managers, policy makers, and others. It
and can significantly aggravate the severity of the must be accepted that the importance of drought lies in
event. Drought is also related to the timing (principal its impacts. Thus definitions should be region-specific
season of occurrence, delays in the start of the rainy and impact- or application-specific in order to be used
season, occurrence of rains in relation to principal in an operational mode by decision makers.
crop growth stages) and the effectiveness of the rains Drought is normally grouped by type into meteor-
(rainfall intensity, number of rainfall events). Thus, ological, hydrological, agricultural, and socioeco-
each drought event is unique in its climatic character- nomic. Figure 2 explains the relationship between
istics, spatial extent, and impacts. The area affected by these various types of drought and the duration of the
1 Natural Climate Variability I

I
I I
Precipitation deficiency High temp., high winds, low
(amount, intensity, timing) relative humidity, greater
sunshine, less cloud cover
Reduced infiltration, runoff,
deep percolation, and Increased evaporation
ground water recharge
I
_ _ _ _ - - _ - _ _ - - - - - - - ____ ____
_ ___ _ - _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
I Soil water deficiency I
I
I Plant water stress, reduced

biomass and vield
Reduced streamflow, inflow to

reservoirs, lakes, and ponds;
reduced wetlands,
wildlife habitat
I Economih Impacts I1- I Environmehtal Impacts I

Figure 2 Relationship between types of drought. (Source: National Drought Mitigation Center.)
DROUGHT 653
event. The impacts associated with drought usually Finally, socioeconomic drought associates the sup-
take three or more months to develop, but this period ply and demand of some economic good or service
can vary considerably, depending on the timing of the with elements of meteorological, hydrological, and
initiation of the precipitation deficiency. agricultural drought. Socioeconomic drought is asso-
Meteorological (or climatological) drought is ex- ciated directly with the supply of some commodity or
pressed solely on the basis of the degree of dryness economic good (e.g., water, hay, hydroelectric power)
(often in comparison to some normal or average that is the result of precipitation shortages. Increases in
amount) and the duration of the dry period. Thus, population can substantially alter the demand for
intensity and duration are the key characteristics of these economic goods over time. This concept of
these definitions. Meteorological drought definitions drought supports the strong symbiosis that exists
must be considered as region-specific since the atmos- between drought and its impacts and human activities.
pheric conditions that result in deficiencies of pre- Thus, the incidence of drought could increase because
cipitation are climate regime-dependent. Most mete- of a change in the frequency of meteorological
orological drought definitions relate actual precipita- drought, a change in societal vulnerability to water
tion departures to average amounts on monthly, shortages, or both. For example, poor land-use prac-
seasonal, water year, or annual time scales. tices such as overgrazing can decrease animal carrying
Agriculture is usually the first economic sector to be capacity and increase soil erosion, which exacerbates
affected by drought because soil moisture supplies are the impacts of and vulnerability to future droughts.
often quickly depleted, especially if the period of
moisture deficiency is associated with high tempera-
tures and windy conditions. Agricultural drought links
Drought Characteristics and Severity
various characteristics of meteorological drought to Droughts differ from one another in three essential
agricultural impacts, focusing on precipitation short- characteristics: intensity, duration, and spatial cover-
ages, differences between actual and potential evapo- age. Intensity refers to the degree of the precipitation
transpiration, and soil water deficits. A plant’s shortfall and/or the severity of impacts associated with
demand for water is dependent on prevailing weather the shortfall. It is generally measured by the departure
conditions, biological characteristics of the specific of some climatic index from normal and is closely
plant, its stage of growth, and the physical and linked to duration in the determination of impact. The
biological properties of the soil. A definition of simplest index in widespread use is the percentage of
agricultural drought should account for the variable normal precipitation, but there are numerous more
susceptibility of crops at different stages of crop complex and more effective indices available. It is
development. For example, deficient subsoil moisture generally recommended that several indices be used to
in an early growth stage will have little impact on final monitor drought onset and development because each
crop yield if topsoil moisture is sufficient to meet early index has its particular strengths and weaknesses. One
growth requirements. However, if the deficiency of of the principal difficulties with this (or any) index is
subsoil moisture continues, a substantial yield loss the choice of the threshold below which the deficiency
may result. of precipitation must fall (e.g., 75% of normal) to
Hydrological droughts are associated with the define the onset of drought and trigger various
effects of periods of precipitation shortfall on surface mitigation actions or response programs. Thresholds
or subsurface water supply (i.e., streamflow, reservoir are usually chosen arbitrarily. In reality, they should be
and lake levels, groundwater) rather than with pre- linked to impact.
cipitation shortfalls. Hydrological droughts are usu- Another distinguishing feature of drought is its
ally out of phase with or lag the occurrence of duration. Droughts usually require a minimum of two
meteorological and agricultural droughts. More time to three months to become established but then can
elapses before precipitation deficiencies are detected in continue for months or years. The magnitude of
other components of the hydrological system (e.g., drought impacts is closely related to the timing of the
reservoirs, groundwater). As a result, impacts are out onset of the precipitation shortage, its intensity, and
of phase with those in other economic sectors. Also, the duration of the event. For example, a dry period
water in hydrological storage systems (e.g., reservoirs, that begins in the late fall and continues through the
rivers) is often used for multiple and competing winter months in the midwestern United States will
purposes (power generation, flood control, irrigation, likely have negligible impacts. However, if dry condi-
recreation), further complicating the sequence and tions persist into the spring and early summer months,
quantification of impacts. Competition for water in agricultural and urban demands for water supplies
these storage systems escalates during drought, and increase dramatically. Pasture growth will be red-
conflicts between water users increase significantly. uced, affecting livestock producers through supply
654 DROUGHT
shortages and increasing prices. Diminished topsoil regions within their borders and develop contingen-
and subsoil moisture will also affect seed emergence cies for the occurrences of such an event. Likewise, it is
and early growth development for grain crops such as important for governments to know the chances of a
corn and soybeans, eventually affecting yield and crop regional drought simultaneously affecting agricultural
production if dry conditions persist throughout the productivity in their country as well as adjacent
summer months. Urban water supplies are often or nearby nations on whom they are dependent for
reduced, forcing water suppliers to impose volun- food supplies. In some instances, a nation’s primary
tary or mandatory water conservation measures. As drought mitigation strategy may be to import food
drought conditions extend over more than one grow- from nearby nations, ignoring the likelihood that a
ing season, impacts magnify substantially as a result of drought may have significant regional impacts on food
declining surface and subsurface water supplies and an supplies. Likewise, the occurrence of drought world-
expanding circle of impacts. wide or in the principal grain-exporting nations, such
Droughts also differ in terms of their spatial as occurred during the ENS0 event of 1982-83, may
characteristics. The areas affected by severe drought significantly alter a developing country’s access to
evolve gradually, and regions of maximum intensity food from donor governments.
shift from season to season. In larger countries, such as
Brazil, China, India, the United States, or Australia,
drought would rarely, if ever, affect the entire country.
The Impacts of Drought
During the severe drought of the 1930s in the United The impacts of drought are diverse and often ripple
States, for example, the area affected by severe through the economy. Thus, impacts are often referred
drought never exceeded 65% of the country (see to as direct or indirect, or they are classified as first-,
Figure 3). By contrast, drought affected more than second- or third-order. Conceptually speaking, the
95% of the Great Plains region in 1934. On the other more removed the impact from the cause, the more
hand, it is indeed rare for drought not to exist in a complex the link to the cause. In other words, a loss of
portion of the United States each year. Thus, the yield resulting from drought is a direct or first-order
governments of larger countries are more accustomed impact of drought. However, the consequences of that
to dealing with water shortages and have established impact (e.g., loss of income, farm foreclosures, gov-
an infrastructure to respond, albeit reactively. For ernment relief programs) are secondary or tertiary
smaller countries, it is more likely that the entire impacts.
country might be affected since droughts are usually Because of the number of affected groups and
regional phenomena. sectors associated with drought, its spatial extent, and
From a planning perspective, the spatial character- the difficulties connected with quantifying environ-
istics of drought have serious implications. Nations mental damages and personal hardships, the precise
should know the probability that drought may simul- determination of the financial costs of drought is an
taneously affect all or several major crop-producing arduous task. It has been estimated that the average
70 Yo
I
I
40
30 I
20
10
0
1895 1905 1915 1925 1935 1945 1955 1965 1975 1985 1995
Figure 3 Percentage area of the United States in severe and extreme drought, 1895-1999. (Source: National Drought Mitigation
Center, based on data provided by the National Climatic Data Center, National Oceanic and Atmospheric Administration, US Department
of Commerce.)
DROUGHT 655
annual impacts of drought in the United States are and short-term assistance programs are often reactive,
$6-8 billion. These figures may be misleading because a kind of ‘band-aid’ approach to more serious land and
drought years often occur in clusters, such as 1931-39, water management problems. Actions of this type
1954-56, 1974-77, 1987-92, and 1996-2000 (see have long been criticized by scientists and government
Figure 3 ) . Impacts during each of these years were officials, as well as by recipients of relief, as inefficient
much above the annual average. During the drought and ineffective. Long-term assistance programs are far
years of 1976-77 and 1988, for example, government fewer in number, but they are proactive. They attempt
estimates of impacts were $36 and $40 billion, to lessen a region’s vulnerability to drought through
respectively. improved management and planning.
The impacts of drought can be classified into three Governmental response to drought includes a wide
principal areas: economic, environmental, and social. range of potential actions to deal with the impacts of
Table 1presents a simplified illustration of the impacts water shortages on people and various economic
associated with each of these areas. Economic impacts sectors. In the United States and other developed
range from direct losses in the broad agricultural and countries, agencies of the federal government typically
agriculturally related sectors, including forestry and respond by making massive amounts of relief available
fishing, to losses in recreation, transportation, bank- to the affected areas. Most of this relief is in the form of
ing, and energy sectors. Other economic impacts short-term emergency measures to agricultural pro-
would include added unemployment and loss of ducers and few, if any, of these assistance measures in
revenue to local, state, and federal government. recent years have been aimed at reducing future
Environmental losses are the result of damages to vulnerability. In developing countries, emergency
plant and animal species, wildlife habitat, and air and response is often provided by donor governments,
water quality; damage from forest and range fires; nongovernmental organizations, and international
degradation of landscape quality; and soil erosion. organizations in the form of food aid, health services,
Although these losses are difficult to quantify, growing access to potable water supplies, and transportation
public awareness and concern for environmental services.
quality has forced public officials to focus greater Because of the unique character of drought, gov-
attention on these effects. Social impacts mainly ernments have been less inclined to invest resources to
involve public safety, health, conflicts between water develop well-conceived mitigation programs and
users, and inequities in the distribution of impacts and contingency plans. This reactive approach to natural
disaster relief programs. disasters is commonly referred to as crisis manage-
As with all natural hazards, the economic impacts of ment. Research has demonstrated that reaction to
drought are highly variable within and between crisis often results in the implementation of hastily
economic sectors and geographic regions, producing prepared assessment and response procedures that
a complex assortment of winners and losers with the lead to ineffective, poorly coordinated, and untimely
occurrence of each disaster. For example, decreases in response. An alternative approach is to initiate plan-
agricultural production result in enormous negative ning between periods of drought, thus developing a
financial impacts on farmers in drought-affected more coordinated and proactive response that would
areas, at times leading to foreclosure. This decreased more effectively address those persons, areas, and
production also leads to higher prices of grains, sectors most at risk. Also, the limited resources
vegetables, and fruit. These price increases have a available to government to mitigate the effects of
negative impact on all consumers as food prices drought could be allocated in a more beneficial
increase. However, farmers outside the drought- manner.
affected area with normal or above-normal produc-
tion or those with significant grain in storage reap the
benefits of these higher prices. Similar examples of
Drought Policy and Planning
winners and losers can be given for other economic
sectors as well. Drought planning is defined as actions taken by
individual citizens, industry, government, and others
in advance of drought for the purpose of mitigating
Drought Response and Preparedness some of the impacts and conflicts associated with its
With the occurrence of any natural disaster come occurrence. Because drought is a normal part of
appeals for disaster assistance from the affected area. climate variability for virtually all regions, it is
For decades, governments have typically responded to important to develop plans to deal with these extended
drought by providing emergency, short-term, and periods of water shortage in a timely, systematic
long-term assistance to distressed areas. Emergency manner as they evolve. This planning process needs to
656 DROUGHT
Table 1 Classification of drought-related impacts (costs and losses)'

_ _ _ _ _ ~ _ _ _ ~
~
Problem sectors Impacts
Economic 0 Loss from crop production

Annual and perennial crop losses; damage to crop quality
Reduced productivity of cropland (wind erosion, etc.)
Insect infestation
Plant disease
Wildlife damage to crops
Loss from dairy and livestock production
Reduced productivity of range land
Forced reduction of foundation stock
Closure/limitation of public lands to grazing
High costhnavailability of water for livestock
High costhnavailability of feed for livestock
High livestock mortality rates
Increased predation
Range fires
Loss from timber production
Forest fires
Tree disease
Insect infestation
Impaired productivity of forest land
Loss from fishery production
Damage to fish habitat
Loss of young fish due to decreased flows
Loss of national economic growth, retardation of economic development
Income loss for farmers and others directly affected
Loss of farmers through bankruptcy
Loss to recreational and tourism industry
Loss to manufacturers and sellers of recreational equipment
Increased energy demand and reduced supply because of drought-related power curtailments
Costs to energy industry and consumers associated with substituting more expensive fuels (oil) for hydroelectric
power
Loss to industries directly dependent on agricultural production (machinery and fertilizer manufacturers, food
processors, etc.)
Decline in food production/disrupted food supply
Increase in food prices
Increased importation of food (higher costs)
Disruption of water supplies
Unemployment from drought-related production declines
Strain on financial institutions (foreclosures, greater credit risks, capital shottfalls, etc.)
Revenue losses to federal, state, and local governments (from reduced tax base)
Deterrence of capital investment, expansion
Dislocation of businesses
Revenues to water supply firms
Revenue shortfalls
Windfall profits
Loss from impaired navigability of streams, rivers, and canals
Cost of water transport or transfer
Cost of new or supplemental water resource development
Environmental Damage to animal species

Reduction and degradation of fish and wildlife habitat
Lack of feed and drinking water
Disease
Increased vulnerability to predation (e.g., from species concentration near water)
Loss of biodiversity
Wind and water erosion of soils
Reservoir and lake drawdown
Damage to plant species
Water quality effects (e.g., salt concentration, increased water temperatures, pH, dissolved oxygen)
Air quality effects (dust, pollutants)
Visual and landscape quality (dust, vegetative cover, etc.)
DROUGHT 657
Table 1 Continued
Problem sectors Impacts
Increased fire hazard

Estuarine impacts: changes in salinity levels, reduced flushing
Social Increased groundwater depletion (mining), land subsidence

Loss of wetlands
Loss of cultural sites
Insect infestation
Food shortages (decreased nutritional level, malnutrition, famine)
Loss of human life (e.g.,food shortages, heat)
Public safety from forest and range fires
Conflicts between water users, public policy conflicts
Increased anxiety
Loss of aesthetic values
Health-related low flow problems (e.g., diminished sewage flows, increased pollutant concentrations, etc.)
Recognition of institutional constraints on water use
Inequity in the distribution of drought impactshelief
Decreased quality of life in rural areas
Increased poverty
Reduced quality of life, changes in lifestyle
Social unrest, civil strife
Population migration (rural to urban areas)
Reevaluation of social values
Increased datdinformation needs, coordination of dissemination activities
Loss of confidence in government officials
Recreational impacts
'Source: National Drought Mitigation Center.
occur at various levels of government and to be Other developed and developing countries are also
integrated between levels of government. making substantial progress in drought preparedness
The purpose of a drought plan is to reduce the through the application of appropriate risk manage-
impacts of drought by identifying the principal sectors, ment techniques. Australia developed a national
groups, or regions most at risk and developing drought policy in the early 1990s and this policy has
mitigation actions and programs that can reduce these undergone revision as experience with the policy has
risks in advance of future drought events. Generally, increased. The Australian policy is focused on the
drought plans have three basic components: monitor- agricultural sector and seeks to improve the self-
ing and early warning; risk and impact assessment; reliance or drought-coping capacity of farmers and to
and response and mitigation. Plans will also improve promote the sustainable use of natural resources. The
coordination within agencies of government and government has invested resources in improving
between levels of government. In the United States, longer-range forecasts and other decision support
there has been a remarkable growth in the number of systems to aid farmers in farm management. South
states with drought plans. In 1982, only three states Africa also developed a national drought policy that
had drought plans. By late 2000, thirty states have emphasizes the same principles as that implemented
drought plans and six additional states are in various by Australia. Many other countries in South America,
stages of plan development (Figure 4). Planning
methodologies are available to facilitate the planning
process (see Further Reading). This trend demon-
strates an increased concern about the potential
impacts of extended water shortages and the com-
plexity of those impacts. Drought plans are at the
foundation of improved drought management, but
only if they emphasize risk assessment and mitigation
programs and actions. Most of the drought plans that
currently exist in the United States still emphasize
emergency response or crisis management, but this Figure 4 Status of drought planning, USA, December 2000.
pattern is slowly changing. (Source: National Drought Mitigation Center.)
658 DRYLINE
Europe, Africa, and Asia are following this trend impacts, its spatial extent, and the difficulties in
toward improving drought management through the quantifying environmental damages and personal
application of the principles of risk management. hardships, the precise calculation of the financial costs
Before developing a preparedness plan, government of drought is difficult. It appears that societal vulner-
officials should first define what they hope to achieve ability to drought is escalating in both developing and
by that plan. Thus, a drought policy statement should developed countries, and at a significant rate. It is
be prepared in advance of a plan. The objectives of imperative that increased emphasis be placed on
drought policy should encourage or provide incentives mitigation, preparedness, and prediction and early
for agricultural producers, municipalities, and other warning if society is to reduce the economic and
water-dependent sectors or groups to adopt appropri- environmental damages associated with drought and
ate and efficient management practices that help to its personal hardships. This will require improved
alleviate the effects of drought. Past relief measures coordination within and between levels of government
have usually discouraged the adoption of appropriate and the active participation of stakeholders.
management techniques. Assistance should also be
provided in an equitable, consistent, and predictable
manner to all without regard to economic circumstances, See also
industry, or geographic region. An objective should also Climate: Overview. Climate Variability: Seasonal to
seek to protect the natural and agricultural resource lnterannual Variability. El Niiio and the Southern Oscil-
base. Degradation of natural resources can result in lation: Observation; Theory. Global Change: Surface
spiraling economic, environmental, and social costs. Temperature Trends; Upper Atmospheric Change.
Hydrology: Ground and Surface Water; Soil Moisture.
Summary Palmer Drought Index. Wildfire Weather.
Drought is an insidious natural hazard that is a normal

part of the climate of virtually all regions. It should not Further Reading
be viewed as merely a physical phenomenon. Rather, Bruins HJ and Lithwick H (1998) The Arid Frontier:
drought is the result of an interplay between a natural Interactive Management of Environment and Develop-
event and the demand placed on water supply by ment. Dordrecht: Kluwer Academic.
human-use systems. Drought should be considered Glantz M H (ed.) (1994) Drought Follows the Plow: Culti-
relative to some long-term average condition of vating Marginal Areas. Cambridge: Cambridge Univer-
balance between precipitation and evapotranspiration. sity Press.
Many definitions of drought exist; it is unrealistic to Vogt JVand Somma F (eds) (2000) Drought and Drought
expect a universal definition to be derived. The three Mitigation in Europe. Advances in Natural and
characteristics that differentiate one drought from Technological Hazards Research. Dordrecht: Kluwer
another are intensity, duration, and spatial extent. The Academic.
Wilhite DA (ed.) (1993)Drought Assessment, Management,
impacts of drought are diverse and generally classified
and Planning: Theory and Case Studies. Natural Re-
as economic, social, and environmental. Impacts sources Management and Policy Series (Dinar A and
ripple through the economy and may linger for years Zilberman D, series eds). Dordrecht: Kluwer Academic.
after the termination of the drought episode. Because Wilhite DA (ed.)(2000)Drought: A Global Assessment, vols
of the large number of groups and economic sectors 1 and 2 . Hazards and Disasters: A Series of Definitive
affected by drought, the nonstructural nature of its Major Works. London: Routledge.
L D Grasso, Colorado State University, Fort Collins, CO, low-level thunderstorm outflows can trigger new
USA thunderstorms. Although different from a boundary,
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. some terrain features are known to be preferred
locations for the initiation of thunderstorms. Another
Introduction type of boundary that can trigger thunderstorms is a
dryline. Drylines have been studied extensively in the
Thunderstorm development is associated with a Great Plains of the United States. Two common
variety of boundaries in the Earth’s planetary bound- methods for studying the dryline are observational
ary layer. Along with cold fronts and warm fronts, field experiments and numerical simulations.
658 DRYLINE
Europe, Africa, and Asia are following this trend impacts, its spatial extent, and the difficulties in
toward improving drought management through the quantifying environmental damages and personal
application of the principles of risk management. hardships, the precise calculation of the financial costs
Before developing a preparedness plan, government of drought is difficult. It appears that societal vulner-
officials should first define what they hope to achieve ability to drought is escalating in both developing and
by that plan. Thus, a drought policy statement should developed countries, and at a significant rate. It is
be prepared in advance of a plan. The objectives of imperative that increased emphasis be placed on
drought policy should encourage or provide incentives mitigation, preparedness, and prediction and early
for agricultural producers, municipalities, and other warning if society is to reduce the economic and
water-dependent sectors or groups to adopt appropri- environmental damages associated with drought and
ate and efficient management practices that help to its personal hardships. This will require improved
alleviate the effects of drought. Past relief measures coordination within and between levels of government
have usually discouraged the adoption of appropriate and the active participation of stakeholders.
management techniques. Assistance should also be
provided in an equitable, consistent, and predictable
manner to all without regard to economic circumstances, See also
industry, or geographic region. An objective should also Climate: Overview. Climate Variability: Seasonal to
seek to protect the natural and agricultural resource lnterannual Variability. El Niiio and the Southern Oscil-
base. Degradation of natural resources can result in lation: Observation; Theory. Global Change: Surface
spiraling economic, environmental, and social costs. Temperature Trends; Upper Atmospheric Change.
Hydrology: Ground and Surface Water; Soil Moisture.
Summary Palmer Drought Index. Wildfire Weather.
Drought is an insidious natural hazard that is a normal

part of the climate of virtually all regions. It should not Further Reading
be viewed as merely a physical phenomenon. Rather, Bruins HJ and Lithwick H (1998) The Arid Frontier:
drought is the result of an interplay between a natural Interactive Management of Environment and Develop-
event and the demand placed on water supply by ment. Dordrecht: Kluwer Academic.
human-use systems. Drought should be considered Glantz M H (ed.) (1994) Drought Follows the Plow: Culti-
relative to some long-term average condition of vating Marginal Areas. Cambridge: Cambridge Univer-
balance between precipitation and evapotranspiration. sity Press.
Many definitions of drought exist; it is unrealistic to Vogt JVand Somma F (eds) (2000) Drought and Drought
expect a universal definition to be derived. The three Mitigation in Europe. Advances in Natural and
characteristics that differentiate one drought from Technological Hazards Research. Dordrecht: Kluwer
another are intensity, duration, and spatial extent. The Academic.
Wilhite DA (ed.) (1993)Drought Assessment, Management,
impacts of drought are diverse and generally classified
and Planning: Theory and Case Studies. Natural Re-
as economic, social, and environmental. Impacts sources Management and Policy Series (Dinar A and
ripple through the economy and may linger for years Zilberman D, series eds). Dordrecht: Kluwer Academic.
after the termination of the drought episode. Because Wilhite DA (ed.)(2000)Drought: A Global Assessment, vols
of the large number of groups and economic sectors 1 and 2 . Hazards and Disasters: A Series of Definitive
affected by drought, the nonstructural nature of its Major Works. London: Routledge.
L D Grasso, Colorado State University, Fort Collins, CO, low-level thunderstorm outflows can trigger new
USA thunderstorms. Although different from a boundary,
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. some terrain features are known to be preferred
locations for the initiation of thunderstorms. Another
Introduction type of boundary that can trigger thunderstorms is a
dryline. Drylines have been studied extensively in the
Thunderstorm development is associated with a Great Plains of the United States. Two common
variety of boundaries in the Earth’s planetary bound- methods for studying the dryline are observational
ary layer. Along with cold fronts and warm fronts, field experiments and numerical simulations.
DRYLINE 659
Observational studies have measured properties of the temperature and dew point temperature decrease.
drylines using a variety of instruments: land-based, In contrast, when a dryline passes an observing
remote sensors, and movable platforms. Land-based station, although the temperature will remain the
mesonetworks were established over much of Okla- same or increase, the dew point temperature usually
homa; each station is capable of measuring tempera- decreases significantly (Figure 1). At times, gusty
ture, dew point temperature, pressure, and wind. In westerly winds develop during the passage of a dryline.
addition, Doppler lidar, Doppler radar, and satellites An environment that may contain a dryline can be
have also been used to make remote observations of observed on the synoptic scale. A decrease of surface
drylines. Movable observing platforms have also been dew point temperatures from 20°C in central Okla-
employed. In some studies, instrumented air planes homa to 7°C near the border of Oklahoma and the
have flown organized paths through drylines. Texas Panhandle to about -2°C within the Texas
Recently, automobiles fitted with instruments have Panhandle is typical: these values were observed on 24
been used to drive through drylines. Along with May 1989. Measurements from dryline field experi-
observational studies, numerical models have also ments have demonstrated that water vapor mixing
been used to study drylines. ratio and dew point temperature gradients - at the
Numerical models are another tool that is com- surface - can respectively be as large as 5 g kg - over a
monly used to study a dryline. Sensitivity studies can 15 km distance and 18°C over a 10 km distance.
be conducted by varying the horizontal and vertical In addition to horizontal variations of both tem-
grid spacing. Terrain slope, soil moisture, and surface perature and water vapor mixing ratio, vertical
vegetation types can all be varied together or one at a variations also occur in the vicinity of a dryline. O n
time to study the impact on dryline evolution. One the moist side of a dryline, the depth of the moist layer
advantage of using a numerical model is the ability to generally increases to the east. In the afternoon,
neglect individual terms in the equations of motion; however, the thickness of the moist layer - immedi-
for example, either turbulent mixing tendencies or ately on the moist side of a dryline - can increase to
advection tendencies can be eliminated. Interpretation several kilometers. O n the dry side of a dryline, the
of data from both dryline simulations and field depth of the boundary layer can grow to a few
experiments provides important information for fore- kilometers. As a consequence, the potentially warmer
casters. air advects eastward over the moist layer, forming a
Forecasting thunderstorm development is an impor- temperature inversion.
tant ‘nowcasting’ issue. At times, the environment on Drylines can appear in the warm sector of an
the moist side of a dryline can support thunderstorms. extratropical low-pressure system. As a region of low
The challenge for forecasters is to determine whether
thunderstorms will develop along a dryline. Some-
times thunderstorms form, at other times thunder-
storms fail to form along the entire length of a dryline.
On some occasions, thunderstorms form along some
parts of a dryline while other parts of a dryline remain
cloud-free. Thunderstorms provide needed rain for
agricultural regions; at the same time, some thunder-
storms can also produce severe weather: large hail and
Warm
damaging outflow winds. In addition, tornadoes are front
frequently spawned; some tornadoes can cause not
only extensive damage to structures and crops, but L- 0
also loss of life.
Dryline Observations
Horizontal and Vertical Structure
Air masses in the planetary boundary layer of the
Earth’s atmosphere are separated by boundaries:
warm fronts, cold fronts, and drylines, for instance.
When a warm front passes an observing station, both Figure 1 Time sequence of temperature ( T ,solid) and dew point
the temperature and dew point temperature increase. temperature (Td, dashed) for the passage of (A) a cold front,
When a cold front passes an observing station, both (B) a warm front, and (C) a dryline.
660 DRYLINE
pressure moves eastward from the Rocky Mountains,

relatively hot and dry air advects north-eastward from
the desert south-west to western Texas, Oklahoma,
and Kansas. South to south-easterly flow from the
- 4 6
Dryline
environment
14
- 16
Gulf of Mexico transports warm moist air over eastern
Texas, Oklahoma, and Kansas. Both air masses - one
hot and dry, the other warm and moist - merge in a
confluence zone over central portions of the southern
plains. That is, the dryline environment contains not
only a horizontal dew point temperature gradient but
also a confluent wind field.
Demarcation of a Dryline
Determination of the location of a dryline, within an
environment where surface dew points decrease from
east to west, has been done in an arbitrary manner. The
following contours have been used to denote a dryline
-- Dryline zone
boundary:
0 The 8 or 9 g k g - ' water vapor mixing ratio (rv) Figure 2 Simplified profile of watervapor mixing ratio (solid lines
isohume. These values are approximately the mini- every 2 9 kg-') showing a dryline environment and a dryline zone
mum needed to support tornadic thunderstorms. To over Oklahoma and Texas.
illustrate, at 950 hPa an r, value of 9 g k g - I repre-
sents a dew point temperature equal to 11.4"C. Even though a dryline environment may exist over a
0 The 355 K isopleth of equivalent potential tempera- region, a dryline zone may be absent. Synoptic surface
ture (ee). observations from 27 May 1991 indicated a dew point
0 The first organized line of veering winds form the temperature decrease from approximately 20°C in
3-hourly synoptic surface charts. central Oklahoma to near 10°C at the Oklahoma-
Texas Panhandle border, to about 1°C in north-west
Some researchers have suggested that a dryline portions of the Texas Panhandle. This profile of dew
boundary is a mixing zone that separates hot and dry point temperatures was similar to 24 May 1989 (see
air from warm and moist air over a distance of above); however, a dryline zone was absent.
approximately 10 km. Observations from 6 May 1995 Although the determination of the location of a
dryline indicated horizontal gradients of water vapor dryline may vary, only one standard symbol is used on
on the scale of 10 km.On the basis of observations of a weather map. Warm fronts are denoted by a thick
the 15 May 1991, 7 June 1994, and 6 May 1995 contour with solid half circles drawn on the
drylines, a conceptual model of a dryline environment contour. Regular spaces are placed between each half
was proposed. In the conceptual model, a dryline zone circle along the length of a warm front. The symbol
separates hot and dry air from warm and moist air. The used to denote the dryline is similar to that for a warm
horizontal length scale of a dryline zone is approxi- front, except that the half circles are unfilled and are
mately 10 km. placed adjacent to each other along the length of the
Observations indicate there are two horizontal dryline.
length scales associated with a dryline: 100 km and
10km. The larger length scale, 100km identifies a
Drylines Worldwide
dryline environment. The dryline environment
describes the gradual decrease of surface dew point Drylines exist in different locations on the globe.
temperatures from 20°C in central Oklahoma to Drylines have been observed in eastern China, west
-2°C within the Texas Panhandle. The smaller-length central Africa, and India; in addition, dryline devel-
scale, 10 km, describes a dryline zone and is contained opment occurs in the Great Plains of the United States.
within a dryline environment. A dryline zone describes Drylines in the Great Plains generally exist as far south
the relatively large horizontal gradient of surface as Texas and as far north as Nebraska; furthermore,
dew point temperatures: 18°C over a 10 km distance, drylines are found east of the Rockies and west of the
for example. Identification of a dryline zone is a Mississippi River. Typically, drylines develop in the
recent method of denoting the location of a dryline spring and early summer. More specifically, observa-
(Figure 2 ) . tions indicate that drylines are present approximately
DRYLINE 661
30% of the days between April and June over the Great movement of a dryline. That is, the eastward move-
Plains -'
ment of the 8 g kg r, isohume can be faster than the
south-westerly winds on the dry side of a dryline.
Influence of Terrain Sloping terrain together with a dry boundary layer
to the west are clues explaining the sudden movement
Unlike the east-west-oriented drylines in Africa,
of a dryline. Owing to the increase in terrain elevation
drylines in the Great Plains of the United States are
over the southern plains, the depth of the moist layer
typically positioned from south to north. Terrain
decreases toward the west. As a result, the western
contours of the Great Plains are approximately
edge of the moist layer is relatively shallow in the
oriented in the north-south direction; that is, surface
morning. While ascending the moist layer, 8 g kg - Y,
elevation generally increases from east to west in the
isohume will exist farther to the east: That is, the
Great Plains. As a result, the depth of the moist layer -
8 g kg-' ry isohume slopes eastward with height. As
flowing northward from the Gulf of Mexico -
the Sun warms the boundary layer, vertical mixing
decreases from east to west. Sloping terrain gives rise
increases owing to the development of turbulent
to a relatively diffuse eastward gradient of dew point
eddies. Eventually, the turbulent eddies will mix the
temperatures along the western edge of the moist layer.
elevated dry air with the shallow moist air below.
In fact, the diffuse dew point temperature gradient is a
When the mixing is complete, the 8 g kg - r, isohume
dryline environment.
will appear to the east of the original location. Vertical
Air on the west side of a dryline originates over the
turbulent mixing can produce the sudden eastward
elevated plateaus of the south-west United States and
movement of a dryline; far enough eastward, however,
northern Mexico. Prior to the onset of the summer
that advective processes are unable to account for
monsoon over the desert southwest, dew point
the horizontal displacement (Figure 3 ) . While the
temperatures are relatively low. Afternoon high
shallow moist layer is mixed upward, a deeper moist
temperatures typically range between 30 and 35°C;
layer exists to the east. When the moist layer is
as a consequence, the boundary layer can become
deep enough, vertical mixing will be unable to mix dry
several kilometers deep. During summer, hot and
air to the surface; consequently, the eastward move-
dry air over the southwestern United States is replaced
ment of the dryline will end, unless another process is
by the cooler and moister monsoonal flow. As
active.
indicated by satellite images, springtime westerly
Besides vertical turbulent mixing, other processes
flow changes to southerly flow during the summer.
can move a dryline eastward. In some situations, one
Because of the moister monsoonal flow, the dew point
section of a dryline can move eastward while other
temperature gradient associated with the dryline
portions remain relatively stationary (Figure 4). After
environment decreases. That is, onset of the summer
some time, an eastward bulge will appear along an
monsoon signals the end of the development of
otherwise north-south-oriented dryline. Unlike verti-
drylines.
cal mixing of the shallow moist layer with dry air
located aloft, high-momentum air, associated with a
Dryline Movement
jet, is transported to the surface. Surface winds behind
Migration of hot and dry air from the south-west to the the dryline can increase to 20 m s - I. Blowing dust
Great Plains can occur gradually under quiescent often results from gusty surface winds on the west side
conditions, or rapidly when an extratropical cyclone of a dryline bulge. In some cases, streaks of blowing
influences the wind field. Generally, hot and dry air is dust appear in satellite imagery. Dust streaks in
advected by west to south-westerly flow, while warm infrared images, for example, will appear cooler than
and moist air is advected by south to south-easterly the surrounding cloud free ground.
flow. Both air masses merge in a confluent zone over Doppler lidar was used to observe the westward-
the Great Plains. Typically, drylines move eastward moving dryline that occurred over western Texas
during the day; during the night, however, drylines during the evening of 21 April 1985. Observations
may move westward. indicated relatively large differences in density on
Generally, the eastward gradient of water vapor in a either side of the dryline. That is, denser air existed on
dryline environment is relatively diffuse in the early the east side of the dryline, while less dense air existed
morning. The dryline - denoted by the 8 g kg water to the west. As a result, the cool moist air moved
vapor mixing ratio isohume - may exist within a westward, lifting the hot and dry air upward. In other
confluent zone. While the Sun warms the boundary words, the cooler air to the east of the dryline acted like
layer, a dryline may suddenly move eastward. Obser- a density current while lifting the warmer air. Upward
vations have shown that, in some cases, advection is motions of approximately 5 m s - were observed at
unable to account for the relatively rapid eastward the interface of the two air masses.
662 DRYLINE
Vertical
t
E9)
.-a
I
I
I
Dryline
(A)
t Vertical
turbulent
E9)
.-a
I
3 3 mixing
I
I
I
I
I
I
-
Dryline
West East -
Figure 3 Vertical turbulent mixing eroding the shallow portion of
bulge
Figure 4 Morning position of a dryline (dashed) (A) prior to the

a moist layer (A) causing the eastward movement of a dryline (5). development of a dryline bulge (5)over Oklahoma.
Sloping terrain is indicated by the thick solid line, while values of
water vapor mixing ratio (every 1 g kg-') are represented by the
thin solid lines.
ture, T,, is written as eqn [l].
T, = T ( l + 0.61 Y,) [I1

Nonlinear Biconstituent Diffusion
Along with vertical turbulent mixing, horizontal To a first approximation T, is constant across a
mixing has been suggested to be important along a dryline. If T, = 312 K, then the air temperature on the
dryline. Cumulus clouds are typically observed along dry side of a dryline, Tdry, will be 31 1.051 29 K for an
drylines; therefore, upward motion along drylines Ydry value of 5 g kg - On the moist side of a dryline,
must exist. Nonlinear biconstituent diffusion has been Tmoist = 309.358 08 K for an rmoist value of 14 g kg- '.
proposed to explain upward motion along drylines. Assuming the mixed air temperature and water vapor
Theoretical arguments show that horizontal mixing of mixing ratio are given by eqns [2] and [3], the mixed
the hot and dry air, to the west of a dryline, with warm virtual temperature, Tvmixed (eqn [4]) will equal
and moist air, to the east of a dryline, produces a rising 312.002 33 K. Because the virtual temperature of the
mixture that is positively buoyant. Virtual tempera- mixed air is greater than the virtual temperature on
DRYLINE 663
either side of the dryline, the mixed air will rise.

Horizontal
mixing
Tmixed = 0.5 (Tdry + Tmoisr)
4
1 Hot 4 Warm
moist
Tvmixed Tmixcd(1 + 0.61Ymixed) [41 I

Nonlinear biconstituent diffusion assumes constant E
.-m
a,
specific heat across a dryline. When variable specific I
heats are taken into account, Tmlxedas computed
above is replaced with eqn [5].
Hmixed 0.5( H d r y + Hmoist) PI

The mixed enthalpy, Hmlxed, is equal to
(1+ Tmlxed.Solving for the mixed tem-
perature in this case yields eqn [6].
(1 0.84ydry)Tdry + (1 + 0.84ymoist)Tmoist I
Horizontal
Tmixed =
+
2.0( 1 0.84~m1xed) mixing
[61
After substitution, Tmlxed= 310.201 51 K. Calcula-
tion of the mixed virtual temperature gives a value of
311.999 13K. As can be seen, the mixed virtual
temperature is smaller than the virtual temperature
on either side of the dryline; therefore, the mixed
air will sink (see Figure 5 ) . As a result, nonlinear
biconstituent diffusion has been discounted as a
possible explanation for upward motion along a
dryline.
Remote Sensing
In addition to airplanes, cars, and surface stations,
remote sensors have also been employed to provide
measurements of a dryline. Some examples of remote
+ West East
Figure 5 Horizontal mixing leads to upward motion along a

-
sensors are weather satellites, National Weather dryline when constant specific heats are used (A), while downward
Service ( N W S ) Doppler radars, and Doppler lidar. motion results from variable specific heats (B).
Each sensor has strengths and weaknesses, but,
additional information is provided about a dryline.
Weather satellites have several different channels the winds in the upper troposphere. Identification of
that are used to study a dryline. Prior to the upper-level jets, along with the axis of maximum
development of cumulus along a dryline, two infrared winds, is possible; in addition, regions of rising and
channels can be combined to improve detection of the sinking motion can be detected. Visible and infrared
location of a dryline. Brightness temperatures in images show regions of blowing dust to the west of
infrared images are reduced by the presence of water drylines. Visible images are also valuable for providing
vapor in a cloud-free field of view. A moist cloud-free a continuous view of cumulus clouds along the entire
boundary layer to the east of a dryline appears cooler length of a dryline; on the other hand, Doppler radars
than the dry cloud-free boundary layer to the west of a will detect only a segment of a dryline. Unlike visible
dryline. Some infrared channels are more sensitive to images, a cloud-free dryline may appear as a fine line
water vapor than others; thus, the combination of when viewed by a Doppler radar.
channels can provide an improved image for detecting Doppler lidar has also been used to study a dryline.
the western edge of a moist layer associated with a A vertically scanning Doppler lidar was used to study
dryline. Water vapor images provide characteristics of the dryline that occurred on 21 April 1985 over
664 DRYLINE
western Texas. Radial speeds along the beam, V,, are can be made with retrieved values of 8:/&,. That is,
related to the horizontal component of the wind the propagation speed of a density current, c, is found
perpendicular to the dryline, U,and the vertical from eqn [12], where k is a constant near 1and h is the
component of the wind, W, by eqn [7]. depth of the denser fluid.
0.5
V, = -U COS 8 + W sin 8 171
c =k ( g h 2 )
where 8 is the elevation angle of the lidar beam. For
small elevation angles, U is diagnosed by neglecting Although more complex, similar steps are used to
W sine. Once the field of U is determined, W is retrieve temperature and pressure from winds mea-
computed from the two-dimensional anelastic con- sured by a horizontally scanning Doppler radar.
tinuity equation [8], where p ( z ) is a vertically varying
reference state density. Diurnal Variation of Water Vapor Gradient
au
-1 apw Observations have indicated a diurnal variation of the
--
- --
ax p(z) az PI water vapor gradient associated with a dryline. During
the early morning, horizontal gradients of water
Both the pressure and temperature fields are vapor, near the surface, are relatively small in a dryline
retrieved once U and W are known. In two dimen- environment. During the afternoon, however, hori-
sions, the anelastic equations of motion can be written zontal gradients of water vapor may increase to several
as eqns [9] and [lo]. grams per kilogram over a horizontal distance of
approximately 10 km. Although the depth of the moist
layer is shallow during the morning, processes in the
afternoon can cause the depth of the moist layer, along
a dryline, to increase to a few kilometers. Despite
eastward advection of hot and dry air from the west
side of a dryline, water vapor isohumes, along a
dryline, are nearly vertically oriented. In addition, the
relatively large horizontal gradient of water vapor
In eqn [9] and [lo], cp, g, eo,, and n are the specific along a dryline is maintained with height. That is, a
heat at constant pressure, the acceleration of gravity,
vertically oriented moist plume exists along a dryline.
the perturbation virtual potential temperature, the
Along with dryline movement, vertical mixing within
reference state virtual potential temperature, and
a dryline zone has been proposed to explain the
Exner function, respectively. The Exner function is
vertically oriented moist plume.
related to pressure, p , by eqn [ l l ] , where POand R d are
Other state variables also display variations during
the reference state pressure and gas constant for dry
the day along a dryline: virtual potential temperature,
air, respectively.
for example. Observations have shown that values of
8, decrease from west to east through a dryline. In
other words, a horizontal gradient of 8” exists along a
dryline. Horizontal gradients of 6, will initiate a
In a system moving with the dryline, aU/at and a W/at transverse circulation along a dryline.
are assumed negligible relative to the advective terms. The north-south component of the vorticity equa-
Advective terms in the cross-dryline direction are tion may be written as eqn [13].
calculated from U and W to diagnose an/ax;follow-
ing integration, n can be determined. Lastly, the
perturbation virtual potential temperature, e:, is
determined from U , W, and n.
Knowledge of temperature and pressure allows for
further study of a dryline. For example, regions of high
and low pressure can be compared with the wind field I n e q n [ W , B = ( l / g ) [ ( & - eo)/e0- ~ ~ ] a n d 8 , , ~ 0 , ~ , ,
to see whether there is a nonhydrostatic contribution and g represent the virtual potential temperature, the
to the pressure field. Retrieval analysis on the 21 April reference state potential temperature, the total
1985 dryline showed the existence of a vertical condensate mixing ratio, and the acceleration of
pressure gradient that aided in lifting hot and dry air gravity, respectively. In addition, T, and T, are the
up and over the cooler and moister air. Comparison of components of the turbulent mixing tendency vector.
the dryline speed with the speed of a density current In a situation where only gradients in 8, exist, Dw,/Dt
DRYLINE 665
will reduce to eqn [ 141. storms can produce outflows in excess of 25 m s - ’,

hail that can damage crops and structures, and
tornadoes. As an example, the severe and tornadic
thunderstorms that formed in Kansas and Oklahoma
That is, decreasing values of 8, from west to east on 26 April 1991 were associated with a dryline.
through a dryline will initiate a transverse solenoidal All of the methods used to denote the position of a
circulation along a dryline. In particular, the circula- dryline may yield drylines in different locations. That
tion will increase the easterly flow, on the moist side of is, a dryline zone may be located several tens of
a dryline, causing an increase of horizontal conver- kilometers east of the 8 or 9 g kg - water vapor
gence. An increase of horizontal convergence along a mixing ratio isohume. If the 8 g kg- water vapor
dryline will develop and maintain the relatively large mixing ratio isohume is used to denote a dryline, then
horizontal gradient of water vapor. An examination of thunderstorm initiation may occur east of a dryline. In
the frontogenetic forcing equation illustrates the other words, because of the arbitrary methods used to
process. denote the location of a dryline, thunderstorms may
The material derivative of Vhrv can be expressed as appear to develop several tens of kilometers east or
eqn [15]. west of a dryline.
Nowcasting thunderstorm development along
a dryline can be challenging. Thunderstorm initia-
-
D(Vhrv) --.- F,e, + Fyey tion can be supressed by the presence of an overlying
Dt
inversion. At times, however, the overlying in-
In eqn [ 151, e, and ey are unit vectors and F, and F y are version might be completely eroded, and convective
given by eqns [ 161 and [ 171. inhibition may be zero. Even though convective
inhibition might be zero, convective development
may continue to be supressed along a dryline. After-
noon vertical profiles of temperature typically exhibit
neutral stability (constant potential temperature) from
the surface to about 2km. Owing to the neutral
stability, upward lifting encounters little resistance.
After the lifting condensation level is reached, a
parcel can rise freely; that is, the lifting condensation
As an approximation, suppose that the dryline level is the level of free convection according to parcel
exhibits negligible variations along the north-south theory.
direction. Furthermore, observations show that the One assumption contained in parcel theory is the
water vapor isohumes are almost vertically oriented; absence of mixing of a parcel with the environment.
in addition, assume that r, of a parcel is constant. Then Consider the following situation. Assume that the
F , = -(au/ax)(ar,/ax)and F , = 0. Because the pre- potential temperature is constant from the surface to
existing gradient of water vapor, &,/ax, is positive at a 2 km; further, assume the water vapor mixing ratio is
dryline, an increase in convergence, au/ax, will cause constant from the surface to 1km followed by a
F, to be positive and thus D(Vhr,)/Dt will be positive. marked decrease in the layer from 1km to 2 km.
In short, a horizontal gradient of water vapor will According to parcel theory, a parcel is lifted from the
increase if acted on by a horizontally convergent wind surface to the lifting condensation level, assumed to be
field. 2 km, while maintaining the surface water vapor
mixing ratio value. In contrast to parcel theory,
Thunderstorm Development
when mixing of a parcel with the environment is
Identification of a dryline environment is an important taken into account, then the potential temperature of
forecasting issue. Processes that occur in the dryline the parcel will remain the same from the surface to
environment may lead to the development of a dryline 2 km: The layer is neutrally stable. Water vapor mixing
zone. Although the dryline zone contains a greater ratio values will also remain the same, but only from
horizontal dew point temperature gradient, relative to the surface to 1km. Within the layer from 1 km to
the dryline environment, the most important char- 2 km, the environmental water vapor mixing ratio
acteristic of the dryline zone is upward motion. If decreases, and mixing of the parcel with the environ-
conditions are favorable, upward motion may lead to ment will decrease the water vapor mixing ratio of the
convective initiation: thunderstorms. Even though the parcel. As a result of mixing the lifting condensation
resulting thunderstorms produce beneficial rains to level will increase, and condensation may occur at
agricultural regions of the southern plains, such a temperature that is cooler than the environment.
666 DRYLINE
In other words, mixing can result in positive values of force in a particular region; furthermore, values of
convective inhibition that upward motion, along a horizontal advection of water vapor can be compared
dryline, is unable to overcome. As a result, thunder- to turbulent mixing of water vapor. Two methods of
storm development will be supressed. analysis exist with a numerical model: Eulerian or
Lagrangian frames of reference. In an Eulerian frame,
comparisons of values of forcing are made at a grid
Dryline Simulations point; in contrast, comparisons of values of forcing
Numerical models can be used to provide additional along trajectories are conducted in a Lagrangian
information about dryline evolution. Field experi- frame. Forcing terms in derived equations can also
ments have provided a wealth of knowledge about be analyzed: the three-dimensional vorticity equation
dryline characteristics and morphology; however, the and the diagnostic equation for pressure are two
use of a numerical model to address some questions examples.
about a dryline might be more practical. For example,
how would the evolution of a dryline change if the Limitations of Numerical Models
slope of the terrain were altered in the Great Plains of Because of the design of a given numerical model and
the United States? Although numerical models have computer resources, simulations of drylines have
limitations, two- and three-dimensional dryline simu- limitations. Computer power can limit the size of a
lations have provided beneficial results. region over which to run a simulation; moreover,
horizontal and vertical grid spacings can also be
Benefits of Numerical Models limited by computer resources. Restrictions placed on
horizontal grid spacings can prevent some processes
Numerical models are beneficial by providing answers
from being resolved. A simulation that has horizontal
to questions that dryline field experiments may be
grid spacings equal to 25 km will represent and resolve
unable to answer. One consequence of the funda-
a dryline environment, but 25 km grid spacing is too
mental theorem of differential equations is that a
large to represent or resolve a dryline zone. The
solution from any simulation is uniquely determined
choices of lateral, top, and lower boundary conditions
from an initial state. As a result, changes in dryline
can influence the quality of a simulation. Because of
evolution with changes in an initial state can be
different time-stepping methods, different methods of
examined with a numerical model. For instance, the
distributing prognostic variables on a grid, and
slope of the terrain can be altered; vegetation
different parameterizations of sub-grid-scale mixing
type along with soil moisture may also be
or microphysics, variations of results from a dryline
varied; and vertical shear can be changed in the initial
simulation can occur. Although limitations exist,
state.
numerical models are a useful tool for understanding
Regions of missing data are unavoidable during a
dryline evolution.
dryline field experiment. Observational data sets are
incomplete in the horizontal direction, in the vertical
Two-Dimensional Dryline Simulations
direction, and in time. Only portions of a dryline can
be sampled by an airplane, an automobile, a radar, a Two-dimensional simulations have been used to
lidar, a satellite, or surface stations. Although a investigate the horizontal gradient of water vapor
numerical model creates a uniformly discrete data associated with a dryline environment. Earlier work
set, the density of simulated dryline data can be demonstrated that vertical turbulent mixing can erode
relatively high. For example, a model can produce data the shallow portion of a moist layer; hence, the
uniformly spaced every 5 km over hundreds of kilo- eastward movement of a dryline (9gkg-1 rv iso-
meters in each horizontal direction. In addition, the hume). Other two-dimensional dryline simulations
relatively large horizontal grid of data could be have shown that surface convergence near a dryline is
generated every 500m above the ground, and the linked to differential surface heating. That is, the
three-dimensional data set can be produced as increase of the horizontal gradient of water vapor
frequently as one simulated time step: every 60s in resulted from solenoidal and frontogenetic forcing on
some cases. That is, a three-dimensional simulation of the scale of 100 km. Additional two-dimensional
a dryline can produce an overwhelming amount of simulations showed not only that downward trans-
data. port of westerly wind by vertical mixing on the dry side
An additional benefit of using a numerical model is of the dryline (8 g kg - rv isohume) moved the dryline
the ability to evaluate individual forcing terms in eastward, but also that easterly winds on the moist side
prognostic or derived equations. For example, Cori- of the dryline were generated by an inland sea breeze.
olis forcing can be compared to a pressure gradient Both processes resulted in horizontal convergence that
DRYLINE 667
maintained the horizontal water vapor gradient on the

scale of 150 km. Typical horizontal grid spacings used
for these simulations were approximately 30 km.
Because of the results from dryline field experi-
ments, horizontal grid spacings were reduced in recent
simulations to resolve the dryline zone. In some
studies, horizontal grid spacings of 5 km were used;
as a result, the increase of the horizontal water vapor
gradient associated with the dryline zone was shown
to be a consequence of solenoids and frontogenetic
forcing on the scale of a few tens of kilometers. In some
cases, horizontally varying soil moisture was found to
be necessary for the development of a dryline zone.
Other two-dimensional simulations showed that soil
moisture was less important than vertical wind shear
and sloping terrain. Although soil moisture was found
to be relatively unimportant, this result might be a
consequence of horizontal grid spacings that were too
large to resolve the dryline zone.
1 Dryline environment
Three-Dimensional Dryline Simulations

Three-dimensional simulations of drylines have been
used to examine the role of horizontally varying soil
moisture on dryline evolution. For example, soil
moisture was varied in a series of sensitivity experi-
ments for the dryline environment that occurred in
early May 1979. Results suggested that variable soil
moisture influenced the evolution of the following:
e Elevated mixed layer (lid)
e Dryline environment
e Easterly ageostrophic flow over east Texas and the
Gulf of Mexico
e
e
Potential instability of air under the lid
Location and intensity of precipitation.
Simulations of the 15, 16, and 26 May 1991 dryline

zones have all suggested that horizontally varying soil
+ West East -
Figure 6 Vertical cross-section of a dryline environment in the
morning (A). Watervapor mixing ratios (thin solid lines) are labeled
moisture would promote the formation of a dryline
zone. That is, simulations have shown that a dryline
’.
every 4 g kg - A thermally forced secondary circulation lifts warm
moist air (thick upward-pointing arrows) to the level of free
zone is thermally forced when a soil moisture field has convection, leading to convective initiation (B).
horizontal variations. In addition, the water vapor
field experiences frontogenetic forcing in response to
the thermally forced secondary circulation. circulation. Warm moist air may rise to the lifting
Convective initiation along drylines has also been condensation level, and forced cumulus clouds can
simulated with a three-dimensional model. Using form. As a result of further ascent by the secondary
several two-way interactive nested grids, horizontal circulation, the level of free convection may be
grid spacings as small as 1km have been achieved. achieved, leading to the initiation of thunderstorms
Studies of this kind have noted that rising motion - (Figure 6).
associated with the secondary circulation within the
dryline zone - is thermally forced; as a consequence,
solenoidal circulations develop along with fronto- See also
genetic forcing. Convective initiation results from the Boundary Layers: Overview. Fronts. Lidar: Doppler.
rising branch of the secondary circulation; that is, Mesoscale Meteorology: Overview. Numerical Mod-
lifting of warm moist air occurs within the secondary els: Methods. Radar: Doppler Radar. Satellite Remote
668 DUST
Sensing: Precipitation; Surface Wind. Severe Storms. Cotton WR and Anthes RA (1989) Storm and Cloud
Turbulence and Mixing. Vorticity. Weather Predic- Dynamics. New York: Academic Press.
tion: Severe Weather Forecasting. Holton JR (1992) Introduction to Dynamic Meteorology,
3rd edn. New York: Academic Press.
Iriabrne JV and Godson WL (1981) Atmospheric Thermo-
Further Reading dynamics. Dordrecht: Reidel.
Bluestein HB (1992) Synoptic-Dynamic Meteorology in Pielke RA (1984)Mesoscale Meteorological Modeling. New
Midlatitudes, vol. 1, Principles of Kinematics and York: Academic Press.
Dynamics. New York: Oxford University Press. Ray PS (1986) Mesoscale Meteorology and Forecasting.
Bluestein HB (1993) Synoptic-Dynamic Meteorology in Boston: American Meteorological Society.
Midlatitudes, vol. 2 , Observations and Theory of Stull RB (1989) An Introduction to Boundary Layer
Weather Systems. New York: Oxford University Press. Meteorology. Dordrecht: Kluwer Academic.
I N Sokolik, University of Colorado at Boulder, CO, USA the threshold wind speed required to initiate the
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. motion of dust particles. Currently, dust emission
schemes allow predicting the threshold wind speed as a
function of the particle size distribution of the parent
Introduction soil and the surface roughness. However, a single value
Mineral aerosols (often referred to as dust) are of the threshold wind speed of 6.5m s-' at 1 0 m
ubiquitous in the atmosphere. Several sources con- height is often used in modeling the dust emission in
tribute to the dust loading such as soil-derived wind- the general circulation models (GCMs), mainly be-
blown dust, industrial processes, and volcanic cause of the lack of input data on surface properties
eruptions. This article deals with soil-derived dust, required for dust production schemes.
describing the properties, distribution, and effects of Once winds are higher than a threshold value, the
mineral aerosols in the atmosphere. movement of dust particles can be initiated either by
aerodynamic forces (called suspension) or by the
impact of saltating soil grains (a process known as
Dust Sources and Production bombardment or sandblasting). Only the finest parti-
cles with diameters below about 60pm can be
Mechanisms suspended and transported upward by turbulent
The major dust sources, which are arid and semi-arid eddies. However, these particles make up only a small
regions of the world, cover approximately 33% of the fraction of soil grains, most of which are present in the
global land area and have a dispersed geographical form of aggregates of larger sizes due to strong
distribution. The largest source is the Sahara-Sahel interparticle cohesive forces. The aggregates typically
region of North Africa, emitting about 500-800 Mt of have diameters from about 60-80ym to several
dust annually. Central Asia is the second largest dust hundred micrometers. The bombardment-induced
source, but less studied than the Saharan sources. It is breakage of aggregates results in the release of fine
estimated that from 6 to 12 Mt year - of Asian dust' particles which is believed to be a major contributor to
are lifted by winds into the atmosphere. Other aeolian dust entrainment. The effects of vegetation
important dust production regions are located in the and surface crusting further complicate the dust
Arabian Peninsula, Australia, south-western United emission. The dust flux measurements reported in the
States, and north-west India. Overall, current esti- literatureareinthe rangeof 10-5to 10-12gcm-2s-1,
mates of global mean dust emission are in the range being centered around 10- * to 10- g c m P 2 s- '.
from 1000 to 5000 Mt year-'. The dependence of dust emission on surface features
The existing uncertainties of dust burden are mainly indicates that modification of surface type (for in-
due to the complexity of dust production mechanisms, stance, by human activities) may alter dust burden.
which exhibit large spatiotemporal variability. The Various human activities (such as agriculture, con-
emission of dust is controlled by both meteorological struction, deforestation, etc.) lead to land surface
characteristics (especially winds) and soil surface disturbances. Disturbed surfaces are likely to be more
features. Roughness, aggregated structure, texture, efficient dust sources than natural undisturbed lands.
mineralogy, and moisture of the soil surface determine The dust fraction produced as a result of human
668 DUST
Sensing: Precipitation; Surface Wind. Severe Storms. Cotton WR and Anthes RA (1989) Storm and Cloud
Turbulence and Mixing. Vorticity. Weather Predic- Dynamics. New York: Academic Press.
tion: Severe Weather Forecasting. Holton JR (1992) Introduction to Dynamic Meteorology,
3rd edn. New York: Academic Press.
Iriabrne JV and Godson WL (1981) Atmospheric Thermo-
Further Reading dynamics. Dordrecht: Reidel.
Bluestein HB (1992) Synoptic-Dynamic Meteorology in Pielke RA (1984)Mesoscale Meteorological Modeling. New
Midlatitudes, vol. 1, Principles of Kinematics and York: Academic Press.
Dynamics. New York: Oxford University Press. Ray PS (1986) Mesoscale Meteorology and Forecasting.
Bluestein HB (1993) Synoptic-Dynamic Meteorology in Boston: American Meteorological Society.
Midlatitudes, vol. 2 , Observations and Theory of Stull RB (1989) An Introduction to Boundary Layer
Weather Systems. New York: Oxford University Press. Meteorology. Dordrecht: Kluwer Academic.
I N Sokolik, University of Colorado at Boulder, CO, USA the threshold wind speed required to initiate the
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. motion of dust particles. Currently, dust emission
schemes allow predicting the threshold wind speed as a
function of the particle size distribution of the parent
Introduction soil and the surface roughness. However, a single value
Mineral aerosols (often referred to as dust) are of the threshold wind speed of 6.5m s-' at 1 0 m
ubiquitous in the atmosphere. Several sources con- height is often used in modeling the dust emission in
tribute to the dust loading such as soil-derived wind- the general circulation models (GCMs), mainly be-
blown dust, industrial processes, and volcanic cause of the lack of input data on surface properties
eruptions. This article deals with soil-derived dust, required for dust production schemes.
describing the properties, distribution, and effects of Once winds are higher than a threshold value, the
mineral aerosols in the atmosphere. movement of dust particles can be initiated either by
aerodynamic forces (called suspension) or by the
impact of saltating soil grains (a process known as
Dust Sources and Production bombardment or sandblasting). Only the finest parti-
cles with diameters below about 60pm can be
Mechanisms suspended and transported upward by turbulent
The major dust sources, which are arid and semi-arid eddies. However, these particles make up only a small
regions of the world, cover approximately 33% of the fraction of soil grains, most of which are present in the
global land area and have a dispersed geographical form of aggregates of larger sizes due to strong
distribution. The largest source is the Sahara-Sahel interparticle cohesive forces. The aggregates typically
region of North Africa, emitting about 500-800 Mt of have diameters from about 60-80ym to several
dust annually. Central Asia is the second largest dust hundred micrometers. The bombardment-induced
source, but less studied than the Saharan sources. It is breakage of aggregates results in the release of fine
estimated that from 6 to 12 Mt year - of Asian dust' particles which is believed to be a major contributor to
are lifted by winds into the atmosphere. Other aeolian dust entrainment. The effects of vegetation
important dust production regions are located in the and surface crusting further complicate the dust
Arabian Peninsula, Australia, south-western United emission. The dust flux measurements reported in the
States, and north-west India. Overall, current esti- literatureareinthe rangeof 10-5to 10-12gcm-2s-1,
mates of global mean dust emission are in the range being centered around 10- * to 10- g c m P 2 s- '.
from 1000 to 5000 Mt year-'. The dependence of dust emission on surface features
The existing uncertainties of dust burden are mainly indicates that modification of surface type (for in-
due to the complexity of dust production mechanisms, stance, by human activities) may alter dust burden.
which exhibit large spatiotemporal variability. The Various human activities (such as agriculture, con-
emission of dust is controlled by both meteorological struction, deforestation, etc.) lead to land surface
characteristics (especially winds) and soil surface disturbances. Disturbed surfaces are likely to be more
features. Roughness, aggregated structure, texture, efficient dust sources than natural undisturbed lands.
mineralogy, and moisture of the soil surface determine The dust fraction produced as a result of human
DUST 669
activities is called anthropogenic dust. Recent esti- 100 pm) become apparent. There remain large uncer-
mates show that anthropogenic dust could be as much tainties in in situ dust particle size distributions and
as 30-50% of total dust load, but this remains concentrations measured from an aircraft, because of
uncertain. uncontrolled particle losses in the inlet. Therefore, an
Natural or anthropogenic climate variability also understanding of dust particle spectra mainly stems
alter dust production. Frequent droughts and strong from ground-based measurements.
winds are likely to enhance dust emission, while rains Near the source regions, dust concentrations could
tend to inhibit it. be as high as a few thousand particles per cubic
centimeter during the dust storm. Gravitational set-
tling and rainout are major removal processes affect-
Dust Composition, Morphology, and ing dust concentration and particle spectra during
long-range transport. Other important processes,
Optical Properties shaping particle spectra and composition, are hetero-
The composition of bed surfaces and mobilization geneous chemistry on dust particle surfaces, cloud
processes determine the initial particle size distribu- processing of dust particles, and interactions with
tion of airborne dust, its composition, and the degree other atmospheric aerosols. All these processes control
of particle aggregation. Both the particle size distri- the lifetime (or residence time) of dust particles in the
bution and composition might be altered during dust atmosphere which is up to 2 weeks. During this time
transport in the atmosphere. period, dust can be carried great distances up to a few
The main species found in dust derived from soil thousand kilometers, affecting large regions. Fre-
surfaces are quartz, various clays (e.g., kaolinite, illite, quently, Saharan dust plumes reach the Caribbean,
montmorillonite), calcite, gypsum, and iron oxides the Gulf of Mexico, and the south-east coast of the
(e.g., hematite, goethite) among others. These miner- United States, while Asian dust can be transported as
als are characterized by very different physical and far as the west coast of the United States. It is believed
chemical properties. For instance, different minerals that long-transported dust particles are mainly in the
have different abilities to adsorb water vapor and size range from about 0.05 to 10 pm, although coarse
other chemically important atmospheric gases. Each particles of 100 pm diameter have been measured at
mineral has distinct spectral optical constants (or distances of several thousand kilometers from the dust
refractive indices). source.
Consequently, the properties of dust as a mixture are Individual particle analysis reveals that dust par-
determined by the relative abundance of various ticles often occur as irregular (nonspherical) aggre-
minerals and their mixing state. Yet numerous climate gates of minerals as well as other species. Yet no
and remote sensing studies have considered dust as a systematic studies of the shapes and sizes of dust
single generic species. This is partly due to the particles have been performed so far. As a result, a
complexity of quantitative determination of the min- simplified assumption that dust particles are spherical
eralogical composition and a lack of data. In addition, is often made.
the dust composition varies from source to source. For The abundance, composition, and morphology of
instance, dust in the Sahelian region is characterized by dust particles determine their optical properties and
a high Fe/A1 ratio due to the abundance of ferralitic hence radiative effects in the atmosphere. Dust parti-
soils. In contrast, soils in the semi-arid regions of cles can scatter and absorb electromagnetic radiation
Central Asia contain less Fe. The difference in the in a wide range of wavelengths from ultraviolet (UV)
amount of iron oxides is of special importance because to infrared (IR).To model dust radiative impacts, one
they control primarily the ability of dust particles to needs to know optical properties of dust such as the
absorb and scatter sunlight. optical depth, single scattering albedo, and scattering
Particle size spectra can have one or several modes, phase function. These characteristics are also func-
each characterized by a specific composition. In tions of the location, time, altitude, and wavelength.
general, various clay particles tend to be smaller in The optical depth of dust plumes is largest near the
size than those made of quartz or carbonates. Ground- source and it decreases farther from the source being
based measurements near dust sources reveal that a controlled largely by particle concentrations. Over the
common mode with the particle distribution centered oceans, the highest optical depths are likely to occur in
around particles whose diameter is 3-6 pm is always regions influenced by dust transport. At visible wave-
present regardless of atmospheric conditions. When lengths, optical depths as high as 10 have been
the wind speed exceeds the threshold value, a submi- measured during dust storms. It has a weak depend-
crometer mode (particle sizes between 0.1 and 1pm) ence on the wavelengths in the visible region, but
and a coarse mode (particle diameters between 20 and various spectral features occur in the IR. Observations
670 DUST
of the optical depth in the IR window region suggest over a given geographical region. For instance, the
that it is about 2-10 times smaller than the visible interannual variability of Asian dust is likely to be
optical depth. linked with shifts in circumpolar vortex dynamics. In
In turn, the single scattering albedo, which is defined addition, precipitation, which is highly variable from
as the ratio of scattering and extinction coefficients, year to year, may affect the atmospheric dust cycle
does not depend on particle concentrations but rather through both a change in soil moisture and by washout
on the particle composition, state of mixing, and sizes. during transport.
It is a key optical characteristic for calculating the Despite the varying frequencies and strengths of
heating or cooling effects of aerosols. Based primarily dust storms, they do occur each year. Also some
on modeling, it is believed that the single scattering seasonal transport patterns are repeatable from year to
albedo of dust is in the range from 0.65 to 0.95 at the year. For instance, the transport of Saharan dust shifts
solar wavelengths and is characterized by strong northward from about 5" N during winter to about
wavelength dependence, increasing from W to near- 20" N during summer. Asian dust outbreaks reach
IR. This remains poorly constrained because of their maximum in spring.
numerous deficiencies found in different techniques Much of our understanding of the large-scale
used to measure light scattering and light absorption features of dust transport is derived from satellite
by dust particles in the atmosphere. In addition, the imagery. Dust plumes are readily observed in W,
single scattering albedo may vary during transport visible, and IR channels of satellite sensors. This
depending on which species aggregate with dust provides a unique tool to characterize the areal
particles. It has been demonstrated that dust particles distribution of dust on a global scale. Figure 1 shows
internally mixed with other atmospheric species can the global distribution of W-absorbing aerosols (such
have drastically different properties from those that as dust and carbonaceous particles) observed by the
are evident at the dust source. satellite Total Ozone Mapping Spectrometer (TOMS)
Although the scattering phase function of dust during April 1998. The large dust plumes covering
particles is crucial for remote sensing applications, North Africa and the tropical Atlantic Ocean are
there are only a few actual measurements. Therefore, clearly seen.
this and other optical characteristics of dust are Although limited, aircraft and lidar measurements
calculated by applying Mie theory for a selected size reveal that dust plumes have a complex multilayered
distribution and spectrum of the refractive index. No structure. For instance, transport of Saharan dust
models have been proposed so far to link the optical occurs at higher altitudes in a layer that typically
properties of dust to its life cycle. reaches 5-6 km, although one or several layers might
be present below it. Concentrations aloft are usually
several times greater than in the marine boundary
Spatial and Temporal Variations of layer. Dust layers are often intermixed with layers of
other aerosols or clouds. The combination of a lidar
Dust Burden in the Atmosphere system planned to fly in 2003 on the CALIPSO mission
Both observations and model simulations indicate that and new multispectral passive sensors on EOS-Terra
dust burden has a complex spatial (horizontal and and EOS-Aqua satellites has the potential to detect the
vertical) and temporal pattern. The dust burden is presence of layers of dust and other aerosols in the
shaped by the source strength, dust production mech- atmosphere together with their specific properties and
anisms, dust removal processes, and transport in the their interaction with cloud layers.
atmosphere.
First of all, initial dust emission has complex spatial
and temporal variations because dust sources have a
Dust Effects
dispersed geographical distribution and their
strengths vary with time. The duration of dust storms Mineral particles, by virtue of their chemical, physical,
is typically a few days having pulses of strong winds of and optical properties, cause the various impacts upon
several hours. The variability of atmospheric transport the atmosphere and overall climate system. Table 1
and size- and composition-dependent removal proc- lists the major effects of dust in the atmosphere and
esses further contribute to the heterogeneous distri- their importance grouped into two broad categories:
bution of dust. direct and indirect radiative effects. They may all
It has been demonstrated that dust transport exhib- result in alteration of the Earth's energy balance and
its substantial seasonal and interannual variability thus influence the climate.
that is mainly controlled by changes in both the The direct radiative effects are caused by the interac-
atmospheric circulation and the hydrological cycle tion of dust particles with atmospheric electromagnetic
DUST 671
Figure 1 Global distribution of UV-absorbing aerosols observed by the satellite TOMS during April 1998.
Table 1 Major effects caused by the presence of dust in the atmosphere

Impact Importance
Direct radiative effects

Causes the radiative forcing at the top of the atmosphere (solar Affects energy balance of Earth's climate system
plus IR)
Causes radiative forcing at the surface (solar plus IR) Affects surface temperature and surface-air exchange
processes
Causes radiative heating/coolingwithin a dust layer Affects temperature profile and atmospheric dynamics
(solar plus IR)
Affects visibility Decreases visibility
Indirect radiative effects

Impacts on clouds
Serves as ice nuclei Affects the properties, amount, and distribution of water and ice
Serves as cloud condensation nuclei clouds and hence the Earth's radiation balance and
hydrological cycle
Impact on atmospheric composition and chemistry

Alters actinic flux (UV) Affects photolysis rates and photochemistry and hence
radiative transfer processes in the atmosphere
Adsorbs chemically important gases Affects the partitioning of semivolatile trace species
Providesparticle surfaces for heterogeneouschemical reactions (such as nitric acid and ammonia): alters the abundance of
radiatively important gaseous species
672 DUST
radiation via scattering, absorption, and emission. In Although beyond the scope of this article, it is
contrast t o greenhouse gases, dust may cause either a important t o point out that dust has other important
positive or negative total (solar plus infrared) radiative diverse impacts. For example, they pose a health
forcing, depending on its optical properties and threat, affect biogeochemical processes in the oceans,
distribution in the atmosphere as well as atmospheric affect terrestrial ecosystems, cause property damage,
conditions and surface reflectance. The radiative affect agricultural production, etc. Large uncertainties
forcing and heating/cooling rates of dust may be of in the chemical, physical, and radiative properties of
different signs in the solar and infrared regions. A dust dust render quantitative assessments of its effects
layer is likely t o heat itself but may cause an overall problematic.
cooling over dark surfaces. Models predict that direct
radiative forcings at the top of the atmosphere and at
the surface have a complex geographical distribution See also
because dust sources and sinks are not uniformly Aerosols: Climatology of Tropospheric Aerosols; Obser-
distributed and because of the short lifetime of mineral vations and Measurements; Physics and Chemistry of
particles in the atmosphere. Thus, the presence of dust Aerosols: Role in Radiative Transfer. Chemistry of the
may enhance greenhouse gas warming in some regions Atmosphere: Principles of Chemical Change. Deposi-
and oppose it in others. tion. Deserts and Desertification. Drought. Global
Dust particles can also affect Earth’s radiation Change: Human Impact of Climate Change. Hydrology:
Soil Moisture. Lidar: Atmospheric Sounding Introduction.
balance indirectly by altering the properties, amount,
Observations for Chemistry (In Situ): Particles. Op-
and distribution of clouds as well as by affecting the
tics, Atmospheric: Optical Remote Sensing Instruments.
radiatively important gases through physicochemical Radiation (Solar). Radiative Transfer: Absorption and
processes. These are termed indirect radiative effects. Thermal Emission: Scattering. Reflectance and Albedo,
Observations reveal that dust particles may serve as Surface. Satellite Remote Sensing: Aerosol Measure-
cloud condensation nuclei for both water and ice ments. Tropospheric Chemistry and Composition:
clouds. The condensation properties of these nuclei are Aerosols/Particles.
likely to be controlled by hygroscopicity and surface
properties of dust particles. Currently, mechanisms of
dust-cloud interactions are not well understood. Further Reading
Although poorly quantified, several mechanisms Charlson RJ and Heintzenberg J (eds) (1995) Aerosol
governing dust-chemistry interactions have been Forcing of Climate. New York: Wiley.
identified. By altering UV radiation, the presence of Denterner F, Carmichael G, Zhang Y, Crutsen P, and Lelifeld
dust affects photolysis rates and hence the photo- J (1996)The role of mineral aerosols as a reactive surface
chemical formation of such a radiatively important in the global troposphere. Journal of Geophysics Re-
species as ozone. In addition, dust particles have a high search 101: 22 869-22 890.
surface area per unit mass which makes them an ideal Guerzoni S and Chester R (1996)The Impact of Desert Dust
across the Mediterranean. Netherlands: Kluwer Aca-
site for heterogeneous chemical reactions. Some
demic Publishers.
atmospheric gases (e.g., nitric acid) can directly Prosper0 J (1996)The atmospheric transport of particles to
condense on dust particles. These mechanisms provide the ocean. In: Ittekkot V, Schafer P, Honjo S, and Depetris
a plausible explanation of the elevated sulfate and PJ (eds.) Particle Flux in the Ocean. New York: Wiley.
nitrate levels associated with dust particles which have Pye K (1987)Aeolian Dust and Dust Deposits. San Diego:
been observed over both the Pacific Ocean and Academic Press.
Atlantic Ocean. However, the quantitative inter- Sokolik IN and Toon OB (1996) Direct radiative forcing
pretation of these observations remains highly un- by anthropogenic mineral aerosols. Nature 381:
certain mainly because the heterogeneous chemical 68 1-68 3.
reactions and adsorption processes on the surfaces of Sokolik IN, Winker D, Bergametti G, Gillette D, Carmichael
dust particles dispersed in the atmosphere have G, Kaufman Y, Gomes L, Schuetz L and Penner J (2001).
Introduction to special section on mineral dust: out-
not been studied until recently. Models indicate
standing problems in quantifying the radiative impact of
that dust-induced changes in atmospheric chem- mineral dust. Journal of Geophysical Research 106:
istry strongly depend on dust particle composition 18015-18028.
and sizes, concentrations of gaseous species, and Tegen I, Lacis AA, and Fung I (1996) The influence on
ambient conditions (e.g., temperature and relative climate forcing of mineral aerosols from disturbed soils.
humidity). Nature 380: 419-422.
DYNAMIC METEOROLOGYI Overview 673
Contents
Overview
Balanced Flows
Potential Vorticity
Primitive Equations
Waves
The latitudinal gradient of potential vorticity pro-

Overview vides the mechanism for generation of global-scale
planetary waves, which are primary features of the
J R Holton, University of Washington, Seattle, WA, USA climate system. Superposed on these global waves are
transient cyclones and anticyclones, whose energy is
derived primarily from the potential energy associated
with the mean Pole-to-Equator temperature gradient.
Study of the development and evolution of transient
Introduction weather disturbances, and of dynamical mechanisms
Dynamic meteorology is the branch of fluid dynamics for producing intraseasonal and interannual climate
concerned with the meteorologically significant variations, are among the principal areas of study in
motions of the atmosphere. It forms the primary dynamic meteorology.
scientific basis for weather and climate prediction, and
thus plays a primary role in the atmospheric sciences.
Most of the meteorologically important motions
The Static Atmosphere
studied in dynamic meteorology are profoundly The vertical distribution of pressure, density and
influenced by the facts that the Earth is a rapidly temperature is determined by the hydrostatic approx-
rotating planet, and that the atmosphere on average imation and the ideal gas law. The hydrostatic
has stable density stratification. These facts make approximation, which is derived from the vertical
the fluid dynamics of the atmosphere very different component of the momentum equation (Newton's
from traditional engineering fluid dynamics. Planetary second law of motion), expresses the balance between
rotation places strong constraints on large-scale the vertical component of the pressure gradient force
horizontal motions; stable stratification places and the force of gravity. The ideal gas law, or equation
strong constraints on vertical motions. These of state, expresses the relationship between pressure,
constraints can be understood by considering the density, and temperature in an ideal gas. These two
fundamental physical laws governing motions of the equations may be combined to form the hyposometric
atmosphere. equation, which determines the thickness of the layer
The motions of the atmosphere are governed by the of temperature T confined between the two pressure
laws for conservation of mass, conservation of mo- surfaces p 2 and p 1 :
mentum, and conservation of thermodynamic energy.
Application of these laws to motions with horizontal
scales of several hundred kilometers or greater leads to
simple relations among the horizontal wind, pressure,
and temperature distributions. These relations form a where @isthe geopotential, Z is geopotential height, T
set of diagnostic relations essential for understanding is the absolute temperature, R (= 287 J kg-' K-') is
the motions that generate weather disturbances. Such the gas constant for dry air, and g (= 9.81 m sP2) is the
motions are generally rotational in character. They acceleration due to gravity. The hypsometric equation
can be characterized by a conservable property shows that atmospheric pressure decreases more
known as the potential vorticity, which is the fluid rapidly in cold air than in warm air. Thus the average
dynamical analogue of spin angular momentum in altitude of a given upper-level pressure surface
solid mechanics. (e.g., the 500 hPa surface) decreases towards higher
674 DYNAMIC METEOROLOGY/ Overview
latitudes owing to the decrease of mean temperature nate. This isentropic coordinate system is useful
with latitude. for analysis of adiabatic motions, since the pre-
Hydrostatic balance requires that pressure decrease diction of atmospheric motions for such conditions
monotonically with height in the atmosphere. Pressure is reduced from a three-dimensional problem to a two-
may thus be substituted for height as a vertical dimensional problem on each isentropic coordinate
coordinate; this has the advantages of eliminating surface. Because diabatic temperature changes asso-
explicit reference to the density field in the equations of ciated with large-scale weather disturbances in the
motion; but it has the disadvantage that pressure extratropics are often much smaller than adiabatic
varies exponentially with altitude so that equal changes, isentropic analysis has proved valuable
altitude increments correspond to rapidly decreasing for the study of air motions associated with such
pressure increments as height increases. For this disturbances.
reason, in dynamical meteorology it is often useful to
use log-pressure coordinates in which the independent
vertical coordinate z is defined by The Equations of Motion
It is convenient to express the basic equations of
dynamic meteorology in a coordinate system rotating
where H = R T o / g is an atmospheric scale-height, with the Earth, and with the log-pressure altitude
po = lo3 hPa (1000 mbar), and TOis a mean temper- defined by eqn [2] as the independent vertical coordi-
ature. Comparing with eqn [I], it is clear that the log- nate. The approximate conservation equations for
pressure coordinate corresponds to actual height for horizontal momentum, mass, and thermodynamic
an isothermal atmosphere at temperature TO.Under energy are then as follows:
most conditions the departure of z from actual altitude
DV
is small enough to be neglected. -= - f k x V - V@+ Fr
In the absence of precipitation, changes in temper- Dt
ature following the motion of individual parcels of air
are controlled primarily by adiabatic expansion and i a
V.V+--(pow) =0
compression as the air parcels move to lower or higher Po az
pressure. The thermodynamic state of such parcels can
be characterized by the potential temperature, 8.
Potential temperature is the temperature that a parcel
of dry air initially at a pressure p and temperature T
would acquire if it were moved adiabatically to the Here, t is time, V is the horizontal velocity, V is the
reference pressure po. It is defined by the following horizontal gradient evaluated at constant log-pres-
relation, which can be obtained from the first law of sure, @ is the geopotential, w D z / D t , is the vertical
thermodynamics: velocity, f = 2 a sin 4 is the Coriolis parameter
(where R = 7.292 x lop5sP1 is the angular velocity
131 of rotation of the Earth and 4 is latitude), k is the
vertical unit vector, Fr is the drag due to small-scale
where cp (= 1004 J kg-' K-l) is the specific heat turbulent fluctuations, p o ,os exp ( - z / H ) is the basic
capacity of dry air at constant pressure. Normally, state density (wherep s is density at the surface, His the
surfaces of constant potential temperature in the scale height defined below eqn [2], J is the diabatic
atmosphere are quasi-horizontal with potential tem- heating rate, and N is the buoyancy frequency whose
perature increasing with altitude. Air parcels displaced square is defined in terms of the height dependence of
vertically conserving potential temperature are then potential temperature as N~ = g(a In e/&).
colder and denser than their surroundings for an In eqns [4] and [ 6 ] , D / D t is the rate of change
upward displacement, and vice versa for a downward following the horizontal motion of a fluid parcel. This
displacement. The atmosphere is then said to be can be related to the rate of change at a given point in
statically stable. When diabatic heat sources (such as space by the expression:
latent heating and radiation) are neglected, 8 remains
constant in time for each air parcel; thus potential E' ($+V.V)
D
temperature is conserved following the motion. Air [71
parcels are then constrained to remain on surfaces of
constant 8, which are referred to as isentropic surfaces. The momentum eqn [4]states that a change in the
In a statically stable atmosphere, potential tempera- horizontal velocity following the motion of an air
ture can be used as the independent vertical coordi- parcel is caused by the net imbalance among three
DYNAMICMETEOROLOGY I Overview 675
forces: the Coriolis force caused by the rotation of the

Earth, the horizontal pressure gradient force (given by P /
the gradient of geopotential on a constant pressure
surface) and the force due to turbulent dissipation
(important mainly near the surface of the Earth). The
equation of mass continuity [5] states that the diver-
gence of the horizontal velocity, which tends to
increase or decrease the horizontal cross-section of a
fluid parcel, must be balanced by vertical motion,
which increases or decreases the depth of the parcel.
The thermodynamic energy eqn [6] states that the rate
of change of temperature following the horizontal
motion is due to the sum of the adiabatic expansion or
compression due to vertical motion, plus the net
diabatic heating by sources such as latent heat release Figure 1 Horizontal plan view of balance of forces for the
and solar or thermal radiative heating. geostrophic wind, V,. P designatesthe pressure gradientforce, Co
designates the Coriolis forces, and @ and @ - 6@ are contours of
constant geopotential on an isobaric surface.
Balanced Flow
When diabatic heating and turbulent dissipation are Equation [8] may be combined with eqn [l]to give
sufficiently small, eqns [4]-[6], together with the the thermal wind relation:
differential form of the hyposometric eqn [l] and
suitable initial and boundary conditions, form a closed
set for prediction of the meteorological fields V, w , a,
and T. Solutions of this complicated set of nonlinear
~ [ v ~- vg(pl)i
( P ~=)R
1: ( ~ x v Tin) P~ 191
partial differential equations can usually only be This relation, which is a consequence of geostrophic
obtained by numerical methods. There are, however, and hydrostatic balance, states that the vector differ-
certain approximate solutions that provide useful ence in the geostrophic wind velocity between two
information on the relations among these field varia- pressure surfaces is proportional to the horizontal
bles for large-scale atmospheric flows in the extra- gradient of the mean temperature in the layer between
tropical regions. the two surfaces. Since large-scale extratropical mo-
When turbulent dissipation is small, which is tions are in hydrostatic balance and are nearly
generally true above the lowest kilometer of the geostrophic, eqn [9] shows that the wind and temper-
atmosphere, large-scale extratropical motions are ature fields are closely coupled. Because temperature
approximately in geostrophic balance, that is, the in the extratropical lower atmosphere generally de-
horizontal pressure gradient force and the Coriolis creases with latitude, the thermal wind relation
force are nearly equal and opposite. The wind velocity indicates that the eastward-directed geostrophic
for which this balance is exact is referred to as the wind increases with altitude and that the strongest
geostrophic wind, V,. It is derived from eqn [4] with upper-level winds will occur where that latitudinal
the acceleration and turbulent dissipation terms temperature gradient is the strongest.
neglected:
Planetary Boundary Layer
In the lowest kilometer of the atmosphere, momentum
transfer by small-scale turbulent eddies becomes an
This states that the geostrophic wind is parallel to lines important component in the momentum balance so
of constant geopotential (or constant height) on a that the geostrophic approximation is no longer valid.
pressure surface, with speed proportional to the The structure of this boundary layer depends strongly
magnitude of the geopotential gradient on the pressure on the vertical stratification in the layer. In regions of
surface. As illustrated in Figure 1, the Coriolis force strong vertical stability (where potential temperature
acts perpendicular to the wind direction (to the right of increases rapidly with height) turbulence is generally
the wind in the Northern Hemisphere), while the weak, except in the lowest few meters above the
pressure gradient force is directed opposite to the surface where strong vertical shear of the wind
Coriolis force. Thus, the geostrophic circulation about provides a mechanical source for turbulent eddies.
the center of a low-pressure system in the Northern Under conditions of strong surface heating, on the
Hemisphere will be a counterclockwise circulation. other hand, the boundary layer may be convectively
676 DYNAMIC METEOROLOGY / Overview
f
changed in proportion to the small difference between
P-26P the Coriolis force and the pressure gradient force,
there is a dynamical quantity that is conserved
following the motion for adiabatic frictionless flows.
This quantity, which is referred to as potential vorti-
P-6P
city, is a fluid dynamical analogue of spin angular
Fr momentum in solid body dynamics. In its simplest
bCo P
form, potential vorticity relates the vertical compo-
nent of vorticity (or spin) of a fluid column confined
between two potential temperature surfaces to the
Figure2 Horizontal plan view of balance of forces in awell-mixed depth of the column (Figure 3).
planetary boundary layer. V is velocity, P designates the pressure Potential vorticity can be expressed mathematically
gradient force, Co designates the Coriolis force, Fr designates the
turbulent drag force, p is pressure, and 6p is a constant pressure
as
interval. (Adapted from Holton JR (1992) lntroduction to Dynamic
Meteorology. New York: Academic Press.) 1 ae
P=(C+f)--
Po az
unstable (potential temperature decreasing with Here, 5 = k . (VxV) is the vertical component of the
height) and strong turbulent eddies may extend relative vorticity due to the horizontal winds and f is
throughout the lowest kilometer of the atmosphere. again the Coriolis parameter, which is the vertical
Over vast areas of the Earth’s surface, however, the component of the vorticity owing to the rotation of the
boundary layer is often near neutral static stability Earth. The sum of the local vertical components of the
(potential temperature constant with height). Al- relative vorticity and the Earth’s vorticity, 5 + f , is
though accurate representation of the force exerted referred to as the absolute vorticity. Equation [lo]
by turbulent eddies in eqn [4]is a challenging fluid states that for a column of fluid confined between two
dynamical problem, it is useful as a first approxima- potential temperature surfaces separated by a fixed
tion to simply assume that turbulent eddies exert a increment of potential temperature 68, the ratio of
drag on the winds so that Fr = -kV, where k is a rate absolute vorticity to the depth of the column 6 z
coefficient typically taken to be about 10 - s - ’. remains constant. Thus, for example, a column
The force balance in the boundary layer is then a moving from a region of high static stability (small
three-way balance among the Coriolis force, the 6z) to a region of low static stability (large 6z) will
pressure gradient force, and the turbulent drag force. stretch vertically, shrink horizontally and spin faster.
This balance is illustrated schematically in Figure 2 . Because potential vorticity is conserved following the
Since the Coriolis force always acts perpendicularly to motion for adiabatic frictionless flow, the evolution of
the wind and the turbulent drag acts in the opposite the field of potential vorticity on isentropic surfaces
direction to the wind, a force balance can be achieved can easily be predicted. The potential vorticity distri-
only if the wind has a component directed across the bution, together with suitable boundary conditions,
isobars towards lower pressure. The component can in turn be used to deduce the three-dimensional
produces net boundary layer inflow into surface low- distribution of wind and temperature.
pressure systems, which by transporting mass to-
wards the pressure minimum acts to ‘spin down’ the
circulation.
Vorticity and Potential Vorticity
Because extratropical flow above the boundary layer is

approximately in geostrophic balance, and thus tends
1f L-’
to be parallel to height contours on pressure surfaces,

the flow is characterized by cyclonic and anticyclonic -6
gyres associated with height (or pressure) minima and
maxima, respectively. Thus, the large-scale flow is
dominated by a rotational flow component, which is
conveniently described in terms of the vorticity
(defined as the curl of the velocity vector). Although
momentum is not conserved for such flows, but is
DYNAMIC METEOROLOGYI Overview 677
Zonally Symmetric Circulations Buoyancy Waves

In the troposphere, temperature normally decreases The vertical stratification of the atmosphere causes a
from Equator to Pole owing to the latitudinal gradient fluid parcel that is displaced vertically to experience a
in solar heating. This heating gradient induces an restoring force due to its buoyancy; the resulting
ageostrophic overturning circulation of air parcels in coherent oscillations in the pressure, temperature, and
the meridional (height-latitude) plane. The overturn- wind fields is called a buoyancy wave, or gravity wave.
ing consists of rising motion in association with The structure of an eastward-propagating buoyancy
convective disturbances in the tropics, poleward drift wave, excited from below, is shown in Figure 4.Phase
in the upper troposphere, slow sinking in the extra- lines, defining the maxima in the perturbation pressure
tropics and an equatorward drift in the planetary and temperature fields, tilt towards the east with
boundary layer. This zonally symmetric parcel circu- height, and propagate eastwards and downwards in
lation is important for transport of water vapor, time. However, the flux of energy (measured by the
momentum, and heat across latitude circles, but correlation between the pressure and vertical velocity
represents only a small deviation from the generally perturbations) is directed upwards, parallel to the lines
west-to-east directed (zonal) winds. of constant phase. There is also an upward flux of
Because the Equator-to-Pole temperature decrease eastward momentum in this case since positive vertical
does not occur uniformly but tends to be concentrated perturbations are in phase with eastward horizontal
in the subtropics, the zonal winds are also concentrat- velocity perturbations, and vice versa. Thus, buoyancy
ed in the jet stream, a narrow band of strong westerly waves provide an important mechanism for vertical
winds in the upper troposphere that encircles the Earth transport of momentum in the atmosphere.
at an average latitude of about 30". The association of The frequency, w , of a buoyancy wave is related to
the jet stream with a strong meridional temperature the angle, E , of the phase lines to the local vertical by
gradient is a consequence of the thermal wind the formula w = N cos a, where N is again the
relationship (eqn [9]), which states that eastward- buoyancy frequency. Thus, high-frequency buoyancy
directed winds must increase rapidly with height waves have energy propagation that is closer to the
where the temperature decreases rapidly in latitude. vertical than do low-frequency buoyancy waves. An
The concentration of the meridional temperature important special case of buoyancy waves are lee
gradient (and potential vorticity gradient) in the waves formed when air in a stable atmosphere is
subtropical jet stream cannot be understood by forced to ascend over a mountain barrier. Oscillations
considering the zonally symmetric circulation forced that are induced by the forced uplift may continue
by solar heating, but rather represents a complex several wavelengths downstream of the mountain.
interaction between the symmetric circulation and the
disturbances that characterize weather and climate.
Thus, the study of atmospheric wave disturbances is
one of the primary areas of dynamic meteorology.
Atmospheric Waves
Waves in the atmosphere are motions that can transmit
energy and momentum without material transport of
air parcels. Most weather disturbances are associated
with one or more types of atmospheric wave. Atmos-
pheric waves result from a balance between the inertia
of the atmosphere and a restoring force. In acoustic
waves, for example, oscillations in the pressure gra-
dient force are balanced by parcel accelerations along
the direction of phase propagation. Such waves are
longitudinal waves in the sense that the fluid parcel
oscillations are parallel to the direction of propaga-
tion. Most meteorologically important waves, how-
Figure 4 Cross-section showing phases of the pressure, tern-
ever, are transverse waves in which the parcel perature, and velocity perturbations for an internal gravity wave.
oscillations are perpendicular to the direction of phase Thin arrows indicate the perturbation velocity field, blunt solid
propagation. Examples of such waves are buoyancy arrows the phase velocity. Shading shows region of upward
waves, inertia waves, and Rossby waves. motion.
678 DYNAMIC METEOROLOGY I Overview
When sufficient moisture is present, condensation may

occur in the updraft portions of the waves, leading to
regularly spaced bands of cloud in the lee of the
mountain. In this case the waves are stationary with
respect to the ground but propagate relative to an
observer moving with the mean wind.
Rossby Waves
The most important class of large-scale atmospheric
waves are called planetary waves, or Rossby waves.
These waves are characterized by oscillations in the
rotational part of the horizontal wind that are parallel
to the horizontal gradient in the potential vorticity.
The simplest example of a Rossby wave occurs in a
barotropic atmosphere. A barotropic atmosphere is Figure 5 Perturbation vorticity field (+ and - ) and induced
one in which potential temperature is constant on each velocity field (dashed arrows) for a meridionally displaced tube of
fluid parcels, showing the mechanism for Rossby wave propaga-
pressure surface so that a6/az in eqn [lo] is independ- tion. Heavy wavy line shows initial perturbation of the tube, light
ent of horizontal position. Potential vorticity conser- wavy line shows westward displacement of the wave perturbation
vation then reduces to conservation of absolute due to advection by the induced velocity field. (From Holton JR
vorticity following the motion: (1992) lnfroducfion to Dynamic Meteorology.New York: Academic
Press.)
Here U is the mean eastward wind, L, is the zonal

The mechanism of Rossby wave propagation can be wavelength and L, is the meridional wavelength.
understood by considering a tube of fluid parcels that Thus, Rossby waves propagate westwards relative to
at time to is motionless and lies parallel to a latitude the mean wind with phase speeds that increase rapidly
circle. In that case [ ( x , t o ) = 0. If the tube is given a as the wavelength increases. For waves corresponding
small sinusoidal meridional displacement ~ ( xt o, ) at to typical extratropical storms with wavelengths of a
the initial time, then from eqn [ l l ] at a later time tl the few thousand kilometers, the Rossby wave phase
relative vorticity will be given by: speed is typically less than 10 m s - which is less than
the mean wind speed. Thus, relative to the ground,
such disturbances tend to move eastwards, but more
slowly than the average eastward wind speed.
where p = df/dy, the rate of increase of the Coriolis Equation [13] also shows that Rossby waves that
parameter with latitude. Since p is positive, it is evident
are stationary relative to the ground (c, = 0) can exist
from eqn [ 121 that the relative vorticity perturbation
only when the mean zonal flow is positive (i.e.,
will be positive for a southward displacement and
eastwards). Stationary Rossby waves are excited in
negative for a northward displacement. Since positive
midlatitudes when westerly winds encounter large-
relative vorticity corresponds to a counterclockwise scale mountain barriers and are displaced meridio-
rotation and negative relative vorticity corresponds to nally during ascent over the barrier. This process is a
a clockwise rotation, the induced perturbation veloc- primary reason why large-scale stationary cyclonic
ities will tend to produce meridional displacement of disturbances are commonly observed to the lee of the
the tube of fluid parcels to the west of the original Rocky and the Himalayan mountain ranges.
displacement, as illustrated in Figure 5 . By this
mechanism, the originally wavy displacement propa-
gates westwards, perpendicular to the north-south Baroclinic Waves
displacement of the fluid parcels.
Barotropic Rossby waves are generated by conversion
Quantitative analysis of the relationship of the wave
of the kinetic energy of the zonal flow into kinetic
frequency to the horizontal scale reveals that Rossby
energy of the waves. They do not involve any
waves are highly dispersive (i.e., their phase speeds are
conversion of potential energy to kinetic energy.
strongly dependent on the wavelength). In Cartesian
Such energy conversion can occur only in the presence
coordinates the zonal (eastward) phase speed relative
of baroclinicity, that is, variations of potential tem-
to the ground has the form:
perature on isobaric surfaces. Baroclinic energy con-
version is responsible for the growth and maintenance
c, = U - p(L2 + L 3 / ( 4 ? 2 ) of most large-scale weather disturbances.
DYNAMIC METEOROLOGY I Overview 679
When the latitudinal gradient of the zonal wind in from the Equator the wind and pressure distributions
the jet stream is sufficiently strong that the meridional in the Rossby-gravity wave are nearly geostrophic, but
gradient of potential vorticity on a constant potential near the Equator there are strong departures from
temperature surface is locally reversed, or when there geostrophic balance. Rossby-gravity waves propagate
is a nonvanishing gradient of potential temperature at westwards relative to the mean flow. They are asso-
the surface of the Earth, the equations of motion ciated with equatorial weather disturbances, and are
linearized about a zonally symmetric basic state have also a significant source of westward momentum for
solutions in the form of exponentially growing wave the equatorial stratosphere.
disturbances. These baroclinically unstable waves
Mesoscale Disturbances
have growth rates, structure, and scales typical of
those observed in developing extratropical cyclones. If an air parcel is saturated, upward displacement will
They are quasi-geostrophic in the sense that the cause water vapor to condense and release its latent
geostropically balanced rotational component of the heat of condensation; potential temperature is then no
wind field strongly dominates over the ageostrophic longer conserved, but increases following the parcel
divergent component. The latter is, however, crucial in motion. If this increase is greater than the potential
the energy cycle by which the waves convert potential temperature gradient of the background atmosphere,
energy associated with the Pole-to-Equator tempera- the atmosphere is said to be conditionally unstable.
ture gradient into disturbance kinetic energy. That is, it is stable with respect to unsaturated parcel
Baroclinic instability provides a mode whereby displacement but unstable with respect to saturated
infinitesimal disturbances may be amplified into parcel displacements. The convective storms associat-
large-amplitude storms. In many situations, however, ed with cumulonimbus clouds can occur only when the
it appears that weather disturbances may develop atmosphere is conditionally unstable, sufficient mois-
rapidly from preexisting upper-level potential vorti- ture is present, and sufficient lifting occurs to bring air
city anomalies in the absence of baroclinic instability.parcels to saturation. The processes that organize
As in baroclinic instability, the growth of storms from convection into mesoscale convective systems are not
upper-level potential vorticity anomalies is associated completely understood. Mesoscale convective systems
with conversion of potential energy to kinetic energy take a variety of forms. Among these are hurricanes,
in association with the ageostrophic secondary squall lines, and supercell thunderstorms. In all cases
flow induced by adjustments towards thermal wind the release of latent heat by convective clouds is the
balance. primary energy source, but the character of the large-
scale environmental flow is generally important for
EquatorialWaves
determining the mode of organization for mesoscale
In the equatorial region there is a special class of systems.
weather-producing waves that combine some of the
characteristics of gravity waves and of Rossby waves.
Equatorial waves are trapped in latitude, that is, they See also
propagate along the Equator with amplitudes decreas- Baroclinic Instability. Coriolis Force. Cyclogenesis.
ing with latitude. In some circumstances they may also Dynamic Meteorology: Waves. Middle Atmosphere:
propagate energy and momentum vertically. Two Gravity Waves. Rossby Waves. Stationary Waves
important examples of equatorial waves are the (Orographic and Thermally Forced). Vorticity. Wave
equatorial Kelvin wave and the Rossby-gravity Mean-Flow Interaction.
wave, The Kelvin wave has pressure and zonal velocity
perturbations symmetric about the Equator, and Further Reading
negligible meridional velocity component. It propa-
gates eastwards, with vertical structure identical to Cushman-Roisin B (1994) Introduction to Geophysical
the eastward-propagating buoyancy wave shown in Fluid Dynamics. London: Prentice-Hall.
Figure 4, and is an important source of eastward Gill AE (1982) Atmosphere-Ocean Dynamics. New York:
Academic Press.
momentum for the equatorial stratosphere. Holton JR (1992) Introduction to Dynamic Meteorology.
Rossby-gravity waves are waves that combine New York: Academic Press.
characteristics of Rossby waves and gravity waves. Pedlosky J (1987) Geophysical Fluid Dynamics. Berlin:
They have meridional wind distributions symmetric Springer-Verlag.
about the Equator and zonal wind and pressure Salby ML (1996) Fundamentals of Atmospheric Physics.
distributions antisymmetric about the Equator. Away New York: Academic Press.
680 DYNAMIC METEOROLOGY/ Balanced Flow
Balanced Flow
M E Mclntyre,Universityof Cambridge, Cambridge, UK balanced, regardless of the wave types involved. The
Copyright 2003 Professor M E Mclntyre
theory of wave-mean interaction is fundamental in
turn to understanding the ‘gyroscopic pumping’ that
drives global-scale stratospheric circulations and
The concept of ‘balanced’ flow is the counterpart, in chemical transports. Indeed, the concept of balanced
atmosphere-ocean dynamics, of the well-known flow enters, implicitly or explicitly, into almost any
concept of ‘nearly incompressible’ or ‘effectively discussion of meteorologically interesting fluid phe-
incompressible’ flow in classical aerodynamics. In nomena, all the way from regional pollutant transport
aerodynamics, a key aspect of such flow - long to planetary-scale teleconnections mediated by
recognized as central to understanding the behavior Rossby wave propagation.
of subsonic aircraft - is that all the significant Balanced flow has analogues in simple mechanical
dynamical information is contained in the three- systems such as the ‘springy pendulum’ composed of a
dimensional vorticity field. This means that the flow massive bob suspended from a pivot by a stiff elastic
has, in effect, fewer degrees of freedom than a fully spring. Such a pendulum has slow, swinging modes of
general flow. More precisely, it means that freely oscillation in which the relatively fast, compressional
propagating sound waves contribute only negligibly to modes of the bob and spring are hardly excited: they
the motion. contribute negligibly to the motion if the spring is stiff
In atmosphere-ocean dynamics there is a corre- enough. The slow, swinging modes correspond to
sponding statement with ‘vorticity’ replaced by balanced flow, and the fast, compressional modes to
‘potential vorticity’, understood in a suitably general sound and inertia-gravity waves. One may describe
sense. The statement applies to a vast set of cases of the swinging modes to a crude first approximation by
rotating, stably stratified fluid flow, for parameter setting the length of the spring equal to a constant - a
values typical of the atmosphere and oceans. It ‘rigid-pendulum approximation’. There is a hierarchy
provides an important key to understanding many of of more accurate approximations that allow the
these cases. If the flow can be considered balanced, spring to change its length in a quasi-static way.
then all the significant dynamical information is In a finite amplitude, two-dimensional swinging
contained in the potential vorticity field, in the oscillation, the spring is longest when the bob is
generalized sense. One may ‘invert’ the potential lowest and shortest when the bob is highest. Such
vorticity field at each instant to obtain the mass and approximations and their ultimate limitations can be
velocity fields. (For a more precise statement, see studied mathematically via techniques ranging all the
Dynamic Meteorology: Potential Vorticity). Again way from two-timing formalisms (method of multiple
this means that the flow has, in effect, fewer degrees of scales) to bounded derivative theory and KAM
freedom than a fully general flow. More precisely, (Kolmogorov-Arnol’d-Moser) theory and other
balance and invertibility mean that not only sound dynamical systems techniques; there is an enormous
waves but also freely propagating inertia-gravity literature.
waves contribute only negligibly to the motion. Thus A quasi-static description may approximate
balanced flows can be much simpler to understand the pendulum motion with remarkable accuracy; the
than fully general flows. error may become exponentially small, or even zero in
Cases of fluid flow describable as balanced come some cases, as the fast-slow time scale separation
under headings such as ‘Rossby waves’, ‘Rossby wave increases. The key point, though, is that in the
breaking’, ‘vortex dynamics’, ‘vortical modes’, ‘vor- quasi-static description the length of the spring evolves
tical flow’, ‘vortex coherence’, ‘blocking’, ‘eddy as if it were functionally related to the elevation of the
transport barriers’, ‘cyclogenesis’, ‘baroclinic and bob. This can be exploited to simplify both the
barotropic instability’, and other shear instabilities, mathematical description of the motion and our
all of which are related to the fundamental Rossby conceptual understanding of it. The functional rela-
wave restoring mechanism or ‘quasi-elasticity’ that tion holds at each instant t, i.e., it holds diagnostically.
exists whenever there are isentropic gradients of More precisely, no derivatives or integrals with respect
potential vorticity in the interior of the flow domain, to t are involved, and values of t do not explicitly
or gradients of potential temperature at an upper or enter into the definition of the functional relation. The
lower boundary. The concept of balanced flow also property of being diagnostic, in this sense, is a
enters into the theory of wave-mean interaction, in crucial part of the mathematical and conceptual
which the mean flow is often considered to be simplification.
DYNAMICMETEOROLOGY/ Balanced Flow 681
In atmosphere-ocean dynamics the defining prop- Here f is the Coriolis parameter, CD(x,t) is the
erty of balance is that an analogous functional relation geopotential height (approximatelygeometric altitude
holds - diagnostic in precisely the same sense. A flow is times gravitational acceleration), and three-dimen-
said to be balanced if the three-dimensional velocity sional position x is specified using pressure altitude.
field u(x,t ) is functionally related to the mass field or Thus the horizontal spatial derivatives a/ax and a / a y
mass configuration, i.e., to the spatial distribution of are taken at constant pressure altitude rather than at
mass throughout the fluid system, presumed to be constant geometric altitude. This qualifies as a balance
hydrostatically related to the pressure field. (Knowl- relation because of the presumption that the hydro-
edge of the mass field then implies knowledge of the static relation also holds, as normally assumed when
temperature and potential temperature fields, hence using pressure as the vertical coordinate. Knowing CD
quantities such as, for instance, the available potential on each constant pressure (isobaric) surface is then
energy and the mass under each isentropic surface.) equivalent to knowing the mass field. So [l] is, as
Such a functional relation between the velocity and required, a diagnostic functional relation between the
mass fields is called a ‘balance condition’ or ‘balance velocity field and the mass field. The vertical derivative
relation’. It provides just enough information to make of [l]is the so-called ‘thermal wind equation’.
the potential vorticity field invertible. The property of The horizontal coordinates x , y are local Cartesian
being diagnostic means that if one knows the mass coordinates in a tangent-plane representation. If we
field at some instant t , but knows nothing about its also take f = constant, giving us the so-called ‘f-plane
time dependence, nor the value of t itself, then the approximation’, then [I] asserts not only that u is
balance relation must nevertheless allow one to slaved to the mass field, but also that it is two-
deduce the complete three-dimensional velocity field dimensionally incompressible or nondivergent, with
u. It must allow the velocity field to be deduced from streamfunction Y = @ / f as expressed by eqn [2].
the mass field and from the mass field alone.
To the extent that a balance relation holds, it
excludes sound waves and inertia-gravity waves from
u(x, t ) = (ay
--
ay ay
-
ax
0)
the repertoire of possible fluid motions. The system
then has too few degrees of freedom to describe such The geostrophic relation [l]-or relations, plural, if
waves. This generalizes the familiar statement in one prefers to think in components rather than vectors
aerodynamics that an incompressibility condition - can be motivated as an approximation to the
excludes sound waves. The reduction in degrees of horizontal momentum equation. The validity of that
freedom is sometimes expressed by saying that some approximation depends on smallness of the Rossby
degrees of freedom are ‘slaved’ to others, or that the number, or, more precisely, on being able to neglect
evolving states of the dynamical system confine relative particle (Lagrangian) accelerations against
themselves to a ‘slow manifold’ in phase space, having Coriolis accelerations, equivalently relative particle
lower dimensionality than the full phase space in accelerations against f times the right-hand side of eqn
which it is embedded. One might say for instance that [l]. The Rossby number, measuring the advective
the velocity field is ‘slaved’ to the mass field. A more contribution to the relative particle acceleration
careful statement would be that in balanced flows the against the Coriolis acceleration, is usually of the
mass and velocity fields evolve as if they were slaved to same order as f - l times a typical magnitude of the
each other, to some useful approximation at least. This relative vertical vorticity aulax - a t q a y = v’Y if eqn
is like saying that the two-dimensional swinging [2] holds. Here u and v are the horizontal velocity
motion of the pendulum evolves as if the length of components corresponding to x and y, and V2 is the
the spring and the elevation of the bob were slaved to horizontal Laplacian.
each other, to some approximation, even though there The geostrophic relation [l] was historically of
is no actual mechanical linkage between the two great importance in early attempts to understand the
variables. dynamics of synoptic-scale weather systems. The
A standard example of a balance condition or history is sometimes discussed under headings such
balance relation is the so-called geostrophic relation, as ‘Buys Ballot’s law’, ‘cyclonic development theory’,
given by eqn [l],which is simple to write and, for and ‘quasi-geostrophic evolution’. Buys Ballot’s law is
typical extratropical parameter values, qualitatively a surface observer’s counterpart of eqn [l] and was
useful though quantitatively not very accurate. discovered empirically through early work with
weather maps.
The modern concept of balance recognizes that, like
aqx, t) a q x , t)
u(x, t ) = -
f (
-~
ay
~
ax > O ) [I1 the rigid-pendulum approximation, eqn [ 11 is merely

the lowest in a hierarchy of more accurate balance
682 DYNAMIC METEOROLOGY/ Balanced Flow
relations. The next member is the relation studied by a physical reality (though not in a way that is
B. Bolin and J. G. Charney in the 1 9 . 5 0in
~ ~connection quantitatively precise), namely the fact that balance
with efforts to develop practical numerical weather is liable to break down through ‘inertial’ and ‘sym-
prediction. The Bolin-Charney balance relation metric’ instabilities near the equator, where f changes
retains [2], even iff varies with latitude, and redefines sign.
Y to satisfy eqn [3], where V is horizontally two- Balance relations still more accurate than [4] can be
dimensional. defined if one is prepared to deal with more
complicated sets of equations. The next relation in
v . ( f V Y ) = VI@ + v . (u.Vu) 131 the hierarchy - to be referred to here as the ‘generalized
Bolin-Charney balance relation’ - is the first in the
Equation [3] is an approximation to the divergence
hierarchy to yield a nonvanishing vertical component
equation, the latter being the result of taking the
of u.It was implicit in the pioneering work of Charney
horizontal divergence of the horizontal momentum
published in 1962, in a famous paper entitled
equation. The relative particle acceleration is now
‘Integration of the primitive and balance equations’.
retained. Its advective part gives rise to the last term of
It starts with [2] and [4] but then adds to the resulting u
[3], while the remaining, a/at part is annihilated when
field a horizontally irrotational correction field gov-
the divergence is taken, because of [2]. It is only
erned by another nonlinear elliptic boundary value
because there are no a/at terms that the relation [3],
problem in the flow domain, a generalization of the
with [2], qualifies as a balance relation.
‘omega equation’ previously developed by N. A.
Again because of [2], the right-hand side of [3] can
Phillips and others. The corrected u field is an
be rewritten in terms of a Jacobian in u and o, as
asymptotically consistent improvement on [ 11, for
V2@- 2a(u, o)/a(x, y), or equivalently a Hessian in
small Rossby number, in the sense that it is one order
y (eqn ~41). more accurate in powers of the Rossby number. The
boundary value problem is derived by taking a / a t of
[4], then eliminating all the resulting time derivatives
using the exact mass conservation and vorticity
equations and the inverse Laplacian of the vorticity
Regarded as an equation for Y when the mass field @ is equation. The vorticity equation expresses V2(aY/at)
given, eqn [4] is not trivial to solve, because of the in terms of diagnostically known, or knowable,
nonlinear terms on the right. Iterative methods need to quantities such as the corrected u field; so the inverse
be used. The problem of finding Y may even become Laplacian is needed in order to eliminate N J / a t from
ill-posed for certain mass fields @, adumbrating, for a p t of 141.
one thing, the fact that there exist mass fields that are This process of eliminating all the time derivatives
not even approximately balanceable by any velocity has to be possible, in principle at least, if the end result
field. A simple thought experiment to make this last is to be a balance relation, which by definition may not
point clear would be to pile up the whole of the Earth’s contain any time derivatives. When the elimination is
atmosphere into a narrow cylinder above the North carried out explicitly, a rather complicated set of
Pole, leaving a vacuum elsewhere. It is obvious that no integro-differential equations results, containing
velocity field u can be in balance with such a mass field. Green’s function integrals whose details depend on
Regardless of the choice of u, the free evolution at the geometry of the flow domain. It may therefore be
subsequent times, in any such thought experiment, computationally more convenient to work with a set
would involve sound and inertia-gravity waves of of equations from which W / a t has not been
enormous amplitude, That is, it would involve gross eliminated, but has been allowed to remain as an
imbalance. unknown that can, in principle, be eliminated. Then
Balance relations are useful in practice only because ‘aY/at’,in quotes, so to speak, must be regarded not as
naturally occurring mass fields, or at least smoothed the actual rate of change of Y but, rather, as an
versions of them are, by contrast, often balanceable to auxiliary variable - better described as a diagnostic
good approximation, as Buys Ballot’s law reminds us. estimate of the rate of change, which must be expected
In most such cases, eqn [4]with suitable boundary to differ, in general, from the actual rate of change of
conditions is a well-posed nonlinear elliptic boundary Y. To avoid confusion over this point a special
value problem in the flow domain, the primary notation is sometimes used, such as Y1 for a diagnostic
exception being flows near the Equator, where Rossby estimate of aY/at, Y2 for a2Y/at2,and so on.
numbers are not small and [4] may fail to be elliptic, as The general form of the functional dependence
can be verified from the theory of Monge-Amp2re defining a balance relation, assuming a balanceable
equations. Again, the failure of ellipticity adumbrates mass field @(x, t ) ,is given by eqn [5], where it is again
DYNAMIC METEOROLOGY I Balanced Flow 683
emphasized that no derivatives or integrals with involves great mathematical subtlety, and full under-
respect to t may appear: it must be possible, in standing has yet to be achieved. Some insight has come
principle at least, to eliminate them all to produce a from studies of a related phenomenon in classical
relation of the form [5], aerodynamics, the weakness of aerodynamic sound
generation or ‘Lighthill radiation’. Recent work at the
u(x, t ) = uB[x; @(., t ) ] PI cutting edge of this problem can be found in papers by
0. Bokhove, 0. Buhler, D. G. Dritschel, R. Ford, J. C.
Time t enters solely via the second argument a(.,t ) of McWilliams, A. R. Mohebalhojeh, S. Saujani, T. G.
uB. The notation a(.,t ) follows mathematical con- Shepherd, J. Vanneste, D. Wirosoetisno, I. Yavneh,
vention and signifies nonlocal spatial dependence. In and others, appearing in the literature from about
other words, the second argument of uB is the whole 2000 onwards.
function, @ of x, over the whole flow domain at the Among other things this recent work has provided a
given instant t - not merely the value of @ at the single clear answer, in the negative, to a question posed in
value of x to which the left-hand side of [SIrefers. Such 1980 by E. N. Lorenz: Could there be an exact balance
nonlocal functions are sometimes called ‘functionals’. relation? Could there be unsteady stratified, rotating
Even the geostrophic relation [1] is enough to illustrate flows that evolve in such a way that freely propagating
the point, though it involves nothing more than inertia-gravity waves are completely absent? More
the behavior of @ in the immediate neighborhood of precisely, is there a slow manifold within the full phase
x - more precisely, it involves enough about that space that is indeed an invariant manifold of the full
behavior to permit the evaluation of the two hor- dynamics?
izontal derivatives. The Bolin-Charney balance rela- The answer in the negative has sometimes been
tions, generalized or not, are fully nonlocal, as is plain viewed with surprise, perhaps because KAM theory
from the occurrence of elliptic partial differential has shown that there are springy pendulum examples,
operators like V2 and, implicitly or explicitly, the and similar examples from other low-order dynamical
associated Green’s function integrals. To find u from @ systems, in which the corresponding question has a
or vice versa, one has to solve elliptic partial positive answer as emphasized in work by 0.
differential equations in the flow domain, as already Bokhove, T. G. Shepherd and others. In dynamical
emphasized, implying for instance that the value of u systems language, there are swinging modes that
at some position x will depend on values @(x’,t ) at confine themselves to invariant manifolds in the
other positions x’ well outside the neighborhood of x. form of ‘intact KAM tori’. In such cases, the swinging
The generalized Bolin-Charney balance relation is motion of the pendulum evolves as if the length of the
often accurate enough for practical purposes, such as spring and the elevation of the bob were exactly slaved
observational data analysis and assimilation, and the to each other.
initialization of the full dynamics for numerical But the negative answer, for atmosphere-ocean
weather prediction. Of fundamental interest, how- dynamics, is now very clear from various lines of
ever, from a theoretical viewpoint, is the fact that the argument beginning with pioneering work of R. M.
pattern of elimination of time derivatives can be Errico and T. Warn, and strongly confirmed by the
extended even further, resulting in balance relations recent work mentioned above. It is also implicit in the
that are more accurate still. The ideas involved seem to nonlocalness, or action-at-a-distance, expressed by
have been first explored by K. H. Hinkelmann in the [5]. Information (about chaotic vortex motion for
1960s, in connection with the initialization problem, instance) cannot in reality travel infinitely fast. Related
and were later approached from another direction, to this is the fact that Lighthill radiation, though often
based on normal mode expansions, by B. Machen- exceedingly weak (accounting for the astonishing
hauer, E Baer, J. Tribbia, and others. accuracy found by Norton) is almost always nonzero.
The most accurate balance relations can, in some In the atmosphere-ocean context, this says that
circumstances, be far more accurate than values of unsteady vortical flow almost always radiates sound
parameters like the Rossby number might ever and inertia-gravity waves, though often very weakly.
suggest; and this accuracy extends over a far wider This in turn relates to dynamical systems concepts
range of parameter values than could reasonably have such as Poincark’s ‘homoclinic tangle’ and the breakup
been expected a priori - with values numerically of of KAM tori into thin ‘chaotic layers’ or ‘stochastic
order unity, and even greater, in some cases. This layers’. Lighthill’s ideas make it overwhelmingly prob-
astonishing fact - discovered by W. A. Norton in the able, even though not yet proven rigorously, that the so-
late 1980s, through ingenious numerical experiments called ‘slow manifold’ is such a stochastic layer. Though
- cannot be deduced by inspection of the momentum astonishingly thin in places - over a far wider range of
equations or other forms of the equations of motion. It parameter values than could reasonably have been
684 DYNAMIC METEOROLOGYI Balanced Flow
expected a priori - it is not a manifold, which by potential vorticity, and from which the exact potential
definition is infinitesimally thin. Though astonishingly vorticity is evaluated. Paradoxical though it may seem
accurate in some circumstances, the concept of at first, all such highly accurate balanced models have
balance is inherently and fundamentally approximate. one velocity field to advect the mass, and another to
The layer is sometimes referred to, therefore, as the advect the potential vorticity. At the highest accuracies,
‘slow quasimanifold’. the two fields differ by only a tiny amount, but differ
(Arguably, a self-contradictory term like ‘fuzzy they must. Related to this is the fact, already mentioned,
manifold’ is best avoided. By its mathematical defini- that diagnostic estimates such as ‘3’1, Y2, . . . differ
tion a ‘manifold’ is a perfectly sharp, smooth hypersur- from true time derivatives such as a’PY/at, a2Y/at2, . . ..
face and not at all fuzzy. Thus ‘fuzzy manifold’ would In all these respects the Bolin-Charney balanced model
add yet another item to the list of self-contradictory has turned out to be wholly exceptional.
terms like ‘variable solar constant’ and ‘asymmetric Velocity splitting was first noticed for Hamiltonian
symmetric baroclinic instability’ - which of course we balanced models constructed from the full dynamics
inevitably have to live with but, perhaps, need not add by the method of R. Salmon. All such models exhibit
to.) velocity splitting, at all levels of accuracy, though in a
One of the most peculiar manifestations of slow slightly different sense: one velocity field advects mass
quasimanifold fuzziness is the phenomenon some- and potential vorticity but another evaluates potential
times called ‘schizophrenia’ or ‘velocity splitting’. This vorticity. As Salmon first showed in the 1980s, the
is a generic property of the most accurate ‘balanced models can be constructed in a systematic way by
models’. imposing the balance relation [ 5 ]as a constraint on the
Just as the swinging modes of the springy pendulum full dynamics within the Hamiltonian framework.
can be described in a simplified yet remarkably Technically speaking, the crucial step that produces a
accurate manner by imposing a functional relation balanced model while preserving Hamiltonian struc-
between spring length and bob elevation, vortical ture is to restrict the ‘symplectic 2-form’ of the full
flows can be described by simplified ‘balanced dynamics (a mathematical object that can contain
models’, constructed by imposing a balance relation both the Hamiltonian flow in phase space and
from the start. This forces a true slow manifold into variations about it) to the in phase space submanifold
existence. The initialization of such a model requires defined by eqn [SI.
only a single scalar field to be specified, such as the As Salmon pointed out, one of the reasons for using
mass field, or the potential vorticity field in the the Hamiltonian framework is that it provides control
generalized sense. This scalar field is sometimes called over conservation principles. The framework, prop-
the ‘master’ field or ‘master’ variable of the balanced erly applied, guarantees that the balanced model will
model, to which all other dependent variables are fully respect the standard conservation principles for
slaved. The model has only one prognostic equation, mass, momentum, and energy, as well as the material
involving only one true time derivative, the rate of conservation (material invariance) of potential vorti-
change of the master field - as distinct from the city. However, there is a fundamental tension between
diagnostic estimates of time derivatives that may be accuracy and conservation. The most accurate
hidden inside the definition of the balance relation [ 5 ] , balanced models cannot be expected to respect
such as the diagnostic estimates Y1, “2, . . . already conservation, beyond the material invariance of
mentioned. potential vorticity. That is because they are trying to
A famous example of such a model is the ‘Bolin- mimic vortical flows that in reality produce Lighthill
Charney balanced model’ or ‘Bolin-Charney balance radiation, which involves wave-induced local mass
model’, or ‘isentropic-coordinate balance equations’, rearrangement, and wave-induced fluxes of energy
so-called, in which either mass or potential vorticity and momentum, none of which can be exactly
can be taken as the master field. Both are advected by described by the balanced model. It is therefore
the velocity field determined via the generalized Bolin- arguable that the most accurate balanced models
Charney balance relation. Here, as implicitly above, will, by that very fact, not respect the standard
the term ‘potential vorticity’ is to be understood in its conservation principles for mass, energy, and momen-
exact (Rossby-Ertel) sense, and is to be evaluated with tum. One cannot have both accuracy and conserva-
the same velocity field, that given by the generalized tion. Something has to give way.
Bolin-Charney balance relation. Within the Hamiltonian framework, which auto-
Now the term ‘velocity splitting’ refers to the fact, matically preserves the conservation principles, what
only recently noticed, that no balanced model more gives way is the concept of a unique velocity field. Less
accurate than the Bolin-Charney model can have a obviously,the same thing happens with non-Hamiltonian
single velocity field that advects both mass and balanced models of the highest possible accuracy
Next Page
DYNAMIC METEOROLOGY/ PotentialVorticity 685
- essentially because the neglect of Lighthill radiation concept itself. For all the foregoing reasons, some
still implies an imperfect representation of local mass authors are beginning to avoid the term ‘geostrophic
rearrangement. This becomes noticeable, even with a adjustment’, instead using the terms ‘Rossby adjust-
non-Hamiltonian balanced model, as soon as one is ment’ or ‘spontaneous adjustment’ as appropriate.
computing with enough accuracy to see the fuzziness Lighthill radiation may also be referred to, therefore,
of the slow quasimanifold. as the ‘spontaneous-adjustment emission’ of sound
and inertia-gravity waves by unsteady vortical flows.
Note on Terminology
The reader is warned that the term ‘geostrophic
balance’, and its shorthand form, ‘geostrophy’, are See also
sometimes used in the literature to mean balance more Buoyancy and Buoyancy Waves: Theory. Coriolis
accurate than geostrophic, i.e., more accurate than [I]. Force. Dynamic Meteorology: Potential Vorticity;
A common example is the phrase ‘geostrophic adjust- Waves. Hamiltonian Dynamics. Instability: Inertial
ment’, which refers to the mutual adjustment of the Instability; Symmetric Stability. Kelvin-Helmholtz
mass and velocity fields to approach balance or to stay Instability. Quasi-geostrophic Theory. Teleconnec-
close to balance - and ‘balance’ of course, in real fluid tions. Vorticity. Wave Mean-Flow Interaction.
flow, means not [I] but the most accurate possible Weather Prediction: Data Assimilation.
balance of the form [5].The example of a circular vortex
adjusting toward ageostrophic, gradient-wind balance
while radiating inertia-gravity waves is enough to
Further Reading
illustrate the point. Gradient-wind balance is the McIntyre ME and Norton WA (2000) Potential vorticity
particular case of Bolin-Charney balance that applies inversion on a hemisphere. Journal of the Atmospheric
to a steady circular vortex. For the circular vortex it Sciences 57: 1214-1235; Corrigendum 58: 949. (Section
holds exactly when fis constant, and is equivalent to [1] 7 describes the only available investigation of a funda-
mental issue neglected above - how to make [5] Galilean
plus a correction term representing relative centrifugal
invariant.)
force. Thus by implication we have another piece of self- Norbury J and Roulstone I (eds) (2002)Large-Scale Atmo-
contradictory terminology, ‘ageostrophic geostrophic sphere-Ocean Dynamics, vol. II: Geometric Methods
adjustment’, unfortunately well established. artd Models. Cambridge: Cambridge University Press.
It may also be noted that the term ‘adjustment’ is (This book is an up-to-date reference on the mathema-
itself used in two distinct senses that are sometimes tical aspects of balanced models, especially Hamiltonian
confused with each other. The first is ‘Rossby’ or balanced models, including a thorough discussion of the
‘initial condition’ adjustment, the mutual adjustment springy pendulum by P. Lynch.)
of the mass and velocity fields toward balance that Saujani S and Shepherd TG (2002) Comments on ‘Balance
occurs primarily because a system is started in an and the slow quasimanifold: Some explicit results’.
Journal of the Atmospheric Sciences 59: 2874-2877.
unbalanced state, an extreme example being the
(Thisis a key to the recent literature on the accuracy of the
thought experiment described above. The second is
balance concept.)
‘spontaneous’ adjustment, the continual mutual Warn T, Bokhove 0, Shepherd TG and Vallis GK (1995)
adjustment of the mass and velocity fields to stay close Rossby number expansions, slaving principles, and
to balance in unsteady vortical flow, even after initial balance dynamics. Quarterly Journal of the Royal
conditions are forgotten. This second process, a far Meteorological society 121: 723-739. (This focuses on
more subtle one, is the process that produces Lighthill asymptotic expansions as one approach to finding high-
radiation. It sets the ultimate limitations of the balance order versions of [SI.)
Potential Vorticity
M E Mclntyre, Universityof Cambridge, Cambridge, UK flow was demonstrated by Carl-Gustaf Rossby in the
Copyright 2003 Professor M E Mclntyre 1930s. Material invariance means constancy on a fluid
particle. The potential vorticity, a scalar field, will be
denoted here by P and can be defined in several ways,
Material Invariance as shown shortly. For dissipationless flow we have eqn
[l],where D/Dt denotes the material derivative.
The significance of the potential vorticity as an exact
material invariant of dissipationless stratified, rotating DP/Dt = 0 [ll
EL NlNO AND THE SOUTHERN OSCILLATION I Observation 713
Contents
Observation
Theory
asked them about atmospheric pressures in their

Observation colonies.
Blanford’s message reached the South Australian
N Nicholls, Bureau of Meteorology Research Centre, Government Astronomer and Meteorologist, Charles
Melbourne, Australia Todd, who noticed that Australian atmospheric pres-
sures were also high, and that the country had been
experiencing a drought at the same time as India.
When another drought struck Australia in 1888, Todd
Introduction realised that India and Australia often experienced
drought at the same time. This synchronism of
During 1877 and 1878 much of China was struck by drought in the two countries is part of the suite of
famine, due to a severe drought. More than nine long-range connections (teleconnections) between
million people perished. In India, at the same time, climate fluctuations in different parts of the world
more than eight million deaths were attributed to that we now call the Southern Oscillation. For the next
famine also caused by a drought. In many districts, a few decades, several meteorologists around the world
quarter of the population died. Drought in the same were occupied in mapping these teleconnections into a
period also caused crop failures, scarcity of food or coherent pattern. Sir Gilbert Walker was the most
even famine, in north-eastern Brazil, Egypt, Indonesia, prominent among these mappers, and it was he who
Fiji, Australia, and southern Africa. In other parts of named the teleconnection patterns the Southern Os-
the world, including Ceylon, the Pacific coast of South cillation. Walker used these teleconnections to develop
America, and Tahiti, many lives were lost from statistical systems for forecasting climate anomalies in
unusual storms or extended periods of heavy rain many parts of the world.
and flood. The El Niiio Southern Oscillation, a major In the middle of the twentieth century, interest in the
pattern of climate variation, links these climatic Southern Oscillation declined. This was partly be-
extremes in different parts of the world; the first major cause the focus of atmospheric scientists shifted to
El Niiio event for which good records exist was the shorter time scales, as computer models exhibited
1877 event. Subsequent El Niiio events have often their ability to forecast weather. A second reason for
reproduced the pattern of climate extremes and the decline in interest was the absence of any theory
societal impacts observed in the 1877 event. We now explaining the teleconnections or the long time scale of
use the phenomenon to make predictions of seasonal the phenomenon. In the early 1960s, Hendrik Berlage
climate variations in many parts of the world. and Jacob Bjerknes separately demonstrated that the
El Niiio and the Southern Oscillation were related.
The term El Niiio originally (at the end of the
nineteenth century) referred to the annual weak
How Was the El Niiio Southern warm current that runs southward along the coast of
OsciIlation Discovered? Peru and Ecuador at the end of the year. Subsequently,
The famine in India in 1877 led to the first scientific scientists applied the term to denote the occasional
attempts to understand and predict monsoon failures large warmings that occur every few years and result in
and drought, and eventually to the mapping of the El major disruptions to the region. Bjerknes developed a
Niiio Southern Oscillation. Henry Blanford, then the theory for how this essentially tropical phenomenon
head of the India Meteorological Department, noticed could affect climate at higher latitudes. This step,
that atmospheric pressures were higher than usual along with the severe ecological and human conse-
over India during the drought. He advised meteorol- quences of the major El Niiio episodes of 1972
ogists in other parts of the British Empire of this and and 1982, revived scientific interest in the study of
714 EL NlNO AND THE SOUTHERN OSCILLATION/ Observation
interannual climate variations and their prediction. Indian Ocean and the west Pacific) associated with
The phenomena are now jointly referred to as the cooler than normal ocean temperatures, while the
El Niiio Southern Oscillation, reflecting their close unusually warm waters in the east bring heavy rains
relationship. and floods to the normally arid Pacific coast of South
America. Figures 2 and 3 show the strong warming of
the east equatorial Pacific that took place during the
What Causes the El Niiio Southern 1997/98 El Niiio. Figure 2 shows the sea surface
temperatures during December 1997, at the peak of
Oscillation? the El Niiio. Warming in the east Pacific at that time
The joint name El Niiio Southern Oscillation, is had almost completely removed the east-west tem-
appropriate because ocean-atmosphere interaction is perature gradient. The east equatorial Pacific warming
the cause of the phenomenon. Easterly winds over the of about 5°C is shown in Figure 3, which exhibits the
eastern and central equatorial Pacific cause oceanic anomalies (deviations from climatology).
'upwelling' (cooler subsurface waters being lifted to How does an El Niiio start? A small change in the
the surface) along the Equator. Southerly winds in the usual sea surface temperature pattern can produce a
eastern Pacific also cause upwelling along the South change in the winds along the Equator. In turn, these
American coast. As a result, the Pacific Ocean is wind changes affect the currents that change the
usually cooler in the east than in the west by several pattern of sea surface temperatures even more. This
degrees. process continues, with ocean temperatures affecting
At tropical latitudes, heavy rains accompany warm winds that affect currents that, in turn, affect ocean
oceans, so the warm west Pacific (including Indonesia temperatures. One important change is related to
and New Guinea) is a heavy rainfall region, while the bursts of westerly winds in the western Pacific. These
cooler east Pacific receives little rainfall. Figure 1 can trigger eastward-moving ocean disturbances that
shows the mean sea surface temperatures for Decem- cause the thermocline (the transition layer between
ber. The relative coolness of the east Pacific, compared warm surface water and cooler, lower waters) to
to the west equatorial Pacific is evident. deepen in the east Pacific. This means that it is harder
This is the 'average' situation, but during an El Niiio for the upwelling to cool the surface (because the
the ocean temperature gradient from one side of the upwelled water is now coming from the upper, warmer
Pacific to the other weakens, and the easterly winds layer), so the east Pacific warms. Eventually, in the
weaken. Droughts occur in the west (around the biggest El Niiio events, the difference in temperature
0" c 4" c 8" C 12°C 16" C 20" c 24" C 28" c 32" C
Figure 1 Climatologicalsea surface temperature for December. (Analysisfrom Bureau of Meteorology, Australia.)
EL NINO AND THE SOUTHERN OSCILLATION / Observation 715
Figure 2 Sea surface temperature for 15-21 December 1997. (Analysis from Bureau of Meteorology, Australia.)
between the west and east equatorial Pacific Ocean circulation across the Pacific and Indian Oceans, and
can disappear altogether. As a result of these major the surrounding continents, is disrupted, with
changes in sea surface temperature and the surface droughts in normally wet areas and heavy rains over
winds, the whole pattern of climate and atmospheric normally arid regions.
Figure 3 Sea surface temperature anomalies (deviations from climatology) for 15-21 December 1997. (Analysis from Bureau of
Meteorology, Australia.)
716 EL NlNO AND THE SOUTHERN OSCILLATION / Observation
The changes associated with the El Nifio often Oscillation are strongly persistent and thus predict-
persist for about a year and then usually collapse quite able. This persistence is greater during the second half
quickly. Sometimes a mirror-image pattern of climate of the calendar year, because El Nifio episodes tend to
disturbances, with flooding in Australia, India, Indo- start around March-May and finish around the same
nesia, northeast Brazil, and dry conditions on the time a year later. Thus, if an event is under way by mid-
Pacific coast of South America, follows. This set of year it is likely to persist through the second half of the
conditions is called La Nifia. La Niiia episodes also year. This means that climate anomalies usually
usually last about a year or so. associated with the presence of an El Nifio at this
As alluded to earlier, the atmospheric variations time can often be predicted well in advance. The
associated with El Nifio and La Niiia events are called tendency for El Niiio events to start around March-
the Southern Oscillation. This name derives from the May is illustrated in Figure 5, which shows east
observation (dating back to the time of Blanford and equatorial Pacific sea surface temperature anomalies
Todd) that, during an El Nifio, atmospheric pressures during the major events of the second half of the
are usually higher than normal over Australian and the twentieth century. In each of the five events, sea surface
Indian Ocean and lower than normal in the southeast temperature anomalies in the east equatorial Pacific
Pacific. During the opposite phase, the La Nifia, the were relatively low at the start of the year, and then
pressure anomalies are reversed. So, in a sense, the increased rapidly from about March, reaching a peak
atmosphere acts like a seesaw, with high or low near the end of the calendar year. The temperature
pressures on either side of the Pacific. anomalies subsequently weakened over the next few
We can monitor this seesaw in atmospheric pressure months.
with the Southern Oscillation Index or SOI. This is the
standardized difference in pressure between Tahiti and What Areas Does the El N i i o Southern
Darwin. When the SO1 is negative, pressures are high
over the Australian region and relatively low in the
Oscillation Affect?
southeast Pacific. This is an indication that the Trade The pattern of climate anomalies seen in the 1877 El
Winds are weak across the Pacific, and these weaker Niiio tends to be repeated each time an El Nifio occurs.
winds result in warm east equatorial Pacific sea surface The typical pattern of rainfall anomalies associated
temperatures - an El Niiio. Figure 4 shows time-series with an El Nifio is shown in Figure 6. The figure
of the Darwin mean sea level pressure and sea surface indicates, for each area consistently affected by the El
temperatures in the 'cold tongue' of the east equatorial Nifio, the months in which the anomalies are most
Pacific (180"-90" W, 6" N-6" S). The close relation- consistent. The pattern of precipitation anomalies
ship between the atmospheric pressure on one side of associated with the other extreme of the El Nifio
the Pacific and sea surface temperatures on the other Southern Oscillation, the La Nifia, is essentially the
side is clear, as is the tendency for El Nifio and La Niiia opposite of that depicted in Figure 6 (i.e., where drier
events to last about 12 months. than normal conditions are usually experienced during
This tendency to last about 12 months means that an El Nifio, then wetter than normal conditions can be
the climate effects related to the El Nifio Southern anticipated during a La Niiia episode).
1950 1960 1970 1980 1990

Year
Figure 4 Time series of annual mean Darwin mean sea level pressure (MSLP) and sea surface temperature in the area 180"-90°W,
6" N-6" S ('Cold Tongue Index', CTI). (Darwin data from Bureau of Meteorology, Australia. CTI data from Todd Mitchell, JISAO, University
of Washington.)
EL NINO AND THE SOUTHERN OSCILLATION/ Observation 717
2-
a
v
1-
F
0 0-
-1 - - 1972
Year=-1 Year = 0 Year=+l - 1982
-2 I I I I I I I , I I I , I , , I I , I , , , 1 , , , , , , , 1 , , 1 , ~
Jan. Apr. Jul. Oct. Jan. Apr. Jul. Oct. Jan. Apr. Jul. Oct.
Month
Figure5 Time-series of monthly CTI during the major El Niiio events of the second half of the twentieth century (1957,1965,1972,1982,
1997). (CTI data from Todd Mitchell, JISAO, University of Washington.)
The El NiAo Southern Oscillation also affects the other hand, tropical cyclones are more frequent
temperatures in some parts of the world. Thus in than usual in the east Pacific during these episodes.
December-February at the peak of an El Niiio,
temperatures are usually above average throughout
central and southern Africa, southern Asia, and the
Prehistoric Behavior of the El Niiio
western Pacific, Canada, and the Pacific coasts of
North and South America. The south-east of the Southern Oscillation
United States tends to be cooler than average. Severe Instrumental records relevant to the study of the El
frosts can occur in places where drought accompanies NiAo Southern Oscillation are available back into the
an El Nifio episode, such as the highlands of Papua late nineteenth century. The study of El Niiio episodes
New Guinea and inland eastern Australia. prior to this depends on documentary records, and
The El NiAo Southern Oscillation also affects paleoclimatic (proxy) records. Documentary evidence
tropical cyclones and some other weather and climate of heavy rains and floods on the Pacific coast of South
extremes. Figure 7 is a time-series of the SO1 and of the America (always associated with El Nifio episodes
number of tropical cyclones around Australia. When during the instrumental era) are available from the
an El NiAo is under way (Le., when the SO1 is strongly sixteenth century. Comparisons of the dates of heavy
negative), fewer than normal tropical cyclones are rains and floods in South America with dates of
observed around Australia. Similarly, Atlantic hurri- droughts in other parts of the world have confirmed
cane activity is reduced during El NiAo episodes. On that the El Niiio Southern Oscillation has been
Figure 6 Schematic of areas with aconsistent precipitation signal associated with El NiAo events. For each region the months are shown
during which it is consistently wetter or drier than normal. In each region the list of months begins in the initial year of the El NiAo (year = 0).
(Reprinted with permission from Cambridge University Press of Trenberth KE (1991), in Glantz eta/.(1991).)
718 EL NINO AND THE SOUTHERN OSCILLATION/ Observation
Australian region (105-165E)

25 24
-2d50 '' I ' 1960

I '
I I 1 I I I ' '
1970 I ' '
Year
I I '
' II
1980 I '' I '' '
1990
1 I ' 1 I '' ' 0
Figure 7 Time-seriesof the SouthernOscillation Index (Sol)( 0) and the number of tropicalcyclones in the Australianregion (0"-15" S,
105"-165" E) (0).(Data from the Bureau of Meteorology,Melbourne, Australia.)
operating for at least hundreds of years. The telecon- observations originally intended for other purposes
nections between droughts and floods in these various have been the main source of information. Atmos-
parts of the world have been similar throughout these pheric pressure, rainfall, and temperature observa-
five centuries, reflecting the effects of the El Niiio tions originally taken for the purposes of weather
Southern Oscillation throughout this period. recording and forecasting, or to determine the 'aver-
Paleoclimatic data, from corals, ice cores in glaciers, age' climate, have been used in studies of how the
tree rings, and marine and lacustrine sediments also phenomenon affects climate variations around the
provide information regarding the occurrence of El globe. Sea surface temperatures recorded by merchant
Niiio episodes prior to the instrumental period. This and other ships have been the main source of
evidence, although not conclusive, suggests that information about the ocean variations associated
El Niiio episodes have been occurring for at least with the El Niiio. In recent decades, however, new and
several thousand years. improved observations, specifically designed for cli-
mate studies, have been initiated. These include
The El Nifio Southern Oscillation satellite observations of rainfall and sea surface
temperature and sea-level, moored buoys monitoring
in the Recent Past the ocean and atmosphere in critical parts of the ocean,
The prominence of the El Niiio Southern Oscillation and subsurface analyses of the ocean thermal struc-
has varied through the instrumental period. Very ture. The analysis of these new data is in its infancy,
strong El Niiio episodes occurred in the first quarter of but the data have already enhanced our ability to
the twentieth century, with only relatively infrequent, monitor, understand, and predict the El Niiio Southern
and weak, events in the period 1925-1950. After Oscillation.
1950, more intense El Niiio and La Niiia events were
observed. Since the mid-l970s, there appears to have
been a shift toward more frequent, or stronger, El Nifio
episodes, with La Niiia episodes becoming relatively See also
infrequent. Some analyses suggest that this behavior is El NiAo and the Southern Oscillation: Theory.
very unusual, given the (admittedly short) historical Monsoon: ENSO-Monsoon Interactions. Walker Circu-
record. lation.
Future Observations of the El Niio Further Reading

Southern Oscillation
Allan R, Lindesay J and Parker D (1996)El Nifio Southern
For most of the period during which the El Niiio Oscillation and Climatic Variability. Collingwood:
Southern Oscillation has been monitored and studied, CSIRO Publishing.
EL NlNO AND THE SOUTHERN OSCILLATION/Theory 719
Diaz HF and Markgraf V (edsj (1992) El Nifio. Historical Glantz MH, Katz RW and Nicholls N (eds) (1991)Telecon-
and Paleoclimatic Aspects of the Southern Oscillation. nections Linking Worldwide Climate Anomalies. Scien-
Cambridge: Cambridge University Press. tific Basis and Societal Impact. Cambridge: Cambridge
Glantz MH (1996)Currents of Change. El Nifio’s Impact on University Press.
Climate and Society. Cambridge: Cambridge University Philander SGH (1990)El Nifio, La Nifia, and the Southern
Press. Oscillation. New York: Academic Press.
P Chang, Texas A&M University, College Station, TX, and how various feedback loops between the atmos-
USA phere and ocean operate and interplay. This under-
S E Zebiak, International Research Institutefor Climate standing provides the theoretical basis for the
Prediction, Palisades, NY, USA development of ENSO prediction systems, which
are critical for operational seasonal-to-interannual
climate forecasting.
Introduction
The Southern Oscillation and Walker
The El Niiio Southern Oscillation is a spectacular,
planetary-scale climate phenomenon that is inherently
Circulation
caused by interactions between the atmosphere and From an atmospheric perspective, the Southern Os-
the ocean. Historically, El Niiio refers to unusually cillation can be viewed as a perturbation about a
warm ocean temperatures that occur every 2-7 years thermally driven east-to-west circulation of the trop-
around Christmas time along Peruvian coast, ical atmosphere across the Pacific Ocean. This circu-
extending into equatorial eastern and central Pacific lation, known as the Walker circulation, is caused by
Ocean. The Southern Oscillation, named by its the sharp contrast in sea surface temperature across
discoverer - Sir Gilbert Walker - on the other hand, the tropical Pacific Ocean. The western tropical Pacific
refers to a ‘seesaw’ of the atmospheric pressure contains the warmest regions of the world’s ocean,
between the Pacific and Indian Oceans. It was not known as the Western Pacific Warm Pool, where the
until the seminal work of Jacob Bjerknes in the sea surface temperature is above 28°C. In contrast, the
late 1960s that scientists realized that these two eastern equatorial Pacific features relatively cold
phenomena are intimately linked. The acronym ocean surface waters, extending from South America
ENSO (El Niiio Southern Oscillation) has now coasts westward along the Equator. This is known as
been widely used to describe this fascinating inter- the Eastern Equatorial Pacific Cold Tongue, where
annual climate fluctuation, emphasizing the inherent the sea surface temperature is 5-10°C colder than the
ocean-atmosphere coupling. surface water of the warm pool. The warm water in the
Although the origins of ENSO lie in the tropical western Pacific creates low surface pressure, which
Pacific, the impact of ENSO is global, owing to causes moisture-laden air to converge into the region.
planetary waves of the atmosphere that redistribute The air rises and the moisture condenses in strong
vorticity from tropics to extratropics. The ‘telecon- convective events, resulting in widespread cloudiness
nection’ of ENSO can disrupt weather patterns around and heavy precipitation. The rising air descends from
the globe. For this reason, ENSO has been recognized the upper troposphere to the surface in the Eastern
as the most important climate phenomenon at inter- Equatorial Pacific Cold Tongue as dry air. Cool
annual time scales. temperatures result in relatively high surface pressure,
Theoretical understanding of the development and divergent flow, and little rainfall. These motions -
evolution of ENSO, and of underlying dynamical rising in the west, sinking in the east - are connected
mechanisms for its irregular oscillation at interannual through easterly trade winds near the surface and a
time scales, goes beyond the boundary of traditional westerly wind aloft, forming the Walker Circulation.
dynamical meteorology and oceanography, because it Fluctuations in the position and intensity of the
requires knowledge about how the tropical atmos- Walker Circulation cause the Southern Oscillation.
phere responds to sea surface temperature changes, When sea surface temperature in the eastern Pacific
how the equatorial ocean adjusts to changes in winds, is warmer than normal, such as during El Nifio
720 EL NlNO AND THE SOUTHERN OSCILLATION/Theory
years, the low atmospheric pressure center normally must move poleward to generate cyclonic vorticity
situated in the Western Pacific Warm Pool moves in order to balance the ambient vorticity change.
eastward, bringing along with it the rising moist The diabatic heating also excites planetary waves
air and heavy precipitation. As a result, the east- which induce remote response outside the region of
west pressure difference across the Pacific is heating. Of particular interest is a Kelvin wave
reduced and the easterly trade winds are weak- propagating eastward at a speed c. It gives an easterly
ened. This produces a weak Southern Oscillation wind symmetric about the Equator which decays
( a negative phase). By the same token, the strength at a rate of r/c per unit distance. This easterly wind
of the Southern Oscillation is enhanced (a positive is reminiscent of the trade winds along the equa-
phase) when sea surface temperature in the eastern torial Pacific as a part of the Walker circulation.
Pacific falls below normal. This sensitivity of the To the west of the forcing region, there are Rossby
tropical atmosphere circulation to sea surface temper- waves which cause the cyclonic flow to expand
ature fluctuations is one of the key elements of westward. The simple atmospheric model gives an
E N S 0 physics. analytical expression linking changes in the equatorial
A secondary process that contributes to equatorial trade winds to changes in sea surface temperature
trade wind fluctuations involves the so-called Hadley through its effect on the diabatic heating of the
circulation, a meridional overturning cell spanning the atmosphere.
tropical Pacific (and the global tropics). This circula-
tion consists of a rising branch, concentrated in a
narrow zone generally north of the Equator, known as
the ITCZ (or intertropical convergence zone), and
Equatorial Ocean Adjustment
sinking motion, with increasingly strong surface From an oceanic perspective, changes in sea surface
easterlies both north and south of the ITCZ. During temperature associated with El Niiio can be under-
El Nifio, as the equatorial surface temperature warms, stood in terms of an equatorial ocean response to
the ITCZ and attendant Hadley circulation shift changes in the Trade Winds. Under normal conditions,
equatorward, leading to a reduction of equatorial an easterly wind stress supplied by the Trade Winds in
easterly winds, beyond that associated with the the central and eastern Pacific acts on the ocean
Walker circulation. surface. This stress is balanced by friction and the
Many of the essential features of the Walker Coriolis force, resulting in poleward surface flow in
circulation (and Hadley circulation) can be captured either hemisphere and thus upwelling in the eastern
by a simple physical model in which the tropical equatorial Pacific and South American coast. The
atmosphere is assumed to be forced by a diabatic westward winds also ‘push’ the relatively warm
heating source and subject to simple dissipation of surface water westward, bringing cold subsurface
momentum and heat (with common decay rate r of water to the ocean surface and lifting the so-called
order of 1-2 days). The diabatic heating is largely thermocline of the ocean in the east. The thermocline
induced by latent heat released by the rising moist air represents a band of water within which temperature
over the warm ocean, and this can be approximated as changes rapidly with depth, thereby separating the
a function of sea surface temperature. Assuming that warm upper ocean from the cold deep ocean. The
the vertical structure of the diabatic heating is fixed Trade Winds thus cause the thermocline to shoal from
and has a simple structure with a single maximum at west to east across the equatorial Pacific Ocean,
midlevels decreasing to near zero at the surface and maintaining the warm pool in the west and the cold
upper levels (which approximates to the heating field tongue in the east. During a negative phase of the
produced by cumulus convection in the tropics), then Southern Oscillation, the easterly Trade Winds are
the entire circulation pattern in the vertical projects weaker than normal, which reduces the upwelling in
primarily onto the so-called first baroclinic mode of the east, deepens the thermocline and causes the sea
the tropical atmosphere, and the horizontal motions at surface temperature to rise. An El Niiio is produced!
each level obey a two-dimensional set of equations (the The opposite occurs during a positive phase of the
so-called shallow water equations). Therefore, the Southern Oscillation, when the Trade Winds are
forced solution of the shallow water equation can give strengthened.
the first-order approximation of the Walker Circula- The close relationship among Trade Winds, therm-
tion. In the region of heating, for example, in the ocline, and sea surface temperature results from the
Western Pacific Warm Pool, the diabatic heating gives rapid adjustment of the equatorial ocean. Because the
rise to a vertical velocity which causes the lower Coriolis force vanishes at the Equator, there is a wave
atmosphere to expand and the vortex to stretch. To guide along the Equator, where a variety of waves are
conserve potential vorticity, the surface air parcels trapped to within a few degrees to either side of it in the
ocean (and a few tens of degrees in the atmosphere). the oceanic ‘memory’ - another key element of ENSO
Two types of waves, the equatorial Kelvin and Rossby physics.
waves, are of particular importance. Kelvin waves are
special gravity waves that propagate eastward with a
speed of approximately 2-3rns-l, and can travel Coupled Dynamics
across the Pacific Ocean in 2 months or so. Rossby If the atmosphere and ocean were decoupled, then
waves are planetary vorticity waves that propagate small perturbations in either the sea surface temper-
westward at a rate of about 0.6-0.8msP1, and can ature or the winds would fade quickly away because of
travel across the Pacific in 6-7 months. Both these dissipation in the oceans and atmosphere. In reality,
waves propagate at a rate that is faster by an order of the tropical atmosphere and equatorial ocean in the
magnitude or more than the planetary waves in Pacific Ocean are tightly coupled, because of the
extratropical oceans. For this reason, equatorial sensitivity of the atmospheric response to sea surface
oceans adjust much more rapidly than extratropical temperature changes and the rapid adjustment of the
oceans in response to changes in the wind stress equatorial ocean to changes in winds. Therefore, a
forcing. Therefore, at interannual time scales, the modest change in either the equatorial sea surface
zonal gradient of the equatorial thermocline and temperature or the trade winds can trigger a chain
overlying Trade Wind stress are approximately at reaction in the coupled ocean-atmosphere system,
balance. involving a positive feedback between the atmosphere
The response of the equatorial ocean to changes in and the ocean.
the zonal wind stress can again be modeled in terms of
a shallow water model. Here, the ocean is approxi-
mated as a two-layer fluid system with a thin, warm Bjerknes Feedback Mechanism
layer on top of a deep, cold layer. The interface
between the two layers represents the ocean thermo- The key ingredients of this positive feedback were first
cline which has an average depth of approximately pointed out by Bjerknes. Consequently, the mecha-
150 m in the equatorial Pacific. Motions of upper layer nism has become known as the Bjerknes hypothesis.
and the interface obey shallow water equation subject Imaging that there is initially a weak westerly wind
to wind stress forcing. The steady-state solution is very anomaly along the Equator that causes the trade winds
nearly a balance between the zonal wind stress and to weaken. From the equatorial ocean adjustment
zonal gradient of thermocline. When a change in the discussed above, we expect that a weak warm sea
trade wind occurs, the ocean is subject to an anom- surface temperature anomaly will occur in the eastern
alous zonal wind stress forcing. A typical wind stress equatorial Pacific, owing to deepening in the thermo-
anomaly associated with the Southern Oscillation has cline depth and weakening in equatorial upwelling.
a spatial structure that has the largest amplitude in the Because the tropical atmosphere is sensitive to changes
western central equatorial Pacific and decays away in sea surface temperature, the small increase in sea
from the Equator. Because of the nonuniform spatial surface temperature in the east will tend to move
structure, off-equatorial wind stress curl is generated. atmospheric convection eastward, reduce the diabatic
The oceanic response to such an anomalous wind heating in the west and weaken the Walker circulation.
stress forcing in the shallow water system takes This causes a further weakening of the trade winds,
place in two stages: first, an equatorial Kelvin which in turn leads to a further warming in the eastern
wave is excited by the strong wind stress anomaly equatorial Pacific, and so on. Key elements of this
at the Equator, propagating eastward and causing feedback loop are illustrated in Figure 1.The Bjerknes
changes in depth of the thermocline and sea surface hypothesis marks the beginning of the formation of
temperature in the eastern equatorial Pacific upon its modern ENSO theory.
arrival. At the time when the Kelvin wave is excited,
Rossby waves of opposite sign to the Kelvin wave are
also generated by the off-equatorial wind stress curl
Coupled Modes
associated with the wind stress anomaly. These Rossby A quantitative understanding of the coupled dynamics
waves propagate westward and reflect at the western can be gained by formulating a simple coupled ocean-
boundary as a second Kelvin wave - now with the atmosphere model, such as the two shallow water
opposite sign to the first Kelvin wave - and thus work systems described above. A stability analysis can then
against the effect brought by the first Kelvin wave in be applied to the coupled system to obtain a set of
the eastern equatorial Pacific with a time delay. It is modes which describe how the coupling between the
this delayed response in the eastern equatorial Pacific ocean and atmosphere can modify free oceanic and
produced by the same wind stress forcing that provides atmospheric wave modes in a wide range of dynamical
722 EL NlNO AND THE SOUTHERN OSCILLATION /Theory
unstable depending on the strength of the feedback.

These modes can be characterized best by a delayed
action oscillator - a prototype model for ENSO. Let T
denote sea surface temperature anomaly in the eastern
equatorial Pacific; then the time evolution of T obeys
a differential-delay equation dT/dt = cT(t) -
bT(t - z), where t is time and z is a time delay
associated with the adjustment of the equatorial
ocean. The behavior of T in this model is determined
by two competing processes: (1) cT represents the
Bjerknes positive feedback, contributing to a growth
of sea surface temperature; (2)bT(t - z) represents the
delayed oceanic adjustment discussed in the previous
section, constituting a negative feedback. This nega-
tive feedback due to the oceanic ‘memory’ effect
counteracts the positive feedback in the eastern
Less enirainment aaker equatorial Pacific with the time delay 5. Physically,
cooling in Trade Winds this is achieved through excitation of the Rossby
eastern basin
\ Weakenb;;fwelling waves by off-equatorial wind stress curl and the
deepens thermocline
reflection at the western boundary. Therefore, the time
delay z is essentially determined by the propagation of
Figure 1 The Bjerknes hypothesis describes a positive feedback the Rossby waves from the western central Pacific to
between the ocean and atmosphere in the equatorial Pacific: a
the western boundary and of the reflected Kelvin wave
weakening in the Trade Winds causes a warming in the eastern
Pacific through a deepening in the thermocline and a weakening in from the western boundary to the eastern equatorial
the upwelling, which in turn leads to a further weakening in the Pacific. This negative feedback process is illustrated in
Trade Winds. (Reproduced with permission from Chang P and Figure 2 .
Battisti DS (1998) The physics of El Nifio. Physics World 11: Because both of the positive and negative feedbacks
41-47.) are comparable in strength, the differential-delay
equation can support oscillations with periods longer
than the wave adjustment time scale z. In fact, when
parameter space. Such an analysis leads to the dis- realistic parameters are used, the differential-delay
covery of a new breed of ‘wave modes’, with charac- equation produces an oscillation with a period of 3-4
teristics that depend on the strength of the air-sea years, which agrees with the average occurrence of El
feedback and the adjustment time scale of the ocean. Nifio in reality. Therefore, the coupled modes of
For a weak air-sea feedback, the adjustment time of delayed-oscillator type are believed to capture the
the ocean is much faster than that imposed by the weak essential physics of ENSO.
air-sea feedback. Therefore, the free waves are not
significantly influenced by the air-sea feedback and
the coupled modes behavior similar to those of free
Nonlinear vs. Stochastic ENSO Theory
waves. For a strong air-sea feedback, the adjustment The delayed oscillator theory offers an explanation for
time of the ocean is much slower than that imposed by the oscillatory behavior of ENSO, but cannot explain
the strong air-sea feedback. The coupled modes the irregularity of ENSO. In nature, an ENSO occurs
depend critically on the air-sea feedback, but is less every 2 to 7 years with considerable variations in its
influenced by the oceanic adjustment. In this limit, the strength. Where does this irregularity come from?
coupled modes behave quite differently from the free Although answers to this question are not entirely
ocean waves and tend to be more unstable because of clear, there are two competing theories that offer
the strong influence of the positive air-sea feedback. different views on this issue.
In between these two extreme limits lies the most One theory relies on inherent nonlinear interactions
interesting dynamic regime, where the adjustment within the coupled ocean-atmosphere system. It
time of the ocean and of air-sea feedback are compa- assumes that the Bjerknes positive feedback is strong
rable to each other. This is the dynamic regime, where enough for the coupled system to reside in an unstable
ENSO is believed to reside in nature. The coupled dynamic regime and maintain a self-sustained oscilla-
modes in this regime have a mixed behavior of free tion at an interannual time scale. This self-sustained
oceanic waves and of the mode that depends primarily oscillation interacts nonlinearly with the annual cycle,
on air-sea feedbacks, and can be either stable or which is driven by the seasonal variation of the solar
Figure 2 Oceanic adjustment associated with the delayed oscillator type of coupled ocean-atmosphere mode. Rossby waves
generated by the relaxationof the Trade Winds in the western central equatorial Pacific propagate westward and are reflected into Kelvin
waves at the western boundary near the Indonesianarchipelago. The Kelvinwaves reach the eastern equatorial Pacific with a time delay T
afterthe relaxationoftheTrade Winds and bring cold watertothe region,which altimately shutsdown thewarn ENSO event.Togetherwith
the Bjerknes positive feedback, this negative oceanic feedback process forms the delayed oscillator mechanismfor ENSO. (Reproduced
with permission from Chang P and Battisti DS (1998) The physics of El Niiio. Physics World 11: 41-47.)
radiation. Because the intrinsic mode of the coupled eigenvectors of A give the dominant coupled modes in
system oscillates at a different frequency from the ENSO system, including the delayed oscillator type of
driving frequency, nonlinear interaction between the modes. However, since the system is linear, all eigen-
two cycles can give rise to complicated behavior of the vectors must be stable, i.e., decaying with time
coupled system response. Depending upon the relative (otherwise the variance would not be bounded).
strength of the driving cycle (the annual cycle) and the Therefore, they cannot support a self-sustained oscil-
intrinsic cycle (the ENSO cycle), the response can be lation and the variability of the system must be
either locked into a periodic cycle (with a period equal maintained by the noise forcing. One important
to a rational number between the period of annual distinction between this view of ENSO physics and
cycle and that of intrinsic coupled mode; this phe- the nonlinear theory is that the evolution of ENSO is
nomenon is known as frequency locking in nonlinear not necessarily dominated by a single mode (the most
dynamics) or chaotic. It is hypothesized that the reality unstable mode, according to the nonlinear theory), but
resides in the chaotic regime, and thus the irregularity rather determined by the interference among many
of ENSO can be attributed partially to the chaos stable modes. Constructive interferences cause sea
generated by the nonlinear interaction between the surface temperature anomalies to grow, whereas
annual cycle and the intrinsic oscillating mode of the destructive interferences cause them to decay. In this
coupled system. theory the irregularity of the ENSO cycle comes
The other, competing theory puts ENSO in a weak naturally because the variability of the system is
feedback regime, so that the coupled system does not maintained by a random forcing.
support a self-sustained oscillation, but is forced Understanding the cause of ENSO irregularity has
externally by ‘weather noise’. Here, ‘weather noise’ important implications for the predictability of this
refers to the high-frequency variability that is not phenomenon. If ENSO evolution is governed by a low-
generated directly by ocean-atmosphere interactions, order chaos, then its predictability limit is determined
but is produced by hydrodynamical instability proc- by the inherent nonlinear dynamics of the coupled
esses of the atmosphere. Although these high-fre- system. On the other hand, if stochastic processes in
quency fluctuations of the atmosphere have coherent the atmosphere is the main cause of ENSO irregularity
spatial structure, in time they can be represented then its predictability depends on not only determin-
approximately as a normally distributed white-noise istic dynamics in the coupled system but also the
process. Under these approximations, ENSO can be nature of noise forcing, which is determined largely by
modeled as a multivariate linear stochastic system, i.e., the internal dynamics of the atmosphere. These
+
(d/dt)s = As Ft, where s is a state vector comprised remain topics of active research.
of sea surface temperature anomalies throughout the
tropical Pacific basin, A a system matrix governing the
deterministic dynamics of ENSO, and F represents the
spatial distribution of the weather noise whose tem- See also
poral fluctuations 5 are represented as a normally Coupled Ocean-Atmosphere Models. El NiFio and
distributed white noise processes. The sea surface the Southern Oscillation: Observation. General Cir-
temperature evolution is then determined by the culation: Models; Overview. Hadley Circulation. Kelvin
properties of the system matrix A and the noise Waves. Ocean Circulation: General Processes.
forcing structure matrix F. In particular, the leading Rossby Waves. Walker Circulation.
724 ELECTRICITY, ATMOSPHERIC / Global Electrical Circuit
Further Reading Neelin JD, Battisti DS, Hirst AC, et al. (1998) ENS0
Theory. Journal of Geophysical Research 103:
Gill AE (1982)Atmosphere-Ocean Dynamics. New York: 14261-14290.
Academic Press. Chang P and Battisti DS (1998) The physics of El Nifio.
Philander SGH (1990)El NiLio, La Nifia and the Southern Physics World 11: 4147.
Oscillation. New York: Academic Press.
Contents
Global Electrical Circuit

Ions in the Atmosphere
Sprites
Historical Development
Global Electrical Circult The development of ideas on the DC global circuit
received great impetus from three giants of research in
E R Williams, Massachusetts Institute of Technology, atmospheric electricity: Benjamin Franklin, William
Cambridge, MA, USA Thomson (Lord Kelvin), and C. T. R. Wilson. Their
Copyright 2003 Eisevier Science Ltd. All Rights Resewed. three contributions, which dominated each of three
successive centuries, are discussed in turn.
Franklin can be credited with the first enunciation of
Introduction
a global flow of moist, electrified air. His concept is
We live in a thin layer of air glued to the Earth’s best discussed around his own picture (Figure l), as
surface by gravity. This gaseous atmosphere is presented to the Royal Academy of Sciences in Paris in
composed largely of neutral molecules of oxygen and 1779. Warm, moist air ascends in the tropics and
nitrogen and as a consequence is an electrical descends in the polar regions. This cloudy air was
insulator. The atmosphere is bathed in radiation - believed to deliver electricity to the cold polar icecap
ultraviolet radiation from the Sun and cosmic on snow, where it would accumulate until breakdown
radiation from deep space. This radiation ionizes of the rarefied upper atmosphere occurred in the form
the atmosphere and makes it a good electrical of the aurora. While this explanation for the aurora is
conductor at upper levels where the radiation is now known to be flawed, the postulated role for the
more energetic. The Earth beneath the atmosphere is tropics and the picture of the large-scale circulation of
abundant in liquid water. All water contains ions in the atmosphere were clearly prescient.
solution, and the ions provide conduction. Three- One hundred years later, Lord Kelvin developed
quarters of the Earth’s surface is covered with potential theory, a mathematical tool needed for
conductive seawater. Liquid water is also present theoretically underpinning the global circuit. Remark-
virtually everywhere on land, permeates the cracks ably, more than 40 years before the conductive
and joints within the Earth’s crust, and makes land- ionosphere was postulated by A. Kennelly and
masses electrical conductors too. The thin layer of 0. Heaviside in 1902, Kelvin advanced the spherical
insulating air sandwiched between these two con- capacitor picture for the global circuit. His expecta-
ductors forms the medium for the global electrical tion for an outer conductor was based on his knowl-
circuit. For the so-called ‘DC’ global circuit, this edge that rarefied air of the upper atmosphere was a
medium is a giant spherical capacitor. For the poor insulator in comparison with air at the Earth’s
‘AC’ global circuit, otherwise called the Schumann surface. He also advocated organized measurements
resonances, the medium is an electromagnetic of the Earth’s electric field, and this suggestion
waveguide. undoubtedly motivated subsequent electrical obser-
Further Reading Neelin JD, Battisti DS, Hirst AC, et al. (1998) ENS0
Theory. Journal of Geophysical Research 103:
Gill AE (1982)Atmosphere-Ocean Dynamics. New York: 14261-14290.
Academic Press. Chang P and Battisti DS (1998) The physics of El Nifio.
Philander SGH (1990)El NiLio, La Nifia and the Southern Physics World 11: 4147.
Oscillation. New York: Academic Press.
Contents
Global Electrical Circuit

Ions in the Atmosphere
Sprites
Historical Development
Global Electrical Circult The development of ideas on the DC global circuit
received great impetus from three giants of research in
E R Williams, Massachusetts Institute of Technology, atmospheric electricity: Benjamin Franklin, William
Cambridge, MA, USA Thomson (Lord Kelvin), and C. T. R. Wilson. Their
Copyright 2003 Eisevier Science Ltd. All Rights Resewed. three contributions, which dominated each of three
successive centuries, are discussed in turn.
Franklin can be credited with the first enunciation of
Introduction
a global flow of moist, electrified air. His concept is
We live in a thin layer of air glued to the Earth’s best discussed around his own picture (Figure l), as
surface by gravity. This gaseous atmosphere is presented to the Royal Academy of Sciences in Paris in
composed largely of neutral molecules of oxygen and 1779. Warm, moist air ascends in the tropics and
nitrogen and as a consequence is an electrical descends in the polar regions. This cloudy air was
insulator. The atmosphere is bathed in radiation - believed to deliver electricity to the cold polar icecap
ultraviolet radiation from the Sun and cosmic on snow, where it would accumulate until breakdown
radiation from deep space. This radiation ionizes of the rarefied upper atmosphere occurred in the form
the atmosphere and makes it a good electrical of the aurora. While this explanation for the aurora is
conductor at upper levels where the radiation is now known to be flawed, the postulated role for the
more energetic. The Earth beneath the atmosphere is tropics and the picture of the large-scale circulation of
abundant in liquid water. All water contains ions in the atmosphere were clearly prescient.
solution, and the ions provide conduction. Three- One hundred years later, Lord Kelvin developed
quarters of the Earth’s surface is covered with potential theory, a mathematical tool needed for
conductive seawater. Liquid water is also present theoretically underpinning the global circuit. Remark-
virtually everywhere on land, permeates the cracks ably, more than 40 years before the conductive
and joints within the Earth’s crust, and makes land- ionosphere was postulated by A. Kennelly and
masses electrical conductors too. The thin layer of 0. Heaviside in 1902, Kelvin advanced the spherical
insulating air sandwiched between these two con- capacitor picture for the global circuit. His expecta-
ductors forms the medium for the global electrical tion for an outer conductor was based on his knowl-
circuit. For the so-called ‘DC’ global circuit, this edge that rarefied air of the upper atmosphere was a
medium is a giant spherical capacitor. For the poor insulator in comparison with air at the Earth’s
‘AC’ global circuit, otherwise called the Schumann surface. He also advocated organized measurements
resonances, the medium is an electromagnetic of the Earth’s electric field, and this suggestion
waveguide. undoubtedly motivated subsequent electrical obser-
ELECTRICITY, ATMOSPHERIC / Global Electrical Circuit 725
current upward to the conductive ionosphere where

it spreads out, to return to Earth in fair-weather
regions, as illustrated in Figure 2. Wilson’s idea led
E J. W. Whipple to compare the universal time (UT)
variation of electric field over the oceans, now referred
to as the ‘Carnegie curve’, with the UT diurnal
variation of thunder areas on a global basis, as shown
in Figure 3. Three major tropical continental zones are
activated sequentially by the surface heating associat-
ed with the passage of the Sun. The similarity in phase
between these two curves has long stood as key
substantiating evidence for the Wilson’s global circuit
hypothesis. Additional support came in 1950 when
0. Gish and G. Wait measured upward currents over
thunderstorms from an airplane. More recent meas-
urements have shown upward currents over shower
clouds.
Wilson’s student T. W. Wormell later extended the
surface measurements of current and formulated a
statistical charge ‘balance sheet’ for the global circuit.
These results showed that point discharge current
dominated over the lightning current in modulating
Figure 1 Benjamin Franklin’s picture of airflow and transport of the negative charge transfer to the Earth’s surface by
electricity from equatorial to polar regions. (Reproduced with electrified storms.
permission from Silverman S (1970) Franklin’stheory of the aurora.
Journal of the Franklin’s lnstitute 290: 177.)
The first coordinated measurements of the global
circuit were made by R. Muhleisen, one set from
Germany and another from a ship in the Atlantic
vations from the research vessels Carnegie and Maude ocean. These balloon measurements integrated the
by the Carnegie Institution, Kelvin undertook his own vertical electric field in the atmosphere to provide the
surface measurements of potential gradient, verified so-called ionospheric potential of the global circuit.
that the Earth carried a negative charge, and conclud- The simultaneous soundings at two locations agreed
ed that the global circuit peaked in winter, a result now to within 5 % in three-quarters of the measurements,
believed to be dominated by local effects. He also providing considerable support for the worldwide
prophesied the use of electrical measurements for nature of the global circuit response. Extensive meas-
purposes of weather prediction: urements of the DC circuit were also carried out by
R. Markson using instrumented aircraft and with
There can be no doubt that electric indications, balloons from stations in Massachusetts and Australia
when sufficiently studied, will be found impor- with similar correlated results, and with diurnal
tant additions to our means for prognosticating variations that closely follow the classical Carnegie
the weather: and the speaker hoped soon to see curve.
the atmospheric electrometer generally adopted
as a useful and convenient weather glass.
Measurements by C. T. R. Wilson of the field

changes associated with lightning in thunderclouds led
him to conclude that the polarity of thunderclouds was
systematically positive in upper levels and negative at
lower levels. Wilson was also engaged with measure-
ments of the currents flowing to Earth during periods
of fair weather. The observation of the transatlantic
propagation of radio waves in 1903 by G. Marconi
verified the presence of the conductive ionosphere.
This collective information led Wilson in 1920 to Figure 2 Simple illustration of the operation of the DC global
formulate his famous hypothesis for the global elec- circuit with electrified clouds as generators and a return current to
trical circuit: thunderstorms are batteries and drive earth in fair weather regions.
726 ELECTRICITY,ATMOSPHERIC/ Global Electrical Circuit
phere is spherically symmetrical. The W radiation

from the Sun is largely removed in causing ionization
at higher altitudes - in the ‘D’ and ‘E’ regions of the
ionosphere - and since this radiation is strongly
anisotropic, the conductivity structure is endowed
0 4 8 12 16 20 24 with a modest asymmetry between the daytime and the
Hour (UT) nighttime hemispheres. As a result, the electromag-
netic waveguide departs slightly from spherical sym-
120
100 ,/‘
..--
+---.
,-The World
,
metry.
In treating the D C global circuit, we are concerned
primarily with the spherically symmetric lowest region
of the atmosphere, where small ions are the dominant
charge carrier. Over the oceans, where the air is
relatively free of aerosol particles, the altitude ( z )
dependence of the electrical conductivity can be
approximated by the exponential function
0 4 8 12 16 20 24
Hour (UT)
Figure 3 Comparisons of the UT variation of electric field over

the ocean (the Carnegie curve) and the UT variation of thunder where = 5x S m-’ , and the e-folding scale
areas worldwide. (Adapted from Whipple JN (1929) On the height zo = 5.0 km. The minimum atmospheric con-
association of the diurnal variation of electric potential gradient in ductivity is found at the Earth’s surface and contrasts
fine weather with the distribution of thunderstorms over the globe. markedly with conductivity values for the oceans and
Quarterly Journal of the Royal Meteorological Society55: 1-1 7.)
typical surface crustal material. This information
establishes quantitatively the medium of the global
circuit - a spherical capacitor with variable resistivity
The first suggestion that the spherical capacitor of
between good conductors.
the DC global circuit also served as an electromagnetic
Current sources for the D C global circuit are in
waveguide appeared when in 1952 W. Schumann
principle any mechanism that separates positive from
postulated the existence of electromagnetic reso-
negative charge vertically between the conductive
nances maintained by global lightning activity. Partial
Earth and the upper atmosphere. Such mechanisms
experimental verification of the resonances was ob-
include the vertical transport of electric space charge
tained by Schumann’s student H. Koenig in Munich in
the late 1950s. M. Balser and C. Wagner of the MIT in the planetary boundary layer, the gravity-driven
Lincoln Laboratory verified the multimodal reso- descent of selectively charged aerosol particles in the
atmosphere, and the separation of electric charge by
nances with the first spectral measurements in 1960.
microphysical processes involving ice in shower
Transient excitations of the Schumann resonances by
clouds and thunderclouds. The C. T. R. Wilson
single extraordinarily energetic flashes were reported
hypothesis postulates the special subset of thunder-
in the early 1970s by D. L. Jones and his colleagues.
storms as the dominant current sources, or batteries,
for the global circuit, though much evidence has
Structure and Operation of the DC appeared that shower clouds without lightning are
also major contributors. The Wilson hypothesis
Global Circuit establishes a firm link between the polarity of charge
The conductivity structure of the giant spherical on the Earth and the polarity of thunderclouds, but
capacitor and electromagnetic waveguide is estab- does not account for the polarity of either entity. Today
lished by a n interplay between ionization from the Sun it seems likely that this polarity is determined by some
and from deep space, and the exponential decline of air fundamental property of ice, still to be identified.
density with altitude above the Earth’s surface. The The behavior of current sources in a medium whose
principal radiation components in this context are the conductivity structure is exponential in the vertical has
galactic cosmic radiation and the ultraviolet (W)and been examined mathematically. The solution for a
X-ray photons from the Sun. The cosmic radiation is point source in an exponential medium infinite in the
the most energetic and serves to dominate the ioniza- vertical is for all current to flow upward toward higher
tion and electrical conductivity at altitudes from 0 to conductivity. When horizontal conductor boundaries
50 km. The cosmic radiation is also largely isotropic and the associated image sources are added, some
and so the conductivity structure of the lower atmos- current flows locally to Earth with the remainder
ELECTRICITY,ATMOSPHERIC / Global Electrical Circuit 727
flowing upward toward larger conductivity. Consist- The steady-state distribution of electric space
ently with these theoretical calculations, observations charge within the resistive lower atmosphere of the
over thunderstorms do show systematic upward spherical capacitor may be determined from Poisson’s
current flow to higher altitudes, in line with the equation,
Wilson hypothesis. In a closed global circuit the
integrated upward current from globally distributed p ( z ) = & o -dz
=-- [41
GOZO
sources, approximately one kiloampere, must ulti-
mately return to Earth, as illustrated in Figure 2. On where EO is the permittivity of free space. According to
account of the spherical uniformity of conductivity these simple electrostatic predictions, positive space
structure, the current flow will also be spherically charge is distributed throughout the atmosphere, with
uniform in the resistive lower atmosphere. According maximum value in the lowest part of the atmosphere,
to Ohm’s Law, this uniform integrated current density, as indicated schematically in Figure 2. An equal and
J o , will set up a vertical electric field given by opposite negative charge resides on the Earth’s surface,
with charge density per unit area given by
giving an electric field at the Earth’s surface of

40 V m - which declines exponentially with increas- and with total negative charge on the Earth
ing altitude. The total voltage drop across the giant
spherical capacitor will then be given as the line
integral of this vertical electric field (where one-
dimensional coordinates are used rather than spherical Note that this charge is large in comparison with the
coordinates, as the vertical extent of the Earth-
ionosphere cavity is small in comparison with the
radius of the Earth, RE):
-
charge transferred by single lightning discharges
( 10 C), which as a consequence have no discernible
effect on the steady-state VI values or the surface
VI = / E ( z ) dz = JOZO
-
00
electric field. The capacitance of the giant spherical
131 capacitor is, by definition,
While the limits of field integration extend from the C=--Qror

171
Earth’s surface to infinity, the 5 km scale height for VI
conductivity and electric field indicate that in three whose value is approaching one farad. The integrated
scale heights the field is reduced to 5% of its surface resistance of the global circuit can also be computed
value and hence more than 95% of the voltage drop is analytically, again ignoring the contribution of the
achieved at 40 km altitude. The integrated potential resistive and highly variable planetary boundary layer,
difference between the Earth and the upper atmos- which is most prevalent over land areas. This global
phere is called ‘ionospheric potential’. resistance is most easily obtained directly from Ohm’s
The numerical evaluation of VI in the simplified law as
treatment of the global circuit includes the experi-
mentally determined current density in fair weather, PI
Jo(2x A m-2), the conductivity scale height zo
(5km), and the (extrapolated) conductivity of air at The time required for the voltage across the giant
the surface (5 x 10- l4 S m - ’). The resulting VI value
capacitor to decay if all sources suddenly ceased is the
from eqn [3] is 200 kV, which is somewhat less than the RC relaxation time:
value of actual measurements (mean value 250 kV).
The primary reason for this discrepancy is the neglect RC = (200 Q)( 1farad) = 200 s 191
of the aerosol-laden planetary boundary layer, the
shallow (-1000 m) zone near the surface in which the Using typical parameters, this gives a relaxation time
conductivity departs from the simple exponential of 200 s, or about 3 min. This time is somewhat longer
form given earlier in eqn [ 11. The local complications if the effect of the resistive planetary boundary layer is
of the planetary boundary layer have been the major included.
obstacle to the measurement of the global circuit from
the Earth’s surface. Measurements at sea, where the Schumann Resonances
boundary layer is substantially less polluted, have
shown records similar to the Carnegie curve on The AC global circuit shares the same Earth-iono-
individual days, but even at sea problems can arise. sphere medium as the DC global circuit but is a
substantially richer and more complex phenomenon. whose fields dominate all other lightning on the planet
Electromagnetic standing waves within the Earth- for periods of several hundred milliseconds. Such
ionosphere cavity, excited by lightning flashes world- maps strongly resemble the maps produced by satellite
wide, are known as Schumann resonances. The observations in Figure 5 in showing strong continental
resonant frequencies are determined by the circumfer- dominance.
ence of the natural waveguide and by the speed of light The majority of lightning flashes are not sufficiently
within the waveguide. The fundamental resonant energetic to stand out above all other events, and the
frequency is 8 Hz. electromagnetic fields of this far larger population of
The effective ionospheric height for the AC global ordinary flashes superimpose to produce the quasi-
circuit is systematically greater than for the DC circuit. steady 'background' resonances. The phase informa-
The propagation of electromagnetic waves requires tion for the integrated activity is lost and the observa-
the displacement current to exceed conduction cur- tions are normally documented as power spectra of
rents; the latter currents damp the waves. A meaning- electric and magnetic fields. Sample spectra, integrated
ful estimate of waveguide height is found by equating for a complete 24-hour period for three field compo-
the displacement current and the conduction current nents (vertical electric, north-south magnetic, and
east-west magnetic), are shown in Figure 4, where
well-defined peaks at 8 Hz for the fundamental mode
are seen as well as higher resonant frequencies at 14,
20,26, and 32 Hz. The widths of the various spectral
peaks are indicative of the quality factor Q of the
WE0 = O(2) [I21
The height dependence of electrical conductivity is
given by [l],and so for a frequency of 8Hz this
condition can be solved for a height which is approx-
imately 50 km. This value is approximately 10 con-
ductivity scale heights and hence is significantly
greater than the height needed to achieve the iono-
spheric potential [3] of the DC global circuit.
The application of Maxwell's equations to a thin
spherical waveguide (assumed uniform for simplicity
of treatment) leads to the normal mode equations for
the electric and magnetic fields resulting from a single
vertical lightning discharge with frequency-dependent
current moment IdS(o), where I is the current in
amperes and dS is the vertical extent of the current-
carrying channel. The role of the ionosphere is treated
HNS
!
with the complex eigenvalue v, which is frequency-
dependent. The parameter a is the radius of the Earth
and h is the height of the waveguide. The P: and P,1 are
complex Legendre polynomials and their derivatives,
respectively. 8 is the distance between the lightning
source and an observer along a great circle path.
I ( o ) dSv(v + 1)P,O(- cos 8 )
N 3.0~10-'~f i/
I
E(w) = II
4Rg&owhsin (nv)
measured in V m-l Hz-' ~ 3 1
I ( @ ) dS P,'( - cos 0)
0.0 I II I I I \II !
H(w)= 0 10 20 30 40 50 60 70
4 R ~ hsin (nv) Frequency (Hz)
measured in A m-' Hz-' ~ 4 1 Figure 4 Schumann resonance power spectra in the vertical
electric (top panel) and two magnetic field components (middle and
These equations have been successfully merged with bottom panel), as measured in West Greenwich, Rhode Island, on
measured electromagnetic fields to construct global 1 Januaty2000. The fundamental mode near 8 Hz is dominant in all
maps of extraordinarily energetic lightning flashes, spectra. Four to five higher order modes are also discernible.
ELECTRICITY, ATMOSPHERIC / Global Electrical Circuit 729
Earth-ionosphere cavity. One definition of Q is simplified earlier development of ideas, a diurnal

variation in integrated current J o will manifest itself as
Q,fo a diurnal variation in VI and in the surface electric field
Af E ( z = 0) on a global basis. As noted earlier, the phase
agreement between the thunder day variations and the
where fo is the center frequency and Af is the frequency Carnegie curve is quite good, but the amplitude
interval between half-power points of the resonance variation of the proxy lightning record is more than
curve. Dimensionless Q values are typically in the twice that of the Carnegie curve. Three major tropical
range of 3-8, indicating that the Schumann resonances storm regions are represented in the thunder day data
are not sharply tuned owing to the lossy nature of the in Figure 3 as separate peaks but such clean distinction
ionosphere. is not apparent in the Carnegie curve. These results
The background Schumann resonances are more strongly suggests that lightning is not the main
difficult to interpret for lightning source properties contributor to the DC global circuit. Consistent with
(location and intensity), because the sources are less this conclusion is the experimental evidence that the
localized than for single events and because lightning local diurnal variation of point discharge current,
near the receiver can dominate the signal and mask rainfall, and rain current are all substantially smaller
more distant activity. Nevertheless, because of the than the local diurnal variation of lightning. Since
overwhelming dominance of lightning in continental point discharge currents, rain currents, and boundary
zones (Figure 5 ) , even the sources for the background layer convective charge transport are present in
resonances exhibit considerable localization. A mod- weather systems other than thunderstorms, it would
ified form of eqns [ 131 and [ 141 has been used to infer appear that a larger collection of weather elements
the global vertical dipole moment change squared per than thunderstorms alone is needed to account for the
unit time based on observations of the electric and modulations of the DC global circuit.
magnetic power spectra. These measurements are Further evidence for diverging behavior between the
broadly consistent with earlier estimates of 100 s - DC and AC global circuits comes from quantitative
for the global lightning flash rate. comparisons of the contributions from each of the
three major tropical zones with the diurnal variation in
Comparative Response of the DC universal time (Figure 3). Africa is frequently the
dominant lightning producer both in the classical
and AC Global Circuit thunder day analysis and in numerous observations of
The first three modes of the electric field of the global lightning from satellites in space (Figure 5 ) . By
circuit are illustrated within the Earth-ionosphere contrast, the Carnegie curve peaks at 19.00 UT when
cavity in Figure 6. Mathematically speaking, the DC the Americas are most active, suggesting that the
global circuit is the zeroth-order mode of Schumann Americas are the dominant tropical contributor to the
resonances, the AC global circuit. On this basis, one DC global circuit.
might expect the ionospheric potential of the DC The annual variation of the global circuit is more
circuit and the global charge moment squared per unit thoroughly documented for the AC component,
time of the AC global circuit to be highly correlated on because of numerous available observations of global
all time scales. While high-quality comparisons of lightning, and shows a maximum in Northern Hemi-
simultaneous behavior are scarce, the available evi- sphere summer. The comparisons also indicate a
dence does not support this prediction. A likely reason similar result to that on the diurnal time scale: the
for different behavior lies with the sources which drive annual variation of lightning (nearly a factor of two) is
the two global circuits, and the different meteorolo- substantially larger than the annual variation of
gical origins of these sources. Table 1summarizes the surface electric field and ionospheric potential.
principal source currents for the DC and AC global
circuit. Charge transfer by point discharge current, by
falling precipitation, and by lightning are all deemed Applications of the Global Circuit
important for the DC global circuit, whereas only A Monitor for Global Change
lightning is believed to contribute substantially to
Schumann resonances. The verification and physical understanding of global
Historically, the behavior of the diurnal variation of change has become a topic of enduring interest
the DC global circuit shown in Figure 3 is viewed as the worldwide. The connection between a currently
cornerstone of atmospheric electricity and a critical increasing COZ concentration and an increasing
test of the C. T. R. Wilson hypothesis that thunder- mean global temperature is controversial and still
storms sustain the source currents. According to the unresolved. The global circuit provides a unique
730 ELECTRICITY, ATMOSPHERIC /Global Electrical Circuit
Figure 5 Global distribution of lightning based on (A) optical observations with the DMSP satellite at local midnight (Orville and
Henderson 1986), and (B) optical observations during both day and night with the Optical Transient Detector. (Courtesy NASA Marshall
Space Flight Center.)
natural framework for monitoring global integrals: on well-recognized local behavior. For example,
the collective current from electrified convection and lightning is more likely in the hot afternoon than
the total lightning activity. The expectation that global during cool nights. At midlatitude locations, lightning
circuit integrals are responsive to temperature is based is more frequent in the hot summer than in the cold
ELECTRICITY,ATMOSPHERIC / Global Electrical Circuit 731
have now been identified on the diurnal, the 5-day,

the 25-day, the 30-60-day Madden-Julian oscillation,
the semiannual, the annual, and the interannual (El
Nifio Southern Oscillation) time scales. The global
circuit sensitivity to surface air temperature on these
various time scales is generally in the range of 10-
100% per "C. For many of these studies, the determi-
nation of a quantitative sensitivity to temperature has
been hindered by the poor sampling of temperature in
regions dominated by deep convective activity (i.e.,
tropical South America and Africa).
The response of the global circuit to a long-term
increase in temperature associated with global warm-
ing remains an open question. Measurements of the
global circuit have been underway for only 50 years
and very sporadically. The Schumann resonance
records are substantially shorter. These records do
not show significant upward trends. Results from a
general circulation model point to an increase in
Figure 6 Illustrations of the behavior of the vertical electric field lightning, but the key quantities for lightning in the
for the first three modes of the Earth-ionosphere cavity. The zeroth models, like cloud buoyancy and convective scale
order mode is the DC global circuit. (Adapted from Wait JR (1996)
Electromagnetic Waves in Stratified Media. Piscataway, NJ: IEEE
updraft strength, are poorly resolved. If CAPE is
Press.) invariant on the long time scale, the only change to be
expected is the increase in moisture with temperature
following the fundamental Clausius-Clapeyron rela-
tion. This dependence is approximately 6% per Kelvin
winter. In the tropics, lightning is more frequent during and is substantially smaller than the global circuit
warm equinox than during cooler solstice. At all sensitivity to temperature change documented on
latitudes, lightning is more common over warmer land shorter time scales. Further work and longer records
than over cooler ocean. In general, in the present are required to discern a meaningful long-term
climate, one finds greater cloud electrification and change.
higher lightning flash rates in regions characterized by The DC global circuit also affords an opportunity to
stronger conditional instability, larger CAPE (convec- investigate changes in boundary layer pollution on a
tive available potential energy), and larger updraft global basis. The strategy would be to distinguish
speeds. In the tropics, such regions are characterized variations in VI caused by source current from
by large wet-bulb potential temperature, a thermody- variations in VI caused by global circuit resistance.
namic quantity representing both temperature and The latter quantity is influenced by the resistance of the
moisture. planetary boundary layer, which is in turn influenced
With the aim of understanding global circuit by the aerosol population that removes small ions and
response to temperature change, a general strategy in reduces the local electrical conductivity. Simultaneous
recent years has been to examine global circuit measurements of the ionospheric potential VI and
behavior on time scales for which the temperature numerous single-station measurements of the air-
variations and the underlying forcings are reasonably earth current density would suffice to compute a
well established. Substantial correlations between global electrical resistance value (recall eqn [SI) whose
global circuit behavior and the underlying global changes could then be monitored. This experiment has
meteorology (which is often temperature-related) not yet been performed, but the requisite observation-
al methods have been established and verified.
Table 1 Dominant current sources for the global circuit

Mesospheric Discharge Phenomena
DC circuit AC circuit
Optical phenomena in the mesosphere high above
Point discharge current Lightning current large thunderstorms (mesoscale convective systems)
Precipitation current have been firmly documented within the last
Lightning current
decade, and named sprites, elves, and trolls. All of
Boundary layer convection
these phenomena appear to be accompanied by a
high-amplitude ‘ringing’ of the Schumann resonances ment, the initiation of dedicated monitoring pro-
by single-energetic cloud-to-ground lightnings. The grams, and a greatly improved understanding of the
application of the normal mode eqns [13] and [14] to global circuit’s relationship with meteorology and
these isolated transients enables the location and ionospheric physics on many time scales. The DC
vertical charge moments to be extracted on the basis global circuit is difficult to measure on a continuous
of ELF (extremely low-frequency) global circuit basis, but possesses a global invariant, ionospheric
measurements. Such observations appear to verify potential, that is well defined and quantitatively
Wilson’s speculations in the 1920s that a sufficiently accessible. The Schumann resonances are relatively
large dipole moment change by lightning in the lower easy to measure on a continuous basis and are
atmosphere can cause a field increase in the upper insensitive to the local variations in the planetary
atmosphere large enough for dielectric breakdown boundary layer, but the global invariant is more
and ensuing optical emission. Global maps of event difficult to extract from single station records. Coor-
locations can be prepared on the basis of ELF dinated measurements with both DC and AC aspects
measurements from a single location. Such maps are are most likely to bear the greatest fruit.
proxy maps for sprites.
Curiously, the great majority of flashes that cause
upper-atmospheric discharges are ground flashes with See also
positive polarity, the opposite of the far more common
negative ground flash that recharges the Earth nega- Chemistry of the Atmosphere: Chemical Kinetics;
tively. Studies of the giant positive discharges with Principlesof Chemical Change. Electricity, Atmospher-
Schumann resonance methods have shown that they ic: Sprites. Middle Atmosphere: Quasi-BiennialOscilla-
tion; Semiannual Oscillation. Tropical Meteorology:
exhibit larger current and transfer greater amounts of
Inter Tropical Convergence Zones (ITCZ).
charge than their negative counterparts. It is not
known whether this asymmetry is related primarily to
local meteorology and differences in the nature of the Further Reading
local charge reservoirs for the two lightning types, or
whether a contribution arises from the fact that Adlerman EJ and Williams ER (1996)Seasonal variations of
the global electrical circuit. Journal of Geophysical
negative flashes are charging the DC global circuit
Research 101: 29679-29688.
and positive flashes are discharging it. Bering EA (1997) The global circuit: global thermometer,
weather by-product, or climate modulator. Rev. Geo-
Diagnostic for the D Region of the Ionosphere phys. Res. Suppl. 845-862, July.
Boccippio D, Williams E, Heckman S, Lyons W, Baker I and
The upper conductive boundaries of both the DC and Boldi R (1995) Sprites, ELF transients and positive
AC global circuit lie in the lowermost regions of the ground strokes. Science 269: 1088-1096.
ionosphere, where the electron density is some four Holzer RE and Saxon DS (1952) Distribution of electrical
orders of magnitude smaller than its daytime maxi- conduction currents in the vicinity of thunderstorms.
mum near 300 km altitude. Even large-aperture radars Journal of Geophysical Research 57: 207-216.
are insufficiently sensitive to detect and monitor the Markson R (1985) Aircraft measurements of the atmos-
lower D region. Measurements in situ by rockets are pheric electrical global circuit during the period 1971-
limited in space and time. Schumann resonance 1984. Journal of Geophysical Research 90: 5967-5977.
Muhleisen R (1977)The global circuit and its parameters. In:
methods may provide a useful global diagnostic for Dolezalek H and Reiter R (eds) Electrical Processes in
ionization in this region. The parameters of greatest Atmospheres, pp. 467-476. Darmstadt: Steinkopf.
interest are the Schumann resonance frequencies, Polk C (1982) Schumann resonances. In: Volland H (ed.)
determined by phase velocities of the resonant waves, Handbook of Atmospherics, pp. 112-178. Boca Raton,
and Q values, both of which depend on conductivity FL: CRC Press.
profiles that change with modulations in ionizing Price C and Rind D (1993) Modeling global lightning
radiation. Recent results have shown systematic distributions in a general circulation model. Monthly
increases in the resonant modal frequencies over the Weather Review 122: 1930-1939.
recent solar cycle that are attributable primarily to Satori G and Zieger B (1996) Spectral characteristics of
ionization increases in the 80-90 km altitude range. Schumann resonances observed in central Europe. Jour-
nal of Geophysical Research 101: 29663-29669.
Wait JR (1996) Electromagnetic Waves in Stratified Media.
Conclusion Piscataway, NJ: IEEE Press.
Whipple F J W (1929) On the association of the diurnal
Renewed interest in the global electrical circuit within variation of electric potential gradient in fine weather
the last decade has spurred new methods for measure- with the distribution of thunderstorms over the globe.
ELECTRICITY, ATMOSPHERIC I Ions in the Atmosphere 733
Quarterly Journal of the Royal Meteorological Society Williams R (1999)Global circuit response to temperature on
5 5 : 1-17. distinct time scales: a status report, In: Hayakawa M (ed.)
Williams ER (1992) The Schumann resonance: a global Atmospheric and Ionospheric Phenomena Associated
rropical thermometer. Science 256: 1184. with Earthquakes, pp. 939-949. Tokyo: Tena.
Williams ER and Heckman SJ (1993) The local diurnal Wilson CTR (1920) Investigations on lightning discharges
variation of cloud electrification and the global diurnal and the electric field of thunderstorms. Philosophical
variation of negative charge on the earth. Journal of Transactions of the Royal Society of London. Series A
Geophysical Research 98: 5221. 221: 73-115,
R G Harrlson and K LAplin, The University of Reading,

Reading, UK
Introduction to Ions in the Global

Atmospheric Electrical Circuit
The terrestrial atmospheric electrical system exchang-
es charge between the planetary surface and the upper
atmosphere, and is maintained by a combination of
charge separation in thunderstorms and vertical ionic
conduction currents in the proportionally rather larger
regions of fair weather. The conduction currents can
only flow because there are charged particles (ions)
present that are sufficiently mobile to be accelerated by
the small electric fields present in the atmosphere. Flgure 1 The role of ions in the atmospheric electrical system.
Consequently, ions have an important part to play in ions are formed throughout the atmosphere by cosmic rays and, in
the maintenance of the global atmospheric electrical the boundary layer, by radon isotopes diffusing from the surface.
circuit. The charge-generating effect of thunderstorms The ions formed may become attached to atmospheric aerosol
leads to a small conduction current throughout fair particies or to water droplets in clouds. Those ions that remain are
accelerated by atmospheric electric fields, originating from charge
weather regions of current density about 2 ~ A r n - ~ .separation occurring in thunderstorms,
The conduction current density J is defined as in
eqn [ 11,where 0 is the conductivity and E the potential
gradient (equal in magnitude to the vertical electric with dimensions of approximately 1nm, have usually
field)I
been referred to as small ions. Large ions (or Langevin
J = oE ions), by comparison, are physically larger (tens to
hundreds of nanometers) and consequently are elec-
If integrated across the surface of the whole planet, the trically less mobile. The historical distinction between
ionic conduction current totals about 2000 A, shown large ions and small ions served to distinguish between
schematically in Figure 1. ions that could be accelerated by atmospheric electric
fields (and therefore directly contribute to the con-
Conductivity of Air
ductivity of air), and those (the large ions) that were
Atmospheric air is rarely regarded as an electrical insufficiently electrically mobile to contribute to
conductor, but it is able to pass a small electric current electrical conduction (Table 1).
because of the ions it contains. The historical devel-
opment of atmospheric electricity spans the discovery
of the electron and the elucidation of the electrical Nature of Atmospheric Ions
structure of matter, which is reflected in the atmos-
Particle Spectrum
pheric electrical usage of the term ion. In general, an
atmospheric ion is a cluster of molecules carrying an Modern measurements have identified small and large
overall charge, i.e., a molecular ion. Such cluster ions, ions as part of the atmospheric aerosol spectrum.
734 ELECTRICITY, ATMOSPHERIC / Ions in the Atmosphere
Table 1 Typical properties of atmospheric particles and droplets Hydration occurs within microseconds of the ioniza-
Particle Radius Mobility
tion (the rate depends on the relative humidity), and
(VJ) (~o-~~v- ~ the
s - ions
~ ) have a lifetime of the order of a hundred
seconds. Positive clusters take the chemical form
Small ions 0.0005-0.0008 0.5-100 I+(HzO), where I+ is a typical atmospheric species,
Large ions 0.008-0.03 0.05-0.5
for example H30'; NO', or NO:; similarly, negative
Aerosol 0.003-30
Cloud droplets 2-30 ions can be represented as I-(HzO), where the ion
Rain drops 30-3000 could be 0; , COT,NH;! or NO,. The average value
of n is typically between 4 and 10. The chemical
differences between the species in the positive and
Aerosol particles span a wide range of sizes in the
negative ions lead to some physical asymmetries in the
atmosphere, from a few nanometers to tens of
properties of the ions, so the negative ions tend to have
micrometers, and the smaller particles are more
fewer water molecules clustered around them and are
prevalent in number concentrations. Large ions are
about 20% more mobile than the positive ions.
therefore effectively submicrometer aerosol particles
that have acquired electric charges. Many of the Electric Charge
smaller aerosol particles are capable of acting as
condensation nuclei, depending on the supersatura- Small ions are singly charged and the dominant forces
tion, and a small fraction can act as cloud condensa- acting on them are electrical. The electrical mobility p
tion nuclei (CCN). It is not thought that charge is a describes the speed that a molecular ion will attain in
significant property of CCN. an electric field. It is the ratio of the magnitude of the
electric field to the ion's drift velocity (The electrostatic
Ion Formation forces acting on the particle are assumed to balance the
drag forces, so that there is no net acceleration). Small
Small ions are formed by the interaction between high- ions have a relatively high mobility of around
energy radiation and air molecules, arising from
10-4m2V-1 s-', and are more strongly influenced
natural radioactivity and cosmic radiation. Atomic by electric fields than are other aerosol particles. The
ions formed by radiolysis are highly polarizing, and a current flowing through the air is almost entirely due
cluster of water molecules rapidly forms at a rate to these small ions, and the conductivity D can be
determined by the availability of water vapor. Large written as eqn [2],where there are number concen-
ions form entirely differently; they result from colli- trations n+ and n- of positive and negative ions,
sions between pre-pexisting aerosol and small ions, with mobilities p + and ,L- respectively, and e is the
which lead to the small ion becoming attached to the
magnitude of the electronic charge.
larger particle. The contribution of the small ion mass
to the combined particle is insignificant, but the
associated charge transfer leads to the resulting
D = e(n+p+ + n-p-) PI
particle becoming charged. Further small ion-large Other, larger, ions exist in the atmosphere, but their
ion collisions can occur, leading to multiple charging electrical mobilities are several orders of magnitude
of large ions, or, if the combining small ion and large less than those of the small clusters, so their contribu-
ion have opposite polarities, to neutralization. The tion to the air conductivity is negligible in comparison.
interaction between small ions and large ions is
principally a result of kinetic factors, and electrical Ion Transport
forces only become significant for highly charged large With typical atmospheric electric fields in undisturbed
ions. Collisions between aerosol and small ions are conditions of 100 Vm-', small ions will typically
important as they serve to remove small ions: in air migrate under electrical forces at about 1ems-' or
containing large quantities of aerosol the small ion less. However the electrical migration is able to occur
content is likely to be correspondingly small, a in regions where there is little dynamical motion, in
principle exploited in the domestic smoke alarm. meteorological terms, and consequently clean air will
This also means that, although the conductivity of air always have fresh ions introduced into it either by
is principally directly due to its small ion content, the electrical migration or by in situ ionization. Large ions
conductivity is indirectly influenced by large ions and and atmospheric aerosol, however, have negligible
aerosol through removal of small ions. electrical acceleration, and their primary motion is due
to advective transport arising from dynamical pro-
Chemical Composition
cesses. The transport of charge may therefore be ionic
Atmospheric small ions consist of clusters of water or particulate (large ions), and the partitioning of the
molecules collected around a singly charged ion. atmospheric space charge (the net amount of charge
Next Page
ELECTRICITY, ATMOSPHERIC / Ions in the Atmosphere 735
per unit volume) between these processes is therefore Bipolar Ion Equations
of relevance to the ultimate destination of the charge.
Equations [3] and [4] describe these ion processes.
Ion Balance Equation dn+

-= q - m+n- -
dt [31
It is possible to express the behavior of atmospheric
ions in terms of a simplified equation in which their
rates of formation and removal are considered 141
separately.
The rate of change of ion concentration is determined
by an ion source term q, and the two loss terms
Source Term described above. Ion-pairs are produced at a rate q per
unit volume. The rate of ion loss by attachment is given
Formation of ions is essentially due to radiolysis of air by nPZ, where p+ and p- are the appropriate ion-
molecules. There are three principal sources near the aerosol attachment coefficients and Z is the aerosol
surface and in the planetary boundary layer: radon number concentration. In these equations, the consid-
isotopes, cosmic rays, and terrestrial gamma radia- erable simplification has been made of assuming that it
tion, and the partitioning between the sources varies is possible to regard the aerosol as monodisperse (Le.,
vertically. Near the surface, ionization from turbulent having a single particle size) so that a single value of 2
transport of radon and other radioactive isotopes is is appropriate. The attachment coefficient varies with
important, together with gamma radiation from aerosol radius, therefore pZ is in fact more accurately
isotopes below the surface. Ionization from cosmic determined for atmospheric particles as an integral
rays is always present, comprising about 20% of the evaluated across the aerosol size distribution. There
ionization at the surface; it increases with increasing are many theoretical approaches to the problem of
height in the atmosphere and dominates above the calculating p, which depend on the aerosol particle
planetary boundary layer, as depicted in Figure 1.The size considered, but p is typically 5 x 10 m3 s for-’
rate of production of ion-pairs per unit volume is aerosol particles with the radii most prevalent by
usually written as q, and this quantity shows consid- number (0.2 pm) in lower regions of the atmosphere.
erable temporal and spatial variability. The increasing
fractional contribution of cosmic ray ionization with Time-dependent Solution
height has already been mentioned, but the temporal It is instructive to simplify the ion balance equation by
variations in q at the surface are also substantial neglecting the ion sign (i.e. n+ n- = a ) , so that the
owning to the effects of turbulence and the transient ion-aerosol equation can be written as eqn [5].
production of ions by cosmic rays. A frequently dn
quoted surface value for q of IO ion-pairs cm-3 s - l is - = q - un2 - n p z 151
a long-term mean. On time scales of seconds, there can dt
be significant departures from this value. In the middle Integrating this equation gives the ion concentration n
and upper atmosphere, the cosmic contribution to q as a function of time t. This time-dependent solution is
dominates. given in eqn [6], from which several instructive points
should be noted.
n ( t )=
Removal Terms
If there are number concentrations n+ and n- of
positive and negative ions, respectively, then the rate at
which they recombine is proportional to the product
n+n-, with a constant of proportionality oc, also known
as the ion-ion recombination coefficient and typically
assumed to be 1.6 x lO-l2m3 s-’. This is the principal
loss mechanism of ions in clean, aerosol-free air. If First, if the ion-pair production rate q is uniform and
aerosol is present, then ions are also lost by attachment the removal rates are also steady, the ion concentration
to the aerosols, and this mechanism can dominate in tends to a steady value for large values oft. Second, the
polluted air. The rate of ion loss is again proportional equation can be simplified according to the situations
to the product of number concentrations, in this case in which attachment or recombination dominates as
the product of the aerosol and the ion number the removal mechanism, according to whether an2 or
concentrations. npZ is the bigger term. In the atmosphere in polluted
ENERGY BALANCE MODEL, SURFACE 747
Huang E, Williams E, Boldi R, et al. (1999) Criteria for systems. Journal of Geophysical Research 101:
sprites and elves based on Schumann resonance obser- 29641-29652.
vations. Journal of Geophysical Research 104: MacGorman DR and Rust WD (1998)T h e Electrical Nature
16943-16964. of Storms. New York: Oxford University Press.
Journal of Atmospheric and Solar-Terrestrial Physics: the Rowland HL (1998) Theories and simulations of elves,
May-June, 1998 issue was dedicated to TLEs and sprites and blue jets. Journal of Atmospheric and Solar-
provides a valuable source of references. Terrestrial Physics 60: 831-844.
Lyons WA (1996) Sprite observations above the U.S. High Uman ML (1987) The Lightning Discharge. New York:
Plains in relation to their parent thunderstorm Academic Press.
T S Ledley, TERC, Cambridge, MA, USA values of different physical quantities on a grid. The
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. physical spacing of that grid defines the resolution of
the model. Models average the physical quantities
over the area of a grid box, which in the simplest
Introduction models could be over an entire spatial dimension.
The climate of the Earth system is constantly changing, Third, models represent processes that occur on scales
and scientists are interested in understanding how and smaller than the grid spacing of the model with
why the climate of the Earth system has changed in the formulas that are based on processes that occur on the
past, what has shaped the climate of the current Earth larger scales. This type of formulation is called a
system, and how the climate of the Earth system might parameterization. With all of these simplifications
change in the future. While it is possible to collect models cannot exactly simulate the climate system.
information from geological records about the cli- The earliest models were also the simplest ones,
mates of the past, and to monitor the Earth system both in the number of processes included and in the
today to determine what the climate is now, these data spatial resolution of the model, since computer
do not allow us to understand how the characteristics resources available to run them were limited. As
and processes in the Earth system produced the computing power increased, models were able to
climates of the past and the present. In addition, we include more of the processes that occur in the Earth
can only speculate about the climates of the future. system and to increase the spatial resolution to finer
In order to develop a better understanding of how grids.
the Earth system has worked in the past and how it The simplest models average over all horizontal and
might evolve into the future, a hierarchy of computer vertical scales to obtain a single value for the
models has been developed as tools to study the Earth temperature of the Earth system. Only the most basic
system. A computer model of the Earth system is a physics acting on the largest scales is incorporated into
mathematical formulation of how that system works these models, and the models can be used only in the
implemented in a computer program. Into each model broadest sense to understand how this physics shapes
is incorporated all current knowledge of how parts of climate. The results cannot be assigned to any
the Earth system interact, after which, using data that particular place on the Earth but only to the system
describe the present climate or a climate of the past as a as a whole.
starting point, the model is employed to simulate the As models become more complex their resolution
climate of the Earth under a wide range of conditions increases, so that variations in latitude, in longitude,
and assumptions. and in the vertical can be examined. With the increase
If all of the current understanding of the climate in resolution comes an increase in the number of
system on all spatial and time scales were included in a characteristics and physical processes included in the
computer model it would be too complex and costly to model. While this increase in complexity increases the
run. As a result even the most complicated models are model’s ability to simulate the climate system and its
relatively simple representations of the real Earth change over time, there are still many characteristics
system. There are basically three ways that models and processes acting on scales smaller than the
simplify the Earth system. First, all models at some resolution of the model that are either completely
level use empirical relationships to represent complex omitted or are parameterized in simple ways. As a
physical processes. Second, models use and compute result these models have a limited ability to simulate or
748 ENERGY BALANCEMODEL, SURFACE
predict climate. However, their simplicity allows Here P, is the fraction of the solar radiation that is not
scientists to examine how the components and reflected that penetrates the surface, cx the surface
processes of the Earth system that are included albedo (the fraction of radiation reflected), F,, the
interact, and to study how changes in those compo- short-wave radiation available at the surface, Fir the
nents and processes might change the climate. long-wave radiation from the surface to the atmo-
The most complex models are the general circula- sphere, Fl, the long-wave radiation from the atmo-
tion models. These models, developed originally for sphere to the surface, Fl the latent heat flux, F, the
the atmosphere and now including the atmosphere, sensible heat flux, and Fcondthe conductive flux from
ocean, biosphere, and cryosphere, have the highest below the surface
spatial resolution and demand large amounts of If the sum of the energy fluxes in eqn [l]does not
computer resources, though the resolution is still too equal zero, the imbalance of energy results in a change
coarse, for example, to resolve individual clouds. As a in temperature defined by
result many of the smaller-scale processes important to
shaping the climate of the Earth system are para-
meterized only simply in these models.
where T is the temperature of the surface, t time, p
density, and c the specific heat.
The Surface Energy Balance
F,, is the incoming short-wave radiation available
The Earth system operates close to an energy balance. at the surface. This radiation is also referred to as solar
This means that an equal amount of energy comes into radiation as most short-wave radiation in the Earth
the Earth system and goes out of it, and as a result the system originates from the Sun. Most solar radiation
temperature of the entire system over a long period of entering the top of the Earth’s atmosphere is trans-
time is relatively constant. However, within the Earth mitted through the atmosphere to the surface or to the
system there are variations over time and over space. top of clouds. At that point it is either reflected back
Some of these are the result of regular cycles such as the through the atmosphere to space, absorbed at the
seasonal cycle, the El Niiio Southern Oscillation surface or in the cloud where it heats the surface or
(ENSO),and glacial/interglacial variations, or regular cloud, or penetrates through the surface or edge of the
changes in location such as the steady decrease in cloud to be absorbed below.
temperature from the Equator to the poles. Some are The albedo, CI, determines how much of the short-
the result of random variations called natural varia- wave radiation that reaches the surface gets reflected
bility, which produce, for example, the day-to-day back to space. cx is expressed as a fraction ranging from
variations we see in our weather, variations in the 0 (no radiation reflected) to 1 (all reflected). So in eqn
strength of yearly monsoons, variations in the number [l](1 - cx) is the fraction of the short-wave radiation,
and the location of landfall of hurricanes, and the not reflected, and (1 - E ) x F,, the amount of the
somewhat irregular intervals of ENSO. Some of the available short-wave radiation that is not reflected.
variations in time and space are the result of changes in P, determines how much of the short-wave radia-
surface conditions such as whether the surface is land tion not reflected back to space is transmitted through
or water, or covered by snow and ice covered. These the surface. Pe is also expressed as a fraction ranging
changes in surface conditions produce changes in the from 0 (no available short-wave radiation not reflected
surface energy balance. The changes in these surface penetrates the surface) to 1 (all available short-wave
conditions affect the amount of energy retained by the radiation not reflected penetrates the surface). So in
Earth system and how it is distributed within that eqn [l](1 - P,) is the fraction of the available short-
system. wave radiation that does not penetrate the surface and
Researchers who have sought to simulate the (1 - P,) x (1 - a)xF,, is the amount of the short-wave
climate of the Earth system and to understand how radiation not reflected that does not penetrate the
and why it changes over time have used models, which surface, or, in other words, the amount of short-wave
in all but the simplest globally averaged cases have radiation absorbed at the surface. This energy flux is
taken into account the surface energy balance. While always directed toward the surface, representing a
more recent work has included the complexity of the gain by the surface, and is never negative.
biosphere in the surface energy balance, the surface Every object radiates energy at a wavelength
energy balance is most simply described by the proportional to the fourth power of its temperature
following equation: in kelvin (K). The temperatures of the Earth’s atmo-
sphere and surface are in a range where the wavelength
of the radiation they emit is in the infrared part of the
electromagnetic spectrum. Since the wavelength of
this radiation is longer than that coming from the Sun, representing a loss to the surface and a gain by the
it is commonly called long-wave radiation. atmosphere.
Fir is the long-wave radiation coming from the In early energy balance models the latent heat flux
surface to the atmosphere. It can be represented in the was computed using a bulk aerodynamic formula of
surface energy balance equation as the form
F1 = c ~~vL (u~
- a,)
where Cdl is the drag coefficient for latent heat flux

where E is the emissivity which ranges from 0 to 1,o the (typical value 1.32 x v the wind speed, L the
Stefan-Boltzman constant, and T, the surface tem- latent heat of vaporization, and a, and a, the absolute
perature. humidities of the surface and the air at the surface
F1, is the long-wave radiation coming from the respectively. Recent models also account for the
atmosphere to the surface. The representation of FI, in impact of vegetation on the latent heat flux.
the surface energy balance equation is mainly deter- The sensible heat flux, F,, is the exchange of energy
mined by the temperature of the air near the surface, between the surface and the atmosphere that results
though it is also affected by the atmospheric humidity from the temperature difference between the surface
and cloudiness. and the atmosphere. The bigger the difference in the
Since Fir and Fl, are both long-wave radiative temperature between the surface and the atmosphere
fluxes, one of which is directed away from the surface the larger is the flux of energy. If the surface is warmer
(Fl,) and one toward the surface (Flw),they partially than the atmosphere then the flux is from the surface to
cancel. In regions or at times when the surface is the atmosphere. If the atmosphere is warmer than the
warmer than the atmosphere, such as over land during surface then the flux is from the atmosphere to the
the day, when the surface is heated by incoming solar surface. Thus this flux can provide either a gain or a
radiation, and over ocean in the high polar latitudes, loss by the surface.
where the water surface is usually warmer than the air, In early energy balance models the sensible heat flux
the net long-wave radiative flux is directed from the was computed using a bulk aerodynamic formula of
surface to the atmosphere. However, when the atmo- the form
sphere is warmer than the surface, as over sea ice
during the Arctic winter and over land surfaces at
night during the winter, the net long-wave radiative where Cds is the drag coefficient for sensible heat
flux is directed from the atmosphere to the surface. flux (typical value 1.41 x v the wind speed,
Thus the net long-wave radiative flux can be directed p the surface air density, Cp the specific heat of dry
to the surface or to the atmosphere, depending on air, and T, and T, the surface and surface air
conditions. temperatures.
The latent heat flux, FI, is the exchange of energy The latent heat flux and the sensible heat flux are
between the surface and the atmosphere that occurs turbulent energy fluxes. This is because these
when water is evaporated from or condenses onto the exchanges of energy between the surface and the
surface. When water is evaporated, the energy atmosphere are affected by the magnitude of the wind
absorbed by it causes it to change state from a liquid speed. As the wind speed increases the energy flux also
to a gas (water vapor) rather than to change the increases.
temperature. The gas then mixes with the rest of the Fcondis the conductive flux from beneath the surface.
atmosphere, carrying the latent heat with it. When the This flux varies, depending on the material the surface
gas changes back into a liquid (condensation),such as is made of and the environmental conditions. If
when clouds form, that heat is released into the temperature decreases with depth, as it generally
environment where the water condenses. As a result does on land near the surface, the conductive flux is
the energy has been moved from the surface, where the directed from the surface to beneath the surface,
water was evaporated, to higher in the atmosphere, representing an energy loss by the surface. However, if
where the water condenses. The strength of this latent the temperature increases with depth, as in the case of
heat flux is dependent mostly on the relative amounts sea ice, which is floating on warmer ocean water, the
of water at the surface and in the air just above the conductive flux is directed toward the surface,
surface. In general, the water, and thus the latent heat, representing an energy gain by the surface.
moves from where there is more water to where there When the amount of energy coming to the surface
is less. In most circumstances the surface contains equals the amount of energy leaving it, the surface is
more water than the atmosphere, so the latent heat said to be in energy balance and the temperature
flux is generally from the surface to the atmosphere, remains constant. However, if more energy is coming
750 ENERGY BALANCE MODEL, SURFACE
into (leaving)the surface than is leaving (coming into) the boundaries of the region. So the net increase or
it then there is a positive (negative) energy imbalance decrease of energy in the region of the air over land (for
and the temperature increases (decreases) in order to example) is determined by the following:
restore the balance.
In order to study the state of the Earth system and Energy exchange Energy exchange
how it varies over time, computer models of the Earth between the land between the air
system, which includes the surface energy balance and air
+ space
equation discussed above, are used to identify which of
A B
the energy fluxes are important in establishing the state
of the Earth’s environment and how the Earth system Energy exchange Energy exchange
responds to changes in the various energy fluxes. between this zone between this zone
+ and zone to south
+ and zone to north
C D
An Energy Balance Model Energy exchange Net change in
between this zone energy content
Figure 1 is a schematic diagram of an energy balance
model that incorporates a full surface energy balance.
+ and air over ocean
-
-
of zone
The diagram shows one latitude zone between
E F [31
latitudes q5 and #I + Aq5, and indicates that the model
treats four distinct regions of the Earth system, where the term is positive if the flux is into the zone and
including air over land, land, air over ocean, and negative if the flux is out of the zone.
ocean. An energy balance is computed for each of these In a more complex model, which has more resolu-
regions. If there is an energy surplus (deficit)in a region tion in the east-west direction, rather than the simple
then the temperature is computed to rise (fall) to land-sea resolution depicted in Figure 1,eqn [3] would
restore energy balance. have to include two terms for the exchange of energy
In an energy balance model the main parameter to within a latitude zone, one for the exchange to the east
be computed is the temperature. In eqns [ 11and [2]the and one for the exchange to the west. In this model,
temperature under discussion is the temperature of the with only two latitudinal zones resolved, we need
surface. In the energy balance model shown schema- consider only the exchange of energy between them.
tically in Figure 1the focus is the temperature of each Term A in eqn [3] is the surface energy balance
of the regions. The temperature of each region is explicitly stated in eqn [l]. If an energy balance
determined by summing the energy crossing each of occurs at the surface then term A is 0, as shown in
Figure 1 Schematic diagram of one latitude zone of a simple energy balance model that employs a full surface energy balance.
(Adaptedwith permission from Ledley TS (1988) Acoupled energy balance climate-sea ice model: impact of sea ice and leads on climate.
Journal of Geophysical Research 93: 15919-1 5932.01 988 by the American Geophysical Union.)
eqn [l].If there is no surface energy balance then By choosing a particular time step, At, over which to
term A is non-zero and it contributes to the net apply eqn [4]one can solve for the new temperature:
change in energy of the region. In recent modeling
T ( t )+ AE*At
work, term A has been developed to include the T ( t + 1) = PI
influence of vegetation on the surface energy balance. PC
Vegetation, with its leaves and root systems, intro-
duces a much more complex picture of the surface The temperature computed here is the temperature
energy balance, which when included in models results representative of the whole layer of air. In order to
in a better simulation of the climate of the Earth determine the surface air temperature, which is
system. generally desirable in order to compute the surface
Term B in eqn [ 3 ] is the net change in energy at the energy fluxes, another equation must be applied that
top of the atmosphere. Since the atmosphere is relates the temperature computed in eqn [5] to the
relatively transparent to solar radiation, most solar surface air temperature.
radiation reaches the surface and is represented as the In using eqn [5] to compute the new temperature
first term in eqn [l]. The only other significant transfer one must be careful that the energy balance equation is
of energy between the atmosphere and space is of valid over the chosen time step size. If the equation is
infrared radiation (long-wave radiation). Since not valid over the chosen time step size then the
the atmosphere is warmer than space, the flux of numerical result of the calculation may not be
infrared radiation is from the atmosphere to space, physically realistic.
representing a loss by the atmosphere (see the IR term The other three regions have similar energy balance
in Figure 1). equations applied to them. The equation for the
Terms C and D in eqn [3] represent the meridional temperature of the air over ocean is different only in
heat transport between the zone under consideration the values of the variables, constants, and coefficients,
and those to the north and south. Since any flux across which are changed so that they represent the character
the pole represents a flux out of the zone under of the ocean surface rather than the land surface.
consideration and then back into it, this flux is The energy balance applied to the land includes
generally set to zero. In the simplest models this terms A, By and F in eqn [ 3 ] .Term A represents the
meridional heat transport is computed on the basis energy exchange between the land and the air over
of the temperature gradient between the zone land. Term B is altered to represent the exchange of
under consideration and those to the north and energy between the land surface layer and deeper
south, and the assumed diffusion coefficients. In layers within the Earth (not shown in Figure 1). This
some cases there may be more than one adjacent energy flux is called the geothermal flux. Term F
zone to the north or south, i.e., air over land in one represents the effect of any imbalance in energy on the
zone may be adjacent to both air over land and air over temperature of the land. In general the resolution of
water to the north or south, depending on the amount energy balance models is too coarse for the horizontal
of land and ocean in each zone. In that case the fluxes exchanges of energy between land in the zone and
from both of these zones must be included in terms C adjacent land or water to have a significant impact on
and D. the temperature of the land in the zone.
Term E in eqn [3] represents the zonal heat transport The energy balance applied to the ocean includes all
between the air over land in the zone under considera- the terms applied to the land with the addition of the
tion and the air over ocean in the same zone. meridional heat transport between adjacent ocean
This can be computed from the temperature gradient regions. The currents in the ocean carry a significant
between the air over land and sea and a diffusion amount of energy both meridionally and zonally, and
coefficient. thus, while the ocean currents are not included
Term F in eqn [ 3 ]represents the net change in energy explicitly in the model, their impact on the energy
in the zone. It can be represented in general by eqn [2]; balance must be included.
however, the temperature now is the temperature of
the air over land and the density and heat capacity are
those for the air over land.
Energy Balance Models as Tools
Energy balance models have been valuable tools in
The temperature of the air over land is computed as
follows: the study of the climate of the Earth system; however,
they are dramatic simplifications of the Earth atmo-
Net change in the energy in the zone = AE sphere system that either exclude or else represent in
only a simple form the real physics and biogeochem-
= (pc)aT/at [4] ical processes that occur. One example of this
752 ENERGY BALANCE MODEL, SURFACE
simplification is that the dynamics of the atmosphere, Table 1 Mean annual sea ice thickness and maximum percen-
i.e., the processes that produce the high- and low- tage of open water or period of ice-free conditions at 75" N and
pressure systems in midlatitudes, monsoons, hurri- 75" S for various cases of specified minimum lead fraction
canes, and tornadoes that are key in producing the 75"N 75"s
exchanges of energy between different latitudinal
Minimum lead Ice thickness (m) 1.42 m 2.02 m
zones and between regions over land and water, are
fraction = 0%
represented only with respect to how they effect energy Maximum YOopen water or period of ice-freeO% 0%
distribution. They are not included in a realistic way. conditions
Thus any results from energy balance models must be
viewed with these simplifications in mind. Minimum lead Ice thickness (m) 1.30m 2.10m
fraction = 1.1%
However, energy balance models have several
Maximum o/o open water or period of ice-free8 weeks 1.7%
advantages. The most important of these are (1)they conditions
are quick, allowing multiple experiments, and (2)it is
relatively easy to analyze model results so that the Minimum lead Ice thickness (m) 1.31 m 2.29 m
causes of a change in the simulated system can be fraction = 2.2%
Maximum % open water or period of ice-free9 weeks 3.4%
traced from the imposed perturbation to the resulting
conditions
change. These advantages allow the study of parti-
cular processes in the Earth system and of how the Minimum lead Ice thickness (m) 1.36 m 2.94 m
Earth system responds to various changes. The results fraction = 4.3%
of these studies can then be used to guide the Maximum Yoopen water or period of ice-free 10 weeks 6.3%
conditions
development of experiments in more complex models
which are more costly to run and more complex to Adapted with permission from Ledley TS (1988) A coupled energy
analyze, but do include the dynamics of the atmo- balance climate - sea ice model: impact of sea ice and leads on
sphere and ocean, a much higher spatial resolution, climate. Journal of Geophysical Research 93: 15919-15932. 0
1988 American Geophysical Union.
and many other physical processes of the Earth system.
An example of how energy balance models can be
used to investigate the relative importance of processes In the study the effect of changing this minimum
that contribute to climate change is an investigation area of open ocean in the winter is investigated.
using a version of the energy balance model described Table 1 shows the mean annual sea ice thickness and
earlier that was conducted to examine how small the maximum percent of open water or period of
variations in the minimum amount of open water in ice-free conditions during the summer for various
the sea-ice-covered polar oceans during the winter specifications of the minimum lead fraction. Figure 2
affect the maximum amount of open water during shows the seasonal cycle of surface air temperature
the summer and the seasonal cycle of surface air zonally averaged over ocean, sea ice, and land as a
temperatures. function of latitude for the control case (minimum lead
The first step in this study is to compare the climate fraction during the winter of 1.1%) and the changes in
simulated with the energy balance model to observa- the seasonal cycle of surface air temperature when the
tions of the current climate to assure that the model is minimum lead fraction is reduced to O%, increased to
able to produce a reasonable simulation of the present 2.2%, increased to 4.3%, and increased to 100%
climate. This includes comparisons of the mean (meaning no sea ice is allowed to form).
annual seasonal cycle of simulated and observed Table 1 shows that when the minimum lead fraction
surface air and surface temperatures, sea ice thickness, is increased from the 1.1% control case, increasing the
area of open ocean, and energy fluxes. The energy amount of open water during the winter, the mean
balance model used for this comparison specifies a annual thickness of the sea ice and the maximum area
minimum lead fraction (theminimum fraction of open of open water or period of ice free conditions during
water in sea ice, a lead being a crack in the sea ice that the summer both increase. The increase in sea ice
exposes ocean water to the atmosphere) of 1.1%. This thickness occurs because the increase in the area of
means that for the winter over the polar oceans, when open ocean during the winter causes an increase in the
the heat loss would produce ice growth in any area of amount of heat lost by the ocean and thus increases the
open water, the model determines the amount of ice production of sea ice. When leads in the sea ice are
that would grow and then mechanically open 1.1% of completely eliminated (minimum lead fraction = 0%)
the ocean area to be ice-free, as would occur as the the sea ice thickness at 75"S decreases owing to the
result of wind stresses and ocean currents in the real decrease in the ice growth rate. However, at 75"N the
word. The climate simulated by this version of the mean annual sea ice thickness is increased because of a
energy balance model is called the control case. decrease in the melting of sea ice during the summer.
T
70
50
30
p"
3
10
.-
4-
5
1
-10 3 -10
1
-30 -30
-50
-70
(ATg0 ( C ) J w D J F M A M J J A S O N
Month
-50 -
&:-%
-50 -
-70 - 0.4
0
0.4 f0;4
Figure 2 (A) The seasonal cycle of surface air temperature zonally averaged over the ocean, sea ice, and land, as a function of latitude for the control case, in degrees Kelvin. (B-D) The
seasonal cycle of the change in the surface air temperature zonally averaged over the ocean, sea ice, and land, as a function of latitude from the controlcase when (B) no leadsare specified, (C)
when the minimum lead fraction is equal to 2.2%,and (D) when the minimum leadfraction is equal to 4.3%. (Adapted with permissionfrom Ledley TS (1988) A coupled energy balance climate-
sea ice model: impact of sea ice and leads on climate. Journal d Geophysjd Research 93:15919-15932.01988 by the American Geophysical Union.)
754 EVOLUTIONOF ATMOSPHERIC OXYGEN
The study goes on to examine the impact of those Further Reading

changes in sea ice thickness and the relative amounts of
Few AA (1996) System Behavior and System Modeling.
open water on surface air temperatures through the Saucalito, CA: University Science Books.
year. Figure 2 shows that the impact on the surface air
Ghil M (1981) Energy-balance models: an introduction. In:
temperatures is rather small during the summer; Berger A (ed.) Climatic Variations and Variability: Facts
however, during the winter, when there is a large and Theories, pp. 4 6 1 4 8 0 . Boston, MA: Reidel.
difference between the surface temperature of ocean Harvey LDD (2000) Global Warming: The Hard Science.
Harlow: Prentice-Hall.
water and of sea ice, the impact of small changes in the
Ledley TS (1985) Sensitivity of a thermodynamic sea ice
area of open ocean on surface air temperature is large.
When areas of open ocean are eliminated, during the model with leads to time step size.Journal of Geophysical
winter, the atmosphere is cut off from a heat source, Research 90: 2251-2260.
Ledley TS (1988). A coupled energy balance climate-sea ice
namely the relatively warm ocean, and surface air
model: impact of sea ice and leads on climate. Journal of
temperatures drop by between 0.4 K and 1.2 K during Geophysical Research 93: 15919-15932.
the winter. When areas of open ocean are increased North GR (1975) Theory of energy balance climate
during the winter, the atmosphere is in contact with anmodels. Journal of the Atmospheric Sciences 32:
expanded heat source, and surface air temperatures 2033-3043
increase by up to 2.0 K when sea ice is allowed to form,
North GR, Cahalan RF and Coakley JA (1981). Energy
and much more when it is not. balance climate models. Reviews of Geophysics and
The use of energy balance models in this kind of Space Physics 19: 91-121.
study permit further investigation to identify whichOjima D (ed.) (1992) Modeling the Earth System.
energy fluxes contribute to the simulated changes in Boulder, CO: UCAWOffice for Interdisciplinary Earth
surface air temperature and under what conditions Studies.
Rosenzweig C and Dickinson RE (1986). Climate-Vegeta-
each energy flux is the most important.
tion Interactions. Boulder, CO: UCAWOffice for Inter-
disciplinary Earth Studies.
Saltzman B (ed.) (1983) Theory of climate. Advances in
See also Geophysics 25: New York: Academic Press.
Air-Sea Interaction: Momentum, Heat and Vapor Schneider SH and Dickinson RE (1974). Climate
Fluxes; Sea Surface Temperature. Boundary Layers: modeling. Reviews of Geophysics and Space Physics
Modeling and Parameterization. Coupled Ocean-Atmo- 12: 4 4 7 4 9 3 .
sphere Models. General Circulation: Models. Land- Sellers PJ, Dickinson RE, Randall DA, et al. (1997)
Atmosphere Interactions: Overview. Mesoscale Modeling the exchanges of energy, water, and carbon
Meteorology: Models. Radiation (Solar). Reflectance between continents and the atmosphere. Science 275:
and Albedo, Surface. Teleconnections. 502-509.
D Catling, University of Washington, Seattle, WA, important gases are at least in part biologically
USA controlled, but oxygen in particular has no significant
K Zahnle, NASA Ames Research Center, Moffett Field, abiotic source. Diatomic oxygen is generated by
CA, USA oxygenic photosynthesis, the biological process in
which water molecules are split using the energy
of sunlight. Today, green plants, single-celled
phytoplankton (free-floating organisms in the ocean),
including cyanobacteria (chlorophyll-containing bac-
teria) all perform oxygenic photosynthesis. Of these,
Introduction
Abundant free oxygen in the atmosphere distinguishes
-
cyanobacteria are the most numerous, with lo2’ in
the oceans, and probably their ancient ancestors were
our planet from all others in the solar system. Earth’s just as plentiful. However, geological differences
oxygen-rich atmosphere is a direct result of life. The between the ancient and modern Earth show that
current atmosphere contains (by volume) 78.09% N2, there was insufficient 0 2 in the early atmosphere to
20.95% 0 2 , 0.93% Ar, 0.036% COZ, and additional leave traces of oxidation that today are ubiquitous,
trace gases. Apart from argon, all of the quantitatively such as the reddening of exposed iron-rich rocks. The
The study goes on to examine the impact of those Further Reading

changes in sea ice thickness and the relative amounts of
Few AA (1996) System Behavior and System Modeling.
open water on surface air temperatures through the Saucalito, CA: University Science Books.
year. Figure 2 shows that the impact on the surface air
Ghil M (1981) Energy-balance models: an introduction. In:
temperatures is rather small during the summer; Berger A (ed.) Climatic Variations and Variability: Facts
however, during the winter, when there is a large and Theories, pp. 4 6 1 4 8 0 . Boston, MA: Reidel.
difference between the surface temperature of ocean Harvey LDD (2000) Global Warming: The Hard Science.
Harlow: Prentice-Hall.
water and of sea ice, the impact of small changes in the
Ledley TS (1985) Sensitivity of a thermodynamic sea ice
area of open ocean on surface air temperature is large.
When areas of open ocean are eliminated, during the model with leads to time step size.Journal of Geophysical
winter, the atmosphere is cut off from a heat source, Research 90: 2251-2260.
Ledley TS (1988). A coupled energy balance climate-sea ice
namely the relatively warm ocean, and surface air
model: impact of sea ice and leads on climate. Journal of
temperatures drop by between 0.4 K and 1.2 K during Geophysical Research 93: 15919-15932.
the winter. When areas of open ocean are increased North GR (1975) Theory of energy balance climate
during the winter, the atmosphere is in contact with anmodels. Journal of the Atmospheric Sciences 32:
expanded heat source, and surface air temperatures 2033-3043
increase by up to 2.0 K when sea ice is allowed to form,
North GR, Cahalan RF and Coakley JA (1981). Energy
and much more when it is not. balance climate models. Reviews of Geophysics and
The use of energy balance models in this kind of Space Physics 19: 91-121.
study permit further investigation to identify whichOjima D (ed.) (1992) Modeling the Earth System.
energy fluxes contribute to the simulated changes in Boulder, CO: UCAWOffice for Interdisciplinary Earth
surface air temperature and under what conditions Studies.
Rosenzweig C and Dickinson RE (1986). Climate-Vegeta-
each energy flux is the most important.
tion Interactions. Boulder, CO: UCAWOffice for Inter-
disciplinary Earth Studies.
Saltzman B (ed.) (1983) Theory of climate. Advances in
See also Geophysics 25: New York: Academic Press.
Air-Sea Interaction: Momentum, Heat and Vapor Schneider SH and Dickinson RE (1974). Climate
Fluxes; Sea Surface Temperature. Boundary Layers: modeling. Reviews of Geophysics and Space Physics
Modeling and Parameterization. Coupled Ocean-Atmo- 12: 4 4 7 4 9 3 .
sphere Models. General Circulation: Models. Land- Sellers PJ, Dickinson RE, Randall DA, et al. (1997)
Atmosphere Interactions: Overview. Mesoscale Modeling the exchanges of energy, water, and carbon
Meteorology: Models. Radiation (Solar). Reflectance between continents and the atmosphere. Science 275:
and Albedo, Surface. Teleconnections. 502-509.
D Catling, University of Washington, Seattle, WA, important gases are at least in part biologically
USA controlled, but oxygen in particular has no significant
K Zahnle, NASA Ames Research Center, Moffett Field, abiotic source. Diatomic oxygen is generated by
CA, USA oxygenic photosynthesis, the biological process in
which water molecules are split using the energy
of sunlight. Today, green plants, single-celled
phytoplankton (free-floating organisms in the ocean),
including cyanobacteria (chlorophyll-containing bac-
teria) all perform oxygenic photosynthesis. Of these,
Introduction
Abundant free oxygen in the atmosphere distinguishes
-
cyanobacteria are the most numerous, with lo2’ in
the oceans, and probably their ancient ancestors were
our planet from all others in the solar system. Earth’s just as plentiful. However, geological differences
oxygen-rich atmosphere is a direct result of life. The between the ancient and modern Earth show that
current atmosphere contains (by volume) 78.09% N2, there was insufficient 0 2 in the early atmosphere to
20.95% 0 2 , 0.93% Ar, 0.036% COZ, and additional leave traces of oxidation that today are ubiquitous,
trace gases. Apart from argon, all of the quantitatively such as the reddening of exposed iron-rich rocks. The
EVOLUTIONOF ATMOSPHERIC OXYGEN 755
transition from the ancient anoxic atmosphere to a An analysis of the average composition of modern
more modern oxic atmosphere appears to have sediments by Heinrich Holland shows that organic
occurred at about 2.3 Ga (where Ga = lo9 years carbon and pyrite burial contribute the equivalent
ago). Although this ‘rise of 02’ has been seen by some to oxygen fluxes of (10.Ok3.3) x 1OI2 mol 0 2 y-’
as controversial, there is more than enough evidence in and (7.8f4.0) x 1 O I 2 mol O2 y - l , respectively.
its favor for it to have come to be regarded as the The reduction of oxidized iron and the burial of
standard history of the Earth’s atmosphere. This ferrous iron (2Fe203 = 4 F e 0 + 0 2 ) also adds a
history has been a major influence on life. All animals, minor flux of oxygen, (0.9k0.4) x 10I2 mol 0 2 y - l .
multicellular plants, and fungi rely on free oxygen to The burial of sulfate minerals in sediments removes
maintain their energy intensive lifestyles. Life on the (0.3f0.1) x 1 O I 2 mol 0 2 y-’ through the oxidation
planet’s surface also became protected from harmful of SO2 in the atmosphere. Summing these fluxes, the
ultraviolet radiation once 0 2 levels exceeded 0.2- total 0 2 production is (18.4 f 7.8) x 10I2 mol 0 2 y-’
0.6%, causing an ozone (03) layer to form in the (Table 1).
stratosphere. But why free oxygen should have On long time scales, oxidation must balance
become abundant on a planet that is overall chemically 0 2 production to leave a steady amount of 0 2 in
reducing, or exactly why it first appeared about half- the atmosphere. About 80-90% of 0 2 production
way through Earth’s 4.5 billionyear history, are still is consumed during oxidative weathering, while
open questions. Nor is there yet any firm understand- 10-20% reacts with reduced gases in the atmosphere.
ing of what regulates 0 2 at today’s level. However, the Both volcanism (associated with the molten rocks) and
modern oxygen cycle can at least elucidate some basic metamorphism (associated with hot rocks that do not
concepts of oxygen production and loss before we turn melt) release reduced gases into the atmosphere. Such
to the history of 0 2 . gases include H2, CO, S02, and H2S, which react with
0 2 , albeit through photochemical intermediates. Con-
sequently, the release of reductants from the solid
The Modern O2 Cycle Earth, either by continental uplift and weathering or
Oxygenic photosynthesis can be summarized by the by geothermal degassing, controls oxygen consump-
following schematic equation [I], where ‘CH20’ tion. Atmospheric 0 2 is not controlled, as is some-
represents the average stoichiometry of organic times believed, by respiration and decay. Measured in
matter.
-
moles, the amount of organic carbon in the biosphere
is IO2 times smaller than the atmospheric reservoir
of O2 and therefore respiration and decay can
modulate no more than 1 % of total amount of
Respiration and decay rapidly consume virtually all atmospheric 0 2 .
(99.9%) of the oxygen produced by photosynthesis,
and regenerate C 0 2 from organic carbon and 0 2 . But
photosynthesis does not completely reverse, because a Table 1 Modern oxygen fluxes in the Earth system
small fraction (0.1%) of organic carbon escapes ~ ~~~ ~~
oxidation through burial in sediments (see Table 1). Oxygen fluxes Magnitude Effect
From eqn [I], the burial of one mole of organic carbon (IO“ mol O2 y- ‘I
will generate one mole of 0 2 . However, the oxygen Organic carbon burial fluxa 10k 3 Production
cycle is complicated by the burial and weathering Pyrite (FeS2) burial fluxa 7.8+3.6 Production
fluxes of two other important redox elements, sulfur Sulfate burial fluxa - (0.3k0.1) Loss
Reduced iron burial fluxa 0.9k0.4 Production
and iron. Pyrite (FeS2) contains both elements in Continental oxidative - (15.5k6.7) LOSS
reduced form. During weathering, reduced minerals weathering flux’
exposed on the continents react with oxygen dissolved Flux of reduced volcanic and - (3k1) Loss
in rainwater. Pyrite oxidizes to form soluble sulfate metamorphic gasesa
(SO:-) and the chemically bound oxygen is then Net O2photosynthetic flux to -0 Net change
the atmosphere (assuming
carried to the oceans in rivers. Bacteria in the ocean use
that burial fluxes and
sulfate and ferric iron (Fe3’) to regenerate pyrite, and oxidative losses are
when pyrite is buried in sediments, oxygen is freed balanced by negative
again (eqn [II]). feedbacks)
Effective oxygen gain from 0.02 Absolute gain
hydrogen escape to space for whole
Earth
aDerivedfrom data in Holland (1978).

756 EVOLUTION OF ATMOSPHERIC OXYGEN
Over the last 350 million years, a continuous Because the oxidation of exposed continental surfaces
record of charcoal in continental sedimentary appears to be saturated, many researchers have
-
rocks suggests that 0 2 has always comprised at
least 15% of the atmosphere, because wood cannot
burn below this threshold. Yet the residence time
favored a negative feedback on the 0 2 source as a
primary control on the modern level of 0 2 . This
control is not thought to act directly, by oxidizing
of 0 2 in the atmosphere-ocean system is far shorter. newly produced organic carbon and preventing its
Dividing the atmosphere-ocean oxygen reservoir burial, but indirectly through phosphorus. Most
(3.78 x l O I 9 mol 0 2 ) (Table 2) by the source flux organic carbon is buried in the ocean on continental
(1.8 x 1013 mol 02 y - l ) gives -2.1 million years shelves. Phosphorus acts as a limiting nutrient for
for the average amount of time an 0 2 molecule marine photosynthesis because the only source of
spends in the atmosphere-ocean system. Evidently, 0 2 phosphorus is from continental weathering and river
levels must be controlled by long-term negative runoff. In the ocean, the burial of phosphorus bound
feedbacks. to iron hydroxides becomes less efficient under anoxic
In broad outline, oxygen is regulated because an conditions. Thus a decrease of oxygen increases the
increase in oxygen increases the consumption of amount of phosphorus available for the production
oxygen and/or decreases the rate of oxygen produc- and burial of new organic matter. However, this
tion. A decrease in oxygen has opposite effects. mechanism is not particularly effective against rising
levels of oxygen. Instead, some workers have hypoth-
esized that above 21 % oxygen the frequency of forest
fires will increase, triggering ecological shifts from
Table 2 Reduced and oxidized reservoirs in Earth's continental
crust.The Earth's exterior contains Fe203and SO,'-that arose via
forest to grassland. Vascular plants like trees amplify
oxidation, and free atmospheric 02.Oxidized species are ex- the rate of rock weathering by about an order of
pressed in terms of the moles of O2 required for their production; magnitude relative to simpler plant life. Rising oxygen
i
e.g., each mole of Fe3+ needed mole O2 to be produced from will thus throttle the supply of phosphorus to the
+
Fez+ via FeO io, = FezO3. Reduced species are expressed ocean, ultimately lowering the rate of organic carbon
in termsof O2moles requiredfortheirconsumption; e.g., each mole
of reduced carbon can be consumed by one mole of O2
burial, the oxygen source. However, while such
controls on modern oxygen concentrations are borne
Species and Magnitude Size (13) out by simple biogeochemical models, they are still
reservoir (1 d mol 0, cornparisone hypothetical.
equivalent)
Oxidized species
O2 in the atmosphere 0.0378 0.07 RAOS Oxygen in the Prebiotic Atmosphere
and ocean
Fe203,SO:-, and O2 in
In all likelihood, the atmosphere started out with an
0.55 RAOS
the atmosphere- oxygen partial pressure p 0 2 I 10 - l 3 bar (10 - Pa)
ocean-sedimentary before life existed because reduced gases overwhelmed
(AOS) system the abiotic source of 0 2 . In the absence of photosyn-
Total Fe3+ in the 1.7-2.6 (3.1- 4.7)R~os thesis, free 0 2 arises only from the photolysis of water
continental crustb
Total oxygen locked up 2.0-2.9 (1.5- 2.2)&dc
and the subsequent escape of hydrogen to space. By
in the continental itself, photolysis of H2O does not provide a net source
crust of oxygen because the photolysis products recombine.
Although Earth's atmosphere has changed over time, a
Reduced species 'cold-trap' for water generally exists at the (tropical)
Reduced carbon in the 0.56
AOS system
tropopause. Because water is cold-trapped to only a
Reduced carbon in felsic c0.78 few parts per million, the rate of production of oxygen
instrusives, from the photolysis of water and escape of hydrogen is
gneisses, schists and very small. The 0 2 so produced can react with
felsic granulites hydrogen through a series of photochemical reactions
Total reduced carbon in
that add up to a net reaction eqn [III].
c1.3 RredC
the continental crust,
RredC 0 2 + 2H2 + 2H20 [I111
aAOS, atmosphere-ocean-sedimentary system: redC, reduced With volcanic outgassing rates similar to those of
carbon in the continental crust.
bMost crustal oxidized iron, Fe3+,resides not in sedimentary rocks -
today, H2 fluxes would be lo2 times larger than the
abiotic 0 2 flux in the primitive atmosphere. If
but in continental basalt. Fe3+ derives from metamorphic or
hydrothermal oxidation processes within the crust. anything, outgassing rates were greater on early Earth
EVOLUTION OF ATMOSPHERIC OXYGEN 757
because of increased heat flow from a hotter, more was not leached. Ferrous iron (Fe2+) is soluble,
radioactive interior. whereas ferric iron (Fe3+)is not. Consequently, iron
will be flushed through a soil if rainwater has little
dissolved 0 2 but will be immobilized otherwise.
Geochemical Evidence for a Rise Another sign of abundant oxygen in soils is that
of O2 about 2.3 Ga cerium is fractionated from the other rare-earth
Once life arose, it must have modulated the cycling of elements. Cerium is oxidized from Ce3+ to Ce4+ to
all gases in the atmosphere-ocean system containing form cerianite (Ce02). The presence of Ce3+-rich
chemical elements of biological importance. The most minerals in paleosols from the Archean eon implies an
dramatic atmospheric change evident in the rock early anoxic atmosphere. Other evidence for ancient
record is an increase in the level of 0 2 around 2.3 Ga anoxic air includes detrital grains from Archean
(Figure 1).This event occurred in the Paleoproterozoic riverbeds, which commonly contain reduced minerals
era (2.5-1.6 Ga), the first of three eras within the that would only survive at low p02. Detrital grains of
Proterozoic eon (2.5-0.57 Ga), which follows the pyrite (FeSZ), uraninite (U02), and siderite (FeC03)
Archean eon (before 2.5 Ga). place upper bounds on Archean pO2 of roughly 0.1
atm, 0.01 atm and 0.001 atm, respectively. In oxic
waters, uraninite dissolves to form soluble U6+ ions,
Evidence from Continental Environments
pyrite oxidizes to sulfate (SO:-) and ferric iron
Paleosols, detrital grains, and red beds, all from (Fe3+), and siderite oxidizes to produce ferric
continental environments, suggest very low levels of iron (Fe3+). Red beds provide further evidence for
0 2 before about 2.3 Ga. Paleosols are ancient surfaces atmospheric redox change. They derive from wind-
that were exposed to the atmosphere during weather- blown dust or river-transported grains coated with
ing. Geochemical studies of paleosols indicate that red-colored hematite (Fe2O3). Before -2.3 Ga, red
around 2.4-2.2 Gay atmospheric p02 rose from beds are very rare, whereas afterwards red beds
<0.0008 atm to >0.002 atm, possibly to >0.03 are ubiquitous. Pre-2.3 Ga red beds result from
atm. Paleosols before -2.4 Ga show that iron was ground water contamination by the post-2.3 Ga
leached during weathering, but after -2.2 Ga, iron atmosphere.
-1 -1
-2 -2
-3 r -3
-4 F -4
-5
I -5
,,** 3T-6
r-13
-14 9)' I ' I ' I I' 1 ' I '
' 155
-5F -14
4.4 3.2 3.0 2.8 2.6 2.4 2.2 2.0 1.8 1.1 0.8 0.6 0.4 0.2 0.0
Time before present (Ga)
Figure 1 The history of atmospheric oxygen. The thick dashed line shows a possible evolutionary path for atmospheric oxygen that
satisfiesgeochemical and biologicalconstraints. Dotted horizontal lines show the duration of geochemical and biologicalconstraints, such
as the occurrence of detrital siderite (FeCO,) in ancient riverbeds. Downward-pointing arrows indicate upper bounds on the level of
oxygen, whereas upward-pointing arrows indicate lower bounds. Unlabeled solid horizontal lines indicate the occurrence of specific
paleosols, with the length of each line showing the uncertainty in the age of each paleosol. Bounds on popfrom paleosols are taken from
Rye and Holland (1998). Biological lower limits on popare based on estimatesfor the requirementsof the marine sulfur-oxidizingbacteria
Beggiafoa,and also the requirements of macroscopic animals that appear around 0.59 Ma. An upper bound on the level of p o p in the
prebiotic atmosphere at c. 4.4Ga (shortly after the Earth had differentiated into a core, mantle, and crust) is based on photochemical
calculations. Similarly, aconstraint of bar ( ~ 0 .Pa)
1 before 2.4Ga is suggested by mass-independentsulfur isotope evidence when
constrained by photochemical models of isotopic exchange.
Evidence from Iron Formations geological time, 20% of the carbon in C02 entering
the ocean-atmosphere system has exited as buried
An atmospheric 0 2 increase at -2.3. Ga is consistent
organic carbon, whereas the remaining 80 % has
with the temporal distribution of banded iron forma-
exited as carbonate.
tions (BIFs). BIFs are laminated marine sedimentary
The largest excursions in 613Ccarb in Earth's history
rocks containing 2 15 wt% iron, usually with alter-
occur between 2.4 and 2.1 Ga, with positive and
nating iron-rich and silica-rich layers. The iron, in
negative oscillations between + 1O%o and - 5%0.
large part, originated from hydrothermal sources in
Three low-latitude glacial, or 'Snowball Earth', epi-
the deep ocean, such as mid-ocean ridges. Today, iron
sodes during 2.4-2.2 Ga together with sparse data
is oxidized immediately and deposited on the flanks
complicate the interpretation of 613Ccarb fluctuations.
of the ridges. In the anoxic Archean oceans, ferrous
Clearly, major perturbations in the carbon cycle and
iron circulated to the continental shelves where it
climate system occurred. The rise of 0 2 would have
was (microbially) oxidized and precipitated. In the
destroyed greenhouse gases like methane, inducing
disappear after -
Paleo-proterozoic, BIFs decline in abundance and
1.8 Ga. This has usually been
attributed to a rise in atmospheric oxygen ventilating
global cooling and decimating any early land biota.
Oxygen would have been catastrophic to some anaer-
obic organisms, for which 0 2 is toxic. In combination
the deep ocean. However, it is also plausible that the
with inducing significant oxidative weathering, rising
deep ocean remained anoxic until the late Proterozoic
oxygen would have dramatically affected the carbon
and that sulfide was responsible for removing iron
cycle.
from the deep ocean. Increased oxidative weathering
Marine sulfur isotopes indicate an increase in
would have led to greater sulfate concentrations
sulfate concentrations at -2.3 Ga consistent with a
in the ocean, promoting microbial sulfate reduc-
rise of 0 2 . Today, sulfate-reducing bacteria (SRB)
tion. Because sulfate reduction produces sulfide
produce most of the sulfide in marine sediments. SRB
( S 2 - ) , the deep ocean could have been swept
reduce 32SO:- in preference to 32SO$- so that the
free of iron by the precipitation of insoluble
sulfide produced is depleted in 34S. However, this
pyrite, FeS2.
fractionation ceases in waters with sulfate concentra-
Evidence from Carbon and Sulfur Isotopes tion < 1mmol 1 - I. Archean sulfides display 34S/32S
in Marine Sediments ratios that cluster around the unfractionated mantle
value, implying Archean oceans with < 1mmol 1-
Marine carbon isotopes indicate drastic environmen-
sulfate, compared to 28.9 mmoll-' in today's surface
tal change in the Paleoproterozoic, consistent with the
sea water. Lack of sulfate is consistent with low 902,
rise of 0 2 . Of the two stable carbon isotopes, 12Cand
which would induce little oxidative weathering of
13C,photosynthetic organisms kinetically concentrate
sulfides, limiting the river supply of sulfate to the
12C into organic matter, leaving inorganic carbonate
oceans. By 2.3-2.2 Ga, sulfides with significant 34S
relatively enriched in I3C. Isotope compositions are
depletions formed ubiquitously in the ocean, reflecting
ex ressed as 613C, where 613C = [('3C/'2C),a,,,,/
F
(l C/12C)standard - 1 1x 1000, in parts per thousand
abundant sulfate and the rise of 0 2 .
Measurements of a third less-abundant sulfur iso-
(%o). From -3.5 Ga, with only a few relatively brief
tope, 33S, in addition to 34S and 32S, show a major
variations, sedimentary organic carbon is found to be
change in the sulfur cycle between 2.4 and 2.1 Ga. In
about 30%0 ( 3 % ) lighter than marine carbonate
modern rocks, 32S, 33S, and 34S obey 'mass-dependent'
carbon that has S13C M O%O. This mainly reflects
fractionation, in which the difference in abundance
biological fractionation. Carbon entering the atmos-
between 33S and 32S is approximately half that
phere-ocean system from volcanism, metamorphism,
between 34S and 32S. Many aqueous chemical and
and weathering has 613CinM -6%0. On time scales
biochemical reactions, such as microbial sulfate
greater than the residence time of carbon in the ocean
reduction, produce mass-dependent isotope fractiona-
( - lo5 years), the same number of atoms entering the
tion. In contrast, the full suite of stable sulfur isotopes
atmosphere-ocean system must exit the system, im-
in pre-2.3 Ga rocks show 'mass-independent' frac-
plying the following balance shown in eqn [l].
tionation, which is thought to result exclusively from
+
6l3Cin = fcarb6l3Ccarb f0rg6'~Corg 111
photochemical reactions such as photolysis of S 0 2 . In
this case, the relative abundance of different isotopes
Here fcarbis the fraction of carbon buried in carbonate deviates from what is expected of mass-dependent
minerals with isotopic composition 613Ccarb,and forg is fractionation. In a high-02 atmosphere, sulfur gases
the fraction of carbon buried in organic carbon with are rapidly oxidized and rain out as dissolved sulfate.
isotopic composition 613Corg.Solving eqn [ 11with the But in the absence of 0 2 , sulfur exits the atmosphere as
observed 613C values gives forg M 0.2. Thus, over sulfide, elemental sulfur, or sulfate, which allows the
mass-independent fractionation produced by anoxic Australia, providing unambiguous evidence of oxy-

photochemistry to be transferred to sediments. genic photosynthesis at 2.7 Ga. Stromatolites, which
are laminated mineral structures produced by micro-
A Second Rise of O2 in the bial growth, provide further evidence. These are
present in 2.72 Ga paleolakes that had no hydrother-
Neoproterozoic Era? mal sources of reductant, implying oxygenic photo-
Some researchers argue that there was a second rise of synthesis as the only means of organic synthesis.
p02 in the Neoproterozoic era (1.0-0.57 Ga) from Isotopic fractionation of carbon between sedimentary
about 1-3 YOto greater than 5-1 8% of present levels. organic carbon and inorganic carbonates is about
From -1.05 to 0.64 Ga, sulfides are increasingly 30%0back to 3.5 Ga and characteristic of photosyn-
found with 34S-depletionsexceeding the fractionation thetic fractionation, which suggests an earlier origin
threshold of sulfate-reducing bacteria. This can be for oxygenic photosynthesis but is not, on its own,
explained if sulfide was reoxidized at the sediment- conclusive.
water interface to SO;- and reduced again by
bacteria, cyclically increasing the isotope fraction-
Theories for the PaleoproterozoicRise of O2
ation. Possibly, 0 2 increased to the point where it
penetrated marine sediments, making the deep ocean Once oxygenic photosynthesizers evolved, they very
aerobic for the first time. This could corroborate an likely dominated the biosphere. In their absence, all
increase of oxygen thought necessary to explain the other carbon-fixing organisms required reductants
appearance of macroscopic animals in the fossil record supplied by geothermal processes, such as H2, to
around 590 Ma. reduce C 0 2 to organic carbon, restricting habitats to
places like hydrothermal vents. In contrast, the source
of hydrogen for oxygenic photosynthesis is ubiquitous
Phanerozoic Changes in p02
H20. Why then, was there a delay of 0.3-1 billion
In the Phanerozoic eon ( 0 . 5 7 Ga to present), p 0 2 has years between the earliest oxygenic photosynthesis
probably always been in the range 0.15-0.35 bar (1 and the rise of 0 2 ?
bar l o 5 Pa), After the land was colonized by plants One explanation is that as geothermal heat declined
around -420 Ma, large amounts of organic carbon owing to the decay of radioactive materials inside the
were buried on the continents in the Carboniferous Earth, the flux of volcanic gases dwindled, lessening
(362-290 Ma). Lignin, a structural compound in trees, the sink on 0 2 . However, increased past volcanic
was difficult to decompose until organisms like fungi outgassing would have also injected proportionately
evolved means of doing so. Because organic carbon more CO2. Carbon isotopes from 3.5 Ga onward
burial is accompanied by a stoichiometric release of show that roughly 20% of the C 0 2 flux into the
- -
0 2 (eqn [I]), Carboniferous organic burial may have biosphere was fixed biologically and buried as organic
led to peak e02 0.3 bar at 300 Ma. This would carbon, with the remainder buried as carbonate.
explain the presence of giant Carboniferous insects, Increased outgassing in the past, on its own, cannot
like dragonflies with 0.7m wingspan, which rely on explain the oxic transition because, as one goes back in
the diffusion of 0 2 for respiration. time, 0 2 production due to organic burial would have
risen in parallel with 0 2 losses. A second hypothesis
Biogeochemical Change: Explaining suggests that large positive carbonate isotope excur-
sions from 2.4 to 2.1 Ga were due to a massive pulse of
the History of Oxygen organic burial that caused the rise of 0 2 . However,
There are several lines of evidence suggesting that given the geologically short residence time of 0 2 , a
oxygenic photosynthesis pre-dates the rise of 0 2 . pulse of organic burial would merely cause atmos-
Biomarkers are organic molecules that are diagnostic pheric 0 2 to rise and decay, not to remain in the
of the organisms from which they were derived. Only atmosphere. A third explanation of the rise of 0 2
cyanobacteria are known to synthesize 2-methyl invokes a gradual shift of volcanic gases from reduced
bacteriohopanepolyols (specific five-carbon-ring to oxidized. This hypothesis is promising, because a
compounds), which are transformed in sediments to secular change would make the atmosphere more
2-methyl hopanes. Similarly, only eukaryotes (single- conducive to higher 0 2 levels. However, data from
celled organisms with cell nuclei) synthesize certain redox-sensitive elements in igneous rocks show that
sterols (four-carbon-ring alcohols) in a process that the mantle's oxidation state, which controls the redox
requires molecular oxygen. Steranes, derived from state of volcanic gases, only permits an increase in H2
sterols, and 2-methyl hopane biomarkers have both relative to C 0 2 by a factor 11.8, which cannot
been found in ancient sedimentary rocks from western account for a sufficient change in the sink on 0 2 .
760 EVOLUTION OF ATMOSPHERIC OXYGEN
The rise of 0 2 can be considered more generally. The escape of hydrogen to space, at - l o 2 times today’s
rate of change of the number of moles of 0 2 , Roz, in escape rate. When hydrogen escapes to space, the
the atmospheric reservoir is given by eqn [2]. Earth as a whole is oxidized. The severe depletion of
noble gases in the composition of the Earth compared
to solar composition shows that the Earth did not
retain gaseous volatiles from the original solar nebula
= Fsource - (Fvolcanic +Frnetamorphic +Fweathering) when it formed. Hydrogen was originally incorporat-
ed into the Earth in an oxidized, solid form such as in
PI ice (H20) or water of hydration ( - OH). If hydrogen
Fsink is the removal flux of 0 2 from the atmosphere (in is lost from such solids, the matter left behind is
moles y - due to numerous oxidation reactions and oxidized. Thus, when hydrogen originates from gases
FSOUrCe is the source flux of oxygen (in moles y - ’)due in the crust (or mantle) and escapes to space, the crust
to burial of organic carbon and pyrite. The sink fluxes (or mantle) is irreversibly oxidized.
are the reaction of 0 2 with reduced volcanic gases Consequently, a plausible explanation for the 2.3
(Fvolcanic ) 9 reduced metamorphic gases (Fmetamorphic 9 Ga rise of 0 2 is that excess reductants scavenged 0 2 in
and reduced material on the continents (Fweathering). the Archean. Hydrogen escape, promoted by methane,
During any particular epoch, Ro, will be in ‘steady then oxidized the Earth, lowering the sink on 0 2 from
state’, which means that 0 2 will have accumulated in gaseous emissions until an oxic transition occurred.
the atmosphere to some value of Roz where the 0 2 sink Explaining a second, Neoproterozoic rise of 0 2 in the
(Fsink) will be equal to the 0 2 source (FSOUrCe).In this same way is problematic, unless methane persisted
case, d(Ro,)/dt = 0. Such a balance of source and throughout the Proterozoic at a moderate level of
sinks is how 0 2 remains stable today. 0 2 levels have order 10 parts per million by volume and shifted a
evolved because the terms on the right-hand side of redox buffer beyond another critical threshold. How-
eqn [2] have altered over Earth history. The sink due to ever, details are still obscure.
oxidative weathering was apparently smaller in the A large uncertainty is the time-integrated export of
Archean, but the source due to organic burial appears reduced and oxidized material from the crust to the
to have been relatively constant on the basis of carbon mantle. Net export of reduced material to the mantle
isotopes. This evidence implies that the sink from could have perhaps oxidized the crust, essentially with
reduced gases was greater in the Archean relative to the same effect as hydrogen escape to space. However,
the source of 0 2 . For 0 2 to rise, this sink must have we know little about how much reduced material has
diminished. been subducted versus how much oxidized material
Unlike volcanic gases, reduced metamorphic gases over Earth history.
provide an oxygen sink that is not determined by the
oxidation state of the mantle and so could have
changed greatly. The metamorphic recycling of more The Emergence of the Ozone Layer
reduced Archean crust would have produced more The Paleoproterozoic rise in oxygen would have
reducing volatiles than the recycling of today’s more created an ozone layer, shielding surface life from
oxidized crust, as a matter of redox conservation. That ultraviolet (W)radiation. UV radiation with wave-
the early crust was more reduced than today is length below about 300 nm biologically harmful.
suggested by evidence showing no oxidized surfaces.
-
A decrease by a factor of 3 in the H2/C02 ratio of
volatile fluxes is all that is required to flip the
Radiation below about 200 nm wavelength is strongly
absorbed by C02 and 0 2 , whereas ozone shields
biologically harmful radiation in 200-300 nm range.
atmosphere from a state dominated in redox terms Photochemical models show that harmful UV is
by hydrogen-bearing species like CH4 to an 02-rich mostly absorbed with an ozone layer that would
state. Crustal redox change, perhaps in combination form with O2 levels about 1-3% of present, similar to
with a small change in mantle redox, could plausibly those after 2.3 Ga.
have effected such a change in the redox state of
gaseous emissions. But why would the crust oxidize?
Just a small excess of hydrogen would have tipped
the redox balance of the atmosphere to an anoxic,
Summary
’hydrogen-rich’state. For thermodynamic reasons, the The most significant biological event in the history of
biosphere is prone to convert metabolically desirable the Earth’s atmosphere was the evolution of oxygenic
hydrogen to CH4, which would accumulate to levels photosynthesis. However, there was a long delay
102-103 times present values. The photolysis of between the appearance of oxygenic photosynthesis
methane in the stratosphere would promote rapid before 2.7 Gay possibly as early as 3.5 Gay and the
oxygenation of the atmosphere at 2.3 Ga. This delay Proterozoic An Eon of Earth history, 2.5 to 0.57 Ga. It
reflects a secular change in the oxygen sink. The long- is composed of three eras, the Paleoproterozoic
term consumption of oxygen is ultimately controlled (2.5-1.6 Ga), the Mesoproterozoic (1.6-1.0 Ga),
by the release of reductants from the Earth’s crust and and the Neoproterozoic (1.0-0.57 Ga).
mantle. A plausible explanation for the 2.3 Ga rise of Red bed Sediment or sedimentary rock that is red or
0 2 is that excess reductants scavenged 0 2 in the early reddish-brown in color owing to the presence of
atmosphere. Such conditions would have stabilized ferric oxide materials usually coating individual
abundant biogenic methane in the atmosphere. Meth- grains.
ane photolysis would have led to significant escape of Reduction A process or environment (reducing envi-
hydrogen to space, oxidizing the Earth’s lithosphere ronment) in which a chemical element or ion gains
and lowering the sink on free 0 2 until an oxic electrons and is changed from a higher to a lower
transition occurred. A second rise of O2 at about valence state.
1.0-0.6 Ga is suggested by increased sulfate concen- Snowball Earth An event in Earth history when low-
trations in the ocean, although the cause remains latitude glaciation occurred and the whole Earth
obscure. The progressive increase of atmospheric may have been covered in ice.
oxygen levels set the stage for multicellular life to Stromatolite A lithified, commonly laminated, sedi-
develop. Ultimately, we owe our own existence to the mentary structure produced as a result of the growth
growth in atmospheric oxygen levels. and metabolic activities of aquatic, bottom-dwell-
ing communities of microorganisms.
Sulfate reduction A process used by some bacteria to
derive energy by reducing sulfate ions (SO$-) to
Glossary H2S.
Archean An Eon of Earth history ending at 2.5 Ga.
The start of the Archean is now generally taken as
when Earth formed -4.5 Ga, although some liter- See also
ature takes the start as 3.8 Ga or 4.0 Ga, where the
Archean is preceded by a ‘Hadean’ Eon. Biogeochemical Cycles: Carbon Cycle; Sulfur Cycle.
Cyanobacteria Single-celled bacteria containing chlo- Chemistry of the Atmosphere: Principles of Chemical
rophyll- a and capable of oxygen-producing photo- Change. Evolution of Earth’s Atmosphere. Gaia Hy-
pothesis.Ozone: Ozone Depletion Potentials; Ozone as
synthesis.
a UV Filter; Role in Climate; Surface Ozone Effects on
Ga l o 9 years ago. Vegetation. Tropospheric Chemistry and Composi-
Hydrogen escape The process in which hydrogen tion: HP. Volcanoes: Composition of Emissions.
atoms in the exosphere exceed the escape velocity of
the Earth and escape into space.
M a lo6 years ago.
Oxidation state The degree of oxidation of an atom, Further Reading
molecule, compound, or quantity of matter. Sub- Gregor CB, Garrels RM, MacKenzie FT and Maynard JB
stances with a low oxidation state have a large (eds) (1988) Chemical Cycles in the Evolution of the
number of available electrons, whereas substances Earth. New York: Wiley.
with a high oxidation state do not. Falkowski PG and Raven JA (1997)Aquatic Photosynthesis.
Oxidizing A process or environment (oxidizing Malden, MA: Blackwell Science.
environment) in which a chemical element or ion Holland HD (1978)The Chemistry of the Atmosphere and
Oceans. New York: Wiley.
loses electrons and is changed from a lower to a
Holland HD (1984)The Chemical Evolution of the Atmos-
higher valence state. pheve and Oceans. Princeton, NJ: Princeton University
Oxygenic photosynthesis The process in which an Press.
organism uses sunlight, water, and carbon dioxide Kump LR, Kasting JF and Crane RG (1999) The Earth
to synthesize organic matter, releasing 0 2 as a waste System. Upper Saddle River, NJ: Prentice-Hall.
product. Rye R and Holland HD (1998)Paleosols and the evolution of
Paleosol A buried, commonly lithified, soil horizon of atmospheric oxygen: a critical review. American Journal
the geological past. of Science 298: 621-672.
Phanerozoic An Eon of Earth history, 0.57 Ga to Walker JCG (1977) Evolution of the Atmosphere. New
present. York: Macmillan.
762 EVOLUTION OF EARTH’S ATMOSPHERE
E J Gaidos and Y L Yung,California Institute of the former include atmospheric photochemistry, vol-
Technology, Pasadena, CA, USA canism, and plate tectonics. The latter include the
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. accretion of new material (impacts of comets or
meteorites), escape of hydrogen to space, and the
sequestration of certain elements (siderophiles) into
The Diversity of Planetary Atmosphere the metallic core.
Although the planets Mercury, Venus, Earth, and Earth’s atmosphere has been profoundly affected by
Mars have masses within a single order of magnitude another process: life. The modern atmosphere, con-
range, they possess atmospheres with extremely taining abundant oxygen in gross chemical disequi-
different properties (Table 1). These bodies may librium with surface organic carbon and gases such as
have initially possessed primordial atmospheres of methane, is testament to life’s ability to efficiently
solar composition whose dominant light gases (hy- convert light energy into chemical energy, some of
drogen and helium) were lost to space and replaced by which is stored in the chemical disequilibrium between
outgassed water, carbon dioxide, and nitrogen (and the atmosphere and surface. Significant disequilibri-
small amounts of other gases) during the final phase of um is not present on the sterile worlds of Venus and
accretion 4.5 billion years (Ga)ago. The divergence in Mars, and it has been suggested that the simultaneous
atmospheric composition seen today may in part presence of pairs of gases like 0 2 and CH4 in an
reflect differences in initial volatile abundance, but atmosphere may serve as a planetary ‘biosignature’
much of the diversity can be attributed to the individ- that reveals the presence of abundant life even at a
ual evolutionary paths of these atmospheres over the distance. Some gases such as C02 ,the principal source
age of the Solar System. Rates of planetary atmos- of biological reduced carbon, are maintained at
pheric evolution have differed markedly: whereas the mixing ratios much lower than the level predicted in
other planets have suffered catastrophic atmospheric the absence of life. The current terrestrial atmosphere
evolution (Mercury has experienced complete loss, is far from the end state reached by Venus, where all of
Venus a runaway greenhouse and devolatilization of the surface volatiles are in the atmosphere (Table 2).
surface rocks, and Mars has lost most of the atmos- Also in contrast to neighboring planets, the terres-
phere to space or the crust), the evolution of Earth’s trial atmosphere maintains conditions suitable for life
atmosphere has been comparatively mild. Both exter- (providing a modest greenhouse effect and a shield
nal processes, such as radiation and the corpuscular against biologically harmful radiation), and has ap-
wind from the Sun and impacts, and internal process- parently done so for 3.5 billion years, despite a 40%
es, such as volcanism and recycling of a planet’s crust increase in solar luminosity, giant impacts, and the
(e.g., plate tectonics) control this evolution. While
some processes drive exchange of compounds between
the atmosphere and reservoirs in the surface, oceans, Table2 The reservoirsof themajorvolatileson the Earth (mantle
or interiors of planets, or the interconversion of quantities are uncertain)
different chemical species, others result in the secular,
Reservoir Size (hPa) Climate role
irreversible evolution of the atmosphere. Examples of
H20 Atmosphere <0.001 Greenhouse gas,
Ocean 26 carbonate sink,
Hydrated crust 10 weathering, biology,
Table 1 The atmospheres of the inner planets Mercury, Venus,
Mantle -100 plate tectonics
Earth, and Mars
Total 136
Planet Mass [Earths] Pressure (hPa) Composition
COP Atmosphere 0.00003 Greenhouse gas
Mercury 0.053 10-~ H, He, Na Ocean 0.0002
Venus
Earth
0.817
1
90 000
1 000
96% CO2, 3% N2,
minor CO, SO2
78% N2, 21Yo 0 2 ,
Carbonate rocks
Mantle
Total
- 4
26
30
1YoAr, minor
Hz0, COP Nz Atmosphere 0.078 Buffer gas, enhances
Mars 0.107 7 95% Cop, 3% N2, Crustal rocks 0.025 effect of greenhouse
2% Ar, minor CO, Mantle -0.1 gases
0 2 Total 0.2
EVOLUTIONOF EARTH’S ATMOSPHERE 763
changing tempo of plate tectonics. This fact is even photochemistry and other energetic processes, also
more remarkable in light of the relatively short operate. All these processes create energetic particles
residence time of most gases: water vapor, the most by photolytic, electron impact, or ionic reactions. In
powerful greenhouse gas in the atmosphere, has a the absence of a magnetic field, direct momentum
residence time of only 10 days before exchanging with transfer may result in knocking atoms from the
the oceans. Even relatively inert nitrogen gas, the exosphere, in a process known as ‘sputtering’. Ions
principal component of the atmosphere, is recycled by may also migrate to the polar regions, where the
biological fixation (reduction), oxidation, and rere- magnetic field lines are open to the tail of the
duction to N2 on a time scale of 10-15 million years. magnetosphere, and they can readily escape along
Thus Earth’s atmosphere is not only out of chemical these field lines. In charge exchange, H atoms lose an
equilibrium, but is dynamically maintained, having no electron to a heavier ion, typically O+, and the
intrinsic buffering capacity on geologic time scales. resulting proton is accelerated by the electric field of
the solar wind to exceed the escape velocity. The
ionization potentials of 0 and H differ by only 0.02 eV.
This difference is smaller than the thermal energies
Evolutionary Processes corresponding to typical upper-atmosphere tempera-
The early atmosphere would have been subject to tures, and as a consequence the charge transfer process
frequent bombardment by planetesimals left over is very efficient.
from early planetary accretion. These impacts may Hydrogen loss from the exosphere is very efficient,
have brought in volatiles, but they may also have but the actual rate (3.0 x 108atomscm-2s-1) of
removed (to space) some of the existing atmosphere escape from the modern oxygen-containing atmos-
either by the momentum of the impact shock wave or phere is limited by the transport of hydrogen from
heating of the upper atmosphere (see below). Impacts deeper in the atmosphere. Hunten’s limiting flux
also turn over the crust and expose fresh surfaces, theorem relates the maximum escape flux of hydrogen
accelerating chemical interactions between the atmos- to the abundance of total hydrogen in the mesosphere,
phere and crustal rocks. Giant impact basins on the determined largely by the chemical sources and
Moon date to 3.9 billion years in age, indicating that a transport. The escape rates of hydrogen in the past
similar bombardment shaped the Earth’s atmosphere were probably much higher than the present rate. The
for the first several hundred million years of its history. primary reason is that the early atmosphere probably
A second process that may have profoundly (perhaps lacked oxygen: on the modern Earth, hydrogen is
catastrophically) affected Earth’s atmosphere was the ‘trapped’ by recombination with oxygen to form water
formation of the iron-nickel core. This removal of in the lower atmosphere before it can reach the upper
metallic iron (and possible fractions of iron-soluble atmosphere and escape. A second reason is that the
elements like sulfur, carbon, and hydrogen) left the Sun was more active in the past, providing a higher flux
Earth’s surface and mantle in a considerably more of EUV radiation to power hydrogen escape. The flow
oxidized chemical state, a condition that would have of hydrogen may have been sufficiently massive as to
been reflected in the oxidation state of the compounds cause the hydrodynamic escape of heavier gases by
that were being outgassed by volcanoes. Thus, as is the momentum transfer.
case today, C 0 2 , N2, and SO2 were the dominant Biology contributes to atmospheric evolution by
components of volcanic gases, rather than CH4, HZS, modulating the interconversion of compounds and the
and NH3. exchange of compounds between the atmosphere and
Thermal escape of light atoms such as hydrogen and other reservoirs. The paramount example of this is
helium can occur from the exosphere (the uppermost oxygenic photosynthesis, which in effect partitions
layer of the atmosphere where the mean free path C 0 2 into organic carbon and oxygen:
between collisions exceeds the scale height). Escape is
efficient if the mean thermal speed is a significant +
C02 H20 + CHzO(organic compounds) + 0 2
fraction of the escape velocity such that a non-
negligible number of molecules in the ‘tail’ of the If the organic carbon is prevented from being
Maxwellian (thermal) distribution have speeds above immediately reoxidized (e.g., by combustion or by
the escape velocity. Extreme and vacuum ultraviolet aerobic respiration in other organisms) then the
radiation from the Sun (mostly in the 121.4 nm Lyman oxygen can enter the atmosphere. At some point in
ct line of hydrogen) is absorbed and converted into heat the past photosynthetic production was able to over-
in the Earth’s thermosphere. The modern thermo- whelm sinks of oxygen and the atmospheric mixing
sphere lies above 400 km and has a temperature of ratio of O2 increased dramatically, an event that would
1200 K. Non-thermal escape mechanisms, driven by never have occurred without the intervention of
764 EVOLUTION OF EARTH’S ATMOSPHERE
biology. Biology is also responsible for repartitioning 300

carbon dioxide between the atmosphere, oceans, and
limestone rocks (kinetically favoring the last) by 0 CI Chondrites
accelerating the process of silicate weathering and d Mars (EETA 79001)
200
carbonate precipitation:
CaSi03 + COZ + CaC03 + Si02
100
Without marine organisms, carbon dioxide levels h
8
would be much higher than current values: it has been v
suggested that the successive development of more E?

G o
sophisticated marine and terrestrial organisms (e.g., 0

plants with vascular roots) has led to a secular decrease
in atmospheric COZ over time. A third process where
life has played an important role is to recycle oxides of
-1 00
nitrogen (specificallynitrates) back to the atmosphere
by denitrification, e.g.:
4HN03(aqueous) + .SCH~O(organics)
1 2 4 ~ 1 ~ 2 6 ~ 1 ~2 8 ~ 1 ~ 3 0 ~1 ~3 2 ~ 1~ 3 4 ~ 1~ 3 6 ~ ~
Mass, m (amu)
Figure I Evidence for loss of an early terrestrial atmosphere:

isotopic composition of xenon in meteorites and in planetary
In the absence of biology, the formation of nitrates atmospheres, plotted relative to terrestrial atmospheric xenon.
by UV and lightning, and their rain out from the Units are parts per thousand deviation from the terrestrial compo-
atmosphere, would have been unchecked, removing sition. The atmospheres of Earth and Mars are depleted in the light
isotopes relative to meteorites, which are thought to reflect the
most N2 (and thus most of the atmosphere!) in about composition of the primordial nebula. The lighter isotopes may
one billion years. have been carried away during the early escape of a hydrogen
Aerosols are a minor component of the terrestrial atmosphere. (After Pepin RO (1989) Atmospheric composition:
atmosphere whose importance to climate is only now key similarities and differences. In Atreya SK, Pollack JB, and
being more fully understood. Aerosols are microme- Matthews MS (eds) Origin and Evolution of Planetary and Satellite
Atmospheres. Tucson, AZ: University of Arizona Press. Reprinted
ter-sized particles of sulfur, organic compounds, and
with permission.)
salts that remain suspended in the atmosphere for
years. They scatter sunlight, effectively cooling the
surface, and serve as cloud condensation nuclei, i.e.,
sites of cloud water droplet formation. Clouds can nation is that these gases were components of a
either cool (by reflecting sunlight back to space) or primordial atmosphere that escaped to space early in
warm (by blocking the escape of infrared radiation). Earth history, perhaps in a hydrogen ‘wind’ driven by
The aerosol content of the atmosphere may have intense ultraviolet radiation from the young Sun or by
evolved because of changes in ocean salinity, the the giant Moon-forming impact. (Terrestrial 40Ar, in
surface area of the continents, and biological produc- contrast to its sister isotopes 36Arand 38Ar,is primarily
tion of sulfur compounds and organic compounds. a product of the radioactive decay of 40K and therefore
an exception.) Models of this escape also predict the
loss of essentially all N- and C-containing compounds
The Record of Atmospheric Evolution from such an atmosphere and imply that modern N
Because most of the atmosphere is recycled on a and C have a separate, later origin, from a long-term
geologically short time scale, it has no ‘memory’ of reservoir in the Earth’s interior or else from a late
anything but the most recent events in Earth history. ‘veneer’ of extraterrestrial material.
The exceptions to this rule are the noble gases Ne, Ar, Our knowledge of the evolution of the other,
Kr, and Xe, elements that are chemically inert and too reactive gases is obtained by examination of the
heavy to escape easily from the Earth. The rare gases in chemical and physical effects the paleoatmosphere
the terrestrial atmosphere are depleted with respect to has had on materials preserved in the rock record. The
other elements compared with their expected values in best record, and the one that has received the most
the primordial nebula that formed the planets. They attention, is that for atmospheric oxygen. Much of this
also exhibit strong elemental and isotopic fraction- work is based on the disparate aqueous solubility of
ation, such that heavier elements and isotopes are compounds of certain elements, particularly Fe, Mn,
more abundant (Figure 1).The most plausible expla- and U, in the presence or absence of atmospheric
EVOLUTION OF EARTH’S ATMOSPHERE 765
oxygen. Under anoxic conditions, iron is in its highly of an oxidation event, the immobility of Fe could be
soluble ferrous state, but insoluble ferric oxides and the indirect result of a rise in atmospheric oxygen and
hydroxides form in the presence of even small amounts the initiation of a ‘modern’ sulfur cycle, in which
of oxygen. Sedimentary units formed in the Archean oxidative weathering of sulfur-bearing minerals on the
(3.8-2.5 Ga) and early Proterozoic (2.5-0.54 Ga) are continents leads to transport of sulfate to the oceans.
characterized by banded iron formations, finely lam- Sulfate reduction (by bacteria) would produce elevat-
inated deposits of iron oxides and carbonates. The ed concentrations of sulfide in deep water and precip-
distribution of these deposits suggests that iron was itation of Fe as pyrite. Sulfur from ancient marine
relatively mobile and hence oxygen was absent in the sediments and evaporite deposits also records a shift
deep ocean during these epochs. Whether oxygen was in the distribution of relative abundance of the two
present in shallower water is not clear: Iron is major isotopes 32S and 34S, suggesting an increase
conspicuously rare in carbonate formations such as in sulfate reduction by bacteria. In Archean rocks there
stromatolites from this time. Conversely, iron concen- is also evidence for mass-independent fractiona-
tration in Archean paleosols (ancient soils) is low, tion among a three-sulfur isotope system that
indicating that Fe remained mobile (and hence remov- includes 33S (Figure 2). This fractionation pattern is
able) during the weathering process. If these soils were unlike those produced by the oxidation of sulfur by
in chemical equilibrium with the atmosphere then the weathering or the reduction of sulfate by biology,
stability of the ferrous state constrained pol levels to where fractionation is proportional to the mass
< l o p 4 of the present atmospheric level (PAL). differences between the isotopes. Mass-independent
Oxidized soils and deposits (redbeds) in which the fractionation could indicate that in the absence of
(ferric) iron is retained do not appear in the geologic atmospheric oxygen during the Archean, certain gas-
record until after 2.4 Ga. phase reactions were significant contributors to the
Reduced forms of mineral sulfur (pyrite) and sulfur cycle.
uranium (uraninite) are insoluble and hence resistant Carbon dioxide is a minor component (preindus-
to removal by aqueous process. These minerals are trial pco, = 280 ppm) of the modern atmosphere, but
seen in Archean deposits, but not in later Proterozoic the principal incondensable greenhouse gas. Since the
ones, again indicating an increase in atmospheric P O , .
;lo I
The ratio of thorium to uranium in mantle-derived
rocks shows a marked increase from samples older
than 2 Ga to younger samples, which can be explained e
e
if recycling of U into the mantle has become less
efficient owing to increased retention of (soluble) U in 1.5
an oxidizing hydrosphere. Manganese is oxidized only
in the presence of abundant molecular oxygen, and on
the modern Earth this occurs only through the
intervention of Mn-oxidizing bacteria. Mn deposits
appear only during certain intervals in the geologic
record. The earliest Proterozoic was a unique period of
Mn deposit formation (some of the deposits are of
great economic importance). If these deposits were
-0.5
-1 .o
1 , , : 1 ;I .
-1.5
formed as oxides (rather than carbonates) then they 0 1000 2000 3000 4000
would be unambiguous indicators of the appearance Age (Ma)
of high oxygen levels. Figure 2 Evidence for a change in the sulfur cycle 2.5 billion
Although sulfur species are not a significant com- years ago: the quantity is the deviation of 633S(the
ponent of the present-day atmosphere, and the bulk of fractionation of the isotopes 33Sand 32Swith respect to a standard
surface S is in the ocean, sulfur deposits may serve as sample) away from a mass-dependent fractionation curve which
relates 633S to 634Sby a line with slope -0.51. All known biological
indicators of past oxygen levels because the rate of and geochemical fractionation processes are mass-dependent,
oxidative weathering of sulfur minerals and the and any samples experiencing only these effects would have
transfer of S as sulfate to the oceans is controlled by A33S = 0. While all samples younger than 2 billion years have
atmospheric 0 2 . The other input of sulfur to the = 0, older samples exhibit mass-independent fractionation
ocean-atmosphere system is SO2 in volcanic gases. which is suggestive of an important role for gas-phase (atmos-
pheric) reactions. This fractionation may be a consequence of an
Variations in atmospheric oxygen will change the
anoxic Archean atmosphere and a lack of oxidative weathering of
relative importance of these two sources. Possible sulfide minerals. (From Farquhar J, Baoh H, and Thiemens M
evidence for such a change is the disappearance of iron (2000) Atmospheric influence of Earth’s earliest sulfur cycle.
formations after 1.8 Ga. Rather than being the result Science 289: 756-758. Reprinted with permission.)
766 EVOLUTION OF EARTH’SATMOSPHERE
amounts of C 0 2 in the oceans and carbonate rocks are more poorly constrained by the geologic record. The
equivalent to 0.023 and 40 atmospheres, respectively, relative abundance of the isotopes I5N and 14N in
there is the potential for substantial evolution in kerogens (organic molecules that are the sedimentary
atmospheric carbon dioxide levels! Higher C 0 2 is product of the breakdown of living organisms) in
suspected as one cause for a warmer climate during the Precambrian rocks indicates a shift in relative abun-
Jurassic and Cretaceous periods. It has been suggested dance from lighter to heavier nitrogen during the late
that there was a secular decrease in atmospheric C 0 2 Archean. One explanation for the shift is an increase in
over time, with levels as high as 300 hpa during the the oxygen content of the atmosphere that led to an
early Archean. The rationale for this hypothesis is the increased production in nitrate, either by atmospheric
observation that the temperature-sensitive weathering processes or by biology. Nitrate compounds are used
of silicate rocks, which produces calcium and magne- by microorganisms, which either convert it to a
sium cations and alkalinity, could have regulated the reduced form (ammonia) to include eventually in
formation of carbonate rocks and sequestration of cellular material, or use it as an oxidant in energy
C 0 2 from the atmosphere. The weathering process generation (denitrification). Reduction of nitrates by
acts as a global thermostat: lower carbon dioxide denitrifying bacteria yields isotopically light N2 gas
levels would bring about lower temperatures, which and isotopically heavy residual nitrates. As a result,
would slow weathering and carbonate formation, surface nitrogen, including that in cells, becomes
allowing CO;? levels to rise. Likewise, lower solar relatively heavy with respect to the atmosphere.
luminosity in the past would have been compensated The shift is small (about lOppm), and it is unlikely
by higher pco,. Although the geochemical logic is that there was a large change in the partitioning
compelling, there is actually little evidence to between the atmospheric reservoir (the largest)
confirm or refute the hypothesis. Under low-oxygen and surface nitrogen between the Archean and
conditions, high C 0 2 concentrations would have Proterozoic. Thus the nitrogen isotopes may be telling
favored the precipitation of minerals such as us more about the history of atmospheric oxygen than
siderite (iron carbonate). Absence of siderite in of nitrogen.
some paleosols may place a constraint on atmospheric
C 0 2 (Figure 3).
The history of nitrogen, the primary but relatively Atmospheric Evolution and the Gaia
inert component of the modern atmosphere, is even
Hypothesis
According to accepted theories of the evolution of
0.0 I I I I I I I main sequence stars, of which the Sun is a typical
member, the solar luminosity has been steadily in-
-0.5
creasing by about 40% since the Sun formed. The
mean surface temperature of a planet like the Earth is
-
w
E
-1.0
determined by energy balance. If the composition of
the atmosphere had remained unchanged, the Earth’s
Y
m mean surface temperature would have been below the
$1.5 freezing point of water before about 2 billion years
P ago. But the sedimentary record shows that liquid
-0
-2.0 water has always been present on Earth. A plausible
resolution of the ‘faint young Sun’ paradox is that the
early atmosphere contained more greenhouse gases
-2.5
(e.g., C 0 2 ) .Two greenhouse gases other than C 0 2 that
have also been considered for this role are ammonia
-3.0 ~
and methane. Both gases have strong absorption
270 280 290 300 310 320 330 bands that fall within the spectral wavelength range
Temperature (K)
or ‘window’ where the Earth reradiates most of the
Figure3 Geochemical constraints on atmospheric pco2from the solar energy it receives back to space, but outside the
rock record. (A) Upper line, one-dimensional climate model primary C 0 2 absorption band at 15 pm. Their abun-
prediction of pco, as a function of global mean annual temperature dances (mixing ratios) in the atmosphere will depend
at 2.75Ga.(B) Lower line, estimate of the maximum pco, in soil
on their rates of production and destruction. The
waters of siderite-free palaeosols as a function of temperature.
(From Rye R, Kuo PH, and Holland HD (1995)Atmospheric largest present-day sources of methane and ammonia
carbon-dioxide concentrations before 2.2billion years ago. Nature are biological, and that is likely to have been the case at
378:603-605. Reprinted with permission.) early times as well. While ammonia is a very effective
EVOLUTIONOF EARTH’S ATMOSPHERE 767
absorber of infrared radiation, it is easily destroyed by See also

UV radiation and is unlikely to have been abundant Aerosols: Observations and Measurements; Role in
enough in the Archean atmosphere. Methane, on the Radiative Transfer. Biogeochemical Cycles: Carbon
other hand, has a chemical lifetime of 10 years: CH4 is Cycle; Nitrogen Cycle. Carbon Dioxide. Climate:
stable with respect to molecular oxygen, but reaction Overview. Gaia Hypothesis. Global Change: Human
with hydroxyl radicals ( O H ) from the photolysis of Impact of Climate Change; Upper Atmospheric Change.
water vapor produces CH3, which does readily react Paleoclimatology: Ice Cores; Varves. Reflectanceand
with 0 2 . In the absence of oxygen to remove the Albedo, Sutface. Thermosphere.Volcanoes: Role in
hydrogen also produced by photolysis, the OH can Climate.
recombine with H and is no longer an effective sink for
methane. The modern production rate can maintain a
CH4 mixing ratio of lo-. in an anoxic atmosphere, a
Further Reading
level that can act in concert with a modest amount of Chamberlain JW and Hunten DM (1987) Theory of Plan-
atmospheric C 0 2 t o maintain Archean temperatures etary Atmospheres. New York: Academic Press.
above freezing. However, the dominant modern Holland HD (1984)The Chemical Evolution of the Atmos-
source of methane is the microbiota in the anoxic phere and Oceans. Princeton, NJ: Princeton University
guts of termites and ruminants. It is not clear what the Press.
Hunten DM (1993)Atmospheric evolution of the terrestrial
production rates would be in the Precambrian before planets. Science 259: 915-920.
animals existed. Hunten DM, Pepin RO and Walker JCG (1987) Mass
It has been suggested that the existence and evolu- fractionation in hydrodynamic escape. Icarus 69:
tion of life on this planet may have had a profound 532-549.
impact on the climate by regulating the amounts of Kasting JF (1997) Habitable zones around low-mass stars
C 0 2 and other greenhouse molecules in the atmos- and the search for extraterrestrial life. Origins of Life and
phere, or by controlling the production of aerosols Evolution of the Biosphere 27: 291-307.
and thereby modulating cloud formation. Observa- Kasting JF (1993) Earth’s early atmosphere. Science 259:
tions of the fundamental role of biology in the cycling 920-926.
of atmospheric gases is one basis for the ‘Gaia’ Lovelock JE and Margulis L (1974) Atmospheric home-
ostasis, by and for the biosphere: the Gaia hypothesis.
hypothesis in which the biosphere is conceived as
Tellus 26: 1-9.
regulating climate to maintain suitable conditions for Melosh HJ and Vickery AM (1989) Impact erosion of the
biological activity. The ability of the biosphere to primordial atmosphere of Mars. Nature 338: 487489.
maintain a global environment that is optimal for Owen T and Bar Nun A (1995) Comets, impacts, and
life is known as homeorrhesis. That the terrestrial atmospheres. Icarus 116: 215-226.
atmosphere has been subject to homeorrhesis by the Porcelli D and Pepin RO (2000)Rare gas constraints on early
biosphere throughout its long history is an intriguing Earth history. In: Canup RM and Righter K (eds)Origin
hypothesis but difficult to prove. Investigations into of the Earth and Moon 11. Tucson, AZ: University of
the evolution of Earth’s atmosphere not only grapple Arizona Press.
with profound questions about our planet’s hab Salby ML (1996) Fundamentals of Atmospheric Physics.
itability but also address the possible existence New York: Academic Press.
Schlesinger WH (1997) Biogeochemistry: An Analysis of
and detection of habitable planets and life around
~ 1 0 6 ~Change.
1 2nd edn. San Diego, CA: Academic
other stars. Human society, through the rapid Press.
combustion of buried organic compounds (fossil Walker JCG (1977) Evolution of the Atmosphere. New
fuels), deforestation, and modulation of the plane- York: Macmillan.
tary albedo, now influences the evolution of the Wayne RP (1991) Chemistry of Atmospheres, Oxford:
atmosphere and climate. What kind of feedback Clarendon Press.
role we will ultimately play in the continuing evolution Yung YL and DeMore WB (1999) Photochemistry of
of our planet’s atmosphere and climate remains to Planetary Atmospheres. New York: Oxford University
be seen. Press.
FLOODING 769
C A Doswell 111, University of Oklahoma, Norman, OK, debris left behind when floodwaters recede can be
USA costly to clean up and also represent a health hazard,
Copyright 2003 Elsevier Science Ltd. All Rights Reserved especially when there are decomposing bodies of
drowned wild and domestic animals in the debris. In
some situations, floods drive wild animals (including
Introduction invertebrates of all sorts) from their normal habitats
and into human habitations near and within
Flooding is arguably the weather-related hazard that is the flooded areas, which can create various pro-
most widespread around the globe. It can occur blems, especially when the animals are venomous or
virtually anywhere. A flood is defined as water aggressive.
overflowing onto land that usually is dry. Flooding is Although flooding has some large negative impacts
often thought of as a result of heavy rainfall, but floods on humans, it is also part of the natural processes
can arise in a number of ways that are not directly shaping the Earth. Floodplains along rivers and
related to ongoing weather events. Thus, a complete streams are among the most fertile regions known.
description of flooding must include processes that Most of the so-called ‘cradles of civilization’ are
may have little or nothing to do with meteorological within floodplains for this very reason (e.g., the Nile
events. Nevertheless, it is clear that in some ultimate River, the Tigris-Euphrates River, among others).
sense, the water that is involved in flooding has fallen Hence, humans have been affected by flooding both
as precipitation at some time, perhaps long ago. The positively and negatively since before historical times,
origins of flooding, therefore, ultimately lie in atmos- whenever they find themselves in the path of these
pheric processes creating precipitation, no matter natural events.
what specific event causes the flooding.
Floods produce damage through the immense
power of moving water and through the deposition
of dirt and debris when floodwaters finally recede.
People who have not experienced a flood may have
little or no appreciation for the dangers of moving
water. The energy of that moving water goes up as the
square of its speed; when the speed doubles, the energy
associated with it increases by a factor of four.
Flooding is typically coupled to water moving faster
than normal, in part because of the weight of an
increased amount of water upstream, leading to an
increase in the pressure gradient that drives the flow. In
most cases, the damage potential of the flood is
magnified by the debris that the waters carry: trees,
vehicles, boulders, buildings, etc. When the waters
move fast enough, they can sweep away all before
them, leaving behind scenes of terrible destruction
(Figure 1).
The effect of the water itself can be devastating on
structures and on the objects within them: books,
furniture, photographs, electronic equipment, and so
on can be damaged simply by being immersed in water,
even if they are not directly damaged by the water
movement. Moreover, floodwaters typically contain
suspended silt and potentially toxic microorganisms
and dissolved chemicals. This means that floods
usually compromise drinking water supplies, resulting
in short-term shortages of potable water, with the Figure 1 Damage resulting from the 1977 Johnstown, Pennsyl-
additional long-term costs in restoring drinking water (c
vania, flash flood event. The Johnsfown Tribune-Democraf,
service to the residents of a flooded area. The mud and used by permission.)
770 FLOODING
Floods as a Direct Result Occasionally, flash floods are created in conditions

of Precipitation that are not favorable for thunderstorms but which
still produce heavy rainfalls. This can occur when
When the waters of a flood arise directly from moist air is forced upward over mountains by the wind
precipitation, atmospheric processes can be identified flow, called orographic precipitation. When the air
as directly responsible for the event. That is, rainfalls forced upward is very moist, the rainfall can be quite
occur that are well beyond the average values for the heavy. The steep, rocky terrain also promotes rapid
affected area. It is only when those rainfalls exceed the runoff of the rainfall. Flooding along the West Coast of
average that land which is usually dry can be affected; the USA or in the European Alps is often of this type;
that is, a flood occurs. Thus, the rainfall amounts that is, not involving thunderstorms.
needed for floods cannot be defined in absolute terms. A characteristic of flash floods is the localized nature
A precipitation event that causes a flood in one of the heaviest rainfall. As shown in Figure 3, the most
location might be well within the bounds of what is intense rainfall is typically confined to a relatively
typical for another location. Generally speaking, the small area. When large amounts of this localized
threshold for flood-producing rainfalls increases as the precipitation fall within a small drainage basin, flash
annual average rainfall for a region increases. floods can occur. Sometimes, the location where flash
flood damage occurs may actually receive little or no
rainfall. That is, the rainfall that causes the problem
Flash Floods
can occur upstream of threatened areas. This separa-
Flash floods are defined as those flood events where the tion between the rainfall and the flood can cause
rise in water is either during or within a few hours of confusion because it may not even be raining in an area
the rainfall that produces the rise. Therefore, flash for which flash flood warnings are issued. Another
floods occur within small catchments, where the factor in the impact of flash floods is that the
response time of the drainage basin is short. Many precipitation causing the event often falls during the
hydrological factors have relevance to the occurrence night, when it can be difficult to get warnings to
of a flash flood: terrain gradients, soil type, vegetative sleeping residents. The central part of the USA is well
cover, human habitation, antecedent rainfall, and so known for its heavy thunderstorm-produced rains
on. In steep, rocky terrain or within heavily urbanized during nighttime hours. Worldwide, thunderstorms
regions, even a relatively small amount of rainfall can are most common during the day, but on the central
trigger flash flooding. These hydrological factors plains of the USA (and in a few other places around the
determine the response of the catchment to the world), the unique geography of the region favors
precipitation event. Thus, a flash flood is clearly the nocturnal thunderstorms. This setting promotes a
result of the concatenation of both meteorological and strong flow of moisture northward from the Gulf of
hydrological circumstances. Mexico, called a low-level jet stream, during the warm
Most flash floods associated with rainfall are months of the year. Moisture carried by the low-level
produced by thunderstorms; that is, deep, moist jet stream helps to maintain thunderstorm systems
convection. A single thunderstorm cell is unlikely to that often begin during daytime hours on the higher
produce enough rainfall to cause a flash flood, so the terrain to the east of the Rocky Mountains. Because of
typical flash flood is the result of several thunder- the low-level jet stream, such storms can persist well
storms moving successively over the same area, known into the nighttime hours, often forming clusters
as ‘training’ thunderstorms (Figure 2), because it of thunderstorms known as mesoscale convective
resembles the passage of cars in a freight train. A systems (Figure 4).
succession of thunderstorms results when new thun- It is the rapidity of the event that makes flash floods
derstorms pass repeatedly over the same place while so damaging and dangerous. Flash floods involve
the overall system of thunderstorms is very nearly rapidly rising, fast-moving waters that can do im-
stationary. The infamous Johnstown, Pennsylvania mense damage; the suddenness of the onset of the flood
flash flood of 19-20 July 1977was produced by such a can result in people being caught unawares. Most
system. Thunderstorms forming in north-western fatalities result from drowning, with perhaps some
Pennsylvania moved south-eastward, only to be traumatic injuries from being carried along in the
replaced by newly formed thunderstorms, a process debris-laden waters and being swept into standing
that went on for several hours. The result was objects. The potential for loss of human life with flash
torrential rainfall concentrated near Johnstown, with floods is high. Debris carried in flash floods can form
amounts exceeding 400mm. The ensuing flood was temporary ‘debris dams’ that typically fail as waters
responsible for 77 fatalities and $550 million (in 1999 back up behind them. Failure of these debris dams then
dollars). results in a ‘wall of water’surging downstream. Debris
FLOODING 771
Figure2 Schematicof the 'training'effect. (A) Atthistime, thereare four numbered thunderstorm cells in variousstages of development.
Cell I is mature, with both updrafts and downdrafts, and heavy rain is about to commence at point X. Cells 11, 111, and IV are still developing,
and have only updrafts. Cell I I has precipitation forming aloft. The hatched contours are radar reflectivity, in standard units of dBZ, which is
related to the rainfall rate. (B) About 15 min later, Cell 1's updraft is dissipated, and it is now dominated by downdraft. Heavy rain continues
at X while Cell II is maturing and developing a downdraft. Cells 111, IV, and now V are still immature. (C) About 15 more minutes have
elapsed. Cell 1's rainfall is continuing but it is now nearly dissipated, while Cell I1 is entering late maturity. It is still raining at X but now the
rainfall is from Cell II, and heavy rain from Cell II is descending from aloft. Now Cell 111 is developing its first precipitation aloft. Cell IV and V
are still immature. (Adapted from Figure 7 in Doswell CA 111, Brooks HE and Maddox RA (1996) Flash flood forecasting: An ingredients-
based methodology. Weather Forecasting 11: 560-581 .)
772 FLOODING
Figure 3 Observed total precipitation (mm) during the Johnstown, Pennsylvania (JST, located by an asterisk) flash flood event. For
reference, Pittsburgh, Pennsylvania (PIT, located by the plus sign) is also shown. (Adapted from Figure 14a in Hoxit LR, Maddox RA,
Chappell CF, Zuckerberg FL, Mogil HM, Jones I, Greene DR, Saffle RE and Scofield RA (1987) MeteorologicalAnalysisof theJohnstown,
Pennsylvania, Rash flood, 79-20 July 7977.NOAA Technical Report ERL 401-APCL 43, NTlS Accession No. PB297412.) NOAA,
National Oceanographic and Atmospheric Administration.
dam failure events can happen repeatedly during the flash floods and landslides associated with its rainfall.
course of the flash flood. Not all flash floods are It was the worst weather disaster in terms of casualties
characterized by a ‘wall of water’ but all of them (by in the Western Hemisphere during the twentieth
definition) involve rapidly rising floodwaters. century.
Because urbanized areas promote runoff of rainfall,
rather than permitting most of the rain to be absorbed
River Floods
into the ground, flash flooding is more likely in cities
than in rural areas surrounding a city. It takes much River floods, in contrast to flash floods, typically
less rainfall in a city to create a flash flood situation unfold over days, or even months. This is because they
than in a rural area of comparable size. occur in large basins involving ‘main stem’ rivers like
Flash floods continue to be a major contributor to the Missouri, or the Nile, and are usually the result of
loss of life, in spite of improved precipitation fore- many individual rainfall episodes spread out over
casting. Some noteworthy examples include events in many days. In fact, within a river flood event, several
the Big Thompson Canyon in Colorado (1976 - 144 flash flood events can occur. Again, hydrological
fatalities) and near the town of Biescas in the Spanish factors often contribute to a river flood, but river
Pyrenees (1996 - 86 fatalities). floods are not so sensitive to them as are flash floods.
Tropical cyclones often create devastating flash Whereas individual thunderstorm systems can cause
floods as a result of torrential rainfalls. In late October flash floods, river floods are usually the result of a
of 1998, Hurricane Mitch caused more than 9000 stagnant synoptic-scale weather pattern. Localized
fatalities (the exact number is not known), mostly in heavy rainfall events occur many times during a period
Nicaragua and Honduras, in Central America, from of days or even months, each contributing its share of
FLOODING 773
Figure4 False-color enhanced infrared satellite image of a mesoscale convective system, with the light red colors indicating the coldest
(therefore the highest) clouds. Note that this image is from 17 August 2000 at 0345, local time, which corresponds to 0845 UTC.
rainfall to the tributaries, which then discharge into period over the lower Missouri and upper Mississippi
the main stem of a river. The river rises gradually in basins. In addition to these factors, considerable
response to all the input rainfall. The river flood rainfall over the region had fallen during the previous
potential of a situation can be increased by concurrent several months, providing a hydrological setting that
snow melt and other factors besides rainfall. favored runoff of the precipitation. This event pro-
The major flooding event during June and July of duced disastrous flooding that persisted for many
1993 was the result of a weather pattern (Figure 5A) weeks.
that produced a storm track across the upper Mid- Owing to the long time scale of the rising waters,
western USA. Abnormally low heights of the pressure river floods pose a lower risk of fatalities; people have
surfaces (associated with cool temperatures) over the more time to take proper actions. Of course, some
northern Plains produced a pattern in which traveling casualties result from waiting until it has become too
weather disturbances intensified in the Midwest after late to respond to the threat. Levee and dam failures, as
crossing the Rocky Mountains. This pattern aloft also well as intentional rapid release of impounded waters
produced an anomalously strong poleward flow of to prevent the catastrophic failure of the flood control
low-level moisture from the Gulf of Mexico into the structures, can produce rapidly rising water situations
Midwest. Mesoscale convective systems developed embedded within a river flood, and these also can
almost every evening during the early summer, typi- contribute to loss of life.
cally persisting through the night. These passed Because of the large scale of river floods, the damage
repeatedly over nearly the same areas, resulting in figures may be enormous; easily into the billions of
widespread significant rainfalls (Figure 5B) for the dollars. Crop losses are a major factor in the costs of
774 FLOODING
Figure5 (A) Map of the 700 hPa heights (thin lines, in dam) and height anomalies (shading, in m) for June/July 1993, and (6)
observed
precipitation for the same period (shading, in mm). Based on data supplied by J. Janowiak of the NOAA Climate Prediction Center.
river floods, whenever large tracts of prime agricul- damage and dislocations along the Upper Mississippi
tural land along floodplains are inundated. Levees are and Lower Missouri basins during the summer floods
often used to protect populated areas, so the failure of of 1993, during which several levees were breached,
those levees can generate major property losses. The illustrate the huge impact such events can have.
FLOODING 775
Floods Arising from Nonprecipitation Societal Impacts and Their Mitigation

Events
The results of floods on society worldwide are
Apart from floods resulting directly from rainfall, substantial. Flooding is responsible for many drown-
there are many ways in which precipitation can cause ing fatalities in tropical cyclones, either from storm
floods, perhaps long after it has fallen. When flowing surges or from freshwater rain-induced flash floods.
water is impounded by the construction of dams, there Flash floods and river floods typically produce more
is some risk that the dams will fail. Johnstown, fatalities every year than either tornadoes or hurri-
Pennsylvania, was inundated by a dam failure during canes in the USA. In many parts of the world, flood
a rainfall event in 1889, for example. Such rapid fatalities are associated with the most significant
releases of stored water can be cataclysmic, manifest- weather-related disasters. Flood damage cost in the
ing themselves as an enormous ‘wall of water’ choked USA is now on the order of several billion dollars
with debris. annually, and this figure continues to rise.
Flood can also arise through the melting of snowfall. Many people now live and play in flood-prone
In situations where the preceding winter’s snowpack is areas: for example, within floodplains of rivers and
deep, a sudden change to warm temperatures in the their tributaries, as well as along coastlines that are
spring can result in abnormally rapid melting and vulnerable to storm-caused flooding from tsunamis,
runoff of the snow melt. The devastating flood created tropical cyclones, and nontropical storms. Develop-
in Grand Forks, North Dakota, in April of 1997 is an ment of flood-prone areas for habitation and recrea-
example. Occasionally, warm rain falls directly onto tion has been increasing, with a corresponding
the melting snow, exacerbating such situations by increase in the risks to life and property. The 1993
speeding the melting process and adding more liquid Upper Mississippi and Lower Missouri River floods
water. provided a grim reminder of the risks of building
Deposits of snow and ice on volcanic peaks can melt permanent structures within floodplains, even when
rapidly during eruptions. The resulting runoff, often flood-control measures have been taken.
turned into a thick slurry by the inclusion of volcanic In the case of flash floods, it is difficult to take
ash, roars down the mountainside and is called a lahar. measures to protect property, owing to the rapidity
A tragic example occurred with the Nevado del Ruiz with which the event happens. However, prevention of
volcano in Colombia on 13 November 1985, which flash flood casualties is possible, provided warnings
killed more than 23 000 people, mostly in the town of can be issued and acted upon properly in a timely
Armero. Another occurred in Iceland during 1996 on fashion. Considerable attention has been paid to
the Vatnajokull glacier, with no fatalities owing to its increasing public awareness of the dangers of driving
remote location. Lahars can continue occasionally for into rapidly rising flood waters, for instance, as a result
years after an eruption, when heavy rains fall onto ash of recent experiences with flash floods. Unfortunately,
deposited by the volcano. situations can still arise where warnings are not issued
During the winter and late spring, when ice can in time. People living and engaging in recreational
build up on rivers in cold climates, the breakup of the activities in places prone to flash floods need to be alert
ice can create ice dams on the river. The ice dams cause during heavy rainfalls and be prepared to seek safety
the waters to back up, sometimes flooding the land even when they do not receive timely warnings.
upstream of the ice dam. Then, the breakup of the ice For river floods and other relatively slow-developing
dam can result in a flash flood wave that surges situations (such as rising snow melt or ice action events),
downstream of the ice dam’s position. it may be possible to reduce the property damage as well
Other flood situations can develop along the shores by removing the contents of structures. Obviously, any
of the world’s oceans and even with large freshwater structures (and their contents) built in flood-prone areas
lakes. Tsunamis, typically caused by underwater are permanently at risk; the only way to guarantee their
earthquakes and landslides, can flood the shorelines protection from floods is to move them out of those
with huge waves that break on the shallow waters near areas. Prevention of fatalities in river flood events is a
the shore. Storms of all sorts, including tropical matter of heeding the warnings of danger and moving
cyclones, can drive the waters before the winds into residents out of the danger areas before the number of
storm surges that inundate shore areas when the options is reduced by the rising waters and by the failure
storms are near the land. Large lakes can experience of levees or other flood-prevention structures.
flooding on their shores due to seiches, which are Forecasting the details of flooding events is an
surges of water (usually oscillatory) within enclosed important part of mitigation. Knowing precisely when
bodies of water. Seiches can be caused by earthquakes and where a flood will occur would no doubt be
or by atmospheric processes. helpful, but it is also important to be able to anticipate
776 FLOODING
the magnitude of the flood. An example of this is the and other structures was dramatically recalled to
tragedy of the 1997 Grand Forks, North Dakota case, public attention. The value of structural methods for
where the river level was only a few feet higher than flood control (levees,flood control dams, breakwaters,
that forecast. Those few feet, however, had a large etc.) remains controversial, but the 1993 floods made
impact, because the flood-control operations were it apparent that structures such as levees can be
based on the lower forecast value. When the river rose breached during major flooding episodes, even though
above that level, the flood-control measures failed they may be able to contain lesser events. Structural
catastrophically. In reality, such a forecast can never be failures create rapidly rising waters (flash floods)
a precise statement; uncertainty is implicitly a part of artificially within a river flood event, increasing the
every forecast, a point that perhaps needs greater hazards to human life as well as destroying property.
emphasis in the future. The decision about when and where to take structural
Flooding, by its very nature, is usually a result of approaches will continue to be a challenge.
both meteorological and hydrologic processes; the Finally, the use of flood-prone areas for human
character of a flood is determined both by the detailed activities puts lives and property at risk, although the
behavior of the precipitation and by the nature of major flood events may be separated by many years.
situation in which the event is likely to occur (soil The long time between events can lead to complacency
conditions, amount of antecedent rainfall, and so on). and subsequent disasters. The choices associated with
It is not likely that precisely detailed forecasts of land use are a continuing challenge, now and in the
flooding events will ever be possible, although it is future. When humans live and play in ways that put
certainly well within our capability to anticipate the them in the path of potential floodwaters, major
possibility of most flood events. The challenge for societal impacts are inevitable.
reducing the social impacts of floods is how best to
make use of the uncertain meteorological and hydro- See also
logical forecasts that are within practical means. The
challenge is to make effective use of whatever fore- Air-Sea Interaction: Momentum, Heat and Vapor Flux-
es. Convective Storms: Convective Initiation; Overview.
casting capability we have, even as we seek to improve
Hurricanes. Hydrology: Modeling and Prediction; Over-
that capability. view. Mesoscale Meteorology: Cloud and Precipitation
Bands; MesoscaleConvective Systems. Palmer Drought
Effects of Human Activities on Index. Predictability and Chaos. Radar: Precipitation
Flooding Radar. Satellite RemoteSensing: Precipitation.Severe
Storms. Weather Prediction: Severe Weather Fore-
In addition to the risks to lives and property that casting.
people take by moving into flood-prone areas, devel-
opment for human use often involves clearing land of Further Reading
its native vegetation and altering the characteristics of
the ground cover. Vegetation works together with the Agnone JC (ed.) (1995)Raging Forces: Earth in Upheaval.
soil to store rainfall, so when that vegetation is cleared, National Geographic Society.
rainfall runoff can increase substantially. Rather than Barry J M (1997)Rising Tide: The Great Mississippi Flood of
1927and How I t Changed America. Simon and Schuster.
being absorbed by the soil and its natural vegetation, Cluckie ID and Collier CG (eds) (1991) Hydrological
in areas where that vegetation has been cleared (either Applications o f Weather Radar. Ellis Horwood.
for construction or for agriculture), heavy rainfall is Dingman L (1994)Physical Hydrology. Upper Saddle River,
more likely to run off and pour into streams and rivers, NJ: Prentice-Hall.
increasing the potential threat from flash floods and Hill CE (ed.) (1986)Nature on the Rampage: Our Violent
river floods. Construction of roads and buildings also Earth. National Geographic Society.
acts to increase runoff, and leads to an increasing Lorenz EN (1993)The Essence of Chaos. Seattle: University
likelihood of localized urban flooding. Such construc- of Washington Press.
tion dramatically increases the fraction of the rainfall Ludlam F (1980) Clouds and Storms. University Park, PA:
that runs off, regardless of antecedent rainfall. Hu- Pennsylvania State University Press.
man-caused fires can also produce at least temporary Pruppacher HR and Klett JD (1997)Microphysics of Clouds
and Precipitation, 2nd edn. Dordrecht: Kluwer Academic
increases in the runoff potential in the headwater Press.
regions of streams and rivers. It is evident that human Ray P (ed.) (1986)Mesoscale Meteorology and Forecasting.
activities are increasing the potential for floods around Boston, MA: American Meteorological Society.
the world. Sarewitz D, Pielke RA Jr and Byerly R (eds) (2000)
Again recalling the Mississippi River floods of 1993 Prediction: Decision-Making and the Future of Nature.
as an example, the issue of flood control through levees Washington, DC: Island Press.
FOG 777
P J Croft, University of Louisiana at Monroe, Monroe, convenience to annoyance and from high costs to
LA, USA deadly consequences. Although mostly negative con-
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. sequences and perceptions are associated with fog,
there are positive benefits as well. These range from a
pleasing esthetic effect to fog harvesting for agricul-
Introduction tural and water supply applications. Thus repercus-
sions can lead to a variety of associated political,
Murky and perhaps even eerie, or pristine and serene, social, and legal implications, depending upon the
are just a few of the descriptions of the often blinding precise impact and the person or peoples - and
white-gray veil that comes to mind when people think economies - affected. The first and most obvious
of fog. Whether over oceans or local waters or over impact of fog is related to the reduction of visibility.
various landscapes, a certain uniformity and blank- This reduction hampers and restricts our navigational
ness is associated with fog. Often hugging the terrain, abilities and thus increases our chances of judgment
or simply masking the landscape, fog compromises the errors in the operation of transportation vehicles.
integrity of our senses that are honed to a fog-less Reduced visibility in fog quickly impacts our ability to
environment. Sometimes it is like an unbending solid drive, move over water, fly, and transit land by train. In
wall, many times vanishing over short distances until each case the inability to see well, or to see an adequate
another patch is encountered, and even the beauty of a distance ahead, is compromised by both fog and the
fogbow belies the significance of fog. speed of motion. It is further compromised by our own
Merely a collection of small-diameter suspended ocular inability to distinguish objects given limited
water droplets in the air, fog may occur in a calm brightness and contrast that occur with fog.
atmosphere that is saturated, or nearly so, and may
also occur in cool and moist air moving quickly past Land Transportation
us. Although fog is commonly thought of as merely ‘a
cloud on the ground’, it is much more than this given Land transportation includes automotives, trucks,
its very dynamics of formation, ‘intensity’ (or thick- and heavy machinery and is prone to disruption and
ness), nature of the droplets that it consists of, and delay when fog is present. Near Windsor, ON (Can-
areal extent and duration. Fog occurs around the globe ada), a highway ‘pileup’ collision during morning fog
for many different reasons and can be elusive when in September 1999 resulted in seven deaths as 62 cars
predicting its exact occurrence. and tractor-trailers collided. In the United States in
Will fog form? Where will it form? When will it Kingsburg, CA in November - and Waynesboro, VA in
form? How ‘thick’ will it be? How long will it persist? April - heavy fog resulted in highway pileups that
These are only a few of the many questions forecasters killed 42 and injured 9 1 people as 40-65 vehicles
and people ask about the enveloping droplets of fog. collided in mountain and valley regions in 1998. The
Yet to examine fog occurrence it is necessary to Virginia pileup is a ‘chain-reaction’ crash in a region
consider first its impact on our activities and then learn prone to ‘heavy’ fog, heavily traveled, and which
more about its characteristics and physical behavior. frequently experiences low visibilities. A tour bus and
Beyond fog impacts, the observation and study of fog truck collided in Asuncion (Paraguay)in March 2000
helps us to define its characteristics more completely while traveling through early morning dense fog and
and thus aids in our understanding of the fog process. 30 of 45 people on board were killed. A caravan of
With this information we are able to better predict its buses transporting college students in Pennsylvania
occurrence, extent, intensity, and duration so as to (United States) traveling through dense fog overnight
avoid or mitigate some of the hazards associated with collided with one another killing two and injuring 106.
fog. In fact, such an examination provides us with an In Bourg-Achard (France) in September 1997,
opportunity to make use of fog in various agricultural, several chain-reaction crashes claimed eight lives and
military, and other applications. The significance of fog injured 63 as over 100 vehicles were involved during a
and fog prediction includes impacts as well as benefits. mid-morning ‘heavy’ fog event. Witnesses and victims
reported that visibility was merely 45 yd (41.148 m)
when the crashes occurred. On the Ivory Coast in
Impacts Abidjan (Africa), ‘thick’ fog combined with dusty
Fog occurrence impacts a wide variety of human winds from the Sahara Desert during the Harmattan
activities worldwide. These impacts range from in- Season in December 1995 killed 14 and injured 86.
778 FOG
News reports indicated that a similar accident in desperately to maintain their scanning of the roadway
August killed 20 and injured 62, and that drivers in this ahead to ensure their own safety.
region are known for speeding and for their reluctance In addition, the variation of fog intensity and
to diminish their speeds even when weather conditions duration create a rapidly changing set of visibilities
are poor. In Lisbon (Portugal), four were killed and 70 during the course of travel and may be further
hurt in a 100 car pileup in January 2000 that halted up enhanced by hilly terrain and/or protected regions.
to 6000 cars in both directions of a 20 mile roadway Although no criteria exist for safe driving in fog, it is
for 5 h. clear that visibilities under 1mile (or 1.6 km) while
In Mobile, AL in the United States, 193 vehicles driving at speeds of near 60miles per hour (Le., one
collided on the Mobile Bay ‘Bayway’ highway in the mile per minute; or 96.54 kph) compromises seriously
country’s worst fog accident ever in March 1995, a driver’s integrity and response time to hidden
sending 91 injured to hospitals and killing, miracu- hazards. This is often exacerbated by the distance
lously, only one person. Insurance losses were esti- between vehicles and curved or inclined sections of
mated at over one million dollars at the time of the roadway. Thus the first lines of defense for navigating
accident, and witnesses and victims report sitting in fog is the reduction of speed, the use of headlights
their cars and listening to the continuing crashes and/or flashers, fog lights or fog-free lenses and
behind them. Some report driving ‘into a wall’ of fog shields, and the ‘stop, rest, and wait’ approach. Other
with visibility immediately reduced from 0.5 mile alternatives include fog dispersion or mitigation tech-
(0.8045 km) to near zero. The roadway was closed for niques discussed later.
hours in both directions of travel. The event led to the Rail transportation may be impacted similarly by
installation of fog sensors in the hopes of avoiding a fog conditions. In Badrshein (Egypt),on the Nile River
repeat of the accident. The same was done for a fog in December 1995 at about 8.00 a.m., one passenger
warning system in Waynesboro at a cost of over five train plowed into an express coach which had slowed
million dollars. due to ‘heavy’ fog. The wreck killed 75 and injured 76
Although many deaths are directly attributable to as five train cars were destroyed and 40 damaged.
collisions, a number are caused by fires ignited during Reportedly the train’s driver could not see even a yard
the collision process. Many factors lead to such serious (approximately 0.9144 m) to the front and had
consequences during fog events. These include poor apparently ‘stuck his head out of the window’ to try
visibility, vehicle speeds (posted as compared to to see better. Rail collisions have also involved motor
traveled), traffic volume, roadway design and surfac- vehicles and marine vessels at various crossings.
ing, driving habits (which include invincibility and
trust of braking systems, e.g., anti-lock systems),
Marine Transportation
roadway conditions (perhaps dry but sometimes wet
due to mist or drizzle, previous rains, or condensation) The operations of ships and barges, pleasure craft, and
that restrict braking ability, and windshield visibility sailboats may all be hampered by fog. Fog events often
effects. Although fog has been cited as the primary slow, and may even stop, marine operations with
cause of an accident in generally less than 1%of all significant economic costs. A well-known event is the
accidents in a given region, it has been cited up to 10% sinking of the Andrea Doria in July 1956 off the mid-
of the time in a fog-prone region, particularly in Atlantic coast of the United States after collision with
multiple collisions. The average claim for one vehicle the Swedish liner Stockholm. One of the most serious
is nearly $8000 (US) and over one million dollars for a incidents occurred in the northern portion of Mobile
multiple-vehicle crash. Bay, Alabama (United States) in September 1993.
The most obvious threat is reduced visibility which During the early morning hours a barge collided with a
restricts a driver’s ability to navigate the roadway. This rail span, moving it out of alignment moments before
is further diminished with increasing speeds and of the arrival of an Amtrak passenger train. Rescue
serious consequence resulting in many deaths and operations were severely hampered due to inability to
injuries every year around the world. The visibility not reach easily the crash location and survivors as 47
only restricts horizontal distance and depth percep- people died. Three years later the country’s Coast
tions but also reduces the ability of drivers to gauge Guard implemented rules requiring towboats to be
their own speed of travel. Through computer simula- equipped with radar, searchlights, radios, compasses,
tions a psychologist was able to determine that and other navigational gear and that the crews knew
although drivers could learn to sustain their speeds how to use them.
in simulation, the addition of fog distorted or de- Shipping operations are stopped or slowed when
stroyed this ability. Ironically, many drivers are prone visibility is as low as 0.25 mile (0.4023km). In a major
not to check their speeds in a fog situation as they seek (or minor) port or cargo region, large financial losses
FOG 779
can occur when operations are halted. In the case of for fog, a delayed or canceled flight may cost an airline
delays, reduced supplies and delivery of critical between $5000 and $25 000 per flight.
elements and products limit the productivity of
industry and commercial interests and their ability to Military, Rescue Operations, and Other Impacts
provide services to their clients. At the same time, ship
Although these are the most common types of impacts
operations require daily operating and maintenance
(Figure l),many other impacts do occur for a variety
costs, salaries, and living expenses which decrease
of military and rescue operations and other activities.
profits. In the case of halting operations, these For example, the D-Day invasion and other theaters of
considerations also include reduction or elimination
engagement have benefited or suffered from the effects
of the viability of product, particularly if perishables
of fog. The deployment of troops in Tuzla Bosnia-
or refrigerated goods are being transported. Losses
Herzegovina by the United Nations was put ‘on-hold’
may range from $10 000 to $25 000 per day per ship
for several days as persistent and continuous fog
and into millions for even a moderately active harbor
claimed the land in December 1995. The rough terrain
or port of call.
is known for its bad weather and thus allows only
small windows of opportunity for flight operations to
Air Transportation be made safely. In addition, rescue operations for a
cargo ship, which had collided with another off the
Aircraft impacted by fog include airplanes, blimps,
South Korean coast, were suspended in June 1996 as
balloons, and helicopters. Although fog-related crash-
persistent fog limited visibility to less than 10yd
es have occurred for large aircraft in the past, this is
(9.144 m). Space shuttle operations, as well as delayed
much less frequent today given instrument flight
launches and landings, have been impacted by fog or
regulations and improved navigational beacons and
the prediction of its occurrence.
technology as part of air traffic control. Smaller planes
Although rare, fog has caused sporting event can-
however are more prone to difficulty as they lack such
celations (baseball and football) or suspensions and
systems. The more typical impact of fog today is the
class delays or cancelations at schools and colleges
delay and diversion of flights. In Hong Kong (China)in
(especially if a large number of commuter students).
March 1998, 10 000 passengers were stranded at the
Fog has also played a role in dangerous air pollution
airport as over 50 flights were canceled, delayed, or
episodes including Belgium in 1930, Pennsylvania in
diverted. Dense fog events in Minneapolis, M N
1948, and London in 1952. More recent events
(United States) in November 1997 and February
include fog combining with forest and oil fires,
2000, and at the Jackson Mississippi (United States)
chemical spills, and other emissions. These events are
International airport in June 1996 resulted in flight
sometimes aided by topographic variations (e.g., a
delays for thousands of travelers during the morning
valley location, river valley, or other water source),
hours. At LaGuardia Airport in New York
regional climate (e.g., coastal), and nature of human
(United States) in February 1996, delays claimed
activities (i.e., industrial or agricultural) in a stable
most travel plans for an entire day as visibility was
atmosphere (e.g., Mexico, Arctic).
reduced to 300ft (91.44m). Some passengers were
Air pollutants can also act as condensation nuclei
shuttled to nearby airports 10-30 miles (16-48 m)
and create lower visibilities through haze and fog. This
away to take alternate flights. Most flight operations, was recognized both in early Roman history as well as
regardless of instrumentation, require at least
prior to that time. Some of the most intense fogs result
0.75 mile (1.207 km) visibility for takeoff and
when high concentrations of pollutants and other
landing.
aerosols are found in the air. London fog and its
Weather has been reported as responsible for one-
combination with pollutants, recognized as early as
fourth to one-half of all aviation accidents in the
the 1660s, was brought to the attention of King
United States, including fatal accidents, with an
Charles I1 and Parliament. In Donora, PA during
average of over 400 lives lost each year. The delays,
October 1948, nearly half the population of 14000
diversions, and cancelations resulting from fog add to
became ill during a prolonged valley air pollution
the cost of major accidents. Delays and diversions may
episode that was accompanied at times by fog. The
result in greater costs due to fuel usage, passenger
enactment of air quality regulations has aided in the
discomfort and complaints, and the shuttling of
abatement, but not the elimination, of such impacts.
passengers to alternate flights or airlines. Cancelations
result in displaced passengers and flight crews and
Legal Implications and Mitigation
create additional costs of lodging, food, and alternate
transportation. Depending on plane size and passen- Although no direct fog insurance exists, the costs
ger loads, and regardless of an airport’s preponderance associated with fog events and disasters may be
780 FOG
Figure 1 Fog transportation factors. The figure shows varying terrain and patches of fog.
covered according to damage, repair, and replacement celations. It will also include those who clear fog at
as well as personal suffering or loss. The liability may airports and other locations if they do not meet their
necessarily carry high rates as a function of the obligations under contract and are taken to court for
numbers of people involved and the nature of the damages.
‘cargo’ being transported, particularly in shipping, or
the nature of the impact, such as in an air pollution
episode. Payment is made based on a judgment of
whether the situation was avoidable, whether actions 0bservations
were prudent in the particular setting, and if the event The most obvious types of observations of fog are its
were predictable or an ‘act of God’ or a ‘once-in-a- occurrence, reduction of visibility and/or fog’s ‘thick-
lifetime’ event. ness’ or ‘intensity’, color (of limited use in reporting),
The sociopolitical decisions and mandates that duration, and extent. Each of these may be assessed
come from these find their way into legislation or locally at an observation site and regionally using a
standards of procedures as evidenced by aviation collection of sites. The traditional determination of
instrument flight rules (IFRs), port regulations for visibility in fog is based on the ability of a human
closure and/or delay/diversion, and the recommenda- observer to see predetermined ground-based targets in
tion (or requirement) that drivers use low-beam all azimuth directions about their site. Electronic
headlights when fog is present. Legal implications methods (e.g., transmissometer) have been applied to
may arise for a farmer who irrigates a crop (for water roadways, harbors, and aerodrome runways to deter-
or for frost protection) near a roadway and thus mine low visibility in the immediate location of the
enhances the fog and which results in an accident. sensor. Various methods and instrumentation used to
Clearly, liability is one of the reasons for more determine visibility are presented in the visibility
conservative decisions on closings, delays, and can- portion of this encyclopedia.
FOG 781
The occurrence and reduction of visibility are easilyspread of up to 5°F (approximately 3°C) and a stable
recorded based on human observation, but are often layer of atmosphere. Although fog occurrence is
electronically derived through the use of transmisso- relatively rare near or below the melting point of
meters and other instruments (e.g., the ASOS used by water, freezing or ice fog is possible. Other definitions
the National Oceanic and Atmospheric Administra- of fog relate the fog’s source and/or method of
tion’s National Weather Service as well as airports formation to its name.
around the world) and through remote sensing plat- Fog may form in place, be transported from one
forms. The duration and extent are considered location to another, and may form in minutes or over
through such platforms as well. The remote sensing an hour, depending upon existing conditions. The two
tools also include satellite, radar, and lidar. Satellite
basic processes responsible for formation, as well as
observations (both infrared and visible) can indicate duration, are radiation and advection (which includes
fog location and extent and infer fog ‘thickness’ in a vertical mixing of air). It is based upon these observa-
manner far superior to a collection of surface obser- tions that fog is often referred to as ‘a cloud on the
vation sites across a region. The use of satellite imagery
ground’ and which consists of visible hydrometeors.
is very helpful in completing the depiction of fog extent
However, the fog formation process has important
and ‘intensity’ for a region. Visible imagery often differences from clouds (including, for example, total
shows a sharp-edged boundary between a dull gray fog moisture content, droplet size distributions, and
region, whereas infrared imagery requires a tempera- chemical contents). By international definition, fog
ture differencing technique to indicate fog areas. In occurs (or is significant) when visibility is restricted to
either case the ‘strength’ or signal ‘brightness’ shown less than 1km (0.62 miles) and is distinguished from
by the image is related to its continuous nature (e.g., the occurrence of mist (or drizzle).
patchy versus widespread) and thickness (strength of Fog droplets range in size because of the various
signal) of the fog area. condensation nuclei with which they form and ac-
However, satellite observations have limitations cording to their resultant wettability and solubility
including minimal temperature contrasts on infrared (and thus their physicochemical composition). The con-
imagery and intervening cloud layers on visible densation process is best illustrated through Kohler
imagery. In addition, satellite observations typically curves that show droplet radius versus saturation level
are not current as processing and dissemination of (i.e., relative humidity of below or near 100% through
images may take an average of 1 0 4 5 m i n . In other supersaturation of up to 0.5%). As the droplet radius
words, satellite imagery is of limited value in terms ofgrows, the saturation vapor pressure decreases and
prediction of occurrence. Radar, although not a the environment becomes supersaturated with respect
detector of fog droplets, is useful in identifying to the droplet. This allows the maintenance of
variations in the refractive index within a limited droplets in an equilibrium state. Because of these
radius of the radar site. This can be used to infer the relationships, maritime nuclei often have very small
presence, or development, of an inversion and its droplet radii and thus little need for supersaturation
height which may indicate potential fog development. conditions.
The use of lidar is designed to detect even low Cited values for fog droplets range from 1to 65 pm
concentrations of small particles and can be used to in diameter with an average diameter of 10-20pm
resolve fog formation and occurrence within a limited most often reported (although some sources state the
distance of an observation site. range as 20-40). Some observational studies differen-
tiate between small droplet fogs (often maritime in
origin) and larger droplet fogs (more typically conti-
Definitions
nental). Maritime fogs are often observed to be more
Fog consists of suspended droplets, some of which continuous and ‘thick’ due to their smaller droplets
may be settling out and/or evaporating, which restrict that allow a greater concentration within a parcel of
visibility and persist for some period of time. Fog may air. The occurrence of fog with light rain or light
vary in depth occurring within the lowest meter of the drizzle is not uncommon and typically limits the fog’s
atmosphere and extending up to 1OOOm in height. ‘thickness’ and duration due to the fall of precipita-
Occasionally the fog layer may be somewhat elevated tion, mixing, and wind flow that accompany them.
above the surface, particularly during the fog disper- People tend to characterize fog as thick or shallow,
sion, breakup, or ‘burnoff’ process. For condensation, and some definition of this is found within the WMO
the relative humidities of the air do not necessarily coded synoptic and METAR observations. Fog is
need to be 100% and in fact may be as low as 80-90% usually distinguished from haze according to visibility
both during and after formation. These conditions are considerations, relative humidity, and content. Haze
mostly associated with a temperature-dew point and fog may occur together and have nearly equivalent
782 FOG
impacts, but fog’s restriction of visibility typically Coast, central Europe, and the Po Valley in northern
predominates. Somewhat by convention, fog ‘thick- Italy (Figure 2).
ness’ (or ‘intensity’) is defined according to the
restriction of visibility. Dense fog produces visibilities
Climatologies
of less than 1km, moderate fog 1to less than 5 km, and
light fog 5 to less than 11km.These definitions rely on Global frequencies of fog are traditionally based on
sight distance as a surrogate for measurement of surface observational data and consider the number of
droplet distributions and are exclusive of the occur- days on which fog was reported. Few climatologies
rence of low stratus clouds. focus on the time of day, extent, duration, or intensity
Fog duration and coverage, although observed in of fog which would make for a useful basis of
time and space, do not have a generally agreed upon worldwide comparison. Fog frequencies tend to be
definition. Fog studies often differentiate between high in locations where moisture is plentiful (oceanic,
long-lasting (i.e., several hours) and short-duration rivedlake, and coastal regions as well as humid or
fogs (less than 2h), and in some cases examine tropical locations) or cooling processes predominate
extended events (i.e., several days). Some studies (mountain-valley locations and ocean currents). As
consider a minimum one-degree ‘square’ of fog to be most frequencies are derived from surface-based
sufficient to depict it as a regional occurrence, with observations, they are necessarily skewed by popula-
smaller areas being defined as local coverage events. tion centers and established habitable regions and thus
Extended events lasting more than a couple of days fog ‘hotspots’ are only approximations of the true
include late winter and springtime events in New global occurrence and distribution. It must be noted
England and the mid-Atlantic states as well as winter- that even a region with a low fog frequency is not
time events over the western US valleys, the Gulf immune to the devastating impacts of fog. Some
Figure2 Observations and definitions. The figure gives aperspectiveview, as in Figure1, with focus on airport location and equipment
and varying fog intensities with spatial scale shown and varying depth of fog layer.
FOG 783
regions may experience an average of 100 days South America as related to El Nifio and La Niiia
per year with fog while others may only average less phenomena).
than 10.
Hemisphere has the majority of land-based and Location
coastal zone fog occurrences given the spread of
A closer look at the common features of fog-prone
continents around the Earth. Many regions in North
areas of the world provides a basic breakdown of
America, Europe, India, Africa, and Asia witness
maritime polar climates, Mediterranean climates (e.g.,
debilitating fog events, some sporadic, others long-
France), and mountainhalley climates (e.g., Chile),
lasting. A variety of climate zones are affected,
Although fog is also more likely to occur in the vicinity
including polar, temperate, tropical and even dry to
of water sources such as ocean currents and river
moist. Locations with the highest frequencies are
valleys and warm water springs (e.g., in cold climates),
found in the vicinity of cold ocean currents and/or
it may also be found in desert climates (e.g., Africa and
upwelling that stabilizes the atmosphere, provides
Australia). Radiational influences are maximized in
local moisture and aerosols, and can cool air quickly -
these desert regions as well as in valley or plateau
and do so for a prolonged period of time - to
regions around the world (e.g., China and Mexico).
condensation. For these locations fog may persist
When advective influences predominate (such as
for several days, cover an extensive area, and then
Newfoundland), various lifting and cooling processes
move en masse to another region. These fog regions
are more important. In addition, the intermediary
may occur any time of day and tend to be more
zone between ocean and land also plays a significant
persistent during the daytime with little diurnal
role in the extent, intensity, and duration of fog.
fluctuations.
Various combinations of these lead many to classify
The Southern Hemisphere also has a variety of land-
fogs and develop a list of fog types that may occur. This
based fog in regions such as South America, Australia,
allows us to distinguish between the climatic predom-
South Africa, and the Antarctic. As strong winds and
inance of fog in some regions versus its origins, and
mixing predominate over much of the Southern
according to the prevailing synoptic flow in real time.
Hemisphere oceans, fog regions tend to persist in the
Clearly, this does not imply that other places do not
coastal zones (e.g., Chile) and/or over the interior
receive fog or that it is unimportant. It does provide a
where terrain varies considerably (e.g., Argentina). In
context in which the scale of fog occurrence and extent
addition, the tropical rain forest regions provide ample
may be considered. These include synoptic-scale
moisture and sufficient radiational cooling overnight
features (e.g., high pressure), mesoscale variations
for the formation of large fog areas. In these cases, the
(e.g., in moisture distributions), and microscale (and
time of year and time of day vary from the Northern
at times mesoscale) features that affect the occurrence,
Hemisphere. Further observations are much more
extent, intensity, and duration of fog (Figure 3).
limited given smaller population bases and/or the lack
of observations or observational equipment in the
Scale
Southern Hemisphere. A wide variety of remote
sensing methods and tools have been used to refine The role of synoptic versus mesoscale and microscale
and localize these values but have not as yet provided a variations is best understood through examination of
complete view. For example, aerosol climatology a site-specific climatology. However, this is less prac-
work and cloud climatologies based on satellite and tical - and less informative -when the same principles
radar observations can more readily complete the are applied to a collection of sites. Instead it is best to
picture of fog occurrence, and thus frequencies and examine the patterns and/or fluxes that are important
climatology, for these parts of the world. at each scale to explore the characteristics of fog.
The importance of the foregoing discussion of fog These are also of use in the identification of fog
climatology is that it helps to identify fog-prone precursors which improve our understanding and
regions, episodic regions, and assist in the forecasting prediction of fog formation, extent, intensity, and
process. It also aids in the identification of common- duration.
alities of fog formation and thus offers clues as to the Synoptic-scale features include high- and low-pres-
evolution and behavior of fog. Some fog regions tend sure areas (with/without rain or recent rainfall),
to be synoptically enriched or dominated and others locations ahead of a warm front, the warm sector,
more by boundary layer processes. Each plays a role in and behind a cold front. In the simplest case, high
the extent, intensity, and duration of fog and helps pressure and sufficient near-surface moisture com-
determine whether fog frequencies are bi- or tri-modal bined with overnight cooling produce fog as radiative
in time and according to global circulation regimes processes dominate. In the case of low-pressure
(e.g., central North America and the west coast of systems and frontal regions, some advective and
Figure 3 Fog climatology. Map of the world showing continents, cold w e a n currents, topographic relief and some of the common locationsfavoring fog occurrence.
FOG 785
radiative processes will dominate. Fog occurrence is advective fogs, occurs in relatively stable layers of the
possible, whether inclusive of precipitation-induced boundary layer. Other relevant factors in fog forma-
and/or cooling effects. The surface character, over tion which may be considered as secondary in nature
which these features pass, then acts to promote or (yet significant in the prediction of fog) include cloud
moderate the cooling and/or lift. These include microphysics, the vertical and horizontal distribution
upslope flow, cold ground (or frozen or snow or ice of temperature and moisture, orographic effects,
cover), and vegetative contributions (in terms of sources and sinks of moisture and heat, and land use
additional moisture in the local boundary layer). and/or surface conditions (Figure 4).
Mesoscale variations modify the imposed synoptic
conditions and may grow with time to be more
Cooling
significant if synoptic flow becomes stagnant (e.g.,
autumn) or blocked (e.g., spring) or simply ‘vanishes’ There are several means of cooling an air mass, or
(e.g., low latitudes). Mesoscale variations in physiog- parcel of air, that may lead to fog formation. The most
raphy and weather conditions are obvious in areas obvious and most prevalent (even in the presence of
such as San Francisco, Death Valley, Salt Lake City, cloud cover) is the diurnal loss of heat by the Earth’s
and the Pacific Northwest (in the United States, West surface (i.e., radiational cooling). Other means include
Coast) and make a great deal of difference in the cooling of an air mass from below, adiabatic
forecasting the location, duration, and intensity of cooling (or mixing), the cooling of an air mass itself
fog. These variations are further complicated by the due to radiational release, and the evaporational
complexity of possibilities on the microscale where cooling of air due to precipitation through a dry air
boundary layer processes dominate. These tend to be layer which may induce cooling to saturation and thus
critical in identifying the precise location of formation,result in fog. Depending on the location, time of year,
timing, intensity, and duration. It is also critical with and moisture availability, these cooling mechanisms
regard to the microphysical aspects that involve may lead to fog formation with varying persistence
aerosols. This immediately distinguishes between and of varying extent and intensity.
maritime and continental concentrations and types Radiational cooling is primarily diurnal in nature
of fog droplets as previously described. and is maximized overnight and during the early
morning hours, with minimum air and surface tem-
peratures often occurring at or near sunrise. Although
Formation Mechanisms the diurnal cooling process occurs year-round, it is
Fog formation requires a variety of factors in different favored during both dry and cold seasons of the year
combinations. Essential to fog formation are sufficient when low-level moisture may be sufficient, relatively
moisture and the process of cooling and/or lifting undisturbed, and the cooling period lengthy. The
(inclusive of mixing). This gives us three basic ‘types’ dryness of the atmosphere is most typically observed
of fog: radiational (cooling), advective (cooling and/or above the boundary layer and allows great radiational
lifting), and combinatorial (cooling and lifting, or losses through an open atmospheric window, even in
mixing). These may occur in several ways from the the presence of middle- or high-level clouds. Radia-
synoptic to the local scale, as has been illustrated by tional fogs may be brief in duration (e.g., less than l h)
the different climates around the world which expe- or may last several hours. The depth and intensity of
rience fog. The more critical factor in the atmosphere these fog events is a function of the cooling time,
is the presence of sufficient moisture, in terms of total extent, and amount of moisture available. It is not
amount and depth, and its horizontal distribution. unusual for such fogs to initiate dew deposition.
Sufficient moisture may also be achieved by increasing The other means of cooling are of varying impor-
its ‘effectiveness’, in other words, making use of the tance to fog formation and duration. For example, the
moisture present and realizing it through cooling cooling of an air mass from below is favored in
and/or lifting processes to cause condensation. In each locations and seasons in which the active surface layer
case, despite synoptic influences, moisture and its is frozen and/or snow covered or when it experiences a
realization is very much a function of mesoscale and greater albedo (e.g., fallow versus the vegetative
microscale conditions and variations. growing season). Such fogs may form and persist for
In addition to moisture and cooling/lifting mecha- hours or days at a time and cover a relatively large area
nisms, fogs are observed to be associated with an with significant intensity. Adiabatic cooling of air is
inherently stable atmosphere. This stability may related to lifting mechanisms and therefore considered
precede or occur after fog formation and often in the next section. The cooling of an air mass itself due
increases with the advent of fog. Even fog that is to radiational release is typically a very slow process
associated with strong winds, as is the case with some and likely to be an important factor for persistent fogs
786 FOG
Figure 4 Fog formation mechanisms.
(e.g., sea fogs). The extent may be great but the these situations involve slow vertical lifting over large
internal variations in coverage and intensity are large horizontal distances or the relatively slow and shallow
given the interactions between the air mass and the vertical mixing of two distinct air masses in the
underlying surface features. Evaporational cooling boundary layer. In the former case, long-lasting,
caused by showery precipitation falling through a extensive, and intense fogs may be expected whereas
layer of dry air may be sufficient to lead to saturation in the latter short term, shallow, and patchy fog of
and fog, but is typically of short duration and of varying intensity occur.
limited intensity. In the case of a synoptic-scale warm Slow vertical lift due to an upslope wind flow,
front, such fog may form and persist for several days parallel with the elevation gradient, will result in
and become quite extensive and intense with minimal discrete levels of cooling and saturation with increas-
local variations. ing distance and transport. Although this process may
be slow in the initial formation of fog depending upon
the amount of moisture available in the air mass (eg.,
Lifting
several hours to nearly 24 h), it is a resilient process
The second basic means of cooling air to achieve or that can produce extended events of widespread dense
sustain fog formation, or for realizing the effectiveness fog (i.e., up to several days are possible). Similarly,
of the moisture present in a parcel of air, is through frontal lifting may produce similar conditions and may
various lifting mechanisms. These include orographic persist for some time dependent upon the rapidity of
lift, frontal lift, adiabatic ascent, and mixing. In many changes in synoptic features. Frontal lift is more
cases, these processes involve advection and thus give commonly warm in nature but may involve cold
rise to advective fog formation and transport. Al- frontal surfaces which are of lesser slope than a typical
though this implies that there are many lifting situa- cold front. In both orographic and frontal cases, the
tions in which fog may form, it is clear that most of formation, duration, extent, and intensity of fog
FOG 787
events is also a function of the underlying surface and over another or whether one type may evolve into
its interaction with the lifted air. For example, the flow another. Since the orographic and frontal lifting
of warm and moist air across frozen or snow-covered processes are typically a gradual cooling process over
ground - or simply upslope - increases the depth, long distances whereas the radiational cooling process
intensity, extent, and duration of fog. is gradual over time and specific to a location, it is
This last process is an important aspect and illus- reasonable to consider various combination fogs in the
trates how two diverse air masses, initially unsaturat- same manner. It is also reasonable to incorporate the
ed, may mix to form a saturated air mass. The use of effects of cloud microphysics, the vertical and hori-
saturation vapor pressure curves can be made to zontal distribution of temperature and moisture,
compare air mass properties as a function of their sources and sinks of moisture and heat, and land use
vapor pressures versus the absolute saturation vapor and/or surface conditions.
pressure for various temperatures and pressures. For example, given the features described above, the
When a cool, and relatively lower vapor pressure, air longest-lasting, most intense, deepest, and potentially
mass combines with a warm air mass with higher most widespread fogs may occur near a coastal region
vapor pressure, their mixing results in saturation. This with a moist onshore flow in the vicinity of a warm
may be seen by plotting the original vapor pressures at frontal (or topographic) or quasi-stationary boundary.
the individual air mass temperatures and connecting This would be further enhanced or favored if the flow
the two points with a straight line. When the line of moisture were sustained, the ground frozen or snow-
crosses the saturation vapor pressure curve, the two covered (and thus the source of cooling maintained),
mixed air masses will form a saturated air mass. The and it was the cool season time of year. The formation
manner in which these two air masses combine may be and advection of sea fog tends to meet these criteria to
through isobaric mixing or weak adiabatic mixing. varying degrees around the world and create some of
Lift that involves the adiabatic ascent and mixing of the foggiest regions known. Regardless of origin, that is
air is greatly dependent upon the existing boundary whether the sea fog formed first through radiational
layer which evolves during the mixing process to cooling or other cooling and lifting processes, it is clear
produce fog. Although of limited extent, turbulent that a variety of factors produce and sustain fog. This is
mixing through adiabatic ascent can result in fog verified by observation of the movement of fog areas
formation which is typically of very short duration and their passage from water to land.
(i.e., less than a few hours), limited depth (e.g., ground
fog), and highly variable in coverage and intensity.
Microphysical and Other Aspects
Such fogs may occur preceding and following the
passage of weak cold fronts with limited pressure and Although the identification of cooling and lifting
air mass differences, and often following the passage mechanisms is significant with regard to the conden-
of scattered showers or light rain, and take place in a sation process for fog formation, alone they are
conditionally stable boundary layer. These fogs tend to insufficient if not considered with regard to the
be infrequent and of short duration as the dynamics nucleation process. A knowledge of drop size distri-
are more likely to lead to low cloud (and ceiling) butions, condensation nuclei associated with fog
formation with drizzle. However, in some cases they droplets, local nuclei populations, and the resulting
may persist and thicken over regions as the frontal atmospheric chemistry are significant with regard to
boundary decays and/or becomes stationary. The the occurrence, extent, intensity, and duration of fog
processes of adiabatic ascent and mixing also play a events. Depending upon the population of condensa-
role in the formation of Arctic Sea Smoke and other tion nuclei, the initiation of fog droplet formation, and
fogs in which the heat flux is rapid and results from a the actual drop size distribution, fog development may
temperature differential rather than a period of occur within 5-15% of the saturation value of an air
radiational cooling. mass (haze is typically within 35%).The role of nuclei
is determined by ‘how active’ they are in encouraging
or discouraging the process of condensation (i.e., how
Lifting and Cooling
hygroscopic or hydrophobic). The presence of con-
Based on the preceding discussion, it is clear that there densation nuclei may be local in origin, advected, or
are many possible combinations which may produce the result of both processes and is significant when fog
fog. It is therefore understandable why so many ‘fog is considered in combination with smoke and pollut-
types’ occur in the literature and are studied around ants. The ‘proper’ combination of nuclei can lead to
the world. For the same reason, it is clear that these long-lasting and devastating fog events.
possibilities raise the question of whether fog is readily For example, there are a number of well-known
predicted and whether one type is readily identified cases in which fog combined with, and its formation
788 FOG
was aided by, industrial emissions. These created (or thickness), extent (or coverage), and duration.
unhealthy and dangerous air quality and low visibili- These are predominantly radiative processes as lifting
ties over several hours (and even days), causing death, mechanisms are essentially cooling processes as well.
illness, injury, and accidental losses. These can be It is assumed that if the proper combination of factors
further modified according to the location in which exists, and that moisture is available in sufficient
they occur. For example, marine environments typi- quantity and/or its effectiveness can be realized, fog
cally produce a large number of small droplets formation is possible. Adequate moisture is provided
whereas continental locations are characterized by through local evaporative fluxes, advective delivery, or
large droplets. In marine environments then it is clear evaporation of falling precipitation. Ideally these are
that while haze is favored in the daytime (and salt all quantified operationally (whether modeled, ob-
nuclei) and a high moisture source with mixing, fog served, or forecast) to make a precise determination of
may still occur if these conditions are overcome (e.g., fog occurrence, intensity, extent, and duration.
at night by cooling and with the introduction of
smaller droplets).
Regardless of the limitations to droplet size by the
concentrations of nuclei, it is the actual droplet
Radiative Considerations Formation -
concentration that determines a fog’s opacity - and
and Growth
that is often referred to as a fog’s intensity or thickness Rapid cooling, but with limited mixing, is best
(or severity). Because of this the lowest visibilities in accomplished through radiative heat loss by the
fog events are associated with high concentrations of surface and by an air mass that is predominantly static
small droplets. Thus again a wide variety of fog in nature and predisposed to stability. These condi-
formations are possible, particularly when consider- tions are favorable to the potential for fog formation,
ing the cooling and/or lifting processes and when assuming sufficient cooling occurs and sufficient
considering the location, transport, and interactions of moisture is available (or its effectiveness realized).
various nuclei across a coastal zone. Commonly, such conditions are associated with clear
(and sometimes dry) air, light winds, and subsidence.
The ‘Family of Fogs’ However, radiative cooling does occur in the presence
of cloud cover and may be enhanced or reduced by the
The foregoing discussion thus provides a ‘family of active surface over which air is present. Each of these is
fogs’ in terms of formation, extent, duration, and maximized within high-pressure areas when winds are
intensity which may be enveloped in a conceptual near calm, the boundary layer is moist, and the mid-
model of fog dynamics. Although similar in manner to and upper layers of the atmosphere are very dry with
the conceptualizations of air mass thunderstorm to strong subsidence. These conditions can produce
supercell, or mesocyclone to wave cyclone, or various widespread, intense, and long-lasting fog. Less opti-
levels of sealland-breeze model conceptualizations, mal conditions (e.g., those present in the midst of a
the family of fogs is, at the moment, quite elusive. weak low-pressure center with little pressure gradient)
Although generous amounts of research have been may also produce fog that is of limited extent,
accomplished and numerous modeling studies com- intensity, and duration.
pleted to reveal more explicitly the cooling and lifting Although radiative and advective process has been
processes that may produce fog, they are as yet initiated, the lowest portion of the boundary layer
incomplete. They do not incorporate adequately the nearest the ground becomes the coolest and thus
atmospheric chemistry and associated microphysics creates a microscale inversion. This inversion grows in
that are significant to the fog process and are poorly tandem with the rate of cooling and the net cooling
understood and not routinely observed. There is also a over time and, as the air reaches saturation, may lead
need to consider the interactions and interface be- to fog formation. As the cooling process continues, the
tween the underlying surface over which fog forms and inversion layer grows deeper and fog may grow or
these microphysical aspects of chemical and physical develop upwards with time to several meters within a
behaviors. few hours. At this point the presence of fog itself begins
to feed back into the radiative balance as warmth from
the surface may be absorbed by the droplets and the
Fog Dynamics rate of cooling slowed. In addition, the upper portion
Given the basic knowledge of fog formation mecha- of the forming fog and the fog layer itself continue to
nisms (or occurrence), and some knowledge of micro- cool, thus strengthening the inversion, while some
physical aspects, it is possible to explore the dynamic deposition and/or reevaporation of fog droplets occurs
processes involved and how they relate to fog intensity nearer the surface. Often at this stage the radiative
FOG 789
processes immediately at the ground have slowed, pressure system, or perhaps in advance of a warm
become less important, and reach a temporary equi- frontal boundary, both in the presence of light winds,
librium in which temperature and moisture conditions this may create large variations in fog extent, intensity,
remain nearly constant. and durations with frequent and rapid variations.
Although these processes dominate in a general These conditions also imply slow transport of a fog
sense, they are complicated by the nature of the surface layer and modification as the fog travels across varying
over which air lies (e.g., soil type), land use and cover, terrain and surfaces and as it encounters variations in
heat and moisture sources, and vehicular and other condensation nuclei. Modifications include other
traffic that create local turbulence. It is the combina- cooling processes, such as orographic lift, which may
tion of these factors that dictates the areal extent and reduce or enhance fog formation, maintenance, and
initial intensity of fog and that accounts for variations growth. For example, a fog that develops in the
as the fog persists. Sand and clay soils radiate at vicinity of weak low pressure may move upslope in
different intensities and thus can contribute to fog response to a weak pressure gradient and either
formation at different rates. In the simplest case, the precipitate out or intensify and deepen. The same fog
ground surface may be conducting heat from subsur- may encounter an industrial area with a variety of
face layers and thus eroding fog formation from the nuclei, which may lead to changes in the drop size
bottom after its initiation. The surface may also be distributions and either accelerate or defer fog forma-
considered active if wet or dry, vegetated or barren, tion.
paved or natural, and frozen or snow-covered. In the
case of a wet ground, more moisture and enhanced Intensity
radiative cooling are possible. For a frozen surface
Once fog has formed, persisted, and grown, the
there is a strong limitation on the radiative cooling of
intensity or thickness of the fog is of greater practical
the ground and a strong enhancement of the cooling of
significance. Fog intensity is a function of the concen-
the air itself. In the case of an asphalt or concrete
tration of small and large drops in terms of their ability
roadway, the radiative rates may favor more rapid fog
to reduce visibility to less than 1km. When fog is in
formation whereas vehicular flow (and turbulence)
place, there is a certain amount of droplet settling and
may discourage or disperse it.
separation with time while cooling and condensation
occur at fog top and dissipation and deposition at fog
Duration and Extent bottom (in many cases). While this process is effective
in maintaining fog, it is also effective in stratifying the
As the radiative cooling persists, the fog layer may
fog layer and generating variations in intensity. In
grow vertically and horizontally with time. In partic-
these situations the middle and upper fog layers
ular, it is not unusual for the fog on the surface to
become the most opaque and thus the most difficult
diminish with time, through deposition (dew or frost)
to navigate in terms of transportation. These effects
and weak induced convective mixing, giving the
may be negated to some extent given a greater degree
appearance of a lifting fog. In this process, the
of turbulent mixing or the presence or introduction of
moisture content and subsequently the dew point are
wind flow (and thus entrainment) near the top of the
decreased within the lowest portion of the fog layer
fog layer.
and may allow for further radiative cooling and fog re-
In addition, the characteristics of condensation
formation. Thus some fogs appear to vary in their
nuclei, can lead to varying intensities. For example, a
occurrence, depth, and intensity with time. More
variety of pollution-enhanced fogs owe their extreme-
importantly, the middle and upper portions of the fog
ly low visibilities to specific chemical species. These
layer are now the most ‘active’ in terms of their
encouraged fog formation prior to saturation, reduced
behavior. In particular, these layers continue to cool,
the amount of deposition, and created a more uniform
resulting in an upward expanse of the fog, and thus the
drop size distribution often characterized by its color
inversion layer, and therefore become somewhat
or smell.
independent of the surface over which they lie.
The horizontal formation and spread of the fog,
Dissipation
initially a function of the radiative properties of the
surface over which the air is found and which account The dissipation of fog, is a function of the processes
for the often patchy nature of fog formation, now that act against cooling and condensational effects
becomes a function of weak circulations or turbulence previously discussed. Therefore dissipation may be
above the ground. In the presence of high pressure and considered in terms of the ‘prevention’ of the processes
light winds, this may create a fairly uniform fog in that favor formation and growth, duration and extent,
terms of intensity and duration. For a weak low- and intensity. Many of these processes may reduce or
790 FOG
eliminate fog within less than 1h without contraven- of these is the recognition of those processes and
tion, but more typically require several hours to factors important to fog formation including cooling,
overcome the inertial presence of the fog layer as lifting, and mixing (cooling and lifting) mechanisms;
well as any underlying or continuing fog formation surface and air mass moisture and characteristics; and
processes. Usually visibility improvements occur the chemical and physical behavior associated with
within the first hour or two as dissipation processes fog droplets. Some of these are summarized by many
become dominant. Fog dissipation is typically longer authors and researchers according to a list of factors
in valley and coastal regions and during the cool such as the prior existence of fog, precipitation areas,
season and over cold waters - and in some cases may soil moisture and cover, temperature and stability,
not occur. boundary layer variations, orographic contributions,
The effects of cooling are mitigated or overcome the synoptic setting and flow regime, vertical and
through direct solar heating of the ground surface (or horizontal wind flows, cloud cover, and advection.
potentially the heating of fog droplets and the air layer Regardless of the list, it is essential to consider the
in which the fog is found, but this is of minimal synoptic-dynamic regime as it determines the charac-
significance) and the destruction of a stratified or ter of the boundary layer - and how that changes with
inversion layer through turbulent mixing. Thus warm- time - to predict adequately fog occurrence, extent,
ing and mixing are vital to dissipate fog and may be intensity, and duration.
accomplished through a variety of frontal or convec-
tive processes. This dissipation will be most effective at Climatological
the bottom and top of the fog layer and around the
One traditional means of fog prediction is based on
edges of the fog area where the air is not near
climatological considerations. This approach, al-
saturation. Thus fog is said to ‘burn off’ or ‘lift’ and
though location specific, is a reasonable first approx-
‘shrink’ with time. The pace of dissipation, which will
imation in identifying and summarizing the factors
be greater under an imposed pressure gradient and
which produce fog in a given region. Typical climatol-
during the warm season, may last a few hours.
ogies assess fog frequencies according to the time of
day and year (or by season), intensity, duration, and
sometimes according to fog occurrence at multiple
Prediction sites. Unfortunately, the climatological approach
tends to oversimplify fog prediction by categorizing
The fog process is quite dynamic and requires a great
events as fog ‘types’with little regard to fog evolution.
deal of physical knowledge and observation to be
In addition, the results are biased for the site at which
understood completely and predicted successfully. In
the climatology is based and can thus lead to predic-
practice however, this is difficult given the limited
tions of no fog for a region which may fail.
nature of our present understanding of fog and the lack
The use of conditional climatology adds some value
of real-time observation, quantification, and modeling
beyond simple climatology in that persistence and
of the chemical and microphysical behaviors that are
contributing factors associated with a fog event can be
involved. This is made more difficult by the lack of
anticipated. Yet this approach is also limited in that it
precise observation and modeling of the detailed
does not address fog dynamics adequately and relies
surface characteristics, the distribution of moisture,
heavily upon the data available. In total a forecaster
temperature, and their interaction. However, if the
may be able to assimilate climatological information
conceptual model of a family of fogs based on the
and estimate local variations with some experience
principles previously described is applied, the chances
for a region, but the process is pragmatic rather
for improved fog prediction may be increased. Ulti-
than scientific and is only slightly more skillful.
mately even a microscale observation network would
Climatological predictions have only limited antici-
not be dense enough to provide essential details to
patory value: knowing the types of situations that
improve prediction. Instead the further application
favor fog.
and refinement of remote sensing tools and numerical
modeling will be necessary to better forecast fog
Numerical Guidance
occurrence, extent, intensity, and duration in a wide
variety of circumstances. Another approach to fog forecasting is through the use
Current forecast practices for fog include climato- of numerical guidance, both raw and processed. Raw
logical approaches, numerical guidance, observation- data from an operational model may be plotted or
al methods, modeling, and other methods (e.g., examined across a region and yield specific informa-
statistical or decision-tree methods and Artificial tion with regard to moisture and cooling with regard
Intelligence - or Expert Systems). Common to each to fog formation. However, model output has a variety
FOG 791
of errors, is often available in only limited time and Modeling

space resolution, and typically does not adequately
Many modeling efforts have been made to simulate as
depict the boundary layer structure or behavior. When
well as to produce a fog environment and fog droplets.
the raw data are processed to produce graphical and
Presently, no operational model is capable of com-
diagnostic analyses, surface and boundary layer plots,
pletely generating these in real time for predictive use.
or to generate statistical forecast guidance, it is of
Modeling has more recently focused on drop size
much greater value. The graphical diagnosis of the
distributions, deposition, and re-creation of the fog
local environment according to model output is useful
environment as well as the physical representation of
in better identifying regions potentially favored for fog
fog and the fog process. Limited studies, including
development (e.g., moisture maxima, cool air pockets)
statistical prediction, have focused on the prediction of
and thus can aid a forecaster in refining a prediction of
the visibility restrictions of fog. The First International
occurrence and extent. This provides a mesoscale
Conference on Fog and Fog Collection is indicative of
prediction which can be partially verified through
the new and renewed importance of fog study.
surface observations and satellite imagery.
Much work is now focused on collection tech-
When used to generate statistical guidance, model
niques, particularly with regard to fog chemistry and
output helps a forecaster assess the confidence level of
its impacts with regard to vegetation. In this regard,
fog formation and extent and allows some speculation
fog modeling has served as an extension of agro-
as to its intensity and duration. This guidance makes
meteorological and climatological study as well as
use of both current observations and model predic-
micrometeorology and cloud physics. Some profes-
tions. For operational models with output statistics
sional consultants and companies now offer fog
from 12 to 60 h into the future, there is some predictive
forecast products, and even assessments, many of
value and skill improvement over climatology in the
which focus on predicting the occurrence and intensity
anticipation of fog. However, most statistical guidance
of fog for a specified location. A variety of other
used in this manner is derived from regression
methods, including statistical or decision-tree meth-
relations that show current observations (or simply
ods and Artificial Intelligence (or Expert Systems),
persistence) to be the most important factor in fog
have found usefulness in the study of fog.
prediction. Thus numerical guidance is of best value in
anticipating fog during the first several hours, or
anticipating the synoptic setting conducive to fog Benefits, Mitigation, and Special
several days in advance.
Cases
Although a variety of severe impacts have occurred
with fog events around the world, efforts to mitigate
Observational
these impacts have been of limited success in part due
The observational approach to fog prediction is to the nature of human response (e.g., driving more
extremely limited in an anticipatory sense, but tends slowly through fog). However, fog has also yielded
to have the greatest skill in terms of fog extent, some beneficial effects in terms of esthetics and
intensity, duration, and ultimate dissipation. This agricultural application. Fog has also been the subject
means that a forecaster has synthesized the prevailing of literature and film and often accompanies the
synoptic environment and its interactions with the depiction of paranormal and Halloween activities.
boundary layer to understand why fog is present (or Fog systems are used routinely to create fog and other
expected) and why it will continue to persist or special effects by the film industry. There has even been
dissipate. Essentially a forecaster in this situation has a survey that found 28% of people engaged in sexual
developed and applied a conceptual model frame- relations more than once when it was foggy as
work, based on theory, knowledge, and experience compared with only 11% who did when it was sunny.
which provides specification that no other current Several facets of fog illustrate how its significance
method can match. The observational approach extends beyond a simple reduction in visibility. For
involves a wide synthesis of sensible surface weather example, fog water use for deposition collection is
elements, middle and upper atmospheric data and critical to agricultural operations in Chile, Mexico,
analyses, and a review of satellite (and perhaps even and other regions of the world. Fog water may be of
radar) images. Practical application of radiative laws use in providing a potable water supply for others.
and parcel theory through the use of sounding Most often fog water is collected for agricultural
diagrams (e.g., a skew-T chart, and even isentropic applications and has been studied with regard to its
analysis) allow a forecaster to ‘run’ a conceptual effects on the growth of giant Redwood trees in
model of fog and to make a successful prediction. California (USA). Study also indicates that fog may
792 FOG
play a significant role in the physical interactions Balloons; Buoys; Kites; Rockets. Observations for
found within plant canopies and their physiological Chemistry (/n Situ): Chemiluminescent Techniques;
and growth conditions. These are also related to fog Gas Chromatography; Ozone Sondes; Particles; Reso-
water pH and the production of acid fogs. These and nance Fluorescence; Water Vapor Sondes. Observations
other issues are being explored through various studies for Chemistry (Remote Sensing): IWFIR; Lidar; Micro-
wave. Operational Meteorology.Orographic Effects:
around the world.
Lee Cyclogenesis; Mountain Waves and Stratospheric
There also exist programs and methods for the Chemistry. Parameterization of Physical Processes:
dispersion, or enhancement, of fog - particularly at Clouds. Radar: Incoherent Scatter Radar; MST and ST
airport locations - and the reduction of acid fogs. The Radars and Wind Profilers; Precipitation Radar; Synthetic
basic methods tend to focus on heating of the fog layer Aperture Radar (Land Surface Applications). Radiative
(to evaporate droplets), downwash mixing (to entrain Transfer: Absorption and Thermal Emission; Cloud-ra-
drier air), hygroscopic treatment (e.g., ice seeding) to diative Processes; Non-local Thermodynamic Equilibrium;
precipitate out, and the use of fog breaks (passive Scattering. Satellite Remote Sensing: Aerosol Meas-
control) to prevent formation or movement into an urements. Static Stability. Synoptic Meteorology:
area. The most effective methods tend to be those that Forecasting; Weather Maps. Thermodynamics: Moist
match the natural dissipative factors (i.e., mixing and (Unsaturated) Air; Saturated Adiabatic Processes. Turbu-
lence and Mixing. Turbulent Diffusion. Weather
evaporation) and that promote improved visibility
Modification: Inadvertant. Weather Prediction: Adap-
within an hour. Unfortunately most techniques are tive Observations; Data Assimilation; Ensemble Prediction;
costly, labor intensive, esthetically intrusive, and must Regional Prediction Models; Seasonal and lnterannual
be maintained until natural processes are capable of Weather Prediction; Severe Weather Forecasting. Wind
continuing dissipation (often for at least several Chill.
hours). These measures are also impractical for
roadways and therefore drivers instead rely on vehicle
fog lights and fog-free lenses or shields. The enhance- Further Reading
ment of fog, although seemingly undesirable, is an Anderson JR (1985) Economic impacts. In: Houghton DD
industry with commercial application for party sup- (ed.) Handbook of Applied Meteorology. New York:
pliers and movie studios. Wiley.
Air Weather Service (1979) General Aspects of Fog and
See also Stratus Forecasting. USAF AWS TR 239.
Croft PJ, Pfost R, Medlin J, and Johnson G (1997) Fog
Agricultural Meteorology and Climatology. Air-Sea forecasting for the Southern Region: a conceptual model
Interaction: Freshwater Flux; Gas Exchange; Momen- approach. Weather Forecasting 12: 545-556.
tum, Heat and Vapor Fluxes; Sea Surface Temperature; Eagleman JR (1991) Air Pollution Meteorology. Trimedia
Storm Surges; Surface Waves. Anticyclones. Arctic Publishing Company.
Haze. Aviation Weather Hazards. Boundary Layers: George JJ (1960) WeatherForecasting for Aeronautics. New
Coherent Structures; Complex Terrain; Convective Bound- York: Academic Press.
ary Layer; Modeling and Parameterization; Neutrally Strat- Houze RA Jr (1993) Cloud Dynamics. New York: Academic
ified Boundary Layer; Observational Techniques In Situ; Press.
Observational Techniques-remote; Ocean Mixed Layer; Mason J (1982) Physics of radiation fog. Journal of the
Overview; Stably Stratified Boundary Layer; Surface Layer. Meteorological Society of Japan 60: 4 8 6 4 9 9 .
Chemistry of the Atmosphere: Chemical Kinetics; Gas Mayer WD and Rao GV (1999) Radiation fog prediction
Phase Reactions; Ion Chemistry; Laboratory Studies; using a simple numerical model. Pure and Applied
Principles of Chemical Change. Cloud Chemistry. Cloud Geophysics, in press.
Microphysics.Coastal Meteorology.Deposition. Di- Schemenauer RS and Bridgman H (ed.) (1998) Proceedings
urnal Cycle. Humidity Variables. Hydrology: Ground of the First International Conference on Fog and
and Surface Water; Modeling and Prediction; Overview. Fog Collection, 19-24 July 1998, Vancouver, British
IsentropicAnalysis. Land-Atmosphere Interactions: Columbia, Canada.
Canopy Processes; Overview; Trace Gas Exchange.
Lidar: Atmospheric Sounding Introduction; Backscatter;
Online reference sources
DIAL; Doppler; Raman; Resonance. Mesoscale Mete-
orology: Overview. Microclimate. Numerical Models:
Chemistry Models; Methods. Observation Platforms: -
http://meted.ucar.edu (COMET Module Radiation Fog)
http://www.cco.net/ trufax/fluoride/fog. html
FORENSIC METEOROLOGY 793
L E Branscome, Climatological Consulting transport of odors from illegal drugs, the death of
Corporation, Palm Beach Gardens, Florida, USA children trapped inside motor vehicles exposed to
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. intense sunlight, and various types of criminal negli-
gence.
In a broader sense, the work of a forensic meteoro-
Introduction logist is not necessarily limited to civil litigation and
criminal prosecutions. For example, a windstorm that
Forensic meteorology is the study of past weather causes property damage may result in a claims dispute
events for the primary purpose of assisting a court of between an insured party and its insurer. A meteorol-
law in adjudicating disputes in which weather may ogist may be hired by one of the parties in the dispute
have been a factor. The task of a forensic meteorologist to ascertain the peak wind speed during the storm. The
is to diligently investigate the weather conditions conclusions of the meteorologist may be sufficient to
pertinent to the litigation and objectively present to the allow the client to make a determination about the
court the most accurate description possible of the claim and settle the dispute with the opposing party,
relevant meteorological events. Expert testimony by a thus avoiding the effort and expense of a lawsuit.
meteorologist often provides crucial information in Similarly, a forensic meteorologist may be hired to
various types of civil and criminal proceedings. provide information and expert opinion about past
Civil litigation is often initiated when a personal weather events to parties involved in administrative or
injury or financial loss occurs, or when a contractual regulatory disputes with government agencies.
dispute cannot be resolved. In many cases the weather
is a contributing factor, or even the primary cause of
the loss or dispute. For example, weather frequently
contributes to maritime cargo losses, aircraft and The Practice of Forensic Meteorology
highway accidents, property damage, construction
General Procedures and Considerations
delays, air pollution emergencies, slip and fall injuries,
and structural failures. The party sustaining the loss or Technical experts are often retained by attorneys
injury may seek compensation through the legal representing the parties involved in a lawsuit. The
system from another party who is alleged to be immediate purpose of retaining a meteorologist is to
responsible for the incident. A judge, jury, or other obtain information and expert opinion about the
adjudicative body must determine the facts of the case weather conditions so that the attorney can evaluate,
and make a decision in favor of one of the parties in the with assistance from the expert, the importance of
litigation, either the plaintiff or the defendant. If the weather to the case. The attorney will consider how
weather was a factor in the incident or dispute on the weather facts and opinions may support or refute
which the lawsuit is based, then expert testimony theories for the causation of the loss that led to the
about the weather conditions relevant to the case must litigation. If the meteorological information and
be presented to the adjudicative body in a courtroom opinions are relevant and helpful, the weather expert
or other judicial setting. Meteorological analysis and may be asked by the attorney to prepare a formal
testimony are sometimes necessary simply to rule out report, evaluate the reports and opinions of the
weather as a contributing factor. Together with other meteorologist (if any) working for the opposing legal
facts, testimony, opinions, exhibits, and evidence, the counsel, assist the attorney in evaluating the merits
adjudicative body considers the relevant weather and demerits of the case from a technical perspective,
evidence and opinions in making its determination. develop exhibits for presentation at trial, and provide
In addition to civil litigation, testimony about the expert testimony and opinions at deposition and trial.
weather is occasionally required in criminal proceed- The expert investigating and testifying about the
ings. For example, the ability of an eyewitness to weather should be a meteorologist with sufficient
identify an alleged assailant in low natural lighting is knowledge, skill, experience, and education to offer an
occasionally an issue in criminal trials. In such cases a opinion about the particular weather conditions
meteorologist may be called to testify about the sky related to the litigation. Although meteorologists are
conditions, phase of the moon, or sunrise and sunset not usually granted professional licenses like engineers
times. Other criminal proceedings have depended on or architects, the American Meteorological Society,
meteorological analysis and testimony as it relates to for example, has a certification program for consult-
the decomposition of the bodies of murder victims, the ants, identified as Certified Consulting Meteorologists,
794 FORENSIC METEOROLOGY
that involves extensive testing and board review. forensic meteorologist cannot offer a reliable opinion
Many attorneys find that the certification assists until weather data are obtained and reviewed. The
them in identifying qualified experts. The forensic nature and quality of the data needed to formulate
meteorologist must also have a high degree of integrity expert opinions depends on the nature of the case and
and composure under pressure, so that the expert testi- the relative importance of weather to the arguments
mony that he or she gives in the courtroom is and theories in the litigation. In the case of a slip and
trustworthy, unbiased and professional. fall injury, it may only be necessary to obtain hourly
The admissibility of expert opinions in the federal rainfall data from a nearby weather station. In the case
courts of the United States is governed by the ‘Daubert of a fatal aircraft accident, a wide variety of data may
test’ in which the judge assesses whether the reasoning be needed to formulate a clear and complete descrip-
or methodology underlying the opinion is scientifically tion of the weather conditions at the time and location
valid and the offered testimony is relevant to the case. of the accident. The data may include surface weather
Many non-federal courts still adhere to the earlier observations, upper-air weather charts, satellite and
‘Frye test’ in which expert opinions must be based on radar images, lightning strike data, aviation weather
principles and techniques that are ‘generally accepted’ forecasts, and pilot reports of weather conditions
as reliable within the relevant scientific community. aloft.
Before accepting work as an expert, the forensic Forensic meteorological investigations often de-
meteorologist should inquire about the nature of the pend on the analysis of severe weather phenomena
case and the parties and attorneys involved, in order to that have significant variations over small temporal
avoid possible conflicts of interest. A conflict may arise and spatial scales. For example, a severe thunderstorm
if the expert has had prior exposure to the facts and can create strong winds and large hail that cause
legal arguments of the case or if the opposing attorney isolated property damage in a few seconds to minutes.
is a regular client of the expert, particularly for cases of The ability to reconstruct past weather for forensic
a similar nature. In some instances the expert should purposes has been greatly enhanced by recent advanc-
decline to accept the work if the case involves a special es in observational techniques that target mesoscale
area of meteorology in which the expert has limited weather phenomena. For example, the National
knowledge or training. For example, an expert in air Weather Service of the United States operates a
pollution meteorology may refer the client to an network of Doppler weather radar stations that
aviation weather expert if the case involves an aircraft provide nearly complete and continuous coverage of
accident in a severe thunderstorm. the country. Similar radar networks are found in other
After accepting the case, the weather expert should countries. The radar monitoring of severe weather and
obtain all the information that the attorney possesses the archiving of the radar data have allowed forensic
that might be relevant to the meteorological investi- meteorologists to develop more accurate and detailed
gation. This information will include, at a minimum, descriptions of severe weather events.
the date, time, location, and nature of the event that Similarly, the weather is monitored continually
led to the litigation. It may also include reports and from space by geostationary and polar-orbiting satel-
statements describing the event that were prepared by lites. Improvements in the resolution of the on-board
investigative agencies, law enforcement officers, or sensors and the addition of more observing channels at
eyewitnesses. For example, the National Transporta- infrared wavelengths allows forensic meteorologists
tion Safety Board of the United States produces to investigate past weather events in greater detail. For
investigative reports that provide information about example, combinations of data from different infrared
the particulars and contributing factors in aircraft channels permits the identification of potential avia-
accidents. The testimony of eyewitnesses regarding the tion hazards such as fog or airframe icing. Other
weather conditions can sometimes provide descriptive satellites are capable of measuring ocean wave heights
details about a weather event that would not be and estimating surface wind speeds at sea, which is
evident in official meteorological data. The expert may valuable information for the investigation of cargo
also find it useful to visit the site of the incident to and ship losses during intense storms.
assess the importance of exposure, surrounding ter- Other advances in meteorology are related to
rain, and site orientation to the analysis and interpre- improvements in computer models of the atmosphere.
tation of the meteorological events. The ability to computationally simulate or analyze
mesoscale weather phenomena has greatly improved.
The global re-analysis of historical weather data with
The Value of Meteorological Data and Analysis
computer models that are also used in operational
Although the attorney retaining the expert may have a weather forecasts has provided another new tool for
substantial amount of information about the event, a the forensic investigator.
Acquisition of Weather Data the end user. Even large data sets, such as satellite and
radar data, can be extracted from computer tapes at
The meteorological data used in forensic work are government archives and delivered to the end-user by
commonly stored in government archives and can be file transfers over the Internet. Many data providers
accessed by the general public for usually modest fees. are also providing large data sets on CD-ROM and
The primary sources of data in the United States are the DVD. Large quantities of data can thus be retrieved
National Climatic Data Center, which is part of the and analyzed on the computer workstations of foren-
National Oceanic and Atmospheric Administration, sic meteorologists. Waiting periods for data delivery
and Regional Climate Centers and State Climatolo- have generally been reduced from weeks and months
gists. Similar government archives are operated by the to minutes and days. While the time required to
weather services of other countries. In addition to acquire and analyze the data on a per-unit basis has
national government sources, a number of local been substantially reduced, the quantity and variety of
government agencies, universities, and private com- data available for forensic work has increased dra-
panies archive meteorological data and operate matically, so that the total amount of effort in a typical
weather observing stations in special monitoring investigation has not diminished. Even so, improved
networks. While anyone involved in litigation can access to data often allows experts to provide their
obtain weather data, a forensic meteorologist is clients with preliminary analysis and opinions in a
usually more efficient in selecting and obtaining the time frame shorter than previously possible.
relevant data and organizing it into a coherent source The physical certification of digitized data is often
of information for the analysis of weather events. impractical, if not impossible. The authenticity of
The collection, archiving, and retrieval of meteor- digitized meteorological data as admissible trial evi-
ological data had undergone considerable change in dence is usually not problematic, provided the testi-
recent years, affecting the practice of forensic meteor- fying expert can demonstrate that the data was
ology. In the past most meteorological data were obtained from reliable sources and is customarily
recorded and archived on paper or film. For example, used by meteorologists in their work.
wind data were recorded on instrument recorder
charts, radar scope images on film, and surface
Reports and Testimony
observations on handwritten forms. The data and
analysis techniques were often limited in quantity and After the weather data has been obtained and ana-
complexity. Forensic meteorologists usually ordered, lyzed, a report is made to the client either verbally or in
and in many cases continue to order, paper copies of writing. A formal report that includes expert opinions is
data and images that bear a fixed certification and seal sometimes required by the client or by the judicial rules
of authenticity from the government agency providing governing the litigation. The report becomes a basic
the data. The physical seal and certification of the data reference document for future testimony by the expert.
copy are still regularly used as a means of authenti- Before a case goes to trial, a discovery period occurs
cating the copy for admission into evidence at trial. when the attorneys in the litigation request copies of
Production and delivery of the hard-copy data and the documents and information used or prepared by
images by the data supplier often resulted in waiting the other side’s technical experts. The information
times of a few weeks to months. Analysis of the data by includes the data gathered and any written reports
a forensic meteorologist was labor-intensive, particu- prepared by the weather expert. Furthermore, the
larly for cases in which years of data had to be expert is questioned in a deposition by the opposing
reviewed to determine normal climatological condi- legal counsel to discover the expert’s opinions and the
tions, since the data had to be extracted from the foundational basis for those opinions. The discovery
certified paper forms. and evaluation of the weather data and expert opinions
The recording, storage, and delivery of data on sometimes encourage a settlement between the oppos-
paper, microfiche, and film is rapidly diminishing. ing parties, particularly if the weather was the primary
Low-cost computer storage of data, data-intensive cause of the loss or dispute that initiated the lawsuit.
observational systems such as Doppler weather radar As the trial approaches, the expert prepares exhibits
and multichannel satellite sensors, complex computer that display the data and information relevant to the
models of the atmosphere and ocean, digital surface issues of the case. The exhibits are often directly
weather sensors, computer networks, and the Internet derived from diagrams, tables, and images found in the
are radically changing the work of the forensic expert’s report. The availability of digitized data has
meteorologist. Many government and private data led to improvements in the presentation of meteoro-
sets are now maintained on computer servers on the logical information in the courtroom. Radar and
Internet and are easily accessible at low or no cost to satellite digital images and computer simulations of
weather events can easily be annotated for trial 4184 were located at the top of the fuselage, the pilots
presentation and stored on a CD-ROM. Using a high- were unable to see the formation of ice on top of the
resolution projector attached to a laptop computer, the wings. Relatives of the deceased passengers sued the
testifying expert can display the images in the court- airline partly over the alleged inaction of the pilots with
room using animation sequences, stop-action frames, respect to the hazardous icing conditions. Attorneys in
zoom views, and various enhancements. Presentations the litigation retained aviation weather experts and
of this kind allow the jury or judge to arrive at a better much of the meteorological investigation focused on
understanding of the weather conditions associated the nature, timing, and severity of the icing conditions.
with the event that led to the litigation. In fact, many lay The microphysics of supercooled drizzle drops was an
people who serve on juries see animated satellite and important aspect of the investigation, since the atmos-
radar images on the weather segments of television pheric conditions were indicative of the presence of
news shows and are somewhat familiar with basic such drops in the area of the holding pattern. These
meteorological concepts related to these observations, drops can be particularly hazardous to aircraft because
prior to seeing such data in the courtroom. they can flow over and freeze behind the icing-
The expert should assist the attorney in preparing prevention devices on the leading edge of the wings
for the expert’s trial testimony. The attorney needs to and create ice formations that seriously disrupt air flow
have a thorough understanding of the expert’s opin- over the wings. Data and images from the National
ions, along with the basis for those opinions, so there Weather Service Doppler radar near Chicago (see
are no surprises in the courtroom. The testimony and example in Figure 1)were carefully studied to deter-
opinions of the expert, together with the trial exhibits, mine when the aircraft most likely encountered super-
are presented during the direct examination by cooled drizzle drops during the holding pattern. The
the attorney who retained the expert. The expert has meteorological analysis was partly intended to deter-
the responsibility of clearly and simply explaining the mine whether any early visible indications of icing were
weather elements of the case to the judge or jury. present that might have given the crew an opportunity
Cross-examination of the expert by the opposing to respond and take evasive action. A substantial
attorney is standard practice. If some points of the amount of time and effort was spent on the preparation
testimony need to be clarified following the cross- of trial exhibits related to the weather testimony, but
examination, the attorney who retained the expert has the litigation was settled shortly before trial.
an opportunity to ask additional questions in redirect Commercial and private aircraft are also subject to
examination. Once the expert’s courtroom testimony hazardous weather associated with thunderstorms.
is finished, the expert’s participation in the litigation The occurrence of microbursts (intense small-scale
usually ends, unless appeals in the matter require a re- downbursts in the decaying stages of a thunderstorm)
hearing of the expert’s opinions. are of particular concern on approach or departure
from an airport. Large changes in wind speed and
Examples of Meteorological direction across a microburst can create sudden and
unexpected changes in lift. The ability of the pilots to
Investigations
recognize and avoid the hazardous conditions associ-
A tragic aviation accident, in which weather played a ated with a microburst are usually at the center of the
critical role, was the crash of a commercial aircraft, litigation in such cases. Microbursts were the primary
American Eagle flight 41 84, near Roselawn, Indiana, cause of several major accidents, such as the Delta Air
on 31 October 1994. The aircraft was in a holding Lines crash at Dallas, Texas, on 2 August 1985 and the
pattern at 3000 m altitude for about 30 minutes while US Airways crash at Charlotte, North Carolina, on
waiting for clearance to land at Chicago’s O’Hare 2 July 1994. Litigation related to these crashes relied
airport. Shortly after it was released by air traffic not only on detailed analysis of the meteorological
controllers from its holding pattern, it descended data but on eyewitness statements from pilots of other
toward 2500m, at which time the pilots could no aircraft and air traffic controllers regarding the weather.
longer control the aircraft and it crashed into a field, Severe turbulence can cause passenger injuries on
killing all 68 people on board. commercial flights and such incidents often lead to
Airframe icing can significantly degrade the per- lawsuits against the airlines operating the flights.
formance of an aircraft. Based on the findings of Typically the plaintiff will allege that the pilots should
detailed meteorological investigations of the Rose- have anticipated and avoided the turbulence. Aircraft
lawn accident, the cause for the loss of control is operation in and around thunderstorms increases the
strongly suspected to have been the accumulation of a likelihood of a severe turbulence encounter. Avoidance
ridge of ice behind the leading edge of the wings. of thunderstorms is made possible through direct
Because the wings of the aircraft operating as Flight visual observation, on-board weather radar, and
Sometimes unofficial weather measurements are at

least as valuable as official records. The construction
of a natural gas pipeline in South America was
interrupted by heavy rain and destructive flooding in
April 1998. The construction company was insured
against losses of this kind and filed a large claim for
extensive damage and delay associated with the event.
The insurer was uncertain about paying the claim since
rainfall totals from official weather stations in the
region were not sufficiently large to have caused the
flood damage claimed by the construction company.
The insured party produced an unofficial measure-
ment of 325 mm of rain in one day from a hotel located
near the project. The rainfall reading at the hotel was
many times larger than any of the official readings. In
order to resolve the validity of the claim, the insurer
hired a meteorologist to investigate the weather
conditions, as well as forensic civil engineers who
inspected the physical damage. An examination of
weather satellite data demonstrated that an isolated
severe thunderstorm with exceptionally high cloud
top heights (Figure 2 ) did occur along the segment of
the pipeline where the most severe damage was found.
A range of probable rainfall amounts was inferred
from satellite data and was generally consistent with
the high rainfall reading at the hotel. Although the
heavy rainfall did not cover as much of the pipeline
project as claimed by the construction company, the
Figure 1 Track of American Eagle Flight 4184 during part of its
insurer concluded that at least part of the claim was
holding pattern near Chicago, Illinois, on 31 October 1994, plotted
on a National Weather Service Weather Surveillance Radar valid.
reflectivity image. Time associated with each aircraft position is in In another example, the owner of a Florida resort
minutes after 21 :00 Universal Coordinated Time. The aircraft hotel claimed that extensive water damage to the
occasionally encountered areas of light precipitation, most likely interior of the building was the result of roof leaks
supercooled drizzle drops, during its hold.
during a very heavy rainfall event. The hotel owner
filed a lawsuit against a construction company that
pre-flight and in-flight weather briefings. Other at- had repaired the roof shortly before the alleged
mospheric phenomena, such as breaking mountain damage. Attorneys for the construction company
waves and strong vertical or horizontal shear in jet retained a forensic meteorologist to determine wheth-
streams, can also cause severe turbulence, but such er a heavy rainfall event actually occurred at the
conditions are usually more difficult to anticipate and property during the period in which the water damage
avoid than convective turbulence. occurred. Rain gauges in the area did not confirm the
The examination of satellite, radar, and upper-air occurrence of a rain event of the magnitude alleged by
data is usually a critical component in meteorological representatives of the hotel. The amounts were far
investigations of turbulence incidents. Two passengers below the quantity of water associated with the
who allegedly suffered turbulence-related injuries interior damage. Attorneys for the hotel asserted that
during a flight across the south-western United States since isolated heavy rainfall often occurs in Florida,
in March 1996 sued the airline. The airline hired an the rain gauges simply missed the ‘deluge’ at the hotel.
aviation meteorologist who concluded that there was Examination of data from the National Weather
no meteorological evidence for the turbulence en- Service Doppler radar station that monitors the region
counter. However, the meteorologist working for the demonstrated conclusively that the rainfall amounts at
plaintiff obtained satellite photographs that showed the hotel were similar to the amounts measured at the
the existence of strong mountain waves at the time and rain gauges. The case reached trial and the meteorol-
location of the incident, consistent with the plaintiffs’ ogist testified about the rainfall and radar analysis.
allegation. The case was settled shortly after the report The jury found in favor of the defendant, i.e.,
of the plaintiff‘s expert was submitted in the litigation. the construction company, concluding that it was
Figure 2 Color-enhanced infrared satellite image from Geostationary Operational Environmental Satellite 8 on 8 April 1998 at 11:45
UniversalCoordinatedTime near Puerto Suarez, Bolivia. Lower temperaturesindicate higher cloud tops. Notethe very cold temperatures
for the heavy thunderstorm near the center of the image. Heavy rain from this thunderstorm caused flooding and extensive damage to a
pipeline construction project.
impossible that rain could have caused the water of meteorological research and testimony with respect
damage to the interior of the hotel. to litigation.
See also
The Outlook for Forensic Meteorology Air-Sea Interaction: Surface Waves. Aviation Weath-
The density and value of physical property is steadily er Hazards. Clear Air Turbulence. Convective
increasing in regions especially susceptible to weather Storms: Convective Initiation. Flooding. Fog. Gust
damage, such as coastal areas. While significant Fronts. Hurricanes. Lee Waves and Mountain
improvements are being made in aviation safety Waves. Microbursts. Radar: MST and ST Radars and
Wind Profilers; Precipitation Radar. Satellite Remote
procedures, particularly related to the distribution Sensing: Surface Wind. Synoptic Meteorology:Weath-
and analysis of weather information for pilots and air er Maps. Weather Prediction: Severe Weather Fore-
traffic controllers, the frequency of air travel is also casting.
expected to grow rapidly. Furthermore, judicial sys-
tems are generally increasing the monetary value
assigned to human life and health. As a result, the
Further Reading
amount of litigation related to weather is likely to
expand, increasing the need for forensic meteorolog- Bradley MD (1983) The Scientist and Engineer in Court,
ical services. Water Resources Monograph Series, 8. Washington DC:
With the expectation of additional enhancements in American Geophysical Union.
the quality and quantity of data from remote sensors Bronstein DA (1993) Law for the Expert Witness. Boca
Raton, FL: Lewis Publishers.
and computer models of the atmosphere and ocean, Falconer PD and Haggard WH (1990) Forensic meteorol-
the work of the forensic meteorologist will become ogy. In: Wecht CH (ed.)Forensic Sciences, Ch. 35B. New
more complex and require continual upgrades in York: Matthew Bender.
professional skills and knowledge. On the other hand, National Climatic Data Center (1999) Weather records in
advances in observational and analytical techniques litigation. Environmental Information Summaries, vol.
will likely also reduce uncertainties regarding past C-1. Asheville, N C : National Oceanic and Atmospheric
weather conditions and, therefore, enhance the value Administration, National Climatic Data Center.
FRONTOGENESIS 799
W Blumen, University of Colorado at Boulder, air parcels and their surrounding environment. Close
Boulder, CO, USA to the Earth, where p x po and where vertical
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. displacements are relatively small, 0 may be replaced
by T. Frontogenesis depends on both the kinematic
properties of the wind field, and on the dynamic and
thermodynamic processes associated with the under-
Introduction lying physics of atmospheric flow. Consider, for
Frontal formation is called frontogenesis and decay or example, an increase in the thermal gradient in the
dissolution is referred to as frontolysis. The intensifi- y-direction from the advection of temperature by the
cation or weakening of the thermal gradient is usually confluent wind field in Figure 1A. Air parcels, confined
considered to be the measure of frontogenesis. The to a horizontal plane, are alternately compressed
instantaneous rate of change of the cross-front tem- along the y-axis and then stretched along the x-axis.
The rate at which the potential temperature gradient,
perature gradient may be expressed as a diagnostic
relationship that encompasses all processes that con-
-ae/ay, increases, following the horizontal motion of
tribute to frontogenesis. Contributions are associated an individual fluid element, D H / D ~is, related to this
with particular configurations of both horizontal and confluent wind field by
vertical wind fields, cloud cover, which can allow a
selective heating or cooling of the Earth’s surface,
either ahead or behind the front, evaporation and
condensation, among other non-adiabatic (i.e., diaba- where a denotes the rate of confluence. This expression
tic) processes and surface heterogeneity, particularly shows that, if rx may be treated as a constant, the
coastal zones. Each contribution may be evaluated potential temperature gradient increases exponential-
independently by means of observations and data ly with time. It is not, however, sufficient to analyze
from numerical model simulations. The relative frontogenesis in this simple manner, because an
importance of each contribution to frontogenesis increase in the horizontal temperature gradient
depends on the type of front, surface-based or upper- requires an increase in the vertical shear of the
air, and its place in the time evolution from inception along-front horizontal wind speed. This relationship
to decay. Model simulations, which are based on a full between the horizontal thermal gradient and the
set of predictive equations, are required for a funda- vertical wind shear is referred to as thermal wind
mental understanding of frontogenesis as an evolu- balance. The maintenance of thermal wind balance
tionary process that involves interactions between requires the development of a thermally direct circu-
different physical processes. lation cell, referred to as a secondary transverse
circulation and displayed in Figure 2. The increase of
Frontogenesis the thermal gradient from confluence is partially
reduced by adiabatic cooling of air which rises on
Frontogenesis is usually defined as an increase in the the warm side of the front, and adiabatic warming of
three-dimensional spatial gradient of the potential air which sinks on the cold side of the front. The
temperature 9. The potential temperature 9 is defined increase in the vertical wind shear, from the net
as increase in the thermal gradient, is associated with the
horizontal branches of the circulation cell. The
Coriolis force deflects motion into the section aloft,
and out of the section below, to effect the necessary
where T is the absolute temperature, p is the pressure, increase in the vertical shear.
po = 1000 hPa is a constant reference pressure, and A shortcoming of the frontogenesis model, ex-
R / c , denotes the ratio of the gas constant to the pressed by eqn [2], is the consideration of only a
specific heat of air at constant pressure. The potential confluent wind field to effect frontogenesis. Typical
temperature is a convenient thermodynamic variable values of a, observed near the Earth’s surface, reveal
for the study of frontogenesis, because it is conserved that more than 2 days are required to increase the
during adiabatic displacements of dry air parcels, and temperature gradient by an order of magnitude. Yet,
changes in 8 may then be associated with diabatic comparable increases have been observed, in some
processes. The latter involve heat exchange between instances, during a period of 1 2 h or less. One way to
800 FRONTOGENESIS
e+Ae e e-Ae
I I I
I I I
Z I I I
4 -1 I I-
l l
I I -
I l l
+ I I I +
I I I
I I I
I I I
Figure 1 Frontogenesis mechanisms: (A) horizontal deformation by confluence and by convergence (thick arrows); (B) horizontal
deformation by horizontal shear and rotation; (C) vertical deformation; (D) vertical divergence; (E) tilting; (F) differential diabatic radiative
heating related to cloud cover.
resolve this discrepancy is to allow the convergent or frontogenesis function, defined as

wind field, also shown in Figure l A , to collaborate
with confluence to enhance frontogenesis. A modified
F = DJVQJ/Dt 141
version of eqn [ 2 ]may be expressed as
where F is the rate of change of the three-dimensional

131 potential temperature gradient ( V is the gradient
operator) following the motion of individual air
where 6 denotes horizontal convergence. The incor- parcels, indicated by the operator D/Dt. Frontogen-
poration of 6 into this simple archetype provides a esis, as defined by F, is produced by physical processes
feedback effect, because the convergent wind field that do not depend on the translation of the front.
increases with the rate of frontogenesis. This mecha- Therefore, the relationship between F and various
nism, among others, contributes to observed values of kinematic and thermodynamic frontogenetical proc-
rapid frontogenesis at lower levels of the atmosphere. esses is a convenient means to describe the relative
Kinematic contributions are not restricted to con- importance of contributions to frontogenesis, even if
fluence and convergence. Consider the frontogenetical evaluation is not made moving with the front.
FRONTOGENESIS 801
0 + 3A0 rotates isotherms in the vertical cross-section, to

, ,0+2AO
increase the temperature gradient along the y-axis.
Finally, cloud cover, precipitation, and surface heter-
ogeneity are all factors that contribute to diabatic (V)
- A0 frontogenesis or frontolysis. The principal physical
processes included in the diabatic contribution are
associated with differential radiative heating, as illus-
trated by the extensive cloud shield in Figure lF, and by
latent heating and evaporative cooling. A secondary
0 - 2A0 factor is differential heating across a frontal zone
t Y
produced by frictional dissipation of kinetic energy
close to the Earth’s surface. The effect of this process
on frontogenesis is unclear since measurements are
Figure 2 Thermally direct circulation cell. The dashed lines are lacking. Turbulent transfer of heat acts to reduce
isotherms of potential temperature, and the horizontaljet flow is into temperature gradients, but molecular heat diffusion is
the section at low levels and out of the section at upper levels. not considered to make a significant contribution to
frontogenesis.
The relative importance of these contributing fac-
A mathematical representation of the atmospheric tors will be examined for three types of frontogenesis.
processes that produce an instantaneous value of F They are surface frontogenesis, coastal frontogenesis,
provides a diagnostic expression, called the fronto- and upper-level frontogenesis. Each may be delineated
genesis equation. The essence of this equation may be with synoptic-scale observations that may be supple-
represented as mented with special observational platforms, such as
instrumented aircraft, to explore relatively small-scale
F = horizontal deformation (I) details that cannot be measured by conventional
+ vertical deformation (11) means.
+ vertical divergence (111)
+ tilting (IV) + diabatic (V)
Surface Frontogenesis
The first four contributions relate to kinematic
configurations of the wind field that interact with the Surface frontogenesis refers to frontogenesis that
temperature field, and the fifth encompasses all occurs in the horizontal cross-front direction over
diabatic processes that contribute to frontogenesis. relatively level terrain, away from coastal zones. Close
The horizontal deformation term (I) is composed of to the ground, where the vertical motion is relatively
four contributions, which are illustrated in Figure 1. small, temperature changes in the cross-front direc-
The confluent flow in Figure lA, restricted to the tion, IvHBl, are primarily associated with horizontal
horizontal plane, concentrates the thermal gradient so deformation (I) and with horizontal gradients of
as to increase F; difluence acts in the opposite sense. diabatic heating or cooling (V). The observed field of
The convergent flow in Figure 1A also increases the horizontal deformation overlaying surface potential
thermal gradient in the horizontal plane as does a temperature isotherms, shown in Figure 3, produced
confluent flow, but horizontal convergence and diver- dry nighttime frontogenesis in the Great Plains over a
gence also produce motion normal to the plane. 5 h period. The idealized patterns of Figures 1A and B
Horizontal shear, as depicted in Figure lB, both all appear in observed low-level deformation flows
concentrates the isotherm pattern and rotates the that produce surface frontogenesis, but are not easy to
alignment to increase frontogenesis in the y-direction. separate into component parts in real flows.
Vertical deformation (11), associated with the vertical A few examples will illustrate situations in which
shear of the horizontal wind, is shown in Figure 1C. differential diabatic heating or cooling contribute to F .
The contribution displayed is frontogenetical, increas- Figure 1F depicts a daytime situation which enhances
ing both the vertical and horizontal temperature frontogenesis. Relatively little temperature change
gradients. Vertical divergence (111)only acts on vertical will occur in cloud-capped cold air behind the front;
gradients of temperature. This term is illustrated by low-level heating is significant on the warm side of
the convergent wind field, in Figure l D , which the front, IvH8) increases. Alternatively, frontolysis
produces frontogenesis along the z-axis. The tilting would be evident at night under the same conditions
term (IV)is a kinematic effect produced by differential shown in Figure lF, since radiational cooling of the
vertical motion. The motion displayed in Figure 1E low-level air on the warm side would take place, with
802 FRONTOGENESIS
approach to frontogenesis is limited. It is necessary to

encompass both thermodynamic and dynamic pro-
cesses, described by the complete set of predictive
equations, to understand completely frontogenesis
and frontolysis.
Coastal Frontogenesis
Coastal frontogenesis is surface frontogenesis that
owes its existence to the particular conditions associ-
ated with a coastal zone. Coastal frontogenesis can be
particularly prominent, for example, along the east
coast of the United States in wintertime. Favorable
conditions are set up by the low-level thermal gradient
between the relatively warm air over the ocean and the
colder continental air inland. The shallow mass of cold
Figure 3 Streamlines of horizontal flow, with isotherms of
potential temperature (K) denoted by dashed lines, during a period
continental air is blocked to the west by the Appala-
of frontogenesis observed in the Great Plains. (Adapted with chian Mountains, which are oriented approximately
permission from Ostdiek V and Blumen W (1997). Journal parallel to the coastline. This topographic constraint is
of the Atmospheric Sciences 54: 1490-1 502. Boston: American called cold air damming.
Meteorological Society.) Differential diabatic heating (V) establishes a pre-
ferred location for the onset of coastal frontogenesis.
The conditions are set up for frontogenesis to proceed
little temperature change in the cold air expected to
when air flows southward from a cold anticyclone,
occur. The formation of clouds also contributes to
which is situated north of the front in the vicinity of the
frontogenesis by means of differential heating associ-
south-eastern coast of Canada. This flow is modified
ated with the release of latent heat. Cumulus convec-
by convective heat flux from the warm ocean and
tion, for example, in advance of a cold front, increases
crosses the United States coastline as an easterly flow.
the cross-front temperature gradient by condensatio-
Convergence occurs along the coastal zone as the low-
nal heating in the warm air. Evaporative cooling will
level airstream encounters rougher inland terrain, and
also produce frontogenesis when precipitation falls
experiences an increased surface drag. Cold air also
into an unsaturated layer of cold air behind the front.
streams southward, parallel to topographic contours,
Consider the vertical deformation (11) term, as
in the corridor between the coast and the Appalachian
displayed in Figure lC, to be associated with positive
chain. As a consequence, the low-level flow turns from
wind shear as a consequence of boundary layer friction
northerly to easterly across the frontal zone, providing
or drag. This effect on frontogenesis would be difficult
a horizontal deformation (I) flow field that reinforces
to measure, since it acts at low levels in consort with
the preexisting thermal gradient. Air convergence in
horizontal deformation (I). Close to the ground,
the frontal zone also provides a vertical uplifting of
however, contributions to frontogenesis by vertical
warm, moist air that produces clouds and precipita-
divergence (111) and tilting (IV) of the front in a
vertical cross-section, which arise from gradients of tion on both sides of the front. The sign and magnitude
of this secondary diabatic (V) contribution to fronto-
vertical velocity, are expected to be relatively small.
genesis may vary from case to case. Further, frontolysis
Yet, these terms become more significant aloft, where
may occur in less than 24h when the horizontal
they contribute to both the frontal slope and to the
deformation field, which reinforces the thermal gra-
vertical component of the temperature gradient. It is
dient, is altered by an evolving large-scale synoptic
not only kinematics of the vertical motion field,
flow pattern.
however, that needs to be addressed. Convergence at
low levels, for example, contributes to frontogenesis
by horizontal temperature advection. At the same
time, rising motion on the warm side of the front
Upper-Level Frontogenesis
would cool by dry adiabatic ascent, and dry adiabatic Upper-level frontogenesis refers to frontogenesis that
descent on the cold side would produce warming. The is initiated aloft in the vicinity of midlatitude jet
temperature change, brought about by this thermo- streams, and may not appear to be directly coupled to
dynamic process, is frontolytical. This example is one frontogenesis accompanying surface-based activity.
of many which illustrates that the purely kinematic Figure 4 displays a synoptic analysis that illustrates a
FRONTOGENESIS 803
100 km
H
km
11.8
--____
9.2
_----
5.6
-- 290
3.0
1.o
(A) 270
100 km
H
_----
290 280 0
Figure 4 Cross-section of an upper-level frontal system during a period of frontogenesis: (A) early stage; (B) mature stage reached 24 h
later. The tropopause is denoted by thick solid lines. The thin dashed lines are isotherms of potential temperature (K), and the thin solid
lines are isotachs (m s- ’). J denotes the axis of the jet stream directed out of the section. Pressure levels and standard heights are shown
on the abscissa. (Adapted with permission from Reed RJ (1955) Journal of Meteorology 12: 226-237. Boston: American Meteorological
Society.)
case of upper-level frontogenesis, which evolved into a axis of the jet stream. The relative position of the
mature front exhibiting a significant thermal gradient circulation cell and its intensity are also determined by
that extended down to the 800mbar level, or about cold air advection along the jet axis. The secondary
2 k m above the Earth’s surface. As in Figure 2 , a circulation that develops in response to changes in the
thermally direct circulation cell is required in order to thermal gradient serves a dual purpose. First, thermal
maintain thermal wind balance at upper levels. The wind balance is maintained in the presence of fronto-
schematic representation, presented in Figure 5, dis- genesis. Second, this secondary circulation converts
plays a prominent clockwise circulation cell, which is potential energy into kinetic energy in order to
confined to the warm side of the frontal zone. The maintain the upper-level jet stream frontal system.
circulation on the cold side may not be well defined The tropopause, which is low over the cold air and
and, accordingly, is omitted from the figure. high over the warm air, represents the relatively sharp
This cross-front circulation represents a response to transition to the lower stratosphere, where large static
horizontal deformation (I), which provides an en- stability resists vertical displacements. Accordingly,
hancement of the thermal gradient upstream, and a the secondary circulation tends to steepen the tropo-
diminution of this gradient downstream, along the pause and draw a relatively thin slice of stratospheric
804 FRONTOGENESIS
upper levels. This process is related to the breakdown

or instability of relatively laminar horizontal flow into
small-scale turbulent eddies that ultimately dissipate
into heat at molecular scale. Flights carried out by
instrumented aircraft reveal that the favored locations
for this process to occur are above and below the jet
i stream and in the frontal zone, where vertical shears of

the horizontal wind are very large. These regions are
more susceptible to large vertical displacements and
overturning eddies, even when the buoyancy force
tends to oppose this type of motion. Small-scale
mixing is usually viewed as a frontolytical process,
acting to smooth out sharp gradients. Yet observations
do reveal the formation of relatively sharp tempera-
Figure 5 Schematic illustration of the transverse secondary ture gradients in the atmosphere that appear to
circulation associated with an upper-level front, as shown in Figure develop from mixing in thin layers separated by
4. (Adapted with permission from Danielson ER (1968) Journal more laminar flow. However, the relative importance
of the Atmospheric Sciences 25: 502-51 8. Boston: American
of this type of differential mixing, and the contribution
made to frontogenesis by dissipation of these turbulent
eddies into heat, has not been adequately determined.
air downward to lower levels and create a tropopause
fold. The upper-level frontal zone is delineated by the
concentrated horizontal and vertical potential tem- Frontogenesis in the Wind Field
perature gradients appearing below the 300 mbar level Surface heterogeneity can fix the position of a front,
in Figure 4. The development and alignment of this and differential diabatic heating can provide a contri-
characteristic thermal field is intimately associated bution to frontogenesis, but ultimately frontogenesis is
with the development and evolving structure of the a process driven by a favorable wind distribution. The
circulation cell. In particular, the subsiding branch of kinematic approach to frontogenesis does not, how-
the circulation provides adiabatic compressional heat- ever, reveal the complex nature of nonlinear interac-
ing below the jet axis, which is responsible for the tions that sharpen gradients. It is necessary to solve the
cross-front thermal gradient, and its descent to low basic equations of dynamic meteorology to develop an
levels of the troposphere. The elapsed period between understanding of how these interactions operate.
the relatively early stage of frontogenesis in Figure 4A Frontogenesis in the wind field may be simply
to the mature stage in Figure 4B is 24 h. illustrated by a prototype, which exhibits a physical
The vertical deformation field (11) provided by the process that has been identified in more complex
circulation cell is frontogenetical at upper levels, but atmospheric models. The evolution of the velocity
becomes frontolytical below where the front weakens normal to the front, u, can be described by
and terminates. The contribution to vertical diver-
gence (111) cannot, however, be easily assessed. Rela- au au
tively large gradients of vertical motion could be -at
+ u - = ax
o PI
expected to exist below the tropopause, where vertical
motion is suppressed. The sign and magnitude of these where t is time and x is the direction normal to the
gradients is only available from model simulations and front. The initial distribution of u, shown in Figure 6 ,
from measurements obtained by instrumented air- displays a convergent wind field. According to eqn [5],
craft. These sources do not provide sufficient infor- each value of u travels at velocity u. At later times, a
mation for a definitive evaluation of the overall effect given value of u is displaced a distance ut from its
of vertical divergence on frontogenesis. The large previous position. This nonlinear self-advection will
magnitude of subsidence on the warm air side of the continue until au/ax = m. A dissipative process in
front is, however, consistent with the frontal slope nature will, however, prevent the occurrence of a
produced by tilting (IV) of horizontally aligned discontinuity in the velocity field. A finite frontal zone
isotherms at upper levels into the more vertical could be expected to occur, exhibiting a balance
orientation that characterizes the mature front in between advection and dissipation.
Figure 4B. The nonlinear process described by eqn [5] is only
The diabatic (V) contribution to frontogenesis has one of many that contribute to frontogenesis in
generally been associated with turbulent dissipation at atmospheric flows. The process of nonlinear
FRONTS 805
See also
Cyclogenesis. Cyclones, Extra Tropical. Fronts.
Synoptic Meteorology: Weather Maps. Turbulence
and Mixing.
Further Reading
Bluestein HB (1993)Synoptic-Dynamic Meteorology
i n Midlatitudes, Vol. 2, Observations and Theory of
Weather Systems. New York: Oxford University Press.
Carlson TN (1991) Mid-Latitude Weather Systems.
London: Harper Collins Academic.
-X
Hoskins BJ (1982)The mathematical theory of frontogen-
esis. Annual Reviews of Fluid Mechanics 14: 131-151.
Figure 6 Horizontal velocity u as a function of the cross-front Keyser D (1986) Atmospheric fronts: an observational
coordinate x . The distribution of u is shown at the initial time (solid) perspective. In: Ray P (ed.)Mesoscale Meteorology and
and at a later time (dashed). Forecasting, pp. 216-258. Boston: American Meteoro-
logical Society.
Keyser D and Shapiro MA (1986)A review of the structure
self-advection is a particularly important feature of and dynamics of upper-level frontal zones. Monthly
small-scale frontogenesis, when the Earth’s rotation Weather Review 114: 452-499.
may be neglected, along-front variability is not signif- Orlanski I, Ross B, Polinsky L, and Shaginaw R (1985)
icant, and vertical velocity is not a factor close to the Advances in the theory of atmospheric fronts. Advances
Earth’s surface. in Geophysics 223-251.
W Blumen, University of Colorado at Boulder, Boulder, fronts in the troposphere, which appear to be uncon-
CO, USA nected to surface-based frontal features.
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. The existence of a front or, equivalently, a frontal
zone, is an intrinsic property of the physical laws that
govern atmospheric motions. Fronts are produced in
Introduction numerical model simulations of the atmosphere in
Atmospheric fronts are three-dimensional zones that which radiation, cloud cover, and surface heterogene-
represent a rapid transition of the thermal field, ity are not a consideration. These latter processes
accompanied by a marked shift in the wind direction. provide, however, some of the interesting variants that
In some cases, a significant transition in the moisture characterize observed atmospheric fronts.
field may also occur across the frontal zone, and an
increase in the magnitude of the wind may occur in
collaboration with the wind shift. The thermal gradi-
Surface Fronts
ent is usually largest at the Earth’s surface, and An atmospheric surface front, in the present context, is
weakens as the zone slopes upward with relatively a well-defined phenomenon that may be distinguished
warm, less-dense air overlying relatively colder air of by the network of global meteorological observations.
higher density. The transition zone may extend over a The frontal intensity, measured by changes in the
distance of 1000 km or more along the Earth’s surface, temperature, humidity, and wind fields that occur
and a typical cross-front to along-front ratio is 1 : l O . across the frontal transition zone, is most pronounced
Fronts are observed at all longitudes in the extratrop- at or near the Earth’s surface, and diminishes with
ical latitude belt and are associated with most of the altitude. Fronts that fall into this category are synop-
significant weather events that occur during the late tic-scale events, which may retain their individual
autumn through the early spring period. Although identities for up to a week or more. The physical
many fronts are surface-based, thermal gradients may processes that give rise to identifiable frontal features
be enhanced at levels up to the tropopause producing are controlled by four principal forces: the buoyancy
FRONTS 805
See also
Cyclogenesis. Cyclones, Extra Tropical. Fronts.
Synoptic Meteorology: Weather Maps. Turbulence
and Mixing.
Further Reading
Bluestein HB (1993)Synoptic-Dynamic Meteorology
i n Midlatitudes, Vol. 2, Observations and Theory of
Weather Systems. New York: Oxford University Press.
Carlson TN (1991) Mid-Latitude Weather Systems.
London: Harper Collins Academic.
-X
Hoskins BJ (1982)The mathematical theory of frontogen-
esis. Annual Reviews of Fluid Mechanics 14: 131-151.
Figure 6 Horizontal velocity u as a function of the cross-front Keyser D (1986) Atmospheric fronts: an observational
coordinate x . The distribution of u is shown at the initial time (solid) perspective. In: Ray P (ed.)Mesoscale Meteorology and
and at a later time (dashed). Forecasting, pp. 216-258. Boston: American Meteoro-
logical Society.
Keyser D and Shapiro MA (1986)A review of the structure
self-advection is a particularly important feature of and dynamics of upper-level frontal zones. Monthly
small-scale frontogenesis, when the Earth’s rotation Weather Review 114: 452-499.
may be neglected, along-front variability is not signif- Orlanski I, Ross B, Polinsky L, and Shaginaw R (1985)
icant, and vertical velocity is not a factor close to the Advances in the theory of atmospheric fronts. Advances
Earth’s surface. in Geophysics 223-251.
W Blumen, University of Colorado at Boulder, Boulder, fronts in the troposphere, which appear to be uncon-
CO, USA nected to surface-based frontal features.
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. The existence of a front or, equivalently, a frontal
zone, is an intrinsic property of the physical laws that
govern atmospheric motions. Fronts are produced in
Introduction numerical model simulations of the atmosphere in
Atmospheric fronts are three-dimensional zones that which radiation, cloud cover, and surface heterogene-
represent a rapid transition of the thermal field, ity are not a consideration. These latter processes
accompanied by a marked shift in the wind direction. provide, however, some of the interesting variants that
In some cases, a significant transition in the moisture characterize observed atmospheric fronts.
field may also occur across the frontal zone, and an
increase in the magnitude of the wind may occur in
collaboration with the wind shift. The thermal gradi-
Surface Fronts
ent is usually largest at the Earth’s surface, and An atmospheric surface front, in the present context, is
weakens as the zone slopes upward with relatively a well-defined phenomenon that may be distinguished
warm, less-dense air overlying relatively colder air of by the network of global meteorological observations.
higher density. The transition zone may extend over a The frontal intensity, measured by changes in the
distance of 1000 km or more along the Earth’s surface, temperature, humidity, and wind fields that occur
and a typical cross-front to along-front ratio is 1 : l O . across the frontal transition zone, is most pronounced
Fronts are observed at all longitudes in the extratrop- at or near the Earth’s surface, and diminishes with
ical latitude belt and are associated with most of the altitude. Fronts that fall into this category are synop-
significant weather events that occur during the late tic-scale events, which may retain their individual
autumn through the early spring period. Although identities for up to a week or more. The physical
many fronts are surface-based, thermal gradients may processes that give rise to identifiable frontal features
be enhanced at levels up to the tropopause producing are controlled by four principal forces: the buoyancy
806 FRONTS
force, the pressure gradient force, the Coriolis force, A midlatitude wave cyclone, a low-pressure system,
and the frictional force, or drag, whose influence is is depicted in Figure 1.A cold front, a warm front and
primarily restricted to the lowest 1-1.5km of the an occluded front emanate from the center of the low
atmosphere, the planetary boundary layer. Gravity at various stages of the 3 6 h period depicted. The
acts downward, toward the Earth’s surface, and is designations, cold front, warm front, and occluded
opposed by an upward directed or vertical pressure front, emerged from the Norwegian cyclone model,
gradient force. Hydrostatic balance occurs when the introduced after World War I, and have their basis in
two forces are equal in magnitude. In most cases there military terminology. In this analogy polar air attacks
is a small imbalance that provides a buoyancy force. southward, the cold front; warmer air counterattacks
The direction of the buoyancy force controls the northward, the warm front. The occluded front
direction of vertical motions associated with fronts. develops when the cold front makes a sharp turn and
The Coriolis force is directly proportional t o the attacks the warm air in its flank and, ultimately, the
Earth’s rotation rate and t o the magnitude of the two fronts merge with warm air rising above the
horizontal wind vector. The Coriolis force, directed to colder, surface-based air. Frontal modifications that
the right of the horizontal wind in the Northern arise from passage over and around relatively large
Hemisphere, is opposed by the horizontal component topographic features, or from the passage across a
of the pressure gradient force, which is directed large body of water, modify this simple picture.
toward low pressure. These forces tend to be in
approximate balance above the planetary boundary
Cold Front
layer (the so-called geostrophic balance), providing a
counterclockwise circulation around a low-pressure A cold front represents the advance of relatively cold,
center. The drag force reduces the wind speed within dense air into a region previously occupied by relati-
the planetary boundary layer, and lessens the magni- vely warm less dense air. The cold front in Figure 1
tude of the Coriolis force, which leads to cross-isobaric is delineated by the triangular teeth that point in
flow toward low pressure. the direction of the advancing cold air. The finite
Figure 1 Stages in the development of an occluded front, and its relationship to a wave cyclone in middle latitudes. (A) Mature wave
cyclone. (B) Partially occluded wave cyclone. (C) Occluded wave cyclone. The thin solid lines in the upper panels are isobars (hPa), and L
designates the low-pressure center. The cold, warm, and occluded fronts are described in the text. (Reproduced with permission from
Lutgens FK and Tarbuck EJ (1992) The Atmosphere, 5th edn. Englewood Cliffs, NJ: Prentice-Hall.)
FRONTS 807
transitional zone, generally 100 km or less, is reduced balance. The appropriate expression is provided by
to a line on synoptic weather maps for two reasons: the
surface observational network is too widely
spaced to delineate the transition zone between
disparate air masses, and the map scale essentially
reduces the transition zone to the thickness of a broad where u is the along-front geostrophic wind, normal
line. to the direction of the temperature gradient, T the
Some cold fronts may be relatively shallow, as is absolute temperature (T is an average value, inde-
indicated in Figure 1,where the extent of the cold air is pendent of y ) and f = 2t2 sin 4, where 0 is the angular
limited to 1-2 km above the ground. In other cases, as velocity of the Earth's rotation and 4 denotes latitude.
in Figure 2 , the cold front retains its identity up The marked vertical change, or shear, of the geo-
through most of the troposphere. The frontal slope is strophic velocity is a direct consequence, according to
designated as dz/dy, where z denotes height above the eqn [I], of the relatively large horizontal temperature
ground and y is normal to the front, increasing toward gradient through the frontal transition zone.
the cold air. Typical slopes of cold fronts are usually in Typical features associated with the passage of
the range dz/dy = 1 : SO-1 : 150, and the frontal a cold front past an observer at a fixed location may
slope in Figure 2 falls within this range. Close to the be described by reference to the first two panels of
ground, however, frontal translation is restrained by Figure 1.The temperature decreases relatively abrupt-
the drag force, which has the effect of steepening the ly as the frontal zones passes, and the wind veers in the
frontal slope and deepening the wedge of cold air clockwise direction. In addition, the surface wind
behind the front. The cold air aloft may, in some cases, speed may increase in response to an increase in the
overrun the warm air ahead of the front, but the nearly pressure gradient force, and it may also become gusty.
hydrostatic balance that usually prevails for midlati- The temperature contrast across the frontal zone is
tude frontal situations requires high-density cold air to maintained by the horizontal advection of tempera-
reside under the sloping transition zone, as displayed ture toward the front, which is consistent with the
in both Figures 1 and 2 . observed wind shift. Figure 1shows that the front lies
The increase in wind through and above the frontal in relative minimum or trough in the surface pressure
zone may be represented by the thermal wind relation, field, which produces a pressure drop as the front
which incorporates both hydrostatic and geostrophic approaches. Then the pressure increases as cold, dense
-56°C _,40------., ,
-64°C
100
250
OM CB NA AT CH
Figure 2 Distribution of isotherms ('C), denoted by thin solid lines, and isotachs (m s-I), denoted by thin dashed lines, in a vertical
cross-section through a cold front. The jet stream axis, denoted by J, is directed into the section. Heavy solid lines indicate the slope of the
surface-based cold front, and the upper-level tropopause. The distance from Omaha (OM) to Charleston (CH) is approximately 1700 km.
(Reproduced with permissionfrom Wallace JM and Hobbs PV (1977) AtmosphericScience,An IntroductorySurvey. New York: Academic
Press.)
808 FRONTS
air moves through the observation site. There is often a pied by relatively cold, high-density air. The leading
decrease in the moisture content of the air, measured edge of the warm front in Figure 1 is delineated by the
by the dew point temperature, through the frontal semicircles that point in the direction of movement of
zone. This feature is particularly noticeable if the the warm air. The receding cold air is not, however,
warm air sector, ahead of the front, is relatively close to pushed northward by the advancing warm air. The
a large body of water, such as the Gulf of Mexico. The north-easterly retreat of the cold air is associated with
extent of cloudiness and precipitation is determined by the dynamics that governs the coupled system of the
the extent of moisture in the air, and by the extent of surface-based wave cyclone and westerly flow aloft.
vertical uplift at and behind the front. Yet the front is The transition zones of warm fronts are not nearly as
not a material surface, and some of the air that sharp and distinct as cold frontal zones, and may
converges at the front in low levels may pass into the extend over a few hundred kilometers. There are at
frontal zone and recirculate without significant uplift. least two reasons for this lack of sharpness. First, the
Some may also rise, reverse direction, and then move receding cold air has progressed much further from its
away from the front, producing little or no cloudiness. source region than the relatively fresh surge of cold air
A characteristic feature, indicated in Figure 1, is for the behind the cold front. As a consequence, exchange of
air to rise, cool, and then condense, forming cumulus- heat and moisture between the surface layer and the air
type cloudiness. The characteristic precipitation above has produced a more significant modification of
patterns are varied, but intermittent shower the receding cold air, and thereby reduced the contrast
activity originating from large cumulus cells and between the adjacent air masses. Second, other air near
banded structures aligned parallel to the front are the surface, embedded in the planetary boundary
not uncommon. Severe activity associated with cold layer, is subject to a drag force. Air motion above the
frontal passages includes hail, lightning, and tornado boundary layer is essentially frictionless flow. The
inceptions. result is to reduce the frontal slope, typically
A backdoor cold front designates a cold front that dz/dy 1 : 300, and the depth of the receding cold
moves opposite to the pattern depicted in Figure 1: its air. As a consequence, the relatively shallow depth of
movement is westward in the Northern Hemisphere. the receding air mass is also more prone to modifica-
In the USA, for example, a New England backdoor tion by vertical exchange with the underlying surface.
cold front is associated with the westward expansion Figure 1A and B also indicate a sequence of change
of a cold surface high-pressure system situated near the as a warm front passes an observer at a fixed location.
North Atlantic coast during winter. The thermal The temperature increases through the transition
gradient is reversed in this case, and the frontal zone zone, but the magnitude of change is less than the
is characterized by fog and low stratiform clouds magnitude generally accompanying cold frontal pas-
fueled by moisture from the ocean. A similar event sages. The wind direction also veers in the clockwise
may also occur in spring and summer, when relatively direction, but the wind speeds are relatively light and
cool maritime air from the Atlantic Ocean moves over the wind shift may be obscured by local topographical
the eastern seaboard. This cold air incursion, also influences on the wind. The circulation must, however,
called a back-door cold front, moves southward along maintain a net horizontal advection of temperature
the eastern slope of the Appalachian Mountains, toward the front in order to maintain frontal con-
which serve as a barrier to inland penetration. trasts; otherwise the front would begin to lose its
The southerly buster or southerly burster, defined as identity. The warm front lies in a trough of low
a squally wind change, is an intense summertime cold pressure, but the sign and magnitude of the pressure
front that arrives at the south-eastern tip of Australia change associated with the frontal passage is depend-
from the Southern Ocean. Arrival of this front in the ent, to a large extent, on the direction of frontal
afternoon can be accompanied by temperature chang- progression. A change in the dew point temperature
es of 10-15°C over a period of a few minutes, but may also be difficult to observe, particularly if an
precipitation is not usually associated with a southerly extensive band of precipitation falls into the cold air
burster. The front travels equator-ward, acquiring a mass. In the absence of precipitation, the dew point
characteristic S-shape as its movement is inhibited by temperature would normally increase after frontal
the east coast mountain chain, but movement inland passage. The warm, less-dense air, which is not
and along coastal waters is less restrained. entrained into a frontal zone circulation, tends to
rise above the surface-based cold air and move with
the upper-level flow more rapidly than the translation
Warm Front
speed of the front. The type of cloudiness that may be
A warm front represents the advance of relatively observed in advance of the front tends to be of the
warm, low-density air into a region previously occu- stratiform variety. An extensive overcast may precede
FRONTS 809
the warm front, with low stratus and fog character- Occluded Front
izing conditions within and near the frontal zone. The
The development of an occlusion or occluded front is a
type and extent of precipitation that may fall from
process that increasingly cuts off the low-level supply
stratiform clouds into the receding cold air depends on
of relatively warm air by vertical uplift, as depicted in
the moisture content, the vertical profile of air
Figure 1.The occluded front is denoted on the surface
temperature, and the northward extent of the cloud
map by alternating triangles and semicircles that point
cover. A particularly treacherous situation occurs
in the direction of movement. According to the
when rain falling into cold air freezes and produces
Norwegian model, depicted in Figure 1, a cold
ice forms that coat structures, such as power lines and
occlusion occurs when the advancing cold air mass,
motorways.
behind the cold front, overtakes the less-cold receding
Northward advection of warm air, as depicted in
air mass ahead of the warm front. The receding cool
Figure 1, is the most common circulation pattern
air is less dense than the advancing cold air and is
associated with warm fronts, but not the only one.
forced to rise, as shown in Figure 1, to maintain a
Westward advection of relatively warm air from the
hydrostatically stable environment. At low levels, cold
Atlantic Ocean, for example, followed by a recircula-
air continues to advance as a cold front and replaces
tion can produce a warm front that progresses in a
cool receding air. As a consequence, the wind temper-
southerly direction, which is referred to as a backdoor
ature and pressure changes across the frontal zone are
warm front. Another type of warm front can be
as with a cold frontal passage, although less pro-
produced by adiabatic warming of air that descends
nounced. The warm front lifted aloft, designated as the
along the eastward-facing lee slopes of the Rocky
occluded front in Figure 1,advances with the advance
Mountains. The leading edge of this descending air,
of the surface-based cold air mass. The precipitation
referred to as a Chinook front, will either remain along
pattern is related to stratiform clouds, as in the case of
the Front Range of the Rocky Mountains or progress
a warm frontal passage, but tends to be displaced over
into the Great Plains as a warm front, if the cold air
the cold surface air. This type of synoptic evolution
ahead of it is displaced eastward.
would tend to be most evident when cold continental
air behind the cold front overtakes receding cold air
Stationary Front that either has been warmed by a longer residence time
over the continent or had originally acquired charac-
A stationary front represents a transition zone be-
teristics of a relatively warmer maritime environment.
tween cold and warm air masses that does not exhibit
The counterpart of the cold occlusion in the
significant motion toward either air mass. It may also
Norwegian model is the warm occlusion, depicted in
be referred to as a quasi-stationary front. This type of
Figure 3. The thermal characteristics are now re-
front may simply reflect a change in the synoptic scale
versed: the advancing cold air is warmer and less dense
circulation pattern that halts the translation of either a
than the receding cold air and, upon merger, the cold
cold or a warm front, or the heterogeneity of the
front is lifted aloft. This process constitutes a warm
Earth’s surface may provide conditions that fix a
front occlusion, and the cold front aloft is designated
frontal transition zone to a preferred location. Coastal
as the occluded front. The surface characteristics tend
zones, zones separating relatively warm oceanic air
to reflect a warm frontal passage, but the cold front
from colder continental air in winter, provide an
environment that promotes the formation of a quasi-
stationary front, referred to as a coastal front. Favored
locations are along the east coast of the United States,
from Maine to the Carolinas, and the Texas coast.
Although these fronts exhibit transition zones that are
comparable to those of a cold front, they are of limited
extent, ranging from 200-600km, and of limited
duration, lasting up to a day or less. The frontal zone is
maintained by convergence of relatively warm moist
air from the ocean toward the relatively dry continen-
tal air, with a clockwise shift in the wind across the
front. This situation often results in a band of
precipitation, parallel to the front, with the maximum
Occluded front
precipitation occurring on the cold side, and possibly a
transition from frozen precipitation to rain on the
warm side. Figure 3 Warm front type of occlusion.
810 FRONTS
aloft may, however, produce more upper-level con- tem. It is defined as

vective activity than is usually associated with a typical
warm front. This type of occlusion could be expected
when cold air of maritime origin enters northerly PI
portions of western coastlines as a cold front, and then
overtakes colder continental air further inland. The where T i s the absolute temperature, p the pressure, p o
Pacific Northwest, and the west-coastal regions of a constant reference pressure, and R/cp denotes the
northern Europe, represent favored locations for this ratio of the gas constant to the specific heat of air at
to occur. constant pressure. Potential temperature is conserved
It has, however, become apparent in recent years under dry adiabatic displacements of air parcels:
that the Norwegian model of an occluded front does temperature increases with descent to higher pressure
not represent the only three-dimensional atmospheric and decreases during ascent to lower pressure. Poten-
structure when cold fronts and warm fronts appear to tial temperature increases with decreasing pressure
merge. Other synoptic processes can provide surface (increase in altitude) to maintain a hydrostatically
signatures that resemble the occlusion depicted in stable atmospheric environment. As a consequence,
Figure 1. A cold front in a marine cyclone, for stratospheric air can be identified with relatively large
example, may separate or fracture from its union values of 8.
with a warm front in the low center, and then travel The mature upper-level frontal structure displayed
along the surface position of the warm front. This in Figure 4, shows a frontal zone depicted by the
process duplicates the increasing separation of the low concentration of potential temperature isotherms. It is
center from the triple point, where the cold, warm, and approximately 100-200 km wide, and exhibits a slope
occluded fronts meet. Another process that may dz/dy 1 : 100. This structure is not described as
appear to be an occlusion occurs when an upper-level either a cold or a warm front, but exhibits qualities of a
front, not originally part of the surface wave cyclone quasi-stationary front that moves parallel to itself.
system, descends into the region where the cold and Superposed on this front and jetstream system is a
warm fronts are starting to merge. Passage of the transverse or cross-front circulation, which is dis-
upper-level front above the receding cold air mass played in Figure 5. This circulation cannot be ob-
provides a synoptic structure similar to a warm front served, because the magnitude of the velocities is
occlusion. The upper-air observing network is not smaller than can be resolved by current measurement
sufficiently dense to provide an unambiguous distinc- techniques. Its existence is based on theoretical work,
tion between the various scenarios presented. Refine- and the position of the circulation cell and its
ment of the Norwegian model of an occlusion by the properties are extracted from numerical model simu-
documentation of other physically consistent physical lations. The circulation on the cold side of the front is
processes that have the appearance of an occlusion, not well defined by these simulations and, accordingly,
their preferred locations, and their frequency of is omitted from the figure. The circulation displayed in
occurrence are topics of current research interest. Figure 5 does provide a consistent explanation of how
the upper-level front is maintained, and why strato-
spheric constituents, such as ozone and high-level
radioactivity from nuclear explosions, can be ob-
Upper- LeveI Front
served in the lower troposphere.
An upper-level front, or an upper tropospheric front, is The subsiding branch of the circulation cell main-
a transition zone exhibiting a sharp thermal contrast tains the most prominent characteristics of the upper-
that may extend from the tropopause down to 2 or level front that are revealed in Figure 4. In particular,
3 km above the ground. Upper-level fronts are not the descent of stratospheric air within a narrow
usually associated with the type of weather that pocket, called a tropopause fold, provides observa-
characterizes surface fronts. Clear-air turbulence and tional evidence of strong subsidence that extends
tropospheric-stratospheric exchange are often defin- downward to low tropospheric levels. This downward
ing characteristics that distinguish upper-level fronts. motion is also responsible for the thermal contrast
Temperature changes within an upper-level frontal across the front and for the slope that characterizes the
system are associated primarily with dry adiabatic fold and the potential temperature isotherms that are
compression; condensation and radiative heating are contained within it. The thermal contrast across the
considered to be secondary factors in the maintenance front is maintained by adiabatic compressional heat-
of the mature front displayed in Figure 4. The potential ing below the jet axis in the subsiding branch of the
temperature 0 is introduced to characterize the dry cell, with relatively less vertical motion and corre-
adiabatic processes associated with this frontal sys- sponding temperature change on the cold side of the
FRONTS 811
100 km
H
290 280 0
Figure 4 Cross-section of an upper-level front. The tropopause is denoted by thick solid lines. The thin dashed lines are isotherms of 0
(K), and the thin solid lines are isotachs (m SKJI)denotes
. the axis of the jetstream, directed out of the section. Pressure levels (hPa) and
standard heights (km) are shown on the abscissa. (Adapted with permission from Reed RJ (1955) Astudy of acharacteristic type of upper-
level frontogenesis. Journal of Meteorology 12: 226-237.)
front. This feature of the circulation has the effect of significant factor, but the nonhydrostatic acceleration
concentrating the thermal gradient within the fold. may be important in the dynamics of these fronts.
The sloping alignment of the isotherms arises because
the differential vertical motion field tilts the potential
Sea Breeze Fronts
temperature isotherms from a horizontal alignment, at
upper levels, into the one that displayed in Figure 4. See breeze fronts are most prominent in the warm part
Thermal wind balance, expressed by eqn [l], of the year, when daytime heating creates higher
provides the explanation for the high-speed jet flow temperatures over the land than over the water. A
above the frontal zone. In this example, the sharp pressure gradient, which develops in response to this
thermal gradient across the front produces a relatively differential heating, drives an onshore flow at low
strong jet that flows southward. Two characteristic levels, with a return flow at about 1-2 km above the
features of an upper-level front exhibited in Figure 4 surface. The sea breeze front is characterized by both a
are cyclonic shear, a wind shift that promotes a sharp temperature drop of a few degrees centigrade or
counterclockwise turning of the wind across the front, more, and a marked increase in the humidity, that can
and the intersection of 0 and p-surfaces, which charac- occur over a horizontal distance of a kilometer or less.
terizes a baroclinic atmosphere. These two features are Inland penetration of the sea breeze front, 10 km or
always associated with the development and enhance- more, may be opposed by an offshore wind ahead of
ment of a prominent upper-level jet and frontal system, the front, and by turbulent convective mixing over
and with surface-based fronts that extend to the upper land, which tends to smooth the temperature and
troposphere (see Figure 2 ) . humidity differences across the frontal zone.
Density Fronts
Small-Scale Fronts Density fronts represent the leading edge of density or
Small-scale fronts occupy a relatively limited horizon- gravity currents, driven by the upstream release of
tal domain, have relatively short lifetimes and are relatively cold high-density air. A pressure gradient
surface-based phenomena. Because they exist only for that is directed away from the source drives the
a few minutes to a few hours, the Coriolis force is not a motion. The situation is similar to the instantaneous
812 FRONTS
Figure 5 Schematic illustration of the transverse secondary circulation associated with an upper-level front, as shown in Figure 4.
(Adapted with permission from Danielson EF (1968) Stratospheric-tropospheric exchange based on radioactivity, ozone and potential
vorticity. Journal of the Atmospheric Sciences 25: 502-51 8.)
release of water from a dam. Two prominent density Drainage front A drainage front is the leading edge
fronts are gust fronts and drainage fronts. of downslope drainage of cold air from high eleva-
tions. It may then be referred to as a density front as it
continues to progress along relatively level terrain.
Gust front The gust front develops from evaporative Drainage currents develop at night, and the most
cooling associated with precipitation below large favorable conditions for occurrence are usually met
convective clouds. Its vertical extent is limited by the during the fall when clear skies and calm or relatively
height of the cloud base, usually below 2 km. Temper- light ambient wind conditions prevail. A pressure
ature changes as high as 10°C over a few tens of meters differential develops between the air over the slope and
can occur during severe thunderstorm activity, and the air at the same level over the level terrain. This pressure
wind gusts may represent a danger to aircraft that gradient force, together with gravity, drives a down-
attempt to land on runways where gust fronts are slope cold current of air with a temperature differen-
evident. tial at its leading edge. The distinguishing charac-
A gust front will last only minutes or tens of minutes teristics of the topography and the depth of radiatively
if the cold air moves away from its source. It may, cooled air along the slope will determine the frontal
however, persist for a few hours or more if the cold air characteristics. Compressional heating during descent
below the cloud base moves with the convective will also modify the cold temperatures behind the
system. Relatively warm, moist air ahead of the front front, but temperature drops of 5°C are not uncom-
moves up and along the frontal surface to the cloud mon after frontal passage. Humidity changes are
base, to provide the necessary fuel to continue the usually not considered a significant factor in this type
convective activity and sub-cloud precipitation that of front, but relatively light and gusty winds of about
ultimately drives the gust front. 5 m s - are often encountered.
FRONTS 813
See also Carlson TN (1991) Mid-Latitude Weather Systems. Lon-

Boundary Layers: Neutrally Stratified Boundary Layer. don: HarperCollins Academic.
Clouds: Cumuliform. Cyclogenesis. Cyclones, Extra
Keyser D (1986) Atmospheric fronts: an observational
Tropical. Density Currents. Frontogenesis. Gust perspective. In: Ray P (ed.) Mesoscale Meteorology and
Fronts. Jet Streaks. Stratosphere-Troposphere Forecasting, pp. 21 6-258. Boston: American Meteoro-
Exchange: Local Processes. Synoptic Meteorology: logical Society.
Weather Maps. Turbulence and Mixing. Keyser D and Shapiro MA (1986)A review of the structure
and dynamics of upper-level frontal zones. Monthly
Weather Review 114: 452-499.
Further Reading Orlanski I, Ross B, Polinsky L, and Shaginaw R (1985)
Bluestein HB (1993) Synoptic-Dynamic Meteorology in Advances in the Theory of Atmospheric Fronts. Advances
Midlatitudes, vol. 2 . Observations and Theory of Weath- in Geophysics, pp. 223-251. New York: Academic
er Systems. New York: Oxford University Press. Press.
GAIA HYPOTHESIS 81 5
T Lenton, Centre for Ecology and Hydrology,

Edinburgh, UK scientist and inventor James Lovelock. Lovelock
was employed by NASA, as part of the team that
aimed to detect whether there was life on Mars.
Lovelock’s interest in atmospheric chemistry led him
to seek a general, physical basis for detecting the
Introduction presence of life on a planet. He recognized that
The Gaia hypothesis postulates that the Earth’s surface most organisms shift their physical environment
is maintained in a habitable state by self-regulating away from equilibrium. In particular, organisms
feedback mechanisms involving organisms tightly use the atmosphere to supply resources and as a
coupled to their environment. The concept is based repository for waste products. In contrast, the atmos-
on several observations: phere of a planet without life should be closer to
thermodynamic equilibrium, in a state attributable to
The atmosphere is in an extreme state of thermo- photochemistry (chemical reactions triggered by solar
dynamic disequilibrium owing to the activities of ultraviolet radiation). Thus, the presence of abundant
life, yet aspects of its composition are remarkably life on a planet may be detectable by atmospheric
stable. analysis.
Present conditions at the surface of the Earth are Such an analysis can be conducted from Earth using
close to optimal for the dominant organisms. an infrared spectrometer (which detects the charac-
Life has persisted for over 3.8 billion years despite teristic absorption due to specific gases) linked to a
increasing solar luminosity and variable exchange telescope. Using this technique with ground-based
of matter with the inner Earth. telescopes, it was discovered that the atmospheres of
The Earth system has repeatedly recovered from Mars, and Venus are dominated by carbon dioxide and
massive perturbations. are relatively close to chemical equilibrium, suggesting
that they are lifeless (Figure 1A). In contrast, the
The Daisyworld model demonstrated that planetary atmosphere of the Earth is in an extreme state of
self-regulation can occur without teleology, in a disequilibrium as a result of the activities of life, in
manner consistent with natural selection. Since the which highly reactive gases, such as methane and
origin of life, organisms have had a profound effect on oxygen, coexist many orders of magnitude from
the Earth’s atmospheric composition and the climate. photochemical steady state (Figure 1A). Large, bio-
The ‘faint young Sun’ was initially counteracted by a genic fluxes of gases maintain this disequilibrium
carbon dioxide and methane ‘greenhouse’ atmos- (Figure 1B). Yet the composition of the Earth’s
phere. The biological amplification of silicate rock atmosphere is fairly stable over geological periods of
weathering has progressively reduced the carbon time. Lovelock concluded that life must regulate the
dioxide content of the atmosphere and acted as a composition of the Earth’s atmosphere.
-
long-term climate stabilizer. Atmospheric oxygen The dominant atmospheric gases, nitrogen and
rose in a stepwise fashion to 21 % of the atmosphere, oxygen, are biological products: atmospheric
about which it has been tightly regulated for the oxygen is the result of past photosynthesis, and
past 350 million years. Feedbacks involving terrestrial denitrifying organisms maintain atmospheric nitrogen
and marine biota also affect the climate over (the thermodynamically stable form of nitrogen in the
shorter time scales. The predominance of positive presence of oxygen should be as nitrate dissolved in the
feedback in the recent glacial-interglacial cycles sug- ocean). The proportions of these gases are particularly
gests that the Earth system is nearing a transition suited to the dominant organisms. Nitrogen serves to
to an alternative state. Eventually, self-regulation will dilute oxygen, which at 21% of the atmosphere
collapse and the Earth will be sterilized, but this is is just below the level at which fires would disrupt
unlikely to occur for at least another 0.5-1.2 billion land life. Yet oxygen is sufficiently abundant to
years. support the metabolism of large respiring animals
such as humans.
The Earth’s climate is close to optimal for the
Earth’s Remarkable Atmosphere dominant organisms, and has always been habitable
The Gaia hypothesis arose from the involvement in the despite major changes in the input of energy and
1960s space program of the British independent matter to the Earth’s surface. Notably, stars on the
816 GAIA HYPOTHESIS
probably richer in greenhouse gases, so that the early

’I n Earth’s surface was actually warmer then than it is
now, in which case cooling of the Earth in the face of
warming from the Sun demands an explanation.
Lovelock suggested that life has been regulating the
Earth’s climate together with its atmospheric compo-
sition.
The idea was named ‘Gaia’ after the Greek goddess
of the Earth, by the novelist William Golding. The first
scientific paper presenting ‘Gaia as seen through
the atmosphere’ was published in 1972. Lovelock
then sought an understanding of the organisms that
might be involved in regulating their planetary envi-
ronment. Lynn Margulis contributed her intimate
Earth Mars Venus knowledge of microorganisms, and the diversity of
chemical transformations that they mediate, to the
development of what became the Gaia hypothesis
104
103
I that ‘the environment at the surface of the Earth is
homeostated by and for the benefit of, the biota’. The
Gaia hypothesis was used to make predictions - for
example, that marine organisms would make volatile
10‘ compounds that can transfer essential elements from
the ocean back to the land. Lovelock and colleagues
N 10 tested this ancillary hypothesis on a scientific cruise
0
7
Y
between England and Antarctica. They discovered
X
3
1 that the biogenic gases dimethylsulfide and methylio-
dide are the major atmospheric carriers of the sulfur
and iodine cycles.
10-1
Later, the Gaia hypothesis was extended to
10-2
include regulation of much of the chemical composi-
(B) With life Abiotic
tion of the ocean. Then evidence began to accumulate
indicating that the Earth has remained habitable
despite major, periodic disruptions, including the
impact of planetesimals (massive meteorites) and
volcanic outbursts. These events appear to have
caused mass extinctions and climate change and yet,
in all cases, diverse, widespread life and a tolerable
climate returned within a short period of geological
Figure 1 The effect of life on the Earth’s atmosphere. (A) time. This supports the notion that the Earth is a self-
Atmospheric compositions of Earth, Mars and Venus (excluding
water vapor and noble gases). (B) Fluxes of gases at the Earth’s
regulating system.
surface with life (preindustry) and without life. (Reprinted with
permission from Lenton (1998).Copyright 1998 Macmillan Mag-
azines Ltd.)
Daisyworld
The Gaia hypothesis was greeted with hostility from
main sequence, such as the Sun, gradually become many scientists and leading scientific journals, partly
more luminous with time as the hydrogen in their core because of its mythological name. The first scientific
is converted to helium (increasing their density and criticism of the hypothesis was that it implies teleol-
accelerating the fusion reaction). The Sun was about ogy, some conscious foresight or planning by the biota.
25% less luminous when life originated on Earth, over Most subsequent criticisms have focused on the need
3.8 billion years ago. This increase in solar output for evolutionary mechanisms by which regulatory
alone should raise the Earth’s surface temperature by feedback loops could have arisen or be maintained.
-2O”C, but the current average temperature is only The Earth is not a unit of natural selection, and hence
15°C. This posed the ‘faint young Sun’ puzzle of why planetary self-regulation cannot have been refined in
the early Earth was not frozen. The atmosphere was the same way as an organism’s physiology. This poses
GAIA HYPOTHESIS 817
~~
the challenge of explaining how planetary self-regu- As the Sun warms, the temperature rises to the point
lation could arise. where white daisies begin to appear in the daisy
The Daisyworld model (Figure 2) was formulated to community. As it warms further, the white daisies gain
demonstrate that planetary self-regulation does not the selective advantage over the black daisies and
necessarily imply teleology. It provides a hypothetical gradually take over. Eventually, only white daisies are
example of climate regulation emerging from compe- left. When the solar forcing gets too much, the white
tition and natural selection at the individual level. daisies die off and regulation collapses. While life is
Daisyworld is an imaginary gray world orbiting a star present, the system is a very effective temperature
like our Sun that gets more luminous with time. The regulator. Solar input changes over a range that should
world is seeded with two types of life, black and white heat the planet’s surface by 55”C, yet it is maintained
daisies. These share the same optimum temperature within a few degrees of the optimum temperature for
for growth of 2 2 3 ° C and limits to growth of 5°C and daisy growth.
40°C. When the temperature reaches 5”C, the first Daisyworld illustrates the importance of feed-
seeds germinate. The paleness of the white daisies back mechanisms for planetary self-regulation. Feed-
makes them cooler than their surroundings, hindering back occurs when a change in a variable triggers a
their growth. The black daisies, in contrast, warm response that affects the forcing variable. Feedback is
their surroundings, enhancing their growth and re- said to be ‘negative’ when it tends to damp the initial
production. As they spread, the black daisies warm the change and ‘positive’ when it tends to amplify it. The
planet. This further amplifies their growth and they initial spread of life is amplified by an environmental
soon fill the world. At this point, the average temper- positive feedback - the warming due to the spread of
ature has risen close to the optimum for daisy growth. black daisies enhances their growth rate. The long
period of stable, regulated temperature represents a
predominance of negative feedback. However, if the
temperature of the planet is greatly perturbed by the
80
removal of a large fraction of the daisy population,
then positive feedback acts to rapidly restore com-
60
fortable conditions and widespread life. The end of
s- 40
Dead planet regulation is characterized by a positive feedback
2 decline in white daisies - solar warming triggers a
4-
cz d reduction in their population that amplifies the rise in
F
20
0
/‘
,I- With daisies temperature.
The modeling approach pioneered in Daisyworld
provided the beginnings of a theoretical basis for
understanding planetary self-regulation. Subsequent-
(A) -20
ly, Lovelock began to refer to Gaia as a theory, in
0.8 I I which self-regulation is understood as a property of
0.7 - Black White the whole system of life tightly coupled to its environ-
h daisies daisies ment. This replaced the original hypothesis that
0.6 -
z regulation is ‘by and for the biota’. The term ‘homeo-
5
-
0.5 - stasis’, which refers to regulation around a fixed set
E0 0.4 - point, was also revised, with Margulis’ more appro-
.-
5 0.3 - priate suggestion of ‘homeorrhesis’, which describes
2 regulation around an evolving point.
LI 0.2 -
0.1 -
0 I I
0.5 0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 1.5 1.6 1.7
Regulation of Atmospheric
(6) Solar luminosity (normalized) Composition over Earth History
Figure 2 The Daisyworld model (Watson and Lovelock (1983)). The Daisyworld modeling approach was adapted to
A thought experiment to demonstrate that planetary self-regulation study the regulation of climate and atmospheric
can emerge from natural selection at the individual level between composition on the early Earth, through the Archean
types of life with different environment-altering traits. The traits are
Eon (4.0-2.5 billion years ago) and the first half of the
‘darkness’ (albedo = 0.25) and ‘paleness’ (albedo = 0.75) of black
and white daisies on a gray planet (albedo = 0.5). (A) Planetary Proterozoic Eon (2.5-1.5 billion years ago). The
temperature as solar luminosity increases, with daisies and without Archean was characterized by chemically reducing
(‘dead planet’). (6) Areal cover of black and white daisies. conditions at the Earth’s surface and the Proterozoic
818 GAIA HYPOTHESIS
by oxidizing conditions. Lovelock proposed that respiration, and weathering that reduces the carbon
methane was the chemically dominant gas in the dioxide content of the atmosphere. Over Earth’s
Archean atmosphere (whereas oxygen dominated sub- history, progressively stronger biological amplifica-
sequently). The model comprised a bacterial ecosystem tion of rock weathering has evolved, culminating in
of oxygen-liberating photosynthesizers (cyanobacteria), the rise of vascular plants over the last 420 million
methanogens, and aerobic consumers, together with years. Biologically amplified weathering has made
atmospheric carbon dioxide, methane, and oxygen, carbon dioxide relatively scarce in the Earth’s atmos-
and global temperature. The carbon fixed in photo- phere (Figure l A ) , and cooled the Earth by 2O4O0C,
synthesis was returned either by aerobic respiration, as thus counteracting the effect of increasing solar
carbon dioxide, or by methanogenesis, as a mixture of luminosity.
methane and carbon dioxide. An atmosphere domi- Over the last -350 million years the oxygen
nated by carbon dioxide and methane with only traces content of the atmosphere has been remarkably stable.
of oxygen was predicted. This could have provided Continuous records of charcoal and vegetation indi-
sufficient ‘greenhouse effect’ to counteract the faint cate that there has been sufficient oxygen to sustain
young Sun. natural fires throughout this time, but fires have never
Although oxygen-liberating photosynthesis origi- been so frequent as to prevent forests regenerating.
nated early in the history of life, oxygen remained This sets bounds of roughly 15-25% on the oxygen
scarce ( < 0.0008 atm; 0.8 hPa) in the Archean atmos- content of the atmosphere. The average amount of
phere, because it was consumed in the oxidation of an time an oxygen atom spends being recycled between
abundant supply of reduced matter that was contin- organisms, atmosphere, and ocean before being re-
ually being replenished by geological activity. The moved in oxidation of rocks is about 3 million years.
weathering profiles of ancient soils indicate that Hence, the whole oxygen reservoir has been replaced
oxygen rose to > 2 hPa and probably > 30 hPa in a over 100 times, while its size has remained close to
global oxidation event 2.2-2.0 billion years ago. constant. This demands that some self-regulating
Lovelock’s model predicted this rise of oxygen as the feedback mechanisms exist. The removal process for
supply of reduced matter began to be exhausted. Once oxygen is saturated: virtually all the reduced matter
there were two molecules of oxygen for each molecule exposed gets oxidized. Hence regulation of atmos-
of methane, oxygen became the chemically dominant pheric oxygen is thought to involve negative feedback
gas. Methane rapidly disappeared from the atmos- on the source of oxygen. Over geological time, the
phere, reducing the greenhouse effect and cooling burial flux of organic carbon in new sediments
the planet, perhaps causing the Huronian glacia- corresponds to the small excess of oxygen liberated
tion which occurred roughly 2.3 billion years ago. in photosynthesis over that consumed in respiration,
However, climate regulation soon recovered in the which provides a net source of oxygen to the atmos-
model, with carbon dioxide as the dominant green- phere. This is balanced by a net sink due to the
house gas. oxidation of organic matter in sedimentary rocks
Since the Archean, long-term climate regulation is exposed on the continents.
thought to have hinged on changes in the carbon The burial of organic carbon can be somewhat
dioxide content of the atmosphere, and the resultant enhanced under anoxic conditions in sediments,
‘greenhouse effect’ on Earth’s temperature. Over probably because anaerobic consumers are less effi-
million-year time scales, the carbon dioxide reservoir cient than their aerobic counterparts. Hence declining
in the atmosphere and ocean is primarily determined oxygen may be counteracted by more efficient organic
by the balance of input from volcanic and metamor- carbon burial, but the effect appears to be too
phic degassing and removal in the process of weath- weak to stabilize atmospheric oxygen. Marine pro-
ering of silicate rocks on land and subsequent ductivity has a more dominant effect on organic
formation of carbonate rocks in the ocean. A chemical carbon burial, and it in turn depends on the supply of
negative feedback mechanism exists whereby increas- nutrients, especially phosphorus, over long time
es in planetary temperature are counteracted by scales. The burial of phosphorus in organic matter
increases in the rate of silicate rock weathering and and bound to iron minerals is less efficient under
the uptake of carbon dioxide. However, the rate of anoxic conditions. Hence, declining oxygen should
rock weathering is greatly enhanced by the activities of cause more phosphorus to be recycled to the water
soil microbes, lichens, mosses, and vascular plants. column, fuelling more productivity and increased
This biological amplification offers the potential for organic carbon burial. However, such mechanisms
more responsive stabilization of the Earth’s tempera- are ineffective against rising oxygen, because it tends
ture. For example, rising carbon dioxide and temper- to remove anoxia from the ocean, thus switching off
ature trigger increased plant growth, microbial the feedback.
GAIA HYPOTHESIS 819
Weathering of phosphorus-bearing rocks is the would have added to regional and planetary cooling
ultimate source of all phosphorus supplied to the and may have been critical for the inception of ice
land and ocean. Vascular plants amplify the rate of sheets.
rock weathering by about an order of magnitude Marine phytoplankton cool the climate by pump-
relative to primitive land biota (e.g., lichen and moss ing down atmospheric CO2 and producing dime-
thylsulfide (DMS), which ultimately increases
cover) and the effect is greatest for trees with their deep
rooting systems. Increasing atmospheric oxygen tends cloud albedo. DMSP (dimethylsulfoniopropionate),
to suppress vegetation by inhibiting photosynthetic the precursor of DMS is produced in widely varying
carbon fixation and increasing fire frequency. Fires amounts by different species of marine phytoplank-
tend to trigger ecological shifts from forest to faster- ton. Its conversion to DMS is catalyzed by the enzyme
regenerating ecosystems such as grassland. By these DMSP lyase and is enhanced by virus infection
mechanisms, rising oxygen should suppress rock and zooplankton grazing. The main reservoir of
weathering and hence reduce the supply of phospho- DMS is in the ocean, where it is consumed by bacteria
rus to the land and ocean, in turn suppressing and oxidized to dimethylsulfoxide (DMSO). A'ir-sea
productivity and organic carbon burial. This mecha- exchange results in a net flux of DMS to the
nism is extremely effective at regulating against rising atmosphere (Figure 1B). In the atmosphere, DMS is
oxygen because of the high sensitivity of fire frequency oxidized in a range of reactions. The main pathway
to rising oxygen. Furthermore, declining oxygen is generates sulfur dioxide, which is further oxidized to
counteracted by increases in plant productivity, rock sulfate, and can ultimately contribute to sulfate
weathering, phosphorus supply, and organic carbon aerosol formation. The aerosol particles grow,
burial. often in combination with another biogenic gas,
ammonia (Figure l B ) , to become cloud condensation
nuclei (CCN).Increases in the number density of CCN
Contemporary Climate Feedbacks make clouds more reflective, increasing the scattering
As well as its role in regulating oxygen and carbon of solar radiation back to space and thus causing
dioxide over long time scales, vegetation also has self- cooling.
sustaining short-term feedback effects on climate. Temperature both directly affects phytoplankton
Globally, plants increase land surface evapotranspira- growth and determines the degree of stratification in
tion and continental precipitation and reduce temper- the ocean water column, and hence the supply of
ature variability. These climatic effects increase net nutrients to the surface layers. Therefore, there is the
primary productivity and biomass, and without them potential for feedback on climate involving the growth
it has been predicted that the boreal, Amazonian, and of DMS-emitting phytoplankton. Originally, a nega-
South East Asian forests would disappear. Such tive feedback was proposed whereby a reduction in
hysteresis of the vegetation-climate system may also temperature and light beneath clouds reduces photo-
exist in the south-western Sahara, where models synthesis and restricts the spread of DMS producers.
predict that vegetation could sustain itself, by main- Subsequent modeling elaborated this proposal with
taining a wetter climate. Vegetation tends to cool the the observation that the formation of a thermocline at
Tropics and temperate regions but warm the high -10°C limits the supply of nutrients to the
northern latitudes. surface ocean, thus setting an effective optimum for
The trees of the boreal forests possess traits of plankton growth. Beneath this temperature lies the
shedding snow and darkness that give them a low originally proposed regime of negative feedback.
albedo (reflectivity) and make them warmer than their Above it, however, an increase in temperature may
surroundings. In this respect they can be likened to the be amplified by a decrease in photosynthetic produc-
dark daisies of Daisyworld. The presence of forest tion, DMS production, and cloud reflectivity, gener-
warms the region, and the Northern Hemisphere, by ating positive feedback. Evidence that DMS
-4°C in winter. The system shows constrained production in the Southern Hemisphere was enhanced
positive feedback that amplifies regional temperature during the last Ice Age indicated that the feedback
changes. Six thousand years ago, orbital forcing may then have been negative but switched to
warmed the high-latitudes and triggered the boreal become positive as temperatures rose at glacial
forest to spread northward and amplify the initial termination.
warming. One hundred and fifteen thousand years ago
-
The cycles of ice ages and interglacial warm periods
the opposite occurred; orbital forcing cooled the high- that have characterized the last 2.5 million years of
latitude summer, triggering a southward spread of the Earth history appear, at first glance, to conflict with the
tundra, replacing the boreal forest. The resulting view that the Earth is self-regulating. The trace gas
increase in albedo because of unmasked snow cover composition of the atmosphere, including CO2, CH4,
820 GAIA HYPOTHESIS
and N 2 0 , has varied. However, the recently extended involving the biological amplification of silicate rock
Vostok ice core record reveals that for the last four weathering appears to be extremely robust to short-
glacial cycles the frequency, bounds, and amplitude of term perturbation. It should continue to gradually
the oscillations have been remarkably constant, de- reduce the C 0 2 content of the atmosphere as
spite highly variable forcing (solar insolation). This is solar luminosity increases, and this will encourage
indicative of a regulatory system, but one that is near plants with CO2-concentrating mechanisms to be-
the limits of its operation, with positive feedback come dominant. Life may eventually perish as a
coming to dominate over negative feedback. The long- result of lack of C 0 2 , overheating by the Sun, or
term climate regulator involving biological amplifica- catastrophic perturbation. However, models based on
tion of silicate rock weathering is near the lower bound the current biota and feedback mechanisms predict
of its operation, having reduced atmospheric C 0 2 that complex life will last at least another 0.5-1.2
near to the lower limit for the growth of most plants billion years.
(which lack a CO2-concentrating mechanism). Posi-
tive feedback is apparent in the onset and termination See also
of ice ages, including the aforementioned changes in BiogeochemicalCycles: Carbon Cycle; Nitrogen Cycle;
boreal forest cover and DMS emissions. Hence, Sulfur Cycle. Carbon Dioxide. Climate Prediction
humans may be perturbing the Earth system when it (Empiricaland Numerical).Evolution of Atmospheric
is unusually vulnerable and has the potential to switch Oxygen. Evolution of Earth’sAtmosphere. Methane.
to a different state. Planetary Atmospheres: Mars; Venus. Reflectance
At present, members of both the marine and and Albedo, Surface. Teleconnections. Tropospher-
terrestrial biota are involved in processes that are ic Chemistry and Composition: Ammonia and Ammo-
removing more than half of the excess carbon dioxide nium; Biogenic Hydrocarbons (inc. isoprene); Sulfur
released to the atmosphere each year by human Chemistry, Organic.
activities. This negative feedback is not sufficient to
prevent the carbon dioxide content of the atmosphere Further Reading
from rising, but it is damping the rate of rise.
Atmospheric C 0 2 and global warming are expected Charlson RJ, Lovelock JE, Andreae M O and Warren SG
to peak some time in the present millennium, the (1987) Oceanic phytoplankton, atmospheric sulphur,
cloud albedo and climate. Nature 326: 655-661.
precise time depending on how fast the fossil fuel
-
reserve is burned. Over the following 10 000 years,
the acidic CO2 added to the atmosphere by human
Lenton T M (1998) Gaia and natural selection. Nature 394:
439-447.
Lenton T M and Watson AJ (2000) Redfield revisited. 2.
activities should be neutralized by the dissolution of What regulates the oxygen content of the atmosphere?
carbonate sediments in the ocean and the weathering Global Biogeochemical Cycles 14: 249-268.
of carbonate rocks on land, processes that increase the Lovelock JE (1979) Gaia: A N e w Look at Life on Earth.
alkalinity of the ocean. However, major reorganiza- Oxford: Oxford University Press.
tions of the climate system could occur in the mean- Lovelock JE (1988) T h e Ages of Gaia: A Biography o f O u r
time. Boreal forests are already amplifying winter Living Earth. Oxford: Oxford University Press.
warming in the northern high latitudes. Global Lovelock JE (1991) Gaia: T h e Practical Science of Planetary
warming and resultant stratification of the ocean Medicine. London: Gaia Books.
Lovelock JE and Margulis LM (1974) Atmospheric home-
may trigger a decline in phytoplankton and their
ostasis by and for the biosphere: The gaia hypothesis.
cooling effect via DMS emissions, providing a further Tellus 26: 2-10.
positive feedback. Schneider SH and Boston PJ (1991) Scientists on Gaia.
Although human perturbation may shift the Earth Boston: MIT Press.
system to a state that is uncomfortable for us as a Watson AJ and Lovelock JE (1983) Biological homeostasis
species, it is highly unlikely to destroy all life on Earth. of the global environment: The parable of Daisyworld.
The mechanism of long-term climate regulation Tellus 35B: 284-289.
GENERAL CIRCULATION / Overview 821
Contents
Overview
Energy Cycle
Mean Characteristics
Momentum Budget
Models
flank and moist, ozone-poor tropospheric air below

Overview and on their equatorward flank. (Tropospheric and
stratospheric air can be distinguished in Figure 1by the
J M wallace,university of Washington, Seattle, WA,
marked difference in lapse rate, which determines the
USA vertical spacing between the isotherms.) The strength
of the jet streams varies with season: They are
strongest during winter, when the north-south tem-
perature gradient in the troposphere is strongest. An
Introduction even stronger westerly ‘polar night jet’ is observed
The atmospheric general circulation encompasses the in the middle and upper stratosphere of the win-
planetary-scale wind systems that shape the Earth’s ter hemisphere, separating the cold polar cap region
climate. Features of interest include the belts of mid- from the less cold sunlit portion of the hemi-
latitude westerlies and subtropical trade winds at the sphere. The Southern Hemisphere polar night jet
Earth’s surface, the jet streams aloft, and the storm is much stronger than its Northern Hemisphere
tracks. counterpart.
The general circulation can be partitioned into Figure 2 shows the zonally averaged mean merid-
zonally symmetric and eddy components, where ional (i.e. nortldsouth) circulations for the same
‘zonally symmetric’ denotes longitudinally averaged seasons as Figure 1 and also for the annual mean.
(i.e., averaged around latitude circles) and ‘eddy’ The trade wind belt is characterized by equatorward
denotes departure from the longitudinal average. The flow at the Earth’s surface and the belt of surface
zonally symmetric flow is a function of latitude and westerlies at higher latitudes by poleward flow. These
height only, and certain properties of the eddies such as low-level meridional flows constitute the lower
their root mean squared amplitude in various fields branches of closed circulation cells extending through
can also be longitudinally averaged and displayed in the depth of the troposphere. By far the strongest of
the form of two-dimensional (latitude versus height) these cells is the tropical ‘Hadley cell’, whose rising
cross-sections. Understanding the climatology of such branch delineates the belt of heaviest tropical rainfall
zonally averaged fields is the most fundamental and whose sinking branch is marked by dry condi-
problem of the general circulation. tions. Weaker, ‘Ferrel cells’ in which air circulates in
Figure 1 shows the zonally symmetric component of the opposite sense to the Hadley cell, are discernible in
the zonal wind and temperature fields for (A) Decem- the mid-latitude circulation.
ber through February (DJF) and (B) June through The rising branch of the Hadley cell is located at
August (JJA).At the Earth’s surface, easterly (i.e., east tropical latitudes of the summer hemisphere, while the
to west) ‘Trade Winds’ prevail equatorward of 30” sinking branch is at cooler subtropical latitudes of the
latitude, while westerlies prevail at higher latitudes. winter hemisphere. Such circulations, characterized
The westerlies strengthen with height and peak at the by the rising of warmer air and the sinking of cooler air,
10 km level just poleward of 30” latitude. These zonal are said to be ‘thermally direct’, regardless of their
wind maxima, commonly referred to as the ‘tropo- configuration or where they occur in the atmosphere.
spheric jet streams’ form part of the boundary between Likewise, circulations like the Ferrel cell, marked by
troposphere and stratosphere, with dry, ozone-rich the rising of cooler air and the sinking of warmer air,
stratospheric air above them and on their poleward are said to be ‘thermally indirect’.
822 GENERAL CIRCULATION / Overview
10 30
25
100 20
a
a h
E
5 Y
v
g 200
S 15 &
.-
v)
E 300
2
10
500
700 5
1000
90"s 60"s 30"s 0" 30"N 60"N 90"N
(6) Latitude
Figure I Pole-to-pole cross-sections showing zonally averaged and time-averaged temperature (dashed contours, 5°C interval) and
zonal wind (solid contours, 5 m s - I interval) for the seasons December through February (A) and June through August (6). The 0°C
isotherm is thickened and regions of easterly winds are shaded. Based on NCEP/NCAR Reanalyses. (Diagram provided by Socorro
Medina.)
The circulation cells pictured in Figure 2 represent Balance Requirements

seasonal averages and zonal averages of wind meas-
The general circulation must satisfy a number of so-
urements at fixed points in space. The circulation
called 'balance requirements' relating to fundamental
patterns derived from such measurements are
conservation laws, posed as statements to the effect
said to be 'Eulerian mean circulations'. In contrast,
that for any specified region such as the tropics or the
if one were t o infer the sense of the mean meridional
Arctic, in the time mean over a sufficiently long time
circulations by tagging millions of air parcels, track-
(such as a season), sources, sinks, and imports or
ing them as they move through the atmosphere
exports of conserved quantities must sum to zero. For
over an extended period, and seasonally and zonally
the atmosphere as a whole, sources must equal sinks.
averaging these air parcel trajectories, one would
obtain quite a different impression of the mean
Available Potential and Kinetic Energy
meridional circulations. In this so-called 'Lagrangian
mean circulation' the Hadley circulations extend into Frictional dissipation observed within the planetary
high latitudes and the mid-latitude Ferrel cells virtu- boundary layer and within patches of turbulence
ally disappear, as indicated on the right-hand side of within the free atmosphere is continually depleting the
Figure 3. kinetic energy of planetary-scale wind systems. Half
GENERAL CIRCULATION/ Overview 823
'available potential energy' inherent in the distribu-

tion of atmospheric mass. Available potential energy
200
300 10 depends upon the height of the atmosphere's center of
mass relative to mean sea level. It is released whenever
500 5
700 the center of mass is lowered through the sinking of
innn colder, denser air and the rising of warmer, less dense
(A)%i"S 60"s 30"s 0" 30"N 60"N 90"N
h 100- "
air in thermally direct circulations like the Hadley
m cell. Such circulations also act to flatten out the
potential temperature surfaces, thereby weakening
existing horizontal temperature gradients on pressure
surfaces.
The kinetic energy released by such thermally direct
circulations is imparted not to the vertical component
100 of the motion, which is so small as to be inconsequen-
115
tial with respect to kinetic energy, but directly to the
horizontal component of the flow, which is pushed
across the isobars from higher to lower pressure by the
horizontal pressure gradient force. Such cross-isobar
flow toward lower pressure is prevalent close to the
(C) 90"s 60"s 30"s 0" 30"N 60"N 90"N
Earth's surface, where the dissipation of kinetic energy
Latitude is most intense. For example, the trade winds in the
Figure 2 Pole-to-polecross sections showing zonally averaged lower branch of the Hadley cell (Figure 2 ) are directed
and time-averaged mean meridional circulations for the seasons down the pressure gradient, out of the subtropical high
December through February (A) and June through August (B), and pressure belt and into the belt of low pressure that
for the calendar year (C). Scaling is as indicated by the arrows at coincides with the rising branch of the Hadley cell in
lower right. Based on NCEP/NCAR Reanalyses.(Diagram provid-
equatorial latitudes. The poleward flow in the upper
ed by Socorro Medina.)
branch of the Hadley cell is also down the pressure
gradient, as evidenced by the existence of westerly
the energy would be gone within a matter of days were winds at that level, which implies (from the geostro-
there not some mechanism continually operating to phic wind equation) that pressure decreases with
restore it. The source of this kinetic energy is the latitude. Thermally direct circulations like the Hadley
subtropical deserts
Lagrangian-mean circulation
Figure 3 The Lagrangian mean circulation.

824 GENERAL CIRCULATION/ Overview
cell, which are characterized by the rising of warmer, The rate of dissipation of kinetic energy is sufficient to
lighter air and the prevalence of cross-isobar horizon- deplete the global reservoir of kinetic energy in only a
tal flow toward lower pressure, release available week, but if available potential energy as well as
potential energy and convert it to the kinetic energy kinetic energy is taken into account, the time scale for
of the horizontal flow. Circulations like the Ferrel cell, depleting (orrecharging) the system is on the order of a
which operate in the opposite sense, are referred to as month.
‘thermally indirect’. For the atmosphere as a whole,
thermally direct circulations are prevalent. In the
Angular Momentum
absence of diabatic heating and friction, the sum of the
available potential and kinetic energy is conserved: The angular momentum of an air parcel is given by
e.g., for a thermally direct circulation, the available (QR cos 4 + u ) R cos +6m, where Q is the angular
potential energy released is equal to the kinetic energy velocity of the Earth’s rotation, R is the radius of the
generated. Earth, 4 is the latitude and 6m is the mass of the air
Since thermally direct circulations are continually parcel. Apart from small tidal interactions with the
depleting the atmosphere’s reservoir of available Moon, the total angular momentum of the atmosphere
potential energy reservoir, something must be operat- plus oceans plus solid Earth is conserved. The angular
ing to restore it. Heating of the atmosphere by momentum of the ocean is very small, owing to the
radiative transfer and by the release of the latent heat inhibition of circumpolar currents by the continents.
of condensation of water vapor in clouds acts to Hence, whenever the atmosphere gains angular mo-
restore the available potential energy in two ways: (1) mentum it tends to be at the expense of the solid Earth,
by warming the atmosphere in the tropics (where the and vice versa. A strong correlation between length of
sum of the condensation heating and the absorption of day and atmospheric angular momentum is, in fact,
incoming solar radiation exceeds outgoing infrared observed on time scales ranging from days up to a few
radiation) and cooling it at higher latitudes, where the years. O n longer time scales, slow motions within the
reverse is true; and (2)by heating the air in the lower Earth’s molten core also affect the length of day.
and middle troposphere, where most of the conden- Westerly winds circulate around the Earth’s axis in
sation heating takes place, and cooling it at higher the same sense as the Earth’s rotation. Hence, air
levels, where infrared cooling to space prevails. parcels in the atmosphere’s westerly wind belts rotate
Mechanism (1)acts to maintain the equator-to-pole more rapidly than the solid Earth does, and air parcels
temperature contrast on pressure surfaces; and mech- in easterly wind belts rotate more slowly. It follows
anism (2) acts to expand the air in the lower tropo- that the frictional drag that is acting to slow the
sphere and compress the air in the upper troposphere, tropical trade winds has the effect of transferring
thereby lifting the air at intermediate levels, which acts angular momentum from the solid Earth to the
to maintain the height of the atmosphere’s center of atmosphere. In a similar manner, frictional drag on
mass against the lowering produced by thermally the surface westerlies that prevail poleward of 30”
direct circulations. Hence, the maintenance of the latitude transfers angular momentum from the atmos-
atmospheric general circulation requires both hori- phere back to the solid Earth. The torques (force times
zontal and vertical heating gradients. The above can be distance from the Earth’s axis) that the atmosphere
summarized in terms of a ‘kinetic energy cycle’ as exerts upon the solid Earth by virtue of the atmospheric
depicted in Figure 4, with generation of available pressure difference between the east and west slopes of
potential energy ( G )by diabatic heating, conversion of large, north-south oriented mountain ranges like the
available potential energy to kinetic energy ( C ) by Rockies and Andes also transfer angular momentum
thermally direct circulations, and the dissipation of between the atmosphere and solid Earth. In the North-
kinetic energy ( D ) . In the long term mean, for the ern Hemisphere this effect is of the same sign and similar
atmosphere as a whole, in magnitude to that of the frictional torques on the
surface winds. Hence, there exists a balance require-
G=C=D ment for a poleward transport of angular momentum
within the atmosphere. The transport must be largest
near 30” latitude, which marks the transition between
G C D
A K b the tropical trade winds and the mid-latitude surface
westerlies, as depicted in Figure 3.
Figure4 Kinetic energy cycle showing available potential energy
Angular momentum can be transported poleward
(A) and kinetic energy ( K ) reservoirs, the generation of available across a latitude circle either by a systematic poleward
potential energy by diabatic heating (G), the conversion due to flux of atmospheric mass, or by ‘exchange processes’
thermally direct circulations (C), and the frictional dissipation (D). in which there is no net mass flux but poleward-
GENERAL CIRCULATION / Overview 825
moving air parcels carry with them more angular

momentum (i.e., they have a stronger westerly wind
component) than equatorward-moving parcels. The
net mass flux in the Earth's atmosphere is very small, as
evidenced by the fact that mean surface pressure over
the region poleward of 30" latitude is not systemati-
cally changing with time and there are no appreciable
internal sources and sinks of atmospheric mass other
than water vapor, which accounts for too small a
fraction of the mass of the atmosphere to yield an
appreciable transport. Hence, exchange processes Longitude
must be responsible for the poleward transport of
angular momentum. Figure 6 Idealized sketch showing the relationship between the
tilt of the eddies and the poleward transport of westerly momentum.
The exchange processes, in turn, may be divided In order for poleward-moving air to have a stronger westerly wind
into two kinds: those involving the Eulerian mean component than the equatorward return flow, the eddies must have
meridional circulations pictured in Figure 2 and those a south-west-north-east tilt. as shown.
involving the eddies, as depicted in Figure 3 upper
right. In the annual average, 30" latitude coincides of air parcels of equal mass, but containing differing
with the boundary between the Hadley and Ferrel amounts of angular momentum, results in a net
cells. Hence the mean meridional motions cannot be upward transport of angular momentum. In a similar
contributing to the required poleward transport across manner, the Ferrel cell is instrumental in transporting
that latitude: The eddies must be responsible. The eddy angular momentum downward from the jet stream
fluxes, whose distribution is shown in Figure 5 do, in level to the mid-latitude surface westerlies where it is
fact, exhibit a systematic poleward transport of lost to the solid Earth via frictional and mountain
angular momentum across 30", which turns out to torques. The resulting upward/poleward/downward
be just enough to satisfy the balance requirement transport of angular momentum is depicted in the
inferred from the frictional torques on the surface upper left quadrant of Figure 3.
winds and the mountain torques. To accomplish the
required transport, the horizontal flow in the eddies
Total Energy
must exhibit a preferred tilt, as pictured in Figure 6 and
such a tilt is, in fact, observed most of the time. It is Averaged over the year, the net radiation at the top of
evident from Figure 5 that most of the poleward flux of the atmosphere is downward at low latitudes and
angular momentum takes place around the jet stream upward at high latitudes. These local imbalances
(- 10 km) level, where the amplitude of the eddies is reflect the differing meridional profiles of insolation
largest and the eastward tilt with increasing latitude is (incoming solar radiation) and outgoing terrestrial
most pronounced. The Hadley cell is instrumental in radiation pictured in Figure 7. The former slopes
transporting the angular momentum acquired by the steeply downward toward the poles, reflecting the
trade winds upward to the jet stream level. The air in strong latitudinal dependence of solar zenith angle,
its rising (equatorial) branch contains much more while the latter mirrors the more gently sloping profile
angular momentum per unit mass than the air in its of the infrared radiation emitted to space from clear
sinking (subtropical) branch by virtue of its larger air, from cloud tops, and from the underlying surface.
distance from the Earth's axis. The vertical exchange Since the equator to pole temperature gradient is not
observed to be increasing systematically with time,
3 100 15-
these local imbalances must be fully compensated by
Q
-?? 200
c
Y
10E
E the poleward transport of energy from the low-
latitude 'surplus' region to the high latitude 'deficit'
3 300
v) .-m region. The transport must be largest in mid-latitudes
500 5 %
a 700 where the curves in Figure 7 cross one another.
1000 The atmosphere and oceans both contribute to the
90"s 60"s 303S 0" 30"N 60"N 90"N
poleward transport of energy. In both media the
Latitude
transport involves exchange processes, with pole-
Figure 5 Annual mean northward flux of westerly momentum due
ward-moving fluid parcels carrying with them larger
to the horizontal motions in eddies, expressed in units of m2s -', To amounts of total energy than equatorward-moving
obtain angular momentum flux, multiply by R cos 4. Based on parcels, by virtue of their having recently been heated
NCEP/NCAR Reanalyses. (Diagram provided by Socorro Medina.) while residing at low latitudes. Both the shallow
826 GENERAL CIRCULATION / Overview
90"s 60"s 30"s 0" 30"N 60"N 90"N L H L H

Latitude Figure 8 Idealized sketch of the thermally direct circulation in the
zonal plane in a developing baroclinic wave. Solid lines represent
pressure surfaces with slopes exaggerated by a factor of about 5:l.
Figure 7 Annual-mean net incoming solar radiation (solid line)
The long dashed lines show the vertical tilt of the wave axes in the
and outgoing terrestrial radiation (dashed line) as a function of
pressure field. Heavy lines indicate the position of the tropopause.
latitude, expressed in units of Wm-'. Distance on the latitude
Short dashed lines labeled P and E show axes of poleward and
scale is proportional to area on the Earth's surface. (Diagram
equatorward moving air respectively. Arrows denote the vertical
provided by Socorro Medina.)
motions and associated cross-isobar flow. (Adapted with permis-
wind-driven gyre circulations and the deeper ther- sion from Wallace JM and Hobbs PV (1977) Atmospheric Science:
An lnfroducfory Survey. New York: Academic Press.)
mohaline circulation contribute to the oceanic trans-
port: The gyres by transporting warm water poleward
in the western boundary currents and equal masses of providing the kinetic energy needed to amplify the
cooler water equatorward on the eastern side of the waves. Baroclinic waves tilt westward with increasing
oceans, and the thermohaline circulation by trans- height: The wave troughs in the pressure field at the
porting relatively warm surface water poleward and 5 km level overlie the cold air masses at the Earth's
cold bottom water equatorward, primarily in the surface, as pictured in Figure 8.
North Atlantic. More than half of the total required In the Northern Hemisphere wintertime, planetary-
transport is accomplished by atmospheric eddies, in scale stationary waves (sometimes also referred to as
which poleward-moving air parcels carry with them 'standing eddies') forced by the large thermal contrasts
greater amounts of sensible and latent heat than do between the cold continents and the warmer oceans
equatorward-moving parcels. and by the existence of the Rockies and the Himalayas
The thermally indirect Ferrel circulation does not also make an appreciable contribution to the poleward
transport energy poleward across middle latitudes: transport of sensible heat (see Stationary Waves
When the gravitational potential energy of air parcels (Orographic and Thermally Forced)). The Icelandic
is taken into account as well as the latent and sensible and Aleutian lows are prominent stationary wave
heat, the equatorward-moving air parcels in its upper features in the wintertime mean sea-level pressure
branch carry with them more energy per unit mass pattern. Poleward flow to the east of these low-level
than the warmer, more moist poleward-moving par- cyclones carries warm air northward, keeping western
-
cels in its lower branch. Eastward-moving baroclinic
waves, with wavelengths of 4000 km, are respons-
ible for most of the poleward eddy transport
Europe and south-east Alaska relatively warm com-
pared to other regions at the same latitude, and the
equatorward flow of cold air to the west of them
of sensible and latent heat across middle latitudes. contributes to the coldness of eastern Canada and
These waves derive their energy from the prevailing Siberia. The Northern Hemisphere wintertime sta-
north-south temperature gradient (see Baroclinic tionary waves tilt westward with height and they
Instability). At the Earth's surface they are marked penetrate upward into the stratosphere. The poleward
by intensifying cyclones (gyres that circulate in the heat transport by the eddies peaks in mid-latitudes, in
same sense as the Earth's rotation) attended by sharp the lower troposphere, as shown in Figure 9 and in the
frontal boundaries that separate warm, poleward- lower left quadrant of Figure 3.
moving air masses on the cyclones' eastern flanks from
cold, equatorward-moving air masses on their western
Water Vapor and Ozone
flanks. The meridional displacements of warm and
cold air masses in the waves serve to sharpen the east- The mass of water vapor in the atmosphere is not
west temperature gradients. Meanwhile, the rising of changing appreciably with time, and the residence
the poleward-flowing warm air masses, in combina- time of individual water vapor molecules in the
tion with the sinking of the equatorward-flowing cold atmosphere is only about a week. Hence, for the globe
air masses, serves to release available potential energy, as a whole, a balance between evaporation and
GENERAL CIRCULATION /Overview 827
A remarkable feature of the distribution of water

vapor is the dryness of the lower stratosphere, believed
to be a consequence of the Lagrangian mean circula-
tion, in which air parcels enter the stratosphere
through the very cold tropical 'tropopause' (Figures
1 and 3) move poleward and eventually re-enter the
60"s 30"s 0" 30"N 60"N 90"N troposphere at higher latitudes in discrete intrusions in
Latitude the vicinity of the jet stream. The same circulation
carries ozone poleward and downward from its
Figure 9 Annual mean northward heat flux due to the horizontal photochemical source region in low latitudes around
motions in eddies, expressed in units of K m s - I . To obtain sensible the 30 km level.
heat flux, multiply by specific heat at constant pressure
1004J k g - ' K - ' . Based on NCEP/NCAR Reanalyses for the
period 1958-1997. (Diagram provided by Socorro Medina.)
What Determines the Character
precipitation is required on time scales as short as
a few days; but on a regional basis there are large
of the General Circulation?
imbalances. Precipitation exceeds evaporation by a To understand why the observed planetary-scale wind
factor of 2 or more in the equatorial rain belts and in systems have the particular geographical and seasonal
the mid-latitude storm tracks and an imbalance in the dependence and amplitude that they do, scientists rely
opposite sense prevails in the subtropical belt, which upon experience gained from working with numerical
encompasses most of the major deserts and cloud-free models based on the same conservation laws that the
maritime anticyclones. The surplus water vapor balance requirements considered in the previous
acquired by the subtropical atmosphere is transported section are based on. When such a model atmosphere
poleward and equatorward, as illustrated in Figure 10. is 'turned on', starting from a state of rest (stably
The eddies are responsible for most of the poleward stratified, a horizontally uniform temperature distri-
transport of water vapor into the mid-latitude storm- bution, flat pressure surfaces, and no wind) the tropics
tracks, while the steadier trade winds that constitute warm and the polar regions cool in response to the
the lower branch of the Hadley cell are responsible for imposed distribution of radiative heating, which is
most of the equatorward transport into the tropical designed to mimic that in the real atmosphere. As the
rain belts, as depicted in Figure 3. tropical atmosphere warms, thermal expansion causes
Regional imbalances between evaporation and pressure surfaces in the upper troposphere to bulge
precipitation also have implications for the oceanic upward relative to the cooling air at the higher
thermohaline circulation. Excess precipitation fresh- latitudes, as depicted in Figure 11A. The sloping of
ens the surface waters, decreasing their density and the pressure surfaces (downward toward the pole)
thereby rendering them more stably stratified. The gives rise to a poleward flow at the upper levels as
impact of the atmospheric hydrological cycle upon the depicted in the figure. The poleward mass flux causes
oceanic circulation is believed to be particularly mass to accumulate and sea-level pressure to rise at
important in the regions of bottom water formation high latitudes, driving a compensating equatorward
over the North Atlantic Ocean and the Weddell Sea. low-level flow. Hence, the initial response to the
heating gradient is the development of a thermally
2.5 1 direct circulation reminiscent of the Hadley cell, but
extending all the way from equator to pole, as shown
in Figure 11B. The conservation of angular momen-
tum imparts a westward component to the equator-
ward flow in the lower branch of the cell and an
eastward component to the poleward flow in the upper
branch. The vertical wind shear between the low-level
easterlies and the upper-level westerlies increases in
90"s 60"s 30"s 0" 30" N 60" N 90" N proportion to the strengthening north-south temper-
Latitude ature gradient, in accordance with thermal wind
balance (Figure 11C).
Figure 10 Annual-mean precipitation (solid) and evaporation
(dashed) as a function of latitude, expressed in units of meters per
If this thermally direct circulation were able to
year. Distance on the latitude scale is proportional to area on the persist, the winds would eventually increase to the
Earth's surface. Based on NCEP/NCAR Reanalyses for the period point where the rate of frictional dissipation equaled
1958-1 997. (Diagram provided by Socorro Medina.) the rate of generation of available potential energy by
828 GENERAL CIRCULATION /Overview
Low pressure
Figure 11 Schematic depiction of the general circulation as it develops from a state of rest in a climate model for equinox conditions in
the absence of land-sea contrasts. See text for further explanation.
the equator-to-pole heating gradient. However, before

such an equilibrium is reached, the simulated circula-
tion undergoes a fundamental change. When the
meridional temperature gradient reaches a certain
critical value, baroclinic instability spontaneously
breaks out in mid-latitudes. Successive generations
of baroclinic waves, developing, evolving through
their life cycles and decaying, modify the general
circulation as indicated in Figure 11D. They produce a
large poleward eddy heat fluxes that oppose the
further buildup of the meridional temperature gradi-
ent across mid-latitudes. As they disperse upward to
the jet-stream (10km) level and thence equatorward
into the tropics, they are attended by a poleward flux
of angular momentum from the tropics into mid-
latitudes. The wave-induced heat and momentum
fluxes give rise to the observed thermally-indirect mid-
latitude Ferrel cell (Figure 2), which maintains the
surface westerlies against frictional dissipation. Were
it not for the waves, there would be no Ferrel cell, and
the broad belt of surface westerlies poleward of 30" Figure 12 Mean zonal wind at the tropospheric jet stream level
( - 12 km) for the season December through February. Contour
would not exist.
'
interval 10 m s - '; the 20 m s - contour is thickened. Based on
NCEP/NCAR Reanalyses for the period 1958-1 997. (Diagram
provided by Socorro Medina.)
Beyond the Zonally Averaged,
Time-Average General Circulation zonally varying features unaccounted for. For exam-
The zonally averaged diagnostics considered in the ple, it is evident from Figure 12 that the westerly wind
foregoing sections deal with only the broad outlines of maximum near 30"N in Figure 1A is a composite
the general circulation: They leave many important made up of intense wintertime jets over Japan and the
GENERAL CIRCULATION / Energy Cycle 829
10 Nor can the statistics that describe the general

circulation necessarily be regarded as perfectly repro-
ducible, year after year. For example, they are
discernibly different during contrasting years of the
El Nifio Southern Oscillation cycle, particularly over
the Pacific sector during the months of January
through April. There are also indications of longer-
term trends, as shown in Figure 13. Such changes in the
general circulation can occur in response to changes in
sea surface temperature, as in the case of El Nifio, or in
2 response to changes in the distribution of radiative
5 100
E heating brought about by changes in the concentration
3
v) of radiatively active trace gases or aerosols.
E
a 200
300 See also

Baroclinic Instability. Climate Variability: Decadal to
500 Centennial Variability; North Atlantic and Arctic Oscillation.
700 Cyclones, Extra Tropical. Middle Atmosphere:
Planetary Waves; Transport Circulation; Zonal Mean
1000 Climatology. Ocean Circulation: General Processes.
30" N 60"N 90" N Operational Meteorology. Planetary Atmospheres:
Latitude Mars; Venus. Satellite Remote Sensing: Temperature
Soundings. Stationary Waves (Orographic and
Figure 13 Linear trend in annual mean, zonal mean wind during Thermally Forced). Stratosphere-Troposphere Ex-
the 30-year period 1970-1999, expressed in m s - ' change over
change: Global Aspects. Tropical Meteorology: Trop-
the 30-year interval. Contour interval 1 ms-'; negative trends are
shaded. Based on NCEP/NCAR Reanalyses. (Diagram provided ical Climates.
by David W. J. Thompson.)
Further Reading
eastern United States, as contrasted against more
diffuse westerlies over other parts of the hemisphere. Grotjahn R (1993) Global Atmospheric Circulations.
Oxford: Oxford University Press.
In a similar manner, the eddy flux cross-sections
Hartrnann DL (1994) Global Physical Climatology.
presented in Figures 5 and 9 tend to be dominated by Cambridge: Cambridge University Press.
well-defined 'storm tracks' over the oceans. Under- Holton JR (1992) Introduction to Dynamic Meteorology,
standing the zonally varying structure of the general 3rd edn. San Diego: Academic Press.
circulation requires consideration of more complex, James IN (1994) Introduction to Circulating Atmospheres.
three-dimensional balance requirements and numeri- Cambridge: Cambridge University Press.
cal simulations that incorporate careful treatment of Peixoto JP and Oort AH (1992) Physics of Climate. New
land-sea thermal contrasts and mountains. York: American Institute of Physics.
R Grotjahn, University of California, Davis, CA, USA changes from one form to another. Energy properties
Copyright 2003 Eisevier Science Ltd. Ail Rights Reserved. can be analyzed to deduce the strengths of circulations,
as well as the rates at which circulations are created,
maintained, or destroyed.
Introduction
The energy cycle provides a physically meaningful
system through which to understand the many con-
straints and properties of the general circulation.
Energy is conserved and can be tracked even as it
-
--
The total energy (TE)is defined by the relationship:
C,T + gZ + Lq + $(u2+ v2 + w 2 ) = TE
DSE
MSE
KE
[l]
830 GENERAL CIRCULATION / Energy Cycle
where C, is the specific heat capacity at constant Conceptual Models

volume, T is the temperature, g is the acceleration due
Two-fluid Model
to gravity, Z is the geopotential height, L is the latent
heat of vaporization or sublimation, q is the specific A fluid flow analog of the pendulum can illustrate
humidity, and u, v , and w are eastward, northward, forms and conversions of energy. Imagine a tank
and upward wind components. The term C,T repre- holding two fluids of different density, separated by a
sents internal energy, gZ gravitational energy, Lq vertical barrier (Figure 1A). The initial state has the
latent energy from the phase changes of water, and highest center of mass and thus the greatest gravita-
i(u2 + v2 + w 2 ) the kinetic energy (KE). Together the tional PE. If the barrier is suddenly removed, the fluids
first two terms define the dry static energy (DSE), begin to move. The motion accelerates until the point
while including the third defines moist static energy in time where the greatest amount of the denser fluid
(MSE). underlies the greatest amount of the less dense
Potential energy (PE) is usually defined as fluid (Figure 1B). The center of mass is now at its
DSE. A tiny fraction of the PE, called the avail- lowest, as is the gravitational PE. Ignoring friction,
able potential energy, APE, can be used to drive mixing, and turbulent effects, KE is maximized at
the KE. The concept of APE is used to explain the this point. As time proceeds further, the fluids over-
links between PE and KE and is usually defined shoot this state, and KE begins converting back to PE
as the difference between the PE and the minimum (Figure 1C).
PE that could be achieved by an adiabatic arrangement APE is defined as the difference between the current
of mass. Sometimes latent heating is included directly PE and the minimum PE. The state with lowest PE is
in the APE, but usually it is treated as a separate the ‘reference state’, which has zero APE. The refer-
diabatic process. PE is useful for global energy ence state definition is somewhat arbitrary. Another
balance. mechanism could possibly occur at some later time to
Solar radiant energy does not reach the Earth lower further the minimum PE, for example a net
equally everywhere. On average, the tropics temperature decrease. However, the size of the con-
receive and absorb far more solar energy annually versions, generation, and destruction are not arbitrary.
than the polar regions. This distribution of absorbed APE is intended to represent the PE available for
energy creates an uneven distribution of temperature. driving motions, so the reference state is usually
Temperature, pressure, and density are related, so the defined by rearranging atmospheric properties in the
PE has an uneven distribution too. The existence of horizontal dimensions so as to reach a state of
APE is essentially due to the horizontal variations in minimum PE. This model reveals that density differ-
density and temperature. APE leads to motions (KE)as ences across a fixed elevation in the tank are propor-
the atmosphere tries to remove these density and tional to the APE.
temperature variations. The motions redistribute The model relates to the atmosphere as follows.
some mass, but mainly the atmosphere transports
heat. The atmospheric circulation becomes a complex 1. Temperature differences create the density differ-
balance between the radiant energy input and ences. The less dense fluid represents the tropics;
output that creates the APE needed to generate the the denser fluid represents polar regions.
KE of circulations, which in turn strive to create a state
2. The reference state has minimum center of mass
of no APE. when the air ‘layers’ are flat. Flat fields of pressure
Available potential energy and KE are defined in and temperature imply no geostrophic or ageo-
formal mathematical ways. The mathematics
strophic winds.
shows interactions from which physical mechanisms
(such as baroclinic instability) can be identified.
The energy equations describe the following chain
of events. Radiation creates APE; some APE is
converted to motions that redistribute the heat
energy; KE in turn is lost by conversion back to APE
and by friction. The forms of energy and the ‘net’
conversions between them can be represented via a (A) (6) (C)
‘box’ diagram. However, the box diagram does
not show the energy cycle in an intuitive sense. Figure 1 Schematic model of fluid motion showing APE and KE
concepts. The tank holds two immisciblefluids with densityp, i p p .
To make the physical mechanisms clear, energy (A) Initial state; (6)state with maximum KE but minimum PE
must be examined regionally and one phenomenon reached during the first oscillation; (C) state where KE is being
at a time. converted back to PE.
3. Density differences (sloping air ‘layers’) have APE The rate of energy release per unit horizontal area, r,
but also produce horizontal pressure gradients by all the air in motion, can be compared to the rate per
that accelerate the air. On the rotating Earth, unit area of energy absorbed from the Sun:
geostrophic winds, and thus KE, are also present.
So, reservoirs, sources, and sinks of APE are not
independent of KE.
where M is the mass in motion, t is the time to complete
Carnot Cycle the circuit, and a is the area of the Hadley cell. For
the schematic circulation indicated in Figure 2A,
The Carnot cycle can be used to estimate KE gener- M zz 1 O I 8 kg, a zz 1 . 5 ~ 1 0 ~ ~and
m ~t, 3 ~ 1 0 ~ s .
N
ation from thermodynamic changes that an air parcel The total rate of energy released by the Hadley cell is
undergoes while completing an atmospheric circuit. MEt-l FZ 5 . 3 1014
~ Js-’. However, the rate per unit
The Hadley cell is a conceptual model for the zonal area is only r N 3.5 Wm-2. The absorbed solar radi-
mean tropical circulation. Air in the lower tropo- ation in the tropics is 100 times larger than r, making
sphere moves equatorward while gaining heat and the atmosphere an ‘inefficient’ heat engine. (Efficiency
moisture from surface fluxes. Near the Equator rapid of the Carnot cycle is often measured in a way
ascent within thunderstorms releases and advects dependent on the temperature, but our estimate is
much latent heat energy. Reaching the upper tropo- related to energy input.)
sphere, air moves poleward, cools radiatively, and The model illustrates these properties:
sinks, completing a circuit.
Kinetic energy generation can be estimated by 1. Warmer air is rising and cooler air is sinking so the
plotting the thermodynamic properties of air parcels center of mass is lowered and KE is created;
on a skew T-log P chart. Aunit area anywhere on the the circuit is counterclockwise and the circula-
chart corresponds to a specific amount of energy tion is ‘thermally direct’. In contrast, the Ferrel
exchange. Figure 2 shows a realistic circuit around an cell is a clockwise circuit that reduces KE to
annual mean Hadley cell. The amount of APE increase PE.
converted to KE by a kilogram of air while it completes 2. A steady state is reached if the frictional losses
the plotted circuit is E 1 . 4 ~ 1 Jkg-l.
N 0~ balance the KE generation.
200
300
e! 400
3
I 3
e!
a 500
-
c
600
700
800
900
1000 I I 1000 I J / I
30 0
(4 Latitude (B) T, 0 -+
Figure2 Interpretationof the Hadley circulation as a Carnot cycle. (A) Meridional cross-section showing the idealized circulation. The
dashed line shows an average path followed by the parcels with numeric labels for each leg. (6)Skew T-In P plot of the thermodynamic
changes along each of the four legs drawn in part (A). The shaded area is proportional to the energy converted from PE to KE.
832 GENERAL CIRCULATION I Energy Cycle
JJA
(6) Latitude
Figure 3 Zonal mean efficiency factor [ E ] for (A) December-February and (B) June-August. [E] is estimated from zonal mean 1979-99
National Center for Environmental Prediction/National Centers for Atmospheric Research (NCEP/NCAR) reanalysis data from the
Climate Diagnostics Center (CDC) in Boulder, CO, USA. The contour interval is 0.03.
3 . The rate of KE generation depends on the area the meridional temperature gradient is stronger
enclosed by the circuit divided by the time to than summer, and so is the Hadley cell.
complete the circuit. 5. In winter, the air motion of the Hadley cell is 5-7
4. The amount of energy converted is proportional to times stronger than that of the Ferrel cell but larger
a circuit integral of T , so it increases as the temperature differences occur along the Ferrel
temperature difference increases between the circuit. So, the net energy conversions are similar
warm and cold stages of the cycle. During winter (see 'The Box Diagram' below).
6. Large energies are involved, but only a tiny fraction Kinetic Energy
of the solar radiation actually drives the observed
Kinetic energy is primarily contained in horizontal winds:
motions.
7. The path followed by air parcels was specified, not
predicted. KE =
s (u2
Kinetic energy has the following properties:

+ u2)dM
Available Potential Energy

1. The distribution of zonal mean KE (Figure 4)has
Available Potential energy is approximated by tern- maxima at upper levels near the subtropical jets.
perature variations on a pressure surface: 2. Kinetic energy is related to atmospheric momen-
tum and torque. Momentum fluxes by the Hadley
APE= s EC,TdM
cells and by midlatitude eddies maintain the KE
maximum near the subtropical jet. However,
“3 ‘cc,pi; p-l{0 - (3
e } ae
- -l
dM [3]
slowing down easterlies increases westerly mo-
mentum, but reduces KE.
E =1 - { ;}K 141
Energy Generation and Conversion
Energy Equations
To understand how energy evolves one needs formulae
where M is mass, 0 is potential temperature, for APE and KE tendencies in a limited domain. The
Po0 = 105Pa, C, is the specific heat capacity at domain may be a unit area in the meridional plane
constant pressure, IC = R C;l, R is the ideal gas (useful for calculating zonal means) or enclosing a
constant, and E is the ‘efficiency factor’. Pr(0) is the single phenomenon to the exclusion of others (e.g., a
reference pressure, which is the average pressure on a single frontal cyclone), Tendency equations for APE
potential temperature surface 8. Available potential ( A ) and KE (K) in a mass M between two isobaric
energy is zero when P = P, everywhere in the domain.
--
surfaces are:
The overbar denotes the horizontal average on an
isobaric surface. aA
-=
at
( E q ) dM + 1 (EOCL) dM
1. For an integral over the depth of the atmosphere, Term: a b

APE differs from PE by the factor E.
2. Observed PE is about a thousand times greater - /EVp.(VpCpT)dM
than estimates of global average APE. .C.
3. Hemispheric PE is greater in summer since the air is
generally warmer than in winter.
4. Hemispheric APE is greater in winter when the
+ / -E) (C,T dM
meridional temperature gradient is stronger, mak- d

ing the term in curly bracket { ) larger than in
summer. The further the atmosphere departs from
the reference state mean, the larger E becomes. at
=- 1 (V, F) dM - 1
(V, V,@) dM
v
5. Diabatic heating or cooling can create APE if it Term: a b
magnifies the departures but the same heating or
cooling can destroy APE if it reduces the depar- - V, (V,K) dM
0
tures. In simplistic terms, APE is generated v

C
by ‘heating where it is hot or cooling where it is
cold’. where q contains all diabatic heating and F is friction.
6. E > 0 in ‘hot’ regions and E < O in ‘cold’ regions. Also, o is pressure coordinate ‘vertical’ velocity; x is a
From Figure 3, E has a positive maximum in the specific volume; CD is geopotential; and V, and Vp denote
tropical middle troposphere and negative minima velocity and gradient evaluated in isobaric coordinates.
in high latitudes. In middle latitudes the sign varies The terms are ordered to match similar processes.
with longitude: E > 0 over oceans during winter or Term ‘a’ in each equation has diabatic sourceknk
warm sectors of frontal cyclones while E < 0 over mechanisms. Term ‘b’ is similar but has opposite sign
continents in winter or behind cold fronts. in the A and K equations and so represents a
Next Page
834 GENERAL CIRCULATION / Energy Cycle
DJF
~
I
80"s 40"S 0" 40" N 80" N
Latitude
JJA
80"s 40"S 0" 40" N 80" N
(B) Latitude
Figure 4 Zonal mean kinetic energy density for (A) December-February and (B) June-August using 1979-99 NCEP/NCAR reanalysis
'.
data from CDC. The contour interval is 50 kg s -'m -
(baroclinic) conversion between these two forms of heat flux, and friction. Figure 5 illustrates how these
energy. Term 'c' is divergence of potential or kinetic are distributed on a zonal mean.
energy flux; it is a conversion between the APE or
1. Solar radiation absorbed. Much more radiation is
KE inside and external to the domain; baroclinic
absorbed ( q > 0) in the tropics ( E > 0) than in polar
or barotropic conversions respectively, appear in
regions ( E < 0), so APE is generated, particularly in
this term.
the winter hemisphere.
2. Terrestrial radiation emitted. The emission ( q< 0)
Diabatic Sources and Sinks of Energy
is greater in the tropics, suggesting destruction of
There are five categories of diabatic processes: solar APE. But APE is generated because the emission in
and terrestrial radiation, latent and sensible surface high latitudes is from cloud tops where E is strongly
GLACIERS 869
Waves. Numerical Models: Chemistry Models; Methods. Symposium in Honor of Professor Akio Arakawa. New
Parameterization of Physical Processes: Clouds; York: Academic Press.
Gravity Wave Fluxes. Quasi-geostrophic Theory. Jacobson MZ (1999)Fundamentals of Atmospheric Model-
Weather Prediction: Regional Prediction Models. ing. Cambridge: Cambridge University Press.
Schlesinger ME (1988) Physically-Based Modelling and
Simu?ation of Climate and Climatic Change. Dordrecht:
Further Reading Kluwer Academic.
Randall DA (ed.)(2000) General Circulation Model Devel- Trenberth KE (ed.) (1993) Climate System Modeling.
opment: Past, Present and Future, Proceedings of a Cambridge: Cambridge University Press.
A C Fowler, University of Oxford, Oxford, UK of which terminated some 10000 years ago) is
Copyright 2003 Elsevier Science Ltd. All Rights Reserved demonstrated by the presence of these and other
such signatures as erratics (isolated boulders trans-
ported by the ice), and glacial striae and scratches in
bedrock, which can even be found in Central Park in
Introduction New York.
Glaciers are rivers of ice, and ice sheets are continental- In sufficiently cold conditions, a network of glaciers
scale domes of ice. Both glaciers and ice sheets flow will coalesce and grow to form an ice cap, or at the
slowly, transporting snow/ice to lower elevations, largest scale, an ice sheet. There are two present-day
where it melts or is released into the sea as icebergs. ice sheets on the Earth, Greenland and Antarctica;
The size and extent of glaciers varies with climate, and whose dimensions are measured in thousands of
they represent the most slowly responding constituent kilometers horizontally and thousands of meters
of the global climate system. Despite this, glaciers and depth. During the ice ages, however, other ice sheets
ice sheets are capable of surprising and dramatic grew to cover much of North America, Scandinavia
effects such as glacier surges, and are thought to have and Britain, and their former presence is betrayed by
been an important causative component of the rapid such relict features as moraines (e.g., Cape Cod),
climate shifts seen in the last ice age. drumlin fields (eg., across the north of Ireland), and
Glaciers are found in regions of high elevation, such glacially excavated fjords and lochs.
as in the Alps or the Himalayas, and in polar regions, Most glacial ice is found in these large ice sheets; the
such as Alaska or Svalbard (Figure 1).Typical depths remainder exists in glaciers and ice caps which occur
are on the order of hundreds of meters, and typical in mountainous regions, largely round the Arctic
lengths are measured in kilometers: the Bering Glacier Ocean basin in the Northern Hemisphere. Extensive
in Alaska is one of the longest, at 200 km. Glaciers systems of glaciers exist in the Andes, Alaska, Norway,
form when snow accumulates to great depth, and is the European Alps, and the Himalayas. Glaciers are
transformed through the effects of pressure to form also found in warmer regions at sufficently high
ice, which then creeps slowly down slope, at rates altitude, for example, in equatorial parts of Indonesia
which are typically measured in tens to hundreds of and Africa; such glaciers are known as tropical
meters per year. The fastest moving glacier on Earth is glaciers.
the outlet glacier Jakobshavn in Greenland, which Glacial extent is thus an indicator of climate (with a
moves steadily at a brisk 8 km y -'. 10 to 100 year response time), and the recession of
Analogous to rivers, glaciers act both as an agent of many glaciers during the twentieth century marked a
water transport and as an agent of erosion. Glaciers, noticeable global warming which terminated a centu-
along with landslides, are the primary agents of ries-long cold period, known as the Little Ice
erosion in high mountain ranges, and they are also Age, which lasted from about AD 1500 to 1900. The
instrumental in forming various landscape features change in climate in Europe over the last hundred
such as U-shaped valleys, terminal moraines, drum- years is evidenced by the shrinking glaciers in the
lins, and eskers. Glacial climate is also associated with Alps, and more poetically by the plentifulness of snow
the formation of permafrost-related features, for in Impressionist paintings of the late nineteenth
example fossil ice-wedge polygons. The former pres- century.
ence of glaciers and ice sheets in North America and Glaciers respond to climate in a similar way to
Northern Europe during a sequence of ice ages (thelast rivers, by the passage of kinematic waves (somewhat
870 GLACIERS
Figure 1 Averticalviewof BakaninbreenGlacierin Svalbard, 1990. Flow isfrom bottom righttowardsthetopleft. Thedarklinesinthe ice
are medial moraines. Bakaninbreen is a surging glacier, and the advancing surge front is clearly visible in the center of the picture (Photo
courtesy Tavi Murray.)
like flood waves on rivers) down the glacier, with a Glaciers exhibit a variety of other wavelike motions,
characteristic speed of transmission of several times which appear to be internally generated and unrelated
the surface speed. Ice sheets also respond to climatic to climate. Wave ogives and ‘Schnellungswellen’, or
change, but on a much longer time scale, of the order of waves of velocity, are seasonal effects. More dramatic
tens of millennia. It seems likely that the time scale of is the glacier surge (Figure 2), a rapid advance of a
the regular occurrence of ice ages over the last several glacier that occurs for a short time and is repeated at
million years, is associated with this response time of regular intervals. A well-documented example of a
the large ice sheets and their effect on climate through surging glacier is the Variegated Glacier in Alaska,
the effect of ice-albedo feedback. which surges rapidly for one to two years, repeating
Figure 2 A closer view of the surging Bakaninbreen Glacier, 1987. The ice attains depths of 300 m, and is 1-2 km wide. The ridge in the
picture is the surge front, a wall of ice some 50m high, propagating down glacier at about a kilometer per year. (Photo courtesy Tavi
Murray.)
GLACIERS 871
Figure 3 Aerial view of the Stancomb Wills ice stream flowing into the Brunt ice shelf, Antarctica. The highly crevassed nature of the
surface is typical of a fast-flowing ice stream. (Photo courtesy British Antarctic Survey.)
this behaviour at intervals of about 20 years. Velocities as the crevasses which are commonly seen at the
during surges can increase by a factor of a hundred, glacier surface. Over a longer time scale, however, ice
and advances of 10 km or more in less than a year have will deform like a viscous fluid due to the stress-
been recorded. induced migration of dislocations within the crystal-
Modern-day ice sheets do not appear to exhibit such line lattice structure. Such dislocation creep can be
collective surge-type behavior, but they do exhibit a measured experimentally, and the effective viscosity is
spatial equivalent to the temporal periodicity of found to depend nonlinearly both on the applied stress
surging valley glaciers, in the existence of concentrated and the temperature: higher stresses and temperature
‘ice streams’ (Figure 3). A well-known example occurs both act to make the ice less ‘sticky’. At a typical
on the Siple Coast in Antarctica, where the ice which glacier stress of one bar (0.1MPa) and a temperature
drains into the floating Ross ice shelf is segregated into near the melting point (273 K), the viscosity is about
five separate ice streams, four of which move much l O I 3 Pas, or about 4 bar year. The shear modulus for
more rapidly than the bounding, relatively quiescent elastic deformation of ice is about 3.5 x l o 9 Pa, and
ice. The Whillans ice stream, for example, moves at the ratio of these, some 3000 s, or 50 min, defines the
speeds of 5 0 0 m y - 1 , as opposed to neighboring ice Maxwell time which separates short term elastic
speeds of less than 1 0 m y - l . The ice streams are behavior from long term viscous behavior. For time
recognizable from the air by their intensely crevassed scales greater than a few hours, ice behaves viscously,
surface, a feature they share with surging glaciers at least at high temperatures. For colder ice at lower
during the active surge phase. In the Siple Coast, they stress, the Maxwell time may be of the order of months
have typical widths of 40 km, and lengths of several or longer.
hundreds of kilometers. There is also evidence of ice Mathematical models of the flow of glaciers and ice
streams in former Northern Hemisphere ice sheets, for sheets take advantage of the fact that they have a high
example the Laurentide ice sheet in North America was aspect ratio, like the atmosphere, so that approxima-
drained by a number of ice streams, amongst them one, tely a form of lubrication, or thin film is appropriate. In
some 200 km wide, which flowed out into the Labrador the simplest version of such models, which assume a
Sea down the Hudson Strait. temperature and moisture-independent flow law, one
derives a convective diffusion equation for the glacier
depth, or a nonlinear diffusion equation for the ice
Physics of Glaciers sheet depth. Unfortunately, such models are unrealis-
Ice is a crystalline solid, and behaves over short time tic because of the strong effect of temperature on the
scales as an elastic medium when subjected to differ- flow law (for glaciers at sub-freezing temperature), or
ential stresses. In particular, ice fractures in because of the effect of moisture on the flow law (for
tension, and these fractures are manifested on glaciers temperate glaciers, at the melting point).
872 GLACIERS
Ice is an insulator, and because of this and the regions. At the other extreme, a temperate glacier is at
warming effects of geothermally derived heat, as well the melting point throughout, and contains a small
as heat generated by stress working in the ice, the amount of liquid water in inclusions. Most glaciers in
temperature, as well as the stress, increases with depth, the Alps are temperate, and contain typically 2 % by
particularly near the base where the effects of advect- volume of water.
ion are generally smaller. At a temperature of -2O”C, The mechanism by which a glacier can be temperate
the viscosity is 40 times greater than at the melting is ascribed to the seasonal variation in temperature. In
point, and at - SO’C, as is appropriate to parts of the temperate climates, snowfall on a glacier during
the Antarctic ice sheet surface, it is 2000 times greater. winter is melted in the summer, and the resulting
A reduction of stress by a factor of 10 near the surface meltwater percolates through the porous upper snow
would cause a further increase by a factor of 100. Thus (or firn, Le., wet snow) where it refreezes, and the
ice has a strongly variable viscosity, and its motion is resulting release of latent heat enables the temperature
more akin to that of a fairly rigid layer being carried to be maintained at the melting point. Various further
along on top of a softer shearing underbelly. Because of classifications can be made in order to allow for the
this, one needs to solve the energy equation for the common situation where a glacier in a polar environ-
temperature also. ment has a surficial cold layer, but the basal ice is
A novelty in ice physics is that the frictional heating warmed by geothermal heat to the melting point
due to viscous dissipation is important. Because the (Figure 4); sometimes a basal temperate layer
heating depends nonlinearly on temperature, the may form, where the ice is internally heated to melt-
temperature and flow equations are coupled via a ing point and may contain water inclusions. Glaciers
positive feedback (faster flow means more heat, hence of this and similar type are often called ‘sub-
higher temperatures, hence reduced viscosity, hence polar’; another common term in current usage is
higher velocity), and the possibility arises of thermal ‘polythermal’ .
runaway occurring (the same phenomenon is the cause
of explosions, and also of the combustion that occurs
in lighting a match). It has been suggested that such an
instability lies at the root of surging glacier behavior,
Basal Sliding
and also of ice streaming, but it is unlikely that this can The effect of basal water on the motion of ice is that it
be the whole story. allows basal slip, or basal sliding, to occur. The
The reason for this is that the rise in temperature mechanism whereby this is thought to occur is a subtle
near the base of a glacier has another important effect. one. When the basal ice is at the melting point, a thin
If the temperature reaches the melting point (and this is (micrometer thick) water layer exists between the ice
often the case), then basal melting starts to occur, and and the underlying bedrock, due to regelation: the ice
water will exist at the glacier bed. In this situation approaching the upstream (stoss)face of a protruding
basal ice motion, or sliding, occurs, and this is thought bump is at higher pressure, and thus melts, because the
to be the cause of much of the fast flow exhibited by ice melting temperature decreases slightly with increasing
streams and surging glaciers. The basal water forms its pressure (the Clapeyron effect); this meltwater forms
own subglacial drainage system, and flows towards the thin film, which then squirts round the bump under
the glacier outlet. It is a common experience to see one the driving pressure gradient, to refreeze downstream
or more outlet streams emerging from the front of a where the pressure is lower. The latent heat necessary
glacier, often from a large portal. In addition, surface is provided by conductive heat transfer through the
rainwater or meltwater often finds its way to the bed rock.
via crevasses or moulins. It is common to see streams Regelation itself allows a mechanism for flow of ice,
on a glacier surface which suddenly disappear down a but also the film lubricates the ice-bed interface, so
hole, presumably to connect to the basal water system. that the ice can simply flow viscously over the bumpy
bed. The bed does offer a resistance because of its
roughness, and the resulting basal shear stress is
related to the sliding velocity by the sliding law.
Thermal Classification Various theories have been proposed for this ‘law’; in
The presence and amount of water in a glacier is general, as one would expect, the stress increases with
associated with a thermal classification of glaciers. the velocity. Sliding at the bed has been observed, and
The basic types suggested by Ahlmann in 1935 are the indeed it is the dominant cause of motion in some
polar and temperate glaciers. As its name suggests, the glaciers.
polar glacier is one which is below the melting Another observed feature of the process of sliding is
temperature throughout, and occurs (if at all) in polar the formation of cavities. As with ordinary hydraulic
GLACIERS 873
Figure 4 Trapridge Glacier, a surge-type sub-polar glacier in the Yukon Territory, Canada. The glacier is 4 km long and the advancing
front is about 70m high. (Photo courtesy Garry Clarke.)
cavities, these will form if the water film pressure several different topologies that have been suggested.
becomes less than the local subglacial drainage pres- The principal one is embodied in the concept of the
sure (which is determined independently). Cavities Rothlisberger channel, which is a cylindrical drainage
will occur if the sliding velocity is high enough, and channel at the bed, cut upwards into the ice. The water
they reduce the resistance of the bed; thus the sliding flows through the channel at a lower pressure than the
law should also depend on the degree of cavitation, overburden ice pressure (the difference between the
and this will depend on the local drainage pressure. In two is called the effective pressure), and consequently
this way the flow of the ice becomes coupled through the channel (like a void in a fluid) tends to close
the basal sliding law to the subglacial hydraulic because of inward creep of the ice. It is maintained
system. open because the water flow through the channel
Although the discussion above refers to a clean generates sufficient frictional heat to melt the channel
interface between ice and bedrock, it is usually the case walls back. A theory to describe this dynamic interac-
that a certain amount of erosional debris is situated at tion leads to a prediction for the effective pressure as a
the bed, often consisting of a mixture of coarse, weakly increasing function of the water flux through
angular rock fragments within a finer grained matrix the channel, and observed borehole water pressures of
of sandy or clayey material. The resultant material is tens of bars below flotation levels can easily be
called till, and when water saturated, it will deform. A explained in this way. The fact that effective pressure
different kind of basal motion can then occur, wherein increases with water flux implies some kind of arterial
ice slides over underlying bedrock via the lubricating drainage system, since a larger channel with larger
effect of the deforming till. This also will lead to an flow rate has a lower pressure than a smaller channel,
effective sliding law, in which basal shear stress and hence will suck water from it; thus it is unlikely
depends on basal velocity and, again on basal water that Rothlisberger channels are uniformly distributed
pressure, since the flow resistance of till also across the bed.
depends on this. Meters thick layers of till underlie On the other hand, there is evidence from borehole
many glaciers, as well as the Siple Coast ice streams, and discharge events that in some circumstances (and
it is thought that it is largely the motion of ice over such particularly during a glacier surge) the drainage
till layers at high water pressure which causes the fast system may indeed become distributed in some kind
flow seen in ice streams, and in some surging glaciers. of anastomosing pattern. One such system occurs if
the channel system closes down, and the water
migrates to the cavities. This forms a ‘linked-cavity’
Subglacial Hydrology system, and was inferred to occur during the 1982-83
In many cases, basal water storage has a profound surge of Variegated Glacier. Another type of distrib-
effect on the dynamics of glaciers and ice sheets, and it uted system that has been suggested is a system of
is important to understand and quantify the way in anastomosing ‘canals’, or a system of connected
which subglacial drainage systems work. There are ‘puddles’. Such distributed systems can support
874 GLACIERS
Figure 5 Aerial view of icebergs forming at the edge of the ice shelf near Halley, Antarctica. (Photo courtesy British Antarctic Survey.)
much higher water pressure than a Rothlisberger ation with outgoing heat loss by long-wave emission,
system, and are consistent in this respect with borehole and sensible and latent heat fluxes. Essentially, the ice
measurements of water pressures near flotation under will take the mean annual air temperature, at least at
ice streams. depths greater than about 1 0 m where the surface
thermal wave does not penetrate. While this statement
would be exact if heat conduction were the only
Mass and Energy Balance transfer mechanism, it fails when the air tem-
perature becomes greater than the melting point.
Glaciers interact with the atmosphere and the oceans
This is because the ice temperature cannot then
through processes of mass and energy exchange.
follow the air temperature, and the resulting
Precipitation occurs as snow in winter and accumu-
melting and refreezing of melt water at depth causes
lates (in the accumulation zone, upstream). Above a
a much more rapid elimination of the winter cold wave
certain elevation (the firn line, or equilibrium line),this
than conduction alone would provide. For these
snow will remain from year to year, and successive
reasons, ice temperature at the glacier surface is
snowfalls lead to an increasing thickening and thus
only approximately equal to the local mean air
compaction of the snow under its own weight, as the
temperature.
air is expelled. In addition, summer melting (where it
Furthermore, surface ablation, while also related to
occurs) allows percolation of meltwater downwards,
surface air temperature, is not a simple function of it,
where it refreezes, and both processes lead to the
and is certainly not dependent on mean annual air
formation of ice, when only isolated air pockets
temperature. It is more directly connected to the mean
remain. As the ice flows downhill, it descends past
annual value of the 'positive air temperature', i.e., the
the firn line to the ablation zone, where net summer
air temperature taken only when it is above the melting
melting outweighs any net snow accumulation, and
point. For a glacier or ice sheet, an equilibrium is
consequently wastage of the glacier occurs. The
obtained when net accumulation balances net abla-
resulting melt runs off the glacier, through surface
tion, thus glaciers act as a climatic indicator of winter
streams or via the basal drainage system, or is
precipitation and summer insolation. The latter ex-
evaporated at the surface. Depending on climatic
plains why the summer insolation curves are used in
conditions, a glacier may flow all the way to the sea,
Milankovitch's ice age theory.
when it is known as a tidewater glacier. Such glaciers
lose mass also by the calving of icebergs (Figure 5 ) .The
net gain of ice (measured as water equivalent) over a
year is known as the mass balance.
CIimatology
Energy interchange at the glacier surface determines Glaciers (and ice sheets) monitor climate variations,
surface temperature (as well as surface melting) in the but climate also responds to changes in glaciation. A
same way as elsewhere on the Earth's surface, via the simple example is that of the thermally induced
balance of incoming short-wave and long-wave radi- katabatic wind. Obviously, the growth of large ice
GLACIERS 875
sheets can potentially exert a thermal and topographic In some way, it would seem that the massive
effect on the general circulation, and thus affect discharge of ice into the North Atlantic can switch
precipitation patterns through the diversion of storm on the global oceanic thermohaline circulation, and
tracks. The other, and more fundamental, effect is lead to sudden dramatic warming of the atmosphere.
through the alteration of land surface albedo. Ice and, The way in which such occasional massive discharges
particularly, fresh snow have a high reflectivity, so that can occur is through periodic surges of the Laurentide
net received surface energy decreases with increasing Ice Sheet through the Hudson Strait ice stream, and
ice cover. This leads to a positive feedback, called the the existence of surging glaciers and ice streams
ice-albedo feedback, which can be used to explain the shows that such behavior is possible. Alternatively
occurrence of ice ages in simple ('zero-dimensional') (or as well), tidewater glaciers are known to undergo
models of the climate. In its most basic form, such a similar cycles of slow advance and rapid (via iceberg
model allows two possible stable, steady states; an ice- calving) retreat, and these might be associated with
free Earth (low reflectivity, high received insolation, the Dansgaard-Oeschger events. In any event, we
high global surface temperature), and an ice-covered see that the dynamics of glaciers and ice sheets are
Earth (high reflectivity, low received insolation, low likely to have been fundamental in the past in driving
global surface temperature). Variations of solar ins- the ice age climate, at all time scales from the
olation due to Milankovitch orbital variations then Milankovitch tuned 100 000-year ice age cycle,
allow a shuttling backwards and forwards between the through the 10 000-year Heinrich events and millen-
two. Although an ice-covered Earth is not attained in nial Dansgaard-Oeschger events, to the sudden dec-
the current sequence of ice ages, there is evidence that adal warmings that terminate them; we are only
such a state was reached during a glacial epoch some beginning to recognize and understand such behavior.
600 million years ago; this is the so-called 'snowball The implication for the study and prognosis of our
Earth' theory. present climate is clear.
Ice sheets are also fundamentally involved in shorter
(millennial) time scale climatic switches. Oxygen
isotope records from deep ice cores in the Greenland See also
ice sheet show repeated switching during the last ice Climate Variability: North Atlantic and Arctic Oscillation.
age between cold and relatively warm conditions. Energy Balance Model, Surface. Hydrology: Ground
These oscillations, with an amplitude of at least S'C, and Surface Water; Modeling and Prediction; Overview.
are known as Dansgaard-Oeschger events, and they Ice Ages (Milankovitch Theory). Katabatic Winds.
take the form of gradual (millennial)cooling, followed Permafrost. Sea Ice.
by abrupt (decadal) warming. It has been suggested
that these temperature cycles reflect alterations in the
oceanic circulation, which is often thought of as a kind Further Reading
of conveyor belt, with downwelling occurring in the Alley RB (2000) The Two-Mile Time Machine. Princeton:
North Atlantic, and upwelling in the Indian and Princeton University Press.
Pacific oceans, and around Antarctica. If the conveyor Benn DI and Evans DJA (1998) Glaciers and Glaciation.
is switched off, then climate becomes cooler, and if it is London: Arnold.
abruptly switched on again, climate can become Colbeck SC (ed.) (1980) Dynamics of Snow and Ice Masses.
abruptly warmer. New York: Academic Press.
Dansgaard-Oeschger events are bunched into long- Hambrey M and Alean J (1992) Glaciers. Cambridge:
er time scale cooling cycles lasting about 10 000 years, Cambridge University Press.
and these are terminated by Heinrich events, which are Knight PG (1999) Glaciers. Cheltenham: Stanley Thornes.
followed by dramatic climatic warming. Heinrich Menzies J (ed.) (1995) Modern Glacial Environments.
Oxford: Butterworth-Heinemann.
events refer to the occurrence in North Atlantic
Paterson WSB (1994) T h e Physics of Glaciers, 3rd edn.
oceanic sediments of layers of ice-rafted debris Oxford: Elsevier Science.
(IRD), and are thought to be due to a massive Post A and LaChapelle ER (2000) Glacier Ice, revised edn.
discharge of icebergs from the Laurentide ice sheet, Seattle: University of Washington Press.
and particularly from the ice stream in Hudson Sharp RP (1988) Living Ice. Cambridge: Cambridge Uni-
Strait, which drained the Hudson Bay ice dome (since versity Press.
the lithic fragments of the IRD largely come from Van der Veen CJ (1999)Fundamentals of Glacier Dynamics.
there). Rotterdam: Balkema.
876 GLOBAL CHANGE / Biospheric Impacts and Feedbacks
Contents
Biospheric Impacts and Feedbacks

Human Impact of Climate Change
Ozone Trends
Surface Temperature Trends
Upper Atmospheric Change
and consume many trace gases, often at rates high

BCot~pherSeImpqMs and enough to affect atmospheric concentrations. Climate
is a key determinant of biogeochemical transforma-
Feedbacks tions, and thereby directly affects rates of trace gas
exchange. Because species differ in their physiological
B A Hungate and G W Koch, Northern Arizona capacities to produce and consume trace gases, the
University, Flagstaff, AZ, USA suite of organisms comprising a biological community
Copyright 2003 Elsevier Science Ltd. All Rights Reserved and the number of organisms present in an ecosystem
both strongly influence biogeochemical cycling and
associated trace gas fluxes. In turn, biogeochemical
Introduction transformations include processes that regulate nutri-
The terrestrial biosphere interacts with the atmos- ent availability to plants, thereby influencing species
phere through exchanges of trace gases and of energy. composition and ecosystem structure. Energy ex-
These interactions are bidirectional, with atmospheric changes between ecosystems and the atmosphere
conditions modifying ecological controls of trace gas include fluxes of sensible heat and latent heat (i.e.,
and energy exchange, and the very exchanges feeding evaporation of water), exchanges that are directly
back to affect atmospheric chemistry and climate mediated by the biophysical properties of ecosystems.
(Figure 1). Here, we review the role of biosphere- Like biogeochemical transformations, biophysically
atmosphere material and energy exchange in atmos- mediated energy exchange is sensitive to climate and
pheric processes, the biological processes involved, to the physiological properties of organisms, mostly
and how biotic responses to ongoing and future global plants. Thus, changes in climate and in the structure of
changes could influence the future chemical and terrestrial ecosystems alter patterns of biosphere-
radiative balance of the atmosphere. Global environ- atmosphere energy exchange.
mental change is already altering biological produc-
tion and consumption of trace gases and energy
partitioning between the biosphere and the atmos- Trace Gases
phere. However, the magnitude and direction of these
Carbon Dioxide
effects are not well constrained, and this uncertainty
contributes to the difficulty in predicting future global The concentration of carbon dioxide (C02) in the
climate. Global changes likely to substantially alter atmosphere is increasing by 0.5% per year (1980-
biosphere-atmosphere trace gas and energy exchange 2000), a rate unprecedented for the last 160 000 years.
include climate change, rising atmospheric carbon A greenhouse gas, C 0 2 is estimated to be responsible
dioxide, acid rain and associated nitrogen deposition, for 60% of current global warming and that predicted
increased W - B radiation due to stratospheric ozone to occur over the next several hundred years. While the
depletion, species invasions, losses of biodiversity, and current increase in CO2 concentrations is clearly
land-use and land-cover change (Figure 1). driven by human activities (fossil fuel burning, defor-
Trace gases in the atmosphere are arbitrarily defined estation, and cement manufacture), two biological
as gas species with concentrations equal to or less than processes strongly influence the concentration of C 0 2
that of carbon dioxide (367ppmv as of this writing). in the atmosphere. Photosynthesis by plants and algae
Through biogeochemical transformations in terrestri- converts atmospheric C 0 2 to organic carbon, and this
al ecosystems, plants and soil microorganisms produce process is the primary entry point of carbon into the
GLOBAL CHANGE / Biospheric Impacts and Feedbacks 877
Climate
change \ Physical climate,

Atmospheric chemistry
3"
i Biophysics
water relations
+
r K
Biogeochemistry
Species composition,
ecosystem structure
Land-use
and land-cover
f r I
Biodiversity Invasive
change
loss species
Figure 1 Interactions between the terrestrial biosphere and atmospheric composition and climate (plain type, thin arrows), and impacts
of global environmental change on thecoupled biosphere-atmosphere system (boldifalics,thickarrows). (Modifiedfrom Sellers PJ etal.
(1991) Charting the Boreal forest's Role in Global Change, EOS, Transactions, American Geophysical Union, Vol. 72, No. 4, 33-40.)
Climate influences biophysical and biogeochemical processes, creating feedback effects to both atmospheric composition and climate
through fluxes of heat, water, and trace gases. Global changes such as N deposition, acid rain, enhanced UV-B radiation, and rising COP
directly affect ecosystem biogeochemistry,feeding back to the atmosphere through altered trace gas fluxes, altering ecosystem structure
by favoring somespeciesoverothers, and affecting biophysically mediatedexchangesof energy by altering physiological processes (e.g.,
stomatal conductance). Other global changes affect ecosystem structure directly, with, for example, invading species introducing new
biological entities with unique contributions to biogeochemical and biophysical properties.
biosphere. Respiration by plants, animals, and micro- reflected in lower atmospheric concentrations of C 0 2
organisms oxidizes organic carbon, returning it to the during the spring and summer. This pattern is most
atmosphere as C02. Rates of C 0 2 exchange between pronounced above the temperate zone in the Northern
the terrestrial biosphere and the atmosphere amount Hemisphere (Figure 2), less so in the Southern Hemi-
to roughly 120 Gt C y - ', compared to only 7.5 Gt y - sphere where land masses are smaller, and essentially
added through the combined anthropogenic effects of absent in tropical regions, where the activities of
fossil fuel burning, land-use change, and cement photosynthesis and respiration tend to be synchronous
manufacture. Thus, predicting the future trajectory throughout the year. Oceanic exchanges of C02, while
of atmospheric C 0 2 concentrations requires an un- large, do not show strong seasonality. These seasonal
derstanding of how global changes might alter the oscillations in atmospheric CO2 concentration reflect
balance between C 0 2 release to the atmosphere versus the metabolism of the terrestrial biosphere and
C 0 2 uptake and storage in ecosystems. underscore the importance of biological control over
Spatial and seasonal changes in atmospheric C 0 2 atmospheric C 0 2 concentrations.
concentration reflect the balance of photosynthesis and Global change can alter these biological controls,
respiration in the terrestrial biosphere. For example, affecting fluxes of photosynthesis and respiration and
photosynthesis is a larger flux than respiration thus the net exchange of carbon between the biosphere
during the growing season, and this imbalance is and atmosphere. Several lines of evidence suggest that
Figure 2 Three-dimensional representation of atmospheric carbon dioxide in the marine boundary layer. The surface represents data
that are smoothed in time and latitude. (Reprinted with permission of the principal investigators: Pieter Tans and Thomas Conway. NOAA
CMDL Carbon Cycle Greenhouse Gases, Boulder CO, (303) 497-6678, ptansQcmdl.noaa.gov.)
such changes are already under way. The amplitude of fertilization increase ecosystem C 0 2 uptake in the
the seasonal oscillations in atmospheric C 0 2 has short term, photosynthesis shows a saturating re-
increased over the past 40 years, particularly above the sponse to rising COZ, and excessive N deposition often
Northern Hemisphere, indicating greater activity of decreases soil fertility and plant production. Because
the terrestrial biosphere in this region. Inverse model- of the slow turnover time of soil organic carbon,
ing of atmospheric C 0 2 concentration and I3C com- increased carbon uptake observed in experiments
position and a number of land-based measurements of tends to overestimate the long-term potential for
biosphere-atmosphere C 0 2 exchange suggest a net ecosystem carbon storage.
sink for atmospheric C 0 2 in the Northern Hemi-
Methane
sphere. While deforestation in the tropics currently
amounts to a substantial source of C 0 2 to the Methane (CH4) is also a greenhouse gas, contributing
atmosphere, on the order of 1 . 7 G t C y P 1 , this is at roughly 15% to the radiative forcing associated with
least partly offset by afforestation of abandoned global warming. As with C 0 2 , two biological pro-
agricultural lands in the Northern Hemisphere and cesses are important in the production and consump-
net uptake of C 0 2 by tropical forests. Experimental tion of atmospheric CH4: methanogenesis and
additions of either C 0 2 or nitrogen (N)often stimulate methanotrophy. Methanogenesis is the largest source
ecosystem carbon uptake, at least in the short term, of CH4 to the atmosphere and is conducted by strictly
suggesting that globally pervasive N deposition and anaerobic microorganisms of the Domain Archaea,
increasing atmospheric CO2 may already be contrib- evolutionarily some of the oldest organisms on Earth.
uting to the current terrestrial C 0 2 sink. Currently, Methanogens thrive in many habitats where oxygen
terrestrial photosynthesis is estimated to be concentrations are low, in wetlands, sediments in
1.9 Gt C y - greater than respiration. However, it is aquatic ecosystems, and in the guts of termites and
unclear whether this terrestrial sink will persist. For ruminant mammals. With intensification and extens-
example, because photosynthesis is less sensitive to ification of livestock farming and rice cultivation,
increasing temperature than is respiration and micro- humans have increased methanogenic activity and
bial decomposition of plant residues, future stores of associated CH4 emissions to the atmosphere. Fossil
organic carbon in soils could decline by up to 11Gt C fuel mining and biomass burning are additional
globally for every 1"C of warming. Particularly a t high important anthropogenic sources of CH4 to the
latitudes, the combination of warming and drying atmosphere. While natural sources add around
could stimulate C 0 2 release from tundra and peat 160 Tg CH4 y - to the atmosphere, anthropogenic
bogs, ecosystems that contain most of the world's soil sources are larger, totaling nearly 375 Tg CH4 y - l.
carbon. Furthermore, while experimental C 0 2 and N Most CH4 is destroyed in the atmosphere through
interactions with O H radicals or is lost to outer space. CH4 uptake, sometimes even causing a shift from net
However, methanotrophs - literally ‘methane eaters’ - methanotrophy to net methanogenesis. Continued
are an additional important global CH4 sink, extensification of agriculture and livestock husbandry
consuming as much CH4 as accumulates in the is likely to further reduce the soil CH4 sink. Acid
atmosphere each year (40Tg) and constituting deposition also decreases CH4 oxidation, owing to
approximately 10% of the global CH4 sink. Methano- physiological responses of methanotrophs to low soil
trophs are bacteria that oxidize CH4 to generate pH, and also, in the case of nitrogen deposition, to
energy and fix C 0 2 into organic compounds, using inhibitory effects caused by increased NH,’ availabil-
CH4 in a manner analogous to a plant’s use of ity in soil. Ammonium-oxidizing bacteria (nitrifiers,
light in photosynthesis. Methanotrophs are wide- see below) have the capacity to oxidize methane
spread in nature, but are quantitatively important because of structural similarities in the ammonia- and
in the global CH4 cycle in the surface soils of terrestrial methane-binding enzymes. For this reason, it is not
ecosystems. Many temperate forests, for example, always apparent which groups of organisms are
are net CH4 sinks, consuming CH4 from the responsible for methane oxidation in soils. Compared
atmosphere. to methanotrophs, however, nitrifiers have a lower
Global change has the potential to increase or affinity for CH4 and therefore oxidize CH4 more
reduce CH4 production in soils. For example, in- slowly, so displacement of methanotrophs by nitrifiers
creased plant production in wetlands enhances sub- in response to N deposition may exacerbate inhibition
strate availability for anaerobic decomposition, thereof CH4 oxidation by ammonium. The inhibitory
by stimulating CH4 production. Generally, CH4 pro- effects of NH: on methane oxidation may also explain
duction in wetlands increases with plant productivity. the reduced capacity for CH4 oxidation following land
Thus, global changes that increase plant production, use change, as NH; availability often increases after
such as elevated C02, are likely to stimulate CH4 land clearing or because of direct application of
efflux from wetlands. Elevated C 0 2 often reduces nitrogenous fertilizers.
transpiration (water loss) from wetland plants, raising
Oxides of Nitrogen
the water table in wetlands and enhancing anaerobic
conditions, again potentially stimulating CH4 efflux. Nitrous oxide ( N 2 0 )is a potent greenhouse gas, with
In the Arctic, thawing of permafrost could increase 300 times the warming potential of CO2 on a molar
CH4 emissions by increasing methanogenic activity in basis. N 2 0 concentration in the atmosphere is in-
previously frozen soils. By contrast, if warming causes creasing a t a rate of 0.3% per year, and it is responsible
widespread drying of wetlands and tundra, anaerobic for around 5% of the radiative forcing associated with
activity could decline and rates of methane oxidation global warming. N 2 0 also plays a critical role in
increase. Experimental evidence in support of both stratospheric ozone depletion. Nitric oxide ( N O ) and
scenarios contributes to the difficulty in predicting nitrogen dioxide (NO2),together referred to as NO,,
biospheric feedbacks to climate change. Widespread are important reactive gases, influencing tropospheric
disappearance of wetlands could diminish the terres- concentrations of 0 3 , OH, H N 0 3 , and CH4. While
trial CH4 source. their effects on atmospheric processes differ funda-
Global changes are likely to reduce methane con- mentally, N 2 0 and NO, are produced and consumed
sumption by soils, diminishing the capacity of terres- by the same groups of organisms, and so these gases
trial ecosystems to consume atmospheric CH4. By are treated together here.
increasing soil water content (through reduced tran- Nitrification and denitrification are the two major
spiration) and thereby reducing CH4 diffusion into biological sources of N 2 0 and NO,. Nitrifying
soil, elevated C 0 2 has been shown to reduce rates of bacteria use ammonia (NH3) as an energy source,
CH4 oxidation by a forest soil. In another case, CH4 oxidizing it to nitrate (NO;). Like methanotrophs,
consumption rates declined in C02-treated forest which oxidize CH4, nitrifiers use their substrate (NH3)
plots even without a change in soil water content or to fix carbon in a manner analogous to a plant’s use of
any other obvious mechanism, raising the possibility light in photosynthesis. Incomplete oxidation of NH3
of reduced efficiency in CH4 oxidation due to changes results in gaseous losses of nitrogen dioxide (NOz),
in the methanotrophic bacteria community. Reduced nitric oxide (NO),and N20; losses of these oxides of
CH4 oxidation has also been observed in grasslands nitrogen to the atmosphere can account for as much as
exposed to elevated C02. Thus, reduced CH4 uptake 5% of the total nitrification flux. Denitrifying bacteria
may be a general consequence of rising atmospheric ‘breathe’ NOS, using it as a terminal electron acceptor
C02. Land use change, too, is likely to reduce the just as animals (and other aerobic organisms) use 0 2 .
terrestrial CH4 sink: converting forests, woodlands, In the process, denitrifiers reduce NO5 to N2 gas, with
and savannahs to cultivated or grazed lands reduces several intermediates, including N 0 2 , NO, and N20.
The final step in the denitrification chain is reduction ing and N deposition are both likely to increase NO,
of N20 to N2, a step that is often incomplete, as fluxes from terrestrial ecosystems.
gaseous N20 escapes from the reaction site. NO,, too,
can be emitted during the earlier steps. Denitrification Other Trace Gases
can also consume both N20 and NO,, and other
reports suggest that NO, consumption by other (non- Carbon monoxide (CO) affects atmospheric chemis-
denitrifying) soil microorganisms may serve a detoxi- try by contributing to tropospheric ozone formation
fying function or may have no clear physiological role. and interfering with methane destruction in the
However, it is not yet clear whether net NLO or NO, stratosphere. Up to 3% of net primary productivity
consumption by soils has any major influence on the can be lost as direct C O emission by plants or as losses
global budget of these trace gases. Plants also play a of volatile organic hydrocarbons (VOCs) subsequent-
role in atmospheric NO,: both uptake and emission of ly oxidized to CO. Soils are both an important source
NO2 by plant leaves are known to occur, but the and sink of atmospheric CO. A number of bacteria are
balance of these processes at the global scale is not capable of oxidizing C O to C 0 2 , including some
known. methanotrophs (also important in CH4 uptake) and
Gaseous losses of NO, and N 2 0 during nitrification nitrifiers (important in N20 and NO, production).
and denitrification represent small proportions of the Selective inhibition experiments indicate that even
total flux of nitrogen through these pathways. Never- eukaryotes may contribute to C O consumption in
theless, the quantities produced are appreciable for the some forest soils. C O oxidation is sensitive to water
global fluxes of these important trace gases. Globally, and organic matter content of soils, but is less sensitive
soils amount to 65% of total atmospheric sources of to temperature than many other ecological processes.
N2O and 40% of NO,. Tropical (NZO), wetland While most soils are net C O sinks, some, particularly
(N20),arid (NO,), and cultivated soils (both) are the in arid regions, are net CO producers. The mechanism
major contributors of these important trace gases. for CO production is unknown, but it is apparently
Rates of N20 production by soils are likely to increase abiotic, as sterilization often converts a soil from a net
with climate warming, because, like most biological C O sink to a source. Compared to other trace gases,
reactions, N20 production is temperature-sensitive the biology of C O consumption is not very well
and has been found to increase in response to understood.
experimental warming in a number of cases. A Plants emit a number of VOCs to the atmosphere,
lowering of the water table in tundra and peatlands including isoprenoids (isoprene), terpenes, and oxy-
associated with warming increases N2O emissions, genated compounds, fluxes totaling 1.2 Pg y-'. These
probably because the slightly more oxidizing condi- compounds influence carbon monoxide, ozone, and
tions favor incomplete reduction of NO; during aerosol formation in the troposphere. About half of
denitrification, increasing the amount of N 2 0 lost in plant VOC emissions to the atmosphere occur as
the reaction chain. Conversion of tropical forests to isoprene. Effects of global change on VOC emissions
pasture, one of the more prevalent land use changes in have not been characterized as well as for other trace
the tropics, usually decreases N 2 0 losses from soils. gases; nevertheless, changes in VOC emissions are
However, pasture reclamation, which often involves likely to occur. Because plant species naturally differ in
application of fertilizer, can cause N 2 0 fluxes to return rates of VOC production, land use changes and other
to or even exceed levels typical of native forests. global changes that result in shifts in species compo-
Additionally, conversion of native ecosystems to sition are likely to profoundly alter VOC emissions to
agriculture often dramatically increases N 2 0 losses the atmosphere. For example, in general, native
from soils. Thus, ongoing global changes are likely tropical forest trees produce more isoprene than either
to increase rates of N20 accumulation in the crops or pasture grasses, such that land clearing for
atmosphere. agriculture or grazing probably reduces VOC emis-
Emissions of NO, to the atmosphere are also sions. Secondary forests, however, contain many
increasing, particularly from cultivated (and ferti- species with high rates of VOC emissions, and net
lized) soils. As with N 2 0 , land clearing, cultivation, emissions from secondary forests may exceed those
and fertilization in agriculture increase soil emissions from primary forests in some cases. For this reason,
of NO,, and global increases in these land use changes land use change could cause VOC emissions to the
are increasing the soil source of NO,. Rising atmos- atmosphere to increase or decline. Isoprene and
pheric CO2 has been shown to reduce NO, fluxes terpene emissions increase exponentially with temper-
associated with nitrification by increasing the demand ature, and are thus likely to increase in response to
for NH3 by nonnitrifying microorganisms, reducing global warming. Global changes that increase plant
substrate availability for nitrifiers. By contrast, warm- production ( C 0 2 , N deposition) are also likely to
Table 1 Trace gases influencing atmospheric chemistry and climate that are produced and consumed by the terrestrial biosphere.
Shown are the gas species of interest, the organisms responsible for producing and consuming each, a rough estimate of the percentage
of total annual production or consumption mediated by the terrestrial biosphere (e.g.,of all global processes that consume COS each year,
-.
plants are estimated to consume 70%),and a brief description of the role of the gas species in atmospheric chemistry and climate
Gas Producing Terrestrial Consuming Terrestrial Atmospheric

organisms production (%) organisms consumption (%) consequences
COP All organisms 40 Plants, autotrophic 40 Radiative forcing

bacteria
cos Sulfur-reducing 21 Plants, soil 72 Aerosol
microorganisms in microorganisms
salt marshes and soils
Non-methane VOCs Plants 80 Soil microorganisms Negligible Reacts with
oxidized
pollutants
' 3 4 Methanogens 70 Methanotrophs 10 Radiative forcing
(nitrifiers)
Nz0 Denitrifiers, nitrifiers 65 Denitrifiers Unknown Radiative forcing
NO and NOp Denitrifiers, nitrifiers, 40 Denitrifiers, plants Unknown Photochemical
plants smog
co Plants 5 Soil microorganisms, 15 O3formation
plants (troposphere);
CH4
destruction
(stratosphere)
increase VOC production. In concert with emissions Water and Energy Exchange
of NO, from combustion and use of fertilizers,
increased VOC emissions could increase concentra- Terrestrial ecosystems interact with the atmosphere
tions of tropospheric ozone and carbon monoxide. through exchanges of energy, moisture, and momen-
Carbonyl sulfide (COS) is the most abundant sulfur tum at all temporal and spatial scales. These interac-
species in the atmosphere. COS is primarily produced tions are functions of key land surface parameters that
abiotically in the oceans; anaerobic production by soil derive from the structure and physiology of terrestrial
microorganisms occurs, particularly in salt marshes, ecosystems. The bidirectional nature of ecosystem-
but is a small source globally. In the stratosphere, COS atmosphere interactions is clearly evident. For exam-
is oxidized by photolysis to sulfate aerosols, forming ple, the gas exchange of leaves responds to variation in
an aerosol layer important in cooling the planet but light, temperature, vapor pressure, and COZ partial
also in enhancing depletion of stratospheric ozone pressure and, in turn, affects the vapor pressure,
through chemical interactions with nitrogen and temperature, and trace gas composition of the atmos-
chlorine species. Biological uptake of COS by soils phere. Over long periods and broad spatial scales,
has been documented, but its significance on the global ecosystem-atmosphere interactions are yet more evi-
scale is not well constrained. Lichens are also capable dent. Water availability and temperature range are the
of COS uptake, probably representing a small but major determinants of the structure and function of
nontrivial sink. Uptake by terrestrial vegetation is the terrestrial ecosystems over decades to centuries, and
major global sink for COS (Table 1).Enzyme inhibi- the paleoecological and paleoclimatological records
tion studies suggest that COS uptake is probably indicate that glacial-interglacial cycles involve cou-
mediated by carbonic anhydrase, an enzyme that pled changes in the distribution of terrestrial ecosys-
occurs in plants, lichens, and bacteria. COS enters tems, surface albedo, biogeochemistry, and climate. It
higher plants through stomata, and there is some is understood that the Earth's climate and ecological
evidence that reduced stomatal aperture impedes COS systems have coevolved over geologic time. Current
uptake by vegetation. Thus, global changes that changes in climate, atmospheric composition, and
reduce stomatal conductance (e.g., increasing atmos- land cover arising from human activities will probably
pheric COZ) may reduce the terrestrial COS sink, continue to affect and be affected by ecosystem-
possibly constituting a negative feedback to global atmosphere interactions.
warming. By contrast, COS uptake in lichens appar- In this section we first explain the biophysical and
ently increases with temperature, a possible positive physiological basis for water and energy interactions
*eedback to global warming. between terrestrial ecosystems and the atmosphere.
882 GLOBAL CHANGE I Biospheric Impacts and Feedbacks
We then examine some of the major ways in which tied to movement of vegetation zones in response to
global change affects ecosystem-atmosphere interac- climate variation.
tions. Much of our current understanding derives from
atmospheric general circulation models ( AGCMs), Sensible heat, latent heat, and evapotranspira-
which have developed rapidly in recent years to tion Ecosystems have a major influence on how the
include realistic parametrizations of the land surface, net radiation received at the surface is balanced by
including the physiological responses of vegetation to losses of sensible and latent heat, which in turn have
multiple environmental factors. Although much un- profound effects on weather and climate. Sensible heat
certainty exists in the magnitude, and in some cases flux from land warms the overlying air and planetary
direction, of ecosystem-atmosphere interactions, boundary layer. Latent heat is the energy required to
modeling and large-scale experimental studies clearly evaporate water from soils and plants, and acts to cool
indicate that the potential feedbacks are large. the surface. Evaporated water can be transported
above the planetary boundary layer, where it may
The Biophysics and Physiology of
release heat during condensation to form clouds
Ecosystem-Atmosphere Interactions
(which affect radiation balance) or precipitation,
The presence of vegetation affects a number of key often at considerable distance from the site of surface
land surface characteristics that are important deter- evaporation. The presence of vegetation generally
minants of the surface energy balance and water and increases latent heat flux relative to sensible heat flux.
energy exchanges with the atmosphere. Of the 111000 km3 of precipitation on land each year,
about 71 000 km3 returns to the atmosphere by
Albedo Vegetation affects the reflectivity, or albedo,
evaporation from soils and transpiration from plants
of the Earth’s surface, the land surface parameter
(together considered as ‘evapotranspiration’, ET), the
having the largest influence on the surface radiation
balance reentering the oceans as river flow. Globally,
budget. Albedo is the integrated reflectance over the
latent heat releases from the surface about three-
solar spectrum (0.0-4.0 pm) and is lower for vegeta-
quarters, and sensible heat one-quarter, of the annual
tion than for bare soil (Table 2) because leaves absorb
average net radiation received by the surface.
strongly in the visible wavelengths (0.4-0.7 pm) useful
The ratio of sensible to latent heat flux densities is
in photosynthesis and moderately from 0.7 to 4.0 pm.
known as the Bowen ratio, p. Average p varies
In contrast, soils have lower average absorption across
inversely with water availability on land, ranging
the solar spectrum. Although leaves of different plant
from about 0.2 for tropical rain forests to 0.4-0.8 in
species vary somewhat in reflectivity, albedo is more
temperate forests and grasslands, 2-6 for semi-arid
sensitive to the leaf area index (LAI), the average
regions, and 10 for deserts. The Bowen ratio is not
amount of leaf area per ground area (m2m - 2 ) . Values
static, however; distinct daily, seasonal, and annual
of LA1 range from 0 for extreme deserts to 1 in arid
trends arise from variation in LAI, soil and plant
regions and up to 5 to 7 or more for forests. The
surface moisture availability, the surface-atmosphere
influence of vegetation on albedo also varies tempo-
vapor pressure deficit (VPD),and the resistance of the
rally. For example, forest albedo generally increases in
plant canopy to transpiration. Transpiration is the
winter when deciduous trees are leafless or evergreens
diffusion of water vapor along the concentration
are snow covered; cropland and grassland albedo
gradient from the saturated interior surfaces of leaves
varies seasonally as the plant canopy develops and
to the surrounding air via microscopic pores known as
then senesces; and extreme events including droughts,
stomata (Figure 3), and is the plant process that most
pest outbreaks, and severe storms can defoliate plants
directly impacts ecosystem-atmosphere exchange of
and increase albedo. Over longer time scales, the
water and latent heat. Transpiration can enhance total
changing distribution of land surface albedo is closely
surface evaporation because plants can extract water
from below the soil surface, which itself develops
high resistance with mild drying, and because plant
Table 2 Reflectivity (albedo) of various surfaces
canopies can present multiple layers (LA1 >1) for
Surface Albedo evaporation. An example of the importance of tran-
Forests 0.05-0.18
spiration to water balance is apparent from figures for
Grassland 0.22-0.28 the Amazon basin estimating that up to 50% of
Crops 0.15-0.26 regional precipitation arises from water transpired
Snow 0.75-0.95 from plants elsewhere in the basin.
Wet soil 0.09i0.04 Stomata act as variable-aperture valves, the resist-
Dry soil 0.19k0.06
Water 0.05to 10.20
ance (or conductance) of which responds to environ-
mental factors including light, temperature, soil
land surface. Aerodynamic resistance varies inversely

with wind speed and the log of the surface roughness
length, the latter being about 10% of the vegetation
height. Thus, for a given wind speed, taller vegetation
such as forest has a lower aerodynamic resistance than
shorter vegetation (e.g., grassland), and promotes
greater turbulent transfer of sensible and latent heat
away from the surface. The vertical structure of the
daytime atmospheric boundary layer is also highly
dependent on partitioning of sensible and latent heat; a
deeper boundary layer develops when sensible heat
flux is large and convection increases.
Terrestrial ecosystems have several additional influ-
ences on land surface properties and processes. Veg-
etation shields soil from radiation inputs, reducing the
Figure 3 Schematic representation of exchange of water vapor magnitude of the components of soil energy budget.
( e ) , carbon dioxide (C), and heat by leaves. Gas flux rates are
determined by the leaf-to-air concentration gradient and the
Plant canopies also intercept precipitation, decreasing
combined stomatal and boundary layer resistance to diffusion. moisture reaching the soil and cooling the canopy
Light-dependent photosynthetic assimilation of COP causes day- surface when intercepted rain evaporates. Finally, the
time C, < C, and a net uptake of COP by the leaf, while in the distribution of roots in the soil profile affects the water
absence of light, leaf respiration causes C, > C, and the net flux is available for evapotranspiration. Studies of variation
from the leaf. Stomatal resistance is under physiological control
in the natural abundance of stable isotopic forms of
and responds to light, temperature, humidity, and COPconcentra-
tion. Subscripts i, s, and a refer to interior, surface, and ambient, water (Hll'O and lHIHO) in precipitation, ground
respectively. (After Sellers et a/. (1997).) water, soils, plants, and the atmosphere are beginning
to reveal the different sources of water that is
evaporated from soils or transpired by plants, and
moisture, VPD, and C 0 2 concentration. Canopy the importance of these atmospheric inputs to local
conductance is the stomatal conductance averaged and regional vapor pressure and precipitation.
over the total canopy leaf area. Because photosyn-
thetic uptake of COZ by land plants shares the same
Terrestrial Ecosystem Effects on Climate
stomatal diffusion pathway as transpiration, stomatal
conductance simultaneously influences water, energy, The land surface parameters described above underlie
and carbon fluxes. Theoretical studies suggest that the influences of ecosystems on climate that are evident
stomatal conductance sometimes varies so as to from observational and modeling studies. Sparsely
optimize the efficiency of carbon gain relative t o water vegetated urban areas feel warmer, and are warmer,
loss. Conditions tending to enhance water stress to than nearby forests or grasslands. In the north central
plants, particularly low soil moisture and high VPD, United States, the rate of increase of mean surface
typically lead to partial stomatal closure, which then temperature during spring slows abruptly as decidu-
causes a decrease in transpiration and latent heat flux ous trees leaf out and increased latent heat flux imparts
and an increase in surface temperature and sensible a cooling effect that counteracts the increasing
heat flux (i.e., /? increases). The potential for ecosys- insolation. The aridity of deserts is reinforced by their
tem-atmosphere feedbacks mediated by stomatal lack of vegetation and, conversely, transpiration by
physiology can be appreciated from the fact that tropical rainforest returns to these regions as precip-
stomatal conductance, through its control on transpi- itation. Avariety of AGCMs have explored vegetation
ration and latent heat, can influence the conditions controls on climate for major regions of the terrestrial
(e.g. vapor pressure deficit) to which it also responds. biosphere. Forest clearing for agriculture in the eastern
This is relevant in the context of global change because and central United States during the 19th century is
several global change factors (altered temperature, estimated to have resulted in cooler summers and
moisture availability, and C 0 2 ) affect stomatal con- autumns, consistent with the instrumental record. The
ductance and thus, potentially, can feed back to either cooling was due to changes in vegetation characteris-
amplify or moderate these changes. tics, in this case primarily reduced leaf area, which
increased albedo and reduced net radiation and
Aerodynamic resistance A third means by which sensible heating of the atmosphere. AGCM simula-
ecosystems affect surface-atmosphere interactions is tions that replaced Amazon forests with grassland
via their influence on the aerodynamic resistance of the estimated substantial reductions (300 mm y - ') in
evapotranspiration and precipitation and increases in warming that may be without precedent in the past
surface temperatures of 3-5°C. Changes in evapo- 10 000 years. Temperature increases for terrestrial
transpiration and precipitation arose from reduced ecosystems should be greater than the global average
LA1 and decreased roughness length, which reduced because the land surface is not as thermally buffered as
turbulent transfer of moisture above the planetary the oceans. High-latitude regions are very likely to
boundary layer. Surface temperatures increased be- warm considerably more than equatorial and mid
cause latent heat flux decreased more ( 1 2 W r n p 2 ) latitudes (a phenomenon known as 'polar amplifica-
than did absorption of solar radiation ( 6W m - 2 ) , the tion') because warming will reduce the extent of snow
latter due to increased albedo. Reduced evapotranspi- and ice cover, decreasing albedo and increasing net
ration following rainforest clearing is also predicted to radiation. Winter warming may exceed summer
decrease cloud cover, which tends to counteract the warming in high latitudes, while in arid regions, where
cooling effect of increased surface albedo. The reduc- soil moisture is low, summer warming may exceed
tion in cloud cover also increases surface insolation, winter warming because of low latent heat loss.
but also allows more outgoing long-wave radiation to Warming will drive greater ocean evaporation, and
be lost by the warmed land surface. globally averaged precipitation is expected to increase
An AGCM scenario of replacing all boreal forests by about 10%. Precipitation is likely to increase in
with bare ground or tundra (short stature vegetation of high latitudes in winter as major storm tracks shift
low leaf area), showed that boreal forests have a strong northward. At mid-latitudes, precipitation is likely to
warming influence during winter and summer relative increase in winter, with rain increasing relative to
to the alternative cover situations. The warming snow, but decrease in summer in continental interiors.
results from the lower albedo of forests than of Aridity is likely to increase in currently arid regions
snow, which is masked by the overlying trees. Under because, although increased surface temperatures will
the deforestation scenario, land surfaces at high drive more evaporation, precipitation will decrease
latitudes were up to 12°C colder in April, and because the water-holding capacity of air increases
remained as much as 5°C colder in July. These nonlinearly with temperature according to the Clau-
terrestrial changes were linked to sea surface condi- sius-Clayperon relationship, and so air vapor pressure
tions; the colder winter air temperature caused by will tend to be farther from saturation. Most AGCM
deforestation reduced sea surface temperature (SST)in projections also forecast increased frequency of ex-
Arctic regions, inducing a thermal lag that inhibited treme weather events including severe storms and
warming of land surfaces in the summer. Lower SSTs droughts.
increased the extent of sea ice, increasing ocean albedo The primary approach to understanding climate-
and reinforcing the cooling effect of deforestation. driven changes in vegetation distribution is based on
These simulated effects of boreal deforestation were current correlations between natural vegetation and
not limited to the boreal region itself; at latitudes as climate, the so-called 'equilibrium' approach. The
low as 30" N simulated air temperatures were up to assumption is that, after climate changes, vegetation
3°C cooler throughout the year in the deforestation will reequilibrate according to the same relationship
scenario compared to current, control conditions. that currently exists. Thus, for example, as high
These examples of ecosystem impacts on climate latitudes warm, boreal forests will migrate northward
illustrate the importance of including land cover and arctic tundra will be compressed into a narrower
change in models of future global climate. latitudinal band. Similarly, vegetation zones should
shift upward in mountainous regions to track increas-
Global Change and Ecosystem-Atmosphere
es in temperature, a pattern of change that would
Interactions
reduce the extent and biodiversity of the uppermost,
Global change, including changes in climate, atmos- alpine tundra ecosystems. Zones of optimal grain
pheric composition, and land use will alter ecosystem- production in north America may shift northward into
atmosphere interactions by changing the abundance, Canada. In some continental regions, predictions are
distribution, and functioning of different terrestrial confounded by large uncertainties in precipitation
ecosystems. For clarity, we discuss ecosystem respons- changes and the relative influence of increased tem-
es and feedbacks to each of these major global changes perature and altered precipitation in increasing or
separately, but recognize that these interactions are decreasing moisture stress to vegetation.
likely to be highly interdependent in the real world. Although simple, an equilibrium approach is po-
tentially misleading. Vegetation is unlikely to migrate
Climate change Global climate change models rapidly enough to track the predicted pace of anthro-
project an increase in globally averaged temperature pogenic climate change, meaning that there will be
of 1.4-5.8"C over the period 1990 to 2100, a rate of transient periods in which vegetation and climate are
GLOBAL CHANGE I Biospheric Impacts and Feedbacks 885
mismatched with respect to the equilibrium models. In assimilation and the larger increases in surface
addition, other determinants of future vegetation temperature come about when leaf nitrogen concen-
distribution, including herbivory, fire, severe storms, tration, which is correlated with levels of photosyn-
and human activities, can strongly affect vegetation thetic enzymes, decreases in elevated COZ and
composition, independently of the climate-vegetation stomatal conductance declines strongly, consistent
correlation. In Australia, for example, the frequency with maintenance of high efficiency of photosynthetic
of fire is expected to increase with climate change water use. Thus, ecosystem biogeochemical responses,
because of an increased probability of the high particularly components of the nitrogen cycle, can
temperatures associated with ignition of bush fires. influence the response of water and energy exchanges
Frequency and intensity of fire may also be affected by to elevated C02.
changes in chemical properties of plant tissues pro-
duced in elevated C02. As evidenced by widespread Land cover and land use change Ongoing human-
fires in Indonesia in 1998, wildfire can have major caused changes in land use and land cover will
impacts on atmospheric properties. There is evidence undoubtedly impact biosphere-atmosphere interac-
that outbreaks of insect herbivores are increasing in tions, water and energy exchanges being no exception.
boreal forests, but their implications for local, region-These impacts arise from the influence of key land
al, or global climate have yet to be explored. Given surface parameters (roughness length, albedo, LAI) on
these complexities, coupled vegetation-AGCM mod- climate, as described above. Predictions of patterns of
els are increasingly including much transient, and in future land cover change carry substantial uncertainty,
some cases nonlinear, behavior of vegetation distribu- because land use depends on many different social and
tion, behavior that can feed back to alter the pattern economic factors. It is safe to assume, however, that in
and pace of climate change. tropical regions the trend toward increased conversion
of forests to grasslands and croplands will continue. In
Rising atmospheric COZ The stomatal response to the developed regions of the temperate zone, there has
C 0 2 is critical to determining overall vegetation been a net increase in forested area in some regions as
feedback to climate because it impacts latent and increased agricultural efficiency has allowed crop-
sensible heat exchange and return of water vapor to lands to be abandoned and to convert back to forests.
the atmosphere in transpiration. Experimental studies Increased rates of timber extraction in some regions
with a large number of plant species indicate that such as Siberia may introduce climatic effects quali-
stomatal conductance is reduced by about 25% with a tatively similar to those estimated in simulations of
doubling of current atmospheric C 0 2 , there being extreme boreal deforestation. Retrospective studies of
considerable variation among species, with coniferous past, well-quantified, patterns of conversion of natural
trees typically responding less than deciduous trees vegetation to various agricultural conditions (pastures
and herbaceous species typically showing the largest and crops and associated irrigation) indicate that land
responses. The stomatal response to C 0 2 can change cover change has influenced regional climate, but
over time as plant physiology adjusts to altered counteracting influences of land use and land cover
ecosystem biogeochemistry and the availability of change in tropical and temperate regions have prob-
essential resources, notably soil nitrogen. Whereas ably caused little change in global temperature.
early AGCM studies addressed only the radiative In general, tropical forest conversion has warmed
effects of increased C 0 2 , current models specifically those regions while temperate forest conversion has
examine the role of vegetation physiology by allowing produced a cooling. This balance is not likely to
stomatal conductance and photosynthesis to respond continue, however, because tropical deforestation has
to climatic conditions and atmospheric COZ. Com- accelerated and temperate deforestation may be
parisons of AGCM simulations for the case of radia- decelerating.
tion only (physical effects of C02) with situations
where vegetation physiology also responds in a real-
Summary and Conclusions
istic manner to increased C 0 2 clearly illustrate the
importance of vegetation feedbacks to the climate Together, the atmosphere and terrestrial biosphere
system. For tropical latitudes, the fully adjusted form a coupled, interactive system, exchanging
physiological response (reduced stomatal conduct- materials and energy in ways that critically regulate
ance) accounts for one-third of the nearly 3°C increase both biological and atmospheric processes (Figure 1).
in surface temperature. Global C 0 2 assimilation by Biological production and consumption of greenhouse
vegetation increased by 11-36% depending on the gases and aerosols influence the radiative balance of
degree of physiological adjustment of levels of the atmosphere, while production and consumption of
photosynthetic enzymes. Lesser increases in C 0 2 reactive gases affect atmospheric chemistry, including
886 GLOBAL CHANGE / Human Impact of Climate Change
formation and destruction of important pollutants. See also

Similarly, plants mediate the partitioning of solar Aircraft Emissions, of Earth,s Atmos-
radiation incident on the terrestrial surface, partition- phere. Gaia Hypothesis. Global Change: Human
ing that affects regional and even global climate. Impact of Climate Change.
Primarily mediated by plants and microorganisms,
biosphere-atmosphere exchanges vary over the globe
because of organisms’ physiological and ecological Further Reading
responses to the environment. Ongoing, human- Field CB and Avissar R (eds) (1998) Special Issue: Bidirec-
caused global changes are already altering bio- tional interactions between the biosphere and atmos-
sphere-atmosphere exchanges, and these impacts are phere. Global Change Biology 4.
large enough that they could substantially shape the Intergovernmental Panel on Climate Change (2001)Climate
responses of the coupled biosphere-atmosphere sys- Change 2001. Contribution of Working Group I: The
tem to future global changes. To better understand the Science of Climate Change. Cambridge, UK: Cambridge
consequences of anthropogenic global change, it will University Press.
Mooney HA and Canadell JG (2002) The Earth system:
be critical to incorporate these complex interactions in
biological and ecological dimensions of global environ-
atmospheric general circulation models. mental change. In: TA Munn (ed.) Encyclopedia of
Global Change, vol. 2. Chichester: Wiley.
Prosser J (ed.) (1997) Special Issue: Microbially Mediated
Acknowledgments Atmospheric Change. Global Change Biology 3(4).
Schlesinger XX (1997) Biogeochemistry: A n Analysis of
We thank Jeff Amthor, Paul Dijkstra, James Holton, Global Change, 2nd edn. New York: Academic Press.
Manuel Lerdau, and Oleg Menyailo for comments on (See especially chapters 3, 5-7, 11-13.)
the manuscript, and Pieter Tans for permission to Sellers PJ, Dickinson RE, Randall DA, et al. (1997)Model-
reproduce Figure 2 here. We thank the editors for ing the exchanges of energy, water, and carbon between
inviting us to write this article. continents and the atmosphere. Science 275: 502-509.
Human Impact of Climate Change

M M Hedger, EnvironmentAgency, Almondsbury, years. But it was during the 1980s that scientific
Bristol, UK evidence about the possibility of global climate change
Copyright 2003 Elsevier Science Lid. All Rights Reserved. and its consequences became an issue of growing
concern for scientists, decision-makers, and the pub-
Introduction lic. It became widely recognized that human-activity
can cause climate changes and can also be impacted by
This report overviews what is known about the human those changes. In 1988, the United Nations Environ-
impacts of climate change. Knowledge in this area is not ment Program (UNEP)and the World Meteorological
well defined, because it depends on the extent of human- Organization ( W M O ) jointly established the Inter-
influenced climate change that occurs, which itself Governmental Panel on Climate Change (IPCC). A
depends on future emissions of greenhouse gases (ghgs), series of international conferences issued urgent calls
and measures which may be taken to reduce them for a global treaty to address the problem. The UN set
(mitigation). Furthermore, some impacts can be amelio- up a negotiating process which led eventually to the
rated if measures are taken to respond to them (adapta- UN Framework Convention on Climate Change
tion). Because the stakes are high, there is considerable (FCCC), which was signed by 155 countries at the
scientific effort being devoted to increasing understand- United Nations Conference on Environment and
ing. In order to help explain the levels of confidence or Development, also known as the Earth Summit, in
otherwise that can be placed in information, sources and Rio de Janeiro in 1992.
methods for generating it are also outlined in the review. Since then, an ongoing political negotiating process
has been underway, designed to strengthen the treaty
and establish positive actions. On the scientific side a
Context parallel institutional process has continued. Through
In the scientific community the greenhouse effect has the IPCC, scientists representing more than 150
been known about and investigated for well over 300 countries have assessed the available information on
GLOBAL CHANGE / Human Impact of Climate Change 887
climate change and its environmental and economic season length over the second half of the twentieth
effects, and have provided the scientific understanding century in many Northern Hemisphere midlatitude
needed to help formulate appropriate responses. A regions.
series of IPCC reports (1990, 1995, and 2001), Shrinkage of glaciers has been detected, and also the
incorporating extensive peer review, has provided later freezing and earlier breakup of ice on rivers and
the most authoritative and comprehensive informa- lakes.
tion available. The assessment is undertaken in three Sea level increased by about 20cm between 1900
groups: on the science of climate change (Working and 2000.
Group I); the impacts (Working Group 11); and the Poleward and altitudinal shifts of plants and animal
economics of and the options for mitigating and/or ranges have been observed, together with declines of
adapting to climate change (Working Group 111). In some animal and plant populations and earlier
addition, the IPCC also provides on request scientific, flowering of trees, emergence of insects, and egg-
technical, and socioeconomic advice to the conference laying in birds.
of the parties to UNFCCC. In 2001 the IPCC In North Africa, there has been a relative dryness
published its most recent information on climate over the past few decades, a phenomenon thought to
change science, which is the main source for this be at least partly related to a change in circulation
review. patterns over the past few decades.
The Sahel has become much drier over the last 25
years.
What Is Meant by Climate Change?
In the Framework Convention in Climate Change, Special mention should be made of the El Nifio
climate change refers to a change in climate that is Southern Oscillation (ENSO). While confidence in
attributed directly or indirectly to human activity that future occurrences of the frequency and magnitude of
alters the composition of the global atmosphere and ENSO events is complicated by the fact that intricate
that is in addition to natural climate variability global circulation models do not capture the phenom-
observed over comparable time periods. In IPCC enon well, there are indications however that the
usage, climate change refers to any change in climate increase in greenhouse gases in the atmosphere has
over time, whether due to natural variability or as a been responsible at least partly for the observed
result of human activity. warming over a broad triangular region in the Pacific
The ultimate aim of the FCCC is to stabilize Ocean associated with ENSO. Associated floods and
greenhouse gas concentrations 'at a level that would droughts are some of the most extreme events expe-
prevent dangerous anthropogenic interference with rienced, with profound implications for agriculture,
the climate system'. This stabilization should be forests (fires), flooding, droughts, human health, and
achieved within a time frame that (1)allows ecosys- natural systems such as coral reefs which are irnpor-
tems to adapt naturally to climate change, (2)ensures tant economic assets. ENSOs have occurred more
that food production is not threatened, and ( 3 )enables often since 1975. Measurements covering the last 120
sustainable economic development to proceed. years indicate that the duration of the 1990-95 event
was the longest on record, and the 1997-98 related
events included extensive flooding in the Horn of
The Impact: What Is Happening Africa, delayed monsoons in Indonesia with pro-
Already? longed fires, and widespread respiratory illness. Trop-
ical forest fires in Brazil, Central America, and
In the year 2000, average global mean surface air Mexico, raged out of control. Summer heat waves
temperature was some 0.6"C higher than temperatures killed hundreds in Florida. The 1998 La Nifia event
at the end of the nineteenth century. The rate and continued the extreme weather, with impacts such as
duration of warming of the Northern Hemisphere in widespread flooding in Bangladesh and China - the
the twentieth century appears to have been unprece- latter displacing 230 million people and - Hurricane
dented in the millennium. Changes observed in a Mitch, which swamped the Central American isth-
diverse set of physical and biological systems in mus. Overall, 1998 proved to be the most costly year
many parts of the world suggest that regional in- on record in weather-related impacts.
creases in climate, particularly increases in tempera-
ture, have already affected physical and biological
systems. Climate Change Science
There has been a near-uniform decrease in the Currently, climate change is not expected to be
number of frost-free days, and an increased growing significantly detectable above natural variability until
888 GLOBAL CHANGE I Human Impact of Climate Change
the 2020s, and its effects will deepen through the rest Assessing the Human Impacts of
of the century. The assessment of its impact therefore Climate Change
involves ambitious long-term future-gazing, and is a
developing area of science. It follows that the methods While there are a number of ways of examining
and contexts in which this endeavor is undertaken are climate change impacts, an integral feature is the
crucial to the outcomes and need to be made explicit. assessment of changes associated with climate scenar-
In this report, as indicated, particular use is made of ios - although the emphasis given to this assessment
the IPCC Third Assessment Report, because it is the does vary. Scenarios are plausible alternative futures
major synthesis of current research and was recently intended to explore possible changes. Other scenarios,
published (2001). particularly those intended to explore social and
In the latest assessment by the IPCC W G on science, economic dimensions in the future, also need to be
averaged surface air temperature is projected in the used for thoroughgoing assessments, but few studies
models to warm by 1.4 to 5.8"C relative to 1990 levels have yet incorporated both climate and socioeconom-
and globally averaged sea level is projected to increase ic scenarios. Most studies of vulnerability of natural
by 0.09 to 0.88 mm by 2100. The ranges derive from systems to climate change follow projections of global
the use of different GCMs and emission scenarios. A change captured in a range of global models. Because
first round of impacts can be confidently predicted: global climate models (GCMs) are constantly being
glaciers and icecaps are projected to continue their refined, and depend on emission scenarios, which are
widespread retreat during the twenty-first century. In also revised intermittently, there is a lag, so the most
addition there will be changes in the variability of recent research on climate change impacts tends not to
climate, and also in the frequency and intensity of be related to the most recent developments in GCMs.
some extreme climate phenomena. These general The impacts of climate change will depend on the
features of climate change act on natural and human sensitivity of the system and to what extent it can
systems, and they set the context for the impact adapt, either through autonomous adjustment or as a
assessment, but they were not available for many of result of direct policy intervention. For this reason, the
the impact studies reviewed in the W G I1 assessment. IPCC conceptual framework, which has been adopted
This is important to note, because the new scenarios widely, differentiates increasingly between the sensi-
have increased temperature rise and much of the tivity of systems, their adaptive capacity, and their
available literature has not yet assessed climate change ultimate vulnerability. However, few studies have yet
impacts, adaptation, and vulnerability associated fully examined adaptive capacity. Climate change can
with the upper range of warming. Most work has have beneficial as well as adverse effects, but it is the
not examined transient change, either. Also, no one adverse effects that are expected to predominate in
knows precisely what will happen once complex much of the world - particularly the developing
changes start, particularly how they will interact countries.
and what feedback mechanisms may kick in - for Sensitivity, Adaptive Capacity, and Vulnerability
example, the release of methane and large-scale
carbon emissions from the die-back of the Amazonian Sensitivity is the degree to which a system is affected,
forest. either adversely or beneficially, by climate-related
The projected temperature and sea-level rises stimuli. Climate-related stimuli encompass all the
associated with a range of emissions, assuming elements of climate change, including mean climate
no policy interventions are made, is set out in characteristics, climate variability, and the frequency
Table 1. and magnitude of extremes. The effect may be direct
(e.g., a change in crop yield in response to a change
in the mean, range or variability of temperature) or
indirect (e.g., damage caused by an increase in the
Table 1 Projected temperature and sea-level rises
frequency of coastal flooding due to sea-level rise).
2025 2250 2100 Adaptive capacity is the ability of a system to adjust
to climate change (including climate variability and
COPconcentration 405-460 ppm 445-640 ppm 540-970 ppm
Global mean 0.4-1.1"C 0.8-2.6"C 1.4-5.8"C
extremes) or to moderate potential damage, to take
temperature advantage of opportunities, or to cope with the
change from consequences.
the year 1990 Vulnerability is the degree to which a system is
Global mean sea- 3-1 4 cm 5-32 cm 9-88 cm susceptible to, or unable to cope with, adverse
level rise from
the year 1990
effects of climate change, including variability and
extremes. Vulnerability is a function of the charac-
GLOBAL CHANGE / Human Impact of Climate Change 889
ter, magnitude, and rate of climate change and anticipated that standards of flood defense are also
variation to which a system is exposed, its sensitivity likely to rise in accordance with projected increases in
and its adaptive capacity. GDP per capita. Assuming no sea-level rise, the
number of people affected by coastal flooding can
Basically, all impacts will increase with temperature therefore be expected to rise marginally between 2020
rise - take the example of biodiversity for example, and 2080 from 10 million to 13 million. With
which serves as an indicator of the health of the planet.
unmitigated emissions and the same levels of sea-
Some highly vulnerable species may be at risk of
defense, however, this number increases by 81 million,
damage or even loss with very small changes, while
particularly as the areas have been growing relatively
greater losses will ensue from higher temperatures
to inland zones.
specifically threatening vulnerable ecosystems such as As many as 55 million of the 81 million are
the Arctic, the Cape Region of South Africa, and projected to be in southern Asia (along coasts from
tropical biodiversity ‘hotspots’. At higher tempera- Pakistan, through India, Sri Lanka, and Bangladesh to
tures coral bleaching will also increase, as will Burma). Over 20 million will be vulnerable in South-
disturbance to ecosystems caused by fire and insect East Asia (Thailand to Vietnam, including Indonesia
damage, and also the loss of coastal wetlands and and the Philippines). Substantial populations will also
shorelines. be affected in eastern Africa (South Africa to Sudan,
including Madagascar).
Impacts on Key Sectors and Places Proportionately, it is the island states which are
most vulnerable to unmitigated emissions. Poorly
Key systems and zones that sustain human societies are developed internal infrastructure and their dependency
sensitive to climate change. Those briefly examined on maritime connections have meant that key infra-
here are: agriculture, coasts, water resources, human structure is often concentrated along the coast (capital
health, and human settlement. cities, ports, and tourist facilities); furthermore, coast-
Agriculture
al areas often provide the most favored sites for new
settlement).
The response of crop yields to climate change varies In addition to sea-level rise, the mean number of
widely depending on the species, cultivar, soil condi- people who would be affected by coastal storm surges
tions, treatment of C 0 2 direct effects, and other increases severalfold if ghg emissions are not reduced.
locational factors. Higher C 0 2 levels can cause a
fertilizer effect. A few degrees of projected warming
Water Resources
are likely to be benign to temperate agricultural
systems, but with larger amounts of projected warm- Confidence in projections for the hydrological cycle
ing most temperate crop yield responses will become and its impacts in different parts of the world is at a
generally negative. lower level than confidence in projections relating to
Adverse impacts are anticipated most of all in the temperature increases, as there are variations between
tropics. An authoritative and innovatory set of studies climate scenarios. One medium-confidence projection
has been undertaken using the Hadley Center GCM. is that of increases in mean annual stream flow in high
Results show that if emissions increase without latitudes and in South-East Asia, with decreases in
reductions from policy interventions, by the 2080s Central Asia, the area around the Mediterranean,
cereal yields in Africa, the Middle East and, particu- southern Africa and Australia.
larly, India may be expected to decrease. While at a Agricultural use of water, mostly for irrigation,
global level, production and prices may be relatively forms the greatest part of water use, with almost 70%
unchanged, some regions, particularly Africa, will be of all withdrawals - domestic use is only about 10% of
adversely affected, experiencing marked reductions in all withdrawals. With higher temperatures and thus
yields; global assessments can therefore mask regional higher losses from evapotranspiration it can be
vulnerability, particularly when people cannot move expected that there will be increased demand for
from subsistence to market-based supply. With lower water for agriculture.
levels of yield and a fall in production, the risk of However, per capita consumption in developing
famine will be exacerbated. countries is declining, owing to population growth
and problems concerning accessibility.
Coasts
The amount of water available from rivers is
Around 20% of the world’s population live within projected to decline substantially in Australia, India,
30 km of the coast, and these communities are growing southern Africa, most of South America and Europe,
at twice the rate of the global trend. It can also be and the Middle East by the 2080s if emissions are not
Next Page
890 GLOBAL CHANGE I Human Impact of Climate Change
mitigated. Increases are expected in North America, most vulnerable groups will suffer the most: the old
Asia (particularly Central Asia), and central eastern and young, the poor, particularly those not on
Africa. Overall water resource stress (the ratio of adequate nutritional levels, and the infirm. Health
national water withdrawals to total national average impacts are expected generally to be adverse, notably
annual runoff, including water imported from up- in the developing world. Changes in the frequency and
stream countries) is predicted to worsen in many intensity of the extremes of cold, heat, flood and
regions, for example, northern Africa, the Middle drought will have knock-on effects on mortality,
East, and the Indian subcontinent. morbidity, and population displacement, and adverse
Water supply systems which rely on snow melt can effects on food production and on fresh water avail-
expect to change, with a shift in the timing of ability and quality, increasing the risks of infectious
streamflow from spring towards winter in many areas disease epidemics, particularly in developing coun-
associated with observed increases in temperature. In tries. Several major climate-related disasters have
some places where water supply is critical, such as the already had major adverse effects on human health -
Pacific Northwest of America, water management including floods in China, Bangladesh, Mozambique,
systems will need to change with reduced water supply and Europe, famines in Sudan, Southern Africa, and
availability in the summer. Glacier retreat - an Central Asia, and Hurricane Mitch. Major forest fires
observed phenomenon - also has implications for have been associated with an increase in respiratory
downstream river flows. Initially glacier melt will and lung symptoms. Although these events cannot be
increase during summer, but as the glaciers decrease attributed confidently to climate change, they indicate
summer flows will no longer be supported and will the susceptibility of vulnerable populations to the
decline below present levels. An increase in surface adverse effects of such events.
temperature is expected to contribute to a rise in the Disruption of populations and of economic activity
snowline in the western Himalayas, which will reduce will also have various health consequences. The effects
the capacity of the natural reservoir, and this, com- of floods can be immediate (drowning),medium-term
bined with any intensification of the monsoon, will (communicable diseases such as cholera, hepatitis A,
increase the risk of flooding in Nepal, Bangladesh, leptospirosis, diarrhoea, and respiratory infections)
Pakistan and North India during the wet season. and long-term (stress-related illnesses and psychiatric
Changes in the hydrological cycle are also expected problems). While these problems are experienced
to lead to the incidence of more droughts in some most in the developing world, they do constitute one
regions. However, with increased possibilities for of the major predictable climate change impacts in the
recharge during winter months, the effects on water developed world too.
resources need to be investigated region by region and, Overall, in temperate countries it is expected that
ideally, catchment by catchment. In Europe, for reduced winter deaths would outnumber increased
example, while most parts get wetter in the winter summer deaths. Increased temperature would proba-
season, in summer there is a strong gradient of change bly be associated with an increase in humidity,
between northern Europe (wetting of up to +2% per aggravating stress. Impacts in large cities would be
decade) and southern Europe (drying of up to -5% worse than in surrounding rural areas, particularly
per decade). because of the urban heat island effect. Deteriorating
The increased frequency of heavy rainfall events, air quality is expected in many large urban areas,
which have already been observed in some places, can assuming that current emission levels continue.
increase runoff and cause increases in flood magnitude Climate change may increase the concentration of
and frequency. In the USA and the UK, more precip- ground-level ozone. Conversely, reductions in fossil
itation is falling in large events than in earlier decades. fuel consumption and reduced particulate emissions
Intense precipitation can cause flood, landslide, ava- have been shown to have beneficial effects on human
lanche, and mudslide damage, as well as ongoing soil health. Changes are expected in vector-borne diseases
erosion. Flood magnitude and frequency are likely to such as mosquito-borne encephalitis, Lyme disease,
increase in most regions, and low flows are likely to and tick-borne encephalitis. For some diseases a net
diminish further. decrease may occur.
Malaria is the most notable vector-borne disease
whose incidence is likely to be significantly altered by
climate change; it is already one of the world’s most
Human Health serious and complex public health problems and is
Significant human health impacts are associated with undergoing a current resurgence. All models show net
the wide range of environmental disturbances brought increases in the potential transmission zone of malaria
about by climate change. It can be expected that the and changes in seasonal transmission at a global scale
GUST FRONTS 915
These are only a few of the many questions that emerge Bruhl C and Crutzen PJ (1988)Scenarios of possible changes
from the realization that the present-day upper in atmospheric temperatures and ozone concentrations
atmospheric structure may be changing. due to man’s activities, estimated with a one-dimensional
coupled photochemical climate model. Climate Dynam-
i c 2:
~ 173-203.
See also Ramanathan V (1988) The greenhouse theory of climate
change: a test by an inadvertent global experiment.
Carbon Dioxide. Chemistry of the Atmosphere: Prin-
Science 240: 293-299.
ciples of Chemical Change. Climate: Overview. Climate Rishbeth H and Roble RG (1992) Cooling of the upper
Variability: Decadal to Centennial Variability; Glacial,
atmosphere by enhanced greenhouse gases: modelling the
Interglacial Variations. Ionosphere. Mesosphere: Metal thermospheric and ionospheric effects. Planetary and
Layers; Polar Summer Mesopause. Noctilucent Clouds.
Space Science 40: 1011-1026.
Ozone: Ozone Depletion Potentials; Role in Climate. Roble RG (1993)‘Greenhouse cooling’ of the upper atmos-
Solar Terrestrial Interactions. phere. €OS 74: 92-93.
Roble RG and Dickinson RE (1989)How will changes in
carbon dioxide and methane modify the mean structure
Further Reading of the mesosphere and thermosphere. Geophysical Re-
Akmaev RA and Fomichev VI (1998) Cooling of the search Letters 16: 1441-1444.
mesosphere and lower thermosphere due to doubling of Thomas GE (1996)Global change in the mesosphere-lower
C 0 2 . Annales Geophysicae 16: 1501-1512. thermosphere region: has it already arrived? Journal of
Brasseur G and Hitchman MH (1988) Stratospheric re- Atmospheric and Terrestrial Physics 58(14): 1629-1656.
sponse to trace gas perturbations: changes in ozone and World Meteorological Organization (1998) Scientific As-
temperature distribution. Science 240: 634-637. sessment of Ozone Depletion: 2998, Global Ozone
Brasseur GP, Orlando JJ and Tyndall GS (1999)Atmospheric Research and Monitoring Project, Report No. 44.
Chemistry and Global Change. New York: Oxford Yung YL and DeMore WB (1999) Photochemistry of Plan-
University Press. etary Atmospheres. Oxford: Oxford University Press.
Theory; MIDDLE ATMOSPHERE: Gravity Waves
R Rotunno, National Center for Atmospheric Research, The basic phenomenon of cold air spreading
Boulder, CO, USA beneath a thunderstorm has been recognized in
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. writing since the early nineteenth century. However,
the term ‘gust front’ started to appear in the literature
only in the early 1960s. Before that time the phenom-
Introduction enon was variously referred to as the pressure jump
line, squall front, micro cold front, or outflow bound-
Evaporation of rain within a thunderstorm produces a ary, among other designations; the latter two are still
ground-based pool of cold air that spreads under its used frequently.
own weight and thereby generates gusty surface
winds. The leading edge of the spreading cold pool is
therefore known as the ‘gust front’. The gust front is an Physics
example of a ‘density current’ (seeDensity Currents),a Evaporation of Rain
phenomenon that occurs in a great variety of geo-
physical and engineering applications. As a character- Choosing the simplest case to illustrate the basic
istic feature of thunderstorms, the gust front plays a physics, imagine a vertically erect thunderstorm with
role in other aspects of thunderstorms, such as storm rain falling through it (Figure 1); since the air beneath
cell redevelopment (see Severe Storms). the thunderstorm is subsaturated, rain falling into it
916 GUST FRONTS
from the Clausius-Clapeyron equation, yields eqn [4]

A
where RHi is the initial subcloud relative humidity.
(L/cp) q v s (Ti)(1 - RHi)
T* - Ti = - 141
L Lqvs(Ti)
1+-----
cp R,Tf
With typical subcloud values of p = 900 hPa, Ti =
293K, and RHi = 0.7, eqn [4] gives Tf - Ti M
-3.5 K. While this is not untypical of observed values,
Ti - Ti N -lOK can be obtained in situations more
,.....
............'.
.....
..... .'. .....
............,,,
.....
.....
complex than that depicted in Figure 1.For example, if
,
.....
*. .'.
..,.....'.
...... .....
......., drier mid-level air is entrained into the thunderstorm,
s!.:.:......
,
)......... .....
RHi in eqn [4] can be much smaller, and so Tf - Ti
,.....
much lower. In any case, the basic idea expressed in
............'.
.'.........., ......,,
....a A\
.....
,.....
I.....'.....,
.'. .....
..... I eqn [4] is the same: Evaporation of rain produces
colder temperatures at low altitude in a thunderstorm.
0
.....I..
* X
Generation of Motion
Figure 1 Schematic diagram of a rain shaft below a thunder-
storm. Evaporative cooling and weight of rain produce negative Fluid motion is governed by Newton's second law,
buoyancy and hence negative circulation C on the circuit shown expressed as in eqn [5], where u is the velocity, p the
(adopting the convention that the line element I on the circuit points pressure, and p the density of the fluid mixture of dry
in the clockwise direction.)
air and water substance; -pg is the external body force
(per unit volume) due to the Earth's gravity, and d/dt is
may evaporate and thereby cool the subcloud layer. the substantial derivative.
Considering an isobaric process in which the rain
evaporates until the subcloud layer is saturated, the du -
_ - - - O1p - g
first law of thermodynamics gives eqn [l]. dt P
Since we are dealing with a fluid mixture of compress-
ible gas and liquid, we begin our development by
recasting the first term on the r.h.s. of eqn [5] as eqn [6].
Here Ti and qvi are the initial temperature and water
vapor mixing ratio (mass of water vapor per unit mass
1
--Vp=-cpO
+
1 E-Iqv
On
of dry air) of the subcloud layer, respectively, Ti is P 1 + 4 v + 41
final temperature at saturation, and qvs is the satura-
tion mixing ratio; the latent heat of vaporization = - C , e [ i + 0.614, - q,]v n [6]
L = 2 . 5 l~o 6J kg-' and cp = 1006J kg-I K-l is the
heat capacity for dry air at constant pressure (theeffect Here rc = ( p / p 0 0 ) ~(poo
~ / ~is~a reference pressure), the
of water vapor on the heat capacity has been ignored). potential temperature 8 = T v 2 , and 41 is the mixing
The saturation mixing ratio is related to the pressure p ratio of liquid water; the perfect gas law p = PdRdT +
and saturation vapor pressure ev(T)through eqn [2]. p,R,T and the fact that p = Pd + pv + p1 have been
used. The second part of eqn [ 6 ] is an approximation
based on the typical smallness of the water substance
mixing ratios.
For clarity of exposition, imagine that the subcloud
Here E Rd/R,, and Rd = 287 J kg-' K-' and R, = air shown in Figure 1 is initially motionless, although
462J kg-I K-' are the gas constants for dry air and rain has fallen into it; the gust front originates with the
water vapor, respectively. To obtain an explicit for-
mula for Ti, one can expand qysin a Taylor series as in
-
circulation C(= f u dl) created around the circuit
shown. From eqns [5] and [6] one can derive equation
eqn ~31. [7] for C
Using the second part of eqn [2] to calculate the Considering as a reference a rain-free (410 = 0)
derivative in eqn [3], and then substituting for de,/dT atmosphere with constant BVo (= 00 0.61q,o) in +
GUST FRONTS 917
hydrostatic balance (dnO/dz = -g/cp8,0), and ex-

pressing the dependent variables as the reference state
value plus a small deviation, eqn [7]may be approx-
imated as eqn [SI. High turbulence
Gust front
boundary
Cold air
The prime denotes the deviation from the reference
state, and terms involving products of water substance
mixing ratios have been neglected. The integrand of
\
eqn [SI is the buoyancy and is composed of three terms: High turbulence
The first term is the thermal buoyancy, which in the Figure 2 Schematic diagram of a gust front based on observa-
case under discussion is negative since 8’ M T’n;l and, tions. (Reproduced with permission from Simpson JE (1997)
as previously demonstrated, evaporative cooling pro- Gravity Currents in the Environment and the Laboratory, 2nd edn.
duces T’<O. The second term represents the contri- Cambridge: Cambridge University Press.)
bution to buoyancy by the presence of water vapor
which, being lighter than air, increases the buoyancy
potential temperature 80 spreading into an environ-
when 4.’ > 0 (as it is in the present case). (Negative
thermal buoyancy produced by evaporation of rain is ment with 8 = 81; g’ z g8i1(01- 6 0 ) is the reduced
numerically much greater than the offsetting effect on gravity.
positive buoyancy of the accompanying increase in
ath + a,h + h a,u =o [IO1
water vapor.) The third term represents the downward
force that liquid water exerts on the air in which it Using the method of characteristics, the solution of
resides. Although the negative buoyancy associated eqns [9] and [lo] for the so-called ‘dam-break’
with rainwater loading can be significant (e.g., the problem was obtained in the nineteenth century and
negative buoyancy associated with q1 = 10 g kg-I is is depicted in Figure 3A. These solutions are charac-
approximately equivalent to that associated with terized by a wave of depression moving into the initial
8’ M -3 K), in most cases thermal buoyancy dominates. reservoir and a constant-speed outflow with depth
The foregoing is simply a precise way of saying that falling to zero. Observations and laboratory tests of
cold, rainy air sinks and spreads out along the ground. the model predictions failed to show the predicted
Understanding exactly how it does so requires a model. time-varying parabolic shape near the leading edge,
Models of Motion
The simplest model is to consider the initial volume of
*
cold air as a rectangle (with unit breadth into the page)
that retains its rectangular shape as it spreads. The
model predicts that at late time the front speed
decreases as tr1I3, while the height decreases as
tr2I3. However, observations (Figure 2) indicate that
outflows from thunderstorms move with roughly
constant speed and height.
A more sophisticated model can be obtained by
considering the atmosphere as composed of two fluid Uf = k(g’hf)”’
hd 1 -t
layers of different potential temperature, and using the

hydrostatic approximation throughout (see Density
Currents). With these assumptions, and restricting (B)
xd Xf
attention to motion in the x - z plane, the full x -

Figure 3 Models of the gust front based on the ‘dam-break‘
momentum equation [5] may be reduced to the so-
called shallow-water momentum equation (eqn [9]).
problem with (A) h + 0 or (B) *= k m at x = x f . The dashed
line indicates the original position of the dense rain-cooled air; xd
and ho indicate, respectively, the position of the edge (Le., the
‘dam’) and the height of the rain-cooled air; hd is the height of the
outflow at the dam site. (Reproduced with permission from Klemp
The equation expressing conservation of mass J, Rotunno R, and Skamarock WC (1994) On the dynamics
becomes eqn [lo], where h and u are, respectively, of gravity currents in a channel. Journal of Fluid Mechanics
the depth and horizontal speed of the cold air with 269: 169-1 98.)
918 GUST FRONTS
but rather a zone of nearly constant height falling

abruptly to zero, as shown in Figure 2 and depicted in
Figure 3B. Recognizing this deficiency, researchers in -U
c
the early 1960s solved eqns [9] and [lo] with the d
condition of eqn [ 111 instead of h + 0 at the front.
uf= &k
a
1 1 t
H 00
P @I
p,=O
The constant k was left to be determined empirically. XS
For k 2 1, eqn [ l l ] produces solutions exhibiting a Figure 4 Moving at the constant density-current speed U, a
zone of constant state behind the leading edge, in control volume analysis of the mass and momentum equations can
which ufand hf are given by eqns [12], where ho is the be done to derive eqn [13],which relates U to H, d and
initial reservoir height (Figure 3B). g’(= go,’ (Or - eo)). (Reproduced with permission from Klemp J,
Rotunno R, and Skamarock WC (1 994) On the dynamics of gravity
currents in a channel. Journal of Fluid Mechanics 269: 169-198.)
Although these solutions have the observed features of Influence of the Gust Front
constant (uf, hi) behind the front, one has no way of
determining k from eqns [9] and [lo]. To make such a In addition to bringing relief from the heat of the day,
determination, one has to return to the full x-momen- the cold outflow from a thunderstorm displaces air in
tum equations [5]. its path upward (Figure 2) and so may regenerate a
The reason why the simple solution shown in Figure new thunderstorm cell. Experience and models have
3A is not realized in a two-layer fluid is that Kelvin- shown that the regenerated cells occur in a preferred
Helmholtz instability produces a nontrivial coupling compass direction if the prevailing wind increases with
between the cold pool and the environment into which height. So, for example, if the prevailing westerly wind
it spreads. Although the details of the instability and increases with height, the gust front from a thunder-
turbulence at the interface are hopelessly complicated, storm produces a new cell on its east side; a collection
one can deduce the relation between (uf and hi) from a of such cells in close proximity, all regenerating cells
simple control volume analysis. Taking as an observa- on their east side, will soon give the system of cells the
tional fact that a dense fluid of height H ( = hi) moves appearance of a line running north-south. The latter is
steadily at speed U(= uf)into an environment of less termed a squall line. Since the squall line by its nature
dense fluid in a channel of depth d (Figure 4),one lives much longer than its constituent cells, these
can deduce eqn [13] from the vertically integrated systems are the producers of copious rain, and
x-momentum equation along with mass conservation. frequently, severe weather such as flash floods, torna-
does and other high-wind phenomena.
(The restriction of the flow to the finite depth d is an

See also
attempt to account for the effect of a stable layer Density Currents. Hydraulic Flow. Kelvin-Helmholtz
located above the generally constant-0 air in the Instability. Severe Storms. Thermodynamics: Moist
subcloud layer.) The steady flow depicted in Figure 4, (Unsaturated) Air; Saturated Adiabatic Processes.
and described by eqn [13], is generally known as a
‘density current’ (or ‘gravity current’), and occurs in a Further Reading
wide range of geophysical and engineering problems
(see Density Currents). With the parameter k deter- Emanuel KA ( 1994) Atmospheric Convection. Oxford:
mined by eqn [ 13], eqn [ 121 now expresses the motion Oxford University Press.
in terms of the external parameters of the problem. Houze RA (1993) Cloud Dynamics. New York: Academic
Press.
The connection described here between the produc-
Klemp J, Rotunno R, and Skamarock WC (1997) On the
tion of buoyancy, circulation, and cold outflow con- dynamics of gravity currents in a channel. Journal of
stitute the minimal model for understanding the origin Fluid Mechanics 135: 169-198.
and nature of gust fronts. Neglected, but potentially Simpson JE (1997)Gravity Currents in the Environment and
important for quantitative prediction, are the effects the Laboratory, 2nd edn. Cambridge: Cambridge Uni-
of surface friction, environmental stratification, and versity Press.
time-dependence of the buoyancy source. Stoker JJ (1957) Water Waves. New York: Interscience.
GUST FRONTS 915
These are only a few of the many questions that emerge Bruhl C and Crutzen PJ (1988)Scenarios of possible changes
from the realization that the present-day upper in atmospheric temperatures and ozone concentrations
atmospheric structure may be changing. due to man’s activities, estimated with a one-dimensional
coupled photochemical climate model. Climate Dynam-
i c 2:
~ 173-203.
See also Ramanathan V (1988) The greenhouse theory of climate
change: a test by an inadvertent global experiment.
Carbon Dioxide. Chemistry of the Atmosphere: Prin-
Science 240: 293-299.
ciples of Chemical Change. Climate: Overview. Climate Rishbeth H and Roble RG (1992) Cooling of the upper
Variability: Decadal to Centennial Variability; Glacial,
atmosphere by enhanced greenhouse gases: modelling the
Interglacial Variations. Ionosphere. Mesosphere: Metal thermospheric and ionospheric effects. Planetary and
Layers; Polar Summer Mesopause. Noctilucent Clouds.
Space Science 40: 1011-1026.
Ozone: Ozone Depletion Potentials; Role in Climate. Roble RG (1993)‘Greenhouse cooling’ of the upper atmos-
Solar Terrestrial Interactions. phere. €OS 74: 92-93.
Roble RG and Dickinson RE (1989)How will changes in
carbon dioxide and methane modify the mean structure
Further Reading of the mesosphere and thermosphere. Geophysical Re-
Akmaev RA and Fomichev VI (1998) Cooling of the search Letters 16: 1441-1444.
mesosphere and lower thermosphere due to doubling of Thomas GE (1996)Global change in the mesosphere-lower
C 0 2 . Annales Geophysicae 16: 1501-1512. thermosphere region: has it already arrived? Journal of
Brasseur G and Hitchman MH (1988) Stratospheric re- Atmospheric and Terrestrial Physics 58(14): 1629-1656.
sponse to trace gas perturbations: changes in ozone and World Meteorological Organization (1998) Scientific As-
temperature distribution. Science 240: 634-637. sessment of Ozone Depletion: 2998, Global Ozone
Brasseur GP, Orlando JJ and Tyndall GS (1999)Atmospheric Research and Monitoring Project, Report No. 44.
Chemistry and Global Change. New York: Oxford Yung YL and DeMore WB (1999) Photochemistry of Plan-
University Press. etary Atmospheres. Oxford: Oxford University Press.
Theory; MIDDLE ATMOSPHERE: Gravity Waves
R Rotunno, National Center for Atmospheric Research, The basic phenomenon of cold air spreading
Boulder, CO, USA beneath a thunderstorm has been recognized in
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. writing since the early nineteenth century. However,
the term ‘gust front’ started to appear in the literature
only in the early 1960s. Before that time the phenom-
Introduction enon was variously referred to as the pressure jump
line, squall front, micro cold front, or outflow bound-
Evaporation of rain within a thunderstorm produces a ary, among other designations; the latter two are still
ground-based pool of cold air that spreads under its used frequently.
own weight and thereby generates gusty surface
winds. The leading edge of the spreading cold pool is
therefore known as the ‘gust front’. The gust front is an Physics
example of a ‘density current’ (seeDensity Currents),a Evaporation of Rain
phenomenon that occurs in a great variety of geo-
physical and engineering applications. As a character- Choosing the simplest case to illustrate the basic
istic feature of thunderstorms, the gust front plays a physics, imagine a vertically erect thunderstorm with
role in other aspects of thunderstorms, such as storm rain falling through it (Figure 1); since the air beneath
cell redevelopment (see Severe Storms). the thunderstorm is subsaturated, rain falling into it
916 GUST FRONTS
from the Clausius-Clapeyron equation, yields eqn [4]

A
where RHi is the initial subcloud relative humidity.
(L/cp) q v s (Ti)(1 - RHi)
T* - Ti = - 141
L Lqvs(Ti)
1+-----
cp R,Tf
With typical subcloud values of p = 900 hPa, Ti =
293K, and RHi = 0.7, eqn [4] gives Tf - Ti M
-3.5 K. While this is not untypical of observed values,
Ti - Ti N -lOK can be obtained in situations more
,.....
............'.
.....
..... .'. .....
............,,,
.....
.....
complex than that depicted in Figure 1.For example, if
,
.....
*. .'.
..,.....'.
...... .....
......., drier mid-level air is entrained into the thunderstorm,
s!.:.:......
,
)......... .....
RHi in eqn [4] can be much smaller, and so Tf - Ti
,.....
much lower. In any case, the basic idea expressed in
............'.
.'.........., ......,,
....a A\
.....
,.....
I.....'.....,
.'. .....
..... I eqn [4] is the same: Evaporation of rain produces
colder temperatures at low altitude in a thunderstorm.
0
.....I..
* X
Generation of Motion
Figure 1 Schematic diagram of a rain shaft below a thunder-
storm. Evaporative cooling and weight of rain produce negative Fluid motion is governed by Newton's second law,
buoyancy and hence negative circulation C on the circuit shown expressed as in eqn [5], where u is the velocity, p the
(adopting the convention that the line element I on the circuit points pressure, and p the density of the fluid mixture of dry
in the clockwise direction.)
air and water substance; -pg is the external body force
(per unit volume) due to the Earth's gravity, and d/dt is
may evaporate and thereby cool the subcloud layer. the substantial derivative.
Considering an isobaric process in which the rain
evaporates until the subcloud layer is saturated, the du -
_ - - - O1p - g
first law of thermodynamics gives eqn [l]. dt P
Since we are dealing with a fluid mixture of compress-
ible gas and liquid, we begin our development by
recasting the first term on the r.h.s. of eqn [5] as eqn [6].
Here Ti and qvi are the initial temperature and water
vapor mixing ratio (mass of water vapor per unit mass
1
--Vp=-cpO
+
1 E-Iqv
On
of dry air) of the subcloud layer, respectively, Ti is P 1 + 4 v + 41
final temperature at saturation, and qvs is the satura-
tion mixing ratio; the latent heat of vaporization = - C , e [ i + 0.614, - q,]v n [6]
L = 2 . 5 l~o 6J kg-' and cp = 1006J kg-I K-l is the
heat capacity for dry air at constant pressure (theeffect Here rc = ( p / p 0 0 ) ~(poo
~ / ~is~a reference pressure), the
of water vapor on the heat capacity has been ignored). potential temperature 8 = T v 2 , and 41 is the mixing
The saturation mixing ratio is related to the pressure p ratio of liquid water; the perfect gas law p = PdRdT +
and saturation vapor pressure ev(T)through eqn [2]. p,R,T and the fact that p = Pd + pv + p1 have been
used. The second part of eqn [ 6 ] is an approximation
based on the typical smallness of the water substance
mixing ratios.
For clarity of exposition, imagine that the subcloud
Here E Rd/R,, and Rd = 287 J kg-' K-' and R, = air shown in Figure 1 is initially motionless, although
462J kg-I K-' are the gas constants for dry air and rain has fallen into it; the gust front originates with the
water vapor, respectively. To obtain an explicit for-
mula for Ti, one can expand qysin a Taylor series as in
-
circulation C(= f u dl) created around the circuit
shown. From eqns [5] and [6] one can derive equation
eqn ~31. [7] for C
Using the second part of eqn [2] to calculate the Considering as a reference a rain-free (410 = 0)
derivative in eqn [3], and then substituting for de,/dT atmosphere with constant BVo (= 00 0.61q,o) in +
GUST FRONTS 917
hydrostatic balance (dnO/dz = -g/cp8,0), and ex-

pressing the dependent variables as the reference state
value plus a small deviation, eqn [7]may be approx-
imated as eqn [SI. High turbulence
Gust front
boundary
Cold air
The prime denotes the deviation from the reference
state, and terms involving products of water substance
mixing ratios have been neglected. The integrand of
\
eqn [SI is the buoyancy and is composed of three terms: High turbulence
The first term is the thermal buoyancy, which in the Figure 2 Schematic diagram of a gust front based on observa-
case under discussion is negative since 8’ M T’n;l and, tions. (Reproduced with permission from Simpson JE (1997)
as previously demonstrated, evaporative cooling pro- Gravity Currents in the Environment and the Laboratory, 2nd edn.
duces T’<O. The second term represents the contri- Cambridge: Cambridge University Press.)
bution to buoyancy by the presence of water vapor
which, being lighter than air, increases the buoyancy
potential temperature 80 spreading into an environ-
when 4.’ > 0 (as it is in the present case). (Negative
thermal buoyancy produced by evaporation of rain is ment with 8 = 81; g’ z g8i1(01- 6 0 ) is the reduced
numerically much greater than the offsetting effect on gravity.
positive buoyancy of the accompanying increase in
ath + a,h + h a,u =o [IO1
water vapor.) The third term represents the downward
force that liquid water exerts on the air in which it Using the method of characteristics, the solution of
resides. Although the negative buoyancy associated eqns [9] and [lo] for the so-called ‘dam-break’
with rainwater loading can be significant (e.g., the problem was obtained in the nineteenth century and
negative buoyancy associated with q1 = 10 g kg-I is is depicted in Figure 3A. These solutions are charac-
approximately equivalent to that associated with terized by a wave of depression moving into the initial
8’ M -3 K), in most cases thermal buoyancy dominates. reservoir and a constant-speed outflow with depth
The foregoing is simply a precise way of saying that falling to zero. Observations and laboratory tests of
cold, rainy air sinks and spreads out along the ground. the model predictions failed to show the predicted
Understanding exactly how it does so requires a model. time-varying parabolic shape near the leading edge,
Models of Motion
The simplest model is to consider the initial volume of
*
cold air as a rectangle (with unit breadth into the page)
that retains its rectangular shape as it spreads. The
model predicts that at late time the front speed
decreases as tr1I3, while the height decreases as
tr2I3. However, observations (Figure 2) indicate that
outflows from thunderstorms move with roughly
constant speed and height.
A more sophisticated model can be obtained by
considering the atmosphere as composed of two fluid Uf = k(g’hf)”’
hd 1 -t
layers of different potential temperature, and using the

hydrostatic approximation throughout (see Density
Currents). With these assumptions, and restricting (B)
xd Xf
attention to motion in the x - z plane, the full x -

Figure 3 Models of the gust front based on the ‘dam-break‘
momentum equation [5] may be reduced to the so-
called shallow-water momentum equation (eqn [9]).
problem with (A) h + 0 or (B) *= k m at x = x f . The dashed
line indicates the original position of the dense rain-cooled air; xd
and ho indicate, respectively, the position of the edge (Le., the
‘dam’) and the height of the rain-cooled air; hd is the height of the
outflow at the dam site. (Reproduced with permission from Klemp
The equation expressing conservation of mass J, Rotunno R, and Skamarock WC (1994) On the dynamics
becomes eqn [lo], where h and u are, respectively, of gravity currents in a channel. Journal of Fluid Mechanics
the depth and horizontal speed of the cold air with 269: 169-1 98.)
918 GUST FRONTS
but rather a zone of nearly constant height falling

abruptly to zero, as shown in Figure 2 and depicted in
Figure 3B. Recognizing this deficiency, researchers in -U
c
the early 1960s solved eqns [9] and [lo] with the d
condition of eqn [ 111 instead of h + 0 at the front.
uf= &k
a
1 1 t
H 00
P @I
p,=O
The constant k was left to be determined empirically. XS
For k 2 1, eqn [ l l ] produces solutions exhibiting a Figure 4 Moving at the constant density-current speed U, a
zone of constant state behind the leading edge, in control volume analysis of the mass and momentum equations can
which ufand hf are given by eqns [12], where ho is the be done to derive eqn [13],which relates U to H, d and
initial reservoir height (Figure 3B). g’(= go,’ (Or - eo)). (Reproduced with permission from Klemp J,
Rotunno R, and Skamarock WC (1 994) On the dynamics of gravity
currents in a channel. Journal of Fluid Mechanics 269: 169-198.)
Although these solutions have the observed features of Influence of the Gust Front
constant (uf, hi) behind the front, one has no way of
determining k from eqns [9] and [lo]. To make such a In addition to bringing relief from the heat of the day,
determination, one has to return to the full x-momen- the cold outflow from a thunderstorm displaces air in
tum equations [5]. its path upward (Figure 2) and so may regenerate a
The reason why the simple solution shown in Figure new thunderstorm cell. Experience and models have
3A is not realized in a two-layer fluid is that Kelvin- shown that the regenerated cells occur in a preferred
Helmholtz instability produces a nontrivial coupling compass direction if the prevailing wind increases with
between the cold pool and the environment into which height. So, for example, if the prevailing westerly wind
it spreads. Although the details of the instability and increases with height, the gust front from a thunder-
turbulence at the interface are hopelessly complicated, storm produces a new cell on its east side; a collection
one can deduce the relation between (uf and hi) from a of such cells in close proximity, all regenerating cells
simple control volume analysis. Taking as an observa- on their east side, will soon give the system of cells the
tional fact that a dense fluid of height H ( = hi) moves appearance of a line running north-south. The latter is
steadily at speed U(= uf)into an environment of less termed a squall line. Since the squall line by its nature
dense fluid in a channel of depth d (Figure 4),one lives much longer than its constituent cells, these
can deduce eqn [13] from the vertically integrated systems are the producers of copious rain, and
x-momentum equation along with mass conservation. frequently, severe weather such as flash floods, torna-
does and other high-wind phenomena.
(The restriction of the flow to the finite depth d is an

See also
attempt to account for the effect of a stable layer Density Currents. Hydraulic Flow. Kelvin-Helmholtz
located above the generally constant-0 air in the Instability. Severe Storms. Thermodynamics: Moist
subcloud layer.) The steady flow depicted in Figure 4, (Unsaturated) Air; Saturated Adiabatic Processes.
and described by eqn [13], is generally known as a
‘density current’ (or ‘gravity current’), and occurs in a Further Reading
wide range of geophysical and engineering problems
(see Density Currents). With the parameter k deter- Emanuel KA ( 1994) Atmospheric Convection. Oxford:
mined by eqn [ 13], eqn [ 121 now expresses the motion Oxford University Press.
in terms of the external parameters of the problem. Houze RA (1993) Cloud Dynamics. New York: Academic
Press.
The connection described here between the produc-
Klemp J, Rotunno R, and Skamarock WC (1997) On the
tion of buoyancy, circulation, and cold outflow con- dynamics of gravity currents in a channel. Journal of
stitute the minimal model for understanding the origin Fluid Mechanics 135: 169-198.
and nature of gust fronts. Neglected, but potentially Simpson JE (1997)Gravity Currents in the Environment and
important for quantitative prediction, are the effects the Laboratory, 2nd edn. Cambridge: Cambridge Uni-
of surface friction, environmental stratification, and versity Press.
time-dependence of the buoyancy source. Stoker JJ (1957) Water Waves. New York: Interscience.
HADLEY CIRCULATION 919
I N James, University of Reading, Reading, UK

Polar
Introduction Mid-latitude
westerlies
The so-called ‘Hadley circulation’ is perhaps the
earliest attempt to account for the global-scale distri- Hadley
bution of winds in the Earth’s atmosphere in terms of cell
basic physical processes. Halley in 1685 and Hadley in
1735 both proposed that the ‘Trade Winds’ that blow
toward the Equator at low latitudes could be under-
stood as the lower branch of an axially symmetric
convection cell driven by the temperature difference
between the Equator and poles of the Earth. Their
ideas were ahead of their time, especially as there was
then no prospect of determining winds at upper levels
of the atmosphere and thus verifying their hypothesis.
Figure 1 A schematic view of the mean circulation of the
When routine upper-air observations became avail- troposphere. The arrows on the globe show the winds near the
able in the mid-twentieth century, the ideas of Halley Earth’s surface. The circulations at the side show the zonal mean
and Hadley were essentially confirmed. Today, the circulation cells at various latitudes.
term ‘Hadley circulation’ refers to the thermally
driven meridional overturning motions in the low-
latitude troposphere. which temperature is plotted against specific entropy.
Figure 1 is the traditional, and somewhat The thermodynamic state of an air parcel - that is, its
oversimplified, schematic view of the global temperature, pressure, density and so on - are repre-
atmospheric circulation that will be found in many sented by a point on the thermodynamic diagram, and
elementary text books. It divides the Earth into a any change of its thermodynamic state by a curve on
set of climate zones, with the Trade Wind regime the diagram. The area under the process curve is
confined to the tropics. The Trade Winds are
simply the low-level part of the overturning ‘Hadley
circulations’, with ascent near the Equator, descent
in the subtropics and a poleward return flow at
upper levels. The more disturbed midlatitudes are
characterized by generally westerly winds, with
irregular growing and decaying eddies, the cyclonic
and anticyclonic weather systems generated by
baroclinic instabilities. When averaged around entire
latitude circles, this turbulent midlatitude flow
averages to a weak ‘Ferrel circulation’, in which
warmer air at lower latitudes sinks and colder air at
high latitude rises. There is some evidence of a very
weak ‘polar cell’ at high latitudes.
The energy that drives the Hadley circulation comes
from the conversion of heat energy to mechanical
energy in the tropical atmosphere: the Hadley circu-
lation is a prototypical example of a thermodynamic
‘heat engine’. Such heat ;engines are ultimately
0 5
respon- 1
5500 5600 5700 5800 5900 6000
sible for maintaining all motions in the atmosphere
Specific entropy (J kg-’ K-’)
against the dissipative effects of friction. The opera-
tion of the atmospheric heat engine is shown in Figure 2 A schematic thermodynamic diagram for the Hadley
Figure 2, which is a classic thermodynamic diagram in circulation.
920 HADLEY CIRCULATION
proportional to the heat energy entering an air parcel. scaled to have units of kg s - They may be thought of
The diagram also shows two different lines of constant as denoting the mass flux across a line from the edge of
atmospheric pressure, one near the Earth’s surface and the plot to that point. The most striking feature is the
one in the tropical upper troposphere. Near the strong rising motion near the Equator, and sinking
surface, air flows toward the Equator, along the motion at latitudes of about 25” Nand S, defining two
segment marked AB, gaining heat from the surface overturning cells, the ‘Hadley cells’, one in each
(this may be in the form of direct or sensible heat, or in hemisphere. However, the actual winds associated
the form of latent heat as water evaporates into the with these circulations are not particularly strong:
air). Near the Equator, it rises almost adiabatically they barely exceed 5 m s - l . The Hadley cells are
(that is, with little heat entering or leaving the air) thermally direct. Weaker, thermally indirect Ferrel
along the segment BC. It then moves poleward along cells are seen at higher latitudes. Considerably weaker
segment CA, cooling (that is, losing entropy) by thermally direct polar cells are seen at high latitudes.
emitting infrared radiation to space and descending. The diagram also reveals that there is a close
Eventually it returns to its original location A. During relationship between the westerly component of the
this cyclic process, more heat is added to the air along wind, shown by the shading, and the meridional flow.
AB than is removed along CA. The excess heat is The westerly component is much stronger, with values
converted to mechanical energy associated with the up to 40 m s - These maximum winds, the so-called
circulation of the tropical air. ‘subtropical jet’, are found in the upper troposphere,
The condition for such an energy conversion to take just where the circulations associated with the Hadley
place is that heat should on average be added at higher cells meet those associated with the Ferrel cells. There
pressure than it is removed. Equivalently, one can say is also a close relationship between the zonal winds
that air must rise on average when it is warmer, and and the temperature fields: they are linked, to a very
descend when it is cooler. A circulation with these good approximation by the therma wind relationship,
properties is called a ‘thermally direct circulation’. A which can be written as eqn [l].
thermally indirect circulation, in contrast, must be
driven by a source of mechanical energy; a refrigerator a p ] - R a[T]
cycle is an example of such a thermally indirect aP P f aY
circulation. In the schematic diagram of Figure 2, the
Hadley circulation is thermally direct, and therefore That is, a strong vertical wind shear is associated with
generates mechanical energy. In contrast, the Ferrel a strong poleward temperature gradient. In the deep
circulation of midlatitudes is thermally indirect and tropics where the Coriolis parameter f is small, this
consumes mechanical energy. relationship indicates that the temperature gradients
The observed annual mean meridional circulation is must be small, whatever the wind field. But in the
shown in Figure 3. The contours are parallel to the subtropics and midlatitudes, the increasing westerly
northward and upward winds averaged around lati- wind with height is associated with the fall of
tude circles and in time. The contour values have been temperature toward the poles.
Figure 3 The annual mean meridional streamfunction. Contour interval 2 x 10’kgs-’. Shading shows zonal winds greater than
20 m s-’. Based on an analysis of 20 years of ECMWF analyses.
The Held-Hou Model

An elegant model due to Held and Hou gives consid-
erable insight into the Hadley circulation and the
factors that determine its extent. Figure 4 illustrates
the model. The atmosphere is divided into two layers.
The lower layer is affected by friction at the ground,
and flow within it is supposed to be generally small.
Friction is effectively zero in the upper layer and so at
this level rings of air conserve their angular momen-
tum as they move poleward. Assuming that such rings
start at the Equator with zero motion relative to the
solid Earth, the wind at higher latitudes in the upper
layer is given by eqn [Z], where Q is the rotation rate of
the Earth, a is the radius of the Earth, and y is the
distance from the equator, proportional to the latitude.
U M = - Qy 2
U
Using the principle of thermal wind balance in the

form of eqn [3], the formula for UM can be used to Figure 5 Solution of the Held-Hou model.
predict the variation of temperature with latitude,
eM(Y). with latitude in the tropics but drops rapidly in the
au - ga aT subtropics and mid-latitudes. The radiative equilibri-
az 2 ~ ayy ~ [31 um temperature has a maximum at the Equator. The
temperature on the equator is set by requiring that
This is to be compared with the hypothetical ‘radiative there be no net heating of air parcels as they circulate,
equilibrium’ temperature distribution OE (y) of an at- that is, that the two shaded areas must be equal. The
mosphere that is not permitted to circulate. Where the poleward limit of the Hadley circulation is at the
actual temperature is less than radiative equilibrium latitude where these curves cross for the second time. A
there is net heating, and vice versa. In a steady state, this formula for the distance of the poleward edge of the
heating and cooling should exactly balance in the Hadley Hadley cell from the equator results from this solution
circulation and this requirement fixes the meridional (eqn ~41).
extent and strength of the Hadley circulation.
Figure 5 illustrates a graphical solution of the Held- [41
Hou model. The actual temperature varies very little
This formula suggests a value for Y of about 2500 km,
in remarkably good agreement with observations
considering the simplicity of the model.
cp The model can be elaborated. For example, the
vertical motions, proportional to the heating in the
regions of ascent and descent, can be estimated.
The model predicts a vertical circulation that is rather
weaker than that observed. The effect of latent heat
release in cumulonimbus clouds, which leads to inten-
sified but narrower regions of ascent, and broad regions
of descent, can be represented. But the basic physics,
which predicts that the Hadley circulations are con-
fined to 2500 km or so of the equator, remains relevant.
Seasonal Effects
The annual mean circulation shown in Figure 3 is in
Figure 4 The configuration of the Held-Hou model. fact the average of two quite different circulation
922 HADLEY CIRCULATION
Figure 6 The mean meridional circulation for (A) the Decemberdanuary-February season and (B) the June-July-August season.
Other details are as for Figure 3.
regimes that persist around the solstices. Figure 6 but at some latitude yo. As well as the latitude of the
shows the circulation for the mean Northern Hemi- northern and southern edges of the Hadley cells, the
sphere winter and summer seasons. In both cases, latitude yc of the streamline that divides circulation
there is a single strong thermally direct Hadley cell into the summer and winter hemispheres, and which is
with rising motion in the summer hemisphere and not the same as yo, must be determined. The algebra is
descent in the winter hemisphere. Weaker, thermally more complicated, but the steps in the argument are
indirect Ferrel cells are seen at middle latitudes in both just the same as for symmetric Hadley cell described in
hemispheres, but there is little or no sign of a Hadley the previous section.
cell in the summer hemisphere. Looking at the mean Figure 7 shows the results. For even small yo, the
circulation for shorter periods reveals that the transi- summer cell shrinks drastically and the winter cell
tion between a circulation like that of Figure 6A and intensifies. Almost all the circulation is associated with
one like that of Figure 6B is quite abrupt. At most ascent in the summer hemisphere and with descent in
times, there is just a single tropical Hadley cell whose the winter. The strength of the circulation is indicated
circulation links the two hemispheres: at some point in by the area between the temperature curves and the
the spring and autumn its direction of circulation radiation equilibrium curve. For yo of only 500 km,
switches abruptly as the temperature maximum the winter cell has intensified by a factor of about 10
crosses the Equator. compared to the symmetric case, while the summer
The Held-Hou model can be adapted to the cell has weakened by a similar factor. The winter cell is
situation where the heating is not symmetric about therefore some 100 times as intense as the summer cell.
the Equator. Assume that the maximum radiative Such a highly nonlinear response to the latitude of the
equilibrium temperature is no longer at the Equator, heating maximum means that the annual mean
A Lagrangian View
The diagrams of the meridional circulation shown so
far have all been based on so-called 'Eulerian aver-
ages'. That is, the winds have been averaged at fixed
points in space to produce the time-mean, zonal-mean
circulation. At all points in space, the winds and
temperatures fluctuate to some degree as weather
systems pass across the observing site. An alternative is
to follow individual elements of fluid as they move
around in the atmosphere, and average their proper-
ties to define a mean circulation. Such a mean is called
the 'Lagrangian mean', and in many ways is a much
preferable way to describe the circulation. For
example, the laws of physics applied to the atmosphere
all refer to the properties of discrete, identifiable lumps
of fluid. However, the Lagrangian mean is very
difficult to calculate in practice, not least because
individual elements of fluid rapidly become distorted
and eventually thoroughly mixed with neighboring
Figure 7 As Figure 5, but for a situation in which the heating
elements.
maximum is located away from the Equator. yc is the latitude of
maximum radiativeequilibrium temperature, yc is the latitudedividing An approximation to the Lagrangian meridional
the winter and summer Hadley circulation cells, yw and ys designate mean circulation can easily be calculated, and is
the limits of the winter and summer Hadley cells, respectively. shown in Figure 8. In constructing this diagram,
the wind data were averaged not on surfaces
meridional circulation is much more intense than the of constant pressure (as in Figures 2-7) but on
circulation derived from the annual mean heating. surfaces of constant 'potential temperature'. The
This is a particularly pointed example of the problem potential temperature of an air parcel generally
of 'nonlinear averaging', which is ubiquitous in the remains more or less constant for periods of less than
study of climate. This result also reconciles the weak a few days. It follows that surfaces of constant
circulations of the Held-Hou model with the stronger potential temperature move up and down in response
observed circulation: we should interpret the annual to the movement of the air. Averaging on potential
mean circulation as the average of the two solsticial temperature surfaces is equivalent, to the degree that
circulations, not as the response to the annual mean potential temperature is indeed conserved, to taking
thermal forcing. the Lagrangian average.
-200
h
2
5 -400
2
3
8
2 -600
L
-800
-1 000
90" S 60"s 30" S 0" 30" N 60" N 90" N
Latitude
Figure 8 The mean meridionalcirculationfor the December-January-February season, but with the data zonally averaged on surfaces
of constant potential temperature rather than on surfaces of constant pressure. Other details are as for Figure 3.
924 HAIL AND HAILSTORMS
Figure 8 differs dramatically from the correspond- See also

ing Eulerian mean circulation shown in Figure 6B. The Boundary Layers: Convective Boundary Layer. Coriolis
tropical Hadley cell is still present, but the mid-latitude, Force. General Circulation: Energy Cycle. Lagrangian
thermally indirect Ferrel is more or less eradicated. Dynamics.
Instead, a thermally direct circulation extends all the
way from the tropics to the pole in the winter
hemisphere. The original picture of the global circula- Further Reading
tion suggested by Halley and Hadley is largely vindi- Hadley G (1735) Concerning the cause of the general trade
cated if one views the circulation in Lagrangian terms. winds. Philosophical Transactionsof the Royal Society of
The thermally indirect Ferrel cell actually tran- London 39: 58-62.
sports heat against the temperature gradient, from Halley E (1686) An historical account of the trade winds,
high latitudes to low latitudes. At the same time, and monsoons, observable in the seas between and near
fluctuations in the flow (often termed ‘eddies’) more the tropicks, with an attempt to assign the physical cause
than compensate by transporting heat down the of the said winds. Transactions of the Royal Society of
temperature gradient, from low latitudes to high. London 16: 153-168.
In fact, the partitioning of the flow into mean and Held IM and Hou AY (1980) Nonlinear axially symmetric
circulations in a nearly inviscid atmosphere. Journal of
eddy parts is arbitrary. The Lagrangian circulation,
the Atmospheric Science 37: 515-533.
dominated by thermally direct circulations at nearly James IN (1995) Introduction to Circulating Atmospheres.
all latitudes, is a more natural and less arbitrary Cambridge: Cambridge University Press.
description. In the Lagrangian view, thermally direct Lindzen RS and Hou AY (1988) Hadley circulations for
Hadley circulations dominate the large-scale zonally averaged heating centred off the equator. Journal
tropospheric circulation. The Lagrangian view is of Atmospheric Sciences 45: 2417-2427.
remarkably like that originally suggested by Halley Peix6to JP and Oort AH (1992) Physics of Climate. New
and Hadley. York: American Physical Society.
C Knight and N Knight, National Center for usually within 5% of that of solid ice, 0.91 gcmP3.
Atmospheric Research, Boulder, CO, USA However, hail may be slushy, containing significant
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. amounts of liquid water, and, especially at small sizes,
the air content may be great enough that the hail is soft.
Soft hail is distinguished from graupel (accumulations
Introduction of rime on snow particles or small frozen water drops)
Hailstones are balls of ice that typically fall from only by size, and since nearly all hail falls through a
cumulonimbus clouds. By convention, they must be thick layer of air above the freezing point before
greater than 5 mm in diameter but their composition, reaching the ground, soft hail is often slushy, because
size, and shape are variable. The largest hailstones can of melting, when it falls. Much rainfall from cumulo-
have longest dimensions of 15cm or more. nimbus clouds in temperate climates is melted graupel
Hailstone amounts are also highly variable, but and small hail.
generally the largest hailstones and heaviest hailfalls
are from the most severe storms; that is, storms with Fundamental Concepts of Hail
the strongest updrafts, tallest cloud tops, and largest
size. Thus hail is correlated with tornadoes, and also
Formation
with lightning, though many storms produce lightning Hail forms by the accretion of water droplets onto ice
with no hail at the ground. Hail is not as well particles falling through supercooled cloud. The basic
correlated with flooding, which often results from elements needed for understanding the principles of
long-lasting and slow-moving precipitation systems hail formation are as follows.
that do not produce hail and may not have exception-
ally strong updrafts.
The Updraft and its Consequences
Hailstones include various amounts of air bubbles,
often in layers that indicate growth stages, but when Humid air rising in the cores of cumulus clouds cools
larger than about 2 cm in diameter their densities are as it rises. The cooling causes the condensation and
Figure 8 differs dramatically from the correspond- See also

ing Eulerian mean circulation shown in Figure 6B. The Boundary Layers: Convective Boundary Layer. Coriolis
tropical Hadley cell is still present, but the mid-latitude, Force. General Circulation: Energy Cycle. Lagrangian
thermally indirect Ferrel is more or less eradicated. Dynamics.
Instead, a thermally direct circulation extends all the
way from the tropics to the pole in the winter
hemisphere. The original picture of the global circula- Further Reading
tion suggested by Halley and Hadley is largely vindi- Hadley G (1735) Concerning the cause of the general trade
cated if one views the circulation in Lagrangian terms. winds. Philosophical Transactionsof the Royal Society of
The thermally indirect Ferrel cell actually tran- London 39: 58-62.
sports heat against the temperature gradient, from Halley E (1686) An historical account of the trade winds,
high latitudes to low latitudes. At the same time, and monsoons, observable in the seas between and near
fluctuations in the flow (often termed ‘eddies’) more the tropicks, with an attempt to assign the physical cause
than compensate by transporting heat down the of the said winds. Transactions of the Royal Society of
temperature gradient, from low latitudes to high. London 16: 153-168.
In fact, the partitioning of the flow into mean and Held IM and Hou AY (1980) Nonlinear axially symmetric
circulations in a nearly inviscid atmosphere. Journal of
eddy parts is arbitrary. The Lagrangian circulation,
the Atmospheric Science 37: 515-533.
dominated by thermally direct circulations at nearly James IN (1995) Introduction to Circulating Atmospheres.
all latitudes, is a more natural and less arbitrary Cambridge: Cambridge University Press.
description. In the Lagrangian view, thermally direct Lindzen RS and Hou AY (1988) Hadley circulations for
Hadley circulations dominate the large-scale zonally averaged heating centred off the equator. Journal
tropospheric circulation. The Lagrangian view is of Atmospheric Sciences 45: 2417-2427.
remarkably like that originally suggested by Halley Peix6to JP and Oort AH (1992) Physics of Climate. New
and Hadley. York: American Physical Society.
C Knight and N Knight, National Center for usually within 5% of that of solid ice, 0.91 gcmP3.
Atmospheric Research, Boulder, CO, USA However, hail may be slushy, containing significant
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. amounts of liquid water, and, especially at small sizes,
the air content may be great enough that the hail is soft.
Soft hail is distinguished from graupel (accumulations
Introduction of rime on snow particles or small frozen water drops)
Hailstones are balls of ice that typically fall from only by size, and since nearly all hail falls through a
cumulonimbus clouds. By convention, they must be thick layer of air above the freezing point before
greater than 5 mm in diameter but their composition, reaching the ground, soft hail is often slushy, because
size, and shape are variable. The largest hailstones can of melting, when it falls. Much rainfall from cumulo-
have longest dimensions of 15cm or more. nimbus clouds in temperate climates is melted graupel
Hailstone amounts are also highly variable, but and small hail.
generally the largest hailstones and heaviest hailfalls
are from the most severe storms; that is, storms with Fundamental Concepts of Hail
the strongest updrafts, tallest cloud tops, and largest
size. Thus hail is correlated with tornadoes, and also
Formation
with lightning, though many storms produce lightning Hail forms by the accretion of water droplets onto ice
with no hail at the ground. Hail is not as well particles falling through supercooled cloud. The basic
correlated with flooding, which often results from elements needed for understanding the principles of
long-lasting and slow-moving precipitation systems hail formation are as follows.
that do not produce hail and may not have exception-
ally strong updrafts.
The Updraft and its Consequences
Hailstones include various amounts of air bubbles,
often in layers that indicate growth stages, but when Humid air rising in the cores of cumulus clouds cools
larger than about 2 cm in diameter their densities are as it rises. The cooling causes the condensation and
HAIL AND HAILSTORMS 925
growth of water droplets to form the cloud, and the 70 -

droplets supercool when the temperature falls below
the freezing point. These water droplets are very small,
and rise in the updraft almost as fast as the air rises, 60 -
typically several tens of meters per second in hail-
storms, because their terminal fall speeds are less than 7 50--
10 cm s - An ice particle in such an updraft that is big E
v)
C, = 0.8
enough to have a higher fall speed collides with and v
b Sea level,
collects supercooled droplets, which freeze upon 'ti
-
40 -- C, = 0.55
impact and stick to it. This is the basic mechanism of
-9
-
hailstone growth, and the initiating particle may be a s
- 30--
snow crystal or snowflake or a frozen water drop. The .-F
other main role of the updraft in hail formation is to be E
strong enough and long-lasting enough to hold the E 20-
1
hailstones aloft, within supercooled cloud above the
freezing level, long enough for them to grow to their
final sizes. If they are to reach the ground as hail, they lo ?pi = 0.5
must be big enough not to melt on the way down. c, = 1.o
Terminal velocities of hailstones are described by 0 1
0 2 4 6 8 10
Diameter (cm)
Figure 1 Terminal fall velocities of hailstones plotted against

diameter, for the drag coefficients (CD) indicated, calculated for a
pressure of 500 hPa (about 6 km above sea level) and a hailstone
In eqn [ l ] , where CD is the drag coefficient, a density of 0 . 9 g ~ m - except
~, where otherwise indicated. (From
dimensionless quantity that expresses how the drag Knight CA and Knight NC (2001). In: Severe Convective Storms.
force (the air resistance) relates to the velocity and the Boston: American Meteorological Society, by permission of AMS.)
fluid properties. The densities of ice and air are
indicated by p ; g is the acceleration due to gravity; 100 and 1000 per cubic centimeter. Hailstones grow
and D is the diameter of the hailstone. Numerical within such clouds.
values are plotted in Figure 1 for different values of CD The droplets freeze when they collide with ice,
and, with CD = 0.55, for both sea level and 500 hPa releasing latent heat ( 334 J g - l ) . However, since the
N
pressure. A pressure of 5OOhPa corresponds very specific heat of water is 4.2 J g - K - a drop super-
roughly to -10°C and 6 k m above sea level, with cooled to -10°C is warmed to 0°C when only about
considerable variability depending upon local condi- 1/8 of it is frozen. If the remaining 7/8 is to freeze, the
tions. The large range of values for CD comes about rest of the latent heat must be absorbed by the
because of the highly variable shapes of hailstones, environment, which in this case is the surrounding
which influence fall speed considerably. Hailstone air, at -10°C. Growing hailstones are thus warmer
diameter is usually defined as the diameter of a sphere than the surrounding air, and in extreme cases their
of equivalent mass. A hailstone growing within an temperature may be 0°C and they may grow as a
updraft may ascend or descend depending upon mixture of ice and water, called spongy hail.
whether its terminal fall speed is less or greater than Hailstone layering consists of shells with varying air
the updraft speed. bubble contents, which arise because of air tempera-
ture variations (changes in growth altitude) or cloud
The Thermodynamics of HailstoneGrowth
water content variations, which cause changes in the
Water remains in the liquid state when cooled below hailstone temperature.
0°C unless it contains a particle of a solid material that
is an ice nucleus, with the property of initiating ice.
Hail Falling Behavior, Shape, and Growth
The content of ice nuclei in the atmosphere is variable,
but as a rough average there might be one that acts at Since hail grows by accreting and freezing water
-20°C per liter of air and only one per many cubic droplets, it grows mostly on its underside where the
meters active at -5°C. Thus clouds that are not too collisions occur. Small hailstones, falling at only
much colder than 0°C are often composed almost ',
10-15 m s - usually maintain a single falling orien-
entirely of supercooled water droplets, since the tation. This often leads to a roughly conical shape,
droplet populations in clouds vary between about with the growth 'center' on top and consecutive,
broadening layers beneath that constitute most of the

bulk. When hailstones grow larger, the higher fall
speed creates more turbulence and they tumble. Often
their growth shapes are then flattened, most of the
collisions occurring around a perimeter because of a
rapid and symmetrical, but complicated, tumbling
motion. Figure 2 shows two sections through such a
hailstone, illustrating the conical center that grew into
a rather thick, somewhat elongated disk. The tum-
bling has important consequences for shape and also
for terminal fall speed and heat exchange, and it may
aid in shedding any unfrozen water that otherwise
might accumulate.
Hailstones larger than a few centimeters in diameter
are often lumpy. When the hail grows wet with liquid
water, lumps may form like icicles, from water flowing
over the surface. Hailstones of this kind are seen in
Figure 3. The lumpy shapes often suggest aggregation
Figure 3 Four hailstones greater than 5 cm in diameter that show
of smaller hail, but sections through lumpy hail never prominent projections. These form as the hail grows while wet, with
have shown this to be the case. The lumpiness makes liquid water flowing over the hailstone surface, and grow in a
the tumbling more chaotic and the shapes more manner similar to the formation of icicles.
complex, and also influences the fall speed by increas-
ing the drag coefficient.
The rate of hailstone growth is expressed as eqn [2], E = 1for the case of simple sweep-out - the collection
where D is the diameter, W is the cloud water content in of all water droplets in the volume of air traversed by
mass per unit volume, and E is the collection efficiency. the hailstone. In reality, E is usually between roughly
0.3 and 0.8 depending upon the sizes of both the
droplets and the hailstone, because some droplets in
the path of the hailstone are carried around it by the
airflow. Figure 4 shows the typical range of values of
c ! 0.3
::
g 0.2
0.1
0 1
Figure 2 Two slices through a large, oblate hailstone: (A) 0 2 4 6 a
perpendicular to the short axis of the stone and (B)parallel to the Diameter (cm)
short and the long axes. The photographs were taken with a bright
background, so clear ice appears white and ice with small air ’,
Figure 4 Hailstone growth velocities, in cm min - as a function
bubbles is darker. The hailstone grew in a constant falling of diameter calculated for the three top curves in Figure 1,
orientation and developed a conical shape up to 2-3cm in length. assuming an effective cloud liquid water content WE of 2.5 g m -3
After that the growth was fastest around the perimeter of the and no shedding of liquid water. Cloud water contents vary from
flattened hailstone, owing to a rapid, and complicated but zero to perhaps twice this value, and growth rates vary accordingly
symmetrical tumbling motion. Note the growth layers (shown by (see eqn [2]). (Reproduced with permission from Knight CA and
differing air bubble content) that signify changes in the growth Knight NC (2001) In: Severe Convective Storms. Boston: American
environment. Meteorological Society.)
HAIL AND HAILSTORMS 927
growth rates as a function of hail size, for WE =

2.5 gmP3 and pi = 0.9gcmP3.
Hailstorms
The necessary and sufficient conditions for hail
formation in a thunderstorm are easy to state in
general terms. The storm needs to provide a strong
updraft containing supercooled water droplets as the
environment for hailstone growth, and into this
environment must come hailstone 'embryos'. These
are particles of ice with terminal fall speeds of a few
meters per second that can grow in a few tens of
minutes into hailstones. Their growth must be fairly Environmental
rapid because the growth environment may disappear
as the updraft weakens (the updrafts in most thunder-
storms are intermittent), or the supercooled water
content of the updraft may become depleted, either by
evaporation as the updraft air mixes with drier
surrounding air or by conversion to many small ice
particles. If the ascent rate of the potential hail embryo
is too great, it may be elevated out of the growth region low levels
before it grows big enough to fall as hail; if the ascent 4)
rate is too slow, it may fall out before it reaches large I I I I I
enough size. The growth region extends from 0 to
- 40°C, the temperature at which all the supercooled
5'15
water droplets freeze spontaneously; a distance of 5 to v)
z
6 km within cloud. E
The formation of hail, then, depends upon a critical s.10
E0,
interrelation between the three-dimensional air flow .-
within storms, the fate of the supercooled droplets that 2 5
occur within the updrafts, and the trajectories of the
potential hail embryos. These factors work together to 0 10 20 30
determine how much of the water vapor that con- (B) km
denses reevaporates in the middle and upper atmos-
Figure 5 Schematic of airflow and some hydrometeor trajecto-
phere, how much of it falls as rain, and how much as
ries deduced in a supercell storm. (A) Plan view, looking down at
hail. The percentage of it that falls as hail is difficult to the storm, showing low-level inflow entering from the south (the
measure, but is usually very small. bottom of the drawing), the main updraft within the storm shown by
the dotted circle, upper-level environmental air flow from the west
and storm outflow to the east. The area within the cloud containing
Hail in Supercell and Multicell Storms: Steady and precipitation-sized hydrometeors is hatched. (B) Cross-section
Time-Dependent Concepts south to north parallel to the inflow, with the upper level flow away
from the reader. The trajectory indicated by open circles in both
There are at least two specific hypotheses for hail views represents direct growth in the main updraft: condensation,
formation that refer to two idealized storm types: the freezing, and some riming leading to small ice particles that travel
supercell storm and the multicell storm. The supercell more or less with the air and out into the upper-level anvil. The
is a type of storm that can last a long time and travel trajectory indicated by the solid circles indicates particles that start
considerable distances, often producing tornadoes in weaker upward flow on the south side of the updraft, grow to
larger sizes and higher fall velocities and fall as small hail or rain to
and often producing long swaths of large, damaging the north of the location of the main updraft. The trajectories
hailstones. The diagrams in Figure 5 represent a indicated by the numbers 1, 2, 3 represent large hail formation:
typical case in the central United States, a horizontal (1) slow growth rising at the west side of the updraft, (2) descent
cross-section view looking down at the storm and a back into the inflow while traveling in the environmentalflow around
the south side of the updraft, and (3) final growth as hail within the
vertical cross-section view, both indicating possible
strong updraft and fallout at its northern edge. (Adapted with
growth trajectories. The humid air entering the storm permission from Browning KA (1977) In: Foote GB and Knight CA
comes in a vigorous flow at the surface from the south (eds.) Hail: A Review of Hail Science and Hail Suppression.
or south-east, rises up through the middle of the storm, Boston: American Meteorological Society.)
and exits to the east because the flow in the upper levels the hailstone size while increasing the number. If the
is from the west. That carries the southern part of the size is reduced enough, most or all of the hail may then
outflow in the upper levels over the top of the inflowing melt before reaching the ground. Ideally, the hail
air at the surface. The updraft in the central part of the would be suppressed and the rainfall increased. There
storm can be 30-50 m s - so small ice particles that has been a great deal of discussion about optimizing
grow within the updraft do not have time to grow very the seeding materials and the timing and location of
large and attain much fall speed before being ejected seeding, but the knowledge of the natural evolution of
into the outflow at the top of the storm, producing the the ice contents within hailstorms is still rudimentary,
extensive anvils characteristic of supercells. However, so it may be many years before a consensus is reached
in the outflow there is little vertical air motion, and the on the prospects of hail suppression.
idea is that some of these particles fall back into the
inflow at low levels. They may become hail embryos,
getting a chance to grow further while rising a second
Hail Climatology
time within the updraft. Falling faster with respect to
the air, they ascend more slowly and have time to grow Hailfall at the ground is a small-scale phenomenon
into fully fledged hailstones. generally affecting areas of one to a few tens of
This is a particularly simple and organized square kilometers, though much larger hailfalls
recycling of precipitation particles to form hail have been documented. It is highly variable and
embryos in a storm: simplified because most supercells poorly resolved by routine weather observation
are not especially at steady state but evolve in various networks. Many small hailfalls are missed altogether.
ways. Thus data on hail climatology are statistical in
Multicell storms, on the other hand, are composed nature and much of the information derives from
of individual convective cells that grow and decay in insurance records.
proximity to one another, sometimes in a systematic Probably the most general climatological correla-
way. The individual cells may last 45-60 min, as their tion for hailstorms is that they often occur at the lee
updrafts increase up to a maximum and then die out. sides of mountain ranges. In South America, the
This can be enough time to produce significant hail United States, and Canada there is a strong tendency
and, if the initial updraft is not too strong, embryos for maximum hail frequencies to occur in the lee of the
may have time to grow and attain a fall speed of several Andes and the Rocky Mountains - locations also of
meters per second without being elevated too high in exceptionally high thunderstorm frequency. In terms
the cloud. Now as the updraft strengthens the embryos of hail damage to crops, however, areas of moderate
may already be there, in the right locations and ready hail frequency may be more important if they contain
to form hailstones. crops that are exceptionally sensitive to hail damage
Multicell hailstorms typically produce spotty, dis- and have high value per unit area, such as wine grapes,
continuous hailfalls, which may be organized within a tobacco, tea, and some fruits and vegetables.
larger hailswath if the multicell is an organized one or Hail becomes rarer toward the poles, simply
may be distributed irregularly. because the frequency of convective storms decreases
with increasing latitude. Hail is also relatively rare in
the tropics and in maritime regions, but the reasons for
this are not as clear-cut. Part of the reason may be a
Hail Suppression thunderstorm frequency effect, but not all. Rain forms
Hailfall is a widespread phenomenon in temperate more readily in clouds in the tropics than in temperate
zones around the world. In some areas it represents a climates because higher humidity and warmer cloud
major loss to agricultural production, and in recent bases provide on the average much more liquid water
years the insurance losses in large, urban areas have in the clouds below the freezing level. Thus rain forms
increased greatly, mostly from hail damage to roofs by the coalescence of water drops without involving
and automobiles. the ice phase, and the fallout of this rain decreases the
Suppression of hail by seeding clouds with artificial amount of supercooled water higher in the clouds
ice nuclei has been practiced in many parts of the where the temperature is appropriate for hailstone
world for several decades. It is still widespread, growth. Likewise, rain forms more easily in maritime
although it is controversial and there has been no cumulus because maritime air contains fewer nuclei
definitive demonstration of positive effects. The main for condensation, which reduces the droplet concen-
idea is that furnishing ice nuclei might increase the trations. A reduced droplet concentration increases
number of potential hail embryos in the clouds, droplet sizes, leading again to a faster formation and
thereby depleting the supercooled water and reducing fallout of rain and a decreasing likelihood of hail.
HAMILTONIANDYNAMICS 929
Hail Detection by Radar behavior, especially if techniques are developed to

provide information about hail size and amount
Radar methods for remote sensing of hail within
within storms. Radar sensing of hail, if developed to
clouds have recently come into use for research but not
be sufficiently quantitative and used routinely, would
yet for the operational radar networks. These rely on
also contribute greatly to hail climatology and hail
varying the polarization of the transmitted pulses and
research.
sensing different polarizations of the backscattered
radiation. Horizontal and vertical, linear polariza-
tions are often used, and sometimes circular polariza-
See also
tion. Intense radar echoes may be caused either by
heavy rain or by hail, and until recently their Cloud Microphysics. Radar: Precipitation Radar. Se-
interpretation has been ambiguous. However, big vere Storms. Snow (Surface).Weather Modification:
raindrops are consistently flattened, with larger hor- Inadvertant. Weather Prediction: Severe Weather Fore-
izontal than vertical dimensions, and this produces a casting.
substantially stronger echo with a horizontally than
with a vertically polarized radar. Hail gives a more Further Reading
nearly equal echo strength at the two polarizations
because small hail is not flattened and larger hail Doswell CC 111 (ed.) (2001) Severe Local Storms. Meteoro-
tumbles so that its elongations are about equally logical Monographs vol. 28, No. 50. Boston: American
Meteorological Society.
distributed it space. Thus the difference, termed the Foote GB and Knight CA (eds)(1977)Hail: A Review of Hail
differential reflectivity, is used as a hail signal when the Science and Hail Suppression. Meteorological Mono-
radar echo itself is intense. graphs vol. 16, no. 38. Boston: American Meteorological
Widespread application of this and other advanced Society.
radar techniques will greatly increase the knowledge Rogers RR and Yau MK (1989) A Short Course in Cloud
of hail production as related to storm structure and Physics, 3rd edn. Oxford: Pergamon Press.
See OPTICS, ATMOSPHERIC: Optical Phenomena
T G Shepherd, University of Toronto, Toronto, Ontario, meteorology consists of the fundamental laws of
Canada physics, including Newton’s second law. For many
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. purposes, diabatic and viscous processes can be
neglected and the equations are then conservative.
(For example, in idealized modeling studies, dissipa-
tion is often only present for numerical reasons and is
Introduction kept as small as possible.) In such cases dynamical
Hamiltonian dynamics describes the evolution of meteorology obeys Hamiltonian dynamics. Even
conservative physical systems. Originally developed when nonconservative processes are not negligible,
as a generalization of Newtonian mechanics, describ- it often turns out that separate analysis of the
ing gravitationally driven motion from the simple conservative dynamics, which fully describes the
pendulum to celestial mechanics, it also applies to such nonlinear interactions, is essential for an understand-
diverse areas of physics as quantum mechanics, ing of the complete system, and the Hamiltonian
quantum field theory, statistical mechanics, electro- description can play a useful role in this respect.
magnetism, and optics - in short, to any physical Energy budgets and momentum transfer by waves are
system for which dissipation is negligible. Dynamical but two examples.

research.
See also
930 HAMILTONIANDYNAMICS
Hamiltonian dynamics is often associated with equation (eqns [ l ] ) .

conservation of energy, but it is in fact much more
than that. Hamiltonian dynamical systems possess dqi - dpi - -E ( i = 1 , . . . ,N) [l]
a mathematical structure that ensures some remark- dt api’ dt aqi
able properties, Perhaps the most important is the X(q,p) is the Hamiltonian function, q = (41, . . . , q N )
connection between symmetries and conservation are the generalized coordinates, and p (el,. . . ,p ~ )
laws known as Noether’s theorem. Well-known the generalized momenta. For so-called natural sys-
examples are the fact that conservation of energy
is linked to symmetry in time, and conservation
+
tems with X = ( /pI2/2m ) U(q), where m is the mass
and U the potential energy, eqns [I] immediately lead
of momentum to symmetry in space. Less well-known to eqn [2], which is Newton’s second law for a
is the fact that material conservation of potential conservative system.
vorticity, so crucial to the theory of dynamical
meteorology, is also connected to a symmetry by
Noether’s theorem, but to a symmetry that is invisible
in the Eulerian formulation of the governing equa-
tions. It turns out that one can exploit the underlying Conservation of energy follows directly from eqns [l],
Hamiltonian structure of a system through the rele- for any IH, by the chain rule (repeated indices are
vant conservation laws even if the explicit form of that summed):
structure is not known, which is useful for applica-
tions. As is shown in detail below, symmetry-based
conservation laws provide a general theory of avail-
able potential energy, and show why it is that Rossby
waves carry negative zonal momentum, thereby [31
explaining both the maintenance of the westerlies
and the stratospheric Brewer-Dobson circulation.
Such laws also provide a powerful way of deriving
stability criteria. Symplectic Formulation
Dynamical meteorologists use a variety of theoret- The theory of canonical transformations suggests that
ical models, ranging from the fully compressible there is nothing special about the qs and p s , and
equations through the hydrostatic primitive, Boussin- Hamilton’s equations [l] can be written in the so-
esq, and quasi-geostrophic equations to the barotropic called symplectic form, eqn [4].
equations. With such a zoo of models, it is crucial to
know the extent to which theories developed
for one model carry over to another. Hamiltonian
dynamics provides this unifying framework. All the
models just mentioned are in fact Hamiltonian, In eqn [4],u = (41,. . . , qN,p1,. . . , p ~ a)n d l is given
and models can be grouped into families according by eqn [5],where I is the N x N identity matrix.
to their Hamiltonian structure. In this way it
becomes immediately apparent, for example, that
the Charney-Stern stability theorem for baroclinic
quasi-geostrophic flow is the counterpart to Rayleigh’s J has certain mathematical properties, including skew-
inflection-point theorem for barotropic flow, and symmetry. More generally, one can take those prop-
that an analogous stability theorem will exist for
erties to be the definition of Hamiltonian structure,
any balanced model having a similar Hamiltonian with] not necessarily of the form of eqn [5]. The skew-
structure, no matter what the definition is of the symmetry of J nevertheless guarantees energy conser-
potential vorticity. Thus, it is precisely through its
vation (eqn [6]).
abstract character that Hamiltonian dynamics has
many powerful applications in theoretical dynamical
meteorology.
There is an important distinction between systems

with a nonsingular (or invertible) J , which can always
Canonicaland NoncanonicalDynamics be transformed into the canonical form of eqn [5],and
In classical mechanics, canonical Hamiltonian dy- those with a singular (or noninvertible) J . The latter,
namical systems are those described by Hamilton’s known as noncanonical systems, possess a special class
of invariant functions known as Casimir invariants. underpins many of its applications to dynamical
These are the solutions of eqn [7] (for canonical meteorology.
systems the solutions are just constants). Casimir invariants are special because 6cu = 0. This
suggests that they correspond to invisible symmetries.
J--=O
ac ( i = 1,. . . For example, in rigid-body dynamics the total angular
'I aui momentum is a conserved quantity in any description
of the motion. In the original canonical description it
That they are necessarily conserved in time then
corresponds to the rotational symmetry of the dynam-
follows from the skew-symmetry ofJ (eqn [SI).
ics, but in Euler's equations, where angles have been
eliminated, it enters as a Casimir because the under-
lying physical symmetry is no longer explicit.
The best-known example of a noncanonical Ham-

iltonian system is Euler's equations for rigid-body Barotropic Dynamics
dynamics. Having an odd number of evolution equa- In what sense are the models of dynamical meteorol-
tions (three in this case), the system is necessarily ogy Hamiltonian? Consider what is probably the
noncanonical because any skew-symmetric matrix of simplest such model, the barotropic vorticity equation
odd dimension must be singular. There is one Casimir (eqn [ 1I]),which describes two-dimensional, nondi-
invariant for Euler's equations, the total angular vergent flow.
momentum.
Noether's Theorem
For a canonical system, if a particular generalized -
Here o ( x ,y, t) = z (V x v) = Q2$ is the vorticity, z is
coordinate q, does not appear in the Hamiltonian, then the unit vector in the vertical direction, v(x, y , t) =
the Hamiltonian is invariant under changes in that i x V$ is the horizontal velocity, $(x,y , t ) is the
coordinate; in other words, there is a coordinate streamfunction, and a( f ,g) fxgr - frgx is the two-
symmetry. Translational and rotational symmetries dimensional Jacobian. The candidate Hamiltonian is
are common examples. Hamilton's equations [ 11then the conserved energy of this system, which is just the
immediately imply that the corresponding generalized kinetic energy. The obvious dynamical variable is the
momentum is conserved: dpjldt = 0. vorticity. In order to cast eqn [ll]in the form of
This connection between symmetries and conserva- eqn [4], we need to regard every point (x,y) in space as
tion laws has a more general and far more power- indexing a degree of freedom analogous to the index i;
ful form. Given a function 3 ( u ) , define 6 ~ u= j the sum over i then becomes an integral over space,
cjij(aF/auj), where E is an infinitesimal parameter; dFu functions become functionals, and partial derivatives
is called the infinitesimal variation in u generated by 3. become functional or variational derivatives. Thus we
(In the canonical case, 6T.u is an infinitesimal canonical write eqn [12].
transformation.) It then follows that the infinitesimal
variation in Fl generated by F is given by eqn [9].
= // V$ . 6V$ dx dy
On the other hand, the time evolution of 3 is given by P P
eqn [lo]. = ]] {V - ($SV$) - $So}dx dy [12]

d 3 w d u j a 3 ax
-- _ _ = -Ijj- 1101 Assuming for now that the boundary terms vanish, we
dt -sui dt auj au, identify the variational derivative as SE/6o = -$.
Using the skew-symmetry of J , eqns [9] and [lo] then The need to integrate by parts reflects the fact that the
imply that 6 ~ 7 -=l 0 if and only if d F / d t = 0. This effect of a vorticity perturbation on the kinetic energy
connects symmetries and conservation laws: the density is nonlocal; thus, partial derivatives at fixed
Hamiltonian is invariant under the variation generat- points in space make no sense and variational deriv-
ed by 3 (i.e., that variation represents a symmetry of atives are essential. Equation [ l l ] can now be cast in
the Hamiltonian) if and only if 3 is a conserved Hamiltonian form as eqn [ 131.
quantity. This result, known as Noether's theorem, is am
-= J -
ax where J = -a(o,.) [13]
one of the central results of Hamiltonian dynamics and at 60
932 HAMILTONIAN DYNAMICS
Note that J is now a differential operator rather than a about the origin.
matrix. It is evidently skew-symmetric: SSfJg dx dy =
- J’gJf dx dy (under suitable boundary conditions)
for arbitrary functions f , g.
M=-//i?udxdy=
// f . ( r x v ) d x d y [18]
Conservation Laws The discussion has so far neglected any contribution

from boundary terms. They are easily included. In the
The form of J in eqn [13] is clearly singular: any presence of rigid lateral boundaries, for a complete
function of w inserted in the argument gives zero. mathematical specification of the problem, eqn [ l l ]
These then represent Casimir invariants of the system: must be supplemented with the conditions [19] on
functionals of the form [14], where C(.)is an arbitrary each connected portion of the boundary.
differentiable function, evidently satisfy
J ( S C / S u ) = 0.
v - n = 0, d t {v.dr =0
Here n is the outward-pointing normal, and s is

the vector arc length along the boundary. The second
The fact that such functionals are conserved in time of eqns [ 191 represents conservation of circulation,
corresponds to the material conservation of vorticity which follows from the underlying momentum
expressed by eqn [ 113. equations but must be included as a separate
To identify the momentum invariants, we need to condition in the vorticity formulation of the dynamics.
apply Noether’s theorem to the various spatial Although the circulation integrals along each
symmetries. Suppose that the domain is unbounded, connected portion of the boundary are constants in
with decay conditions at infinity, so that there is time, they are independent dynamical variables
symmetry in all directions. The variation in w corre- and are needed to determine v from w. The
sponding to a translation by SX in the coordinate x is Hamiltonian formulation of eqn [13] may easily
given by 6w = -(au/ax)6x. Setting E = 6x, we then be extended to include the circulation integrals in
need to solve for the momentum invariant M accord- addition to w as dynamical variables. The Casimir
ing to eqn [15]. invariants then include functions of these circulation
integrals. With regard to the momentum invariants,
of course, the rigid boundaries must respect the
same symmetries; a zonal channel flow with walls
To within the addition of a Casimir, the solution of eqn at constant y breaks the translational symmetry
[ 151 is given by SM/Sw = y. Hence we may choose M in y and the rotational symmetry, leaving only the
as in eqn [16], where v = (u,v ) . zonal impulse of eqn [16] as an invariant. The
final equality of eqn [16] is then no longer strictly
true, but the impulse and momentum differ only by
terms involving the circulations along the channel
walls, which are Casimirs. Since symmetry-based
= /J u dx dy
invariants are only defined to within a Casimir in
any case, the impulse and momentum are essentially
equivalent.
The first, elementary form of M given by eqn [16] is A simplified model of barotropic dynamics is the
known as Kelvin’s impulse. It represents the y ‘center- point-vortex model, where the vorticity is concentrat-
of-mass’ of the vorticity distribution, and is in many ed in Dirac delta functions. The point-vortex model
ways the preferable form because it is local in 0.The has been used to study two-dimensional turbulence
final form, however, shows that the invariant M and certain kinds of atmospheric flow structures. It
corresponding to symmetry in x is ultimately just the also turns out to be Hamiltonian, and is in fact a
x-momentum, as expected. canonical system: the Casimirs are built into the model
The same argument applied to translation in the as parameters through the choice of the point-vortex
coordinate y yields eqn [171. strengths.
Other Balanced Models

Finally, rotational symmetry leads to eqn [18], where The barotropic vorticity equation has a mathematical
r (x, y) and r = lrl, which is the angular momentum structure that is analogous to that of many models of
HAMILTONIAN DYNAMICS 933
balanced, or potential-vorticity-driven, flow (see Dy- the equation for the interior potential vorticity (the
namic Meteorology: Balanced Flows) and the results latter being eqn [Ill, with q in place of w , applied at
derived above extend in an obvious way to such every value of z; thus, the advection of q remains
systems. Inclusion of the beta effect means simply a purely horizontal). The Eady model is an extreme case
change from w to the potential vorticity q = w By. + where the interior potential vorticity is uniform and
Since 6 y = 0 (recalling that the coordinate y is like an the flow is driven entirely by the temperature distri-
index), 6q = 6w and eqns [11]-[14] go through butions on the upper and lower boundaries; the
unchanged with q in place of w. However the beta dynamical structures driven from each boundary are
effect breaks translational symmetry in y and rota- known as Eady edge waves. Since these temperature
tional symmetry, leaving only the translational sym- distributions also evolve according to eqn [ l l ] , with
metry in x represented by the zonal impulse invariant the QG temperature IC/, in place of w, it is not
of eqn [16]. Strictly speaking the latter should be surprising that the same kind of Hamiltonian structure
written with q in place of w , but the integrals differ by a also applies to this model. The energy is given by eqn
constant and so represent the same invariant. Inclu- [201.
sion of topography is no more difficult; one simply
includes an additional topographic term h( x ,y ) in the
definition of q. This will generally break all spatial
symmetries, leaving only the energy 'FI and Casimirs C
In eqn [20], the reference-state density p,(z) and
as invariants. This illustrates a general and important
stratification function S(z) = N 2 / f 2 are both pre-
point, namely, that symmetry-based invariants are
scribed, with N ( z ) the buoyancy frequency and f the
fragile: a slight change in the conditions of the problem
Coriolis parameter, and where V is still just the
destroys their conservation properties. In contrast, the
horizontal gradient operator. With the potential
energy and the Casimirs are robust invariants (robust
vorticity given by eqn [21], where f and P are
within the conservative context, of course) that
constants, eqn [22] follows.
survive such perturbations.
Stratification is most easily introduced in the con-
text of the quasi-geostrophic ( Q G )model (see Quasi-
geostrophic Theory). Layered QG models are com-
pletely trivial extensions of the barotropic system:
their evolution is determined by the potential vorticity
qi(x,y , t )in each layer i, governed by eqn [ I 11with qi in
place of w , together with conservation of circulation
along any rigid lateral boundaries that may be present.
These are then the dynamical variables. The energy
now includes available potential as well as kinetic [221
energy, but, apart from some geometric factors repre-
This is like eqn [12], but with an additional term
senting the layer depths, one still recovers 6'FI/6qi =
involving the temperature variations 6$z at the upper
- $ i in each layer as well as eqn [13] with qi in place of
and lower boundaries. Including these as independent
w . The various invariants follow in the obvious way
dynamical variables, in addition to q (and possibly
with the spatial integrals summed over the different
also circulation terms), the governing equations can be
layers. The same considerations, incidentally, apply to
cast in the symplectic form of eqn [13]. The Casimirs
layered non-QG 'intermediate' models that still have
the form of eqn [I11 - namely, nondivergent now involve integrals of arbitrary functions of the
horizontal advection of the potential vorticity qi temperature on the upper and lower boundaries, in
within each layer, with the flow in each layer driven addition to integrals of arbitrary functions of potential
vorticity in the interior (eqn [23]).
by the potential vorticity in all layers (as described by
the particular definition of qi).
With continuous stratification and with upper and
lower boundaries (at z = 1and z = 0, say), there is an
additional effect. It is well known that the temperature
distribution along the upper and lower boundaries is
equivalent to potential vorticity (see Baroclinic Insta-
bility), and independent evolution equations for these
temperature distributions are required to fully specify The momentum invariants similarly extend in
the continuously stratified QG system, in addition to obvious ways: for example, the zonal impulse
invariant is given by eqn [24]. The dynamical variables are v and h, for which eqns
[27] hold.
63.t 63.t 1
- = hv,
6V
= 31Vl2 + gh 1271
Note that no integration by parts is necessary in this

The semi-geostrophic (SG) model (see Frontogene- case; this is characteristic of velocity-based represen-
sis) is widely used in mesoscale dynamics because of its tations of the dynamics. It can easily be verified that
ability to represent realistic frontal structures. It turns eqns [25] may be cast in the symplectic form (aulat) =
out that the SG model can also be cast in the form of J(63.t/6u) with u = ( u ,u, h ) with J given by eqn [28],
eqn [ 111, and hence in the symplectic form of eqn [131, where q = -
(f + i V x v)/h is the potential vorticity of
provided the equations are written in isentropic- the shallow-water system.
geostrophic coordinates. However, in these coordi-
0 4
nates rigid boundaries appear to move in time. The SG
equations, in contrast to the QG equations, make no I=(:;
-ax
,: -2) PI
geometrical distinction between horizontal and verti-
cal boundaries - this is why they are also useful for the The matrix (28) is evidently skew-symmetric; the
study of coastal dynamics in physical oceanography - signs on the derivative terms are indeed correct,
and the same kind of independent dynamical degrees of since first-order differential operators are themselves
freedom encountered in the QG system on upper and skew-symmetric, as with the J in eqn [13]. The
lower boundaries also appear on lateral boundaries. In zonal (absolute) momentum invariant is given as
the special case of channel walls, these degrees of expected by eqn [29], for which it is easy to verify
freedom correspond to coastal Kelvin waves and are that J(6M/6u)= -au/ax in line with Noether’s
analogous in some respects to the Eady edge waves theorem, and the other momentum invariants follow
represented by both the QG and SG systems. They similarly.
must be taken into account in the variational calcula-
tions, and enter into many of the resulting expressions.
M = //h(u-fy)dxdy ~ 9 1
Unbalanced Models The Casimirs are given by eqn [30] for arbitrary
Balanced models are controlled by the advection of functions C( .).
potential vorticity (perhaps augmented by the advec-
tion of isentropic surfaces on rigid boundaries), so
for such models it is natural to seek a Hamiltonian
C= // hC(q)dx dy [301
description analogous to eqn [131. However, models Thus, potential vorticity still plays a crucial role in the
that include a representation of gravity waves or other Hamiltonian description of the dynamics. Special
high-frequency oscillations, called unbalanced mod- cases of Casimirs are total mass (C = 1) and total
els, do not fit within this framework. They necessarily circulation (C = q ) .
have additional degrees of freedom. For such models, a Stratification is easily incorporated. The hydrostatic
description in terms of the velocity field is a more primitive equations can be cast in Hamiltonian form
natural way to reflect the Hamiltonian structure. For isomorphic to that of eqn [28] when expressed in
example, the rotating shallow-water equations [25] isentropic coordinates. Even the fully compressible
withv(x, y, t ) = ( u ,v ) the horizontalvelocity, h ( x ,y, t ) stratified Euler equations, which form the most
the fluid depth, g the gravitational acceleration, and general system imaginable for (dry) dynamical mete-
with constant f , conserve the energy (eqn [ 2 6 ] ) . orology, can be cast in an analogous form, although
there are now additional dynamical variables associ-
av + ( f i + V x V) x v + V
- ated with compressibility. The Casimirs are in this case
at given by eqn [31], where p(x, y, z , t ) is the density,
+
d ( x , y, z , t ) is the potential temperature, and q = [ ( f z
ah
-+V.(hv)
at
=o -
V x v) V d ] / p is the Ertel potential vorticity, with v
and V now acting in all three space dimensions.
3.t = // +
{h/vI2 gh2} dx dy 1261
The invariance of the Casimirs is of course evident but not positive definite. One would like both prop-
directly from the dynamical equations (eqn [32]) and erties in order to define normal modes, spectra, etc.
reflects the material invariance of q and 8. Another problem, at first sight unrelated, arises with
momentum. The momentum of a wave would appear
84 ae to be zero (the average of a sinusoid is zero), yet waves
-+v+
at
= 0, -+v.ve
at
= 0:
can certainly transfer momentum; this is what drives
the quasi-biennial oscillation in the tropical strato-
a- +P V . ( p v ) = 0 sphere, for example (see Middle Atmosphere: Quasi-
at
1321 Biennial Oscillation). How is one to describe this wave
momentum?
The fully compressible stratified Euler equations In canonical Hamiltonian mechanics, the distur-
are, in fact, a straightforward expression of Newton's bance energy about an equilibrium is always quadrat-
second law, without constraints such as hydrostatic ic; from this one assesses stability and defines normal
balance, provided they are expressed in Lagrangian modes. There is no ambiguity. So why are things not
coordinates (see Lagrangian Dynamics). In Lag- equally clear for fluid dynamics? The answer lies in the
rangian coordinates, the dynamical variables are the noncanonical Hamiltonian structure of virtually every
positions and momenta of fluid elements, which are fluid dynamical system in the Eulerian representation.
natural canonical variables. The thermodynamic fields If u = U is a steady solution of a Hamiltonian system,
can be expressed in terms of these variables: p can be then eqn [33] holds.
written in terms of the Jacobian of particle positions
(which describes the compression of the fluid), while
can just be chosen as one of the Lagrangian coordi-
nates. In this way, the fully compressible stratified
Euler equations represent a canonical Hamiltonian For a canonical system, the invertibility of J then
system. But there are six dynamical variables in the implies that 63-1/6u = 0 at u = U. This means that U is
Lagrangian description, compared with only five in a conditional extremum of 3-1, and X[u] - X[U] is
the Eulerian description; in transforming to Eulerian quadratic in the disturbance. However, for a non-
coordinates, a reduction of the phase space takes canonical system none of this follows and the distur-
place. This is where the potential vorticity comes in. In bance energy is generally linear in the disturbance.
Lagrangian coordinates, the potential vorticity is still
Pseudoenergy
materially conserved; but what symmetry does it
correspond to? The answer is a particle-relabeling Hamiltonian structure provides the solution to this
symmetry: if one rearranges fluid elements while quandary. Equation [33] is locally the same as the
preserving the same Eulerian fields, then the dynamics equation defining the Casimirs, which means that
is unchanged. There is just enough freedom to do this, 63-1/6u is locally parallel to 6C/6u for some C (a
because there is one more Lagrangian than Eulerian different C for each choice of U). In other words, there
variable. Upon reduction to the Eulerian description, exists a Casimir C such that eqn [34] holds.
this additional degree of freedom disappears, and the
particle-relabeling symmetry becomes invisible. That
is why potential vorticity conservation then appears in
the form of a Casimir invariant.
Now, both 3-1 and C are invariants, and the com-
bined invariant 3-1 + C satisfies the extrema1 condition
Disturbance Invariants d(3-1 + C) = 0 at u = U. We have thus constructed
what we wanted, namely a disturbance quantity that is
Probably the most powerful application of Hamilto-
conserved and is locally quadratic in the disturbance
nian dynamics to dynamical meteorology arises in the
context of studying the properties of disturbances to
basic states. In fluid dynamics, the question of how to
define the energy of a wave has often been a point of
+ +
A = (3-1 C)[U] - (3-1 C)[U] WI
confusion if not contention. For example, in the case of This quantity is known as the pseudoenergy. Provided
a basic flow, if the wave energy is defined as the energy one has a complete set of Casimirs, eqn [34] can
in the frame of reference moving with the basic flow, always be solved for a Hamiltonian system and the
then it is positive definite but not conserved. On the pseudoenergy can always be constructed according to
other hand, if it is defined as the difference energy eqn [35]. This is one of the great attractions of
relative to the basic-flow energy, then it is conserved Hamiltonian dynamics: it provides systematic recipes
in abstract terms, which can be worked out for any This is self-evidently positive definite for dpo/dz < 0
particular application. and has the small-amplitude quadratic approximation
A particularly illuminating application is the subject [421.
of available potential energy, highly useful in energy
budget analyses. Me demonstrate the method in the
case of the three-dimensional stratified Boussinesq 1421
equations. The energy is given by eqn [36].
Equation [41] is the exact, finite-amplitude expression
for the available potential energy (see General Circu-
lation: Energy Cycle) of disturbances to a stably
stratified, resting basic state p o ( z ) ,while eqn [42] is its
Here ps is the constant reference-state density, and the more familiar small-amplitude counterpart, widely
dynamical variables are v and p , for which eqns [37] used in the theory of internal gravity waves (see
hold. Buoyancy and Buoyancy Waves: Theory). Similar
constructions can be performed to define the available
potential energy of any stratified fluid system.
Although the small-amplitude expression of eqn
The term pgz in eqn [36] is the gravitational potential [42] appears to be singular in regions where
energy, and is linear in the dynamical variables. dpo/dz = 0 , the finite-amplitude expression of eqn
Now consider disturbances to a stably stratified, [41] remains perfectly well-defined in such regions.
resting basic state v = 0, p = po(z). Although the Pseudomomentum
Casimirs of this system include functions of the
potential vorticity, because the basic state is at rest, The same kind of reasoning can be applied for
67-t/6v = 0 at v = 0 and this dependence is unneces- disturbances to zonally symmetric (x-invariant) basic
sary, so we may consider Casimirs of the form of states, assuming that the underlying system possesses
eqn [38]. the same symmetry. For such states, with aU/ax = 0,
Noether’s theorem implies that the zonal impulse or
C= /// C ( p )dx dy dz with
6C
- = C’(p)
6P
[38] momentum invariant satisfies eqn [43].
Equation [34] then leads to the condition

C’(po) = -gz. This is the defining relation for the
function C(.).Thus, one has to express gz in terms of But just as with eqn [33], there is a Casimir C such that
the same argument po. This can be done by inverting 6(M + C) = 0 at u = U; with this C, one may imme-
the functional dependence p o ( z ) to obtain Z ( p o ) , diately construct the invariant [44], which is quadratic
where Z ( p o ( z ) )= z. This is always possible provided to leading order in the disturbance.
p o ( z ) is monotonic, which is the case for a stably
stratified basic state. One then has eqn [39].
A = (M + C)[U] - ( M + C)[U] P41
This quantity is known as the pseudomomentum.
C b ) = - / p g z ( P ) dP [391 We calculate the pseudomomentum for the case of
barotropic flow on the beta-plane. Suppose we are
From this the pseudoenergy of eqn [35] takes the form given a monotonic basic state q o ( y ) . From eqns [14]
and [16], with q in place of o,we have eqn [45].
1401.
The extrema1 condition 6(M + C) = 0 at q = qo then

leads to C’(q0) = -y. This is now isomorphic to the
construction of the available potential energy, replac-
The first term in the spatial integrand is the kinetic ing gz with y and p with q. If we define the function
energy and is positive definite; the last two terms can Y (.) by Y (qo(y)) = y, then evidently eqn [46] holds.
be rewritten as in eqn [41].
1411
The small-amplitude approximation to the spatial to leading order in the disturbance quantities. In fact,
integrand is given by eqn [471. their quadratic approximations are exactly conserved
by the linearized dynamics. (The quadratic approxi-
mation to the pseudoenergy is the Hamiltonian of the
linearized dynamics.) When either of these quantities
is sign-definite for a given basic state, it follows that
Equations [46] and [47] are evidently negative definite that basic state is stable to normal-mode instabilities.
for dqo/dy > 0, which is the case when qo is dominated Indeed, in order to reconcile exponentially growing
by by. These rather peculiar expressions have no disturbances with conservation of pseudoenergy and
obvious relation to zonal momentum at first sight, but pseudomomentum, the latter quantities must vanish
they nevertheless explain why it is that Rossby waves for such disturbances. This fact provides a useful
always exert an eastward (positive) force when they constraint on the structure of normal-mode instabi-
leave a source region, and a westward (negative) force lities, as well as a powerful unifying framework
when they dissipate and deposit their momentum in a between different models.
sink region: they carry negative pseudomomentum. This simple framework accounts for virtually every
The general nature of the derivation ensures that known stability theorem in dynamical meteorology.
exactly the same expressions hold for any balanced For resting, stratified basic states in unbalanced
model having the basic form of eqn [13]. If the basic models, with pseudoenergy like eqn [40] for the
state qo is chosen to be the zonal mean 4, then the zonal Boussinesq model, the condition of positive definite
mean of eqn [47] becomes eqn [48], where q’ = q - 4. pseudoenergy is the statement of static stability (see
Static Stability). For basic flows in axisymmetric or
symmetric stratified unbalanced models, the same
condition is the statement of symmetric stability (see
Instability: Symmetric Stability), which reduces to
In the case of stratified QG dynamics, the negative of Rayleigh’s centrifugal stability theorem in the special
eqn [48] is known as the Eliassen-Palm wave activity case of axisymmetric homogeneous flow (see Instabil-
(see Middle Atmosphere: Zonal Mean Climatology), ity: Inertial Instability). These stability theorems are
which has been widely used in dynamical meteorology all quite analogous t o static stability. A different
to assess the effect of Rossby waves on the zonal mean situation arises for balanced models. There, the
flow. It is such an effective diagnostic precisely because pseudoenergy can take either sign depending on the
it represents negative pseudomomentum. Moreover, basic flow. The positive-definite and negative-definite
and importantly, its use is not restricted to waves. The cases correspond respectively to Arnol’d’s first and
exact, finite-amplitude expression of eqn [46] ensures second stability theorems. (They are analogous to the
that the concept of pseudomomentum applies to fully stability of a rigid body rotating about an axis of
nonlinear, even turbulent disturbances. symmetry corresponding respectively to a maximum
The robust negative definiteness of the pseudomo- or minimum moment of inertia.) In the special case of a
mentum of balanced disturbances explains a great deal parallel basic flow, Arnol’d’s first theorem states that
about the general circulation of the atmosphere. the flow is stable if uo/(dqo/dy)<O, which is the
Propagation of synoptic-scale Rossby waves away FjGrtoft-Pedlosky theorem.
from their source region in the baroclinic storm tracks With regard to pseudomomentum for balanced
implies an eastward force in the storm track regions, models, eqn 1461 is sign-definite whenever dqoldy is
accounting for the maintenance of the westerlies sign-definite. For barotropic flow with q = w , this
(see General Circulation: Momentum Budget). corresponds to Rayleigh’s inflection-point theorem; on
The westward momentum deposition associated the beta-plane with q = co -k by, to the Rayleigh-Kuo
with breaking planetary-scale Rossby waves in the theorem; and for stratified QG flow with q given either
stratosphere drives the poleward Brewer-Dobson by its multilevel forms qr or by eqn [21] in the
circulation (see Middle Atmosphere: Zonal Mean continuously stratified case, to the Charney-Stern
Climatology), which is responsible for the observed theorem. For stratified QG dynamics in the presence of
distribution of ozone and other chemical species in the a lower boundary, the second terms of eqns [23] and
stratosphere. [24] become relevant and there is an additional
contribution to the pseudomomentum involving the
temperature distribution on the lower boundary; it is
Stability Theorems isomorphic to the interior eqns [46]-[48], replacing q
The pseudoenergy and pseudomomentum are, by their with I),. Since the climatological temperature gradient
construction, conserved quantities that are quadratic along the Earth’s surface is towards the Equator, the
pseudomomentum associated with surface distur- pseudoenergy d of eqn [40].

bances is generally positive. In this case the Charney-
Stern stability criterion is not satisfied for observed
flows; on the other hand, normal-mode instabilities are
generally required to involve both temperature distur-
bances on the lower boundary and potential-vorticity
disturbances in the interior, in order to create a
disturbance with zero total pseudomomentum. The With the choice 6 = &&E, eqn [ 5 3 ] establishes
Charney model of baroclinic instability (seeBaroclinic Liapunov stability in the norm defined by eqn [52].
Instability) is the best-known example of this. In the Conservation of pseudoenergy is clearly central to the
presence of an upper boundary, there is a further proof.
contribution to the pseudomomentum, with opposite The finite-amplitude stability of stably stratified flow
sign to the lower contribution in accord with eqn [24]. is not too surprising; it corresponds to physical intui-
Thus in the Eady model of baroclinic instability, where tion, and indeed motivates the very concept of available
the potential vorticity is uniform and the interior potential energy, which has a long pedigree. What is
contribution to the pseudomomentum disappears, the perhaps more surprising is that exactly the same kinds
instability can arise from the interaction of distur- of constructions can be made for all of the stability
bances on the upper and lower boundaries that theorems mentioned above, and for virtually any model
together add up to zero total pseudomomentum. within the same family. They can also be used to obtain
These statements all concern normal-mode stability. rigorous upper bounds on the saturation of normal-
But what can be said about stability goes much further mode instabilities, by considering the initial unstable
than this. The existence of finite-amplitude distur- flow (plus infinitesimal disturbance) to be a finite-
bance invariants suggests the possibility of nonlinear, amplitude disturbance to a stable basic state.
or Liapunov stability: namely, that small disturbances
stay small for all time, where small is defined in terms See also
of some disturbance norm. Mathematically, we say
that a basic state U is Liapunov stable to disturbances Baroclinic Instability. Buoyancy and Buoyancy
u’ in a given norm l\u’IIif for all E > 0 there exists a Waves: Theory. Dynamic Meteorology: Balanced
6 ( ~ >) 0 such that eqns [49] hold. Flows; Potential Vorticity. Frontogenesis. General Cir-
culation: Energy Cycle; Momentum Budget. Instability:
Ilu’(O)il<S +~Iu’(~)II<E for all t [49] Inertial Instability; Symmetric Stability. Lagrangian Dy-
namics. Middle Atmosphere: Quasi-Biennial Oscilla-
Let us see how this applies to static stability for tion; Zonal Mean Climatology. Quasi-geostrophic
the Boussinesq model considered earlier. Suppose that Theory. Static Stability.
the basic state has dpo/dz< 0 and that furthermore the
basic-state density gradients are bounded according to
[50] for some constants G I , c2.
Further Reading
Arnol’d VI (1989) Mathematical Methods of Classical
Mechanics, 2nd edn. New York: Springer-Verlag.
Benjamin TB (1984) Impulse, flow force and variational
principles. IMAJournal ofApplied Mathematics 32: 3-68.
Then eqn [41] for the available potential energy is Holm DD, Marsden JE, Ratiu T and Weinstein A (198.5)
bounded from above and below according to eqn [5l]. Nonlinear stability of fluid and plasma equilibria. Physics
Reports 123: 1-116.
Landau LD and Lifshitz EM (1976) Mechanics, 3rd edn.
L L New York: Pergamon Press.
Morrison PJ (1998) Hamiltonian description of the ideal
Define the disturbance norm by eqn [52], with fluid. Reviews of Modern Physics 70: 467-521.
c1 I 1. 5 c2. Salmon R (1988) Hamiltonian fluid mechanics. Annual
Review of Fluid Mechanics 20: 225-2.56.
II(V> P - Po)/I2 Salmon R (1998)Lectures on Geophysical Fluid Dynamics.
New York: Oxford University Press.
Shepherd TG (1990) Symmetries, conservation laws, and
521 Hamiltonian structure in geophysical fluid dynamics.
Advances in Geophysics 32: 287-338.
Then using eqn [51] we obtain the chain [53 of Shepherd TG (1993) A unified theory of available potential
inequalities, valid for any time t, involving the energy. Atmosphere-Ocean 31: 1-26.

research.
See also
Hamiltonian dynamics is often associated with equation (eqns [ l ] ) .

conservation of energy, but it is in fact much more
than that. Hamiltonian dynamical systems possess dqi - dpi - -E ( i = 1 , . . . ,N) [l]
a mathematical structure that ensures some remark- dt api’ dt aqi
able properties, Perhaps the most important is the X(q,p) is the Hamiltonian function, q = (41, . . . , q N )
connection between symmetries and conservation are the generalized coordinates, and p (el,. . . ,p ~ )
laws known as Noether’s theorem. Well-known the generalized momenta. For so-called natural sys-
examples are the fact that conservation of energy
is linked to symmetry in time, and conservation
+
tems with X = ( /pI2/2m ) U(q), where m is the mass
and U the potential energy, eqns [I] immediately lead
of momentum to symmetry in space. Less well-known to eqn [2], which is Newton’s second law for a
is the fact that material conservation of potential conservative system.
vorticity, so crucial to the theory of dynamical
meteorology, is also connected to a symmetry by
Noether’s theorem, but to a symmetry that is invisible
in the Eulerian formulation of the governing equa-
tions. It turns out that one can exploit the underlying Conservation of energy follows directly from eqns [l],
Hamiltonian structure of a system through the rele- for any IH, by the chain rule (repeated indices are
vant conservation laws even if the explicit form of that summed):
structure is not known, which is useful for applica-
tions. As is shown in detail below, symmetry-based
conservation laws provide a general theory of avail-
able potential energy, and show why it is that Rossby
waves carry negative zonal momentum, thereby [31
explaining both the maintenance of the westerlies
and the stratospheric Brewer-Dobson circulation.
Such laws also provide a powerful way of deriving
stability criteria. Symplectic Formulation
Dynamical meteorologists use a variety of theoret- The theory of canonical transformations suggests that
ical models, ranging from the fully compressible there is nothing special about the qs and p s , and
equations through the hydrostatic primitive, Boussin- Hamilton’s equations [l] can be written in the so-
esq, and quasi-geostrophic equations to the barotropic called symplectic form, eqn [4].
equations. With such a zoo of models, it is crucial to
know the extent to which theories developed
for one model carry over to another. Hamiltonian
dynamics provides this unifying framework. All the
models just mentioned are in fact Hamiltonian, In eqn [4],u = (41,. . . , qN,p1,. . . , p ~ a)n d l is given
and models can be grouped into families according by eqn [5],where I is the N x N identity matrix.
to their Hamiltonian structure. In this way it
becomes immediately apparent, for example, that
the Charney-Stern stability theorem for baroclinic
quasi-geostrophic flow is the counterpart to Rayleigh’s J has certain mathematical properties, including skew-
inflection-point theorem for barotropic flow, and symmetry. More generally, one can take those prop-
that an analogous stability theorem will exist for
erties to be the definition of Hamiltonian structure,
any balanced model having a similar Hamiltonian with] not necessarily of the form of eqn [5]. The skew-
structure, no matter what the definition is of the symmetry of J nevertheless guarantees energy conser-
potential vorticity. Thus, it is precisely through its
vation (eqn [6]).
abstract character that Hamiltonian dynamics has
many powerful applications in theoretical dynamical
meteorology.
There is an important distinction between systems

with a nonsingular (or invertible) J , which can always
Canonicaland NoncanonicalDynamics be transformed into the canonical form of eqn [5],and
In classical mechanics, canonical Hamiltonian dy- those with a singular (or noninvertible) J . The latter,
namical systems are those described by Hamilton’s known as noncanonical systems, possess a special class
of invariant functions known as Casimir invariants. underpins many of its applications to dynamical
These are the solutions of eqn [7] (for canonical meteorology.
systems the solutions are just constants). Casimir invariants are special because 6cu = 0. This
suggests that they correspond to invisible symmetries.
J--=O
ac ( i = 1,. . . For example, in rigid-body dynamics the total angular
'I aui momentum is a conserved quantity in any description
of the motion. In the original canonical description it
That they are necessarily conserved in time then
corresponds to the rotational symmetry of the dynam-
follows from the skew-symmetry ofJ (eqn [SI).
ics, but in Euler's equations, where angles have been
eliminated, it enters as a Casimir because the under-
lying physical symmetry is no longer explicit.
The best-known example of a noncanonical Ham-

iltonian system is Euler's equations for rigid-body Barotropic Dynamics
dynamics. Having an odd number of evolution equa- In what sense are the models of dynamical meteorol-
tions (three in this case), the system is necessarily ogy Hamiltonian? Consider what is probably the
noncanonical because any skew-symmetric matrix of simplest such model, the barotropic vorticity equation
odd dimension must be singular. There is one Casimir (eqn [ 1I]),which describes two-dimensional, nondi-
invariant for Euler's equations, the total angular vergent flow.
momentum.
Noether's Theorem
For a canonical system, if a particular generalized -
Here o ( x ,y, t) = z (V x v) = Q2$ is the vorticity, z is
coordinate q, does not appear in the Hamiltonian, then the unit vector in the vertical direction, v(x, y , t) =
the Hamiltonian is invariant under changes in that i x V$ is the horizontal velocity, $(x,y , t ) is the
coordinate; in other words, there is a coordinate streamfunction, and a( f ,g) fxgr - frgx is the two-
symmetry. Translational and rotational symmetries dimensional Jacobian. The candidate Hamiltonian is
are common examples. Hamilton's equations [ 11then the conserved energy of this system, which is just the
immediately imply that the corresponding generalized kinetic energy. The obvious dynamical variable is the
momentum is conserved: dpjldt = 0. vorticity. In order to cast eqn [ll]in the form of
This connection between symmetries and conserva- eqn [4], we need to regard every point (x,y) in space as
tion laws has a more general and far more power- indexing a degree of freedom analogous to the index i;
ful form. Given a function 3 ( u ) , define 6 ~ u= j the sum over i then becomes an integral over space,
cjij(aF/auj), where E is an infinitesimal parameter; dFu functions become functionals, and partial derivatives
is called the infinitesimal variation in u generated by 3. become functional or variational derivatives. Thus we
(In the canonical case, 6T.u is an infinitesimal canonical write eqn [12].
transformation.) It then follows that the infinitesimal
variation in Fl generated by F is given by eqn [9].
= // V$ . 6V$ dx dy
On the other hand, the time evolution of 3 is given by P P
eqn [lo]. = ]] {V - ($SV$) - $So}dx dy [12]

d 3 w d u j a 3 ax
-- _ _ = -Ijj- 1101 Assuming for now that the boundary terms vanish, we
dt -sui dt auj au, identify the variational derivative as SE/6o = -$.
Using the skew-symmetry of J , eqns [9] and [lo] then The need to integrate by parts reflects the fact that the
imply that 6 ~ 7 -=l 0 if and only if d F / d t = 0. This effect of a vorticity perturbation on the kinetic energy
connects symmetries and conservation laws: the density is nonlocal; thus, partial derivatives at fixed
Hamiltonian is invariant under the variation generat- points in space make no sense and variational deriv-
ed by 3 (i.e., that variation represents a symmetry of atives are essential. Equation [ l l ] can now be cast in
the Hamiltonian) if and only if 3 is a conserved Hamiltonian form as eqn [ 131.
quantity. This result, known as Noether's theorem, is am
-= J -
ax where J = -a(o,.) [13]
one of the central results of Hamiltonian dynamics and at 60
Note that J is now a differential operator rather than a about the origin.
matrix. It is evidently skew-symmetric: SSfJg dx dy =
- J’gJf dx dy (under suitable boundary conditions)
for arbitrary functions f , g.
M=-//i?udxdy=
// f . ( r x v ) d x d y [18]
Conservation Laws The discussion has so far neglected any contribution

from boundary terms. They are easily included. In the
The form of J in eqn [13] is clearly singular: any presence of rigid lateral boundaries, for a complete
function of w inserted in the argument gives zero. mathematical specification of the problem, eqn [ l l ]
These then represent Casimir invariants of the system: must be supplemented with the conditions [19] on
functionals of the form [14], where C(.)is an arbitrary each connected portion of the boundary.
differentiable function, evidently satisfy
J ( S C / S u ) = 0.
v - n = 0, d t {v.dr =0
Here n is the outward-pointing normal, and s is

the vector arc length along the boundary. The second
The fact that such functionals are conserved in time of eqns [ 191 represents conservation of circulation,
corresponds to the material conservation of vorticity which follows from the underlying momentum
expressed by eqn [ 113. equations but must be included as a separate
To identify the momentum invariants, we need to condition in the vorticity formulation of the dynamics.
apply Noether’s theorem to the various spatial Although the circulation integrals along each
symmetries. Suppose that the domain is unbounded, connected portion of the boundary are constants in
with decay conditions at infinity, so that there is time, they are independent dynamical variables
symmetry in all directions. The variation in w corre- and are needed to determine v from w. The
sponding to a translation by SX in the coordinate x is Hamiltonian formulation of eqn [13] may easily
given by 6w = -(au/ax)6x. Setting E = 6x, we then be extended to include the circulation integrals in
need to solve for the momentum invariant M accord- addition to w as dynamical variables. The Casimir
ing to eqn [15]. invariants then include functions of these circulation
integrals. With regard to the momentum invariants,
of course, the rigid boundaries must respect the
same symmetries; a zonal channel flow with walls
To within the addition of a Casimir, the solution of eqn at constant y breaks the translational symmetry
[ 151 is given by SM/Sw = y. Hence we may choose M in y and the rotational symmetry, leaving only the
as in eqn [16], where v = (u,v ) . zonal impulse of eqn [16] as an invariant. The
final equality of eqn [16] is then no longer strictly
true, but the impulse and momentum differ only by
terms involving the circulations along the channel
walls, which are Casimirs. Since symmetry-based
= /J u dx dy
invariants are only defined to within a Casimir in
any case, the impulse and momentum are essentially
equivalent.
The first, elementary form of M given by eqn [16] is A simplified model of barotropic dynamics is the
known as Kelvin’s impulse. It represents the y ‘center- point-vortex model, where the vorticity is concentrat-
of-mass’ of the vorticity distribution, and is in many ed in Dirac delta functions. The point-vortex model
ways the preferable form because it is local in 0.The has been used to study two-dimensional turbulence
final form, however, shows that the invariant M and certain kinds of atmospheric flow structures. It
corresponding to symmetry in x is ultimately just the also turns out to be Hamiltonian, and is in fact a
x-momentum, as expected. canonical system: the Casimirs are built into the model
The same argument applied to translation in the as parameters through the choice of the point-vortex
coordinate y yields eqn [171. strengths.
Other Balanced Models

Finally, rotational symmetry leads to eqn [18], where The barotropic vorticity equation has a mathematical
r (x, y) and r = lrl, which is the angular momentum structure that is analogous to that of many models of
balanced, or potential-vorticity-driven, flow (see Dy- the equation for the interior potential vorticity (the
namic Meteorology: Balanced Flows) and the results latter being eqn [Ill, with q in place of w , applied at
derived above extend in an obvious way to such every value of z; thus, the advection of q remains
systems. Inclusion of the beta effect means simply a purely horizontal). The Eady model is an extreme case
change from w to the potential vorticity q = w By. + where the interior potential vorticity is uniform and
Since 6 y = 0 (recalling that the coordinate y is like an the flow is driven entirely by the temperature distri-
index), 6q = 6w and eqns [11]-[14] go through butions on the upper and lower boundaries; the
unchanged with q in place of w. However the beta dynamical structures driven from each boundary are
effect breaks translational symmetry in y and rota- known as Eady edge waves. Since these temperature
tional symmetry, leaving only the translational sym- distributions also evolve according to eqn [ l l ] , with
metry in x represented by the zonal impulse invariant the QG temperature IC/, in place of w, it is not
of eqn [16]. Strictly speaking the latter should be surprising that the same kind of Hamiltonian structure
written with q in place of w , but the integrals differ by a also applies to this model. The energy is given by eqn
constant and so represent the same invariant. Inclu- [201.
sion of topography is no more difficult; one simply
includes an additional topographic term h( x ,y ) in the
definition of q. This will generally break all spatial
symmetries, leaving only the energy 'FI and Casimirs C
In eqn [20], the reference-state density p,(z) and
as invariants. This illustrates a general and important
stratification function S(z) = N 2 / f 2 are both pre-
point, namely, that symmetry-based invariants are
scribed, with N ( z ) the buoyancy frequency and f the
fragile: a slight change in the conditions of the problem
Coriolis parameter, and where V is still just the
destroys their conservation properties. In contrast, the
horizontal gradient operator. With the potential
energy and the Casimirs are robust invariants (robust
vorticity given by eqn [21], where f and P are
within the conservative context, of course) that
constants, eqn [22] follows.
survive such perturbations.
Stratification is most easily introduced in the con-
text of the quasi-geostrophic ( Q G )model (see Quasi-
geostrophic Theory). Layered QG models are com-
pletely trivial extensions of the barotropic system:
their evolution is determined by the potential vorticity
qi(x,y , t )in each layer i, governed by eqn [ I 11with qi in
place of w , together with conservation of circulation
along any rigid lateral boundaries that may be present.
These are then the dynamical variables. The energy
now includes available potential as well as kinetic [221
energy, but, apart from some geometric factors repre-
This is like eqn [12], but with an additional term
senting the layer depths, one still recovers 6'FI/6qi =
involving the temperature variations 6$z at the upper
- $ i in each layer as well as eqn [13] with qi in place of
and lower boundaries. Including these as independent
w . The various invariants follow in the obvious way
dynamical variables, in addition to q (and possibly
with the spatial integrals summed over the different
also circulation terms), the governing equations can be
layers. The same considerations, incidentally, apply to
cast in the symplectic form of eqn [13]. The Casimirs
layered non-QG 'intermediate' models that still have
the form of eqn [I11 - namely, nondivergent now involve integrals of arbitrary functions of the
horizontal advection of the potential vorticity qi temperature on the upper and lower boundaries, in
within each layer, with the flow in each layer driven addition to integrals of arbitrary functions of potential
vorticity in the interior (eqn [23]).
by the potential vorticity in all layers (as described by
the particular definition of qi).
With continuous stratification and with upper and
lower boundaries (at z = 1and z = 0, say), there is an
additional effect. It is well known that the temperature
distribution along the upper and lower boundaries is
equivalent to potential vorticity (see Baroclinic Insta-
bility), and independent evolution equations for these
temperature distributions are required to fully specify The momentum invariants similarly extend in
the continuously stratified QG system, in addition to obvious ways: for example, the zonal impulse
invariant is given by eqn [24]. The dynamical variables are v and h, for which eqns
[27] hold.
63.t 63.t 1
- = hv,
6V
= 31Vl2 + gh 1271
Note that no integration by parts is necessary in this

The semi-geostrophic (SG) model (see Frontogene- case; this is characteristic of velocity-based represen-
sis) is widely used in mesoscale dynamics because of its tations of the dynamics. It can easily be verified that
ability to represent realistic frontal structures. It turns eqns [25] may be cast in the symplectic form (aulat) =
out that the SG model can also be cast in the form of J(63.t/6u) with u = ( u ,u, h ) with J given by eqn [28],
eqn [ 111, and hence in the symplectic form of eqn [131, where q = -
(f + i V x v)/h is the potential vorticity of
provided the equations are written in isentropic- the shallow-water system.
geostrophic coordinates. However, in these coordi-
0 4
nates rigid boundaries appear to move in time. The SG
equations, in contrast to the QG equations, make no I=(:;
-ax
,: -2) PI
geometrical distinction between horizontal and verti-
cal boundaries - this is why they are also useful for the The matrix (28) is evidently skew-symmetric; the
study of coastal dynamics in physical oceanography - signs on the derivative terms are indeed correct,
and the same kind of independent dynamical degrees of since first-order differential operators are themselves
freedom encountered in the QG system on upper and skew-symmetric, as with the J in eqn [13]. The
lower boundaries also appear on lateral boundaries. In zonal (absolute) momentum invariant is given as
the special case of channel walls, these degrees of expected by eqn [29], for which it is easy to verify
freedom correspond to coastal Kelvin waves and are that J(6M/6u)= -au/ax in line with Noether’s
analogous in some respects to the Eady edge waves theorem, and the other momentum invariants follow
represented by both the QG and SG systems. They similarly.
must be taken into account in the variational calcula-
tions, and enter into many of the resulting expressions.
M = //h(u-fy)dxdy ~ 9 1
Unbalanced Models The Casimirs are given by eqn [30] for arbitrary
Balanced models are controlled by the advection of functions C( .).
potential vorticity (perhaps augmented by the advec-
tion of isentropic surfaces on rigid boundaries), so
for such models it is natural to seek a Hamiltonian
C= // hC(q)dx dy [301
description analogous to eqn [131. However, models Thus, potential vorticity still plays a crucial role in the
that include a representation of gravity waves or other Hamiltonian description of the dynamics. Special
high-frequency oscillations, called unbalanced mod- cases of Casimirs are total mass (C = 1) and total
els, do not fit within this framework. They necessarily circulation (C = q ) .
have additional degrees of freedom. For such models, a Stratification is easily incorporated. The hydrostatic
description in terms of the velocity field is a more primitive equations can be cast in Hamiltonian form
natural way to reflect the Hamiltonian structure. For isomorphic to that of eqn [28] when expressed in
example, the rotating shallow-water equations [25] isentropic coordinates. Even the fully compressible
withv(x, y, t ) = ( u ,v ) the horizontalvelocity, h ( x ,y, t ) stratified Euler equations, which form the most
the fluid depth, g the gravitational acceleration, and general system imaginable for (dry) dynamical mete-
with constant f , conserve the energy (eqn [ 2 6 ] ) . orology, can be cast in an analogous form, although
there are now additional dynamical variables associ-
av + ( f i + V x V) x v + V
- ated with compressibility. The Casimirs are in this case
at given by eqn [31], where p(x, y, z , t ) is the density,
+
d ( x , y, z , t ) is the potential temperature, and q = [ ( f z
ah
-+V.(hv)
at
=o -
V x v) V d ] / p is the Ertel potential vorticity, with v
and V now acting in all three space dimensions.
3.t = // +
{h/vI2 gh2} dx dy 1261
The invariance of the Casimirs is of course evident but not positive definite. One would like both prop-
directly from the dynamical equations (eqn [32]) and erties in order to define normal modes, spectra, etc.
reflects the material invariance of q and 8. Another problem, at first sight unrelated, arises with
momentum. The momentum of a wave would appear
84 ae to be zero (the average of a sinusoid is zero), yet waves
-+v+
at
= 0, -+v.ve
at
= 0:
can certainly transfer momentum; this is what drives
the quasi-biennial oscillation in the tropical strato-
a- +P V . ( p v ) = 0 sphere, for example (see Middle Atmosphere: Quasi-
at
1321 Biennial Oscillation). How is one to describe this wave
momentum?
The fully compressible stratified Euler equations In canonical Hamiltonian mechanics, the distur-
are, in fact, a straightforward expression of Newton's bance energy about an equilibrium is always quadrat-
second law, without constraints such as hydrostatic ic; from this one assesses stability and defines normal
balance, provided they are expressed in Lagrangian modes. There is no ambiguity. So why are things not
coordinates (see Lagrangian Dynamics). In Lag- equally clear for fluid dynamics? The answer lies in the
rangian coordinates, the dynamical variables are the noncanonical Hamiltonian structure of virtually every
positions and momenta of fluid elements, which are fluid dynamical system in the Eulerian representation.
natural canonical variables. The thermodynamic fields If u = U is a steady solution of a Hamiltonian system,
can be expressed in terms of these variables: p can be then eqn [33] holds.
written in terms of the Jacobian of particle positions
(which describes the compression of the fluid), while
can just be chosen as one of the Lagrangian coordi-
nates. In this way, the fully compressible stratified
Euler equations represent a canonical Hamiltonian For a canonical system, the invertibility of J then
system. But there are six dynamical variables in the implies that 63-1/6u = 0 at u = U. This means that U is
Lagrangian description, compared with only five in a conditional extremum of 3-1, and X[u] - X[U] is
the Eulerian description; in transforming to Eulerian quadratic in the disturbance. However, for a non-
coordinates, a reduction of the phase space takes canonical system none of this follows and the distur-
place. This is where the potential vorticity comes in. In bance energy is generally linear in the disturbance.
Lagrangian coordinates, the potential vorticity is still
Pseudoenergy
materially conserved; but what symmetry does it
correspond to? The answer is a particle-relabeling Hamiltonian structure provides the solution to this
symmetry: if one rearranges fluid elements while quandary. Equation [33] is locally the same as the
preserving the same Eulerian fields, then the dynamics equation defining the Casimirs, which means that
is unchanged. There is just enough freedom to do this, 63-1/6u is locally parallel to 6C/6u for some C (a
because there is one more Lagrangian than Eulerian different C for each choice of U). In other words, there
variable. Upon reduction to the Eulerian description, exists a Casimir C such that eqn [34] holds.
this additional degree of freedom disappears, and the
particle-relabeling symmetry becomes invisible. That
is why potential vorticity conservation then appears in
the form of a Casimir invariant.
Now, both 3-1 and C are invariants, and the com-
bined invariant 3-1 + C satisfies the extrema1 condition
Disturbance Invariants d(3-1 + C) = 0 at u = U. We have thus constructed
what we wanted, namely a disturbance quantity that is
Probably the most powerful application of Hamilto-
conserved and is locally quadratic in the disturbance
nian dynamics to dynamical meteorology arises in the
context of studying the properties of disturbances to
basic states. In fluid dynamics, the question of how to
define the energy of a wave has often been a point of
+ +
A = (3-1 C)[U] - (3-1 C)[U] WI
confusion if not contention. For example, in the case of This quantity is known as the pseudoenergy. Provided
a basic flow, if the wave energy is defined as the energy one has a complete set of Casimirs, eqn [34] can
in the frame of reference moving with the basic flow, always be solved for a Hamiltonian system and the
then it is positive definite but not conserved. On the pseudoenergy can always be constructed according to
other hand, if it is defined as the difference energy eqn [35]. This is one of the great attractions of
relative to the basic-flow energy, then it is conserved Hamiltonian dynamics: it provides systematic recipes
in abstract terms, which can be worked out for any This is self-evidently positive definite for dpo/dz < 0
particular application. and has the small-amplitude quadratic approximation
A particularly illuminating application is the subject [421.
of available potential energy, highly useful in energy
budget analyses. Me demonstrate the method in the
case of the three-dimensional stratified Boussinesq 1421
equations. The energy is given by eqn [36].
Equation [41] is the exact, finite-amplitude expression
for the available potential energy (see General Circu-
lation: Energy Cycle) of disturbances to a stably
stratified, resting basic state p o ( z ) ,while eqn [42] is its
Here ps is the constant reference-state density, and the more familiar small-amplitude counterpart, widely
dynamical variables are v and p , for which eqns [37] used in the theory of internal gravity waves (see
hold. Buoyancy and Buoyancy Waves: Theory). Similar
constructions can be performed to define the available
potential energy of any stratified fluid system.
Although the small-amplitude expression of eqn
The term pgz in eqn [36] is the gravitational potential [42] appears to be singular in regions where
energy, and is linear in the dynamical variables. dpo/dz = 0 , the finite-amplitude expression of eqn
Now consider disturbances to a stably stratified, [41] remains perfectly well-defined in such regions.
resting basic state v = 0, p = po(z). Although the Pseudomomentum
Casimirs of this system include functions of the
potential vorticity, because the basic state is at rest, The same kind of reasoning can be applied for
67-t/6v = 0 at v = 0 and this dependence is unneces- disturbances to zonally symmetric (x-invariant) basic
sary, so we may consider Casimirs of the form of states, assuming that the underlying system possesses
eqn [38]. the same symmetry. For such states, with aU/ax = 0,
Noether’s theorem implies that the zonal impulse or
C= /// C ( p )dx dy dz with
6C
- = C’(p)
6P
[38] momentum invariant satisfies eqn [43].
Equation [34] then leads to the condition

C’(po) = -gz. This is the defining relation for the
function C(.).Thus, one has to express gz in terms of But just as with eqn [33], there is a Casimir C such that
the same argument po. This can be done by inverting 6(M + C) = 0 at u = U; with this C, one may imme-
the functional dependence p o ( z ) to obtain Z ( p o ) , diately construct the invariant [44], which is quadratic
where Z ( p o ( z ) )= z. This is always possible provided to leading order in the disturbance.
p o ( z ) is monotonic, which is the case for a stably
stratified basic state. One then has eqn [39].
A = (M + C)[U] - ( M + C)[U] P41
This quantity is known as the pseudomomentum.
C b ) = - / p g z ( P ) dP [391 We calculate the pseudomomentum for the case of
barotropic flow on the beta-plane. Suppose we are
From this the pseudoenergy of eqn [35] takes the form given a monotonic basic state q o ( y ) . From eqns [14]
and [16], with q in place of o,we have eqn [45].
1401.
The extrema1 condition 6(M + C) = 0 at q = qo then

leads to C’(q0) = -y. This is now isomorphic to the
construction of the available potential energy, replac-
The first term in the spatial integrand is the kinetic ing gz with y and p with q. If we define the function
energy and is positive definite; the last two terms can Y (.) by Y (qo(y)) = y, then evidently eqn [46] holds.
be rewritten as in eqn [41].
1411
The small-amplitude approximation to the spatial to leading order in the disturbance quantities. In fact,
integrand is given by eqn [471. their quadratic approximations are exactly conserved
by the linearized dynamics. (The quadratic approxi-
mation to the pseudoenergy is the Hamiltonian of the
linearized dynamics.) When either of these quantities
is sign-definite for a given basic state, it follows that
Equations [46] and [47] are evidently negative definite that basic state is stable to normal-mode instabilities.
for dqo/dy > 0, which is the case when qo is dominated Indeed, in order to reconcile exponentially growing
by by. These rather peculiar expressions have no disturbances with conservation of pseudoenergy and
obvious relation to zonal momentum at first sight, but pseudomomentum, the latter quantities must vanish
they nevertheless explain why it is that Rossby waves for such disturbances. This fact provides a useful
always exert an eastward (positive) force when they constraint on the structure of normal-mode instabi-
leave a source region, and a westward (negative) force lities, as well as a powerful unifying framework
when they dissipate and deposit their momentum in a between different models.
sink region: they carry negative pseudomomentum. This simple framework accounts for virtually every
The general nature of the derivation ensures that known stability theorem in dynamical meteorology.
exactly the same expressions hold for any balanced For resting, stratified basic states in unbalanced
model having the basic form of eqn [13]. If the basic models, with pseudoenergy like eqn [40] for the
state qo is chosen to be the zonal mean 4, then the zonal Boussinesq model, the condition of positive definite
mean of eqn [47] becomes eqn [48], where q’ = q - 4. pseudoenergy is the statement of static stability (see
Static Stability). For basic flows in axisymmetric or
symmetric stratified unbalanced models, the same
condition is the statement of symmetric stability (see
Instability: Symmetric Stability), which reduces to
In the case of stratified QG dynamics, the negative of Rayleigh’s centrifugal stability theorem in the special
eqn [48] is known as the Eliassen-Palm wave activity case of axisymmetric homogeneous flow (see Instabil-
(see Middle Atmosphere: Zonal Mean Climatology), ity: Inertial Instability). These stability theorems are
which has been widely used in dynamical meteorology all quite analogous t o static stability. A different
to assess the effect of Rossby waves on the zonal mean situation arises for balanced models. There, the
flow. It is such an effective diagnostic precisely because pseudoenergy can take either sign depending on the
it represents negative pseudomomentum. Moreover, basic flow. The positive-definite and negative-definite
and importantly, its use is not restricted to waves. The cases correspond respectively to Arnol’d’s first and
exact, finite-amplitude expression of eqn [46] ensures second stability theorems. (They are analogous to the
that the concept of pseudomomentum applies to fully stability of a rigid body rotating about an axis of
nonlinear, even turbulent disturbances. symmetry corresponding respectively to a maximum
The robust negative definiteness of the pseudomo- or minimum moment of inertia.) In the special case of a
mentum of balanced disturbances explains a great deal parallel basic flow, Arnol’d’s first theorem states that
about the general circulation of the atmosphere. the flow is stable if uo/(dqo/dy)<O, which is the
Propagation of synoptic-scale Rossby waves away FjGrtoft-Pedlosky theorem.
from their source region in the baroclinic storm tracks With regard to pseudomomentum for balanced
implies an eastward force in the storm track regions, models, eqn 1461 is sign-definite whenever dqoldy is
accounting for the maintenance of the westerlies sign-definite. For barotropic flow with q = w , this
(see General Circulation: Momentum Budget). corresponds to Rayleigh’s inflection-point theorem; on
The westward momentum deposition associated the beta-plane with q = co -k by, to the Rayleigh-Kuo
with breaking planetary-scale Rossby waves in the theorem; and for stratified QG flow with q given either
stratosphere drives the poleward Brewer-Dobson by its multilevel forms qr or by eqn [21] in the
circulation (see Middle Atmosphere: Zonal Mean continuously stratified case, to the Charney-Stern
Climatology), which is responsible for the observed theorem. For stratified QG dynamics in the presence of
distribution of ozone and other chemical species in the a lower boundary, the second terms of eqns [23] and
stratosphere. [24] become relevant and there is an additional
contribution to the pseudomomentum involving the
temperature distribution on the lower boundary; it is
Stability Theorems isomorphic to the interior eqns [46]-[48], replacing q
The pseudoenergy and pseudomomentum are, by their with I),. Since the climatological temperature gradient
construction, conserved quantities that are quadratic along the Earth’s surface is towards the Equator, the
pseudomomentum associated with surface distur- pseudoenergy d of eqn [40].

bances is generally positive. In this case the Charney-
Stern stability criterion is not satisfied for observed
flows; on the other hand, normal-mode instabilities are
generally required to involve both temperature distur-
bances on the lower boundary and potential-vorticity
disturbances in the interior, in order to create a
disturbance with zero total pseudomomentum. The With the choice 6 = &&E, eqn [ 5 3 ] establishes
Charney model of baroclinic instability (seeBaroclinic Liapunov stability in the norm defined by eqn [52].
Instability) is the best-known example of this. In the Conservation of pseudoenergy is clearly central to the
presence of an upper boundary, there is a further proof.
contribution to the pseudomomentum, with opposite The finite-amplitude stability of stably stratified flow
sign to the lower contribution in accord with eqn [24]. is not too surprising; it corresponds to physical intui-
Thus in the Eady model of baroclinic instability, where tion, and indeed motivates the very concept of available
the potential vorticity is uniform and the interior potential energy, which has a long pedigree. What is
contribution to the pseudomomentum disappears, the perhaps more surprising is that exactly the same kinds
instability can arise from the interaction of distur- of constructions can be made for all of the stability
bances on the upper and lower boundaries that theorems mentioned above, and for virtually any model
together add up to zero total pseudomomentum. within the same family. They can also be used to obtain
These statements all concern normal-mode stability. rigorous upper bounds on the saturation of normal-
But what can be said about stability goes much further mode instabilities, by considering the initial unstable
than this. The existence of finite-amplitude distur- flow (plus infinitesimal disturbance) to be a finite-
bance invariants suggests the possibility of nonlinear, amplitude disturbance to a stable basic state.
or Liapunov stability: namely, that small disturbances
stay small for all time, where small is defined in terms See also
of some disturbance norm. Mathematically, we say
that a basic state U is Liapunov stable to disturbances Baroclinic Instability. Buoyancy and Buoyancy
u’ in a given norm l\u’IIif for all E > 0 there exists a Waves: Theory. Dynamic Meteorology: Balanced
6 ( ~ >) 0 such that eqns [49] hold. Flows; Potential Vorticity. Frontogenesis. General Cir-
culation: Energy Cycle; Momentum Budget. Instability:
Ilu’(O)il<S +~Iu’(~)II<E for all t [49] Inertial Instability; Symmetric Stability. Lagrangian Dy-
namics. Middle Atmosphere: Quasi-Biennial Oscilla-
Let us see how this applies to static stability for tion; Zonal Mean Climatology. Quasi-geostrophic
the Boussinesq model considered earlier. Suppose that Theory. Static Stability.
the basic state has dpo/dz< 0 and that furthermore the
basic-state density gradients are bounded according to
[50] for some constants G I , c2.
Further Reading
Arnol’d VI (1989) Mathematical Methods of Classical
Mechanics, 2nd edn. New York: Springer-Verlag.
Benjamin TB (1984) Impulse, flow force and variational
principles. IMAJournal ofApplied Mathematics 32: 3-68.
Then eqn [41] for the available potential energy is Holm DD, Marsden JE, Ratiu T and Weinstein A (198.5)
bounded from above and below according to eqn [5l]. Nonlinear stability of fluid and plasma equilibria. Physics
Reports 123: 1-116.
Landau LD and Lifshitz EM (1976) Mechanics, 3rd edn.
L L New York: Pergamon Press.
Morrison PJ (1998) Hamiltonian description of the ideal
Define the disturbance norm by eqn [52], with fluid. Reviews of Modern Physics 70: 467-521.
c1 I 1. 5 c2. Salmon R (1988) Hamiltonian fluid mechanics. Annual
Review of Fluid Mechanics 20: 225-2.56.
II(V> P - Po)/I2 Salmon R (1998)Lectures on Geophysical Fluid Dynamics.
New York: Oxford University Press.
Shepherd TG (1990) Symmetries, conservation laws, and
521 Hamiltonian structure in geophysical fluid dynamics.
Advances in Geophysics 32: 287-338.
Then using eqn [51] we obtain the chain [53 of Shepherd TG (1993) A unified theory of available potential
inequalities, valid for any time t, involving the energy. Atmosphere-Ocean 31: 1-26.
HUMIDITYVARIABLES 939
J Curry, Georgia Institute of Technology, Atlanta, GA, A rigorous derivation of the saturation vapor
USA pressure can be accomplished from the second law of
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. thermodynamics. For the liquid-vapor equilibrium,
we can write the Clausius-Clapeyron equation as
follows,
Introduction
des - -
hes
Atmospheric humidity is the amount of water vapor
._
d T - R,T2 PI
present in the atmosphere. Both the rate of evaporation
and the time and place that condensation of cloud water where L1, is the latent heat of vaporization (energy per
occurs are controlled by humidity. Determination of the unit mass). At any given temperature, eqn [2] states
atmospheric humidity is important for determining that there is one and only one pressure at which water
surface evaporation, atmospheric radiative transfer, vapor is in equilibrium with liquid water, with the
and certain chemical reactions in the atmosphere. saturation vapor pressure increasing approximately
exponentially with increasing temperature.
Vapor Pressure Integration of eqn 121 is made difficult owing to the
variation of the latent heat of vaporization with
The partial pressure of water vapor, e, is used as the
temperature. Additionally, application of the Clau-
fundamental measure of water vapor content in the
sius-Clapeyron equation to determining the satura-
atmosphere. Using Dalton’s law of partial pressures,
tion vapor pressure in the atmosphere is not strictly
we can use the ideal gas law to write the following
valid because of the presence of other gases. Hence
expression for water vapor pressure:
empirical values of the saturation vapor pressure are
e = p,R,T PI typically used (shown in Table 1).
These empirical values of saturation vapor pressure
where p, is the water vapor density (often referred to can be represented by a sixth-order polynomial:
as the absolute humidity), R, is the specific gas
constant for water, and T is the atmospheric temper- 7
ature. The atmospheric relative humidity is defined as e, = a1 t x a , ( T - T,,)’-’ [31
the ratio of the atmospheric vapor pressure to the n=2
equilibrium (or saturation) vapor pressure at the
temperature of the air, e,. where T,,, the so-called ‘triple point’ of water, is
To understand what is meant by the saturation 273.15K, and the coefficients for the saturation
vapor pressure, we must first examine the equilibria vapor pressure over water and over ice are as given
between phases. Consider a closed container half full in Table 2. Note that, at a given temperature, the
of pure water and overlain by dry air. As the water saturation vapor pressure over ice is smaller than the
begins to evaporate from the water surface, a small saturation vapor pressure over liquid water; this is
increase in pressure is detected in the air above, because the latent heat of sublimation (associated with
resulting from the motion of the water vapor mole- the phase transition between vapor and ice) is larger
cules that are added to the air through evaporation. As than the latent heat of vaporization.
more and more molecules escape from the water
surface in the closed container, the steadily increasing
vapor pressure in the air above forces more and more
Humidity Variables
of these molecules to return to the liquid. Eventually Values of the saturation vapor pressure are used in the
an equilibrium is reached where the number of water determination of some of the commonly used atmos-
vapor molecules returning to the surface balances the pheric humidity variables. A wide variety of different
number leaving. At that point the air is said to be humidity variables are used in atmospheric science.
saturated with water vapor, and the partial pressure of The reasons for this are partly historical and partly
the water vapor is equal to the saturation vapor related to the different methods by which atmospheric
pressure. The relative humidity is equal to unity (or humidity is measured. Given information about the
100%). If the temperature of the water in the closed ambient atmospheric temperature and pressure, one
container were increased, more water would evapo- humidity variable can be used to determine each of the
rate before a balance was reached. other humidity variables.
940 HUMIDITY VARIABLES
Table 1 Saturation pressures over pure liquid water and pure ice as a function of temperature
- 50 0.0635 0.0393 - 24 0.8826 0.6983 1 6.565 26 33.606

- 49 0.0712 0.0445 - 23 0.9647 0.7708 2 7.054 27 35.646
- 48 0.0797 0.0502 - 22 1.0536 0.8501 3 7.574 28 37.793
- 47 0.0892 0.0567 - 21 1.1498 0.9366 4 a. 128 29 40.052
- 46 0.0996 0.0639 - 20 1.2538 1.032 5 8.71a 30 42.427
- 45 0.1111 0.0720 - 19 1.3661 1.135 6 9.345 31 44.924
- 44 0.1230 0.081o - 18 1.4874 1.248 7 10.012 32 47.548
- 43 0.1379 0.0910 - 17 1.e183 1.371 a 10.720 33 50.303
- 42 0.1533 0.1021 - 16 1.7594 1.505 9 11.473 34 53.197
- 41 0.1704 0.1 145 - 15 1.9114 1.651 10 12.271 35 56.233
- 40 0.1891 0.1283 - 14 2.0751 1.810 11 13.118 36 59.418
- 39 0.2097 0.1436 - 13 2.2512 1.983 12 14.016 37 62.759
- 38 0.2322 0.1606 - 12 2.4405 2.171 13 14.967 38 66.260
- 37 0.2570 0.1794 -11 2.6438 2.375 14 15.975 39 69.930
- 36 0.2841 0.2002 - 10 2.8622 2.597 15 17.042 40 73.773
- 35 0.3138 0.2232 -9 3.0965 2.837 16 ia.171 41 77.798
- 34 0.3463 0.2487 -8 3.3478 3.097 17 19.365 42 82.01 1
- 33 0.3817 0.2768 -7 3.6171 3.379 18 20.628 43 86.419
- 32 0.4204 0.3078 -6 3.9055 3.684 19 21.962 44 91.029
- 31 0.4627 0.3420 -5 4.21 42 4.014 20 23.371 45 95.850
- 30 0.5087 0.3797 -4 4.5444 4.371 21 24.858 46 100.89
- 29 0.5588 0.4212 -3 4.8974 4.756 22 26.428 47 106.15
- 28 0.6133 0.4668 -2 5.2745 5.173 23 28.083 48 111.65
- 27 0.6726 0.5169 -1 5.6772 5.622 24 29.829 49 117.40
- 26 0.7369 0.5719 -0 6.1070 6.106 25 31.668 50 123.39
- 25 0.8068 0.6322 - - - - - - -
The relative humidity, H, is defined as commonly multiplied by 100 and expressed as a

H = -e percentage. Relative humidity can be changed if
~41 moisture is added or removed (which changes e ) , or
e,
if the air temperature changes (which changes e s ) .At
and Hi, the relative humidity with respect to ice temperatures below O"C, it is necessary to specify
saturation, is defined as whether the relative humidity is being evaluated
H. e
-
I -
- relative to the saturation vapor pressure over liquid
esi water or over ice. Table 3 shows that an atmosphere
where e,i is the saturation vapor pressure over ice. The saturated with respect to liquid water is supersaturat-
relative humidity is the ratio of the actual partial ed with respect to ice, and that the degree of supersat-
pressure of water vapor in the air to the saturation uration increases with the supercooling.
vapor pressure, and is a function only of e and T . It is The water vapor mass mixing ratio, w,, is the ratio
of the mass of water vapor (m,) present to the mass of
Table 2 Coefficients of the sixth-order polynomial fits to satura- dry air (md).It is thus defined, after substituting from
tion vapor pressure for the temperature range - 50" to 50°C for the ideal gas law [l],as
both liquid water and ice
Coefficient Liquid water Ice
a1 6.111 76750 6.109 526 65

a2 0.443 986 062 0.501 948366
Table 3 Variation of H i with Tfor constant H = 1
a3 0.143053301 x i o - ' 0.186288989~i o - '
a4 0.265027 242 x 10 - 0.403 488 906 x 10 - T("C) H Hi
a5 0.302246994 x 0.539797852 x
a6 0.203886313 x 0.420713632 x 0 1.o 1.o
a7 0.638780966~1 0 - l ~ 0.147271071 x 1 0 - ~ - 10 1.o 1.10
- 20 1 .o 1.22
After Flatau PJ, Walko RL, Cotton WR (1992) Polynomial fits to - 30 1 .o 1.34
saturation vapor pressure. Journal of Applied Meteorology 31: - 40 1.o 1.47
1507-1 513.
HUMIDITY VARIABLES 941
where E = M,/Md = 0.622 is the ratio of the molec- or equivalently

ular weight of water to the molecular weight of the e = es(TD) 1131
mixture of dry air gases, Pd is the density of dry air, and
p is total atmospheric pressure. We can determine the dew point temperature by
A value of the saturation mass mixing ratio, w,, is inverting either eqn [12] or eqn [13], which can be
given by done using eqns [3] and [6]. Although the dew point
e, temperature is expressed in Kelvin, the dew point
w, = &- 171 temperature is a measure not of temperature, but of
P - e,
atmospheric humidity. The term T - TD is called the
Expressions for e and e, from eqns [6] and [7] can be dew point depression. Dew point depression is in-
substituted into eqn [4]. Since p>>e and p<<e,, versely proportional to relative humidity; a relative
W,
HZ- 181 humidity of 100% corresponds to a dew point
WS depression of zero.
is an approximate definition of the relative humidity. Analogously to the dew point temperature, we
The specific humidity is defined as the ratio of the define the frost point temperature as the temperature
mass of water vapor to the total mass of atmospheric at which ice saturation occurs. The frost point
gases. The water vapor mixing ratio can be related to temperature, TF,is thus defined as
the specific humidity, q,, as
e = esi (TF) ~ 4 1
m, e - wv
qv = = E 191 or equivalently as
md+mv p - ( l - ~ ) e 1+w,
wv = wsi (TF) [I51
Since both wv and qv are always smaller than 0.04 in
the Earth's atmosphere, q,xw,. where wsi is the saturation mass mixing ratio over ice.
The total mass of water vapor in a column of unit Note that for a given atmospheric mixing ratio,
cross-sectional area extending from the surface to the T F< T D ,since esi( T )< e, ( T ) .
top of the atmosphere is called the precipitable water,
W, :
oc
ExampIe CaICulations
w, = P,dZ Air has temperature of 30°C, a relative humidity of
SO%, and a pressure of 1000 hPa. This information
where z is height above the Earth's surface. The term can be used to determine each of the humidity
'precipitable water' is used because, if all the vapor in variables for these conditions:
the column were to be condensed into a pool of liquid
at the base of the column, the depth of the pool would partial pressure of water vapor, e = 21.2 hPa
be equal to Wv/pI, where pI is the density of liquid water vapor mixing ratio, w, = 0.0135
water. To obtain a relationship between precipitable specific humidity, q, = 0.0133
water and specific humidity, we can write eqn [ 101 in
terms of pressure by incorporating the hydrostatic dew point temperature, TD= 291.8 K
equation:
See also
Climate Variability: Glacial, Interglacial Variations.
where p o is the surface pressure, corresponding to Satellite Remote Sensing: Water Vapor. Thermo-
dynamics: Moist (Unsaturated) Air; Saturated Adiabatic
z = 0 and pa is the air density. Processes.
The dew point temperature, denoted by TD, is
defined as the temperature to which moist air must be
cooled, the atmospheric pressure and water vapor Further Reading
mixing ratio remaining constant, such that it becomes
just saturated with respect to water. Obviously, at T D Curry JA and Webster PJ (1999) Thermodynamics of
the mixing ratio of the air becomes its saturation Atmospheres and Oceans. San Diego: Academic Press.
Flatau PJ, Walko RL and Cotton WR (1992)Polynomial fits
mixing ratio
to saturation vapor pressure. Journal of Applied Meteo-
1121 rology 31: 1507-1513.
942 HURRICANES
F D Marks, Hurricane Research Division, Miami, FL, intensify to a hurricane, while others do not, are not
USA well understood. Neither is it clear why some tropical
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. cyclones become major hurricanes, while others do
not. Major hurricanes produce 80-90% of the United
States hurricane-caused damage despite accounting
Introduction
for only one-fifth of all landfalling tropical cyclones.
‘Hurricane’ is the term used in the Western Hemi- Only two category 5 hurricanes made landfall on the
sphere for one of the general class of strong tropical mainland United States (Florida Keys 1935 and
cyclones, including western Pacific typhoons and Camille 1969). Recent major hurricanes to make
similar systems, that are known simply as cyclones in landfall on the United States were Hurricanes Bonnie
the Indian and southern Pacific Oceans. A tropical and Georges in 1998, and Bret and Floyd in 1999.
cyclone is a low-pressure system which derives its As with large-scale extratropical weather systems,
energy primarily from evaporation from the sea in the the structure and evolution of a tropical cyclone is
presence of 1-minute sustained surface wind speeds dominated by the fundamental contradiction that
> 17m s - and the associated condensation in con- while the airflow within a tropical cyclone represents
vective clouds concentrated near its center. In contrast, an approximate balance among forces affecting each
midlatitude storms (low-pressure systems with air parcel, slight departures from balance are essential
associated fronts) get their energy primarily from for vertical motions and resulting clouds and precip-
the horizontal temperature gradients that exist in itation, as well as changes in tropical cyclone intensity.
the atmosphere. Structurally, the strongest winds As in extratropical weather systems, the basic vertical
in tropical cyclones are near the Earth’s surface balance of forces in a tropical cyclone is hydrostatic
(a consequence of being ‘warm-core’ in the tropo- except in the eyewall, where convection is superim-
sphere), while the strongest winds in midlatitude posed on the hydrostatic motions. However, unlike in
storms are near the tropopause (a consequence of extratropical weather systems, the basic horizontal
being ‘warm-core’ in the stratosphere and ‘cold-core’ balance in a tropical cyclone above the boundary layer
in the troposphere). ‘Warm-core’ refers to being is between the sum of the Coriolis ‘acceleration’ and
warmer than the environment at the same pressure the centripetal ‘acceleration’, balanced by the hori-
surface. zontal pressure gradient force. This balance is referred
A tropical cyclone with the highest sustained wind to as gradient balance, where the Coriolis ‘accelera-
speeds between 17 and 32 m s - is referred to as a tion’ is defined as the horizontal velocity of an air
tropical storm, whereas a tropical cyclone with parcel, u, times the Coriolis parameter, f . (f is the
sustained wind speeds 2 33 m s - is referred to as a Coriolis parameter (f = 2Rsin q5), where R is the
hurricane or typhoon. Once a tropical cyclone has angular velocity of the Earth (7.292 x 10 - s - and
sustained winds 2 50 m s - it is referred to as a major $Jis latitude. The Coriolis parameter is zero at the
hurricane or super typhoon. In the Atlantic and equator and 2R at the pole.) Centripetal ‘force’ is
eastern Pacific Oceans hurricanes are also classified defined as the acceleration on a parcel of air moving in
by the damage they can cause using the Saffir-Simpson a curved path, directed toward the center of curvature
scale (Table 1). of the path, with magnitude u 2 / r , where z, is the
The Saffir-Simpson scale categorizes hurricanes on horizontal velocity of the parcel and r the radius of
a scale from 1to 5, with 1the weakest and 5 the most curvature of the path. The centripetal force alters the
intense. Major hurricanes correspond to categories 3 original two-force geostrophic balance and creates a
and higher. The reasons that some disturbances nongeostrophic gradient wind.
Table 1 Saffir-Simpson scale of hurricane intensity
Category Pressure (hPa) Wind (m s - ’) Storm surge (m) Damage
> 980 33-42 1 .O-1.7 Minimal

979-965 43-49 1.8-2.6 Moderate
964-945 50-58 2.7-3.8 Extensive
944-920 59-69 3.9-5.6 Extreme
< 920 2 70 25.7 Catastrophic
HURRICANES 943
Figure 1 NOAA-14AVHRR multispectralfalse color image of Hurricane Floyd at 2041 UTC, 13 September 1999 about 800 km east of
southern Florida. (Photo courtesy of NOAA Operationally Significant Event Imagery website: http://www.osei.noaa.gov/.)
The inner region of the tropical storm, termed the the axial wind velocity, f the Coriolis parameter, and r
cyclone ‘core’, contains the spiral bands of precipita- the radius from the storm center. An approximate
tion, the eyewall, and the eye that characterize tropical breakdown of regimes is: Ro< 1, geostrophic flow;
cyclones in radar and satellite imagery (Figure 1).The Ro > 1, gradient flow; and Ro > 50, cyclostrophic
primary circulation - the tangential or swirling wind - flow. When the Rossby number significantly exceeds
in the core becomes strongly axisymmetric as the unity, the balance in the core becomes more cyclo-
cyclone matures. The strong winds in the core, which strophic, where the pressure gradient force is almost
occupies only 1 4 % of the cyclone’s area, threaten completely balanced by the centrifugal ‘force’. The
human activities and make the cyclone’s dynamics time scales are such that air swirling around the center
unique. In the core, the local Rossby number is always completes an orbit in much less than a pendulum day
> 1 and may be as high as 100. The Rossby number (defined as l / f ) .
indicates the relative magnitude of centrifugal ( U / T ) When the atmosphere is in approximate horizontal
and Coriolis ( f )accelerations, Ro = V / f r ,where V is and vertical balance, the wind and mass fields are
944 HURRICANES
tightly interconnected. The distribution of a single ward). While consistent with the simple PV distribu-
mass or momentum variable may be used as a starting tion, the wind and mass fields are also in horizontal
point to infer the distribution of all other such and vertical balance. The tropical cyclone being a
variables. One such variable is potential vorticity warm-core vortex, the PV inversion dictates that the
(PV), approximately equal to the vorticity times winds that swirl about the center decrease with
the thermal stratification, which is related to the increasing height, but they typically fill the depth of
three-dimensional mass and momentum fields the troposphere. If the PV reaches values 2 10 P W ,
through an inverse second-order Laplacian-like oper- the inner region winds can become intense, as in
ator. The benefit of such a relationship is that PV Hurricane Gloria (Figure 3). Gloria had PV values
variations in a single location are diagnostically exceeding 500 P W just inside the radius of maximum
related to variations in mass and wind fields at a winds of 15 km where the axisymmetric mean tan-
distance. Areas of high PV correspond locally to low gential winds exceeded 65 m s '.
mass, or cyclones, while areas of low PV correspond to Many features in the core, however, persist with
anticyclones. little change for (pendulum) days (mean life span of a
Typical extratropical weather systems contain tropical cyclone is about 5-10 days). Because these
high PV values around 0.5 x 1Op6rn2s-'Kkg-' long lifetimes represent tens or hundreds of orbital
(0.5PW) to 5 P W , whereas typical values in the -
periods ( 1 h), the flow is nearly balanced. Moreover,
tropical cyclone core are 2 10 P W . Figure 2 shows the
wind and mass fields associated with an idealized
axially symmetric tropical cyclone PV anomaly with
at winds > 3 5 m s p 2 , the local Rossby radius of
deformation is reduced from its normal lo3km to
a value comparable with the eye radius. The Rossby
-
the PV concentrated near the surface rather than in a radius of deformation is the ratio of the speed of the
vertical column. The cyclonic anomaly (positive in the relevant gravity wave mode and the local vorticity, or,
Northern Hemisphere) is associated with a cyclonic equivalently, the ratios of the Brunt-Vaisala and
circulation that is strongest at the level of the PV inertial frequencies. This scale indicates the amount
anomaly near the surface, and decreases upward. of energy that goes into gravity waves compared with
Temperatures are anomalously warm above the PV inertial acceleration of the wind. In very intense
anomaly (isentropic surfaces are deflected down- tropical cyclones, the eye radius may approach the
15 km
10 km
5 km
0
0 2000 km
i 1 1 5 km
I 0 2000 km
Figure 2 v (gradientwind) (m s-' ) and 0' (perturbationpotentialtemperature) (K, top panel); and h' (geopotentialheight perturbation)
(dm) and [ / f (bottom panel) for a warm core, lower cyclone.The tropopauselocationis denoted by the bold solid line, and the label0 on the
horizontalaxis indicatesthe core (and axis of symmetry) of the disturbance.The equivalentpressure deviationat the surface in the center
of the vortex is - 31 hPa. (Reproduced with permission from Thorpe AJ (1986) Synoptic scale disturbances with circular symmetry.
Monthly Weather Review 114: 1384-1 389; 0American MeteorologicalSociety.)
HURRICANES 945
250
350
-a 450
(I)
L
3
f 550
!?
650
750
Distance from vortex centre (km)
Figure 3 Radial-heightcross-section of symmetric potential vorticity for Hurricane Gloria, 24 September 1985. Contours are 0.1 PVU.
Values in the data-sparse region, within 13 km of vortex center, are not displayed. (Reproduced with permission from Shapiro LJ and
FranklinJL (1 995) Potentialvorticity in Hurricane Gloria. Monthly Weather Review1 23: 1465-1 475; C American MeteorologicalSociety.)
depth of the troposphere (15km), making the aspect Ocean the maximum numbers of tropical cyclones
ratio unity. Thus, the dynamics near the center of a occur in August through October, while in the South
tropical cyclone are so exotic that conditions in the Pacific and Australia regions the maxima are in
core differ from the Earth's day-to-day weather as February and March. In the South Indian Ocean, the
much as the atmosphere of another planet does. peak activity occurs in June. In the western North
Pacific, Bay of Bengal, and South Indian Ocean regions
tropical cyclones may occur in any month, while the
Climatology other regions at least one tropical cyclone-free month
occurs per year. For example, in the North Atlantic,
There are 80-90 tropical cyclones worldwide per year, there has never been tropical cyclone activity in
with the Northern Hemisphere having more tropical January.
cyclones than the Southern Hemisphere. Table 2 Some general conclusions can be drawn from the
shows that of the 80-90 tropical cyclones, 45-50 global distribution of tropical cyclone locations
reach hurricane or typhoon strength and 20 reach (Figure 4A). Tropical cyclone formation is confined
major hurricane or super typhoon strength. The to a region approximately 30"N and 30"S, with
western North Pacific (27 tropical cyclones), eastern 87% of them located within 20" of the Equator. There
North Pacific (17 tropical cyclones), south-west Indi- is a lack of tropical cyclones near the Equator, as well
an Ocean (10 tropical cyclones), Australiahouth-west as in the eastern South Pacific and South Atlantic
Pacific (10 tropical cyclones), and North Atlantic (10 basins. From these observations there appear to be at
tropical cyclones) are the major tropical cyclone least five necessary conditions for tropical cyclone
regions. There are also regional differences in the development.
tropical cyclone activity by month with the majority of
the activity in the summer season for each basin. 0 W a r m sea surface temperature (SST) and large
Hence, in the Pacific, Atlantic, and North Indian mixed-layer depth (i.e., the thickness of the mixed
946 HURRICANES
Table 2 Mean annual frequency, standard deviation (G) and percentage of global total of the number of tropical storms (winds
2 17 m s- I),
hurricane-force tropical cyclone (winds 2 3 3 m s- I),
and major hurricane-force tropical cyclone (winds 2 5 0 m s - I ) . Dates
in parentheses provide the nominal years for which accurate records are currently available
Tropical cyclone basin Tropical storm % of Hurricane annual % of Major hurricane % of total
annual total frequency (0) total annual
frequency (G) frequency (G)
Atlantic (1944-00) 9.8 (3.0) 11.4 5.7 (2.2) 12.1 2.2 (1.5) 10.9
NE Pacific (1970-00) 17.0 (4.4) 19.7 9.8 (3.1) 20.7 4.6 (2.5) 22.9
NW Pacific (1970-00) 26.9 (4.1) 32.1 16.8 (3.6) 35.5 8.3 (3.2) 41.3
N Indian (1970-00) 5.4 (2.2) 6.3 2.2 (1.8) 4.6 0.3 (0.5) 1.5
SW Indian (30-100" E) 10.3 (2.9) 12.0 4.9 (2.4) 10.4 1.8(1.9) 9.0
(1969-00)
AustralianKE Indian 6.5 (2.6) 7.5 3.3 (1.9) 7.0 1.2(1.4) 6.0
(100-1 42" E) (1969-00)
AustralianKW Pacific 10.2 (3.1) 11.8 4.6 (2.4) 9.7 1.7(1.9) 8.5
(142" E) (1969-00)
Global (1970-00) 86.1 (8.0) 47.3 (6.5) 20.1 (5.7)
layer, defined as the depth of the sharp temperature speed as the environmental flow in which it is
inversion (also referred to as the thermocline) embedded then its heating remains over the distur-
between the cooler bottom water and the warmer bance center. However, if it is moving slower than
near surface water). Numerous studies suggest a the mean wind at upper levels then the heating in the
minimum SST criterion of 26°C for development. upper troposphere is carried away by the mean flow.
The warm water must also have sufficient depth Recent analysis suggests that the effect of shear is to
(i.e., 50m). Comparison of Figures 4A and 4B, the force the convection into an asymmetric pattern
annual mean global SST, shows the strong correla- such that the convective latent heat release forces
tion between regions where the SST is > 26°C and flow asymmetry and irregular motion rather than
annual tropical cyclone activity. An SSTof >26°C is intensification of the symmetric vortex. Thus, if the
sufficient but not necessary for tropical cyclone vertical shear is too strong ( > 16 m s-') then exist-
activity, as is evidenced by the regions with tropical ing tropical cyclones are ripped apart and new ones
cyclone activity when the SST <26"C. Some of the cannot form.
discrepancy exists because storms that form over Low atmospheric static stability. Static stability is
regions where the SST is >26"C are advected the ability of a fluid to become turbulent (unstable)
poleward during their life cycle. However, tropical or laminar (stable) due to the effects of buoyancy
cyclones are observed to originate over regions neglecting all other inertial effects of motion. The
where the SST <26"C. These occurrences are not troposphere must be potentially unstable to sustain
many, but the fact that they exist suggests that other convection for an extended period. Typically meas-
factors are important. ured as the difference between the equivalent po-
Background Earth vorticity. Tropical cyclones do tential temperature, Be, at the surface and 500 hPa,
not form within 3" of the Equator. The Coriolis instability must typically be > 10 K for convection
parameter vanishes at the Equator and increases to to occur. This value is usually satisfied over tropical
extremes at the poles. Hence, a threshold value of oceans.
Earth vorticity ( f )must exist for a tropical cyclone 0 Tropospheric humidity. The higher the midlevel
to form. However, the likelihood of formation does humidity, the longer a parcel of air can remain
not increase with increasing f . Thus, nonzero Earth saturated as it entrains the surrounding air during its
vorticity is necessary, but not sufficient to produce ascent. Vigorous convection occurs if the parcel
tropical cyclones. remains saturated throughout its ascent. A relative
Low vertical shear of the horizontal wind. In order humidity of 50-60% at lower to midlevels
for tropical cyclones to develop, the latent heat (700-500 hPa) is often sufficient to keep a parcel
generated by the convection must be kept near the saturated during ascent. This condition is regularly
center of the storm. Historically, shear was thought evident over tropical oceans.
to 'ventilate' the core of the cyclone by advecting the
warm anomaly away. The ventilation argument These conditions are usually satisfied in the summer
suggests that if the storm travels at nearly the same and fall seasons for each tropical cyclone basin.
HURRICANES 947
60"N
55O N
50" N
45" N
40"N
35" N
30° N
25" N
20" N
15" N
10" N
05O N
0
05" S
10" s
15" S
200 s
25"S
30" S
.............. .......
............ ........
35" s
40"S
................ .........
..........
45"
50° S
s
.......................... ................
55" s
(A) """,",IPE 50"E 70"E 9O"E 110'E 130"E 150"E 1 7 O o E O 1 7 0 " W 150"W 130"W 11O"W 9O"W 70"W 50"W 30"W 1O"W
Figure 4 (A) Frequency of tropical cyclones per 100 years within 140 km of any point. Solid triangles indicate maxima, with values
shown. Period of record is shown in boxes for each basin. (B) Annual sea surface temperature distribution ("C).
However, even when all of the above conditions are ical storm strength, whereas in the inactive phase
favorable, tropical cyclones don't necessarily form. In practically none of the disturbances intensify. The two
fact, there is growing evidence for significant inter- hurricanes and unnamed depression in Figure 5
annual variability in tropical cyclone activity, where represented the second 10-day active period during
numerous tropical cyclones form in a given basin over the summer of 1999. An earlier period in mid-August
a week to 10 days, followed by 2-3 weeks with little or also resulted in the development of three hurricanes
no tropical cyclone activity. Figure 5 shows just such (Brett, Cindy, and Dennis), two of which were major,
an active period in the Atlantic basin in mid-September as well as a tropical storm (Emily).There is speculation
1999, where two hurricanes (Floyd and Gert), both that the variability is related to the propagation of a
major, and an unnamed tropical depression formed global wave. Because the SST, static stability, and
within a few days of each other. During these active Earth vorticity don't vary that much during the season,
phases almost every disturbance makes at least trop- the interannual variability is most likely related
948 HURRICANES
Figure 5 GOES multispectral false color image of Hurricanes Floyd and Gert and an unnamed tropical depression at 1935 UTC, 13
September 1999. (Photo courtesy of NOAA Operationally Significant Event Imagery website: http://www.osei.noaa.gov/.)
to variations in tropospheric relative humidity and formation of tropical cyclones. Another effect of
vertical wind shear. the El Nifio circulation is warmer SST in the eastern
It has long been recognized that the number of South Pacific. During such years, tropical cyclones
tropical cyclones in a given region varies from year to form closer to the Equator and farther east. Regions
year. The exact causes of this remain largely specula- such as French Polynesia, which are typically
tive. The large-scale global variations in atmospheric unfavorable for tropical cyclones owing to a strong
phenomena such as the El Nifio Southern Oscillation upper-level trough, experience numerous tropical
(ENSO) and the Quasi-Biennial Oscillation (QBO) cyclones. The eastern North Pacific is also affected
appear to be related to annual changes in the frequency by the El Nifio through a displacement of the ITCZ
of tropical cyclone formation, particularly in the south to near 5"N. Additionally, the warm ocean
Atlantic Ocean. The ENSO phenomenon is charac- anomaly of El Niiio extends to near 20"N, which
terized by warmer SSTs in the eastern South Pacific enhances the possibility of tropical cyclone formation.
and anomalous winds over much of the equatorial The result is an average increase of two tropical
Pacific. It influences tropical cyclone formation in the cyclones during El Niiio years. Cyclones also develop
western North Pacific, South Pacific, and even the closer to the Equator and farther west than during a
North Atlantic. normal year.
During the peak phase of the ENSO, often referred The QBO is a roughly 2-year oscillation of the
to as El Nifio (which usually occurs during the months equatorial stratosphere (30-50 hPa) winds from east-
July-October), anomalous westerly winds near the erly to westerly and back. The phase and magnitude of
Equator extend to the dateline in the western North QBO are associated with the frequency of tropical
Pacific acting to enhance the intertropical convergence cyclones in the Atlantic. Hurricane activity is more
zone (ITCZ)in this area, making it more favorable for frequent when the 30-hPa stratospheric winds are
HURRICANES 949
westerly. The exact mechanism by which the QBO For a tropical cyclone to occur, evaporation must
affects tropical cyclones in the troposphere is not clear; speed up and the equilibrium enthalpy at the sea
however, there are more North Atlantic tropical surface temperature must rise through a lowering of
cyclones when the QBO is in the westerly phase than the surface pressure. Tropical cyclones are thus finite-
when it is in the easterly. amplitude phenomena. They do not grow by some
linear process from infinitesimal amplitude. The
normal paradigm of searching for the most rapidly
growing unstable linear mode used to study midlati-
Tropical Cyclogenesis tude cyclogenesis through baroclinic instability fails
Enthalpy is a thermodynamic state function defined here. The surface wind has to exceed roughly 20 m s -
for an ideal gas as the temperature times the specific before evaporation can prevail against downdraft
heat at constant pressure plus a constant. For a system cooling.
like the atmosphere which consists of a mixture of How then do tropical cyclones reach the required
components the total enthalpy is the mass-weighted finite amplitude? The answer seems to lie in the
sum of the enthalpies of each component. Thus, the structure of tropical convection. As explained previ-
total enthalpy for a system consisting of a mixture of ously, behind a squall line the lower troposphere
dry air, water vapor, and liquid water is defined as a (below the 0°C isotherm at -5 km) is dominated by
constant plus the temperature times the sum of the precipitation-driven downdrafts which lie under the
specific heats at constant pressure for each component ‘anvil’ of nimbostratus and cirrostratus that spreads
times the masses of each component, respectively. In behind the active convection. Above 5 km, a combi-
an adiabatic, reversible process, the total enthalpy is nation of differential radiative fluxes at the top and
conserved, although the component enthalpies may bottom of the anvil and residual condensational
not be due to the exchange of enthalpy between heating from the main updraft maintains weak rising
components in phase changes. Most of the energy motion. This updrafts-over-downdrafts arrangement
needed for tropical cyclones to form and maintain requires horizontal convergence centered near 5 km
themselves is realized through the difference in en- altitude to maintain mass continuity. The important
thalpy between the warm near-surface waters of the kinematic consequence is formation of patchy shallow
tropical ocean and the tropospheric column. The vortices near the altitude of the 0°C isotherm. The
process of bringing the late-summer tropical tropo- typical horizontal scales of these ‘mesovortices’ are
sphere into thermodynamic equilibrium with the sea tens to hundreds of kilometers. If they were at the
surface at 28-30°C, mainly through the irreversible surface or if their influence could be extended down-
energy transfer from the ocean to the air by evapora- ward to the surface then they would be the means to
tion, can produce hydrostatic pressures as low as the get the system to the required finite amplitude.
minimum sea-level pressures of the most intense The foregoing reasoning defines the important
tropical cyclones. Thus, much of the tropical oceans unanswered questions: (1) how do the midlevel
contain enough moist enthalpy to support a major mesovortices extend their influence to the surface,
hurricane. and (2) what are the detailed thermodynamics at the
Throughout most of the Trade Wind regions, air-sea interface during this process? Leading hypoth-
gradual subsidence causes an inversion that traps eses for (1)are related to processes that can increase
water vapor in the lowest kilometer. Sporadic convec- the surface vorticity through changes in static stability
tion (often in squall lines) that breaks through the and momentum mixing, both horizontally and verti-
inversion exhausts the moist enthalpy stored in the cally. However, the answers to these questions await
near-surface boundary layer quickly, leaving a wake of new measurements that are just becoming available
cool, relatively dry air. This air comes from just above through improved observational tools.
the inversion and is brought to the surface by down-
drafts driven by the weight of hydrometeors and
cooling due to their evaporation. If the squall line does Basic Structure
not keep moving it quickly runs out of energy. A day, or
Primary and Secondary Circulations
even several days, may pass before normal fair-
weather evaporation can restore the preexistent moist Inner core dynamics have received a lot of attention
enthalpy behind the squall. The reasons why squall over the last 40 years through aircraft observations of
line convection generally fails to produce hurricanes the inner core structure. These observations show that
lie in the limited amount of enthalpy that can be stored the tropical cyclone inner core dynamics are dominat-
in the sub-inversion layer and the slow rate of ed by interactions between ‘primary’ (horizontal
evaporation under normal wind speeds in the trades. axisymmetric), ‘secondary’ (radial and vertical)
950 HURRICANES
circulations, and a wavenumber one asymmetry As the air converges towards the eye and is lifted in
caused by the storm motion. The primary circulation convective clouds that surround the clear eye, it
is so strong in the cyclone core that it is possible to ascends to the tropopause (the top of troposphere,
consider axisymmetric motions separately, if account where temperature stops decreasing with height). As
is taken of forcing by the asymmetric motions. The shown in Figure 6, the convective updrafts in the
primary circulation is in near-gradient balance, and eyewall turn the latent heat into sensible heat through
evolves when heat and angular momentum sources the latent heat of condensation to provide the buoy-
(often due to asymmetric motions) force secondary ancy needed to loft air from the surface to tropopause
circulations, which in turn redistribute heat and level. The updraft entrains midlevel air, promoting
angular momentum. mass and angular momentum convergence into the
Figure 6 shows that the primary circulation is core. It is the midlevel inflow that supplies the excess
sustained by the secondary circulation that consists of angular momentum needed to spin up the vortex. The
frictional inflow that loses angular momentum to the thermodynamics of a storm can be modeled as an
sea as it gains moist enthalpy. (Angular momentum idealized heat engine, running between a warm heat
M = Vr + f?/2, where V is the tangential wind reservoir, the sea, at around 300K, and a cold
velocity, f the Coriolis parameter, and r the radius reservoir, 15-18 km up in the tropical troposphere,
from the storm center.) The inflow picks up latent heat at about 200 K. The net energy realized in the whole
through evaporation, and exchanges sensible heat process is proportional to the difference in tempera-
with the underlying ocean, as it spirals into lower ture between the ocean and the upper troposphere.
levels of the storm under influence of friction. The Storm-induced upwelling of cooler water reduces
evaporation of sea spray adds moisture to the air, while ocean SST by a few degrees, which has a considerable
at the same time cooling it. This process is important in effect on the storm’s intensity.
determining the intensity of a tropical cyclone. Near As shown in Figure 7, the secondary circulation also
the vortex center, the inflow turns upward and brings controls the distribution of hydrometeors and radar
the latent heat it acquires in the boundary layer into reflectivity. It is much weaker than the primary
the free atmosphere. Across the top of the boundary circulation except in the anticyclonic outflow, where
layer, turbulent eddies cause significant downward the vortex is also much more asymmetric. Precipita-
flux of sensible heat from the free atmosphere to the tion-driven convective updrafts form as hydrometeors
boundary layer. The energy source for the turbulent fall from the outward sloping updraft. Condensation
eddies is mechanical mixing caused by the strong in the anvil causes a mesoscale updraft above the 0°C
winds. The eddies are also responsible for downward isotherm and precipitation loading by snow falling
mixing of angular momentum. Hence, these turbulent from the overhanging anvil causes a mesoscale down-
eddy fluxes fuel the storm. draft below 0°C isotherm. The melting level itself
Figure0 Schematic of the secondary circulation thermodynamics. (Reproduced with permission from Willoughby HE (1999) Hurricane
heat engines. Nature 401 : 649-650; 0Macmillan Magazines Ltd.)
HURRICANES 951
primary circulation cause updrafts to pass through the

convective heat sources, because the path of least
resistance for the warmed air lies along constant
angular momentum surfaces. Similarly, horizontal
temperature gradients due to vertical shears cause
the horizontal flow to pass through momentum
sources, because the path of least resistance lies along
isentropes (potential temperature or 0 surfaces).
Although the flow lies generally along the angular
momentum or isentropic surfaces, it has a small
component across them. The advection by this com-
ponent, not the direct forcing, is the mechanism by
which the primary circulation evolves.
Some of the most intense tropical cyclones exhibit
‘concentric’ eyewalls, Le., two or more eyewall struc-
tures centered at the circulation center of the storm. In
much the same way as the inner eyewall forms,
convection surrounding the eyewall can become
organized into distinct rings. Eventually, the inner
eye begins to feel the effects of the subsidence resulting
from the outer eyewall, and the inner eyewall weak-
ens, to be replaced by the outer eyewall. The pressure
rises due to the destruction of the inner eyewall are
usually more rapid than the pressure falls, due to the
intensification of the outer eyewall, and the cyclone
itself weakens for a short period of time. This
mechanism, referred to as the eyewall replacement
cycle, often accompanies dramatic changes in storm
intensity. The intensity changes are often associated
with the development of secondary wind maxima
outside the storm core.
A good example of contracting rings of convection
effecting the intensification of a hurricane is shown in
Figure 7 (A) Schematic of the radius-height circulation of the Figure 8 for Hurricane Gilbert on 14 September 1988.
inner core of Hurricane Alicia (1983). Shading depicts the reflec-
Two convective rings, denoted by intense radar
tivity field, with contours of 5, 30, and 35dBZ. The primary
circulation (m s - ’ ) is depicted by dashed lines and the secondary reflectivity, are evident in Figure 8A. The outer ring
circulation by the wide hatched streamlines. The convective is located near 80-90 km radius and the inner one at
downdrafts are denoted by the thick solid arrows, while the 10-12km radius. Figure 8B shows that both are
mesoscale up- and downdrafts are shown by the broad arrows. (B) associated with maxima in tangential wind and
Schematic plan view of the low-level reflectivity field in the inner
vorticity. Figure 9 shows that in the ensuing 12-24 h
core of Hurricane Alicia superimposed with the middle of the three
hydrometeor trajectories in (A). Reflectivity contours in (b) are 20 the storm filled dramatically. However, it is not clear
and 35 dBZ. The storm center and direction are also shown. In (A) how much of the filling was caused by the storm
and (B) the hydrometeor trajectories are denoted by dashed and moving over land and how much by the contracting
solid lines labeled 0-1-2-3-4 and 0’-1’-2’. (Reproduced with outer ring and decaying inner ring of convective
permission from Marks FD and Houze RA (1987) Inner core
activity.
structure of Hurricane Alicia from airborne Doppler radar observa-
tions. Journal of the Atmospheric Sciences 44: 1296-1 317; A process has been proposed whereby: (1)nonlinear
OAmerican Meteorological Society.) balanced adjustment of the vortex to eddy heat and
angular momentum sources generated by some envi-
ronmental interaction in the storm’s periphery pro-
marks the height of maximum mass convergence. duces an enhanced secondary circulation; (2) a
Inside the eye, dynamically driven descent and mo- secondary wind maximum develops in response; and
mentum mixing leads to substantial pressure falls. ( 3 ) the wind maximum contracts as a result of
In order for the primary circulation to intensify, the differential adiabatic warming associated with the
flow cannot be in exact balance. Vertical gradients of convective diabatic heating in the presence of a inward
angular momentum due to vertical shears of the radial gradient of inertial stability. Under these
952 HURRICANES
circumstances, understanding the intensification of

the tropical cyclone reduces to determining what
mechanisms can produce an enhanced secondary wind
maximum.
Inner core - eyewall and eye The most recognizable

feature found within a hurricane is the eye (Figure 10).
It is found at the center and is typically between
20-50km in diameter. The eye is the focus of the
hurricane, the point about which the primary circu-
lation rotates and where the lowest surface pressures
are found in the storm. The eye is a roughly circular
area of comparatively light winds and fair weather
found at the center of strong tropical cyclones.
Although the winds are calm at the axis of rotation,
strong winds may extend well into the eye. As seen in
Figure 10, there is little or no precipitation and
sometimes blue sky or stars can be seen. The eye is
the region of warmest temperatures aloft - the eye
temperature may be 210°C warmer at an altitude
of 12 km than the surrounding environment, but only
0-2°C warmer at the surface.
The eye is surrounded by the eyewall, the roughly
circular area of deep convection associated with the
0 20 40 60 80 100 120
up-branch of the secondary circulation and the highest
(B) Distance from hurricane center (km)
surface winds. The eye is composed of air that is slowly
Figure 8 (A) Composite horizontal radar reflectivity of Hurricane sinking and the eyewall has a net upward flow because
Gilbert for 0959-1025 UTC, 14 September 1988; the domain is of many moderate -occasionally strong -updrafts and
360 km x 360 km, marked every36 km. The line through the center
is the WP-3D aircraft flight track. (B) Profiles of flight-level angular
downdrafts. The eye's warm temperatures are due to
velocity (0, solid) tangential wind (short dash), and smoothed warming by compression of the subsiding air. Most
relative vorticity (c, long dash) along the southern leg of the flight soundings taken within the eye are similar to that for
track shown in (A). (Reproduced with permission from Kossin JP, Hurricane Hugo in Figure 11. They show a low-level
Schubert WH, and Montgomery MT (2000) Unstable interactions layer which is relatively moist, with an inversion
between a hurricane's primary eyewall and a secondary ring of
enhanced vorticity. Journal of the Atmospheric Sciences 57: 3893-
above, suggesting that the sinking in the eye typically
3917; 8 American Meteorological Society.) does not reach the ocean surface, but instead gets only
within 1-3 km of the surface. An eye is usually present
only in hurricane-strength tropical cyclones.
The general mechanisms by which the eye and
eyewall are formed are not fully understood, although
observations shed some light on the problem. The
calm eye of the tropical cyclone shares many qualita-
tive characteristics with other vortical systems such as
tornadoes, waterspouts, dust devils, and whirlpools.
Given that many of these lack a change of phase of
water (i.e., no clouds and diabatic heating are
involved), it may be that the eye feature is a funda-
mental component to all rotating fluids. It has been
12 13 14 15 16 17
hypothesized that supergradient wind flow (i.e.,
September (1988)
swirling winds generating stronger centrifugal 'force'
Figure 9 Hurricane Gilbert's minimum sea-level pressure than the local pressure gradient can support) present
(MSLP) and radii of the inner and outer eyewalls as a function of near the radius of maximum winds causes air to be
time, September 1988. Solid blocks at bottom indicate times over centrifuged out of the eye into the eyewall, thus
land. (Reproduced with permission from Black ML and Willoughby
HE (1999) The concentric eyewall cycle of Hurricane Gilbert. accounting for the subsidence in the eye. However,
Monthly Weather Review 120: 947-957; 8 American Meteoro- others found that the swirling winds within several
logical Society.) tropical cyclones were within 1 4 % of gradient
HURRICANES 953
Figure 10 Eyewall of Hurricane Georges, 1945 UTC, 19 September 1998. (Photo courtesy of M. Black, NOANOAWAOML Hurricane
Research Division.)
balance. It may be thought that the amount of Another feature of tropical cyclones that probably
supergradient flow needed to cause such centrifuging plays a role in forming and maintaining the eye is the
of air is only on the order of a couple of percent and eyewall convection. As shown in Figure
- 12, convec-
thus difficult to measure. tion in developing tropical cyclones is organized into
HUGO 21463 HUGO 21463

1839UTC 15SEP89 17.4 N 54.8 W 1839UTC 15SEP89 17.4 N 54.8 W
-20 -1 0 0 10 TUU I l l 1
400 0 10 20 30
Mixing ratio (gm kg-l)
I I
500
h
m
a
600
e!
3
v)
E 700
850
800
1000 1000
10 20 30 40 50 340 360 380 0 50 100 150
(A) Temperature ("C) (B) @e (K) P - PSAT( h W
Figure 11 (A) Skew T l g p diagram of the eye sounding in Hurricane Hugo at 1839 UTC, 15 September 1989, 17.4" N, 54.8"W.
Isotherms slope upward to the right; dry adiabats slope upward to the left; moist adiabats are nearly vertical curving to the left. Solid and
dashed curves denote temperature and dew point, respectively.The smaller dots denote saturation points computed for the dry air above
the inversion, and the two larger dots temperature observed at the innermost saturated point as the aircraft passed through the eyewall.
(B) Be, water vapor mixing ratio, and saturation pressure difference, P-PsAT,as functions of pressure at 2123 UTC. (Reproduced with
permission from Willoughby HE (1998) Tropical cyclone eye thermodynamics. Monthly Weather Review 126: 3189-321 1; Q American
954 HURRICANES
Q pressure falls across the band since warm air is lighter

than cold air. Because of the pressure falls on the
inside, the tangential winds around the tropical
cyclone increase, owing to an increased pressure
gradient. Eventually, the band moves toward the
center and encircles it and the eye and eyewall form.
The circulation in the eye is comparatively weak
and, at least in the mature stage, thermally indirect
(warm air descending), so it cannot play a direct role in
the storm energy production. On the other hand, the
temperature in the eye of many hurricanes exceeds that
90 100 110
which can be attained by any conceivable moist
Radial distance from storm center (km) adiabatic ascent from the sea surface, even accounting
for the additional entropy (positive potential temper-
ature, 0, anomaly) owing to the low surface pressure in
the eye (the lower the pressure, the higher the 0 at a
given altitude and temperature). Thus, the observed
low central pressure of the storm is not consistent with
that calculated hydrostatically from the temperature
distribution created when a sample of air is lifted from
a state of saturation at sea surface temperature and
pressure. The thermal wind balance restricts the
amount of warming that can take place. In essence,
the rotation of the eye at each level is imparted by the
eyewall, and the pressure drop from the outer to the
inner edge of the eye is simply that required by gradient
balance.
Because the eyewall azimuthal velocity decreases
Figure 12 (A) Schematic of the rainband in radius-height
coordinates. Reflectivity, B e , mesoscale (arrows), and convective
with height, the radial pressure drop decreases with
scale motions are shown. (B) Plan view. Aircraft track, reflectivities, altitude, requiring, through the hydrostatic equation,
cells, stratiform precipitation, 150 m flow, and Be values are shown a temperature maximum at the storm center. Thus,
(Reproduced with permission from Barnes GM, Zipser EJ, given the swirling velocity of the eyewall, the steady-
Jorgensen DP, and Marks FD (1983) Mesoscale and convective state eye structure is largely determined. The central
structure of a hurricane rainband. Journal of the Atmospheric
Sciences 40:21 25-21 37; 0American Meteorological Society.)
pressure, which is estimated by integrating the gradi-
ent balance equation inward from the radius of
maximum winds, depends on the assumed radial
profile of azimuthal wind in the eye.
long, narrow rainbands that are oriented in the same In contrast, the eyewall is a region of rapid variation
direction as the horizontal wind. Because these bands of thermodynamic variables. As shown in Figure 13,
seem to spiral into the center of a tropical cyclone, they the transition from the eyewall cloud to the nearly
are sometimes called spiral bands. The earliest radar cloud-free eye is often so abrupt that it has been
observations of tropical cyclones detected these bands, described as a form of atmospheric front. Early studies
which are typically 5-50 km wide and 100-300 km were the first to recognize that the flow under the
long. Along these bands, low-level convergence is a eyewall cloud is inherently frontogenetic. The eyewall
maximum, and therefore upper-level divergence is is the upward branch of the secondary circulation and
most pronounced. A direct circulation develops in a region of rapid ascent that, together with slantwise
which warm, moist air converges at the surface, convection, leads to the congruence of angular mo-
ascends through these bands, diverges aloft, and mentum and moist entropy (e,) surfaces. Hence, the
descends on both sides of the bands. Subsidence is three-dimensional vorticity vectors lie on Oe surfaces,
distributed over a wide area outside of the rainband, so that the moist PV vanishes. As the air is saturated,
but is concentrated in the small inside area. As the air this in turn implies, through the invertibility principle
subsides, adiabatic warming takes place, and the air applied to flow in gradient and hydrostatic balance,
dries. Because subsidence is often concentrated on the that the entire primary circulation may be deduced
inside of the band, the adiabatic warming is stronger from the radial distribution of 0, in the boundary layer
inward from the band, causing a sharp contrast in and the distribution of vorticity at the tropopause.
Next Page
HURRICANES 955
17:22:00 17:23:00 17:24:00 17:25:00 17:26:00 17:27:00 17:28:00 17:29:00

Time (UTC)
Figure 13 Time series plots of tangential wind ( Ve),radial wind ( Vr),vertical velocity (w),and Be in Hurricane Hugo at 1721-1 730 UTC,
15 September 1989. The aircraft flight track was at 450 m. Thick dashed vertical lines denote the width of the eyewall reflectivity maximum
at low levels.
In the classic semigeostrophic theory of deforma- bility is that the basic state vorticity gradient must have
tion-induced frontogenesis, the background geo- both signs in the domain of interest.) This instability
strophic deformation flow provides the advection of leads to frontal collapse as a result of radial diffusion
temperature across surfaces of absolute momentum of momentum into the eye, and also may explain the
that drives the frontogenesis, whereas in the hurricane ‘polygonal eyewalls’ where the eyewall appear on
eyewall, surface friction provides the radial advection radar to be made up of a series of line segments rather
of entropy across angular momentum surfaces. Also than as a circle. It may also explain intense mesoscale
note that the hurricane eyewall is not necessarily a vortices observed in the eyewalls of Hurricanes Hugo
front in surface temperature, but instead involves the of 1989 and Andrew of 1992.
0, distribution, which is related directly to density in Once the radial turbulent diffusion of momentum
saturated air. driven by the instability of the primary circulation
There is likely a two-stage process in eye formation. becomes important, it results in a mechanically
The amplification of the primary circulation is strong- induced, thermally indirect (warm air sinking) com-
ly frontogenetic and results, in a comparatively short ponent of the secondary circulation in the eye and
time, in frontal collapse at the inner edge of the eyewall. Such a circulation raises the vertically aver-
eyewall. (Frontal collapse is an increase in the aged temperature of the eye beyond its value in the
horizontal gradient of an airmass property, principally eyewall and allows for an amplification of the entropy
density, and the development of the accompanying distribution. Feedbacks with the surface fluxes then
features of the wind field through the secondary allow the boundary layer entropy to increase and
circulation that typify a front.) The frontal collapse result in a more rapid intensification of the swirling
leads to a dramatic transition in the storm dynamics. wind. Thus, the frontal collapse of the eyewall is an
While the tropical cyclone inner core is dominated by essential process in the evolution of tropical cyclones.
axisymmetric motions, hydrodynamic instabilities are Without it, amplification of the temperature distribu-
potential sources of asymmetric motions within the tion relies on external influences, and intensification of
core. In intense tropical cyclones the wind profile the wind field is slow. Once it has taken place, the
inside the eye is often U-shaped, in the sense that the mechanical spinup of the eye allows the temperature
wind increases outwards more rapidly than linearly distribution to amplify without external influences
with radius (Figure 13). The strong cyclonic shear just and, through positive feedback with surface fluxes,
inside the eyewall may result in a local maximum of allows the entropy field to amplify and the swirling
absolute vorticity or angular momentum, so that the velocity to increase somewhat more rapidly.
profile may actually become barotropically unstable.
(This refers to the hydrodynamic instability arising Outer structure and rainbands The axisymmetric
from certain distributions of vorticity in a two- core is characteristically surrounded by a less sym-
dimensional nondivergent flow. It is an inertial insta- metric outer vortex that diminishes into the synoptic
bility in that kinetic energy is the only form of energy ‘environment’. In the lower troposphere, the cyclonic
transferred between the current and perturbation. A circulation may extend more than 1000 km from the
well-known necessary condition for barotropic insta- center. As evident in Figure 14 the boundary between
966 HYDRAULIC FLOW
storm’s inner core dynamics, and ( 3 ) the structure of Vorticity. El Niiio and the Southern Oscillation: Obser-
the synoptic-scale upper-tropospheric environment. vation; Theory. Middle Atmosphere: Quasi-Biennial
Even if we could make a good forecast of the landfall Oscillation. Severe Storms. Tropical Meteorology:
position and intensity, our knowledge of how a Inter Tropical Convergence Zones (ITCZ).
tropical cyclone’s structure changes as it makes
landfall is in its infancy, because few hard data survive Further Reading
the harsh condition. To improve forecasts, develop-
ments to improve our understanding through obser- Elsberry R (ed.) (1995) Global Perspectives of Tropical
vations, theory, and modeling need to be advanced Cyclones. World Meteorological Organization Report
together. No. TCP-38. Geneva: WMO.
Emanuel KA (1986) An air-sea interaction theory for
tropical cyclones. 1. Steady-state maintenance. Journal
of the Atmospheric Sciences 43: 585-604.
See also Ooyama KV (1982) Conceptual evolution of the theory and
Convective Storms: Overview. Cyclogenesis. Dyna- modeling of the tropical cyclone. Journal of the Meteor-
mic Meteorology: Balanced Flows; Overview; Potential ological Society of Japan 60: 369-380.
R B Smith, Yale University, New Haven, CT, USA mathematical models of the midlatitude atmosphere
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. including the Coriolis force. On smaller scales,
following the pioneering work of R. R. Long and M.
Tepper, a variety of atmospheric phenomena have
The study of hydraulic flow is one branch of a broader been treated with hydraulic models. Cool outflowing
field of fluid mechanics dealing with the dynamics of air from thunderstorms, sea breeze fronts, and the
density stratified flow under the influence of a gravity leading edges of cold fronts all behave like gravity
field. It has a natural application to the stratified currents. Existing cool layers beneath marine inver-
atmosphere and ocean. The field of hydraulics is sions and frontal layers behave hydraulically in
distinguishable from other studies of stratified flow by mountainous areas, causing barrier jets, gap jets,
its emphasis on layered flow and the use of the hydraulic jumps, severe downslope winds, and wake
hydrostatic or longwave approximation. Typically, in eddies. Cold high terrain can generate layered cold air
hydraulic flow formulations, the fluid system is com- avalanches and katabatic winds.
posed of one or more homogeneous fluid layers, In oceanography too, hydraulic theory has found
separated by sharp interfaces with density discontinu- wide application. Basin to basin exchange of water
ities. This formulation, together with the hydrostatic masses is limited by hydraulic control at sills and
assumption, insures that the velocity is nearly uniform straits. The propagation of tidal currents and tsunamis
with height within each layer. In this way, a continuous is controlled by the long-wave speed. Turbidity
problem is reduced to a problem with one or more currents slump into the deep ocean according to
discrete layers; this results in a vast reduction in the gravity current dynamics. Coastally trapped currents
number of degrees of freedom. The possibilities for obey a modified set of hydraulic equations. Even large-
mathematical analysis, numerical computation, and scale wind-driven ocean currents are often modeled as
physical conceptualization are greatly enhanced by the two layers, defined by the thermocline, with wind
simple formulation of hydraulic theory. stress and the Coriolis force playing dominant roles.
Historically, the field of hydraulics arose out of, and The theory of hydraulic flow is based on a few
is still largely involved in, the study of natural river fundamental definitions and concepts. These are:
flow and engineering problems related to water flow in reduced gravity, the long-wave speed, Froude number,
channels. Its application to atmosphere and ocean hydraulic control, conjugate states, the hydraulic
dynamics is more recent. Beginning in the 1950s, a jump, and gravity or density current. Reduced gravity
growing number of atmospheric applications have (g’)is a measure of the effective magnitude of gravity
been suggested. O n large scales, C. G. Rossby, G. acting on layers of different density. It defined
Benton, and N. A. Phillips developed two-layer as the product of the acceleration of gravity
HYDRAULIC FLOW 967
( g = 9.81 m s P 2 ) times the relative density difference of a layer), where energy may be dissipated by
( A p / p ) between the two superposed layers; thus turbulence, but mass and momentum are conserved.
g' = ( A p / p ) g . In the compressible atmosphere, the Jumps are commonplace events in rivers, and related
relative density difference is approximately the differ- phenomena have been identified in the atmosphere
ence in potential temperature (@), so that g' = and ocean.
(A@/@)g. For example, if the air above an inversion A so-called gravity current or density current occurs
is 3" warmer than the air below, and the average when a new fluid pushes its way over or under an
potential temperature is 300 K, the effective gravity is existing denser or less dense fluid, under the influence
g' = (3/300)g = 0.0981 m s - 2 . The lon wave speed of gravity. While it resembles a hydraulic jump, a
for a single layer is given by C = (g'H)'j: where H is density current is not a sudden thickening of a pre-
the depth of the layer. If a cool marine layer of air has existing layer, but the introduction of a new fluid layer.
an effective gravity of g' = 0.1 and a depth of 1000 m, The literature is not clear on whether all layered
long gravity waves will propagate along it at a speed of formulations of stratified fluid mechanics should be
C = (O.IXIOOO)''~ = 1 0 m s - l . classified as 'hydraulics'. When friction or Coriolis
The Froude number plays a central role in hydraulic forces dominate, the term hydraulics is less often used.
theory. It is defined as the ratio of flow speed ( u ) to
long-wave speed (C),i.e., FY = u / C . It is related to the
ratio of kinetic to potential energy in a layer, but is
more useful as a measure of whether waves can move See also
upstream against the current. Flows are categorized as Convective Storms: Convective Initiation. Density Cur-
subcritical, critical, or supercritical according to rents. Downslope Winds. Dynamic Meteorology:
whether the Froude number is less than, equal to, or Overview. Gust Fronts. Katabatic Winds. LandSea
greater than unity. In supercritical flow (FY > l ) ,long Breeze. Lee Waves and Mountain Waves. Mesoscale
waves cannot move upstream or even stand steady, Meteorology: Overview. Microbursts. Mountain Me-
against the fluid flow. As the long waves are usually the teorology. Solitary Waves. Static Stability. Valley
fastest waves in the system, information carried by Winds.
waves cannot then be felt upstream of a disturbance.
The nature of a fluid response to any disturbance is Further Reading
highly sensitive to the Froude number.
Hydraulic control occurs whenever a layered flow is Baines PG (1995) Topographic Effects in Stratified Flows.
forced to transition from subcritical to supercritical Cambridge: Cambridge University Press.
Hughes RL (1989)The hydraulics of local separation in a
flow by the narrowing of a channel or valley, the rising
coastal current with application the the Kuroshio mean-
of a sill or mountain, or the alteration of some other der. Journal of Physical Oceanography 19: 1809-1820.
geometric or external parameter (e.g., Coriolis force, Jackson PL and Steyn DG (1994) Gap winds in a fjord. 2.
coastal slope, etc.). Downstream of the control point, Hydraulic analog. Monthly Weather Review 122: 2666-
i.e., the point where the Froude number is unity, 2676.
supercritical flow prevents information from propa- Pratt LJ and Lundberg PA (1991) Hydraulics of rotating
gating upstream. As a result, the amount of flow strait and sill flow. Annual Review of Fluid Mechanics 23:
through the channel cannot be altered from down- 81-106.
stream. Only the upstream conditions and control Seitter KL (1987)Numerical study of atmospheric density
point characteristics have an influence. current motion including the effects of condensation.
Conjugate states are defined as multiple states of Journal of the Atmospheric Sciences 43: 3068-3076.
Smith RB (1985)On severe downslope winds.Journa1 of the
flow, defined by fluid speed and layer depth, with
Atmospheric Sciences 42: 2597-2603.
identical mass and momentum flux. They can be Smith RB and Smith DF (1995) Pseudoinviscid wake
computed easily in the hydraulic formalism. If a flow formation by mountains in shallow-water flow with a
has a conjugate state, it can, in principle, jump drifting vortex. Journal of the Atmospheric Sciences 52:
spontaneously to its other state, without the loss or 436454.
gain of mass or momentum. Most commonly, this Yih C-S (1965)Dynamics of Nonhomogeneous Fluids. New
occurs in a hydraulic jump (Le., an abrupt thickening York: Macmillan.
968 HYDROLOGY I Overview
Contents
Overview
Ground and Surface Water
Modeling and Prediction
Soil Moisture
Hydrologic science has an important place in the

Overview field of water resources, especially freshwater resourc-
es, which are the subject of intense concern and study.
R C Bales, University of Arizona, Tucson, AZ, USA In arid and semi-arid regions, the fair allocation and
wise use of water are significant societal challenges,
affecting relations between nations, states, cities, and
individual users. As a global resource, water appears
abundant. Locally and regionally it is often taken for
Introduction granted. However, the twentieth century has wit-
Hydrology is the science that encompasses the study of nessed a tremendous growth in the use of water,
water on the Earth’s surface and beneath the surface of resulting in limits on both its availability, due to
the Earth, the occurrence and movement of water, the human exploitation, and its quality, owing to con-
physical and chemical properties of water, and its tamination.
relationship with the living and material components
of the environment. Ultimately, many hydrologic
questions involve the transport of solutes, nutrients, Water Cycle
energy, sediment, or contaminants, as well as the
fluxes of water itself. A fundamental concept of hydrology is the hydrologic
As a science, hydrology has both basic and applied cycle, which can be described at many different scales
aspects. The first relates to questions about the Earth of space and time. At the global scale, the hydrologic
system, and specifically about the role of water in cycle is the endless recirculatory process linking water
natural processes, particularly as related to the Earth’s in the atmosphere, on the continents, and in the
biosphere. The second relates to using scientific oceans. We can think of this recirculatory process in
knowledge to provide a sound basis for wise usage of terms of reservoirs or compartments that store water
water resources. The development of hydrologic (e.g., oceans, atmosphere, glaciers, ice sheets, ground
science in recent years is based on both of these water) and the movement of water between them.
aspects, which are equally important and intimately Movement of water from one compartment to another
linked. can occur in any of the three phases. For example, the
Water is central to most natural processes. Water movement of water between the land surface and the
weathers, then transports sediment and solutes to atmosphere occurs in the vapor phase (evaporation
lakes and oceans, thereby shaping the landscape. The and condensation), liquid phase (rain),and solid phase
land-based part of the Earth’s water cycle is important (snowfall).
for transporting carbon from the continents to the Solar energy and gravity are the main forces driving
ocean. The high capacity of water for storing thermal the hydrologic cycle. The dynamic processes of water
energy and the large amount of heat required to vapor formation and transport of vapor and liquid in
change between solid, liquid, and vapor forms of the atmosphere are driven largely by solar energy.
water strongly influence the global energy balance. Precipitation and the flow of water on and beneath the
The distribution of atmospheric water and its regula- Earth’s surface are driven primarily by gravity. Within
tion by oceanic and land-surface processes make it a partially dry soil, gravitational pressure and capillary
central aspect of climate. Water vapor is the most forces are responsible for the movement of water.
important greenhouse gas. In short, life depends on The hydrologic cycle can be considered to start
water. anywhere, but it is often convenient to consider
HYDROLOGY I Overview 969
atmospheric water first. The dominant hydrologic the reservoir (e.g., in m3) divided by the flow through
process involving atmospheric water is precipitation the system (e.g., in m3 yr - ’).Water in the oceans has a
of water to the land surface. Condensation of water residence time approaching 3000 years, less than half
from the atmosphere to the land surface (e.g., dew, of the residence time for ice sheets, while in the
frost) and fog deposition can be important locally, in atmosphere it has a residence time of only 0.02 years or
the absence of precipitation. Precipitation that reaches about 8 days; the residence time in rivers is 0.05 years
the land surface as snow or ice can be retained on or about 1 7 days.
vegetation and subsequently evaporate or fall to the Most of hydrology deals with scales smaller than
ground, accumulate in seasonal snowpacks, and later global; however, the same concepts of fluxes and
melt or evaporate or accumulate in glaciers, ice caps, reservoirs apply. A catchment is an area in which water
or ice sheets. Water is also lost from glaciers and ice falling on or flowing across the land surface drains into
caps by evaporation and melting; pieces of ice can also a particular stream or river and flows ultimately
break off from the edges of glaciers and ice sheets (this through a single point or outlet. Thus a catchment is
is known as calving) and be returned directly to the defined relative to a specific location and includes all of
ocean, in the form of icebergs. A portion of rain or the land area that drains to that point; i.e., it can be
snowfall can be retained temporarily on vegetation considered to catch the water that flows past that
surfaces, and subsequently evaporate or fall to the point. Clearly, then, any number of catchments can be
ground. Rain or snowmelt can also collect in surface defined for a particular river (corresponding to any
depressions, enter into the soil (infiltration), or flow as location along the river). Some special locations of
runoff over the land surface into small rivulets and interest for defining catchments are points correspond-
ultimately into larger streams, lakes, and rivers. Water ing to dams, stream gauges, cities, and a river’s mouth.
that infiltrates into the soil can also follow different Ground water reservoirs (aquifers),on the other hand,
paths. Some returns to the atmosphere by evaporating are defined by subsurface geologic structure. Aquifers,
from the soil or being transpired by plants (transpira- or water-bearing formations, are bounded by material
tion), the sum of which is termed evapotranspiration. of low permeability, i.e., material with a very small
The remaining water continues to move downward ability to transmit water.
through the soil and recharges the saturated portion of
the subsurface, becoming groundwater. Ground water
discharges into streams and rivers, or directly to
Precipitation
the ocean. Water evaporates from the surface of the
oceans and thereby replenishes the water in the Precipitation is the deposition of liquid water droplets
atmosphere. and ice particles that have formed in the atmosphere
Much effort in hydrology goes toward estimating and grown to a size sufficient to fall to the Earth’s
the amounts of water in the various storage compart- surface by gravity. Precipitation is classified according
ments and the magnitudes of the various flows to and to the phase it is in when it reaches the ground, Le.,
from these compartments at local, regional, and global solid (snow, sleet, and hail) or liquid (rain and freezing
scales. Nearly 97% of all water on the Earth is stored rain). Other deposition processes (e.g., direct deposi-
in the oceans, while only about 0.001% is stored in the tion of dew and fog), though generally small, can
atmosphere. Fresh water accounts for about 2.5% of however be important in terms of chemical fluxes (e.g.,
the total storage, 70% of which is contained in the two acidic fog).
polar ice sheets and 30% is found in ground water. Most of the precipitation falling over North Amer-
Only about 0.4% is found in glaciers and ice caps. The ica originates from the bordering oceans, even in the
fresh water in lakes, streams, rivers, and marshes interior of the continent. However, over the Amazon
represents only 0.26% of all fresh water and 0.008% basin, a significant fraction of the precipitation is
of all water on Earth. That is, if the Earth’s ocean derived from within-basin evapotranspiration. In
were represented as a 16 1 (4 gallon) bucket, the fresh other continental basins, local evapotranspiration
water fraction would be equal to a little over 1 m l does have some influence on local precipitation, but
(tteaspoonful). most of the precipitated water must be transported
Another useful concept for enumerating reservoirs significant distances across the continents from the
and the flows of water through them is residence time, oceans.
which is a measure of how long, on average, a Average precipitation over the continents is ex-
molecule of water spends in that reservoir before tremely variable geographically, reflecting the influ-
moving on to another reservoir of the hydrologic cycle. ence of a number of important physiographic factors.
For a system at steady state, i.e. with inflow and First, precipitation increases with elevation owing to
outflow the same, residence time is equal to the size of orographic cooling. Second, precipitation is typically
higher on windward than on leeward sides of moun- rainfall, snowfall has a delayed effect on river flow and
tain ranges. Third, precipitation tends to drop off as air hydrology. Accumulation occurring during winter
masses move further inland, away from the ocean. months becomes all-important during spring runoff,
Fourth, the temperature differences between adjacent which provides much of the streamflow, ground water
land and ocean influence moisture transport. Fifth, recharge, and soil moisture for wide areas of the Earth.
prevailing wind direction has local to regional effects. Melt waters can also cause serious floods, particularly
Global average precipitation is about 1000 mm yr - '. when compounded with spring rainfall.
In the continental US, average annual precipitation
ranges from about 40mmyr-' at Death Valley,
California (in the Mojave Desert), to over 3000
mmyr-' in parts of the Pacific Northwest. In the
Catchment
Atacama Desert of northern Chile rainfall is infre- Much effort in hydrology focuses on water balances at
quent, averaging under 1mm yr - '. the scale of a catchment, be it a 1km2 headwater basin
Rainfall and snowfall are measured at a point by or a 100 000 km2 regional basin. Hydrologists often
collectors of very simple construction. Essentially, any think of a catchment as functioning like a group of
receptacle with a reasonable opening can serve to reservoirs that store and release water. Precipitation
estimate the precipitation per unit area. In the US, the and snowmelt are the inputs to the catchment reser-
standard gauge has a 20 cm diameter opening. Wind is voir, and evapotranspiration and runoff are the
probably the single most important factor in rain outputs. Integrating the various processes that affect
gauge accuracy. Updrafts resulting from air moving up runoff involves quantitative partioning of the different
and around the instrument reduce the catch, which has fluxes and reservoirs into downstream fluxes and
led to the development of elaborate shields to mitigate reservoirs. For example, a fraction of precipitation is
wind effects and detailed correction procedures. intercepted by vegetation, with the remainder falling
In a few areas, radar is also used for precipitation to the ground. Intercepted rain and snow then either
measurement, although rain gauges are generally evaporates or eventually falls to the ground. Snow on
required for calibration. Radar emits electromagnetic the ground partitions between snowmelt and subli-
energy in narrow bands, which upon hitting precipi- mation. Rainfall or snowmelt reaching the ground
tation is partially absorbed, scattered, and reflected. either infiltrates or runs off, depending on the infiltra-
Some of the reflected energy returns to the transmitter, tion capacity of the soil. A portion may flow down-
and the attenuation of the signal indicates the intensity stream in the subsurface and reemerge at a lower point
of rainfall. In principle, the same method could be used in the catchment. Much infiltrating water may even-
for snowfall. tually be transpired by plants or evaporated, though
Snow accumulation at a point is more accurately some will travel downward sufficiently far to enter, or
measured, either in situ by snow pillows equipped with recharge, a ground water aquifer. Of all these quan-
pressure transducers or manually. Snow accumulation tities, surface runoff in streams and rivers, or dis-
is reported in terms of equivalent water depth rather charge, is most often measured.
than snow depth, the proportionality factor being the Discharge is estimated by continuously measuring
density. Snow accumulation can also be measured the stage, or height of a stream at a point using a
from aircraft in terms of the extinction of gamma or pressure transducer or a mechanical float. This point
other types of radiation. Satellites are used for measurement is then related to the volumetric dis-
estimating snowpack depth and areal coverage. A charge through a calibration, or rating, curve, which is
combination of visible and infrared signals are used to developed by measuring the volumetric discharge at
extract snow-covered area. Over flat terrain, passive the gauged point for many different stages. A time
microwave signals indicate snow water equivalent. series graph of discharge is commonly referred to as a
Because much snow falls in heterogeneous terrain and hydrograph.
is thus not uniformly distributed, accurate measure- A streamflow hydrograph is often defined in terms
ments of snow water equivalent remain a major of two components, quickflow and baseflow. Separa-
challenge. Snow measurement networks in the west- tion of a hydrograph into two components suggests
ern US consist of index sites, for which correlations that water is being routed through two different
with streamflow have been developed. They do a storage reservoirs. During and after rainfall and
relatively poor job of providing direct estimates of snowmelt events, water moves through the catchment
snow volume. into the stream channel and the discharge increases
Snowmelt plays a major role in the hydrology of (quickflow). The resulting peak in the hydrograph is
midlatitudes, as many rivers originate in mountains generally defined as a flood, regardless of whether the
where snow dominates the precipitation. In contrast to river actually leaves its banks and causes damage.
HYDROLOGYI Overview 971
Background discharge between floods (baseflow) is practice of irrigation in agriculture, which accounts
supplied by inflow of ground water, which may lag the for about two-thirds of global water use. Understand-
occurrence of precipitation by days, weeks, or even ing the movement of soil water, its uptake by plants,
years. and its loss through evapotranspiration and recharge
Water flowing across the ground surface is termed to the groundwater system is essential in this regard.
overland flow. For this to occur, water must accumu- Hydrologists have traditionally recognized three
late at the surface rather than infiltrate into the soil. divisions within the unsaturated zone: the capillary
This happens for three main reasons: (1)the catchment fringe, the intermediate belt, and the belt of soil water.
surface may be nearly impermeable owing to the The capillary fringe is a zone in which the pressure is
presence of exposed bedrock or pavement; ( 2 ) the less than atmospheric, overlying the zone of saturation
instantaneous rate of infiltration through the pervious and containing capillary interstices, some or all of
surface may be exceeded by the instantaneous rate of which are filled with water that is continuous with the
rainfall (or snowmelt) onto the catchment surface, water in the zone of saturation but is held above that
causing ponding of water at the surface, and ( 3 ) the zone by capillarity forces acting against gravity. That
catchment soil upon which the rainfall is precipitated is, the capillary fringe is a saturated zone above the
may be saturated to the soil surface, causing ponding water table where water is affected by capillary forces.
because the precipitated water cannot infiltrate into an Above that is a zone of soil water from which water is
already saturated soil. Overland flow in catchments is discharged to the atmosphere by the action of plants or
one of the most rapid paths that water can follow to by evaporation. For the most part, plants extract water
the stream channel. from a portion of the soil near the surface (the ‘root
Water that has infiltrated the soil surface continues zone’). Depending on the depth of the vadose zone and
to be influenced by gravity, so that it percolates the plant, roots may lie only in the upper part of the
downward through the soil profile. In general, the vadose zone, or extend into the water table. For
ability of the soils and rocks of a catchment to conduct example, most grasses have roots extending only a few
water (hydraulic conductivity) decreases with depth; centimeters to tens of centimeters, whereas some trees
water percolating downward has thus been observed in semi-arid regions have roots that extend through
to back up, causing local areas of saturation in the soil.vadose zones that are tens of meters thick and reach the
In these instances, water may move laterally toward a water table. The volumetric moisture content (volume
stream by a process known as shallow subsurface of water per bulk volume of soil sample) in the
stormflow. Some of the water in subsurface stormflow capillary fringe is the saturation value. In other words,
moves at a relatively slow pace through the soil and the pores are completely filled with water. As water
contributes to the baseflow of streams, particularly drains or is withdrawn by plants, soil moisture content
during wetter winter and spring periods. Subsurface decreases from saturation to a fairly constant value,
stormflow also may occur quite rapidly along pre- termed the field capacity.
ferred flow pathways or macropores (e+, soil cracks, Rates of removal of water from the unsaturated
old animal burrows, and decayed root channels). zone by evapotranspiration are controlled by a
number of factors, including the wetness of the soil
itself. If a vegetated surface is supplied with plenty of
water (e.g., a well-watered lawn), evapotranspiration
Vadose will be controlled by atmospheric conditions, e.g.
In most areas, the water table is some distance below solar radiation, wind speed, and humidity. That is,
the ground surface. Between the ground surface and evapotranspiration will proceed at the maximum rate
the water table is a region in which the pore spaces of (potential evapotranspiration). As a soil dries, evapo-
the rock or soil may be partly filled with air and partly transpiration will proceed at the potential rate for
with water. This region is referred to as the unsatu- some time, but ultimately the rate will drop. As water
rated zone or vadose zone, and water in this zone is is pulled from the soil near a plant root, the moisture
referred to as soil moisture. Hydrologists want to be content in the soil surrounding the root decreases. In
able to describe the amount of water in the unsatu- order to maintain a steady flow of water to the plant
rated zone and fluxes through the zone for two main root, the plant must exert ever greater suction (ever
reasons. First, ground water recharge occurs through greater negative capillary pressure heads). At some
this zone. Second, most terrestrial plants extract water point, the plant cannot sustain this battle with a drying
from the vadose zone. Plants wilt when soils become soil and the transpiration rate falls below the potential
too dry because the tension forces holding the water in rate.
the soil are too great to allow the plants access to the Most plants have openings (stomata)on their leaves
water. Related to the water balance of plants is the to allow them to take up carbon dioxide from the
atmosphere. When the stomata are open, plants al. The water in a confined aquifer is under pressure
transpire water. Unlike evaporation, transpiration is and, in a well penetrating the aquifer, will rise above
not controlled solely by physical conditions because the top of the aquifer. The height to which water rises
plants regulate the rate at which water is released in in a well defines the piezometric surface, or pressure of
transpiration in a manner that varies by plant type and water in the aquifer.
ecological conditioning. Of the water taken up by In areas where the water table is sufficiently close to
plant roots, most is transpired through the stomata. A the ground surface, ground water levels are influenced
few percent is concerted to biomass through photo- directly by transpiration. During the day, when
synthesis. Hence, to first order, the water taken up by transpiration is high, water movement is upward
the roots is converted to vapor and lost to the from the water table and the level declines. At night,
atmosphere. The degree of stomatal restriction varies transpiration is reduced, groundwater flows laterally
considerably across plant species and even throughout from locations relatively unaffected by direct transpi-
the year for a given species. ration effects, and the water table recovers.
Recharge to aquifers can occur from direct infiltra-
tion of rainfall or snowmelt past the root zone and to
the water table. Water can also seep from surface
Ground Water water bodies, such as rivers, ponds and lakes, into the
The largest accessible fresh water reservoir for human ground. Artificial recharge (recharge induced by
exploitation is ground water, or water that is present in activities of people as opposed to that which occurs
the fractures and interstitial spaces in subsurface naturally) can be implemented by introducing water
geologic materials. In contrast to the vadose zone, in into recharge wells or by routing water into infiltration
ground water the void spaces are completely filled with basins in permeable material.
water. An aquifer is a saturated geological formation The specific yield is a hydrologic parameter that
that contains and transmits significant quantities of determines the response of the water table to changes
water under normal field conditions. ‘Significant’ is a in inputs and outputs. In the case of an increase in
vague term, but the implication is that aquifers are evapotranspiration, the change in water table level
formations that can be used for water supply. Many may be fairly uniform over a given area, although
aquifers are unconsolidated materials, mainly gravel variations will occur due to the lateral movement
and sand. Examples of this type of aquifer include of groundwater and spatial variations in evaporation
those in coastal plains and intermontane valleys. rate and vegetation. Pumping a well has a different
Limestones, partially cemented sandstones and con- effect. Pumping produces a decrease in hydraulic
glomerates, and permeable volcanic and igneous rocks head at a point, which increases the hydraulic
are also important as aquifers. gradient toward the well. The change in water
An aquitard is a formation of relatively low perme- level in the pumping well, or in observation wells
ability, and includes both formations that contain nearby, is referred to as drawdown. The amount
water but do not transmit significant quantities (e.g., of this drawdown will decrease as one moves away
clays and shales) and formations that neither contain from the pumping well, and the pattern produced is
nor transmit significant quantities of water (e.g., called a cone of depression because of its characteristic
unfractured crystalline rocks), shape.
Aquifers are classified according to hydraulic con-
ditions as well as type of material. Ground water by
definition refers to water in the saturated zone of the
subsurface; one type of aquifer is an unconfined or
Integration
water table aquifer. Deeper in the soil profile, satu- The need for an integrated understanding of water
rated conditions prevail (the saturated zone). The cycles and linked chemical cycles is critical for a
water table is defined as a surface of zero (gauge) number of reasons. For example, a quantitative
pressure within the subsurface, and separates the understanding of how ground water recharge changes
saturated and unsaturated zones. Water will flow into with precipitation in a variable or changed climate
an excavation or well up to this level; the water table is requires an understanding of how catchment processes
equivalent to a free surface. An aquifer with the water respond to translate precipitation into recharge. Un-
table as the bounding surface at its top is an unconfined derstanding how evapotranspiration and ecosystem
aquifer. functioning respond to change in land use depends on
The second type of aquifer is a confined or artesian how water and nutrient fluxes change. Hydrologic
one. This is found when permeable material (the science is at the center of many pressing issues in other
aquifer) is overlain by relatively impermeable materi- natural sciences.
HYDROLOGY I Ground and Surface Water 973
See also Fetter CW (1988) Applied Hydrogeology. New York:

Macmillan.
Climate Variability: Seasonal to lnterannual Variability.
Freeze RA and Cherry JA (eds) (1979)
Hydraulic Flow. Hydrology: Modeling and Prediction;
Englewood cliffs, NJ: Prentice-Hall.
Soil Moisture. Land-Atmosphere Interactions:
Hornberger GM, RaffenspergerJP, Wiberg PL and Eshleman
Overview
KN (eds) (1988)Elements of Physical Hydrology. Balti-
more, MD: Johns Hopkins University Press.
Maidment DR (1992)Handbook of Hydrology. New York:
Further Reading McGraw-Hill.
Bras RL (1990)Hydrology - A n Introduction to Hydrologic National Research Council (1991) Opportunities in Hydro-
Science. Reading MA: Addison-Wesley. logic Sciences. Washington, DC: National Academy Press.
Ground and Surface Water

S Ge, University of Colorado, Boulder, CO, USA history as numerous local and regional conflicts over
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. water resources have erupted. Evidence of early
attempts to harness water for human purpose has
been documented by archaeologists. For example,
Introduction clever water usage for irrigation can be traced as far
Water is one of the most precious and indispensable back as 4500 years ago in the Middle East. Also, a
natural resources for civilization. As important com- remarkable water-collecting tunnel system dating
ponents of the hydrological system on Earth, ground from around 500 BC in Egypt has been unearthed. As
water and surface water impact numerous aspects of world population grows at an unprecedented rate, the
the Earth’s processes and many facets of our daily lives. demand for water is expected to continue, particularly
Water at the Earth’s surface interacts directly with the in arid developing countries. The global withdrawal of
atmosphere, and water in the subsurface continuously water in the last century is shown in Figure 1;a sharp
redistributes thermal energy and dissolved minerals in increase is seen in the last 50 years. The total daily fresh
the Earth’s crust at a variety of temporal and spatial water withdrawal in the United States from 1950 to
scales. Hydrology encompasses the study of the the present follows the general global trend (Figure 2).
occurrence and movement of water both at the land The decrease in the 1980s is primarily due to increased
surface and in the subsurface environment. Water irrigation efficiency and a reduction in water con-
stored in the subsurface is known as ground water and sumed in the thermoelectric power industry as a result
is either in a moisture state in near-surface soils or in a of improved power plant technologies and efficiencies.
liquid state in rock voids at greater depth. Some texts The decrease in industrial water usage, however, is
exclude moisture from the broad definition of ground
water.
Focusing on the physical aspect of the dynamics of
ground water and surface water, this article reviews
our current knowledge of water as a resource, and
explains the main concepts associated with ground
water and surface water storage. The fundamental
principles and material properties governing water
movement are presented. Finally, the applied aspects
of hydrology as it relates to water contamination, land
subsidence, and geological processes are discussed.
Water as a Resource 1900 1920 1940 1960 1980 2000

Year
While the presence of abundant liquid water makes the
Earth a unique planet in the solar system, that Figure 1 Annual global water withdrawal (data from Gleick
abundance has been challenged throughout human 2001).
974 HYDROLOGY I Ground and Surface Water
1600 primarily owing to its easy accessibility. The remain-

-I
I +,- Groundwater ing water is distributed in the atmosphere and
biosphere.
Hydrological Cycle
Powered by solar energy, the hydrological cycle is the
endless movement of water from one reservoir to
another in the Earth system (Figure 4).Water evapora-
tes into the atmosphere from open waters such as
oceans and lakes, from soil moisture in the unsaturat-
ed zone, and from the water table. Plants lose water to
0 : I I I I
I I
the atmosphere through the process of transpiration.
1950 1960 1970 1980 1990
These two processes, evaporation and transpiration,
Year
are collectively known as evapotranspiration. Water
Figure 2 Total daily fresh water withdrawal in the United States falls back to the Earth’s surface as precipitation in the
(data from Solley et a/., 1998). form of snow or rain. Upon reaching the surface, water
flows overland as runoff to streams or infiltrates to the
subsurface to become ground water. In the subsurface,
water infiltrates through soils, recharges the ground
accompanied by a continual increase in domestic
water table and joins the ground water flow system.
water consumption.
Ground water takes its course through geological
basins of various scales, and some eventually makes its
Global Water Distribution way to the oceans while some accumulates in inland
Figure 3 shows the global distribution of water. Of the aquifers. The rates of water flow between reservoirs
water on Earth, 97% is stored in the ocean and is too within the hydrological cycle vary spatially and
salty to be used directly for human consumption. Ice temporally in the Earth’s system. As a result, the
caps and glaciers, the next largest water reservoir, hold residence time of water - the time water remains in a
approximately 2% of the global water, which counts reservoir since recharge - in different reservoirs varies
for 94% of the total fresh water. The vast domain of from hours in the near-surface soil to tens of thousands
the subsurface holds only about 1%of the water on of years in rocks several kilometers deep in the crust.
Earth. Furthermore, only the water in the upper few
kilometers of the Earth’s crust is accessible and fresh Water Budget Balance
enough for human consumption. The salinity of The primary objective in studying ground and surface
ground water increases with depth and often becomes water is to understand the spatial and temporal
too high for the water to be useful as a resource below variations of water storage and movement. One basic
4 or 5 km. Surface waters, including streams, lakes, principle governing these variations is the conserva-
and rivers, hold only 0.01% of the global water. tion of mass, or water balance. This principle requires
Seemingly a small fraction of the water budget, surface that the amount of water entering minus the amount of
water has served as the main resource for civilization, water exiting equals the change in storage of a
controlled volume during a specified time period. A
watershed, or drainage basin, is a geographic region
over which the principle of water balance is applied.
b Ice caps, glaciers (2.1%)

24 000 000 km3
Ocean (97%)
1340000000 km3
The watershed is defined as the area over which all the
water flows into a given section of a stream. Watershed
delineation is based on topography, where topo-
I
Ground water (0.6Oh) graphic divides, or ridges, form the boundaries of a
8 400 000 km3 watershed. Equation [l]gives a water balance aver-
Surface water (0.017%) aged over a long period.
177250 km3
Soil moisture (0.005)
67 000 km3
In eqn [l], P is the precipitation, Gin is the ground
Figure 3 Global water distribution (data from Herschy and water inflow, Q is the stream outflow, ET includes
Fairbridge 1998). The lengths of the bars are not to scale. evaporation and transpiration, and Go,, is the ground
HYDROLOGYI Ground and Surface Water 975
Precipitation
Water table --- Evaporation
Figure 4 The hydrologic cycle. The water table is the boundary between the unsaturated zone above and the saturated zone below.
Upon reaching the land surface, precipitation either infiltrates soil to replenish ground water or flows overland as runoff to open water
bodies. Water evaporates from open water bodies at the Earth’s surface, soil moistures in the unsaturated zone, and the water table.
Transpiration occurs over vegetated lands. Ground water flows through the vast domain of the subsurface and returns to the oceans.
water outflow. All quantities have the dimension useful basic information on the direction of ground
[L t - 7. water flow.
Subsurface Hydrological System Porosity, Permeability, and Hydraulic Conductivity

Characterization The primary factors controlling ground water occur-
rence and movement in the subsurface are the hydro-
logical properties of the geological material of interest
Water Table
and the hydrological driving force. The most impor-
In near-surface soils, water is in the form of moisture in tant material properties are porosity and permeability.
the unsaturated zones where void spaces are partially Porosity, defined as the fraction of void space in a unit
filled with water. At deeper locations, void spaces are volume, may vary from 0-5% for tight igneous and
completely filled with water, which forms saturated metamorphic rocks to 25-50% for sands or fractured
zones. The boundary between the unsaturated and the rocks. Clay can have a porosity as high as 70%.
saturated zone marks the water table. The shape of a Porosity is a direct measure of the geological material’s
water table generally follows the topography, near the capacity for holding water.
surface in low lands but deeper in high elevations. The Permeability, on the other hand, is a measure of the
depth of the water table varies in both space and time. material’s ability to transmit water. It is defined as
In humid areas, the water table can be at or near the follows: k = Cd2,where k is the intrinsic permeability
surface, whereas in drier regions the water table can be [L2], Cis a constant of proportionality related to grain
hundreds of meters below the land surface. The water size, sediment sorting, and packing arrangement, and
table can rise in wet seasons as precipitation percolates d is the average grain diameter [L]. The ease of flow
through unsaturated zones to the water table, and can differs depending on the type of fluid flowing through a
drop in dry seasons when more water is lost through system. One can imagine that a fluid that is sticky like
evaporation and transpiration. The most reliable way honey would flow much more slowly than would clean
to locate water table depth is to drill wells. Because water in the same medium. It is necessary to consider
ground water beneath the water table flows in the not only the medium but also the fluid properties.
general down-slope direction of the water table, Consequently, a hydraulic conductivity is introduced,
knowledge of water table shape and position gives defined as K = kpg/p, where K is the hydraulic
conductivity [L t - '1, p is the fluid density [M L - '1, g is unsaturated zone is conventionally expressed in neg-
the acceleration [L t - 2 ] , and ,u is the dynamic viscosity ative values, reflecting the use of atmospheric pressure
'
of the fluid [M L - t - '1. Values of hydraulic conduc- as the zero reference pressure. The pore pressure
tivity vary over several orders of magnitude from distribution and the rate of moisture movement vary
ms-' for unfractured rocks to l m s - ' for spatially depending on soil types and weather condi-
sands and gravels. The most reliable means of obtain- tions and temporally in response to rainstorms,
ing the hydraulic conductivity values is irt situ well seasonal changes, and long-term climate change.
testing. When water is withdrawn from or injected Infiltration is an important process in the unsatu-
into a well, the rate of water level recovery in the well rated soil zones and involves downward movement of
can be monitored by a pressure transducer and moisture under wet climatic conditions. The infiltra-
recorded by a data logger. The rate of water level tion rate over a small area can be measured using a ring
recovery, as an indicator of the material permeability, infiltrometer. This is a portable cylindrical ring, with a
is used to compute the hydraulic conductivity of the diameter from a few centimeters to 20 cm, extending
material surrounding the well. Laboratory tests on several centimeters above and below the surface of the
core samples are conducted to obtain hydraulic soil. The rate of water dissipating from the ring
conductivity values valid at centimeter scales. Com- infiltrometer into the soil can easily be converted to an
puter modeling is also commonly practiced as an infiltration rate.
indirect means of inferring hydraulic conductivity at In contrast to infiltration, evaporation and transpi-
different scales. ration draw moisture upward under dry climatic
conditions. Evaporation causes water loss from sur-
Aquifers, Aquitards, and Aquicludes face waters, such as lakes and rivers, and from
A geological unit that is highly permeable and that can shallow-depth soils. Water evaporates as a vapor
transmit a significant amount of ground water is called diffusion process that is largely controlled by the
an aquifer. When an aquifer is confined between two energy exchange between radiation or sensible heat
less-permeable units, it is called a confined aquifer. from the atmosphere or ground, and the heat energy
When bounded at the top by the water table, the change in the evaporating body. A direct method for
aquifer is called an unconfined aquifer. Water pressure determining the evaporation rate has been developed
in confined aquifers is usually higher than in uncon- and is known as the pan-evaporation approach. It
fined aquifers. Thus, when a well is drilled into a involves exposing a cylindrical pan of water to the
confined aquifer, the water level in the well can rise to atmosphere in clearings where precipitation can be
above the land surface to form artesian flow. An monitored. The standard US National Weather Service
aquitard is a less permeable geological unit, while the Class A pan is 1.22 m in diameter and 25.4 cm deep.
term aquiclude refers to an impermeable unit. As Transpiration is a process whereby water is lost to the
absolutely impermeable geological material is rare, the atmosphere through the vascular systems of plants.
term aquiclude is infrequently used in hydrological The transpiration process works by absorption of
characterizations. Aquifer and aquitard are more water by plant roots, translocation of liquid through
commonly used to characterize geological systems. the plant vascular system, and transpiration into the
For distinction, aquifers and aquitards are often atmosphere through openings in the leaf surface.
identified in a relative sense. Gravel, sandy materials, Although transpiration is also considered a diffusion
or highly fractured rock make good aquifers, whereas process, water is first pulled through the plant by a
clay-rich, poorly sorted sediments, and unfractured potential energy gradient before diffusing into the air
rocks often form aquitards. in response to a vapor pressure difference.
Hydraulic Head
Physical Hydrological Processes
Hydraulic head is one of the key parameters describing
Water in Soils
the mechanical energy state of a hydraulic system. It
Near the land surface at shallow depths, soils are often represents the mechanical energy per unit weight of the
partially saturated; such a region is known as the fluid in the system. Hydraulic head is defined as
unsaturated zone. The degree of saturation is defined h = h, + h,, where h is the hydraulic head, h, is the
as the fraction of pores that contain water, and varies pressure head, and h, is the elevation head. All three
from 0, representing a dry condition, to 1,representing quantities have the dimension [L]. The pressure head
a fully saturated condition. The water in the partially represents the energy due to pore fluid pressure, and
saturated soils clings to soil particle surfaces and is the elevation head represents the gravitational energy
sustained by suction or tension. Pore pressure in the arising from elevation. Water flows from high to low
HYDROLOGYI Ground and Surface Water 977
hydraulic heads. Knowledge of the hydraulic Wells

head distribution in an area allows one to estimate
Wells serve as a window to the subsurface in the study
the direction and rate of flow according to Darcy's
of ground water. The primary need for wells is to
law, which is described in the following section.
withdraw water from the ground. Some wells are used
In situ measurement of hydraulic head is
for monitoring water levels and also for sampling
accomplished by measuring the water levels in
water for chemical analyses. When a well is pumped, a
wells, using electric sounding water tapes or similar
depression cone forms around the well as the water
devices.
level declines. Water level response can be quite
different depending on whether the pumping occurs
Darcy's Law
in a confined or in an unconfined aquifer. Under the
The basic theory governing ground water movement is same pumping rate, a larger water level decline is
Darcy's law, named after Henry Darcy, a French expected in a confined aquifer and a smaller decline in
engineer. In Dijon, France, in 1856, Darcy performed a unconfined aquifers. This is because less water is
water purification experiment using a simple setup available in confined aquifer storage, while uncon-
involving water flow through a cylindrical sand fined water may receive water from drainage of the
column. The experimental data led to an empirical water table.
relationship between water flux and the experimental
parameters. That relationship later became the well- Surface Water Dynamics
known Darcy's law. In a one-dimensional system,
Although small in volume in the global water distri-
Darcy's law is expressed as eqn [2].
bution, surface water has been the main water source
for our civilization. A stream is a body of water
dh
Q=-AK- flowing down-slope along a natural passage. A river is
dx a stream with a significant amount of flow. A stream
with no tributaries is designated as a first-order
In eqn [2], Q is the volumetric flow rate [L3t-'1, A
stream; the confluence of two first-order streams is
is the cross-sectional area of flow [L2], K is the
the beginning of a second-order stream; two second-
hydraulic conductivity [L t - '1, h is the hydraulic
order streams are the beginning of a third-order
head [L], and x is the distance coordinate [L]. The
stream, and this pattern can continue to form higher-
dimensionless quantity dhldx is called the hydraulic
order streams. The branching patterns of stream
gradient. It is clear that the hydraulic gradient is
orders have been studied through a fractal approach
the primary driving force for ground water flow.
that provides a mathematical framework for treat-
The negative sign in the equation denotes that ground
ment of similar geometric characteristics over a range
water flow is in the direction of decreasing hydraulic
of scales. Streams are vital geological agents that incise
head.
valleys, transport tonnes of solid and dissolved mate-
rials to the sea, and reshape our landscape over time.
Ground Water Flow Patterns
Stream behavior is affected by a variety of factors such
The movement of ground water normally occurs as as the steepness of the stream, the cross-sectional area
slow-moving seepage through the pore spaces in soils of the stream, the velocity of water in the stream, and
and rocks or as relatively fast flow through rock sediment load. The Reynolds number, Re, is a
fractures. The velocities are generally much slower convenient parameter for describing the state of flow
than stream flow, and may be as low as 10 - m s - A as laminar or turbulent. The dimensionless Reynolds
ground water velocity on the order of 10 - m s - ' number is defined as Re = p U Y / p , where p is the
would be considered high. The low velocity of ground density of water [ML-3], U is the average flow
water has important implications for geological proc- velocity [L t - '1, Y is the average flow depth [L], and p
esses such as metamorphisms and ore formation as is the dynamic viscosity of the fluid [ML-'t-'].
well as for contaminant movement because it leads to Laminar flow occurs when Re is less than 500.
longer residence times. The extent of ground water Turbulent flow occurs when Re is greater than 2000
systems varies from local hill slopes at meter scales to and circulating eddies form in turbulent regions.
regional geological basins at kilometer scales. Flow Transitional flow lies between laminar and turbulent
paths are shallower in local flow regimes and deeper in flows. Actual stream flow is seldom laminar, but when
regional flow systems. As a result, the residence time of the degree of turbulence is small, flow is often
ground water also varies significantly ranging from considered in the laminar range. The following math-
hours or days in shallow and small systems to tens of ematical formulation can then be used. The equation
thousands of years in deep and large geological basins. of motion can be expressed as z, = K,h(dh/dl), where
is the flow velocity [ Lt-l], Kch is the flow channel neutralize and reduce the stream’s acidity to tolerable
conductivity [L t - ‘1, and dh/dl is the head gradient, levels for aquatic organisms.
the change in head h with respect to change in distance
1. The head represents the energy of the system and the
gradient provides the driving force. The energy Applied Aspects of Hydrology
includes gravitational, kinetic, and pressure compo- While the scientific aspect of hydrology seeks under-
nents. For turbulent flow, the flow dynamics is more standing of the role that water plays in natural
complex and nonlinear and the equation of motion processes and mechanisms of water movement in the
becomes z/ = K,h(dh/dl)1’2. Earth system, the applied aspect relates to using the
scientific knowledge to understand the impact and
consequences of improper water use, and to better
Surface Water and Ground Water Interaction protect water resources. Although only two areas are
discussed below, applied hydrology impacts on our
In order t o effectively manage water resources, we society and environment far beyond what is included
must understand the interaction between surface in this section.
water and ground water. Conjunctive use of surface
and ground water has increasingly become common
practice in water management, particularly in arid
Water Contamination
regions. The basic concept of conjunctive use is to
store water in aquifers under wet climate conditions Water contamination has increasingly become a con-
when streams flows are high and t o withdraw water cern in modern times. Application of pesticides and
from the aquifers under drier conditions when demand fertilizer in agricultural practices can result in areal
is high but stream flows are low. The success of a contamination as excess irrigation water percolates
conjunctive-use project depends heavily on the dy- through soils and carries the chemicals into the ground
namics of the interaction between the surface water water systems. Contamination sources that spread
and the ground water. Streams can either gain water over large areas are known as nonpoint sources;
from or lose water t o a ground water system. In reality, sources confined to small areas are called point
many streams do both at different locations and at sources. Multiple closely spaced point sources can
different times. The rate and direction of flow into or form a nonpoint source. Landfills are major point
out of the stream can also vary as the elevations of the sources of water contamination. Aging and leaking
water table and stream surface fluctuate. Pumping of liners around landfills allow leachate, a mixture of
ground water can change the quantity and direction of water and dissolved chemicals, to leak into water
flow between an aquifer and a stream. As a result, systems. Wastes from mines and nuclear facilities and
stream flow is reduced, which can be of particular petroleum spills have all contributed to the contam-
concern during droughts. Moreover, the interaction ination of streams and ground water, thus degrading
between ground water and surface water affects water water quality. While it is necessary to identify sources
quality. When the ground water in shallow aquifers is of contamination, understanding the contaminant
contaminated (for example, from agricultural prac- transport is equally important so that the future
tices of applying fertilizer and pesticides) the shallow behavior of contaminant plumes can be predicted
aquifers can contaminate surface water as the ground and effective remediation procedures can be designed.
water flows toward a stream. The opposite can also Three major mechanisms control the transport of
occur when a stream is heavily contaminated (for contaminants in water: dispersion, advection, and
example, from mine waste drainage in mountainous retardation. Dispersion involves the dissolved con-
regions); and relatively clean ground water can dilute taminants diffusing into and mixing with clean waters.
the contaminant level as the ground water enters the Advection - the transport of contaminants by moving
stream from the gaining sections of the contaminated water - is often the dominant mechanism once the
stream. Mixing of ground and surface water also has contaminants make their way into permeable rocks.
an effect on other natural environments such as While dispersion and advection are the physical
wetlands and aquatic environments when acidity, aspects of transport, retardation takes into consider-
temperature, and dissolved oxygen are altered by the ation of the chemical reactions occurring between the
interaction. Streams may become acidic as they receive contaminant and the media during its migration.
atmospheric deposition of chemicals, such as sulfate Slowdown of the advancing plume can be caused
and nitrate. Acidic precipitation directly affects the by various chemical reactions such as sorption,
well-being of aquatic ecosystems. However, signifi- precipitation, oxidation, ion exchange, or biological
cant ground water flow into a stream may help activities.
HYDROLOGY I Ground and Surface Water 979
Land Subsidence locations, a process known in petroleum system

studies as secondary migration. The mechanical inter-
Extraction of ground water plays a direct role in land
action between ground water and rock deformation has
subsidence. Uneven subsidence of the historic tower of
been thought to contribute to the triggering of earth-
Pisa in Italy has created a tourist attraction. However,
quakes. As pore pressures in faults and the surrounding
in many cities and towns subsidence has been a
area increase, faults become lubricated and the fault is
problem when building foundations and road surfaces
weakened, setting the stage for an earthquake. The
become cracked and tilted as the ground subsides.
best-known example is that of the documented earth-
When a large amount of water is withdrawn from the
quakes between 1962 and 1972 in the Denver area
ground, void spaces in rock or soil collapse, the soil is
when liquid waste was injected underground into
compacted, and the ground subsides. The Santa Clara
fractured granites a few kilometers deep. The time
Valley and nearby San Joaquin Valley in northern
and frequency of the earthquakes were correlated
California have experienced regional subsidence due
strongly with the time and volume of waste injection.
to excessive ground water pumpage to sustain pro-
ductive agriculture. As Las Vegas Valley turns into a
fast-growing metropolitan area, ground water is See also
rapidly being depleted and this area has also suffered Agricultural Meteorology and Climatology. Climate:
problems with land subsidence and ground fissures Overview. Hydrology: Modeling and Prediction; Over-
due to horizontal movement. Subsidence may also view. Palmer Drought Index.
occur from drainage of soils that are rich in organic
carbon as microbial decomposition converts organic
carbon t o carbon dioxide gas and water. Subsidence at Further Reading
the rate of 20-80 mm per year has been observed as a AlleyWM, Reilly TE and Franke OL (1999)Sustainability of
result of the decomposition of the remains of shallow- Ground-Water Resources. US Geological Survey Circular
water sedges and reeds in California’s Sacramento- 1186. Washington, DC: US Geological Survey.
San Joaquin Delta and in Florida’s Everglades. More Dingman SL (1994) Physical Hydrology. New York:
catastrophic subsidence takes place with the forma- Macmillan.
tion of sinkholes associated with localized collapse of Domenico PA and Schwartz FW (1998) Physical and
subsurface cavities. Often triggered by a decline in the Chemical Hydrogeology. New York: Wiley.
ground water level, sinkholes typically form in areas Fetter CW (2001)Applied Hydrogeology. Englewood Cliffs,
NJ: Prentice-Hall.
underlain by carbonates (e.g., limestone and dolo- Fetter CW (1999) Contaminant Hydrogeology. Englewood
mites) and evaporites containing easily dissolved Cliffs, NJ: Prentice-Hall.
minerals such as salt, gypsum, and anhydrite. Freeze AR and Cherry J (1979) Groundwater. Englewood
Cliffs, NJ: Prentice-Hall.
Gleick PH (2001) Making every drop count. Scientific
Role of Ground Water in Geological American February, p. 41.
Processes Herschy RW and Fairbridge RW (1998) Encyclopedia
of Hydrology and Water Resources. Boston: Kluwer
Mater exists in pore spaces in soil and rock from near Academic.
the surface to a depth of more than 10 km. The amount Hornberger GM, Raffensperger JP, Wiberg PL and Eshleman
of pore water decreases with depth in response to a KN (1998) Elements of Physical Hydrogeology. Balti-
general decrease of rock porosity with depth. Ground more, MD: The Johns Hopkins University Press.
water plays an essential role in mineral dissolution and Rodriguez-Iturbe I and Rinaldo A (2001) Fractal River
precipitation, and thus impacts metamorphic process- Basins. New York: Cambridge University Press.
es by altering the mineral compositions of rocks. As Solley WB, Pierce RR and Perlman HA (1998) Estimated
water flows through deep sections of the crust or Use of Water in the United States in 2 995. US Geological
passes through thermally active regions, such as in the Survey Circular 1200. Washington, DC: US Geological
vicinity of cooling igneous rocks solidifying from Survey.
magma, the heated waters become hydrothermal Tindall JA and Kunkel JR (1999) Unsaturated Zone
Hydrology. Englewood Cliffs, NJ: Prentice-Hall.
waters. Hot springs emerge at the locations of hydro- Vorosmarty CJ, Green P, SalisburyJ and Lammers RB (2000)
thermal water discharge. Ground water carries dis- Global water resources: vulnerability from climate
solved minerals and transports them to ore-forming change and population growth. Science 289: 284-288.
locations. Petroleum is matured, in natural gas or Winter T (1998) Ground Water and Surface Water,
crude oil states, at a greater depth of burial. Ground A Single Resource. US Geological Survey Circular
water may then transport petroleum to shallow 1139. Washington, DC: US Geological Survey.
Next Page
980 HYDROLOGY I Modeling and Prediction
Modeling and Prediction

Z Yu, University of Nevada-Las Vegas, Las Vegas, NV, the representation of physical processes in these
USA models is often too crude and the scales of measure-
Copyright 2003 Elsevier Science Ltd. All Rights Reserved ment for many hydrological parameters are incom-
patible with the scales used in the models. Studies have
indicated that the uniform effective parameters (e.g.,
Introduction saturated hydraulic conductivity) cannot represent the
ensemble heterogeneity, resulting in a significant
Hydrological models have become an indispensable underestimation of hydrological variables (e.g.,
tool for study of hydrological processes and the impact streamflow). These physically based distributed mod-
of modern anthropogenic factors on the hydrological els can be referred to as conceptual lumped models in
system. Mathematical models that are governed by the some sense. The physically based models are more
laws for conservations of mass and momentum are complex and require more computing time for solving
used to describe the temporal and spatial variation of a PDEs numerically and considerable effort to master all
hydrological system in the field on the basis of their intricacies, such as model calibration. For sim-
information concerning climate, land use and land ulating the hydrological response (e.g., soil moisture)
cover, and hydrology. Modeling the hydrological to climate forcing (e.g., storms and human-induced
response to various natural and anthropogenic forc- global warming), these models currently offer no
ings has the potential to contribute to the understand- advantage over the traditional conceptual lumped
ing of these physical processes, such as flow and water-balance models.
transport in the surface and subsurface and the Physically based hydrological models are an impor-
atmosphere-land surface interaction. tant evolutionary step in representing hydrological
Two types of hydrological models have been used in processes and spatially distributed data. At present,
most applications: lumped conceptual models and the ability to represent spatially varying processes is
physically based models. A lumped model is generally outstripping our ability to collect various data sets.
applied in a single point or a region for the simulation The need for the research on the better representation
of various hydrological processes. The parameters of physical processes in space and time is evident given
used in the lumped model represent spatially averaged the availability of digital products (e.g., distributions
characteristics in a hydrological system and are often of elevation, soil, vegetation) and remotely sensed data
unable to be directly compared with field measure- (e.g., soil moisture, vegetation), along with new
ments. In general, lumped models use simple book- technologies for measuring temporal and spatial
keeping procedures to quantify physical processes by variability in precipitation.
simulating the temporal variation of various physical Research on data assimilation and analysis, sub-
processes in a hydrological system. The advantage of grid-scale variability in precipitation and hydraulic
these models over physically based models is that the parameters, sub-grid-scale model calibration, predic-
conceptual parameterization in the models is simple tion uncertainty analysis, and inclusion of fine-reso-
and computation is efficient. With the availability of lution surface and subsurface hydrology in various
spatially distributed digital and remotely sensed data hydrological and climatic models is being conducted
sets of features such as precipitation, elevation, veg- in various research institutes to facilitate an interactive
etation, etc., many distributed lumped models have link between the hydrology and atmosphere. Simula-
been developed in recent years. These kinds of models tion and prediction of various hydrological systems
have been widely used in most climate and meteoro- are among the principal areas of study in the current
logical studies to model hydrological processes. fields of hydrology and climate.
Many physically based distributed-parameter mod-
els have been developed to facilitate various hydro-
logical and climatic applications over recent years. Hydrological Simulation in
These models represent hydrological processes in a
Meteorological and Climatic Models
physically rigorous manner because they use process-
based partial differential equations (PDEs) to describe Climate variability has substantial impact on hydro-
the spatial variability of hydrological processes. logical, biological, and ecological systems such as
Within the framework provided by these equations, water availability and quality, floods, and droughts.
it is possible to estimate governing parameters through The effects of climate variability on economic vitality
field studies. One disadvantage of such models is that and the quality of life (e.g., the 1988 North American
ICE AGES (MILANKOVITCHTHEORY) 995
M F Loutre, Universite catholique de Louvain, Louvain- latitude through the year, is forcing climate. Four main
la-Neuve, Belgium steps can be identified in an astronomical theory
Copyright 2003 Elsevier Science Ltd. All Rights Reserved (Figure 1):
e the theoretical computation of the long-term vari-
Introduction ations of the Earth’s orbital parameters and related
insolations;
An astronomical theory of paleoclimate aims to e the design of climate models to study climate
provide an understanding of the relationship between changes related to insolation changes;
insolation and climate on the global scale. One such e the collection of geological data and their interpre-
theory was developed by Milankovitch who first tation in terms of climate; and
computed the radiation received by the Earth from e the comparison of these proxy data to the simulated
the Sun (insolation).This quantity undergoes diurnal, climatic variables.
seasonal, as well as millennia1 scale (long-term)
variations. Milankovitch argued that insolation Therefore astronomical theory is at a crossing point of
changes induce surface temperature changes and that geology, astronomy, physics, chemistry, biology, and
long-term insolation changes are sufficient to produce geophysics. Moreover, it is related to processes of the
ice ages by changing the geographic and seasonal solid Earth, the atmosphere, the hydrosphere, and the
distribution of sunlight received by the Earth. He ocean.
estimated the relationship between summer radiation
and the altitude of the snowline, which allowed him
to compute the latitude of the ice sheet margin for the
last 650 000 years. Milankovitch’s curves were found The Ice Ages
to be remarkably consistent with the reconstruction of Our planet Earth was born some 4.5 billion years ago.
the glacial period made by Penck and Briickner. Many events have marked its evolution; in particular,
However, Milankovitch theory is not the only astro- some periods have experienced markedly colder
nomical theory of paleoclimate. For example, in the climate than others. Up to 0.9 billion years ago the
nineteenth century, Croll developed an astronomical Earth was apparently mostly ice-free, despite low solar
theory of paleoclimate, which was later rejected luminosity, although there is some evidence for a first
because it did not fit with the evidence of that time. glaciation approximately 2.5 billion years ago. At
Today astronomical theories are still a focus of least three major phases of glaciation occurred be-
research. Sophisticated climate models are used to tween 900 and 600 million years ago. Glacial deposits
test how insolation, and in particular its distribution in from the late Proterozoic have been found at low
Figure 1 Schematic outline of an astronomical theory of paleoclimate.

996 ICE AGES (MILANKOVITCHTHEORY)
latitudes, suggesting that at that time ice sheets the other during Glacial Epochs, when eccentricity is
covered the Earth from pole to pole. This is the so- large. According to Croll, the last Glacial Epoch began
called ‘snowball Earth’ hypothesis. The return to some 250 000 years ago and ended about 80 000 years
warmer conditions would then have resulted from the ago. Later, he also took into account the role of the tilt
accumulation in the atmosphere of COZ from volcanic of the Earth’s axis of rotation. He hypothesized that an
activity. The large cap carbonate found in Namibia, ice age would be more likely to occur when the tilt was
for example, could be the witness of this time. small because the polar regions would then receive less
However, this hypothesis is still debated. From 600 heat. However, he acknowledged that orbital changes
to 100 million years ago mild climates prevailed, could only be a triggering mechanism. He identified
punctuated by several major phases of ice growth. the albedo-temperature feedback and the change in
These long geological cold periods, such as the late ocean currents as two mechanisms responsible for
Precambrian Ice Age, the late Ordovician-Silurian Ice amplifying the direct climatic effect of the astronom-
Age, and the Permo-Carboniferous Ice Age, are also ical forcing. Meanwhile, geologists collected evidence
called Ice Ages. A gradual cooling and drying of the around the world that several ice ages had occurred in
globe started some 50 million years ago. The Antarctic the past, separated by nonglacial epochs, as predicted
ice sheet started to grow some 40 million years ago, by Croll’s theory. However, with time the majority
whereas the Greenland and midlatitude ice sheets of geologists in Europe and America became opposed
formed later (4-2.4 million years ago). The Quater- to Croll’s theory as more and more new evidence
nary Ice Age, the cold period in which we are living, is suggested that the last Glacial Period ended later
characterized by a succession of colder and warmer than 15000 years ago, instead of 80000 years
periods, the glacial-interglacial cycles. During colder ago as required by Croll’s astronomical theory. By
phases (or ice ages) the ice sheets spread out in the high the end of the nineteenth century, the theory fell into
latitudes. The purpose of astronomical theory is to disfavor.
explain these broad climatic features, which charac- The attention of the scientific world was drawn
terize not only the Quaternary, but also older periods back to the ice age problem with the publication in
including the Early Mesozoic, more than 150 million 1924 of Milankovitch’s theory. This was the first full
years ago. astronomical theory of the Pleistocene ice ages,
including the computation of the orbitally induced
changes in the insolation and climate. According to
A Brief History of the Astronomical Milankovitch’s theory, the summer in northern high
latitudes had to be cold to prevent the winter snow
Theory of Paleoclimates from melting. In turn this would allow a positive value
The inspiration for the pioneering work on the in the annual budget of ice, and a positive feedback
astronomical theory of paleoclimate was probably cooling would be initiated over the Earth through a
Louis Agassiz’s lecture about his theory of a Great Ice further extension of the snow cover and subsequent
Age at a meeting of the Swiss Society of Natural increase of surface albedo. This hypothesis requires
Sciences, held in Neuch2tel in 1837. Only a few years that summer insolation is a minimum in the high
later, Joseph Adhkmar proposed an explanation of the latitude Northern Hemisphere. One of Milankovitch’s
existence of ice ages based on the precession of the first major contributions consisted of radiation curves
equinoxes. At the same time the French astronomer, showing past insolation changes at high northern
Urbain Leverrier, calculated the changes in the Earth’s latitudes (Figure 2).He claimed that each minimum of
orbit over the last 100 000 years. James Croll would these curves should cause an ice age. Comparing
eventually take Adhkmar’s idea and put it into an Milankovitch’s curves with the Penck and Briickner
astronomical theory of climate. Croll’s major hypoth- curve for the succession of European ice ages, Koppen
esis was that the critical season for producing an ice and Wegener concluded that the theory matched the
age is winter. He determined that the precession of the facts amazingly well. Although the timing of the ice
equinoxes must play a decisive role in the amount of ages and the radiation minima did not agree in detail,
sunlight received during winter. Indeed, winter would the general pattern of the two curves was quite similar.
be warmer if it occurred when the Earth were closer to But by 1955, the astronomical theory was rejected by
the Sun and it would be colder if the Earth were farther geologists. Indeed, using new techniques for dating
from the Sun. Moreover, the shape of the Earth’s orbit Pleistocene fossils (radiocarbon dating) they showed
could reinforce this effect. He concluded that periods that there were more glacial advances during the past
of greater eccentricity could lead to exceptionally 80 000 years (or at least the time interval believed to be
warm or cold winters. He argued that ice ages 80 000 years) than could be explained by the Milan-
therefore occurred alternately in one hemisphere or kovitch theory.
60"
W
-0
3
._
c
L
m 65"
-
c
c
-Wm
> 70"
._
3
(T
(B) 600 550 500 450 400 350 300 250 200 150 100 50 0
500
450
400
600 550 500 450 400 350 300 250 200 150 100 50 0
(C) Time (ky BP)
Figure 2 The Milankovitchamplitude of the secular variations of the summer radiationat 65" N,(A) after Stockwelland Pelgrim's, and (B) after Le Verrier and Miskovitch's astronomical solution.
The ordinate axis in (A) and (B) gives the latitude that receives the same radiation as 65" N for the last 600 ky. Part (C) gives the mean irradiance (W m -*) according to Berger's computation.
The theory was still largely disputed until the early In eqn [l], a, the semimajor axis of the orbit, gives its
1970s. Nevertheless, progress was being made toward size. The value of a is constant through geological time
a better understanding of the ice ages, in particular the to a very good accuracy. The eccentricity, e, is a
Pleistocene. New dating techniques were developed measure of the departure of the ellipse from a circle,
and accurate climatic interpretation was given to and the true anomaly, u, of the Earth is a measure of the
geological records, such as variation of the oxygen position of the Earth in its orbit. The true anomaly is
isotope records of forams in deep sea cores, or measured counterclockwise from perihelion (where
sequences of soils and loess. New evidence was put the Earth is nearest to the Sun). Usually the angles that
forward that major climate changes have accompa- describe the position of the Earth in its orbit are not
nied variations in obliquity and precession over the measured from the perihelion, but rather from the
last 500 000 years. This finding stimulated the revival spring equinox (SE). Therefore, we have u = 2 - o,
of the astronomical theory. Vernekar, Bretagnon, where 1 is the longitude of the Earth in its orbit
Berger, and others refined the calculations of orbital measured from the spring equinox of the year, or
history; geologists obtained new records of past moving spring equinox, and w is the longitude of the
climate; the improved dating techniques clarified the perihelion relative to this same moving spring equi-
detail of the Quaternary time scale; global past climate nox. Alternatively, the position of the perihelion
changes were reconstructed with better accuracy; and (0) is often measured from the First Point of Aries
finally, climate models were designed to test this (vernal point). This point on the Earth’s orbit gives the
theory. Within a few years it became increasingly clear direction of the Sun as seen from the Earth at the spring
that major changes in global climate were caused by equinox. Consequently, we have v = i- 6 - 7c. In
changes in the astronomically driven insolation forc- addition, the Earth’s axis of rotation is tilted with
ing. Moreover, the importance of mechanisms and respect to the orbital plane. The obliquity, E , is the
processes such as the carbon cycle, vegetation change, angle between the Earth’s axis of rotation and the
ocean circulation, and dynamics of the cryosphere was perpendicular to the orbital plane.
also acknowledged. The astronomical parameters, e, E , and o,experi-
ence long-term variations. These variations can be
obtained by solving two sets of equations, one set for
Orbital Parameters the mutual gravitational forces in the planetary system
and the other for the Sun-Earth-Moon system.
The German astronomer Johannes Kepler (1571-
1630) formulated the three laws of planetary motion,
which are satisfied with a high accuracy not only by the
system of planets and Sun, but also by the system of
each set of satellites moving about their primary
planet. They are:
0 The orbit of each planet is an ellipse with the Sun at

one focus.
0 The line (the radius vector) joining the Sun to each
planet sweeps out equal areas of its ellipse in equal
times. WS
0 The cubes of the semimajor axis of the planetary
P
orbits are proportional to the squares of the planets’
periods of revolution.
While Kepler gave a description of the orbital motion,
Sir Isaac Newton (1642-1727) defined the law of
gravitation, which is the basis for explaining
the planetary motion. In particular, during its Figure 3 Elements of the Earth’s orbit around the Sun (S). Some
travel counterclockwise on its orbit around the Sun reference points are drawn on the orbit, Le., the perihelion (P) and
(Figure 3), the Earth is sometimes nearer to and the aphelion (A), the spring equinox (SE), the summer solstice
sometimes farther away from the Sun. The distance (SS), the autumn equinox (AE), and the winter solstice (WS). The
vernal point is given by y . 1. is the longitude of the Earth in its orbit
from the Earth to the Sun ( Y ) can be computed through measured from the spring equinox of the year, or moving spring
the equation of the ellipse, given here as eqn [ 11. equinox; (I) is the longitude of the perihelion relative to this same
moving spring equinox; and v is the true anomaly measured
r = [ a ( l - e’)] x (1 + e cos v)-l 111 counterclockwise from the perihelion.
ICE AGES (MILANKOVITCH THEORY) 999
Different methods, from fully analytical to fully largest amplitude terms correspond to periods of 41
numerical, were developed following the first calcula- (by far the largest term), 54, and 39 ky. The variations
tions by Leverrier in the nineteenth century. Moreover, of the climatic precession ( e sin 0j reflect two oppos-
the accuracy of these solutions has been greatly ing motions, i.e., the counterclockwise motion of the
improved. However, the orbital motion of the planets perihelion along the ecliptic with a period of 100 kyN
in the solar system is chaotic, i.e., the solution is and the clockwise motion of the vernal point along the
strongly dependent on the initial conditions, which ecliptic with a period of 25 700 years. The two
N
limits the possibility of obtaining an accurate solution effects taken together result in the climatic precession,
for the astronomical parameters of the Earth over which corresponds to the slow shift of the seasons
more than 35-50 million years. about the Earth's orbit relative to the perihelion. This
Figure 4 shows the long-term variations of the three motion has a mean quasi-period of 21 ky, which is
orbital parameters ( e , E , e sin 0)from 400 000 years derived from periods at 1 9 and 23 ky.
Before Present (400ky BP) to 100000 years After
Present (100ky Ar), a time slice over which the
solution is very accurate. The eccentricity of the orbit
Insolation
varies between near circularity ( e = 0) and small The amount of solar radiation received at a mean
ellipticity ( e = 0.07). These variations are quasi-peri- Earth-Sun distance, r,, by a surface perpendicular to
odic with a mean period of about 100 ky. However, a the incident radiation, is about 1370 W m-2 (thisis the
longer period of about 400 ky can also be discerned. In so-called total solar irradiance, SO).However, Y, varies
practice, the analytical solution for the eccentricity, over time according to the eccentricity. Therefore,
expressed in trigonometrical series, puts forwards instead of SO,it is often preferred to use the related
many terms having different periods. The major quantity, S, defined at the constant distance a from
periods in the development are 404, 95, 124, 99, Sun. As the solar energy decreases according to the
and 131 ky (in order of decreasing amplitude). The square of the distance from the Sun, the amount of
obliquity (tilt of the Earth's axis) varies between 22" solar radiation received by the Earth on a unit surface
and 25" with a very clear quasi-period of 41 ky. The perpendicular to the Sun's rays at a distance r from the
0.06
0.05
0.04
0.03
0.02
0.01
0.00 0.06
0.04
0.02
0.00
-0.02
-0.04
25 -0.06
24
23
22
100 50 0 -50 -100 -150 -200 -250 -300 -350 -400
(kyr AP) Time (kyr BP)
Figure 4 Long-term variations of eccentricity, climatic precession and obliquity from 400 000 years ago to 100 000 years into the future
(from Berger).
Sun is given by W = s ( ~ / r )Moreover,

~. the incident time in the year 1, using eqn [4].
radiation is usually not perpendicular to the Earth’s
surface, but it is tilted according to the elevation of the sin 6 = sin isin E [41
Sun above the horizon. The elevation angle at a given This shows that the energy (irradiance) available at
point on the Earth is measured by the zenith distance, any given latitude 4 on the Earth (on the assumption of
z , which is the angular distance from the zenith (the a perfectly transparent atmosphere) is a single-valued
point vertically overhead) along the vertical circle function of the total solar irradiance, S, the semimajor
through the point. The instantaneous insolation (irra- axis, a, of the Earth orbit about the Sun, its eccentricity
diance) received by a unit horizontal surface at a e, its obliquity E, and the longitude of the perihelion
particular time characterized by a zenith distance, z, is measured from the moving spring equinox, w.There-
given by eqn [2]. fore it appears that the irradiance varies only accord-
ing to three astronomical parameters, i.e., the
w = S ( a / r ) 2 cos z PI
eccentricity (e), the climatic precession (e sin a)and
the obliquity ( E ) . Moreover, climatic precession and
Spherical trigonometry applied to the astronomical
eccentricity on one side, and obliquity on the other
coordinates on the celestial sphere for the Earth’s
side, appear in two distinct factors in the formulation
orbital motion provides cos z (Figure 5 ) , expressed as
for the irradiance. Each of these factors has a physical
eqn [31. meaning. The distance factor ( p = r / a ) depends on the
cos z = sin 4 sin 6 + cos 4 cos 6 cos H [3j precession and eccentricity, and the inclination factor
(cos z ) is solely a function of the obliquity.
Here 4 is the latitude on the Earth, 6 is the declination The daily irradiation is the irradiance integrated
(the angular distance from the Equator along the over a whole day, either from sunrise to sunset or over
meridian), and H is the hour angle (measured clock- 24 h, in case of no sunset. The 24 h mean irradiance
wise on the Equator from the meridian of the observer (wd),Le., the average daily irradiation over 24 h,
to the meridian of the Sun). The declination depends expressed in M m- 2 , is often preferred. The value of
on the obliquity ( E ) . It can be computed for any given wd depends on the latitude.
Figure 5 Position of a celestial body (the Sun for example) on the celestial sphere. The different variables are explained in the text.
ICE AGES (MILANKOVITCH THEORY) 1001
0 For the latitudes where there is a daily sunrise and affected by variations in precession, although the
sunset, Le., 141 5 71/2 - 161,wd is expressed by eqn obliquity plays a relatively more important role for
[517 high latitudes, mainly in the winter hemisphere. The
variations in the obliquity are perceptible in the same
o 4 sin 6
wd = S / ~ t p - ~ ( Hsin way in both hemispheres (Figure 6Aj, i.e., an increase
+cos 4 cos 6 sin Ho) 151 in the obliquity induces an increase in the insolation
during the local summer (March to September in the
where Ho, the absolute value of the hour angle at Northern Hemisphere and September to March in the
sunrise and sunset, is given by eqn [6]. Southern Hemisphere) and an insolation decrease
during local winter. Consequently the seasonal con-
cos Ho = - t a n 4 tan 6 161 trast in daily irradiation is reinforced. A change in the
0 For the other latitudes, Le., 141 > 71/2 - 1 61: climatic precession (Figure 6Bj such that the June
Either it is the long polar night (Ho = Oj, i.e., summer solstice is moving from the perihelion to the
$6 5 0, in which case wd is given by [7] aphelion (Le., close to the present-day situation)
induces a decrease of insolation over the whole Earth
wd = 0 [71 (Northern and Southern Hemispheres j simultaneous-
ly over one half of the year (Northern Hemisphere
Or it is the long polar day (Ho = nj, i.e., $6 > 0, in
summer season and Southern Hemisphere winter
which case Wd is given by [ 81.
season, i.e., from March to September). Climatic
W, = s p p 2 sin 4 sin 6 precession plays an opposite role in both hemispheres.
181
At present, perihelion occurs in early January. This
The daily irradiation varies through the year as well situation favors mild winters and cool summers in the
as according to the latitude. Moreover, it also exhibits Northern Hemisphere, and cold winters and warm
long-period variations caused by the secular variations summers in the Southern Hemisphere.
in the elements of the Earth's orbit and rotation. Comparison between changes in the orbital param-
Some features of the long-term variations in eccen- eters and/or in the solar radiation received by the Earth
tricity, obliquity, and climatic precession can be with geological reconstruction of past climate changes
discerned in the insolation variations. Low values of is also used to provide a clock for dating the records. In
the eccentricity are mirrored in the small amplitude of this case it is assumed that the quasi-periods observed
the insolation change, such as for the recent past and in the data are a direct response to the quasi-periodic
near future; conversely, large values of e induce large variations of the forcing. Consequently the astronom-
amplitudes of the insolation change, for example, at ical chronology is directly applied to the geological
about 100 ky BP (Figure 2Cj. Solar radiation is mostly data, possibly with a time lag.
Figure 0 Variation of the seasonal contrast of the mean irradiance (W m-') following (A) an increase of the obliquity from 22.5" to 25"
(e = 0.05 and winter at perihelion) and (B) a change in the climatic precession (from summer solstice at perihelion to summer solstice at
aphelion; e = 0.05 and E = 25").
Paleoclimate Modeling a northward shift of the main regions of monsoon

precipitation over Africa and India.
Climate models are simplifications of reality, designed The conceptual models are simple models designed
to describe the complexity of the interactions within to assess whether a climate process can explain
the climate system. These numerical climate models past climate changes. For example, the simple thresh-
can be used to test the astronomical theory of old (or multistate) climate model due to Paillard
paleoclimate, i.e., to study whether astronomically simulates ice volume increase as a function of a
induced changes in insolation are able to drive smoothed truncation of the insolation. The
the climate system into glacial-interglacial cycles model distinguishes three distinct states (interglacial,
similar to these recorded in geological data. The mild glacial, and full glacial) and the transition
modeling effort has led to a better understanding of the between them occurs when insolation and ice volume
physical mechanisms involved in the climate system cross prescribed thresholds in insolation and ice
response to the astronomically forced changes in the volume. This model reproduced reasonably well the
pattern of incoming solar radiation. Such mechanisms succession of glacial-interglacial cycles over the late
are related, in particular, to the ice sheets, the Pleistocene.
lithosphere, the hydrological cycle, the cloud proper- Models of intermediate complexity are the only
ties, the albedo temperature feedback, the land-sea ice climate models to be able to simulate the time-
temperature gradient, the C 0 2 cycle, and the ocean dependent behavior of the fully coupled climate
circulation. The different parts of the climate system, system over a time interval long enough to test the
i.e., the atmosphere, the hydrosphere, the cryosphere, astronomical theory of paleoclimate. Earth system
the biosphere, and the lithosphere, are becoming Models of Intermediate Complexity (EMICs) include
convincingly modeled separately, and work is going most of the processes described in comprehensive
on towards the design of comprehensive coupled models, in particular the slow-response climate vari-
models including several parts, if not all of them. A ables such as ice volume, bedrock depression, deep-
hierarchy of models, climate models of different ocean temperature, and atmospheric concentration of
complexities that differ in their degree of spatial and greenhouse gases. They also simulate the interactions
temporal resolution, are used for paleoclimate between the different parts of the climate system.
purposes. Moreover, they are simple enough to allow for long-
General Circulation Models (GCMs) are primarily term climate simulations (several glacial-interglacial
used for simulating geographic features of paleocli- cycles).
mates. Their major limitation is their high computing The LLN 2D NH climate model (two-dimensional
cost. For this reason they are used for simulations climate model developed in Louvain-la-Neuve) is one
covering a few thousand years at maximum. They of these EMICs. It was designed in order to understand
provide a ‘snapshot’ view of the climate in equilibrium the response of the climate system to astronomical
with the boundary conditions. forcing. It links the atmosphere, the upper mixed layer
At the last interglacial, some 125 ky BP, modeling of the ocean, the sea ice, the continents, the ice sheets,
experiments led to warmer conditions, especially in and their underlying lithosphere. It is forced by
the high latitudes, reduced sea-ice extent, enhanced computed insolation and reconstructed atmospheric
northern tropical monsoon and northward displace- C 0 2 concentration. It considers only the Northern
ment of the tundra and taiga biomes, in good agree- Hemisphere (the Southern Hemisphere is not consid-
ment with geological reconstruction. However, the ered) and it has no explicit representation of the
strong cooling induced by changes in the orbital thermohaline circulation. It has been able to simulate
parameters at 115 ky BP are not sufficient to initiate many of the different situations that characterize the
glaciation, at least if vegetation changes are not last 3 million years: the entrance into glaciation
properly taken into account. This clearly puts forward around 2.75My BP, the dominance of the obliquity
the importance of the vegetation-albedo-temperature cycle during the late Pliocene-early Pleistocene, the
feedback. Several GCMs have been used to simulate emergence of the 100 ky cycle around 900 ky BP, and
the climate of the Last Glacial Maximum, some 20 ky the glacial-interglacial cycles of the last 600 ky. The
ago. Again, important processes at work at that time climatic changes over the Northern Hemisphere, in
were identified. They are related to the C 0 2 concen- particular the continental ice volume, simulated by the
tration, sea ice, ocean temperature, and land albedo. LLN 2D NH climate model during the last 400 ky
As part of the Paleoclimate Modeling Intercomparison shows a broad good agreement with reconstruction
Project (PMIP), several GCMs performed the same (Figure 7).
simulation of the mid-Holocene climate ( 6ky BP). However, a major discrepancy in this model is the
Some robust features have been identified, for instance too frequent melting of the ice sheets during the
ICE AGES (MILANKOVITCHTHEORY) 1003
Figure 7 Comparison of records and modeled data over the last 400 000 years. (A) Variation in the mean irradiance in July at 60" N (red
full line) and the Con concentration (dashed green line). (B)Proxy records for the variation of continental ice volume, Le., stacked,
smoothed oxygen isotope record as function of age in the SPECMAP time scale (full dark blue line), 6 l 8 0 record from the oceanic core
MD900963 (long-dashed blue line), and reconstructed sea level from benthicforams in the oceanic core V19-30 (short-dashed blue line).
(C) Northern Hemisphere continental volume as simulated by the LLN 2D NH climate model.
interglacial. The largest difference between the simu- See also

lated and the reconstructed Northern Hemisphere Carbon Dioxide. Climate Variability: Glacial, Intergla-
continental ice volume appears between 180 and cial Variations. Energy Balance Model, Surface. Gen-
15Oky BP. Moreover an unusual feature shows up eral Circulation: Models. Glaciers. Numerical
between 400 and 350ky BP. This time interval is Models: Methods. Paleoclimatology: Ice Cores; Varves.
characterized by a very long interglacial, which
does not seem to be recorded in data. This behavior Further Reading
is possibly caused by the interplay between insolation Berger A, Imbrie J, Hays J, Kukla G and Saltzman B (eds)
forcing and CO2 concentration forcing. This (1984) Milankovitch and Climate. NATO AS1 Series C,
model also confirms that the orbital forcing acts vol. 126. Dordrecht: Reidel Publishing Company.
as a pacemaker for the glacial-interglacial cycles Bradley R (1999) Paleoclimatology. Reconstructing
and that the climate response to orbital forcing is Climates of the Quaternary. New York and London:
amplified by CO2. Moreover, important processes in Academic Press.
climate change were identified, such as albedo- Imbrie J and Imbrie KP (1979)Ice Ages. Solving the Mystery.
temperature feedback, water vapor-temperature feed- Cambridge, M A and London, England: Harvard Univer-
back, the snow aging process, and the isostatic sity Press.
rebound. Milankovitch MM (1941) Kanon der Erdbestrahlung
und seine Anwendung auf des Eizeitenproblem.
New observation techniques, accurate dating meth- R. Serbian Acad. Spec. Publ. 132, Sect. Math. Nat.
ods, improved transfer functions, and comprehensive Sci., 33. Beograd: Koninglich Serbische Akademie.
climate models will lead to increasingly accurate Reprinted in English: Canon of Insolation and the
knowledge of the past evolution of the atmosphere Ice-Age Problem. Zavod za udzbenikb i nastavna
and the oceans, the waxing and waning of the ice sredstva, Beograd (1998).
sheets, and the growth and retreat of the forests and Roy AE (1978) Orbital Motion. Bristol, Philadelphia and
deserts. New York: Adam Hilger.
1004 INSTABILITY/ Inertial instability
Contents
Inertial Instability
Symmetric Stability
Wave-CISK
Inertial Instability
J A Knox, University of Georgia, Athens, GA, USA Inertial instability is the geophysical equivalent of
Copyright 2003 Elsevier Science Ltd. All Rights Resewed. centrifugal instability and occurs when angular
momentum decreases as one moves outward from
the axis of rotation of the flow. This can be visualized
Introduction on the global scale (Figure 1B) by imagining the
Inertial instability is a fundamental, but infrequently roles of the cylinders being played by latitude lines,
documented, hydrodynamic instability characterized with the Equator serving as the inner cylinder. The
by strongly divergent anticyclonic flow. It is the latitude at which angular momentum is nondecreasing
geophysical analog to the centrifugal (or Taylor- outward from the rotation axis forms the outer
Couette) instability in fluid dynamics examined by cylinder.
Rayleigh and Taylor nearly a century ago. Bergen This analogy between centrifugal and inertial
School meteorologists seeking an explanation for instabilities is correct only if the rotation in the
cyclogenesis pioneered inertial instability research in Taylor-Couette apparatus is equated to the total
the 1930s and 1940s. However, the triumph of vertical rotation of the flow in the geophysical
baroclinic instability theory and balanced dynamics case. In meteorological terminology, the rotation in
in the 1950s and 1960s sharply curtailed interest in Figure 1B is the sum of the planetary and relative
inertial instability. A revival of research during the past vertical vorticities. From this viewpoint, inertial
two decades, particularly in middle-atmosphere and instability should be expected where the relative
mesoscale dynamics, has led t o a growing appreciation vorticity is opposite in sign to the planetary vorticity,
of the role of inertial instability in geophysical flows. and at least equal in magnitude to it. The threshold for
Today, inertial instability arises in a wide range of inertial instability - the latitude corresponding to the
subjects: the dynamics of mesoscale convection and outer cylinder in Figure 1B -is thus the location of zero
monsoons, wave generation and breaking in the absolute vorticity. With some restrictions, this is
stratosphere and mesosphere, and the maintenance proved mathematically in the next section.
of jets in planetary atmospheres and equatorial Inertial instability, like centrifugal instability, leads
oceanography. Below, we examine what inertial to roll-like vortical motions. The flattened, ‘pancake’-
instability looks like physically, how it is represented like divergent circulations of inertial instability
mathematically, and how it is manifested geophysi- (Figure 2 ) attempt to reduce the anticyclonic local
cally. Those new t o the subject may wish to begin with rotation via momentum transports so that the total
the latter. rotation is the same sign throughout the domain.
An inertially unstable flow can exist on a variety of
scales. As illustrated in Figures 1 and 2, the flow can be
Physical Description
as large as a zonally symmetric latitudinal ring of air
Centrifugal instability occurs in the Taylor-Couette around the Earth. It can also occur on scales as small as
problem when the angular momentum of a fluid a mesoscale vortex. When viewed from the perspective
contained between two rotating cylinders (Figure 1A) of a circular vortex, this instability is triggered when a
decreases radially outward, violating Rayleigh’s sta- very strong outward pressure gradient force and the
bility criterion. Parcels then rearrange themselves to centrifugal force combine to overwhelm the Coriolis
achieve a stable radial profile of angular momentum. force and lead to the breakdown of balance.
INSTABILITY/ Inertial Instability 1005
In layman’s terms, the salient point of inertial

instability research is that a planet’s rotation sets a
local rotational ‘speed limit’ beyond which violations
are corrected via inertial instability. This ‘speed limit’
can be indirectly inferred from weather charts in a
number of ways: for example, the strong tendency for
nonnegative absolute vorticity in the Northern Hemi-
sphere, and also the complete absence of intense
‘bull’s-eye’high-pressure centers analogous to ‘bomb’
cyclones.
Inertial instability depends on planetary rotation
and the horizontal shear and/or curvature of the
horizontal wind. In this sense, it shares an affinity with
barotropic instability. However, the rapid, strongly
divergent motions associated with inertial instability
invite analogies with another fundamental, strongly
ageostrophic instability: static instability. This ana-
logy is strengthened by the close similarity in the
derivation of these two instabilities’ criteria, discussed
below.
Mathematical Criteria for Instability

Inertial instability represents a large departure from
geostrophic balance. As a result, inertial instability
theory cannot rely on quasi-geostrophic or nonlinear
balances to facilitate a generalized theory, making it a
far less tractable problem than barotropic or barocli-
nic instability. In fact, no fully three-dimensional
theory for inertial instability exists currently. Instead,
the classic derivation relies on the parcel-mode
approach, an extremely simplified flow geometry,
and ignores nonlinear, frictional, and diabatic effects.
Even so, the result is remarkably useful when applied
to observed and simulated flows.
We begin by assuming that we have a perturbed
parcel embedded in a purely zonal geostrophic flow, as
illustrated in Figure 3. The horizontal equations of
motion describing the parcel are:
f+(=O
In these equations, the horizontal Lagrangian

Figure 1 (A) Centrifugal instability in the Taylor-Couette
derivative D,Dt = a/at + ua/ax + ua/ay and f is the
experiment and (B) inertial instability near the equator. (Adapted Coriolis parameter (assumed to be constant for
with permission from Hua BL, Moore DW and Le Gentil S (1997) simplicity). In eqn [2], the meridional pressure
Inertial nonlinear equilibration of equatorial flows. Journal of Fluid gradient term has been rewritten in terms of the
Mechanics 331 : 345-371 .) geostrophic wind.
These two equations can be combined into one
equation for the meridional ageostrophic wind in the
following manner. Taking the Lagrangian derivative
1006 INSTABILITY / Inertial Instability
Figure2 Schematic view of inertially unstable circulations in (y,2).DIV and CONV refer to regions of divergence and convergence due
to the horizontally divergent inertial circulations, shown in heavy bold arrows. Conservation of mass leads to the vertical motions shown in
the lighter arrows, which adiabatically create the warm and cold temperature anomalies labeled W and C respectively. (Adapted with
permission from Dunkerton TJ (1981) On the inertial stability of the equatorial middle atmosphere. Journalof the Atmospheric Sciences
38: 2354-2364, and Hayashi H, Shiotani M and Gille JC (1998) Vertically stacked temperature disturbances near the equatorial
stratopause as seen in cryogenic limb array etalon spectrometer data. Journal of GeophysicalResearch 103: 19469-19483.
of eqn [2] yields that the parcel will not locally change the geostrophic
wind but it can advect geostrophic relative vorticity
meridionally; thus the analysis is non-quasi-geos-
131 trophic but omits the two-way interaction of the
geostrophic adjustment problem.
The second term in the parenthesis in eqn [3] can be Insertion of eqns [l] and [4] into eqn [3] and
replaced with the right-hand side of eqn [I],coupling rearrangement leads directly to
the two equations of motion. The first term in the
parenthesis can also be related to the meridional
ageostrophic wind via the following approximation:
This second-order differential equation bears a close

141 resemblance to the stability equation for static
instability, with the meridional ageostrophic wind
The relation in eqn [4] derives from the steady, replacing the vertical displacement and f(f + Cg)
zonally uniform mass field in Figure 3. It also assumes replacing the buoyancy frequency. As in the static
L Note that the inertial oscillation period in eqn [6]
- us
P- AP reduces to the usual textbook form if ig = 0; the more
general form in eqn [6] has been applied successfully to
constant-pressure radiosonde trajectory periodicities.
Weak inertial stability is analogous to weak static
stability, in which forcing leads to a larger response
than in strongly stable conditions.
yL X
The criterion in eqn [8] can be interpreted in a
variety of ways. In the Northern Hemisphere, it is
Figure 3 Schematic illustrating the physical situation posed in equivalent to negative geostrophic absolute vorti-
the mathematical derivation of the inertial instability criterion (eqn city. For statically stable conditions, eqn [8] implies
[8]), in which a parcel (small circle) in a background flow with
negative potential vorticity in a geostrophic flow in the
horizontal shear is subject to a perturbation (heavy arrow).
Northern Hemisphere. If eqn [8] is calculated on an
isentropic surface then it is identical to the criterion for
stability problem, we assume a wave solution and symmetric instability (see Instability: Symmetric
obtain the following cases: Stability). Interestingly, the necessary criterion for
barotropic instability is the meridional derivative of
f (f + Cg) > 0 eqn [81.
271 The e-folding times of inertial instability are, by eqn
stable inertial oscillation with period [8], dependent on the latitude and the magnitude of the
[f (f + i g ) l anomalous absolute vorticity. Observations and mod-
eling studies suggest e-folding times as short as a few
hours in the mesoscale midlatitide troposphere and
f (f + rg) = O inertial neutrality 171 around one day in the equatorial middle atmosphere.
These time scales are much longer than for static
instability but are usually shorter than for barotropic
1 instability.
inertial instability with e-folding time Extensions beyond eqn [8] are possible in some
[-f (f + cases. For a circular vortex, the criterion in eqn [8] is
modified only slightly, with the sum of the shear and
Range from -1.3828446e-09 to 2.5954147e-07 s-' Contour = 1.50000e-09
Figure 4 The inertial instability criterion (eqn [8]), calculated from Limb Infrared Monitoring of the Stratosphere (LIMS) geopotential
heights for the period 12-17 December 1978. Negative values, corresponding to inertial instability according to eqn [8], are shaded.
(Reproducedwith permissionfrom Knox JA (1997) Generalizednonlinear balancecriteriaand inertial stability. JournaloffheAtmospheric
Sciences 54: 967-985.)
1008 INSTABILITY/ Inertial Instability
Figure 5 Vertically layered 'pancake structures' in the lower mesosphere in LlMS eddy temperatures (contour interval 1 K, negative
regions shaded) at the international dateline on 10 December 1978, at the beginning of a week-long episode of inertial instability. Note that
the temperature anomalies are strongest just south of 10" N, where the inertial instability is centered in Figure 4, and that anomalies of the
opposite sign exist near 30" N, as predicted by the schematic in Figure 2.
curvature vorticities replacing the geostrophic relative eddy diffusion damps these scales out and leads to a
vorticity. Extension of the analysis to the beta plane preferred intermediate vertical scale. However, the
does not alter the instability criterion, although the instability itself may be the source of the turbulent
condition for stability becomes necessary, no longer eddies that in turn select the vertical scale of the
sufficient. On the sphere, the metric terms due to the instability. Thus the eddy diffusion hypothesis for
Earth's curvature can alter the criterion, but the vertical scale selection, which depends on externally
change is significant only for high wind speeds and1 prescribed approximations such as Rayleigh drag, is
or high latitudes. The vertical equation of motion may incomplete.
be incorporated into the analysis via the thermal wind
law; the resulting instability criterion is eqn [8], with
the absolute vorticity replaced with the Ertel potential Observed Phenomena Related to
vorticity. Friction in the form of Rayleigh drag Inertial Instability
stabilizes the flow linearly, i.e., the more drag there Inertial instability is possible wherever the antic-
is, the less unstablelmore stable the flow is. Extension yclonic relative vorticity rivals the Coriolis parameter
of the theory to zonally asymmetric flows has been in magnitude. Therefore, it is not confined to any one
achieved but does not yield an alternative criterion; in region of the planet, although it is likeliest where the
general eqn [8] is used locally with good results, except Coriolis parameter is smallest, i.e., the tropics.
in very strong anticyclones.
Unfortunately, the assumptions in the analysis
Middle Atmosphere
above preclude any insight into the vertical scale of
the instability. The growth rates for inertial instability Large-scale inertial instability is observed in the
are greatest at smallest scales; it is hypothesized that equatorial lower mesosphere, often lasting about a
INSTABILITY / Inertial Instability 1009
Figure 6 Climatology of inertial instability using National Centers for Environmental Prediction daily geopotential height analyses
(horizontal resolution = 381 km at 60’ N) at 250 mb for December 1966-December 1994 (29 Decembers). Contours indicate the number
of analyses in which eqn [8] was satisfied over the 899 days in the study. (Figure courtesy Russ Schumacher, Colorado State University,
and David Schultz, National Severe Storms Laboratory.)
week and occurring within a week or two of the boreal

winter solstice. The instability develops in regions of
negative absolute vorticity (Figure 4)on the poleward
flank of the stratopause semiannual oscillation east-
erlies; there is also evidence for it around the summer
solstice. Its hallmarks are layered, nearly stationary
‘pancake structures’ in the eddy temperature field
(Figure 5 ) that have a latitudinal width of 10-20
degrees of latitude and a vertical wavelength of
roughly 10 km.
Numerical simulations and observations have
shown that Rossby wave breaking in the tropics acts
as a trigger for the onset of equatorial inertial
instability and influences its zonally asymmetric
Figure 7 Eta model initial analysis of absolute geostrophic nature in winter. The divergent character of the
vorticity (shaded, only negative regions shown: scale is
x l o 5 s - ’), geopotential height (heavy contours; in decameters),
instability, and some recent observational work,
and horizontal divergence (light contours; scale is x 105s ’) at suggest a role for inertial instability in the latitudinal
300 mb over Texas and Oklahoma valid at 0000 UTC on 4 May transport of tracers in the tropical middle atmosphere.
1999. The centroid of tornado reports at this time is indicated with a Recent research strongly indicates that Rossby-
IargeT. Shaded regions, correspondingto inertial instabilityviaeqn wave-triggered inertial instability excites the two-day
[8], are colocated with severe weather and with high values of
divergence. The inertially unstable region over Texas was
wave of the equatorial stratopause region. The
associated with 29 severe weather reports. (Figure courtesy David instability may also play a role in some stratospheric
Schultz, National Severe Storms Laboratory.) sudden warmings and other high-latitude phenomena.
Figure 8 National Weather Service Doppler radar image from Sullivan, Wisconsin (circle at center of image) at 1133 UTC on 14 July
1995. The banded echoes extending east-southeast across central Wisconsin correspond to elevated convection at dawn during a deadly
heat wave.
Upper Troposphere closely related to eqn [SI and is infinite in the case of
inertial instability). However, the evidence for its role
Inertial instability has been sought for in jet stream in both hurricane outflow and severe thunderstorms is
analyses since the 1940s. A long-term climatology of mixed. In Figure 7, the upper-tropospheric flow pattern
inertial instability based on NCEP geopotential for the devastating 3 May 1999 Oklahoma City,
heights (Figure 6 ) reveals why this has been a difficult Oklahoma, tornado outbreak is depicted. Inertially
search: the criterion for the instability is achieved in unstable regions are colocated with strong divergence
the data-rich midlatitudes only once every few years at aloft and severe weather at the surface in this case, but
most. (The occurrence rate is of course higher for not in others. The relationship between inertial
higher-resolution data sets, but is still rare.) However, instability and convection is still not well understood.
Figure 6 suggests that inertial instability is a fairly Inertial instability may also have a connection to
common phenomenon in the subtropical upper tropo- hazardous weather from an aviation perspective. The
sphere, particularly on the equatorward flank of the gravity wave radiation expected from strongly antic-
East Asian jet. Recent work has shown layered yclonic regions, such as inertially unstable circula-
disturbances in low-PV regions near the tropopause tions, may lead to some otherwise unexplained
just south of Japan. instances of clear-air turbulence.
Inertial instability can enhance the outflow from
Lower-to-Middle Troposphere
mesoscale convective systems such as thunderstorms,
‘tropical plumes’, and hurricanes (the Rossby radius of Inertial instability below the jet stream level is
deformation, an estimator of the outflow width, is confined to the tropics and a few rare instances in the
Figure 9 Potential vorticity (.025 PV unit contour interval, only negative values contoured) and equivalent potential temperature (2 K
contour interval) at 4 km over the upper Midwest United States, as determined by a 24-hour forecast from the University of Wisconsin non-
hydrostatic model (UW-NMS; horizontal resolution 6.67 km, vertical resolution 200-1000 m) valid at 1200 UTC 14 July 1995. Note the
region of negative PV extending east-southeastwardacross central Wisconsin. The equivalent potential temperature field indicates the
near-horizontalcharacter of the flow over the region.
midlatitudes. The cross-equatorial Asian monsoon convection and the mean latitude of the Intertropical
circulation appears to accelerate toward the coast of Convergence Zone.
India under the influence of inertial instability. The Intense extratropical anticyclones may possibly
divergence - convergence couplets of inertial instabil- exhibit inertial instability. This is the best explanation
ity appear to determine the location of near-equatorial for an unforecast pre-dawn elevated convection event
10 km
4 km
Axis of maximum inertial instability

Figure 10 Cross-section of 24-hour forecast vertical motion (contoured every 0.025ms-’ = 2.5cms-’) from UW-NMS valid 1200
UTC 14 July 1995. The cross-section slices NNW-SSE through the region of maximum inertial instability in Figure 9 (vertical mark below
horizontal axis). The checkerboard pattern of vertical motion in the vicinity of the inertially unstable region should be compared with the
schematic in Figure 2.
near the center of the mid-July 1995 Chicago heat Some outstanding research issues involving
wave high-pressure system. Elongated bands of light inertial instability include observation and theoretical
precipitation (Figure 8) coincided closely in time and explanation of its onset and three-dimensional
space with a narrow region of negative potential structure; elucidation of its relationship to wave
vorticity (Figure 9). Very high-resolution model dynamics, other instabilities, and balanced dynamics;
simulations of this event, shown in Figure 10, indicate and further investigation of its role in mixing on a wide
a checkerboard pattern of vertical motion (and thus range of scales, from convection to planetary-scale
divergence and convergence) strongly suggestive of flows.
Figure 2 . It appears that the rising motions induced by
the instability were substantial enough to cause
condensation and elevated convection where strong
subsidence would normally be expected. Glossary
Taylor-Couette instability Also known as centri-
Other Geophysical Fluids fugal instability, this is a fundamental fluid
dynamical instability that occurs between
Like the atmosphere, the oceans may also contain differentially rotating cylinders. The bibliographic
inertially unstable flows wherever the anticyclonic entry for Donnelly should satisfy the non-expert’s
current is fast and the effect of the planetary rotation is curiosity.
weak. The layered structure of subthermocline equa- Rayleigh drag An expedient approximation to dis-
torial ocean currents, reminiscent of the middle sipation used in theory and modeling, it assumes
atmosphere ‘pancake structures’ and confined to that dissipation is proportional to the wind speed
within a degree or two of the Equator, has been linked times a constant, i.e., it is a linear drag on the wind
with inertial instability. Furthermore, anticyclonic speed.
ocean eddies, even at high latitudes, can satisfy the Ertel potential vorticity A generalized version of
criterion in eqn [SI. For example, the overwhelming potential vorticity that is valid for baroclinic flows
tendency for ‘spiral eddies’ on the scales of a few and is conserved for adiabatic and frictionless
kilometers to rotate cyclonically has been attributed to conditions.
the limiting effects of inertial instability on small-scale ‘PV unit’ A way of reducing the complexity of the
anticyclones. units of Ertel potential vorticity; 1 PV unit = 10 -
Farther afield, the atmospheres of Mars and the
large gaseous planets are likely venues for inertial
2
, -1
’.
K kg- Values below zero indicate inertial
instability according to extensions to eqn [SI; values
instability owing t o their strong jet structures at low above approximately 1.5 units usually indicate
latitudes. Modeling studies suggest that the very stratospheric air.
strong shears on the equatorial flanks of the Martian Rossby radius of deformation A fundamental length
jets should be inertially unstable, even in long-term scale of the atmosphere that is, among other things,
means. the distance over which the height field adjusts
during the geostrophic adjustment process. It is
Summary also a measure of the radially confined nature of a
vortex and is smallest for cyclones, largest for
After several decades of off-and-on attention, inertial anticyclones.
instability now seems firmly ensconced in the lexicon
of geophysical fluid dynamics. The instability is at its
largest and most observable in connection with
breaking Rossby waves in the tropical middle atmo-
See also
sphere, but it can happen anywhere anticyclonic shear Angular Momentum of the Atmosphere. Anticy-
and/or curvature becomes unusually intense. Its purely clones. Dynamic Meteorology: Balanced Flows;
horizontal origins and flattened pancake circulations Waves. Instability: Symmetric Stability. Laboratory
make it orthogonal to static instability and much more Geophysical Fluid Dynamics. Mesoscale Meteoro-
logy: Mesoscale Convective Systems. Middle Atmo-
difficult t o observe. However, as a strongly ageos-
sphere: Planetary Waves: Semiannual Oscillation; Trans-
trophic instability it otherwise shares much in port Circulation. Monsoon: Dynamical Theory. Ocean
common with static instability. It is perhaps not too Circulation: General Processes. Planetary Atmo-
much of a stretch to call inertial instability by the spheres: Jupiter and the Outer Planets; Mars. Rossby
nickname ‘horizontal convection’, while keeping in Waves. Tropical Meteorology: Inter Tropical Conver-
mind that rotation, not density, is at the heart of gence Zones (ITCZ). Turbulent Diffusion. Vorticity.
inertial instability. Wave Mean-Flow Interaction.
Further Reading Knox JA (1997) Generalized nonlinear balance criteria and

inertial stability. Journal of the Atmospheric Sciences 54:
Andrews DG, Holton JR and Leovy CB (1987) 9 67-9 85.
Middle Atmosphere Dynamics. New York: Academic Knox JA (1997)Possible mechanisms of clear-air turbulence
Press. in strongly anticyclonic flows. Monthly Weather Review
Angel1 JK (1962) The influence of inertial instability upon 125: 1251-1259.
transosonde trajectories and some forecast implications. Mecikalski JR and Tripoli GJ (1998) Inertial available
Monthly Weather Review 90: 245-251. kinetic energy and the dynamics of tropical plume
Bjerknes J (1951) Extratropical cyclones. In: Malone TF formation. Monthly Weather Review 126: 2200-2216.
(ed.) Compendium of Meteorology, pp. 577-598. Bos- Munk W (2001)Spirals on the sea. Scientia Marina 65(Supp.
ton, MA: American Meteorological Society. 2): 193-198.
Blanchard DO, Cotton WR and Brown JM (1998) Orsolini YJ, Limpasuvan V and Leovy CB (1997) The
Mesoscale circulation growth under conditions of tropical stratopause in the UKMO assimilated analyses:
weak inertial instability. Monthly Weather Review 126: Evidence for a 2-day wave and inertial circulations.
118-140. Quarterly Journal of the Royal Meteorological Society
Boyd JP and Christidis ZD (1982) Low wavenumber 123: 1707-1724.
instability on the equatorial beta-plane. Geophysical O’Sullivan DJ and Hitchman M H (1992)Inertial instability
Research Letters 9: 769-772. and Rossby wave breaking in a numerical model. Journal
Clark PD and Haynes PH (1996) Inertial instability on an of the Atmospheric Sciences 49: 991-1002.
asymmetric low-latitude flow. Quarterly Journal of the Rayleigh Lord (1916) On the dynamics of revolving fluids.
Royal Meteorological Society 122: 151-182. Proceedings of the Royal Society of London Series A 93:
Donnelly RJ (1991) Taylor-Couette flow: the early days. 148-154.
Physics Today 44(11): 32-39. Rodwell MJ and Hoskins BJ (1995) A model of the Asian
Dunkerton TJ (1981) On the inertial stability of the summer monsoon. Part 11: Cross-equatorial flow and
equatorial middle atmosphere. Journal of the Atmo- PV behavior. Journal of the Atmospheric Sciences 52:
spheric Sciences 38: 2354-2364. 1341-1356.
Dunkerton TJ (1993)Inertial instability of nonparallel flow Rosier SM and Lawrence BN (1999) The January 1992
on an equatorial fl-plane. Journal of the Atmospheric stratospheric sudden warming: A role for tropical inertial
Sciences 50: 2744-2758. instability? Quarterly Journal of the Royal Meteorologi-
Emanuel KA (1979) Inertial instability and mesoscale cal Society 125: 2575-2596.
convective systems. Part I: Linear theory of inertial Sato K and Dunkerton TJ (2002) Layered structure
instability in rotating viscous fluids. Journal of the associated with low potential vorticity near the tropo-
Atmospheric Sciences 36: 2425-2449. pause seen in high-resolution radiosondes over Japan.
Hayashi H, Shiotani M and Gille JC (1998) Vertically Journal of the Atmospheric Sciences 59: in press.
stacked temperature disturbances near the equatorial Sawyer JS (1949) The significance of dynamic instability in
stratopause as seen in cryogenic limb array etalon atmospheric motions. Quarterly Journal of the Royal
spectrometer data. Journal of Geophysical Research Meteorological Society 75: 364-374.
103(D16):19469-19483. Smith AK and Riese M (1999) Cryogenic Infrared Spectro-
Hayashi H, Shiotani M and Gille JC (2002)Horizontal wind meters and Telescopes for the Atmosphere (CRISTA)
disturbances induced by inertial instability in the observations of tracer transport by inertially unstable
equatorial middle atmosphere as seen in rocketsonde circulations. Journal of Geophysical Research 104(D16):
observations. Journal of Geophysical Research 10.10291 19171-191 82.
2001JD000922 31 July 2002. Stevens DE (1983) On symmetric stability and instability of
Hitchman MH, Leovy CB, Gille JC and Bailey PL (1987) zonal mean flows near the equator. Journal of the
Quasi-stationary zonally asymmetric circulations in the Atmospheric Sciences 40: 882-893.
equatorial lower mesosphere. Journal of the Atmospheric Stevens DE and Ciesielski PE (1986) Inertial instability of
Sciences 44: 2219-2236. horizontally sheared flow away from the equator. Journal
Hoskins BJ (1974) The role of potential vorticity of the Atmospheric Sciences 43: 2845-2856.
in symmetric stability and instability. Quarterly Taylor GI (1923) Stability of a viscous liquid contained
Journal of the Royal Meteorological Society 100: between two rotating cylinders. Philosophical Transac-
480-482. tions of the Royal Society of London Series A 223:
Hua BL, Moore DW and Le Gentil S (1997) Inertial 289-343.
nonlinear equilibration of equatorial flows. Journal of Tomas RA and Webster PJ (1997) The role of inertial
Fluid Mechanics 331: 345-371. instability in determining the location and strength of
Hunt BG (1981) The maintenance of the zonal mean state near-equatorial convection. Quarterly Journal of the
of the upper atmosphere as represented in a three- Royal Meteorological Society 123: 1445-1482.
dimensional general circulation model extending up Wilson RJ (1997)A general circulation model simulation of
to 100 km. Journal of the Atmospheric Sciences 38: the Martian polar warming. Geophysical Research
2172-21 86. Letters 24: 123-126.
1014 INSTABILITY/ Symmetric Stability
Symmetric Stability
H 6 Bluestein, University of Oklahoma, Norman, wet-bulb potential temperature instead of surfaces of
OK,USA potential temperature), then the atmosphere is in a
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. state of conditional symmetric instability (CSI). CSI is
analogous to conditional instability for air parcels
lifted vertically. Since CSI involves forces that are both
Introduction horizontally and vertically directed, the process by
which the instability is released is also referred to as
Symmetric stability is a state of the atmosphere in slantwise convection. When a layer of moist air that is
which an inviscid, dry-air parcel displaced from its initially symmetrically stable is lifted to saturation,
equilibrium position with respect t o some axis along and the vertical displacement of air itself creates the
or about which the flow has no variations, i.e., along or conditions for slantwise convection, then the process is
about an axis of symmetry, experiences a restoring referred to as potential symmetric instability (PSI),
force which makes it oscillate about its original which is analogous to potential instability for
position. For axially symmetric displacements in a upright convection. At saturation, CSI and PSI are
ring about an axially symmetric vortex the wave equivalent.
motions are called inertial or centrifugal waves. The importance of CSI is that it is thought to be
Centrifugal oscillations are like buoyancy waves responsible for the formation of some mesoscale bands
with the horizontal centrifugal (inertial) force playing of precipitation that are oriented along the thermal
the role of buoyancy (gravity). wind. Since the thermal wind is oriented along the
Similar oscillations can also occur in a statically elongated zone of strong temperature gradient asso-
stable, rotating atmosphere when the thermal ciated with fronts and is quasi-two-dimensional, CSI
wind shear vector is unidirectional and does not vary may be triggered in response to slantwise, ageostro-
along the direction it is oriented. Parcels in the form phic, frontal circulations initiated by confluence/
of a tube are displaced in a vertical plane normal to diffluence acting on a cross-frontal temperature gra-
the thermal-wind vector. In this case the axis of dient. It is also thought that CSI may be responsible for
symmetry is the axis along which the thermal wind is eyewall rainbands in some tropical cyclones.
directed.
If potential temperature increases with height and
the geostrophic absolute vorticity is anticylonic then
the atmosphere is inertially unstable; if the potential The Parcel Theory of Symmetric
temperature decreases with height and the geostrophic Instability in an Inviscid, Dry
absolute vorticity is cyclonic then the atmosphere is Atmosphere
gravitationally (or statically) unstable. If the geo-
The analysis of symmetric stability is simplified by
strophic absolute vorticity is cyclonic and potential
temperature increases with height then the atmos- using a parcel approach analogous to that used in
phere is both inertially stable and gravitationally the parcel theory of upright convection. Consider a
stable; however, if infinitesimal displacements in the Cartesian coordinate system in which there is a
plane normal t o the vertical shear are accompanied by temperature gradient in the y-p plane and that a/&
forces that move the air parcel farther away from its of all variables is zero (this choice of an axis of
equilibrium position then the atmosphere is symme- symmetry is arbitrary; sometimes the y-axis is chosen
trically unstable. Since the atmosphere is baroclinic, to be the axis of symmetry). For simplicity, the
dynamics are described for the Northern Hemisphere.
owing t o the thermal wind, this instability is a special
case of baroclinic instability for a flow in which there is Consider the quantity
no temperature gradient component along the axis of
symmetry. m=u-fy [I1
When tubes of moist, unsaturated air are lifted in a
symmetrically stable atmosphere t o a level at which where u is the x-component of the wind and f is the
condensation occurs (and water and ice loading are Coriolis parameter. In inviscid flow in, the absolute
not significant or are neglected) and thence t o a level at momentum or pseudo-angular momentum, is con-
which the atmosphere is symmetrically unstable with served; it is attributed to an infinitesimal tube of air
respect to saturated processes (i.e., when vertical extending through some point (y,p) infinitely off in
trajectories follow surfaces of constant equivalent or both the +x- and -x-directions.
INSTABILITY / Symmetric Stability 1015
The inviscid momentum equation in the y-direction, or d, and if aO/az > 0 and am,/ay<O. If aO/az<O
with height as the vertical coordinate, is (Figure 1D) or if am,/ay > O (Figure l E ) , then the
atmosphere is statically unstable or inertially un-
stable, respectively, and not symmetrically unstable.
Figure 1A depicts neutral stability and Figure 1C
where u is the y-component of the wind, p the density, depicts absolute stability.
The thermal wind relation in terms of potential
p the pressure, D/Dt the total (material) derivative,
and the geostrophic absolute momentum temperature is to a good approximation
mg = ug - fY 131 151
where ug is the geostrophic component of the wind
in the x-direction. Therefore the net force in the The slope of a surface of constant 0 is therefore
y-direction on a tube is proportional to the difference
between the m of the tube, which is conserved, and the
mgof the environment into which the tube is displaced.
It is assumed for simplicity that the tube does not mix and the slope of a surface of constant mg is
with its environment. The original value of m of the
tube is just its geostrophic value at its equilibrium
point in the y-p plane and that a/ax plane. Therefore 171
there will be a net force in the y-direction on the tube if
it moves into an environment where mg is different It follows that the necessary condition for symmetric
from that of its equilibrium, starting location. Vertical instability is
gradients of mgare associated with thermal wind shear
in the x-direction (i.e., with temperature gradients in
the y direction); gradients of mg in the y-direction are
associated with geostrophic absolute vorticity.
The inviscid vertical equation of motion is where Ri is the Richardson number for the geostrophic
wind and c,is the geostrophic vorticity, which for
[41 symmetric flow (a/ax = 0) is -aug/ay. In typical
synoptic-scale flow in midlatitudes the geostrophic
where w is the vertical velocity, g the acceleration of vorticity is an order of magnitude smaller than f ; then
gravity, 0 the potential temperature of the environ- the necessary condition for symmetric instability is
ment, and 0’ the potential temperature of the tube. If that Ri< 1. In the vicinity of fronts where geostrophic
the flow is adiabatic and there is no diffusion of heat vorticity is much larger, Ri must be smaller.
then 0’ is conserved following the motion of the tube. It Ertel’s potential vorticity for an atmosphere in
is assumed that the environment is not disturbed by the geostrophic and hydrostatic balance is
tube’s motion so that there is no vertical perturbation
pressure-gradient force. Therefore there will be a net 191
force in the vertical on the tube if it moves into an
environment where 0 is different from that of its
equilibrium, starting location. It follows from eqn [8] that an equivalent necessary
Whether or not there is a restoring force on the tube condition for symmetric instability is that Ertel’s
potential vorticity for the geostrophic wind is negative
that brings it back to its equilibrium point about which
it undergoes a stable oscillation (symmetric stability) (anticyclonic in either hemisphere). Since
or whether is continues to move in the direction of its
displacement (symmetric instability) depends on how
(I,+ f),= (c, f L -
+ f
the surfaces of mg and 0 are oriented and what the
direction of displacement is with respect to the where (5, + f ) , is the geostrophic absolute vorticity
surfaces (Figure 1).Symmetric instability is possible evaluated on an isentropic surface and (5, f),= +
(Figure 1B) if the slope of the 0 surfaces is greater than f - au,/ay is the geostrophic absolute vorticity evalu-
the slope of the mgsurfaces and if the tube is displaced ated on a surface of constant height, then nega-
infinitesimally along a plane whose slope is interme- tive (anticyclonic in either hemisphere) isentropic
diate between that of the 0 surfaces and that of the mg geostrophic absolute vorticity is also an equivalent
surfaces (i.e., along paths a or c, but not along paths b necessary condition for symmetric instability.
1016 INSTABILITY / Symmetric Stability
Ri <-
ig+ f
>-
I
Ri
ig+ f
[$>O] %>O
aY
Figure 1 Stability criteria for symmetric flow @/ax = 0) in terms of the slope of the mg (solid lines) and potential-temperature B (dashed
lines) surfaces. Displacements in the directions a. b. c , and dare referred to in the text. (A) Neutral stabilityfor displacements along the B
and m, surfaces; otherwise stable for infinitesimal displacements. (B) Symmetric instability: unstable for infinitesimal slantwise
displacements intermediate in slope between that of Q surfaces and m, surfaces; otherwise stable. (C) Absolute stability: stable for any
infinitesimal slantwise displacement. (D) Gravitational instability: unstable for any infinitesimal slantwise displacement not along a Q
surface. (E) Inertial instability: unstable for any infinitesimal slantwise displacement not along an mg surface. (Adapted with permission
from Bluestein HB (1993) Synoptic-Dynamic Meteorology in Midlatitudes. Vol. I/: Observations and Theory of Weather Systems. New
York: Oxford University Press.)
It can also be shown that the necessary condition for tions. Thus, balanced frontal circulations are possible
symmetric stability is equivalent to the ellipticity only if the atmosphere is symmetrically stable.
condition for the Sawyer-Eliassen equation, which
describes the vertical circulation about a front forced
by geostrophic confluence/diffluence and differential The Parcel Theory of Slantwise
diabatic heating and whose dynamics are governed by Convection in an Inviscid, Moist
the geostrophic-momentum approximation. Since the
Sawyer-Eliassen equation is a second-order, constant-
Atmosphere
coefficient, partial differential equation, the condition The analysis of symmetric instability in a moist
of ellipticity is necessary for it to have unique solu- atmosphere is complicated by latent heat release, and
INSTABILITY/ Symmetric Stability 1017
water and ice loading. The governing momentum Z

equation remains eqn [2]. The governing vertical
equation of motion, on the other hand, is different
from eqn [4] since it must account for latent heat
release, and if there is condensate, for water and ice
loading also. Surfaces of constant entropy that
account for latent heat release and for condensate
loading replace potential temperature in eqn [4]. If
both the environment and the tube are unsaturated
and there is no condensate, then virtual potential
temperature (e,) may be used in place of potential
temperature. If the tube is saturated and the environ-
ment unsaturated, and condensate ignored, then
potential temperature of the tube may be replaced by
equivalent virtual potential temperature (eev);if both
the tube and the environment are saturated, and
Figure 2 Idealized example of a vertical cross section in the
condensate ignored, then the potential temperature of Northern Hemisphere, normal to the thermal-wind shear vector,
both the tube and environment may be replaced by showing surfacesof constant m, (solid lines), and constant 8,.Oev.
equivalent virtual potential temperature. (Wet-bulb and eWvIn this example 8,. Oev. and Ow (dashed lines) increase with
virtual potential temperature (ewv) may be used height (gravitational and conditional stability), mg decreases with
increasing y (inertial stability), and 6, decreases with y (baroclinic
instead of equivalent virtual potential temperature.) atmosphere, with colder air at larger values of y). Lifting conden-
For the purpose of illustration consider an atmos- sation level (LCL); level of free slantwise convection (LFS); level of
phere that is unsaturated and has no condensate, but is neutral buoyancy (LNB). Below the LCL the slope of the 8, surface
moist. Suppose that the distribution of e,, Bey, and mg is less than that of the mg surface. Note that the slope of Be, and Ow,
is as shown in Figure 2. Since the slope of the surfaces surfaces is greater than the slope of 8, surfaces because the lapse
rate of a 8, surface is greater than that of a Oev or Ow surface, and
of constant 0,. are not steeper than the surfaces of
8,.8,,. and Ow, decrease with y. (Adapted with permission from
constant m g at low levels, the atmosphere there is Bluestein HB (1993) Synoptic-Dynamic Meteorology in Midlati-
symmetrically stable or even neutral with respect to tudes, Vol. 11: Observations and Theory of Weather Systems. New
unsaturated displacements. However, there are re- York: Oxford University Press.)
gions aloft where the slope of surfaces of constant Qev
are steeper than the surfaces of mg.In these regions, if
condensate loading is ignored then the atmosphere is saturated and the slope of the surfaces of constant
is symmetrically unstable with respect to infin- Qev is greater than the slope of the surfaces of constant
itesimal saturated displacements. Since the stability mg because the tube has undergone a finite displace-
depends upon whether or not a tube is saturated or ment rather than an infinitesimal displacement.
unsaturated, the symmetric instability condition is If the tube is lifted further, however, so that
conditional. eventually it crosses to the left of the original m g
Suppose an unsaturated tube at low levels is lifted a surface, and it is displaced upward and to the right at a
finite distance along a surface of constant Qv (e.g., by slope intermediate between that of the Bey and mg
the ascending branch of a frontal circulation or more surfaces, then according to eqns [2] and [4] it would
slowly as a result of quasigeostrophic forcing) until it continue to accelerate upward and to the right if it
reaches its lifting condensation level (LCL) and that were not forced any more. The level at which it would
condensate is ignored. If lifted any further, it follows a first realize symmetric instability is called the level of
surface of constant Be,. Thus far the tube is neutrally free slantwise convection (LFS), in analogy with the
buoyant. Owing to the inclusion of the effects of latent level of free convection (LFC) for upright convection.
heat, the surfaces of constant Be, have slopes different Eventually the tube will reach a level at which it
from that of the surfaces of constant e,. The m of the crosses back to the right side of the original m gsurface
tube is greater than that in its environment everywhere where the slope of the 8," surfaces is now less than that
to the right of the original mg surface; therefore of the mgsurfaces. Above this level, the level of neutral
according to eqn [2] the tube is symmetrically stable buoyancy (LNB) for slantwise convection, the atmos-
because it feels a restoring force that has a component phere is symmetrically stable. The LNB tends to
to the left; were the tube forced no further, it would be near the tropopause, where 0 surfaces are more
become negatively buoyant and move back down and horizontally oriented owing to the strong static
to the left towards its original equilibrium position. stability (eqn [6]) in the lower stratosphere and where
The tube is symmetrically stable even though it mg surfaces are more vertically oriented owing to the
1018 INSTABILITY / Symmetric Stability
lack of horizontal temperature gradient and thermal- of Ug in midlatitudes, the horizontal scale of CSI
wind shear (eqn [ 7 ] at
) the tropopause. precipitation bands is on the order of 100 km, which is
The amount of kinetic energy it takes to lift a tube to mesoscale.
its LFS is called the slantwise convective inhibition
(SCIN).The potential energy in the environment that
is converted into kinetic energy of the tube while it is Observations of Conditional
symmetrically unstable is called the slantwise convec- Symmetric Instability
tive available potential energy (SCAPE). SCIN and Slantwise convection may be triggered suddenly after
SCAPE are the analogs to the convective inhibition a gradual buildup in SCAPE by synoptic-scale or
(CIN) and convective available potential energy mesoscale processes, or it may be in statistical equi-
(CAPE) in the parcel theory of upright convection. librium so that the SCAPE is nearly zero and constant.
The SCAPE is equivalent to CAPE computed along a Since the latter is the frequently observed case, it is not
surface of constant m,. In the geostrophic coordinate easy to use SCAPE as a forecast tool because its
system used in semigeostrophic theory, m gsurfaces are absence does not preclude the possibility that CSI is in
parallel to the geostrophic coordinate, which is fact being released; the existing observational network
directed opposite in direction to the horizontal tem- cannot resolve the space and time scales of the
perature gradient. Thus, SCAPE is CAPE computed in production of CSI and its release when the atmosphere
geostrophic coordinates. In the limit of vanishing is in statistical equilibrium.
baroclinicity, m g surfaces become vertically oriented Bands of precipitation ahead of warm fronts may
(see eqn [7]when au,/az -+ 0) and SCAPE is identical be due to CSI. The bands are important in that
to CAPE. large amounts of precipitation may accumulate in
If the atmosphere is saturated then an equivalent mesoscale regions, which makes the forecasting of
necessary condition for CSI is that Ertel’s potential floods and heavy snow difficult; synoptic-scale
vorticity for saturated moist processes is negative dynamics, on the other hand, can explain only how
(anticyclonic in either hemisphere). If the effects of lower amounts of precipitation accumulate over
condensate are accounted for then the criteria for CSI broader regions. The slanted convection in the
are more complicated. The loading depends on what eyewall of some rapidly intensifying tropical cyclones
phase of water substance is present; condensate that
might also be a result of CSI. In this case, the
falls out does not follow air parcel motion and may
temperature gradient is directed toward the center of
evaporate into unsaturated air. the tropical cyclone.
Symmetrically unstable tubes undergoing slant- When Ertel’s potential vorticity for moist processes
wise ascent must be compensated for by regions of
is very small, i.e., when the atmosphere is nearly
slantwise-descending air. The regions of slantwise-
neutral with respect to CSI, the atmosphere’s response
descending air modify the environment so as to to frontogenetical forcing is enhanced. Since frontal
make the tubes less symmetrically unstable, just as secondary circulations and the secondary circulations
compensating subsidence around buoyant clouds in tropical cyclones themselves can produce preci-
(according to the slice method in the theory of up-
pitation bands it is therefore not always easy to
right convection) warms the environment and lessens distinguish between bands forced by the secondary
the buoyancy in the clouds (i.e., the CAPE is dimin- circulations and the bands forced by CSI.
ished). The narrower the slantwise ascending branch
and the wider the slantwise descending branch, the
less is the slantwise acceleration (i.e., the less is the See also
SCAPE). The most unstable configuration is one of
Dynamic Meteorology: Overview. Instability: inertial
thin, relatively rapidly slantwise-ascending layers
Instability. Mesoscale Meteorology: Cloud and Precip-
of saturated air sandwiched in between thick, less itation Bands.
rapidly slantwise-descending layers of unsaturated air
that are being cooled evaporatively from precipitation
falling out from above. If the most unstable configu- Further Reading
ration is the one most likely to occur, then CSI
Bluestein HB (1993) Synoptic-Dynamic Meteorology in
precipitation bands should be relatively narrow and Midlatitudes, Vol. 11: Observations and Theory of
widely spaced. Weather Systems. New York: Oxford University Press.
The horizontal scale of CSI precipitation bands Emanuel KA (1994) Atmospheric Convection. New York:
estimated from the horizontal extent of a sloping m g Oxford University Press.
surface (eqn [ 7 ] )is U,/f,where U , is the change in Holton JR (1992) A n Introduction t o Dynamic Meteoro-
geostrophic wind in the layer of CSI. For typical values logy. New York: Academic Press.
INSTABILITYI Wave-CISK 1019
Lilly DK (1986) Instabilities. In: Ray PS (ed.) Mesoscale Thorpe AJ (1999) Dynamics of mesoscale structure associ-
Meteorology and Forecasting, pp. 259-271. Boston MA: ated with extratropical cyclones. In: Shapiro MA and
American Meteorological Society. Gronas S (eds)T h e Life Cycles of Extratropical Cyclones,
Schultz DM and Schumacher PN (1999) The use and pp. 285-296. Boston, MA: American Meteorological
misuse of conditional symmetric instability. Monthly Society.
Weather Review 127: 2709-2732. Boston, MA:
American Meteorological Society.
Wave-ClSK
C S Bretherton,University of Washington, Seattle, WA, integrated horizontal moisture convergence. The
USA Arakawa-Schubert scheme, a typical ‘quasi-equilibri-
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. um’ closure in which the convective mass flux is chosen
so as to regulate the local convectively available
potential energy, does not support wave-CISK, as
Introduction shown by Stark. Neither does the Betts-Miller scheme,
a typical moist adjustment convective parameterizat-
In 1964, Charney and Eliassen introduced the term
ion, as shown by Neelin and Yu, although Hayashi and
CISK (conditional instability of the second kind) to
Golder showed that CISK could be excited if the
describe a positive feedback between deep moist
convective adjustment turns on and off frequently. In
convection and a large-scale circulation. They hy-
recent years, moisture-convergence-based convective
pothesized that hurricane intensification was such a
parameterizations have been criticized for using a
process, in which surface friction helps to induce low-
nonlocal measure (moistureconvergence) to regulate a
level convergence into a vortex, resulting in deep
local thermodynamic process (convection), allowing
convection and latent heating that amplify the vortex.
unrealistic soundings to develop. Furthermore, the
In the late 1960s, Yamasaki and Hayashi first consid-
development of wave-CISK is strongly influenced by
ered the feedbacks between deep convection and large-
the vertical profile of convective heating perturbation
scale equatorial waves. In their models, convection
selected by the parameterization. ‘Top-heavy’ heating
could intensify (destabilize) the wave in some cases
perturbations concentrated in the upper troposphere
through purely inviscid processes not dependent on
are most favorable for wave-CISK instabilities.
feedbacks with surface drag or surface thermodynam-
ic fluxes. In 1974, Lindzen termed this destabilization
wave-CISK, to distinguish it from Charney and
A Simple Wave-CISK Model
Eliassen’s frictional CISK, and put forth perhaps the
most expansive view of the role of wave- In its simplest form, wave-CISK can be phrased in
CISK, implicating it in the development of tropical terms of a nonrotating inviscid gravity wave interact-
circulations from squall lines to the Madden-Julian ing with a simple parameterization of moist convec-
Oscillation. tion. Wave-induced perturbations in surface heat flux
Wave-CISK has proved a somewhat slippery hy- and radiative fluxes are neglected. Classically, a linear
pothesis to test, and has largely fallen from favor stability analysis of a small-amplitude wave is used to
among specialists in convective dynamics. The pre- assess the convective feedback.
dicted instabilities are very sensitive to the represen- We consider the mathematical structure of an
tation of cumulus convection. For simple models the extremely simple wave-CISK model based on small-
fastest-growing instabilities have very short wave- amplitude (linear) two-dimensional inviscid hydro-
lengths, and are not clearly separable from conven- static motions of a nonrotating atmosphere. The
tional conditional instability of individual cumulus pressure velocity w is assumed to be zero at the mean
clouds. However, wave-CISK is a mode of instability surface pressure p = p s . At the tropopause pressure pt,
permitted by many convective parameterizations, a similar boundary condition or a boundary condition
including some used in climate models, so it can be a that allows upward-propagating gravity waves to
useful concept in interpreting model output even if radiate out of the domain may be applied.
physically dubious. The mean convection is assumed to maintain the
The class of convective parameterizations that tend same profile of water vapor mixing ratio q ( p ) every-
to lead to wave-CISK instabilities are those that where. This is a very strong simplifying assumption on
diagnose convective mass flux based on column- the column moisture budget, and one of the weakest
1020 INSTABILITY / Wave-CISK
links in simple wave-CISK models. In addition, most Raymond has given a simple physical description of
such models have not explicitly attempted to maintain wave-CISK by regarding the convective heating as a
a consistency between the mean temperature and series of pulses, each of which acts as a vertically
moisture profiles that ensures that boundary layer air distributed source of gravity waves. He then showed
will be conditionally unstable, and hence able to that for a wave moving at a particular phase speed, the
convect, but will not penetrate much above the gravity wave generated by the pulse heating would
tropopause. Such consistency requires a basic state in generate low-level horizontal convergence in phase
which the moist static energy is similar at the tropo- with the preexisting wave, causing the wave to
pause to that in the boundary layer, which relates the amplify.
assumed q(ps)to the assumed troposphere-mean static An important feature of the eigenvalue problem is
stability. that the horizontal wavenumber k cancels out of the
The perturbation convective heating in any column equation. This implies that if an instability is present,
is assumed to be caused by the conversion of the shorter wavelengths will grow fastest (no short-wave
converged moisture into rainfall; the resulting latent cutoff), since the growth rate is proportional to
heating is redistributed through the column by turbu- Im(kc). This suggests that wave-CISK might manifest
lent convection according to a fixed vertical heating itself at the shortest scale at which the model assump-
profile yj(p), which is normalized to have a mean value tions are still plausible. Since the model assumes an
of unity averaged over the depth of the atmosphere. ensemble convective heating response, this scale
The thermodynamic equation for geopotential pertur- would have to be somewhat larger than the spacing
bation +(x,p:t ) is eqn [l]. between convective clouds or cloud groups. However,
on such scales other mechanisms such as cold-pool
dynamics seems to play a more important role in
organizing convection into mesoscale systems.
The short-wavelength behavior is related to the
In eqn [l],x is the horizontal coordinate, t is time, 0 is
difficulty of clearly separating wave-CISK in this
the static stability parameter, M(x, t ) is the perturba-
model from conventional conditional instability of
tion column-integrated horizontal moisture conver-
individual cumuli. For a given temperature profile,
gence, R and Cp are the gas constant and isobaric
moistening the near-surface air will render the profile
specific heat of air, and L is the latent heat of
more conditionally unstable to individual cumuli. Our
vaporization for water vapor. Using the continuity
simple CISK model exhibits a distorted version of this
equation, the moisture convergence can be written as
same mechanism, rendering dubious the notion of
eqn 121. CISK as a distinct instability of cumulus cloud
ensembles. To see this, we can integrate eqn [ 2 ] by
parts as in eqn [4].
Applying the horizontal momentum and continuity

141
equations results in a linear, separable equation for
+,e, t ) (eqn ~31). This expresses the parameterized moisture conver-
gence, and hence the cumulus-induced heating rate, as
proportional to the vertical velocity weighted by
dq/dp . This heating counteracts the adiabatic cooling
With a rigid-lid boundary condition, w(ps) = o(pt) associated with rising motion, reducing the effective
= 0. A radiation boundary condition would relate w(pr) static stability of the lower and mid-troposphere. If the
to its vertical derivative. In either case, this equation near-surface air is sufficiently humid, this effective
admits normal modes of the form o = &(p) static stability can become negative, promoting short-
exp(ik[x - ct]). In the case of constant CJ, Cho and wavelength CISK instabilities. Within the framework
Pendlebury showed how the complex phase speed c of our CISK model, this condition of negative effective
can be calculated by a transcendental equation most static stability plays the same role as conventional
easily derived by Fourier-analyzing &(e) and the conditional instability does for growth of individual
normalized heating profile q ( p ) . Unstable modes cumuli.
(Im(c) > 0 ) exist for a variety of heating profiles. Use The stability analysis is more involved for convec-
of a radiation boundary condition instead of a rigid lid tive parameterizations that do not employ a moisture
at the tropopause usually has a minor effect on the convergence closure, and only a few such studies
unstable modes. have been published. The complication is that the
INSTABILITY/ Wave-CISK 1021
right-hand side of eqn [3], which involves the hori- mulus heating, his theory also predicted new classes of
zontal Laplacian of the heating, is not usually express- boundary-trapped CISK modes. Emanuel found a
ible purely in terms of w.Although a similar eigenvalue baroclinic wave-CISK mode varying perpendicularly
problem for c can often still be formulated, it usually to the wind shear, somewhat akin to symmetric
must be solved numerically by vertical discretization instability, which can be excited in a broader set of
and may now have a short-wave cutoff. conditions than classical wave-CISK modes. These
theories await decisive testing against observations
and more sophisticated numerical models.
Elaborations on the Basic Model
Many elaborations on the above model have been Current Status of Wave-CISK
proposed. In 1970, Hayashi extended a similar model
to continuously stratified motions on an equatorial Neither observations nor current cloud-resolving nu-
beta-plane using separation of variables in the merid- merical model simulations clearly show classical
ional direction. In particular, the equatorial Kelvin wave-CISK-like modes. Furthermore, the theoretical
wave has a similar zonal structure and growth rate to a models that predict wave-CISK are based on dubious
nonrotating gravity wave. This has led to many parameterizations of cumulus convection. However,
theories that rationalize the tropical Madden-Julian there are many intriguing indications that moist
(intraseasonal) oscillation as a wave-CISK mode. convection may in fact help destabilize some large-
More sophisticated general circulation model simula- scale waves through mechanisms not considered in
tions using moisture-convergence-based convective classical wave-CISK. Some of these include: (i) the
parameterizations, starting with Hayashi and Sumi in effect of wave-associated surface wind perturbations
1986, have also frequently produced intraseasonal on the surface fluxes and boundary layer structure
oscillations that have been interpreted as wave-CISK. (wind-induced surface heat exchange, or WISHE); (ii)
In 1979, Davies obtained a short-wavelength radiative feedbacks on the wave associated with
cutoff by assuming a short delay between the mois- convectively produced anvils or moisture redistribu-
ture convergence and the convection. With a tion; and (iii) feedbacks between the convection and
30-60-minute delay, the fastest growing wavelength the humidity profile in the convecting column. The
is a few hundred kilometers, corresponding to a typical importance of these convective/large-scale feedbacks
size of a mesoscale convective system. In 1987, Lau in producing transient variability in the tropics and
and Peng considered ‘positive-only heating’, in which parts of the mid-latitudes on all time scales remains an
only upward motion (creating moisture convergence) active and very important topic of research.
produces perturbation heat release. This can be
thought of as a simple way to represent waves suffi- See also
ciently strong to suppress all convection in their
subsiding branches, but still weak enough to be Baroclinic Instability. Dynamic Meteorology: Poten-
tial Vorticity. Hamiltonian Dynamics. Instability: Inertial
approximated by linear dynamics. Such models pro-
Instability; Symmetric Stability. Kelvin-Helmholtz Insta-
duce unstable modes with a propagating narrow band
bility. Lagrangian Dynamics. Quasi-geostrophic
of ascent surrounded by a broad subsidence region. Theory. Rossby Waves. Vorticity.
This could be considered as a parameterized repre-
sentation of the circulation around a single intense
cumulonimbus. Further Reading
Additional physical feedbacks have been consid- Charney JG and Eliassen A (1964) On the growth of the
ered. In 1990 Wang and Rui considered frictional hurricane depression. Journal of the Atmospheric Sci-
wave-CISK, the impact of surface friction and con- ences 21: 68-75.
vective heating on an equatorial wave, and found that Cho H-R and Pendlebury D (1997)Wave CISK of equatorial
the surface drag could stimulate a pattern of convec- waves and the vertical distribution of cumulus heating.
tive heating that helps destabilize an equatorial Kelvin Journal of the Atmospheric Sciences 54: 2429-2440.
wave. Mak considered the feedback of cumulus Davies HC (1979) Phase-lagged wave-CISK. Quarterly
Journal of the Royal Meteorological Society 105:
convection (represented via eqn [2]) with an Eady
325-353.
model of mid-latitude baroclinic instability. He Emanuel KA (1982) Inertial instability and mesoscale
showed that the most unstable Eady mode becomes convective system. Part 11: Symmetric CISK in a baro-
shorter and intensifies more rapidly when moderate clinic flow. Journal of the Atmospheric Sciences 39:
cumulus-driven latent heating is included; these effects 1080-1 092.
can be interpreted in part as consequences of a reduced Hayashi Y (1970) A theory of large-scale equatorial waves
effective static stability. With sufficiently strong cu- genearated by condensation heat and accelerating the
1022 IONOSPHERE
zonal wind. Journal of the Meteorological Society of Mak M (1994) Cyclogenesis in a conditionally unstable
Japan 48: 140-160. moist baroclinic atmosphere. Tellus 46A: 14-33.
Hayashi Y and Golder DG (1997)United mechanisms for the Neelin JD and Yu J-Y (1993)Modes of tropical variability
generation of low- and high-frequency tropical waves. under convective adjustment and the Madden-Julian
Part I: Control experiments with moist convective oscillation. Part I: Analytical theory. Journal of the
adjustment. Journal of the Atmospheric Sciences 54: Atmospheric Sciences 51: 1876-1894.
1262-1276. Raymond DJ (1983)Wave-CISK in mass-flux form. Journal
Hayashi Y-Y and Sumi A (1986)The 30-40 day oscillation of the Atmospheric Sciences 40: 2561-2572.
simulated in an ‘aqua-planet’ model. Journal of the Stark TE (1976)Wave-CISK and cumulus parameterization.
Meteorological Society of Japan 64: 451-467. Journal of the Atmospheric Sciences 33: 2383-2391.
Lau K-M and Peng L (1987) Origin of low-frequency Yamasaki M (1969) Large-scale disturbances in the condi-
(intraseasonal) oscillations in the tropical atmosphere. tionally unstable atmosphere in low latitudes. Papers in
Part I: Basic theory. Journal of the Atmospheric Sciences Meteorology and Geophysics 20: 289-336.
44: 950-972. Wang B and Rui H (1990) Dynamics of the coupled moist
Lindzen RS (1974)Wave-CISK in the tropics. Journal of the Kelvin-Rossby wave on an equatorial beta plane.Journa1
Atmospheric Sciences 31: 156-179. of the Atmospheric Sciences 47: 397-413.
M C Kelley, Cornell University, Ithaca, NY, USA step, and by the time one gets to the transition from gas
Copyright 2003 Elsevier Science Ltd. All Rights Reserved, to plasma, quite a lot of energy is required. In this final
step the phase change required actually rips an
electron away from the gaseous atom or molecule,
Introduction leaving a positive ion behind. Since electric charge is
The early Greeks thought that all material was created conserved, the new state of matter remains neutrally
from four elements: air, earth, fire, and water. We now charged on average (equal numbers of positive ions
know that the elements are actually combinations of and negative electrons), but these constituents may
protons, neutrons, and electrons and that all matter is seldom run into each other and hence have little
made from various combinations of these atomic chance t o recombine to the atomic (gaseous) state. If
building blocks. However, when cataloguing the this is so, a plasma is born.
possible states of matter the ancients were not so far H o w much energy is needed t o rip apart atoms? The
off. In our daily lives, the states of matter referred to as response is ‘a few electron volts’. This is surprising at
solid, liquid, and gas are clearly related to Aristotle’s first, since we are all familiar with the batteries in our
earth, water, and air. radios and automobiles, which operate at voltages of
But what about fire? What about a fourth state of 1.5-12 V. We must remember, however, that batteries
matter? Indeed, there is a fourth state of matter, the run on chemical reactions that themselves involve
state called ‘plasma’, which could equally well have exchanges of electrons between atoms and molecules,
been called ‘fire’, since the hotter a flame, the closer it so the volt is a natural-sized unit for ionic bonds. How
comes to the plasma state. We dwell for a moment here can we relate this unit t o temperature? Suppose we
on the plasma state itself, since the Earth is surrounded have a pure gas like hydrogen; how hot does it need to
by just such a medium - a region called the ionosphere be to become a plasma? Suppose the gas is already hot
- which is the topic of this article. In fact, we see enough that the H2 molecules have separated into pure
throughout this volume that the atmosphere of the hydrogen as the bonds are broken due t o collisions of
Earth itself includes all four states of matter when one molecules with each other. The proton-electron pair
includes raindrops, snow, ice, and the ionospheric that makes up a hydrogen atom has a binding energy
plasma, in addition t o the gaseous component. requiring 13.5 electron volts (eV) to separate the
The list of earth, water, and air (solid, liquid, and particles. Thus to have significant numbers of atom
gas) can be reordered according t o the common collisions result in ionization as in eqn [I], the average
knowledge that when a solid is heated it becomes a energy of the colliding H atoms must be the order of
liquid and then a gas in processes called ‘change of 13.5 eV.
phase’. At each phase change, bonds are broken t o
form the next phase. More energy is required at each H+H-+H+H++e- [I1
1022 IONOSPHERE
zonal wind. Journal of the Meteorological Society of Mak M (1994) Cyclogenesis in a conditionally unstable
Japan 48: 140-160. moist baroclinic atmosphere. Tellus 46A: 14-33.
Hayashi Y and Golder DG (1997)United mechanisms for the Neelin JD and Yu J-Y (1993)Modes of tropical variability
generation of low- and high-frequency tropical waves. under convective adjustment and the Madden-Julian
Part I: Control experiments with moist convective oscillation. Part I: Analytical theory. Journal of the
adjustment. Journal of the Atmospheric Sciences 54: Atmospheric Sciences 51: 1876-1894.
1262-1276. Raymond DJ (1983)Wave-CISK in mass-flux form. Journal
Hayashi Y-Y and Sumi A (1986)The 30-40 day oscillation of the Atmospheric Sciences 40: 2561-2572.
simulated in an ‘aqua-planet’ model. Journal of the Stark TE (1976)Wave-CISK and cumulus parameterization.
Meteorological Society of Japan 64: 451-467. Journal of the Atmospheric Sciences 33: 2383-2391.
Lau K-M and Peng L (1987) Origin of low-frequency Yamasaki M (1969) Large-scale disturbances in the condi-
(intraseasonal) oscillations in the tropical atmosphere. tionally unstable atmosphere in low latitudes. Papers in
Part I: Basic theory. Journal of the Atmospheric Sciences Meteorology and Geophysics 20: 289-336.
44: 950-972. Wang B and Rui H (1990) Dynamics of the coupled moist
Lindzen RS (1974)Wave-CISK in the tropics. Journal of the Kelvin-Rossby wave on an equatorial beta plane.Journa1
Atmospheric Sciences 31: 156-179. of the Atmospheric Sciences 47: 397-413.
M C Kelley, Cornell University, Ithaca, NY, USA step, and by the time one gets to the transition from gas
Copyright 2003 Elsevier Science Ltd. All Rights Reserved, to plasma, quite a lot of energy is required. In this final
step the phase change required actually rips an
electron away from the gaseous atom or molecule,
Introduction leaving a positive ion behind. Since electric charge is
The early Greeks thought that all material was created conserved, the new state of matter remains neutrally
from four elements: air, earth, fire, and water. We now charged on average (equal numbers of positive ions
know that the elements are actually combinations of and negative electrons), but these constituents may
protons, neutrons, and electrons and that all matter is seldom run into each other and hence have little
made from various combinations of these atomic chance t o recombine to the atomic (gaseous) state. If
building blocks. However, when cataloguing the this is so, a plasma is born.
possible states of matter the ancients were not so far H o w much energy is needed t o rip apart atoms? The
off. In our daily lives, the states of matter referred to as response is ‘a few electron volts’. This is surprising at
solid, liquid, and gas are clearly related to Aristotle’s first, since we are all familiar with the batteries in our
earth, water, and air. radios and automobiles, which operate at voltages of
But what about fire? What about a fourth state of 1.5-12 V. We must remember, however, that batteries
matter? Indeed, there is a fourth state of matter, the run on chemical reactions that themselves involve
state called ‘plasma’, which could equally well have exchanges of electrons between atoms and molecules,
been called ‘fire’, since the hotter a flame, the closer it so the volt is a natural-sized unit for ionic bonds. How
comes to the plasma state. We dwell for a moment here can we relate this unit t o temperature? Suppose we
on the plasma state itself, since the Earth is surrounded have a pure gas like hydrogen; how hot does it need to
by just such a medium - a region called the ionosphere be to become a plasma? Suppose the gas is already hot
- which is the topic of this article. In fact, we see enough that the H2 molecules have separated into pure
throughout this volume that the atmosphere of the hydrogen as the bonds are broken due t o collisions of
Earth itself includes all four states of matter when one molecules with each other. The proton-electron pair
includes raindrops, snow, ice, and the ionospheric that makes up a hydrogen atom has a binding energy
plasma, in addition t o the gaseous component. requiring 13.5 electron volts (eV) to separate the
The list of earth, water, and air (solid, liquid, and particles. Thus to have significant numbers of atom
gas) can be reordered according t o the common collisions result in ionization as in eqn [I], the average
knowledge that when a solid is heated it becomes a energy of the colliding H atoms must be the order of
liquid and then a gas in processes called ‘change of 13.5 eV.
phase’. At each phase change, bonds are broken t o
form the next phase. More energy is required at each H+H-+H+H++e- [I1
IONOSPHERE 1023
Just what does a volt signify? The units of a volt are well as into ionizing the air, resulting in the temper-
joules per coulomb (J C - I ) . This means that if a 1-volt ature also rising to values much higher than in any part
battery is capable of storing one coulomb of charge, a of the dense atmosphere below. Life on Earth is thus
total of one joule of energy is available (enoughto lift a protected by its upper atmosphere from these danger-
mass of one kilogram to a height of 100cm on the ous photons, just as the ozone layer absorbs the lower-
Earth's surface). A typical car battery can supply a energy, but still harmful, ultraviolet component of the
current of 100 amperes (100 coulombs per second) for Sun's spectrum.
an hour, although not continuously, so it stores about We compare and contrast the atmosphere and
360 000 C. Thus the 1 2 V car battery stores about ionosphere in Figure 1. The most important atmos-
4.3 million joules (12J C - l x 360 000 C). Since an pheric parameter is temperature, which is plotted
electron has a change of only 1.6 x 10 - l 9 C, an energy versus height in (A).The key ionospheric parameter is
of 13.5 eV corresponds to about 2.2 x 10-l'J. Now the number of electrons (which equals the number of
we need to relate energy to temperature. Kinetic positive ions) per cubic centimeter. This is plotted in
theory shows that the average energy of a particle in (B) for typical nighttime and daytime conditions.
a gas is equal to $kBT where kg is Boltzmann's con- As anticipated above, the atmospheric temperature
stant (1.38 x 1 0 - 2 3 J K - 1 ) and T is the absolute rises from its lowest value near the mesopause (near
temperature in Kelvin. If we set this expression 200 K) to well over 1000 K in the thermosphere in the
equal to 2.2 x 10-l'J and solve for T, we obtain same height range where the daytime ionosphere is
T = 106 280 K! Such a high temperature shows why it produced. A glance at Figure 1B shows that the
is difficult to produce and control plasmas in the ionosphere does not entirely disappear at night, even
laboratory or in fusion machines. though the sunlight is no longer present to create new
The Sun is powered by nuclear fusion at its core and ionization. This is one of the key characteristics of the
is therefore very hot; hence it follows that much of its Earth's ionosphere and explains, for example, how
matter is ionized. Gravity controls this fiercely hot Marconi was first able to send wireless signals across
object. The Earth is much cooler, and hence it is not the Atlantic Ocean at night. To understand why some
obvious that there would be a plasma state in its of the ionosphere remains through the night, we must
environs. However, there is a plasma surrounding the consider the ion chemistry of the region.
Earth called the ionosphere. The fundamental pro- At high altitudes (>300km), production ( P ) and
duction and loss mechanisms for the Earth's iono- loss ( L ) of ionospheric plasma are both small. The
sphere are described next. This is followed by a balance between diffusion and gravitation results in
description of more exotic sources of the plasma the so-called hydrostatic equilibrium in which the
surrounding the Earth, including the solar wind, plasma pressure ( p ) is of the form of eqn [l].
magnetic storms, meteors, and the auroras. These
sources are localized in time and space and can be
linked under the umbrella of weather processes in
space or, in short, space weather. Space weather is also In eqn [I],where e is the base of the natural logarithms;
influenced by sources of energy and momentum from h is height above some reference; po is the pressure at
the Earth, the dense atmosphere below, and sources the reference altitude; and Hp is the plasma scale
including waves from severe storms, orographic fea- height, given by eqn [2].
tures, and earthquakes, as well as the release of stored
energy via plasma instabilities.
Sources and Fundamental Features In eqn [2], M is the average ion mass and g is the
gravitational acceleration. According to eqn [ 11, the
of the Ionosphere pressure falls by a factor of about 2.7 for each altitude
The ionosphere is formed primarily when the most increase of Hp.M is quite close to the average mass of
energetic component of the solar spectrum - the X- the neutral atmospheric particles surrounding the
rays and extreme ultraviolet (EUV) light - impact the plasma. The factor of 2 comes from the fact that the
illuminated side of the Earth. These high-energy average plasma mass is half the ion mass, since the
photons strike the daytime side of the Earth, ionizing electron mass is so tiny. The neutral atmosphere
the upper atmosphere and losing energy in the process. behaves like eqn [l]except that the neutral scale height
As the beam penetrates the atmosphere, the ionizing H , is half as large. One conclusion from the above is
beam becomes weaker and weaker, leaving behind a that, because the electrons are so light, the ionosphere
layer of ionization. Part of the energy goes into heat as extends higher into space than the neutral atmosphere
1024 IONOSPHERE
Neutral gas Ionized gas
1000 1000
-
Y
E
100 100
v
0
-0
3
tu
.-
I
3 10 10
- Day
- - - Night
1 I I I 1
0 400 800 1200 1600 io3 i o 4 i o 5 i o 6
(A) Temperature (K) (6) Plasma density ( ~ m - ~ )
Figure 1 Typical profiles of neutral atmospheric temperature (A) and ionospheric plasma density (6)
with the various layers designated.
(Reprinted with permission from Kelley MC (1989). Copyright 1989 by Academic Press.)
surrounding it. For reference, H,, is about 50 km and glowing corona seen during an eclipse. By chance,
Hp is about 100 km in the middle ionosphere. hydrogen and oxygen have almost identical ionization
At these altitudes the composition of the atmos- potentials, so charge exchange is a very easy process,
phere is no longer similar to the surface composition as shown in eqn [II].
(which is 79% N2, 20% 0 2 + minor constituents).
The atmosphere is no longer mixed, and lighter atoms H + O+=O + H+ [I11
can reach higher altitudes. Also, 0 2 is photodissoci-
ated into free oxygen atoms. Figure 2 shows the Thus, if O+ is surrounded by H gas, after a while an
composition in terms of various atoms, molecules, and oxygen ion will give up its charge to form a hydrogen
ions versus height for the mid-latitude ionosphere/ ion (H+).This explains why O t ions formed at low
thermosphere. We see that oxygen becomes dominant altitudes during daytime become Ht ions at very high
at 200 km and hydrogen above 700 km. Similarly, the altitudes.
ionosphere is primarily made up of Ht (with some Gravity and pressure are not the only forces with
H e + ) at very high altitude, Ot in the height range near which the ionosphere must deal. The Earth’s dipole
the peak density, and a mixture of O:, N:, and NOt magnetic field lines force the hydrogen ions to travel
in the lower thermosphere. Hydrogen is so light that it along closed trajectories between the hemispheres,
can escape the Earth’s gravity and form the Earth’s since following the magnetic lines is easy but moving
geocorona, a halo of hydrogen analogous to the Sun’s across them is not. The particle motion is helical, the
1000
150
100
102 103 io4 105 io6 io7 io8 io9 io1O ioi1
Number density ( ~ m - ~ )
Figure 2 International Quiet Solar Year (IQSY) daytime atmospheric composition. (Reprinted with permission of the MIT Press from
Johnson CY (1969). Ion and neutral composition of the ionosphere. Ann IQSYS. Cambridge, MA: MIT Press. Copyright 1969 by MIT.)
IONOSPHERE 1025
An alternative route is ion-atom interchange followed

by [VIb].
0' + N2 + NO' +N WaI
Both [Val and [VIb] leave oxygen in an excited state,

which emits both red (630 nm) and green (557.9 nm)
Figure 3 A toroidal region of high plasma density exists around light that is visible from the ground to sensitive
the Earth on averagewithin the region shown. These magnetic flux
tubes are filled with plasma of ionospheric origin during the day and
cameras. Such emissions are called airglow and
discharge only slowly at night. provide a tool for visualizing the ionosphere, as shown
in the next section.
To summarize thus far: the ionosphere is created
particles moving in circles around the magnetic field during the daytime by X-rays and EUV from the Sun,
lines while freely moving parallel or antiparallel to the which are absorbed while heating and ionizing the
direction of the field lines. The result of this motion is outer layer of the atmosphere. This heats the gas to
that the entire region, in a toroidal shape (shown in temperatures over 1000 K, explaining why it is called
Figure 3), becomes filled with a hydrogen plasma the thermosphere. The plasma, which is primarily O+
during the daytime (whose source is sunlight ioniza- above 200 km, diffuses upward against gravity, reach-
tion of oxygen coupled with charge exchange).During ing high enough that charge exchange with the
the night, this region - called the plasmasphere - starts geocorona converts the ionosphere to a Ht plasma,
to unload downward into the ionosphere by the which can escape gravity. The plasma is constrained by
reverse process, tending to maintain the oxygen the dipole magnetic field to a toroidal configuration,
plasma in the ionosphere during the night with the filled during the day and emptied at night. Molecular
whole process starting over the next day. The reason ions dominate in the lower thermosphere, but they
the plasmasphere abruptly ends at about 4 earth radii disappear quickly after sunset, leaving a slowly
(60" magnetic latitude) is very interesting and is decaying O f layer.
discussed below.
Refilling from above is not the entire reason that the A Day in the Life of the Midlatitude
ionosphere lasts all night, however. It turns out that a
charged atom cannot easily recombine with an elec-
Ionosphere
tron, since in a reaction such as [111] it is very difficult The most powerful single tool for ionospheric studies
simultaneously to conserve both energy and momen- is an incoherent scatter radar that detects the micro-
tum, and the reaction rate is very small. scopic fluctuation due to thermal motions in the
ionosphere. The first such instrument, and still the
Of + e - +0 [I111 largest of the 11 now in use worldwide, is near
Arecibo, Puerto Rico. The dish is 1km in circumfer-
However, in reactions [IVa] and [IVb] there are two
ence and can 'see' the ionosphere out to several
end products, and this difficulty does not arise.
thousand kilometers altitude. Figure 4 shows the
0; + e - + 0 + 0 [IVaI plasma content measured by the radar over a full day.
The earlier discussion explains the basic character of
NO' + e +N +0 [IVb] this plot, but not at all its details. We see that the
Reactions [IVa,b] are called dissociative recombi- density is high during the day, even at 100 km. At night
nation and are very fast. This explains why the the lower ionosphere rapidly - and almost completely
molecular ions (seen in Figure 2) at low altitudes - disappears, so much so that to see anything at all we
disappear at night, leaving the Of plasma above as the have to change the altitude scale and gray scale. Then
distinct nighttime layer (shown in Figure 1).In fact, slowly during the night, the density decays to low
reaction [111] is so slow that O+is actually lost through values just before sunrise, when the cycle begins again.
a two-step process such as charge exchange [Val But what about the wiggles? Why does the layer move
followed by [Vb]. up and down? And what is the origin of the weak
ionization layers seen at low altitudes?
o+ + 0 2 + 0; + 0 [Val These effects in large part are due to horizontal
winds and waves in the thermosphere. With a very hot
atmosphere on the daytime side (>1000K) in full
1026 IONOSPHERE
Figure 4 Ionospheric plasma density over Arecibo during a 24-h period (16-17 September 1999; time is Atlantic Standard Time). The
lack of plasma below250 km at night is due to recombinationof molecular ions. The high-altitude plasma and interestingthin layers are due
to ions such as O+,Mg+, and Fe+, which have long lifetimes.
sunlight and a very cool one at night (<800 K), it is not The more abrupt changes in height may be due to the
surprising that strong winds blow from day to night electrical forces that act on the ionosphere. These
continuously all over the globe. Unlike the thick lower electric fields have associated voltages as high as
atmosphere, the thermosphere has no thermal inertia. 200000V and ionospheric currents as great as
The winds simply blow continuously, acting as a huge 1000 000 A, yielding power levels of 2 x 10" W:
atmospheric thermal tide. Speeds of 200 m s - more power than any man-made generator on Earth.
(720 km h-') are not uncommon. It is very hard to Two major generators provide this electrification and
move the plasma across the magnetic field lines, but both involve motion of a conductor across a magnetic
such winds easily move the ionosphere up and down field, in exactly the manner by which generators
the magnetic field lines. At Arecibo, the direction of the convert mechanically rotation of machines into elec-
magnetic field is at an angle of 4.5" to the vertical tric energy. The solar wind is the most powerful of the
(pointing northward and downward). When the two, generating hundreds of kilovolts across the
wind blows southward, the ionosphere moves up- Earth's polar regions and causing one of nature's
ward, much like a ping-pong ball held up against most spectacular visual displays, the aurora borealis
gravity on an inclined plane by a hair dryer, as shown and aurora australis (see the next section). The second
in Figure 5. generator is the motion of the Earth's atmosphere,
At high altitude there is very little neutral gas and described earlier. Tides, winds, and gravity waves in
recombination is weak. If the wind ceases or blows the atmosphere all drive currents and generate electric
north, the ionosphere falls owing to gravity into fields by the dynamo effect. Because the magnetic field
regions where reactions [Va,b and VIa,b] can eat away lines behave like conducting wires, the only voltage
at it. Thus a southward wind not only elevates the easily allowed is across magnetic field lines. Figure 6
ionosphere, but it also keeps it high out of reach of the illustrates what happens if a single positive particle is
losses due to the thermosphere. Some of what we see at subjected to orthogonal electric and magnetic fields.
night in Figure 4 is due to these winds. Initially, the particle is accelerated parallel to the
Figure 5 An illustration showing the analogy between the mid-latitude ionosphere on the right, with atmospheric winds pushing the
ionosphere up the magnetic field, and a light object suspended on an inclined plane.
IONOSPHERE 1027
- 4 Ion path
Rolling wheel V= -
IEl
IBI
Electron path
Figure 6 In crossed electric and magnetic fields in vacuum, ions and electrons exhibit the motion shown schematically. The ion path is
specifically shown to be similar to that of a dot on the rim of a rolling wheel.
electric field by the force qE. But once it attains a back-lit CCD (charge-coupled device) chip illuminat-
velocity, the magnetic force, qV x B, deflects it to the ed by a fisheye (all-sky) lens. A narrow (630k 1nm)
right. Eventually it comes to rest and the cycle starts filter was inserted into the path and the chip was
over. The trajectory is the same as that of a dot on the exposed for 90 s. There were 1024 x 1024 pixels in the
edge of a rolling, nonslipping wheel moving to the image, which, at a height of 250 km, covers a circle of
right at a velocity E/B. A negatively charged particle 1000 km diameter. The image has been corrected for
(electron) starts out in the opposite direction, but is the lens effects, vignetting, etc., and projected as if the
also deflected to the right, and drifts on average at the viewer were above the Earth looking down rather than
same speed as the positive ions. Since there are equal up (hence the map of the Caribbean in its usual
numbers of positive ions and electrons there is no net geometry). We see intricate patterns of light and dark
current, just a net velocity. Above about 150km, regions, with one of the dark zones positioned right
collisions are so rare that Figure 6 describes the over the Arecibo Observatory. The ionosphere is
motion quite well - electric fields are one-to-one highly structured on this night and is very different
related to the motion of the ionosphere across the from what would be predicted if only production, loss,
magnetic field lines, whereas winds, gravity, and gravity, and diffusion were operating. The sequence of
diffusion dominate along the direction of the magnetic images taken on this night show that the dark bands
field. An eastward electric field over Arecibo, for surged poleward from well south of Puerto Rico and
example, causes an (E x B)BP2 drift northward and then drifted toward the west. This unexpected be-
upward at a 45" angle. Some of the abrupt height havior shows we have much to learn about even the
changes visible in Figure 4 are due to such electric field- most well-behaved regions of the ionosphere.
induced motions.
The abrupt height changes might be temporal or
spatial or a combination of both; it is difficult to tell Fire and Ice: The High-Latitude
with a single measurement. But the fact that red light is
emitted in reactions [Val and [VIb] allows us to
Ionosphere
visualize the plasma in two dimensions. The data in Other entries in the Encyclopedia discuss the auroras
Figure 7 were obtained on the same night using a bare, and the Earth's magnetosphere at some length. Here
Figure 7 The 630 nm airglow for 17-1 8 February 1999, superimposed on a map of the Caribbean islands with Puerto Rico in the center.
The regions depleted of airglow (shown black) commenced in the SE and surged to the NW on this night. (Reproducedby permissionof the
American Geophysical Union from Kelley et a/. (2000) Caribbean Ionosphere Campaign, Year One: airglow and plasma observations
during two intense midlatitude spread-f events. Geophysical Research Letters).
1028 IONOSPHERE
we discuss some of the more striking features of the

ionosphere in this region. First of all, the light
emissions we call the auroras also are due to the
impact of energetic particles, primarily electrons, on
the atmosphere. Some of these leak in from the solar
wind, but most are accelerated from the background
plasma at a height near 5000 km above the Earth. The
Earth has its own ‘cosmic ray’ generator, just as do
pulsars, Jupiter, and other magnetized objects in the
cosmos. Figure 8 shows a view of the Earth from
800 km during a spectacular aurora. Close inspection
of the city lights shows that the aurora reached
Madison, WI, and covered most of the Great Lakes
region.
Since intense light must be accompanied by the
production of plasma, the aurora is a highly dynamic
source of the ionosphere wherever and whenever it
appears. Figure 9 gives some insight into its global
character. The view is from 5000 km and shows rings
of light circling the top (and bottom) of the Earth’s
Figure 9 The sequence begins at 0529 UT on 2 April 1982

(upper left image) as the NASNGSFC spacecraft Dynamics
Explorer 1 first views the auroral oval from the late evening side
of the dark hemisphere at low northern latitudes near apogee (3.65
Earth radii altitude) and then from progressively greater latitudes as
the spacecraft proceeds inbound over the auroral oval toward
perigee. The poleward bulge at onset of the auroral substorm is
observed beginning at 06:05 UT (fourth frame). In successive
12-minute images the substorm is observed to expand rapidly in
latitude and longitude. (Photograph courtesy of LA Frank, JD
Craven, and RL Rairden, University of Iowa. Reprinted by
permission of Academic Press from Kelley (1989). Copyright
1989 by Academic Press.)
polar regions like a halo. This auroral oval waxes and

wanes in size, sometimes (as seen in Figure 8) reaching
highly populated regions, but usually limited to the
mid-arctic zone.
During certain conditions in the solar wind (when
the interplanetary magnetic field is parallel to the
Earth’s magnetic dipole), the solar wind electric field is
transmitted very efficiently along the Earth’s magnetic
field and throughout both polar ionospheres. Mag-
netic storms often occur at these times and great
auroras result. Since the Earth’s magnetic field is nearly
vertical, near the pole the (E x B ) F 2 drifts are nearly
horizontal. Thus, two huge circulation cells often
occur in the ionosphere. If the magnetic storm contin-
ues for more than an hour or so, the neutral atmos-
phere can also be put into motion. It is remarkable that
Figure 8 A spectacular view of the eastern half of the United the coupled d a r wind and ionosphere actU-allY Put the
States during a major auroral display. Earth’s atmosphere in motion! Heat is also generated
IONOSPHERE 1029
Figure 10 A space weather radar map showing regions of highly turbulent ionospheric plasma over Peru. The dark regions are
analogous to clear air and thunderstorm-related turbulence in the troposphere. (Reproduced by permission of the American Geo-
physical Union from Kelley et a/. (1981) Gravity wave initiation of equatorial spread F: a case study. Journal of Geophysical Research
86: 9087.)
by the electric currents, which change the global based systems. Some of these aspects involve the
circulation. ionosphere and are described here. Others are related
The polar circulation rips away at the plasma- to magnetospheric phenomena, such as the killer
sphere, reducing its size and compressing it to lower electrons of the radiation belt that create havoc in
latitudes. After the storm and over a few days, the satellite systems, and solar proton events that create
region refills with cool dense plasma out to about 4 severe radiation levels for astronauts building the
earth radii and 60" magnetic latitude. This altitude- International Space Station.
latitude region is where the solar wind and earthly Rapidly changing magnetic fields due to the vast
wind dynamos have approximately equal control of ionospheric currents flowing in a major auroral event
the ionosphere-plasmasphere-magnetosphere sys- create electrical voltages at the surface of the Earth,
tem. During a great magnetic storm the composition just as they do in an electrical transformer. Vast power
of the thermosphere can be significantly modified, grids are the perfect detectors of these voltages, which
even worldwide. This creates great negative iono- are unexpected and can thus trigger the unnecessary
spheric storms wherein the ionosphere virtually dis- shutdown of elements along the grid. During the great
appears for a day, even in full sunlight. The storm of the last solar cycle, the Province of Quebec
atmosphere changes so much that the Earth acts like went dark for 12 hours because of such a surge. Today
some other planet, one with very little oxygen. The our power grids are even more interconnected, and
solar cycle maximum of 2001 has yielded many predictions of such conditions are becoming both of
exciting scientific discoveries as we continue to more practical importance and of some practical
instrument the Earth and its surrounding near-space feasibility. Such predictions are among the first chal-
regions. We also expect that space weather will lenges of the fledgling National Space Weather Pro-
become more relevant to mankind, as discussed next. gram in the United States as well as the global
counterpart of this new program.
Some important space weather effects are strictly
due to the Earth's dynamic atmosphere without help
Space Weather
from the solar wind. The mid-latitude weather dis-
There are many aspects of space weather that are cussed earlier is of this type and is fairly rare. But near
relevant to human habitation, particularly as we the magnetic equator (where the magnetic field lines
become more dependent on technology and space- are exactly horizontal), severe convective storms occur
1030 ISENTROPICANALYSIS
night after night in some seasons and longitudes. The onset of severe space weather, much as the meteorol-
ionosphere is so dense at the Equator that such storms ogy community has come to predict severe storms in
create havoc with communication systems using radio the lower atmosphere. However, the scale of the
waves, which must propagate through the ionosphere interacting system stretches from the Sun to the Earth
from satellite to the ground. In brief, in such storms the and is much more variable than the solar constant
satellite signals ‘twinkle’, just as starlight does passing that directly heats the Earth’s lower atmosphere.
through the turbulent lower atmosphere. This creates In addition, this vast region of space has few observ-
deep fades and distortions in satellite signals, which ing stations, and the challenges of prediction are
disrupt communications. The higher the radiowave enormous.
frequency, the less does ionospheric turbulence create
problems. However, even at the high frequencies of the
Global Positioning System satellites (>1GHz), iono-
spheric effects can occur. See also
Figure 10 is a space weather radar map obtained Dynamic Meteorology: Waves. Electricity, Atmos-
over the magnetic equator near Lima, Peru. The dark pheric: Ions in the Atmosphere. Global Change: Upper
areas reveal where ionospheric turbulence exists. O n Atmospheric Change. Magnetosphere. Mesosphere:
this night severe weather erupted just after sunset and Polar Summer Mesopause. Radiation (Solar). Satellite
lasted for several hours, its effects extending to over Remote Sensing: GPS Meteorology. Solar Winds.
1000 km altitude. Radio signals propagating through Thermosphere. Turbulent Diffusion.
this region would be seriously degraded.
Further Reading
Summary Chen FF (1984) Introduction to Plasma Physics and Con-
We have learned much about the ionosphere since its trolled Fusion. 2nd edn. New York: Plenum Press.
discovery near the beginning of the twentieth century, Davies K (1990) Ionospheric Radio. Exeter, UK: Peter
Peregrinus.
when Marconi first skipped radiowaves off it all the
Hargreaves JK (1992) The Solar-Terrestrial Environment.
way across the Atlantic. The space age rocket and Cambridge: Cambridge University Press.
satellite probes, along with powerful ground-based Kelley MC (1989) The Earth’s Ionosphere. San Diego:
radars, have revealed much about its properties. New Academic Press.
things are still being discovered, but we are also Rishbeth H and Gariott OK (1969) Introduction to lono-
entering an age of prediction. The US National Space spheric Physics, International Geophysical Series, vol.
Weather Program has as a goal the ability to predict the 14. New York: Academic Press.
J R Holton, University of Washington, Seattle, WA, USA adiabatic. Thus, changes in the internal energy of
Copyright 2003 Elsevier Science Ltd. All Rights Resewed air parcels owing to exchange of heat with the
surroundings are small compared with changes asso-
ciated with adiabatic expansion or compression
Introduction caused by pressure changes. Air parcels moving
Atmospheric motions are generally three-dimension- adiabatically are confined to isentropic surfaces.
al; air parcels change altitude (and pressure) as they are Analysis of the motion relative to isentropic surfaces
advected by the winds. Since vertical velocities asso- then reduces the three-dimensional problem of trajec-
ciated with large-scale motions tend to be small, and tory analysis to a problem of two-dimensional parcel
are not directly observed, the computing of the advection on the isentropic surfaces. For this reason,
trajectories of air parcels utilizing conventional mete- isentropic analysis has become a popular technique for
orological analyses on isobaric surfaces is a difficult analysis of the transport of long-lived trace constitu-
task. Isentropic analysis seeks to simplify the ents. The method is especially useful in the strato-
calculation of air parcel trajectories by utilizing the sphere where the flow can often be considered to be
fact that, outside regions of active precipitation, approximately adiabatic for time scales of a week or
large-scale motions in the atmosphere are quasi- longer.
night after night in some seasons and longitudes. The onset of severe space weather, much as the meteorol-
ionosphere is so dense at the Equator that such storms ogy community has come to predict severe storms in
create havoc with communication systems using radio the lower atmosphere. However, the scale of the
waves, which must propagate through the ionosphere interacting system stretches from the Sun to the Earth
from satellite to the ground. In brief, in such storms the and is much more variable than the solar constant
satellite signals ‘twinkle’, just as starlight does passing that directly heats the Earth’s lower atmosphere.
through the turbulent lower atmosphere. This creates In addition, this vast region of space has few observ-
deep fades and distortions in satellite signals, which ing stations, and the challenges of prediction are
disrupt communications. The higher the radiowave enormous.
frequency, the less does ionospheric turbulence create
problems. However, even at the high frequencies of the
Global Positioning System satellites (>1GHz), iono-
spheric effects can occur. See also
Figure 10 is a space weather radar map obtained Dynamic Meteorology: Waves. Electricity, Atmos-
over the magnetic equator near Lima, Peru. The dark pheric: Ions in the Atmosphere. Global Change: Upper
areas reveal where ionospheric turbulence exists. O n Atmospheric Change. Magnetosphere. Mesosphere:
this night severe weather erupted just after sunset and Polar Summer Mesopause. Radiation (Solar). Satellite
lasted for several hours, its effects extending to over Remote Sensing: GPS Meteorology. Solar Winds.
1000 km altitude. Radio signals propagating through Thermosphere. Turbulent Diffusion.
this region would be seriously degraded.
Further Reading
Summary Chen FF (1984) Introduction to Plasma Physics and Con-
We have learned much about the ionosphere since its trolled Fusion. 2nd edn. New York: Plenum Press.
discovery near the beginning of the twentieth century, Davies K (1990) Ionospheric Radio. Exeter, UK: Peter
Peregrinus.
when Marconi first skipped radiowaves off it all the
Hargreaves JK (1992) The Solar-Terrestrial Environment.
way across the Atlantic. The space age rocket and Cambridge: Cambridge University Press.
satellite probes, along with powerful ground-based Kelley MC (1989) The Earth’s Ionosphere. San Diego:
radars, have revealed much about its properties. New Academic Press.
things are still being discovered, but we are also Rishbeth H and Gariott OK (1969) Introduction to lono-
entering an age of prediction. The US National Space spheric Physics, International Geophysical Series, vol.
Weather Program has as a goal the ability to predict the 14. New York: Academic Press.
J R Holton, University of Washington, Seattle, WA, USA adiabatic. Thus, changes in the internal energy of
Copyright 2003 Elsevier Science Ltd. All Rights Resewed air parcels owing to exchange of heat with the
surroundings are small compared with changes asso-
ciated with adiabatic expansion or compression
Introduction caused by pressure changes. Air parcels moving
Atmospheric motions are generally three-dimension- adiabatically are confined to isentropic surfaces.
al; air parcels change altitude (and pressure) as they are Analysis of the motion relative to isentropic surfaces
advected by the winds. Since vertical velocities asso- then reduces the three-dimensional problem of trajec-
ciated with large-scale motions tend to be small, and tory analysis to a problem of two-dimensional parcel
are not directly observed, the computing of the advection on the isentropic surfaces. For this reason,
trajectories of air parcels utilizing conventional mete- isentropic analysis has become a popular technique for
orological analyses on isobaric surfaces is a difficult analysis of the transport of long-lived trace constitu-
task. Isentropic analysis seeks to simplify the ents. The method is especially useful in the strato-
calculation of air parcel trajectories by utilizing the sphere where the flow can often be considered to be
fact that, outside regions of active precipitation, approximately adiabatic for time scales of a week or
large-scale motions in the atmosphere are quasi- longer.
ISENTROPIC ANALYSIS 1031
Isentropic Coordinates of relative vorticity evaluated on an isentropic

The first law of thermodynamics for adiabatic condi- surface, and f is the Coriolis parameter. For adiabatic
tions can be expressed as motions P is conserved following the motion on 6
surfaces:
D
-(c,
Dt
In T - R In p) = 0 PI [41
where R , c,, p, and T stand for the gas constant for dry
air, the specific heat at constant pressure, the atmos-
pheric pressure, and the temperature, respectively.
If this expression is integrated from a state at pre- Applications
ssure p and temperature T to a state with pressure In practice, isentropic analysis requires interpolation
ps E 1000 hPa, the temperature becomes of meteorological fields from isobaric surfaces, on
which upper-air data is normally reported, to surfaces
of constant 0. A latitude-height section showing the
relationship between isentropic surfaces and isobaric
The quantity 8 in this expression is referred to as the surfaces is given in Figure 1. In the troposphere
potential temperature. It represents the temperature the isotherms slope downward toward the poles,
which an air parcel located anywhere in the atmos- because temperature decreases with height and with
phere would acquire through adiabatic compression a latitude. The isentropic surfaces, on the other
pressure of 1000 hPa. Thus, as long as the motion is hand, slope upward toward the poles in the tropo-
adiabatic, potential temperature is conserved follow- sphere. This slope is substantial, especially in the
ing the motion, and surfaces of constant potential winter hemisphere. Note also that for potential
temperature represent isentropic surfaces. temperatures between about 300 and 350 K, the
Normally, potential temperature increases mono- isentropic surfaces cross the tropopause, so that air
tonically with height and thus can be used as an can be exchanged between the low-latitude tropo-
independent vertical coordinate just as pressure is used sphere and the extratropical lowermost stratosphere
as a vertical coordinate in standard synoptic analysis. through adiabatic motions. In the stratosphere there is
When the equations of motion are transformed into a downward slope of the isentropes from the equato-
the isentropic coordinate system (with 8 as the vertical rial region to the polar region in the summer hemi-
coordinate) the horizontal pressure gradient force is sphere and to the midlatitudes in the winter
represented by the horizontal gradient of a quantity hemisphere. The flattest 8-surface tends to occur
called the Montgomery stream function, defined as near the 350 K level.
Y c,T + @. Thus, Isentropic coordinates have occasionally been used
1 in weather forecast models and global climate models.
--vp = -DY However, partly because of the complexity of the
P
where the gradient of Y is evaluated by taking the
partial derivatives with 8 held constant.
In isentropic coordinates an infinitesimal control
volume of horizontal cross-section 6A and vertical
extent 68 has a mass 6M = p 6A S z = a SA 68, where
a=-- g-'ap/a0 is the 'density' in ( x , y, 0) space (i.e., it
is the quantity that when multiplied by 6A60 gives
the mass 6M.For adiabatic motions a satisfies the
equation
aa + Dlj
- (aV) = 0
at 131
Thus, o is changed locally only if there is a convergence Figure 1 Longitudinal average temperature (dashed contours)
or divergence of the mass flux in an isentropic layer of and potential temperature (solid contours) for January 1993. The
heavy solid line denotes the tropopause. The shaded area denotes
infinitesimal depth.
the lowermost stratosphere, the region in which isentropes span
Motions on isentropic surfaces are governed by the tropopause. (Reproduced with permission from Holton JR,
conservation of potential vorticity, defined as Haynes PH, Mclntyre ME et a/. (1995) Stratosphere-troposphere
P G (is + f ) o - l , where [e is the vertical component -
exchanae. Reviews of GeoDhvsics
.~ 33: 403-439.)
Isentropic analysis is particularly valuable for

studies of trace constituent transport in the strato-
sphere. Radiative heating and cooling are the
only significant diabatic processes in the stratosphere.
The time scale for radiative temperature adjust-
ments is of order 3 weeks in the lower stratosphere
and 1 week in the upper stratosphere. Thus, for
time scales of a week or so motions in the lower
stratosphere can be approximated as adiabatic, and
long-lived trace constituent transport can be approx-
imated by the transport along the isentropes.
An example of isentropic analysis is shown in
Figure 2. The figure compares satellite obser-
vations of nitrous oxide with predictions based on
8-day trajectory calculations. In general the trajectory
model captures the changes in the hemispheric distri-
bution of nitrous oxide during each of the 8-day
periods. If accurate heating and cooling rates are
available then the motion of trace constituents
through isentropic surfaces can be included to improve
the results of such trajectory studies. Isentropic anal-
ysis is likely to remain a preferred methodology for
interpretation of atmospheric trace constituent obser-
vations.
See also
Dynamic Meteorology: Overview. Middle Atmos-
phere: Transport Circulation. Stratosphere-Tropo-
sphere Exchange: Local Processes. Thermodynamics:
Moist (Unsaturated) Air.
Figure 2 Synoptic maps of the nitrous oxide mixing ratio on the

465 K potential temperature surface. The left column shows Further Reading
observations from the Cryogenic Limb Array Etalon Spectrometer
(CLAES) on the Upper Atmosphere Research Satellite and the Bluestein HB ( 1993) Synoptic-Dynamic Meteorology in
right column shows predictions based on trajectories initialized on Midlatitudes. Vol. 11. Observations and Theory of
14 February 1993. Solid contours on observational plots show Weather Systems. New York: Oxford University Press.
potential vorticity contours. Note that the trajectory calculations Holton JR (1992) Introduction to Dynamic Meteorology.
maintain the observed low nitrous oxide mixing ratios in the interior
San Diego, CA: Academic Press.
of the polar vortex as the vortex rotates and becomes elongated.
(Courtesy of Dr Gloria Manney; adapted from Manney GL, Zurek
Hoskins BJ, McIntyre ME and Robertson AW (1985)On the
RW, Lahoz WA, et a/. (1995) Lagrangian transport-calculations use and significance of isentropic potential vorticity
using uars data. Pt 1. Passive tracers. Journal of the Atmospheric maps. Quarterly Journal of the Royal Meteorological
Sciences 52: 3049-3068.) Society 111: 877-946.
Salby ML (1996) Fundamentals of Atmospheric Physics.
New York: Academic Press.
boundary condition where @-surfaces intersect the Shaw N (1933)Manual of Meteorology. Vol. 3 . T h e Physical
ground, the use of isentropic coordinates in numerical Processes of Weather. Cambridge: Cambridge University
models has been rather limited. Press.
ISOTOPES, STABLE 1033
C A M Brenninkmeijer, Max Planck Institute for

Chemistry, Mainz, Germany Table 1 Minor isotope abundances
Copyright 2003 Elsevier Science Ltd. Ail Rights Reserved Element Minor isotopes (%) Main reference materials
H 'H or D: 0.15 V-SMOWa

Introduction C 1%: 1.11 V-PDBbsC,NBS-19'
(ai3C = 10.95)
Chemical reactions and phase transitions induce small N "N: 0.37 Atmospheric N2
but measurable changes in the stable-isotope abun- 0 I7O:0.037 V-SMOW, or
" 0 : 0.204 atmospheric O2
dances of the elements in atmospheric gases and
S 33S:0.76 CDT~
chemical compounds in aerosol. These natural varia- 34S:4.22
tions do not affect chemistry or transport but their 36S:0.014
investigation can deepen our understanding of certain CI 37CI:24.47 Not available
atmospheric processes. The most important underly- Br "Br: 49.46 Not available
~~
ing principles are that different sources of trace gases Vienna-Standard Mean Ocean Water.
have different isotopic signatures, and that the isotopic bOriginal material exhausted.
compositions of trace substances are changed by their 'Calcium carbonate.
removal processes. Application of isotopic analysis in "troilite phase of the Canyon Diablo meteorite.
atmospheric chemistry has revealed a rather wide-
spread class of new isotope effects, the causes of which
negligible. Note that atmospheric methane inciden-
are not yet understood. Analytical progress has been
tally has an excess of deuterium owing to releases from
rapid and nowadays very small quantities of subs-
heavy-water nuclear power plants. Its multideuteri-
tances can be analyzed rapidly, thus turning isotope
um-substituted isotopic composition is peculiar and
analysis into an increasingly useful tool in the atmos-
nonrandom and of little practical use.
pheric sciences.
For any two reference materials A and B,
the conversion for a sample X is made according to
Notation eqn PI.
With the exception of hydrogen, variations in stable-
isotope ratios occurring in nature usually amount to
only several percent. Therefore, the isotope abun-
dances for a particular element are expressed on a For mixing a molar fraction f having a delta value 6,
relative scale that is defined by one or more standard into a reservoir with 6 b , the resulting isotopic compo-
reference materials. For a given sample (SA), the sition will be given by eqn [ 3 ] .
isotopic value relative to the standard material (ST)is
given in parts per thousand (%o); using carbon-13 as an
6, = f d a f (1- f ) a b 131
example with R = 13C/12C, eqn [I]gives the relative For convenience, the isotopic abundance of the
isotopic value. elements most used in atmospheric studies are given
in Table 1.
Reference materials for isotopic analysis are avail-
able from the International Atomic Energy Agency
(IAEA) in Vienna, and the National Institute for
This definition is based on the ratio of the minor to the Standards and Technology (NIST) in Gaithersburg,
more abundant isotope. It excludes the use of molec- MD, USA.
ular ratios, so that, for instance, 613CH4 is not defined
in the literature; thus the notation 613C(CH4) is
Measurement
unambiguous. For the deuterium content of a sample
of methane the appropriate equation corresponding to Measurement is in most cases based on mass spect-
eqn [l]is used. The molecular ratio in the sample, rometry, by which a precision of up to 0. 0l%0can be
(CDH3)/(CH4),will be four times larger. Given the low achieved. This is ample for all atmospheric applica-
abundance of deuterium, the fraction of methane tions. The isotopic composition of a sample, at times in
molecules with two or more deuterium atoms is micromole quantities, is repeatedly compared to that
1034 ISOTOPES,STABLE
of a mass spectrometer standard under near-identical Table 2 Reaction rate constants at 298K in cm3
conditions. The analytical principle is electron impact ’ ’
molecules- s
ionization, magnetic separation and focusing, and CI OH
Faraday collectors. Conversion to the appropriate
reference material is basically done using eqn [2], after 100 6.4
74 5.2
ion current corrections have been made. For in-
46 3.4
stance for CO2, mass 45 represents a combination of 23 2.1
1 3 ~ 1 6 0 1 6 0 plus 12 17 16
C 0 0. A recent development 6 0.9
has revolutionized atmospheric applications. Instead
of transferring the sample material under high vacuum
into the mass spectrometer, a flow of helium is used as mixing in the air above the liquid, the depletion of the
carrier gas. This has allowed the coupling of a gas water vapor is determined by the difference in diffu-
chromatograph, which can separate atmospheric trace sion speeds of the isotopically substituted molecules
gases from the air matrix, with the mass spectrometer. (isotopomers), and can be calculated readily.
One advantage is that nanomole quantities can be Kinetic isotope effects are common in chemical
analyzed. The other is that the speed of analysis is reactions. The heavier isotopes have stronger bond
increased and automation is straightforward. This strengths, and heavier molecules react more slowly.
on-line analysis already accounts for most of the Here the calculations are more difficult and transition
atmospheric applications. It is expected that the state theories with several refinements, including
number of isotope-based atmospheric studies will tunneling effects, are often used. For the atmospher-
increase rapidly because the analytical effort has ically relevant reaction between chlorine radicals (Cl)
become comparable to the abundance measurement and CH4, calculations are based on the intermediate
of, for instance, hydrocarbons. complex C1-HCH3. The theory and experimental
In a few cases, optical techniques are used. Tunable results do not always conform. The systematics of
diode laser absorption spectroscopy (TDLAS), for isotope substitution for the reactions of CH4 is-
instance, allows comparison of absorption by CDH3 otopomers with O H and C1 are shown in Table 2.
and CH4. The advantage is that no sample preparation Note that the reaction rate constants are determined
is necessary: the methane simply has to be separated by the large difference between the reactivity of D
from air. When using mass spectrometry, the methane compared to H. For C1 the reaction rate constant is
has first to be chemically converted to CO2 and H2. A almost entirely determined by the degree of substitu-
TDLAS instrument has been used in aircraft measure- tion because the reaction rate for D is much smaller.
ment of water vapor isotopes. Occasionally Fourier For atmospheric applications, the overall reaction
transform Infrared (FTIR) spectroscopy is used for rates are used. Thus, whether, for instance, DCl or HCI
laboratory measurement of C 0 2 and N 2 0 is- is the product is not of importance.
otopomers. Optical techniques have the major advan- The fractionation factor k (or a ) is defined as the
tage that remote sensing is possible. Infrared isotopic ratio in one compound or phase relative to
spectroscopy is used for stratospheric water vapor that of the other. No strict convention is in operation.
and ozone isotope measurements. Often water vapor is assayed relative to the liquid, and
k < 1, (Le., depletion). The accompanying fraction-
ation constant E is defined as a - 1. For kinetic isotope
effects (KIE), the fractionation factor is the ratio
Kinetic and Equilibrium Isotope between the reaction rate constants for the minor and
Effects major isotopomers. In most cases a will be positive as
A distinction is made between kinetic and equilibrium molecules substituted with the heavier isotopes gener-
isotope effects, as for water vapor in contact with the ally react more slowly. A notable exception is CO, for
liquid in an enclosed volume. The depletion of the which C l 8 0 reacts more rapidly with O H than does
vapor in D and l80can be calculated on the basis of C l 6 0 , and accordingly a< 1. The various isotope
the vapor pressure differences. When C 0 2 is in oxygen fractionation processes of relevance to atmospheric
isotopic equilibrium with H20, its l80content at processes are shown in Table 3.
25°C is 41%0above that of the water. This effect can be
calculated like most other (thermodynamicj equilib-
rium effects, using statistical mechanics and the
Mass-independent Fractionation
partition function. Stable-isotope investigations in the atmosphere have
A kinetic isotope effect also occurs during net revealed one or more types of fractionation processes
evaporation of water. In the absence of turbulent that are not yet understood and have been pooled
Table 3 Fractionation processes pheric research has changed this; recently MIF has
been detected for 0 3 , C 0 2 , CO, N20, H202, SO$-,
Process Example
and even 0 2 . A measure for MIF is the excess in "0
Phase change Evaporation expressed as: A i 7 0 = 6"O - 0.52 6 l 8 0 .
Diffusion COa in leaf stomata
Dissolution COP in water
Gravitational settling Ice core studies Atmospheric Applications
Atmospheric escape H2
Chemical reaction CH4 + OH General
Photolysis Na0
Isotope effects do not affect atmospheric chemistry
under most circumstances. The reasons are that
isotope effects are nearly always small, in the percent
under the name 'mass-independent fractionation' range, and the abundances of the minor isotopes are
(MIF).The rationale for this name was that the effect, low. Even though the isotope effects for deuterium
first observed for ozone (03), was attributed to may be large because of the large relative mass
molecular symmetry effects, totally independent of difference, its low abundance (Table 1) excludes
mass. Retrospectively, the name is an unfortunate effects on chemistry. For chlorine, the less-abundant
choice, but it is often used along with the expression isotope is comparatively rather abundant, but here
'anomalous fractionation'. Because of its practical and fractionation effects of only a few percent do not make
fundamental importance, MIF is discussed here. much impact on chemistry. Thus, in general, isotope
Figure 1 shows the relationship between 6 1 7 0 and effects are useful as tracers but have no direct
6"O for a wide range of substances. The slope of the implications.
mass-dependent fractionation line thus defined is close Given a certain atmospheric trace gas, isotopic
to 0.52. The theoretical basis for this correlation using variations are due to two factors. One is the isotopic
statistical mechanical calculations is well understood. composition of its sources. In principle, the relative
For instance, the equilibration constant for an isotope source strengths can be calculated using eqn [3],
exchange reaction is proportional to the vibrational provided there are only two isotopically distinct and
frequencies, and therefore the reciprocal masses. For well-defined sources. The other factor, which disturbs
diatomic molecules for 0 isotopes ("0,"0 and l60) this simple picture, is isotope fractionation in the
one obtains (1/16 - 1/18)/(1/16 - 1/17) = 1.89, removal of the gas; however, this does give informa-
the inverse giving 0.53. In fact, because of this tion about the degree of removal. If one sink has a
omnipresent strict mass dependence, almost no atten- distinctly stronger fractionation, its impact can be
tion has been paid to the analysis of 170in oxygen- detected and, under certain conditions, quantified.
containing compounds for several decades. Atmos- For an isolated, well-mixed amount of gas being
depleted by a loss process (e.g., reaction with O H or
photolysis), the isotopic composition will evolve in
time according to a Rayleigh distillation process.
100
Using m/mo as the ratio of the actual to the initial
90
80
mixing ratio, eqns [4a] or [4b] apply.
70
60
3 50
0' 40
< 30
20
10
0
-10
-20 Ozone, 0 3
-40 -20 0 20 40 60 80 100 120
6'80 60) Ozone has been analyzed in situ optically and by
balloon-borne mass spectrometry, and via sample
Figure 1 The 6 l 8 0 and corresponding 6"O values for some
important reservoirs, showing mass-independent fractionation
collection. Ozone is formed via O(3P) + 0 2 + M --f
effects as deviations from the 'mass-dependentfractionation' line 0 3 + M and has a remarkable isotopic composition
of slope 0.52. Note that molecular oxygen has a small negative throughout the atmosphere. It is considerably
value of - 0.15%0,not visible on this scale. enriched in "0 relative to molecular oxygen by
1036 ISOTOPES, STABLE
net flux into the terrestrial biosphere, which favours

Table 4 Relative reaction rates for the formation of ozone from
l60(O),I7O(P),and '*O(Q) l2CO over 13C0 by about 25%0,is compensated by the
Reaction Relative rate return flux. The bicarbonate is buffered by the large
carbonate reservoir. The activity of the biosphere,
o+oo 1 which is dominated by the Northern Hemisphere
P+OO 1.03
biota, induces a seasonal cycle of up to l % o at high
Q+00 0.93
0 + PP 1.23 northern latitudes at the surface. The seasonal mini-
P + PP 1.02 mum is in spring when C 0 2 from respiration, oxida-
Q + PP 1.03 tion, and fossil fuel emission have peaked. Fossil fuels,
O+QQ 1.53 and plant organic matter have 613C values that
P+QQ 1.31
average approximately -25%0. To what degree the
Q+QQ 1.03
effect of El Niiio can be extracted from the two or three
globally accurate isotope time-series is debated, but
dissolution into the oceans depends on temperatures
70-110%0 and by almost the same amount in "0. In and wind and causes a small fractionation. NIST has
fact, A i 7 0 E 33%0. No other substance known on produced gaseous standards for the required high-
Earth has this degree of excess of 170.Via this mass- precision isotope analysis of atmospheric C02. A
independent enrichment (the excess " O ) , certain useful application of 613C(C02)has been the identi-
other trace gases (e.g., stratospheric C02) exhibit fication of a large Northern Hemispheric terrestrial
MIF. The cause for the complex isotopic anomaly of carbon sink in 1992 and 1993.
O3 is only slowly becoming understood. The reaction "O( C 0 2 ) is determined by the exchange with the
rate constants for permutations of the isotopes, and main water reservoirs. Isotopic exchange with cloud
their temperature and pressure dependence, have been droplets is far too slow in comparison to the atmos-
measured (examples are given in Table 4).Despite the pheric residence time of C02. No exchange with water
many experimental data, there is not yet a theoretical vapor is possible because bicarbonate formation is
explanation. required. However, exchange with soil water, leaf
The isotopic composition of stratospheric O3 has water, and ocean water is important. In this, C 0 2
also been debated because some extreme enrichments strives toward isotopic equilibrium accompanied by a
have been reported, though these have not been temperature-dependent enrichment of about 40%0.
confirmed. Improved measurement of O3 collected Accordingly, the 6 l 8 0 value of C 0 2 is close to 40%0
from the stratosphere up to 35 km shows that 6 l 8 0 is (V-SMOW, Vienna-Standard Mean Ocean Water).
between 70%0 and 110%0. Laboratory experiments Note that the standard V-SMOW-C02 is often used,
show that this range of 6 values corresponds to thus giving a value of near zero. Because leaf water is
temperatures of 270-200 K. Thus, the enrichment of enriched owing to the equilibrium and kinetic isotope
stratospheric ozone is a measure of temperature. effects accompanying stomatal evaporation, C 0 2 in
Similar results are obtained using 170.Tropospheric equilibrium with leaf water is more strongly enriched.
O3 has been collected and analyzed in an urban The "0 isotope effect of emission of C 0 2 by fossil fuel
environment: the 6l8O values ranged between 80%0 combustion ("0 = 23.5%0)has only a small influence
and 100%0 and were not correlated with the mixing owing to the size of other large exchange fluxes. When
ratios between 5 and 90nmol/mol. The delta values C 0 2 from two sources is mixed, each with a given
agree with laboratory experiments. 1sO/160ratio, one can calculate the (low) abundance
of C180180.This calculated ratio does not necessarily
correspond with the real C 1 801 80 ratio, because it
Carbon Dioxide, COP would require scrambling of "0 between the C 0 2
molecules, which does not occur (no gaseous ex-
Troposphere
change). Analytical problems in determining the ratio
Quantifying fluxes of C 0 2 between the terrestrial of C1sO1sO to C160160 have prevented practical
carbon reservoirs is of great importance, and isotopic application of this generally overlooked effect.
analysis is useful for this. The combustion of fossil
fuels has decreased 613C of atmospheric C 0 2 from a
Stratosphere
preindustrial value of about - 6 %O to presently about
- 7.8%0V-PDB. The preindustrial value is determined The seasonality of 613C and C 0 2 mixing ratios
by the equilibrium fractionation (-9%0 at 15OC) propagates with delay and attenuation into the stra-
between gaseous C 0 2 and ocean bicarbonate tosphere. This allows accurate stratospheric measure-
I3C E 2%0). The fractionation associated with the ments to be used as a clock. In the stratosphere, C 0 2
acquires MIF from O3 via quenching of O(lD), at 250 K and plays no significant role. 6"O increases
+
according to 0 3 + hv + 0 2 O(lD) followed by from zero at the tropopause to 10%0at 40 km.
+
o('D) + co2-+ co3*+ o ( ~ P ) C O ~ .A ~ ~in- O
creases accordingly from zero to 14%0.This interesting
signal is a new gaseous tracer for air entering the Methane, CH4
troposphere from the stratosphere, such as O3 and The isotopic compositions of atmospheric methane
14~0.
and its sources are shown in Figure 2; a large depletion
is observed. The actual atmospheric composition
Water, H20 deviates from the average source composition because
of the isotopic fractionation mainly in the sink
Troposphere reaction of CH,+OH. Also, the smaller soil sink
tends to enrich the methane that is left in the
The International Atomic Energy Agency coordinates atmosphere. Note the large difference for deuterium.
the global monitoring of the isotopic composition of Despite the considerable spread in the source values,
precipitation for hydrological applications (isotope inverse modeling using the isotopic composition has
hydrology). There are large seasonal and geographical helped to further constrain the methane budget. Both
differences. Water evaporating from the oceans is 6I3C and 6D show seasonal cycles with a total
depleted, but the condensing water is again enriched amplitude of approximately 0.5%0 depending on
relative to the vapor. With increasing distance into the latitude and can be resolved with difficulty. There are
continents, D and "0 decrease (distance effect) presently few data on deuterium available, but the
following the mentioned Rayleigh fractionation. advances in isotope mass spectrometry that have been
With increasing latitude and decreasing temperatures, mentioned will change this. Interannual variations in
vapor and precipitation cause further depletion. In the Southern Hemisphere have been associated with
Antarctica, 6D can be as low as -400%0, and 6'*0 biomass burning. Moreover, at times small increases in
reaches -50 %o. Isotopic analysis of ice cores is one of 613C have been detected that were attributed to the
the main tools in paleoclimatology. Global precipita- reaction of C1 + OH; as Table 5 shows, this reaction
tion D and "0 isotope values define the 'meteorolog- has a characteristic large 13C kinetic isotope effect.
ical water line' 6D s 86'*0 + 10%0.Hailstones have The strong increase in atmospheric CH4 induced an
been assayed layer by layer for study of their formation isotopic disequilibrium in which the isotopic compo-
process. Relatively little attention has been paid to sition came closer to that of the averaged source.
tropospheric water vapor analysis. Analysis of CH4 extracted from firn air reveals a clear
minimum 6D value about 75 years ago. After a
Stratosphere weakening in the increase of CH4 over the last decade,
the equilibrium was approached more closely and 6D
Because of the many roles of atmospheric water
increased again.
through latent heat, cloud albedo, heterogeneous
Stratospheric chemistry produces large isotope
chemistry, and the radiation budget, efforts are made
changes in CH4 because of the greater role of C1 and
to use isotopic analysis. In the stratosphere, gas-phase
O ( l D ) as sinks. Table 5 shows that O ( l D ) has a large
chemical interactions affect the isotopic composition.
isotope effect, which was not expected on the basis of
Through cooling and concomitant condensation, wa-
the rapid kinetics of this reaction. The isotopic
ter vapor reaching the tropopause attains 6D =
composition of CH4 in the stratosphere can be
-670%0 and 6 l 8 0 = During transport into
successfully modeled using a 2D model, incorporating
the middle atmosphere, isotope exchange and addi-
the fractionation factors from Table 5. Such results are
tion of H20 from methane oxidation take place.
of further use for calculating 6D of H20 and 6I3C of
Exchange of oxygen is via the HO, family of reactions
CO.
involving oxygen and ozone. These processes enrich
the isotopically depleted vapor imported from the
troposphere. Above 40 km, 6D reaches about
-400%0, and stays constant at that level. 6 l 8 0
Hydrocarbons and Methyl Chloride
increases by as much as 100%0. There is a clear The recent introduction of coupled gas chromatogra-
deviation from the meteoric water line. Furthermore, phy-isotope mass spectrometry has allowed the anal-
stratospheric H20 obtains MIF largely from 0 3 via ysis of nanomole amounts of substances, bringing the
+ +
H 0 3 + O H 0 2 and the subsequent hydrogena- low-abundance trace gases within reach of isotopic
tion of OH. The oxygen isotope exchange between analysis. Values obtained for ethane, ethene, and
O H and H20 has a rate constant of 6 x 10 - "cm3 s - ' propene in background air in New Zealand yielded
1038 ISOTOPES, STABLE
Figure 2 Overview of the 13C and deuterium (D'H) isotope ratios of the main methane sources. The arrow indicates the kinetic isotope
effect.
613C values of - 22%0to - 29%0,which is typical for -32%0. One has to assume either that additional
organic matter. Methyl chloride was strongly depleted isotope fractionation occurs in the formation of CO
at -43.5%0. Samples collected south of Japan, from the photolysis of formaldehyde, or that the yield
between 5" and 3 5 " N showed the following 613C of C O from CH,+OH, is lower than the generally
values: ethane, propane, n-butane, n-pentane, i-bu- accepted 85%. For 13C there is a pressure effect. The
tane, and i-pentane between - 30 and - 23%0;ethyl- KIE maximizes at 5%0 at 1013 hpa, and turns into a
ene and propylene between about - 30 and - 10%0; negative effect at 300 hpa. The annual average 613C
acetylene between - 20 and - 2%0; methyl chloride value in the Southern Hemisphere is approximately
between - 40 and - 30%0. - 29%0;in the mid-latitude Northern Hemisphere it is
- 27%0.The annual cycles are dominated by the CH4
source effect, not the KIE of C O + OH.
For "0 there is a negative isotope effect of nearly
Carbon Monoxide, CO - 10%0, almost independent of pressure. Generally,
The shorter lifetime of C O results in larger concen- the farther away from C O sources, the more negative
tration and isotopic changes (Figure 3). Concer- the 6 l 8 0 becomes, reaching - 10%0in the lowermost
ning source signatures, C O from high-temperature stratosphere. There are only few stratospheric data for
combustion processes adopts the 6"O value of CO, but air samples collected during ozone hole
atmospheric oxygen ( +23.5%0), without much frac- conditions in the Antarctic lowermost stratosphere
tionation. This gives a useful clear signal for the
technological source. Also distinctive is the low 613C
value of C O derived from CH4 oxidation. This value is Table 5 Kinetic isotope fractionation factors (KIE) for reaction
believed to be about - 51%0,composed of the - 47%0 with CH4 at 296 K
of CH4 further lowered by the kinetic fractionation of I3c D
4%0 in C H 4 + O H (Table 5 ) . There are problems in
closing the isotope budget for I3CO in the Southern OH 1.004 1.294
CI 1.065 1.50
Hemisphere using this information, because atmos- 1.013 1.06
WD)
pheric 613C(CO)values appear rarely to fall below
Alert
o Spitsbergen
A IzaAa
3 Baring head
A Scott base
613C(%o V-PDB)
Figure3 Overview of the 13Cand l80isotope ratiosof the main carbon monoxidesources. The arrow indicates the kinetic isotope effect.
Observational values for some Northern Hemisphere locations are shown. The spread in the data is mostly due to the seasonal cycle.
(NMHC, non-methane hydrocarbons.)
yielded CO with S13C as low as -43%0. Modeling position of N 2 0 has been studied intensively. The
confirmed that CO from CH4+C1 was the cause of 6"N and Sl8Ovalues of tropospheric N2O show little
this. Another application of the large KIE in CH,+Cl scatter around approximately 7%0and 21%0relative to
is the estimation of free chlorine during Arctic low- atmospheric N2 and 0 2 , respectively. The main
ozone events at the surface. Small downward excur- sources of N 2 0 are based on microbial nitrification
sions in S l 3 C ( C 0 )were observed during such events. and denitrification in soils and the ocean, and are
After MIF had been detected for O3and C02, it also generally depleted relative to the atmosphere. The
was discovered in CO. MIF in CO has two causes. One range of 6 values is considerable. Soil gases form the
is that unsaturated organic compounds react with 03. most depleted source type, with averages of about
CO from these reactions is a small source of CO - 13%0and l O % O , respectively. A source identified
enriched in "0 and exhibiting MIF derived from 0 3 . with enrichment was N 2 0 from denitrification in
However, the main cause lies in the important reaction upwelling deep water in the Arabian sea. Enriched
CO + O H + CO,+H. Assuming for the moment that N2O was also found in the Pacific. These sources do
ClSOreacts 10%0faster than C l 6 0 , CI7O should react not account for the atmospheric enrichment. The
about 5%0 faster if the normal widespread mass- enrichment of atmospheric N 2 0 has been explained
dependent fractionation applies. In contrast, Ci 70 on the basis of its main stratospheric sink, i.e.,
reacts nearly as rapidly as Ci60. This then results in an photolysis. N 2 0 that escapes photolysis and is reim-
excess of I7O in atmospheric CO. There is not yet a ported into the troposphere causes enrichment relative
theoretical explanation for MIF in CO. Because all to the average sources. Theory and experiment of
major CO sources are mass-dependent, 6170(CO)is a photolysis show a qualitative agreement. Photolysis
unique signal indicating the 'aging' of C O by reaction experiments over the entire range of wavelength of
with OH. interest still has to be performed. Analysis of strato-
spheric N2O samples confirms that enrichment in-
creases with altitude.
Increased interest in N 2 0 was sparked by the
Nitrous Oxide, N20 awareness that in this linear molecule, NNO, frac-
Despite its long lifetime and concomitant small tionation through photolysis is not identical for the
isotopic and abundance variability, the isotopic com- two different N atoms. Instrumental innovation in
1040 ISOTOPES,STABLE
Remaining N,O fraction (%) 0 2 , its isotopic composition should be strictly mass-
o,loi--: 20 40 60 80 oi dependent.
However, a very small fraction of 0 2 is continuously
-+
7
0.08 converted to 0 3 ,which exhibits MIF. Thus a small
excess of "0 is withdrawn. In the stratosphere, this
0
0.06 MIF signal is transferred from the O3pool via O ('D) to
1 become incorporated into C02. The ensuing step
'D2. 0.04 in the process is that COf reenters the troposphere,
-c where it isotopically exchanges with H20. In this
0.02 way, C 0 2 loses its MIF signature continuously to the
large reservoir of H2O. Another process transferring
0.00
MIF is the reaction O('D)+H,O, which also leads
-2.5 -2.0 -1.5 -1.0 -0.5 0.0
to a net loss of I7O. Both processes remove a small
In (remaining N,O fraction)
excess of "0 from the 0 2 reservoir. Given the long
Figure4 A Rayleigh-typeof plot forthe change in the I5N isotopic lifetime of atmospheric oxygen of roughly 1000 years,
composition of N 2 0 subjected to photolysis at 193 nm.The enrich- it acquires a deficit of "0, resulting in the value
ment for I4Ni5NO is considerably larger than for 1 5 N i 4 N 0 . 6170 = -0.15%0. Although this value is small, there
are applications in limnology and oceanography in
which atmospheric oxygen can be distinguished from
photosynthetic oxygen.
mass spectrometry now allows the detection of the
isotopic ratios for both positions using only small
amounts of sample. Figure 4 shows the result of
N 2 0 photolysis at 193nm for I5N. The fraction-
ation factors for 15Ni4N0 and 14Ni5N0 are Aerosols
10.9%0 and 35.7%0 respectively. Thus 14Ni5N0 is The oxidation of SO2 by O3 or H202 in the liquid
less likely to be photolyzed. The same applies for phase leads to a small degree of MIF in atmospheric
14N14N180, for which at this wavelength the frac- sulfates. This can possibly be applied to trace the
tionation factor is 17.3%0. For the other important oxidation pathway of S02. There is one report of
stratospheric reaction, O ( l D )+N2O, only few data H202 possessing MIF, for which the reason is not
are available. known. Sulfur isotope measurements on marine sul-
Possible gas-phase sources of N20 have been fate aerosol particles support the hypothesis that
proposed in the literature, and the occurrence of dimethyl sulfide can be a source of non-sea-salt sulfate.
MIF, albeit at a low level (6170 E l % o ) , has intensified Some research has also been dedicated to chlorine
this interest. N o confirmation has been provided, and isotopes in aerosols.
the cause of MIF in N 2 0 remains unknown.
Molecular Oxygen, O2 See also

In view of the sheer size of this reservoir, no usable Aerosols: Physics and Chemistry of Aerosols. Carbon
isotope effects were originally contemplated to occur. Dioxide. Evolution of Atmospheric Oxygen. Evolu-
Notwithstanding, it has been established that even tion of Earth's Atmosphere. Global Change: Ozone
atmospheric oxygen itself possesses a small degree of Trends. Land-Atmosphere Interactions: Trace Gas
MIF, 6170 = -0.15%0, which offers interesting appli- Exchange. Methane. Planetary Atmospheres: Jupiter
cations. The "0 isotopic enrichment of 0 2 relative to and the Outer Planets: Mars; Venus. Stratospheric
Ozone Recovery. Stratospheric Water Vapor. Trop-
ocean water (V-SMOW) is 23.5%0.The cause for this
ospheric Chemistry and Composition: Carbon Mon-
enrichment is well understood. Respiration processes oxide.
in plants and soils favor the use of the lighter of the
isotopic 0 2 molecules. In addition to this is that the
leaf water in plants, which forms the substrate for the
oxygen appearing in photosynthetic products, is Further Reading
enriched. These processes keep atmospheric oxygen Gonfiantini R, Stichler Wand Rozanski K (1995)Standards
enriched relative to the very large reservoir of terres- and intercomparison materials distributed by the Inter-
trial water. Because of this exchange between the national Atomic Energy Agency for stable isotope meas-
large terrestrial water reservoir and atmospheric urements. In: Reference and Intercomparison Materials
for Stable Isotopes of Light Elements, IAEA-Techdoc- Thiemens MH (1999) Mass-independent isotope effects in
825, pp. 13-29. Vienna: IAEA. planetary atmospheres and the early solar system. Science
Kaye JA (1987) Mechanisms and observations for 283: 341-345.
isotope fractionation of molecular species in Richet P, Bottinga Y and Javoy M (1977) A review of
planetary atmospheres. Reviews of Geophysics 25: hydrogen, carbon, nitrogen, oxygen, sulphur, and chlo-
1609-1658. rine stable isotope fractionation among gaseous mole-
Kaye JA (ed.) (1992) Isotope Effects in Gas-Phase Chemis- cules. Annual Review of Earth and Planetary Science 5:
try. Proceedings of a symposium by the Division of 65-110.
Physical Chemistry, 201st National Meeting of the Rockmann T, Brenninkmeijer CAM, Saueressig G, et al.
American Chemical Society, Atlanta, Georgia. ACS (1998) Mass-independent oxygen isotope fractionation
symposium Series 502. Washington DC: American in atmospheric C O as result of the reaction C O + O H .
Chemical Society. Science 281: 544-546.
JET STREAKS 1043
P Cunningham,Florida State University, Tallahassee, General Characteristics of Jet Streaks

FL, USA
D Keyser, University at Albany, State University of
Jet streaks are observed along the polar-front and
New York, Albany, NY, USA
subtropical jet streams in both the Northern and
Copyright 2003 Elsevier Science Ltd. All Rights Reserved Southern Hemispheres and can be either mobile or
stationary with respect to the stream in which they are
embedded. Like atmospheric fronts, to which they are
Introduction closely related, jet streaks are often significantly
anisotropic, exhibiting disparate length scales in the
Jet streaks, defined as localized wind speed maxima along-jet and cross-jet directions, with the former
along the axis of a jet stream, are generally found in being up to an order of magnitude larger than the
association with frontal zones in the upper tropo- latter. In the case of the polar-front jet stream, jet
sphere, and together these features are often referred streaks typically have length scales of 2000 km or less
to as ‘jet-front systems’. In many cases, jet streaks are in the along-jet direction and progress eastward
also accompanied by a localized downward intrusion through the meanders of the stream. Some polar-front
of air from the stratosphere into the troposphere jet streaks are observed to translate continuously
known as a ‘tropopause fold’. The interest that jet along the jet stream, whereas others travel discretely
streaks have attracted in the fields of synoptic and and appear to jump between the north-westerly and
dynamic meteorology stems from their importance in south-westerly flow inflections of the wave pattern in
the development and behavior of synoptic-scale and the stream. In the case of the subtropical jet stream, jet
mesoscale weather systems, including extratropical streaks possess along-jet length scales of 4000-
cyclones and mesoscale convective systems. In addi- 8000 km and tend to be approximately stationary
tion to their association with these weather systems, jet and located in the ridges of the wavelike meanders in
streaks are often implicated with other phenomena the stream.
and processes of interest to atmospheric science, such Jet streaks often play a critical role in the develop-
as clear air turbulence, stratosphere-troposphere ex- ment of extratropical cyclones and in the organization
change, and a class of motions referred to as ‘unbal- of mesoscale convective systems, and as such have
anced’ (e.g., inertial and buoyancy oscillations). The become an important part of mainstream synoptic
basic structure of jet streaks and the role that these meteorology and weather forecasting. The association
features play in the evolution of extratropical cyclones of jet streaks with these phenomena is related to the
and of mesoscale convective systems are often por- observation that the divergence of the horizontal wind
trayed schematically in terms of conceptual models (hereafter, this quantity will be referred to as the
that relate jet streaks in various flow configurations to horizontal divergence, while the terms divergence and
characteristic patterns of horizontal divergence and convergence will be used to distinguish between
vertical motion. Dynamical interpretation of jet positive and negative values of this quantity, respec-
streaks has proceeded along two lines of thinking: in tively) is frequently large in the vicinity of jet streaks.
the first, referred to as ‘balanced’, it is assumed that a The relationship of horizontal divergence in the upper
dynamical relationship constrains the wind and geo- troposphere to changes in sea level pressure and to
potential fields; in the second, referred to as ‘unbal- patterns of vertical motion in the middle troposphere
anced’, no such constraint exists. In the balanced case, is well known. Indeed, a fundamental principle of
jet streaks may be intepreted either as an integral part early synoptic meteorology states that the develop-
of upper-tropospheric baroclinic waves, in which case ment of surface cyclones requires upper-level diver-
their evolution is controlled by energy dispersion gence slightly in excess of lower-level convergence,
associated with the wave, or in terms of the interaction which results in a net reduction in mass in the column
between coherent vortices and the jet stream, in which and a decrease in the sea level pressure. The vertical
case their evolution occurs in response t o the motion circulations associated with patterns of horizontal
and evolution of the constituent vortices. In the divergence in the upper troposphere may be inferred
unbalanced case, the dynamics are strongly coupled using the equation of mass continuity formulated in
to a category of buoyancy oscillations referred to as pressure coordinates, given by V V = -aco/ap, where
inertia-gravity waves, which may strongly modulate V is the horizontal gradient operator evaluated on an
weather systems on short time scales. isobaric surface, Vis the horizontal wind velocity, and
1044 JET STREAKS
co is the vertical velocity in pressure coordinates. The as determined by the configuration of the 60 m s -
pressure-coordinate vertical velocity is related ap- isotach.
proximately to the height-coordinate vertical velocity, A vertical cross-section through the jet streak under
w , via w FZ -pgw, where p is air density, g is the consideration in the cross-jet direction is shown in
acceleration due to gravity, and positive (negative) o Figure 2, which illustrates the components of a
corresponds to downward (upward) motion. The jet-front system. The tropopause, depicted as the
pressure-coordinate form of the vertical velocity is thick solid line, is defined ‘dynamically’ as the
adopted for the remainder of this article. Since the 2 P W potential vorticity surface (1P W =
vertical velocity is generally close to zero near the 10 - m2 s - K kg - I). Here potential vorticity, P ,
tropopause, divergence (convergence) in the upper is given by P = -g(a€J/ap)(f + c0),where f is the
troposphere will be associated with ascending (de- Coriolis parameter, 0 is the potential temperature, and
scending) motion in the middle troposphere. This is the vertical component of the relative vorticity
relationship between upper-tropospheric horizontal evaluated on an isentropic surface. In the vicinity of
divergence and middle-tropospheric vertical motion the jet streak, the tropopause transects the core of
will be alluded to frequently in the remainder of this maximum wind speed (i.e., the ‘jet core’) and slopes
article. steeply, being considerably lower on the poleward side
Although the observed structure and behavior of jet than on the equatorward side. The downward and
streaks can vary considerably from case to case, equatorward protrusion of the tropopause into the
several properties are common to these features and middle troposphere beneath the jet streak corresponds
are illustrated through the analysis of a jet streak to an incipient tropopause fold. In a well-developed
observed over the north central United States on 3 fold, the tropopause generally exhibits an S-shaped
November 1995. In Figure 1, which depicts the wind pattern, such that there is a significant region in which
speed and geopotential height fields at 300 hPa, the jet the tropopause height is multivalued. Above the jet
streak is located in the south-westerly flow inflection core, the contours of constant potential temperature
of a synoptic-scale trough-ridge pattern. In this jet (i.e., isentropes) slope upward. Below the jet core, the
streak, the maximum wind speed is approximately isentropes slope downward, such that the potential
’,
75 m s - and the aspect ratio expressed in terms of temperature gradient evaluated on an isobaric surface
along-jet to cross-jet length scales is approximately 4:l in this region is locally large. This area of locally
Figure 1 European Centre of Medium-Range Weather Forecasts (ECMWF) analysis on the 300 hPa pressure surface, valid at 1200
UTC 3 November 1995, depicting wind speed (values greater than 50m s - ‘ shaded as indicated) and geopotential height (contour
interval 120m; solid lines).
JET STREAKS 1045
Figure2 Cross-section along A-A' indicated in Figure 1, valid at 1200 UTC 3 November 1995, depicting wind speed normal to the plane
of the section (values greater than 50 m s - ’ shaded as indicated), potential temperature (contour interval 4 K; thin solid lines) and the
dynamic tropopause (thick solid line).
enhanced horizontal potential temperature gradient The applicability of the thermal wind relation
defines the upper-tropospheric frontal zone, which in referred to above to jet-front systems implies the
this case extends well into the middle troposphere and validity of geostrophic balance, which will now be
joins with the upward extension of a surface frontal employed in considering the dynamics of jet streaks.
zone. The fan-shaped pattern of the isentropes ob- Geostrophic balance is valid to a first approximation
served in conjunction with this jet streak is a conse- in many jet streaks, such that the so-called ageo-
quence of the thermal wind relation, which links the strophic wind (defined as Vag = V - V,, the vector
vertical shear of the geostrophic wind to the horizontal difference between the actual horizontal wind and the
potential temperature gradient. geostrophic wind) is relatively small in comparison to
Jet streaks in the upper troposphere and their the geostrophic wind. The magnitude of the departure
associated frontal zones can also be preferred loca- from geostrophic balance is quantified by the Rossby
tions for phenomena not normally thought of as number, Ro, which is defined as the ratio of the
related to weather. In the regions of large vertical shear characteristic scales for the velocity acceleration and
above and below the jet streak, the Richardson the Coriolis force, and is given by Ro = V/fL,where L
number, Ri, defined as Ri = (g/B)(ae/az)/jaV/az12, and V may be taken to be the along-jet length scale and
can become small. When this is the case, the flow can the maximum wind speed of the jet streak, respective-
break down into turbulent eddies. Since this break- ly. If Ro is small with respect to unity, the jet streak will
down often occurs in regions of background subsid- be in approximate geostrophic balance. In such cases
ence and relatively cloud-free air, this phenomenon is the so-called quasi-geostrophic system, an approxi-
generally referred to as ‘clear air turbulence’, which mation to the full equations of motion valid for small
can be hazardous to aircraft. Moreover, the region of Ro, provides a useful foundation for understanding
turbulent mixing tends to be localized in the vicinity of the structure and behavior of jet streaks. In fact, the
the tropopause fold, which contains stratospheric air quasi-geostrophic system often is applicable under
protruding downwards into the troposphere. In the circumstances beyond the limits of its strict validity, so
region of mixing, the tropopause, which ordinarily that it may be employed qualitatively to interpret the
corresponds to a material surface separating strato- structure and behavior of jet streaks in which Ro is not
spheric from tropospheric air, is highly porous, particularly small. Many of the existing dynamical
allowing for a vigorous two-way exchange of air concepts regarding jet streaks are based on deductions
and trace constituents between the stratosphere and derived from the quasi-geostrophic system. Neverthe-
troposphere. less, sometimes Ro can be so large in the vicinity of jet
1046 JET STREAKS
streaks that the quasi-geostrophic system is not even ically curved, may be super-geostrophic. Explanations
qualitatively useful. In such jet streaks, unbalanced for employing the terminology of ‘entrance’ and ‘exit’
motions may be important to the evolution of the with regard to jet streaks and for the presence of sub-
streak and the full equations of motion, also known as geostrophic and super-geostrophic along-jet flow will
the primitive equations, are required to elucidate the be provided subsequently.
properties of the flow and the behavior of the streak.
The ageostrophic wind at 300 hPa for the jet streak
under consideration, shown in Figure 3, displays a
four-gyre circulation pattern that is cyclonic upstream
Conceptual Models of Jet Streaks
and downstream of the jet core and anticyclonic on the Since the horizontal divergence is often a small
flanks of the streak. Hence, in the region upstream of residual between two larger kinematic quantities
the jet core, known as the entrance region, the that are nearly equal in magnitude but opposite in
ageostrophic wind is directed toward the poleward sign, uncertainties arise in calculating the upper-
(i.e., cyclonic-shear) side of the streak, corresponding tropospheric horizontal divergence directly from con-
to lower geopotential height; in the region down- ventional upper-air data. These uncertainties may be
stream of the jet core, known as the exit region, the particularly significant if the jet streak is located in a
ageostrophic wind is directed toward the equatorward data-sparse region, or if the streak is so narrow that the
(i.e., anticyclonic-shear) side of the streak, corre- observational network is unable to resolve accurately
sponding to higher geopotential height. There is also a the structure of the streak. Consequently, the horizon-
significant along-jet component of the ageostrophic tal divergence has been inferred using air parcel
wind directed upstream, indicating that the actual arguments applied to the vector momentum equation
wind in this jet streak is sub-geostrophic (i.e., the and to the vorticity equation, the results of which have
actual wind speed in the core of the jet is smaller in been portrayed schematically in terms of conceptual
magnitude than the geostrophic wind speed), a prop- models relating horizontal divergence patterns to
erty that may be related to the cyclonically curved various aspects of the structure of jet streaks. Several
orientation of the jet streak. Nevertheless, this partic- well-known conceptual models of jet streaks
ular property is not generic; some jet streaks are are reviewed below: (i) models of straight and
approximately geostrophic in the along-jet direction, curved jet streaks; (ii) models of jet streaks coupled
whereas others, particularly those that are anticyclon- with other upper-tropospheric jet streaks or with
Figure3 ECMWFanalysis on the300 hPa pressure surface, valid at 1200 UTC3 November 1995, depicting wind speed (values greater
than 50 m s - ’ shaded as indicated) and ageostrophic wind (arrows; vector scale indicated at bottom of diagram).
JET STREAKS 1047
lower-tropospheric frontal zones; and (iii) a model facing downstream). To infer the nature of the
describing the life cycle of a jet streak embedded within ageostrophic flow and the horizontal divergence in
an evolving synoptic-scale baroclinic wave. this and subsequent conceptual models, it is conven-
ient to adopt a natural coordinate representation of
Straight Jet Streaks the frictionless vector momentum equation (eqns [ 1J
and [2]).
Perhaps the best known conceptual model of jet 1DV
Vagn = - __ 111
streaks, illustrated in Figure 4, depicts a straight f Dt
isotach maximum in which the isentropes are assumed
to be oriented parallel to the jet axis. This conceptual K , V2
model is often referred to as the ‘four-quadrant model’ uags = --
because the upstream (entrance) and downstream
f
(exit) regions of the streak are divided into left and Here vagn and vags are the cross-stream and along-
right regions by the jet axis (with left and right defined stream components of the ageostrophic wind in a
Indirect B’
(8) circulation circulation
Figure 4 Schematic illustration, applicable to the Northern Hemisphere, of ageostrophic circulations and vorticity patterns for a straight
jet streak. (A) Transverse ageostrophic wind components and associated patterns of convergence (CONV) and divergence (DIV) in the
entrance and exit regions at the level of maximum wind. (B) Transverse ageostrophiccirculations in entrance (cross-section A-A’) and exit
(cross-section B-B’) regions of jet streak depicted in (A), along with schematic isentropes (dotted lines) and location of jet core (J). (C)
Relative vorticity and associated advection patterns, with NVA and PVA indicating negative (anticyclonic) and positive (cyclonic) vorticity
advection, respectively. (From Uccellini LW and Kocin PJ (1987) The interaction of jet streak circulations during heavy snow events along
the East Coast of the United States. Weather and Forecasting 2: 289-308. American Meteorological Society, Boston.)
1048 JET STREAKS
'right-handed' natural coordinate system such that regions will experience an increase in cyclonic relative
n = k x s, s = V/V, V is the horizontal wind speed, vorticity, and an air parcel traveling downstream
and K , is the parcel trajectory curvature. At the level of through the right entrance or left exit regions will
the jet core, the vertical velocity is assumed to be small experience an increase in anticyclonic relative vorti-
and may be neglected; hence the Lagrangian rate of city. Equation [3] indicates that convergence is ex-
change D/Dt (= a / a t + V . V) may be defined follow- pected in the left entrance and right exit regions,
ing the horizontal flow. whereas divergence is expected in the right entrance
In general, jet streaks travel at a speed that is slower and left exit regions.
than the maximum wind speed in the jet core. Con- The conceptual model shown in Figure 4 assumes
sequently, air parcels travel through the streak, enter- that isentropes are parallel to the jet axis. For many
ing upstream and exiting downstream - hence the observed jet streaks, however, this assumption is
terms entrance and exit regions introduced previously. overly restrictive and the isentropes are generally
An air parcel in the entrance region will experience aligned at an angle to the jet axis. The effect of the
increasing wind speed (DVIDt > 0), and an air parcel resultant along-jet temperature advection on the
in the exit region will experience decreasing wind patterns of horizontal divergence and vertical motion
speed (D V/Dt < 0). Since the curvature is negligible accompanying a straight jet streak is illustrated in
for the straight jet streak depicted in Figure 4A the Figure 5 for two different alignments of the isentropes
along-stream component of the ageostrophic wind is relative to the jet axis. When there is cold advection
small (see eqn [ 2 ] )the
; cross-stream component of the along the jet (Figure 5A), the transverse vertical
ageostrophic wind is directed from higher to lower circulations are shifted toward the equatorward
values of geopotential height in the entrance region (anticyclonic-shear) side of the jet streak in the
and from lower to higher values in the exit region (see entrance region and to the poleward (cyclonic-shear)
eqn [l]).Away from the jet axis, where speed accel- side in the exit region. Conversely, along-jet warm
erations are weaker, the ageostrophic flow decreases in advection (Figure5B) results in a shift of the transverse
magnitude and thus there is convergence in the left
entrance and right exit regions, and there is divergence
in the right entrance and left exit regions. The vertical
circulations transverse to the jet streak, depicted in
Figure 4B, are thermally direct in the entrance region
and thermally indirect in the exit region. The sense of
these circulations is consistent with the conversion
from potential to kinetic energy required for an air
parcel to increase its speed in the entrance region of the
streak, and with the conversion from kinetic to
potential energy required for an air parcel to decrease
x =15" @ +A@
its speed in the exit region.
It is also possible to deduce the patterns of horizon-
tal divergence illustrated in Figure 4A using the
equation governing the evolution of the vertical
component of the relative vorticity evaluated on an
isobaric surface, 5, again at the level of the jet core,
given by eqn [3].
(B) @+A@ ~=-15" @+A@

In addition to the neglect of the tilting, vertical
Figure 5 Schematic illustration of geopotential height (heavy
advection, and friction terms, the Lagrangian rate of solid lines), wind speed (heavy dashed lines), and potential
change of planetary vorticity is here assumed to be temperature (thin solid lines) associated with a straight jet streak
small in comparison to the rate of change of relative for isentropes aligned at a small angle to jet axis: (A) along-jet cold
vorticity. As illustrated in Figure 4C, a straight jet advection; (B) along-jet warm advection. Arrows indicate the sense
streak exhibits cyclonic relative vorticity on its pole- of the cross-front ageostrophic wind component, and the plus and
minus signs indicate maxima and minima in the middle-tropo-
ward side and anticyclonic relative vorticity on its spheric pressure-coordinate vertical velocity. Recall that positive
equatorward side. Hence, an air parcel traveling (negative) values of this quantity correspond to downward
downstream through the left entrance or right exit (upward) motion. (Adapted from Shapiro (1982).)
JET STREAKS 1049
circulations to the poleward (cyclonic-shear) side in the ageostrophic wind is directed downstream (uagS> 0)
the entrance region and to the equatorward (anticy- and the actual wind is super-geostrophic. In the
clonic-shear) side in the exit region. As depicted in inflections between the trough and ridge, the along-
these illustrations, these shifts in the vertical circula- stream ageostrophic wind is zero; hence, divergence is
tions can lead to the maximum vertical motions being expected between the trough and ridge, whereas
located beneath the jet axis, rather than to either side convergence is expected between the ridge and trough.
as in Figure 4. In the cold advection case, such a Consequently, the pattern of horizontal divergence for
configuration may be important to the development of a jet streak embedded in such a curved jet stream may
tropopause folds, which requires the presence of be modified significantly from that shown in Figure
significant subsidence localized beneath the jet axis. 4A, depending on where the streak is located with
respect to the trough or the ridge. For example, for a jet
Curved Jet Streaks streak located near the base of the trough, the
convergence and divergence in the left entrance and
Although evidence exists supporting the idealized left exit regions of the streak will be enhanced by the
structure of isolated straight jet streaks such as those convergence upstream and the divergence down-
depicted in Figures 4 and 5, jet streaks exhibiting a stream associated with the jet stream. As a result, the
four-quadrant pattern of horizontal divergence and horizontal divergence may exhibit a two-cell pattern,
vertical motion are the exception rather than the rule. rather than the four-cell pattern associated with a
The rarity of the four-quadrant pattern is a conse- straight jet streak. The two-cell pattern appears to
quence of the fact that few jet streaks are immune to be more common, given the frequent association of
the effects of curvature: many jet streaks have curved jet streaks with so-called ‘short-wave troughs’ in
axes, and even jet streaks that have straight axes tend the upper troposphere, which are characterized by
to be embedded in a large-scale jet stream that may cyclonic curvature.
itself be curved. The influence of curvature on the
horizontal divergence patterns associated with jet
streaks may be considered by appealing to the sche- Coupled Jet Streaks
matic depiction of a steady curved jet stream shown in Sometimes two separate jet streaks in the upper
Figure 6 , which exhibits uniform wind speed in the troposphere may come into such close proximity
along-stream direction such that DV/Dt 3 0. Assum- that the vertical circulations associated with these
ing that parcel trajectory curvature may be approxi- features become coupled. Such a configuration, de-
mated by streamline curvature, eqn [2] implies that in picted schematically in Figure 7, is common to
the trough, where the curvature is cyclonic (K,> 0 in cyclogenetic events that produce heavy snow along
the Northern Hemisphere), the ageostrophic wind is the East Coast of the United States. In this configura-
directed upstream (vags< 0) and the actual wind is sub- tion, the left exit region of the equatorward jet streak
geostrophic, whereas in the ridge, where the curvature and the right entrance region of the poleward jet streak
is anticyclonic (K, < 0 in the Northern Hemisphere), become colocated, such that the divergence in the
upper troposphere and the ascent below are enhanced.
This coupling of the vertical circulations can lead to
the organization of heavy precipitation and to the
rapid development of the surface cyclone.
The vertical circulations associated with upper-
tropospheric jet streaks can also interact with the
corresponding circulations associated with lower-
tropospheric jets and frontal zones. A hypothetical
example depicting unfavorable and favorable config-
urations of upper-level and lower-level systems is
provided in Figure 8. In the unfavorable configuration
Figure 6 Schematic illustration of geopotential height (solid
(Figures 8A,B), the right exit region of an upper-
lines), wind speed (dashed lines: region of maximum wind speed tropospheric jet streak is located above a surface
shaded), ageostrophic wind (arrows) and associated patterns of frontal zone. The subsidence associated with the jet
convergence(C0N) anddivergence(D1V)in thevicinityof acurved streak in this region overlies the ascent associated with
jet stream exhibiting uniform wind speed in the along-stream the surface frontal zone, acting to stabilize the air
direction. (Adapted from Shapiro MA and Kennedy PJ (1981)
Research aircraft measurements of jet stream geostrophic and above the front and to suppress convection. In the
ageostrophic winds. Journal of the Atmospheric Sciences 38: favorable configuration (Figures 8C,D), the left exit
2642-2652.American Meteorological Society, Boston.) region of the jet streak is located above the surface
1050 JET STREAKS
Figure 7 Schematic illustration of surface cold and warm fronts, high and low pressure centers, sea level isobars (dotted lines),
precipitation (shaded, with asterisks representing snow and dots representing rain), upper-level flow (arrows), upper-level trough axes
(dot-dashed lines), and jet streaks (cross-hatched) associated with a 'typical' heavy snow event along the East Coast of the United States.
(From Uccellini LW and Kocin PJ (1987) The interaction of jet streak circulations during heavy snow events along the East Coast of the
United States. Weather and Forecasting 2: 289-308. American Meteorological Society, Boston.)
frontal zone, and the ascent associated with the jet downstream of the trough (Figure 9D).Although this
streak can couple with the ascent accompanying the scenario is seldom observed in its entirety, it is
surface front, leading to the release of convective common for an evolving jet streak to progress through
instability and to the outbreak of severe convective at least one or several of the stages depicted in Figure 9
storms. during some portion of its life cycle.
Jet Streak Life Cycles

The foregoing conceptual models treat jet streaks as
Dynamics of Jet Streaks
static entities in the sense that their structure and The conceptual models of jet streaks presented above
amplitude remain steady, although they can translate are based on hypothetical configurations of wind,
or propagate. In reality, jet streaks undergo life cycles geopotential height, and potential temperature fields,
while interacting with the background-flow environ- and on deductions based on air parcel arguments.
ment. A hypothetical jet streak life cycle is illustrated However, it is not immediately obvious that these
in Figure 9, in which the jet streak is traveling through configurations correspond to realizable solutions,
and interacting with an evolving synoptic-scale baro- obtained using either analytical or numerical methods,
clinic wave. In this scenario, the jet streak forms in the to consistent sets of dynamical equations governing
confluence zone situated within a trough-over-ridge the evolution of the flow. To describe and understand
pattern (Figure9A). As the jet streak migrates into the the processes by which jet streaks evolve, it is desirable
north-westerly flow inflection between the ridge and to refer to such solutions. In the balanced case, the
the trough (Figure 9B), the baroclinic wave amplifies. dynamics of jet streaks may be described in terms of
The jet streak subsequently advances toward the base linear waves or nonlinear coherent vortices using the
of the trough (Figure 9C), at which time the wave is quasi-geostrophic system, whereas in the unbalanced
fully developed and after which time the wave decays case, the dynamics of jet streaks are strongly coupled
as the streak travels through the south-westerly flow to inertia-gravity waves.
JET STREAKS 1051
-1 00 m
k, ' ' *ids B
,
Figure 8 Schematic illustration of vertically uncoupled (A,B) and vertically coupled (C,D) upper- and lower-tropospheric jet-front
systems. (A) Upper-troposphericjet exit region (isotachs, heavy solid lines: jet axis, solid arrow) aligned along and displaced toward cold
side of surface frontal zone (isentropes,dashed lines; cold front marked conventionally) and lower-troposphericjet (jet axis, open arrow).
(B)Cross-section along A-A' indicated in (A), depicting upper- and lower-tropospheric jets (isotachs, thick dashed lines), upper-and
lower-troposphericfrontal zones (bounded by thin solid lines), tropopause (double solid lines), moist boundary layer (shading) capped by
lid, and streamlines of transverse ageostrophic circulation (solid arrows, strength of circulation proportionalto width). (C) As in (A) except
for upper-troposphericjet exit region aligned across surface frontal zone. (D) As in (B) except for cross-section along B-B' indicated in (C).
(Adapted from Shapiro (1982).)
Jet Streaks and Baroclinic Waves ing downstream development is shown in Figure 10.
The evolution of the jet streaks in this process is
It was noted previously that jet streaks often play a
consistent with the observation that many streaks
significant role in extratropical cyclogenesis. In fact,
appear to 'jump' between the north-westerly and
there is evidence suggesting that jet streaks are an
south-westerly flow inflections of the wave, rather
integral part of the upper-tropospheric baroclinic
than translate continuously through the wave pattern,
wave associated with cyclogenesis and of a dynamical
as is the case for the conceptual life cycle shown in
process known as 'downstream development', which
Figure 9.
accounts for the growth and decay of many such
waves. In the downstream development process,
Jet Streaks and Coherent Vortices
kinetic energy propagates from an upstream trough
that is decaying toward a new trough that is growing Observations suggest that, in some cases, jet streaks
downstream. This energy is maximized in the inflec- are associated with monopolar and dipolar vortices of
tions between the troughs and ridges, and these kinetic mesoscale dimensions (length scales of approximately
energy maxima correspond to jet streaks. An idealized 500 km) that are embedded within a larger-scale jet
picture of the structure of a baroclinic wave undergo- stream. These features differ from the synoptic-scale
1052 JET STREAKS
:..c . ........:;.
2'
'2:
:. *q
...........
.... ...:.'...
I
'.; : ;
... :... :';.
1 ; '.
i.,;, :. ......................... .. ..
.................
.. . .. .. ... '..,.
. . . .'.*. . :
,
. :
(A)
.....
-I .... I
-
... :
I.. ..... ....i : ..........
,. ..
.r:
A,;.
.....
........ .:
Figure 9 Schematic illustration of geopotential height (heavy solid lines), wind speed (heavy dashed lines), and potential temperature
(thin dashed lines) associated with the migration of an upper-troposphericjet-front system through asynoptic-scalebaroclinicwaveover a
72 h period. (A) Jet-front forming in the confluence zone between middle- and polar-latitude currents. (B) Jet-front situated in the north-
westerly flow inflection of amplifying wave. (C) Jet-front at the base of the trough of fully developed wave. (D) Jet-front situated in the
south-westerly flow inflection of decaying wave. (From Shapiro (1982).)
JET STREAKS 1053
Super geostrophic that are large compared with unity, and by large values
of horizontal divergence and Lagrangian rates of
* Q2
change of this quantity. A well-documented example
Ly maxima /I m
$50 Developing
Exist-
upstream
trough
- Sub geostrophic
downstream
trough
Figure 10 Schematic illustration of the relationship between the

components of a synoptic-scale baroclinic wave. The upper-level
geopotential field is indicated by two contours labeled Qi and Qz.
The geopotential anomaly, d’,relative to the time mean is positive/
negative in the ridgehrough. Wave-relative air flow is indicated by
heavy solid arrows and the ageostrophic wind is indicated by open
arrows. Centers of maximum vertically integrated eddy kinetic
energy are shown as ellipses. (From Orlanski I and Sheldon JP
(1995) Stages in the energetics of baroclinic systems. Tellus 47A:
605-628. Munksgaard International Publishers, Copenhagen.)
baroclinic waves described above in that they do not

disperse their energy; instead they travel coherently,
maintaining their identities for periods of up to several
weeks. A straight jet streak represented in terms of a
symmetric vortex dipole solution to the quasi-geo-
strophic system is depicted schematically in Figure 11;
this schematic may be compared with its observation-
ally based counterpart in Figure 4. In the representa-
tion depicted in Figure 11, the flow associated with the
cyclonic vortex advects the anticyclonic vortex and
vice versa, while the combined flow of both vortices
results in a localized wind speed maximum (i.e., a jet
streak) in the region between the vortices (Figure 11A).
Hence the jet streak moves downstream in conjunc-
tion with the dipole, but at a slower speed than the
maximum wind speed in the jet core, consistent with
observations. The rotational component of the ageo-
strophic flow (Figure l l B ) , which dominates the total
ageostrophic flow in the vortex dipole solution,
exhibits a four-gyre circulation pattern consistent
with the total ageostrophic flow pattern depicted in
Figure 3. The divergent component of the ageostroph-
ic flow (Figure 11C) is associated with a four-quadrant Figure 11 Schematic illustration of a straight jet streak (region of
pattern of horizontal divergence consistent with that maximum wind speed shaded) represented in terms of a quasi-
depicted in Figure 4A. geostrophic vortex dipole in a uniform zonal background flow. (A)
Quasi-geostrophic potential vorticity, q, and geostrophic velocities
associated with respective constituent vortices, such that solid
UnbalancedFlow in Jet Streaks arrows indicate flow induced by cyclonic vortex ( q > 0, solid line)
and dotted arrows indicate flow induced by anticyclonic vortex
The foregoing discussion of the dynamics of jet streaks cas (cas
( q < O , dotted line). (6)Ageostrophic vorticity, > 0, solid
is predicated on the assumption that the flow satisfies line; Lag t o , dotted line), and induced rotational ageostrophic
velocity given by solid arrows. (C) Horizontal divergence 6 (6 > 0,
some balance relationship connecting the wind and
solid line; d<O, dotted line) and induced divergent ageostrophic
geopotential fields. In some cases, however, jet streaks velocity given by solid arrows. Vector scales of arrows are
do not satisfy any particular balance relationship; such equivalent in (6)and (C), and an order of magnitude smaller
jet streaks are often characterized by Rossby numbers than in (A).
1054 JET STREAKS
of an unbalanced flow configuration is that of an ie vertical component of relative

upper-tropospheric jet streak located in the south- vorticity evaluated on an isen-
westerly flow inflection between a trough and ridge in tropic surface (s - '1
the circumstance where the distance between the 6 potential temperature (K)
trough and ridge is contracting. It has been hypothe- P air density (kg m - 3,
sized that this upper-tropospheric configuration is 4' geopotential anomaly (m2s - 2 )
conducive to the generation of large-amplitude iner- @ geopotential (m2s - 2 ,
tia-gravity waves in the lower troposphere, which may w vertical velocity in pressure co-
be accompanied by hazardous weather. Moreover, ordinates (Pa s - ')
inertia-gravity waves associated with this configura- V horizontal gradient operator
tion may be important in the redistribution of evaluated on a n isobaric sur-
momentum and energy between the troposphere and face ( m - ' )
the stratosphere and mesosphere above, and as such v.v horizontal divergence evalua-
may play a significant role in the general circulation of ted on an isobaric surface (s -')
the middle atmosphere.
Nomenclature See also

Buoyancy and Buoyancy Waves: Theory. Clear Air
f Coriolis parameter (s-') Turbulence. Cyclogenesis. Cyclones, Extra Tropi-
g acceleration due t o gravity cal. Dynamic Meteorology: Overview. Fronts. Meso-
(ms-2) scale Meteorology: Mesoscale Convective Systems;
k unit vector in vertical direction Models. Quasi-geostrophic Theory. Severe Storms.
K, parcel trajectory curvature Stratosphere-Troposphere Exchange: Global As-
(m-7 pects; Local Processes. Synoptic Meteorology: Fore-
L horizontal length scale ( m ) casting; Weather Maps. Tropopause.
n unit vector normal to horizon-
tal velocity vector
pressure (hPa) Further Reading
potential vorticity (m2s - ' Bjerknes J (1951) Extratropical cyclones. In: Malone TF
Kkg-') (ed.) Compendium of Meteorology, pp. 577-598.
quasi-geostrophic potential Boston: American Meteorological Society.
vorticity (s - '1 Bluestein HB (1986) Fronts and jet streaks: a theoretical
Ri = (g/e)(aO/az)/ Richardson number perspective. In: Ray PS (ed.) Mesoscale Meteorology and
1 av/azl Forecasting, pp. 173-215. Boston: American Meteoro-
RO = V / f L Rossby number logical Society.
S unit vector parallel to horizon- Bluestein HB (1993) Synoptic-Dynamic Meteorology in
tal velocity vector Midlatitudes. Volume II: Observations and Theory of
t time (s) Weather Systems. New York: Oxford University Press.
V horizontal wind speed ( m s - ') Carlson TN (1991) Mid-Latitude Weather Systems.
London: HarperCollins Academic.
V horizontal wind velocity ( m s - ')
Holton JR, Haynes PH, McIntyre ME, et al. (1995)
v, geostrophic wind velocity Stratosphere-troposphere exchange. Reviews of Geo-
(ms-') physics 33(4):403-439.
Vag = V - Vg ageostrophic wind velocity Keyser D (1986) Atmospheric fronts: an observational
(ms-') perspective. In: Ray PS (ed.) Mesoscale Meteorology
W vertical velocity in height coor- and Forecasting, pp. 216-258. Boston: American Mete-
dinates (m s - I) orological Society.
height coordinate ( m ) Keyser D and Shapiro MA (1986)A review of the structure
horizontal divergence (s - ') and dynamics of upper-level frontal zones. Monthly
vertical component of relative Weather Review 114: 452-499.
vorticity evaluated on an iso- Kocin PJ and Uccellini LW (1990) Snowstorms along the
northeastern coast of the United States: 1955 to 1985.
baric surface (s - I)
Meteorological Monographs, vol. 22, no. 44. Boston:
ageostrophic vorticity, defined American Meteorological Society.
as the vertical component of Palmhn E and Newton CW (1969)Atmospheric Circulation
relative vorticity of the ageo- Systems: Their Structure and Physical Interpretation.
strophic wind velocity (s-') New York: Academic Press.
JET STREAKS 1055
Riehl H and collaborators (1952) Forecasting in middle Volume, pp. 167-191. Boston: American Meteorological
latitudes. Meteorological Monographs, vol. 1, no. 5. Society.
Boston: American Meteorological Society. Uccellini LW (1990) Processes contributing to the rapid
Shapiro MA (1982) Mesoscale Weather Systems of the development of extratropical cyclones. In: Newton C W
Central United States. University of Colorado, Boulder: and Holopainen EO (eds) Extratropical Cyclones, The
Cooperative Institute for Research in Environmental Erik Palmin Memorial Volume, pp. 81-105. Boston:
Sciences (CIRES)/National Oceanographic and Atmos- American Meteorological Society.
pheric Administration (NOAA). Uccellini LW and Koch SE (1987) The synoptic setting and
Shapiro MA and Keyser D (1990)Fronts, jet streams and the possible energy sources for mesoscale wave disturbances.
tropopause. In: Newton CW and Holopainen EO (eds) Monthly Weather Review 115: 721-729.
Extratropical Cyclones, The Erik Palme'n Memorial
KATABATICWINDS 1057
T R Parish, University of Wyoming, Laramie, WY, USA Katabatic winds occur most frequently during
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. nocturnal conditions, and encompass a wide range
of time and distance scales. Cold air drainage flows
along the sides of valleys having a distance scale on the
Introduction order of 1km and the broad continental-scale gravity
flows over the great ice sheets of Antarctica and
Katabatic winds are a class of atmospheric motion in Greenland are each classified as katabatic. The
which air flow is directed down a topographic incline following discussion will focus on the cold, dense
such as a mountainside or glacier. The term ‘katabatic’ downslope wind features for which the term katabatic
is Greek in origin from katabatikos, meaning to go is most commonly applied.
down, and has been used in meteorological literature
since World War I. Reference to katabatic winds is
reserved for air motion in the lower atmosphere,
extending from the surface to a height typically below
Dynamics of Katabatic Winds
1km. Topography is the controlling factor for both the
In the most general sense, any wind blowing down intensity and the direction of katabatic winds. This is
an incline can be classified as a katabatic wind. The illustrated schematically in Figure 1,which represents
broad definition has allowed a host of local wind the atmosphere as a simple two-layer fluid. Although
features to be defined as katabatic. As air moves highly idealized, the layer model approach in Figure 1
downslope, it is subject to adiabatic warming due to can explain many of the characteristics of katabatic
compression since the pressure increases during the winds. Each layer has a constant density. The lower
descent. Two general categories of downslope winds layer represents the cold, dense layer nearest the Earth
exist, distinguished from each other by the relative and the upper layer represents the atmosphere undis-
temperature of the air stream. A wind that is warmer turbed by the cooling process. A negatively buoyant
than the air being displaced along and at the bottom of lower layer most commonly develops at night when
an incline is referred to as a ‘foehn’. This term has its the sky is clear as strong radiative cooling of the
origin in the warm winds descending the slopes of the inclined surface takes place. Horizontal pressure
Alps in Europe. Similar downslope winds east of the differences are the fundamental cause for all atmos-
Rocky Mountains of North America have been called pheric motions. In the case of katabatic winds the
‘Chinooks’, which is taken from the Native American horizontal pressure gradient force is proportional to
word meaning ‘snow eater’ to describe effects of such the difference in pressure between the cooled air near
wind events. The occurrence of these winds requires the surface ( p l ) and at the same horizontal level in the
particular weather conditions in the ambient atmos- ambient atmosphere overlying the cooled layer ( p 2 ) .
phere above the downslope flow that direct the air Through the hydrostatic relationship, the vertical
motion. change in pressure is dependent on the air density
Use of the term ‘katabatic’ to describe warm wind and the thickness of the air column. For a given height
features was common throughout much of the early in a vertical column, pressure decreases more rapidly
and middle part of the twentieth century, but has with height for cold air than for warm air. Applying the
decreased over the last few decades. Discussion of hydrostatic principle, p 1 will be greater than p 2 if no
katabatic winds now invariably refers to the second horizontal pressure difference exists above the cooled
type of downslope flow, which is a cold wind. Some air layer, and thus horizontal acceleration will occur in a
streams originate over high plateaus that are subjected downslope sense. By use of the hydrostatic equation,
to intense cooling, such as over the interior of the the horizontal pressure gradient ( P ) in the lower layer
Antarctic or Greenland ice sheets. These air currents can be defined as in eqn [l].
remain colder than the surrounding environment
during the descent despite compressional heating. p=- A p aH 1 ap
Gravity is the driving force for such flows since the [11
P1 as P 1 astop
density of the cold air is greater than that of the air it
displaces. The cold downslope winds are dependent on In eqn [l],g ( = 9.81msP2) is the acceleration of
the local terrain slope characteristics and relative gravity, A p is the density difference (pl - p 2 ) between
density of the air stream, although weather conditions the bottom and top layers, p1 is the density in the
in the ambient atmosphere are often important as well. bottom layer, H is the height of the lower layer above
1058 KATABATIC WINDS
the depth of the cooled lower layer.
The first term in eqn [3] represents the effect of a

density contrast over sloping terrain. It can be seen
that the size of this term is dependent on the magnitude
of the density contrast and the slope of the terrain. This
term is sometimes called the buoyancy force. The
second term represents the influence of the gradient of
thickness of the cooled layer column along the slope.
For a situation in which the thickness of the lower
layer decreases, as is shown schematically in Figure 1,
Figure 1 Cross-sectional view of two-layer fluid representative of
the cold, katabatic wind layer (lower layer) beneath the undisturbed
this term will act to accelerate the downslope compo-
ambient atmosphere; p designates pressure, p designates density nent of the flow. For cases in which the horizontal
( p , > p 2 ) , H is the height of the katabatic layer above a fixed scales are larger than a few kilometers, the second term
reference, h is the depth of the katabatic layer above the ground, is generally smaller than the first term and can be
and z is the terrain height above a fixed reference. neglected. This simplification can also be made if it is
assumed that the inversion interface follows the
terrain. Both remaining terms, the buoyancy force
some fixed reference level, and s is a horizontal length and the horizontal pressure gradient force in the
scale. Two terms act to force atmospheric motion in ambient atmosphere, are important in the forcing of
the lower layer. The first term represents the effects of katabatic winds. Although developed for the two-
cooling over the sloping surface to produce a nega- layer model shown in Figure 1,eqn [3] can be applied
tively buoyant air stream. The second term depicts the directly to the atmosphere if potential temperature is
pressure gradient force in the atmosphere above the substituted for temperature to incorporate effects of
lower layer in the horizontal direction along the compressibility.
terrain slope. It is thus evident that the horizontal Figure 2 is a nomogram illustrating the relationship
pressure gradient force in the atmosphere above the between the inversion strength, terrain slope and
cooled layer near the surface is transferred downward ambient pressure gradient. It can be shown that a
through the fluid. temperature inversion over even gentle slopes can
Using the equation of state, eqn [l]may be rewritten produce significant accelerations. Typical horizontal
in a more convenient form as eqn [2]. pressure gradients in the ambient atmosphere are on
the order of 100Pa over a horizontal distance of
100 km, which is equivalent to a geostrophic wind of
approximately 8 m s - in middle latitudes. As can be
seen from Figure 2 , this is the same magnitude as the
Here AT is the temperature difference between the two buoyancy force arising from a lower layer that is
layers (TI - T2) and TI is the temperature in the diabatically cooled by 10°C over an incline of
bottom layer. In this form, AT can be viewed as the approximately M O O . For steeper inclines with modest
strength of the temperature inversion between the temperature inversions, the buoyancy force can be
cooled layer and the ambient overlying atmosphere. In seen to be considerably larger than typical values of the
ordinary atmospheric situations, temperatures de- horizontal pressure gradient force in the ambient
crease with height. Under conditions of a temperature atmosphere. Because of the large magnitude of the
inversion, such as occurs during nocturnal conditions, buoyancy force and stable stratification of the atmos-
the coldest air is situated closest to the ground and phere, katabatic winds respond to the local topo-
temperatures increase with height. The strength of the graphic configuration, often regardless of the pressure
temperature inversion can be determined from the gradient in the free atmosphere.
measurement of temperature in the vertical direction, Newton’s second law governs the response of a free
known as a sounding. body to external forces acting on that body. In the case
Recognizing that the height H of the density of the Earth’s atmosphere, which rotates following the
interface is just the sum of the terrain height and planetary angular velocity, it is advantageous to write
thickness of the lower layer, P can also be expressed as Newton’s second law in a coordinate system fixed to
in eqn [3], where z is the height of the terrain and h is the Earth. A common method is to express the
KATABATIC WINDS 1059
wind (m s-') (Pdloo km)

2 --
0.0001 - - 20
strength
(K) 0.0002 - -- 30
30 --
25 --
0.0005 - 4 -- 50
20
15
--
-- 0.001
7 -- 90
;:
-1-- 80
lo :I
--
0.002 - 8 -- 100
8 9 -
6 -- 0.005 - 10 --
5 -- .- 150
--
4 0.01 =
3 --
l5 :-200
0.02 - 20 r
2 --
25 i- 300
1.5 -- 0.03 1
30 -- 400
1 -- 0.1 =
40 -- 500
50 -- 6oo
-1 igg
0.5 - 60
70 -
Figure 2 Nomogram showing the relationship between inversion strength, terrain slope and the horizontal pressure gradient force.
Extending a straight line between the inversion strength and terrain slope provides a measure of the pressure gradient force associated
with katabatic winds. Calculations assume a katabatic layer potential temperature of 273 K; geostrophic wind magnitude is valid at 43' N.
acceleration of an air parcel in terms of the specific rotation cannot be neglected, air streams are deflected
forces (force divided by mass) to form the equation of across the terrain gradient, often at angles exceeding
motion. Atmospheric forces acting to accelerate air 45"from downslope. These conditions are found over
include the pressure gradient, gravity, and friction. the great continental ice sheets of Antarctica and
The Coriolis force is an apparent force arising from the Greenland. Observations show that the katabatic
rotation of the Earth. The horizontal vector equation wind is deflected at angles across the fall line vector
of motion for katabatic flow relative to the rotating of the terrain that are consistent with effects of the
Earth can be expressed as eqn [4]. Coriolis force.
DV Geographical Distribution of Katabatic

-= P - f k x V + F [41
Dt Winds
Here, D/Dt represents the time rate of change Every continent on Earth experiences katabatic winds.
following a fluid parcel, V is the horizontal velocity, Occurrences are most evident during nocturnal con-
tis time, f is the Coriolis parameter (2R sin 4, where R ditions. Along the highlands and slopes of mountain-
is the Earth's angular velocity of 7.29 x 10 - 5 rad s ~ ', ous terrain, rapid cooling of the surface takes place
and 4 is the latitude), and F is the frictional drag. For after sunset during periods of fair weather as the
small horizontal scales of motion of a few kilometers Earth's surface radiates energy that escapes to space.
or time scales on the order of a few minutes, the effect Air in contact with the ground also becomes cooled
of the Earth's rotation is insignificant and acceleration and hence becomes denser than air at the same
of katabatic flow is subject to the pressure gradient horizontal level situated away from the terrain. The
force P and friction. Under these conditions, the air cold, dense air then begins to move downslope.
accelerates directly downslope, in absence of strong Similarly, radiative cooling of air over an elevated
forcing from the ambient environment. For larger plateau, especially one that is covered by snow, results
scales of motion on the order of hundreds of kilome- in the development of a pool of cold air, and a shallow
ters or time scales on the order of a few hours, the dome of high pressure becomes established. Addition-
Coriolis force needs to be considered. Acceleration al accumulation of cold air or passage of extratropical
provided by the Coriolis force is directed to the right cyclones may act to trigger the release of cold air
(left) of the wind vector in the Northern (Southern) through mountain gaps or valleys in the form of
Hemisphere. For katabatic winds in which the Earth's katabatic winds.
1060 KATABATIC WINDS
Descriptive names for katabatic wind events have the scalar equations of motion from eqn [4] can be
been given to certain episodes that produce profound written as eqns [5] and [6].
changes in the local weather or present peril to
agricultural or other commercial activities. Nearly k
O=P-fv- -vu 151
all are the combined result of both cooling of the air h
stream and the horizontal pressure gradients in the
ambient atmosphere. The mistral is a cold, northerly k
O=fu--vv
wind that originates in the elevated regions of the Alps, h
and descends into the Rhone valley and other low-
Equation [5] is the component along the direction of
lying regions along the south of France. It is best
the horizontal pressure gradient force P, which is
developed when low pressure forms in the Gulf of
typically in the downslope direction. Equation [6] is
Genoa. The cold air often results in frost damage to
the equation of motion perpendicular to P. Again, Vis
vineyards. Bora is a name originally used t o describe
the magnitude of the wind; u = V cos p and is directed
the cold north-east winds observed along the coast of
along P ; v = V sin /3 and is directed to the left of and
the former Yugoslavia. This air stream originates over
normal to P. The term fi is the angle between P and V .
Russia and crosses the Carpathian and Alps mountain
ranges and descends onto the usually warm shores of The force balance arising from eqns [ 5 ] and [6] to
produce the steady Antarctic katabatic wind is shown
the Adriatic Sea. Bora cases are associated with a
schematically in Figure 3. Equations [5] and [6] can
pressure field consisting of a cyclone near the Black Sea
then be solved for V and /3 (eqn [ 7 ]provided
) estimates
and an anticyclone over the European continent. Both
mistral and bora are strong and gusty winds with of the strength and depth of the temperature inversion,
maximum frequency of occurrence during winter. terrain slope, and pressure gradient force in the
ambient environment are known.
Other local katabatic winds include the oroshi, a cold
wind found on the Pacific side of the mountains on the
island of Japan and the coho, a westward surge of cold v = / T = f s i n p 171
air off the elevated plateau regions through the
Columbia Gorge along the north-west coast of the For a given terrain slope, increasing values of P are
United States and southern British Columbia. Kata- accompanied by an increase in V and a decrease in p,
batic winds are not restricted to middle and high and hence the flow is faster and directed more
latitudes. Cold drainage flows known as tehuantepe- downslope. The observed behavior of Antarctic
cers occur in the Gulf of Tehuantepec along the pacific katabatic winds qualitatively matches that prescribed
coast of Mexico. Katabatic winds have also been by this simplified model. Assessment of the steady-
documented in South America along the eastern slopes state wintertime Antarctic katabatic wind regime from
of the Andes from northern Chile to central Colombia. eqn [7] is illustrated by streamlines in Figure 4. The
Katabatic winds are best-developed over the great force balance between pressure gradient, Coriolis
ice sheets of Antarctica and Greenland. The geograph-
ical position of the ice sheets and high reflectivity of the
ice fields combine to limit the heating of the surface.
Temperature inversions in the lowest 100 m over the P
elevated interior sections of Antarctica commonly 2-A2
exceed 25°C during nonsummer months. Observa-
tions show the surface winds t o be consistently of a
V I
katabatic nature for much of the year. Surface winds
over Antarctic ice slopes are nearly unidirectional,
with approximately 90% of the observations from a
30" sector. The katabatic wind systems over Antarc-
tica and Greenland are among the most persistent on
Earth.
To a first approximation, the near-steady flows over
C / \4F
z+Az
the ice sheets can be envisaged as a balance between
the horizontal pressure gradient force, the Coriolis Figure 3 Horizontal view of force balance in the lower atmos-
force, and the friction force. A simplified but useful phere for katabatic winds over Antarctica; z designates the terrain
height, V is the katabatic wind vector, /? is the angle between the
approximation for the friction force is F = -k VV/h, wind and the downslope direction of the terrain, P is the horizontal
where k is a dimensionless constant having a magni- pressure gradient force associated with atmospheric cooling over
tude between and For steady conditions, sloping terrain, C is the Coriolis force, and F is the friction force.
KATABATIC WINDS 1061
Figure 4 Streamlines of wintertime katabatic winds over Antarctica. Terrain contours (units of kilometers) of Antarctic ice sheet denoted
by thin lines.
force, and friction results in a katabatic wind vector See also

that is directed across the height contours of the terrain Antarctic Climate. Boundary Layers: Stably Stratified
at angles of 30-60" over the interior of the continent. Boundary Layer. Cold Air Damming. Coriolis Force.
Winds near the coast are directed more downslope. Downslope Winds. Dynamic Meteorology: Overview.
A counterclockwise, divergent katabatic wind LandSea Breeze. Mountain Meteorology. Valley
pattern is present over the Antarctic ice sheet. A Winds.
similar pattern can be found over Greenland, except
that the motion is in a clockwise sense. By mass
continuity, sinking motions must exist over the great Further Reading
ice sheets. Avertical circulation becomes established in Atkinson BW (1981)Meso-Scale Atmospheric Circulations.
the troposphere above the katabatic layer in response New York: Academic Press.
to the continental outflow. The strongest winds are Barry RG (1981) Mountain Weather and Climate. London:
found over the steepest ice slopes, which are generally Methuen.
near the coast. King J C and Turner J (1997) Antarctic Meteovology and
Antarctic katabatic winds are not uniform, as can be Climatology. Cambridge: Cambridge University Press.
seen in Figure 4.Convergence of drainage streamlines Schwerdtfeger W (1970) The climate of the Antarctic. In:
Orvig S (ed.) World Survey of Climatology. vol. XIV,
is present at various locations about the periphery of
Amsterdam: Elsevier.
Antarctica. Cold, negatively buoyant air from a broad Schwerdtfeger W (1984) Weather and Climate of the
horizontal area becomes concentrated into a restricted Antarctic. Amsterdam: Elsevier.
pathway. These 'confluence zones' represent areas Simpson JE (1994)Sea Breeze andLocal Winds. Cambridge:
of enhanced cold air supplies available to feed kata- Cambridge University Press.
batic winds downstream. The most intense and Yoshino M M (ed.) (1976) Local W i n d Bora. Tokyo:
persistent katabatic winds are observed along and University of Tokyo Press.
downstream of the axes of the confluence zones Whiteman CD (2000) Mountain Meteorology. Oxford:
depicted in Figure 4. Oxford University Press.
1062 KELVINWAVES
B Wang, University of Hawaii, Honolulu, HI, USA and maintenance of the Madden-Julian Oscillation.
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. Oceanic Kelvin waves play a critical role in tidal
motion, in the adjustment of the tropical ocean to
wind stress forcing, and in generating and sustaining
Introduction the El Nifio Southern Oscillation.
The Kelvin wave is a large-scale wave motion of great
practical importance in the Earth’s atmosphere and Boundary-Trapped Kelvin Waves
ocean. Discovered by Sir William Thompson (who
later became Lord Kelvin) in 1879, the Kelvin wave is a Surface Kelvin Waves
special type of gravity wave that is affected by the
Earth’s rotation and trapped at the Equator or along The mechanism and properties of the Kelvin wave can
lateral vertical boundaries such as coastlines or be illustrated by considering a horizontally propagat-
mountain ranges. The existence of the Kelvin wave ing Kelvin wave in a rotating fluid of uniform finite
relies on ( a ) gravity and stable stratification for depth H , where H is small compared to the horizontal
sustaining a gravitational oscillation, (b) significant extent of the fluid. The fluid has homogeneous density
Coriolis acceleration, and (c) the presence of vertical and a free surface, and is confined by a vertical lateral
boundaries or the equator. An important feature of the boundary. Such an idealized model is referred to in
Kelvin wave is its unidirectional propagation. The geophysical fluid dynamics as a shallow water model.
Kelvin wave moves equatorward along a western The lateral bounding wall prohibits flow across the
boundary, poleward along an eastern boundary, and boundary, and this absence of transverse motion with
cyclonically around a closed boundary (counterclock- respect to the lateral boundary is a defining character-
wise in the Northern Hemisphere and clockwise in the istic of Kelvin waves (Figure 1). Fluid parcels (ele-
Southern Hemisphere). The wave amplitude is largest ments) are constrained to move in a vertical plane
at the boundary and decays exponentially with
distance from it. At the Equator, Kelvin waves always
propagate eastward, reaching their maximum magni-
tude at the Equator and decaying exponentially with
increasing latitude.
There are two basic types of Kelvin waves: bound-
ary trapped and equatorially trapped. Each type of
Kelvin wave may be further subdivided into surface
and internal Kelvin waves. Surface, or barotropic,
waves penetrate the entire depth of the fluid. Kelvin
waves also appear within the stably stratified ocean
and atmosphere, and are called internal, or baroclinic,
Kelvin waves. Internal Kelvin waves are often found in
a layer with large density gradients; the density
gradient acts as an interface that allows the existence
of internal gravity waves. Examples of such density
gradients are the oceanic thermocline (a layer of large
vertical temperature gradient separating a shallow
layer of warm surface water about 50-200 m deep and
a much deeper layer of cold water below) and the Figure 1 Northern hemisphere Kelvin waves on opposite sides
of a channel that is wide compared with the Rossby radius. In each
lower edge of an atmospheric inversion, a layer in vertical plane parallel to the coast, the currents (shown by arrows)
which temperature increases with height. Like gravity are entirely within the plane and are exactly the same as those for a
waves, Kelvin waves can also propagate vertically in a long gravity wave in a nonrotating channel. However, the surface
continuously stratified geophysical fluid. elevation varies exponentially with distance from the coast in order
Atmospheric Kelvin waves play an important role to give a geostrophic balance. This means Kelvin waves move with
the coast on their right in the Northern Hemisphere and on their left
in the adjustment of the tropical atmosphere to con- in the Southern Hemisphere. (Reproduced with permission from
vective latent heat release, in the stratospheric Gill AE (1982) Atmosphere-Ocean Dynamics. New York: Aca-
quasi-biennial oscillation, and in the generation demic Press.)
KELVIN WAVES 1063
~~~
parallel to the lateral boundary in the neighborhood of Internal Kelvin Waves

the boundary. Thus, the horizontal longshore (along
The horizontally propagating internal, or baroclinic,
the boundary) component of the Coriolis force must
Kelvin wave behaves in the same manner as the surface
vanish. Consequently, the wave motion at the lateral
wave except that the motion varies with depth. Most
boundary, and in any parallel vertical plane, is exactly
frequently, it occurs at a stable interface (or a thin layer
the same as a hydrostatic gravity wave in a nonrotating
with large vertical density gradients) separating two
system, i.e., the shallow water gravity wave (Figure 1).
relatively homogeneous layers. For internal Kelvin
The wave travels along the boundary with the shallow
waves, the pressure gradient force normal to the lateral
water gravity wave speed C determined by the square
boundary arises from the tilt of the interface and is
root of the product of the gravitational acceleration
balanced by the Coriolis force associated with the
(g) and the depth of the fluid, C = (gH)lI2.The shape
vertical differential flow parallel to the boundary. The
of the wave in the longshore direction is arbitrary and
internal Kelvin wave speed depends on the density
is conserved as the wave travels. This implies that the
difference across the interface and is normally much
surface Kelvin wave is nondispersive, and that the
slower than that of surface Kelvin waves. In the ocean,
wave energy is transmitted at the speed of the shallow
the typical speed for internal coastal Kelvin waves is
water gravity wave. Because the Kelvin wave solution
of the order of 1m s - and the Rossby radius of
in any vertical plane parallel to the lateral boundary is
deformation is of the order of lOkm in the mid-
identical to that of the nonrotating case, the energy of a
latitudes. Evidence for coastal Kelvin wave propaga-
Kelvin wave is partitioned equally between kinetic and
tion along the eastern boundary of the Pacific has been
potential energy.
observed in coastal sea level and temperature records.
A fundamental difference between the Kelvin wave
In the atmosphere, boundary trapped Kelvin waves
and the two-dimensional gravity wave is that the
occur primarily in the form of internal waves. They are
Kelvin wave can propagate in only one direction,
often found along the edge of a plateau or a mountain
rather than in two opposite directions. This is due to
range, such as the coast of South Africa, the west coast
the constraints of the Earth’s rotation and the presence
of California, and the eastern flank of the Tibetan
of the lateral boundary. Rotation modifies the flow by
Plateau. These internal Kelvin waves are created near a
piling up fluid against the lateral boundary, producing
stable inversion layer (often located at the top of the
an offshore (normal to the boundary) pressure gradi-
boundary layer) against steep topography. The elevat-
ent force associated with the surface elevation. Since
ed plateau or mountain range rises above the inver-
offshore motion is prohibited by the presence of the
sion, forming a lateral boundary for the air of the
boundary, the offshore pressure gradient force must
lower layer. Energy is prevented from escaping verti-
balance the Coriolis force associated with the long-
cally by the inversion and prevented from escaping
shore flow. For this reason, the motion is referred to as
laterally by the topography. The maximum distur-
semigeostrophic (Le., geostrophic balance is reached
bance intensity of these waves is found near the coast
in only one direction). A direct consequence of this
or plateau and decreases exponentially in intensity
geostrophic balance is the exponential decay of the
away from the coast or plateau. The offshore extent of
longshore velocity and surface height with distance
the waves depends on both the thermal structure and
from the lateral boundary (Figure 1 ) .The Kelvin wave
topography. When the topography is steep, the depth
amplitude is significant only within an e-folding
of the inversion determines the Rossby radius for the
distance of the order of the Rossby radius of defor-
atmospheric Kelvin wave. A typical value for atmos-
mation ( R )from the lateral boundary. This important
pheric internal Kelvin waves is on an order of
length scale is defined by the ratio of the gravity wave
1000 km.
speed, C, over the absolute value of the Coriolis
parameter (f).Over this characteristic distance, the
tendency of the gravitational force to flatten the
free surface is balanced by the tendency of the
Equatorially Trapped Kelvin Waves
Coriolis force to deform the surface. This is possible Matsuno in 1966 showed that the eastward-propa-
only for a wave traveling in the direction along which gating Kelvin wave is a possible free solution to the
the lateral boundary (where the wave has maximum perturbation equations of the shallow water model on
amplitude) is always on the right in the Northern an equatorial p-plane ( p is the meridional gradient of
Hemisphere and on the left in the Southern Hemi- the Coriolis parameter), provided that the meridional
sphere (Figure 1).Therefore, the effects of rotation and velocity vanishes. This type of wave is called an
the lateral boundary determine the unidirectional equatorial Kelvin wave, so named because it is
propagation and the trapped behavior of the extremely similar in character to coastally trapped
Kelvin wave. Kelvin waves, with the Equator serving as a boundary.
1064 KELVIN WAVES
Like the coastal Kelvin wave, the propagation of an equatorial Kelvin waves are eastward-propagating
equatorial Kelvin wave is unidirectional, Le., eastward and have zonal velocity and pressure perturbations
only. In each vertical plane parallel t o the equatorial that vary with latitude as Gaussian functions centered
vertical plane, the motion of fluid particles is precisely on the Equator (Figure 2). The e-folding distance
the same as that in a shallow water gravity wave for decay with increasing latitude is given by
(Figure 2). The Kelvin wave propagates without Rc = (C/2P)'l2, where C = (gH)l12 is the gravity
dispersion at the speed C = (gH)1'2,as for nonrotat- wave speed and 8 is the meridional gradient of the
ing long gravity waves. Because the Coriolis parameter Coriolis parameter at the equator. Rc is called the
changes sign at the Equator, eastward flow occurring equatorial Rossby radius of deformation, because of
on both sides of the Equator would induce equator- its relationship with the decay scale for the case of
ward Ekman mass transport, piling up fluid at the constant Coriolis parameter. The same analysis can be
Equator and generating a meridional pressure gradient applied to baroclinic waves in both atmosphere and
force. In this sense, the Equator acts as a lateral wall. ocean, with H being interpreted as the equivalent
The Earth's rotation links the motion in each latitu- depth. Typical values of the internal gravity
dinal plane, because momentum conservation in the wave speed, C, for the tropical atmosphere are
north-south direction requires a geostrophic balance 20-80 m s - ', giving an equatorial Rossby radius of
between the eastward velocity and the meridional between 6 and 12 degrees of latitude. For baroclinic
pressure gradient force. This geostrophic balance ocean waves, appropriate values of C are typically in
results in the perturbation zonal velocity reaching a ',
the range of 2-3 m s - so that the equatorial Rossby
maximum on the Equator and decaying with increas- radius is 200-250 km.
ing distance from the Equator (Figure 2). This is
possible only for eastward-traveling waves. Thus,
Vertically Propagating Kelvin Waves
In general, the Earth's rotation traps planetary-scale
gravity waves in the troposphere unless the frequency of
the wave is greater than the Coriolis frequency (about
(1day) - '). For this reason, mid-latitude synoptic
waves are generally unable to penetrate significantly
into the stratosphere. However, near the Equator, the
dramatic decrease in the Coriolis parameter allows
these longer-period waves to propagate vertically.
Vertically propagating Kelvin waves have been identi-
fied in both the equatorial atmosphere and ocean.
Vertically propagating Kelvin waves can be illus-
trated by considering a continuously stratified fluid
with a constant buoyancy frequency in a semi-infinite
vertical domain near a lateral boundary or in the
vicinity of the Equator. For simplicity, consider a linear
equatorial 8-plane model. Solutions can be obtained
using the normal-mode technique by neglecting me-
ridional perturbations. In a vertical section along the
equator, or in any parallel vertical plane, the motion of
vertically propagating equatorial Kelvin waves shares
the properties of an ordinary vertically propagating
Figure 2 The theoretical equatorially trapped Kelvin wave gravity wave. A vertical cross-section of the perturba-
solution to the linear shallow water equations on an equatorial p- tion motion, pressure, and temperature structure for
plane for a nondimensional zonal wavenumber 1, which is vertically propagating Kelvin waves is shown in
constructed from the theory of Matsuno (1966). Hatching is for
convergence and shading for divergence, with a 0.6 unit interval
Figure 3. The local change in temperature is due to
between successive levels of hatching or shading, and with the adiabatic warming or cooling, so that the temperature
zero divergence contour omitted. Unshaded contours are for oscillation leads the vertical (zonal)wind and pressure
geopotential, with a contour interval of 0.5 units. Negative contours oscillations by a quarter-cycle.
are dashed and the zero contour is omitted. The largest wind vector Since stratospheric Kelvin waves are forced from
is2.3 units, as marked. (Reproduced with permission from Wheeler
M, et a/. (2000) Large-scale dynamical fields associated with
below by disturbances in the troposphere, the wave
collectively coupled equatorial waves. Journal of the Atmospheric energy propagation must have an upward component.
Sciences 57(5): 613-640.) According to theory, Kelvin waves become dispersive
KELVIN WAVES 1065
Rossby radius for barotropic Kelvin waves is about

3000 km. Continental shelf regions normally extend
about a hundred kilometers seaward; hence, a steep
f continental slope is practically indistinguishable from
a vertical boundary at the scale of the Rossby radius.
Thus, a barotropic Kelvin wave extends far from the
coast and occupies a substantial fraction of a typical
ocean. Much of the energy of tide waves traveling
along continents is transmitted in the form of baro-
tropic Kelvin waves with a speed of about 200 m s-'.
For instance, along the coast of California more than
two-thirds of the semidiurnal and half the diurnal tidal
amplitudes can be accounted for by traveling baro-
tropic Kelvin waves. For shallow seas and coastal
waters, the Rossby radius is about 200 km. When a
I / / / I Kelvin wave moves through a region in which the fluid
w Longitude E- depth or the Coriolis parameter varies and the wave
Figure 3 Longitude-height section along the Equator showing
energy flux remains constant, the amplitude of
pressure, temperature, and wind perturbations for a thermally the wave varies in proportion to ( f / H ) 1 ' 2 . Thus,
damped Kelvin wave. Heavy wavy lines indicate material lines; wave amplitude increases when Kelvin waves move
short blunt arrows show phase propagation. Length of the small into shallow water. In coastal regions, Kelvin waves
thin arrows is proportional to the wave amplitude, which decreases can also be generated as storm surges are diffracted by
with height owing to damping. The large shaded arrow indicates the
net mean flow acceleration owing to the wave stress divergence.
vertical boundaries and scattered by irregular coast-
(Reproducedwith permission from Holton JR (1992) lntroduction to lines. Variable longshore winds and atmospheric
Dynamic Meteorology, 3rd edn. San Diego: Academic Press.) pressure gradients acting on the sea surface are also
possible energy sources for oceanic Kelvin waves.
Internal coastal Kelvin waves can be generated by
in the presence of vertical propagation, and the vertical wind-induced, time-dependent coastal upwelling.
component of their phase velocity is always opposite Coastal upwelling (downwelling) is caused by an
to that of their group velocity (the velocity at which the Ekman mass flux transported offshore (onshore) and
wave energy is transmitted). Thus, the phase velocity forced by longshore winds. The disturbances can then
must have a downward component. The condition of propagate along the coast as boundary-trapped
eastward propagation due to equatorial trapping internal Kelvin waves. Therefore, the amount of
requires that the vertical wavenumber be negative. upwelling depends not only on local wind forcing
Hence, an eastward- and downward-propagating but also on the forcing that generated the waves at an
equatorial Kelvin wave has constant phase lines that earlier time. In a lake, during strong wind forcing, the
tilt eastward with height (Figure 3 ) . region of upwelling progresses cyclonically (in the
Because the variation in zonal wind depends on the Northern Hemisphere) at the speed of an internal
pressure gradient force, the highest zonal pressure Kelvin wave.
gradient precedes the largest westerly acceleration by a Equatorial Kelvin waves play a critical part in
quarter-wavelength, and the zonal wind and pressure thermocline adjustment. As the equatorial ocean
waves coincide. This creates an upward flux of wave circulation responds to wind stress forcing, equatorial
energy. An individual parcel moves up along the tilted Kelvin waves transmit signals rapidly from the west-
phase line, bringing westerly momentum upward, and ern to the eastern extremities of the ocean basin.
moves down, bringing easterly momentum down- Internal Kelvin waves propagating along the ther-
ward. Thus, the Kelvin wave transports westerly mocline take about two months to cross the entire
momentum upward. equatorial Pacific. These equatorial Kelvin waves play
an essential role in the El Niiio Southern Oscillation
(ENSO) cycle. First, they support a positive feedback
Significance of Kelvin Waves in the between the central Pacific zonal wind and eastern
Atmosphere and Ocean Pacific sea surface temperature (SST) anomalies. A
westerly wind anomaly excites downwelling Kelvin
Oceanic Kelvin Waves
waves, which propagate into the eastern Pacific,
Kelvin waves are essential in the description of ocean suppressing the thermocline and causing the SST to
tides. For a deep ocean ( H = 5 km) at 30"N, the rise; this, in turn, enhances the central Pacific westerly
1066 KELVIN WAVES
wind anomaly by increasing the eastward pressure Atmospheric Equatorial Kelvin Waves
gradient force in the atmosphere. This positive feed-
The atmospheric equatorial Kelvin wave is one of the
back provides a development mechanism for ENSO
critical wave motions in the response of the tropical
SST warming. Second, the cyclonic wind stress curl
atmospheric circulation to a heat source (Figure 4).
associated with the central Pacific westerly wind
When an imposed heating centered on the Equator is
anomaly can induce upwelling oceanic Rossby waves
switched on at some initial time, Kelvin waves carry
that propagate westward. These waves are eventually
information rapidly eastward, thereby creating east-
reflected at the western ocean boundary, generating
erly trade winds in that region and forming a Walker
upwelling equatorial Kelvin waves, which propagate
cell (rising motion over the heat source region and
into the eastern Pacific and offset the warming by
sinking motion to its east) (Figure 4C). Internal
enhancing vertical cold advection. This negative
equatorial Kelvin waves traveling with typical speeds
feedback provides a mechanism for turning the
of 20-80 m s - are an effective means by which the
coupled system t o its opposite (La Niiia) phase and
equatorial atmosphere becomes homogenized in the
sustaining the ENSO cycle. In addition, the atmos-
zonal direction. The easterly winds are in geostrophic
pheric intraseasonal wind forcing continuously gen-
equilibrium, so that there is a trough along the
erates equatorial Kelvin waves whose nonlinear
Equator, with the winds along the Equator flowing
rectification t o the mean state may also contribute to
directly down the pressure gradient (Figure 4). The
the eastern Pacific warming.
-1 0 -5 0 5 10 15
(A) X
-
-1 0 r
d 10 15
(ii)
Figure4 Solutionforheatingsymmetricabouttheequatorin theregion 1x1 <2fordecayfactor~= 0.1, (A) Contoursofverticalvelocity w

(solid contours are 0,0.3,0.6, broken contour is - 0.1) superimposed on the velocity field for the lower layer. The field is dominated by the
upward motion in the heating region where it has approximately the same shape as the heating function. Elsewhere there is subsidence
with the same pattern as the pressure field. (8)Contours of perturbution pressure p (contour interval 0.3) which is everywhere negative.
There is a trough at the equator in the easterly regime to the east of the forcing region. On the other hand, the pressure in the westerlies to
the west of the forcing region, though depressed, is high relative to its value off the equator. Two cyclones are found on the north-west and
south-west flanks of the forcing region. (C) The meridionally integrated flow showing (i) stream function contours, and (ii) perturbation
pressure. Note the rising motion in the heating region (where there is a trough) and subsidence elsewhere. (Reproduced with permission
from Gill AE (1980) Some simple solutions for heat-induced tropical circulation. Quarterly Journal of the Royal Mefeorological Sociefy.
106: 447-462.)
KELVIN WAVES 1067
westward-propagating Rossby wave regime to the propagating Kelvin waves carry westerly momentum
west of the forcing region is about one-third the size of upward, they are damped by radiative cooling, small-
the Kelvin wave regime because Rossby waves travel scale turbulence, and critical level interaction. As the
at one-third of the Kelvin wave speed (Figure 4A). The waves are damped, they lose momentum and acceler-
equatorial westerlies between the symmetric Rossby ate the westerly mean flow. The damping depends on
waves provide inflow into the heating region, and are the Doppler-shifted wave frequency. As the Doppler-
in geostrophic equilibrium, so that a relative ridge shifted frequency decreases, the vertical component of
appears along the Equator (Figure 4B). Meanwhile, the group velocity also decreases, and a longer time is
the flow converges toward the Equator. If atmospheric available for the wave energy to be damped. Hence,
damping is taken into account, this simple model can westerly Kelvin waves tend to be damped preferen-
largely explain the steady-state atmospheric response tially in the westerly shear zone where their Doppler-
to an imposed heat source. shifted frequencies decrease with height. The associ-
Kelvin and mixed Rossby-gravity waves are the ated momentum flux convergence produces westerly
predominant disturbances in the equatorial strato- acceleration of the mean flow, causing the westerly
sphere, and play a critical role in the stratospheric shear zone to descend. A similar argument is valid for
circulation through their vertical transport of energy the downward propagation of the easterly phase of the
and momentum. Stratospheric Kelvin waves are QBO through the action of Rossby-gravity waves.
excited by oscillations in the large-scale convective A peak in the variability of the tropical atmosphere
heating pattern in the troposphere, and are a source of appears in the 30- to 60-day period range, and is
westerly momentum for the QBO (see Middle Atmos- known as the Madden-Julian oscillation (MJO).
phere: Quasi-Biennial Oscillation). The QBO is a Madden and Julian found that a 30- to 60-day
zonal wind oscillation in the equatorial stratosphere, oscillation in zonal winds is in approximate geo-
and propagates downward with a period of about 24- strophic balance with varying pressure maxima and
30 months. Figure 5 shows an example of the zonal minima centered at the Equator. A low-level low-
wind oscillations caused by the passage of Kelvin pressure anomaly is accompanied by low-level easterly
waves near the Equator. The descent of the westerly anomalies. The pressure and zonal wind are out of
phase of the QBO is shown in the figure. At each level, phase in the upper troposphere. The wave patterns
there is an increase of the zonal wind with time. move eastward along the Equator. At the Equator, the
Superposed on this secular trend is a fluctuating Kelvin meridional winds appear to be insignificant. The
wave component with a period of about 12 days and a amplitude of the oscillation decays with distance away
vertical wavelength of about 10-12 km. As vertically from the Equator. These features are similar to those of
Figure 5 Time-height section of zonal wind at Canton Island (3’s).Isotachs at intervals of 5 m s - ’ . Westerlies are shaded.
(Reproduced with permission from Holton JR (1992) introduction to Dynamic Meteorology, 3rd edn. San Diego: Academic Press; original
courtesy of J. M. Wallace and V. E. Kousky.)
1068 KELVIN-HELMHOLTZ INSTABILITY
internal equatorial Kelvin waves except that the large Further Reading
vertical scale of the MJO implies a faster phase speed Cushman-Roison B (1994)An Introduction to Geophysical
than is observed. Arguments involving coupling with Fluid Dynamics. Englewood Cliffs, NJ: Prentice-Hall.
equatorial westward-traveling Rossby waves and Gill AE (1982) Atmosphere-Ocean Dynamics. New York:
interaction with the release of latent heat in the Academic Press.
disturbances as well as viscous damping have been Holton JR (1992) Introduction to Dynamic Meteorology,
invoked to explain the observed slow phase speed. 3rd edn. San Diego: Academic Press.
LeBlonde PH and Lawrence AM (1978)Waves in the Ocean.
New York: Elsevier.
See also Matsuno T (1966) Quasi-geostrophic motion in the equa-
Dynamic Meteorology: Waves. El Niho and the South- torial area. Journal of the Meteorological Society of
ern Oscillation: Observation; Theory. Middle Atmos- Japan 44: 25-42.
phere: Quasi-Biennial Oscillation. Ocean Circulation: Pedlosky J (1987) Geophysical Fluid Dynamics. Berlin:
Surface-Wind Driven Circulation. Rossby Waves. Trop- Springer-Verlag.
ical Meteorology: Equatorial Waves; Intra-seasonal Os- Philander SG (1990) El Nifio, La Nifia and the Southern
cillation (Madden-Julian Oscillation). Oscillation. New York: Academic Press.
P G Drazin, University of Bath, England, UK mechanism of billow clouds, clear air turbulence, and
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. other similar phenomena in the atmosphere.
The visionary British meteorologist Lewis Richard-
son recognized in the 1920s that atmospheric turbu-
Introduction lence could be maintained only if the inertial
Kelvin-Helmholtz instability is the name given, since instability due to shear could overcome the static
the 1940s, to an instability of a shear layer in a fluid, stability due to heavier air being beneath lighter air.
which is the mechanism of many phenomena observed The essence of his argument can be recapitulated in
in the atmosphere and oceans. It is said that in 1868 the terms of the energetics of the instability of a horizontal
German physiologist and physicist Hermann von shear layer in a stratified fluid, as follows. Suppose
Helmholtz first recognized the instability of a shear then that a basic flow of an incompressible inviscid
layer, by writing that ‘every perfect geometrically fluid of variable density has velocity U(z)i and density
sharp edge by which a fluid flows must tear it asunder p ( z ) , where i is a horizontal unit vector and z is the
and establish a surface of separation, however slowly height. The essential mechanism of the instability is
the fluid may move’, although this remark may seem to the conversion of the available kinetic energy of
denote merely recognition of separation of a flow at an relative motion of the horizontal layers of the fluid
edge. However, in 1871 the British physicist William into kinetic energy of a perturbation, overcoming the
Thomson, Helmholtz’s friend who was later ennobled potential energy needed to raise or lower fluid when
as Lord Kelvin, posed mathematically, and solved d p l d z 5 0 everywhere, that is when light fluid is
fully, a prototypical problem of linear instability of a always above heavier. Thus shear tends to destabilize
horizontal vortex sheet between the uniform moving and buoyancy to stabilize the flow. To quantify these
layers of two fluids of different densities, in an attempt tendencies, suppose that two neighbouring fluid par-
to model the formation of ocean waves by the wind. +
ticles of equal volume, at heights z and z 6z,are
Later, Helmholtz developed this model and applied it to somehow interchanged. Then the increment of work
the formation of billow clouds. At the same time the 6 W per unit volume needed to overcome gravity and
British physicist Lord Rayleigh was developing the effect this interchange is
theory of the instability of a shear layer, that is a parallel 6W = -g6p6z
flow in which the fluid speed varies across the layer.
These ideas were developed, extended, and applied where g is the acceleration due to gravity and
in the twentieth century. An especially important 6p = (dp/dz)6z. In order that the horizontal momen-
extension is the instability of a horizontal shear layer tum of the inviscid fluid is conserved in the inter-
in a stratified fluid, that is a fluid whose density varies change, the particle initially at height z will plausibly
with height, because this models more realistically the have final velocity intermediate between the velocities
internal equatorial Kelvin waves except that the large Further Reading
vertical scale of the MJO implies a faster phase speed Cushman-Roison B (1994)An Introduction to Geophysical
than is observed. Arguments involving coupling with Fluid Dynamics. Englewood Cliffs, NJ: Prentice-Hall.
equatorial westward-traveling Rossby waves and Gill AE (1982) Atmosphere-Ocean Dynamics. New York:
interaction with the release of latent heat in the Academic Press.
disturbances as well as viscous damping have been Holton JR (1992) Introduction to Dynamic Meteorology,
invoked to explain the observed slow phase speed. 3rd edn. San Diego: Academic Press.
LeBlonde PH and Lawrence AM (1978)Waves in the Ocean.
New York: Elsevier.
See also Matsuno T (1966) Quasi-geostrophic motion in the equa-
Dynamic Meteorology: Waves. El Niho and the South- torial area. Journal of the Meteorological Society of
ern Oscillation: Observation; Theory. Middle Atmos- Japan 44: 25-42.
phere: Quasi-Biennial Oscillation. Ocean Circulation: Pedlosky J (1987) Geophysical Fluid Dynamics. Berlin:
Surface-Wind Driven Circulation. Rossby Waves. Trop- Springer-Verlag.
ical Meteorology: Equatorial Waves; Intra-seasonal Os- Philander SG (1990) El Nifio, La Nifia and the Southern
cillation (Madden-Julian Oscillation). Oscillation. New York: Academic Press.
P G Drazin, University of Bath, England, UK mechanism of billow clouds, clear air turbulence, and
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. other similar phenomena in the atmosphere.
The visionary British meteorologist Lewis Richard-
son recognized in the 1920s that atmospheric turbu-
Introduction lence could be maintained only if the inertial
Kelvin-Helmholtz instability is the name given, since instability due to shear could overcome the static
the 1940s, to an instability of a shear layer in a fluid, stability due to heavier air being beneath lighter air.
which is the mechanism of many phenomena observed The essence of his argument can be recapitulated in
in the atmosphere and oceans. It is said that in 1868 the terms of the energetics of the instability of a horizontal
German physiologist and physicist Hermann von shear layer in a stratified fluid, as follows. Suppose
Helmholtz first recognized the instability of a shear then that a basic flow of an incompressible inviscid
layer, by writing that ‘every perfect geometrically fluid of variable density has velocity U(z)i and density
sharp edge by which a fluid flows must tear it asunder p ( z ) , where i is a horizontal unit vector and z is the
and establish a surface of separation, however slowly height. The essential mechanism of the instability is
the fluid may move’, although this remark may seem to the conversion of the available kinetic energy of
denote merely recognition of separation of a flow at an relative motion of the horizontal layers of the fluid
edge. However, in 1871 the British physicist William into kinetic energy of a perturbation, overcoming the
Thomson, Helmholtz’s friend who was later ennobled potential energy needed to raise or lower fluid when
as Lord Kelvin, posed mathematically, and solved d p l d z 5 0 everywhere, that is when light fluid is
fully, a prototypical problem of linear instability of a always above heavier. Thus shear tends to destabilize
horizontal vortex sheet between the uniform moving and buoyancy to stabilize the flow. To quantify these
layers of two fluids of different densities, in an attempt tendencies, suppose that two neighbouring fluid par-
to model the formation of ocean waves by the wind. +
ticles of equal volume, at heights z and z 6z,are
Later, Helmholtz developed this model and applied it to somehow interchanged. Then the increment of work
the formation of billow clouds. At the same time the 6 W per unit volume needed to overcome gravity and
British physicist Lord Rayleigh was developing the effect this interchange is
theory of the instability of a shear layer, that is a parallel 6W = -g6p6z
flow in which the fluid speed varies across the layer.
These ideas were developed, extended, and applied where g is the acceleration due to gravity and
in the twentieth century. An especially important 6p = (dp/dz)6z. In order that the horizontal momen-
extension is the instability of a horizontal shear layer tum of the inviscid fluid is conserved in the inter-
in a stratified fluid, that is a fluid whose density varies change, the particle initially at height z will plausibly
with height, because this models more realistically the have final velocity intermediate between the velocities
KELVIN-HELMHOLTZ INSTABILITY 1069
of the ambient fluid at its initial and final levels, say

( U + kGU)i, so that the other particle has final velocity
+
[U (1 - k)GU]i in order to conserve linear momen-
tum, where 6U = (dU/dz)Gz and k is some number
between 0 and 1.Then the increment of kinetic energy
6T per unit volume released by the interchange is
6T =$pU2 + $(p + 6 p ) ( U + 6U)2 - $p(U +

- $(p + 6 p ) [ U+ (1- k)SUl2
= k( 1 - k ) p ( 6 q 2 + kU6U6p
on neglecting higher-order terms in 6p( 6 U)2. For small

increments,
6T = k(1 - k ) p ( 6 U ) 2
L
on neglectingthe inertial effects of the variation of density
(this is a good approximation for instability in the
atmosphere because the buoyancy effects of the variation
of density there are almost always much greater). Thus,
P1
1P(4
0 P2
4.
with equality holding for k = Now a necessary
condition for there being enough energy to effect this
interchange, and hence for instability to occur, is that
6 W 5 6T, and therefore that
(B)
do 1 /dUI2
Figure 1 Sketch of (a) velocity and (b) density profiles in Kelvin’s
basic flow of two horizontal layers of fluid.
somewhere in the field of flow, or that
Ri(z) I 3
Theory and Experiments
where the local Richardson number is defined as the
dimensionless quantity Now let us go back to Kelvin’s problem, and see a few
of its details. Suppose that a basic horizontal flow of an
incompressible inviscid fluid is given by
a
This criterion, namely that Ri(z) 5 somewhere in
the flow, is a necessary condition for instability of the
3
given basic flow, and so Ri(z) > everywhere in the
flow is a sufficient condition for stability; it is called
Richardson’s criterion. The above argument for an
incompressible inviscid fluid may be adapted for a
perfect gas in adiabatic motion, and so for air in the
atmosphere, by replacing the density by the potential
density, and this leads again to Richardson’s criterion
but with the Richardson number redefined as
Ri(z) = g
T dz (”+ I-)/rg) 2
where T is the absolute temperature and r the

Figure 2 Kelvin’s cat’s-eye pattern. This shows the streamlines
adiabatic lapse rate. For a perfect gas = g/cp, where near the level where the phase velocity of the waves equals the
cp is the specific heat at constant pressure. In fact, basic velocity of a smoothly varying profile of a homogeneous
r M 8 K per kilometer in the troposphere. inviscid fluid.
This gives a vortex sheet at z = 0, as sketched in where k and l are given horizontal wavenumbers in the
Figure 1. Kelvin was motivated by the special case with x - and y-directions (so the wavelengths in these
p2 << p1 to model wind blowing on an ocean current. directions are 2nlk and 27111, respectively). He de-
We may anticipate the occurrence of internal gravity duced, by solving an eigenvalue problem, that
waves in the special case with U l = U1. In any event,
Kelvin took an irrotational flow coupled to a pertur-
k2P1P2(U1 - U2I2
bation of the profile of the vortex sheet, and resolved s=-ik
the perturbation into independent normal modes, with P1 + P2 (PI + P2I2
its flow quantities proportional to (or rather the real
parts of functions proportional to)
exp[i(kx + l y ) + st]
- g(k2 + 1 2 ) 1 / 2 ( P l - P 2 )
P1 + P2 1
Figure 3 (A) Photograph of instability of a shear layer. The lower stream of water moves leftwards faster than the dyed upper stream.
Photograph of F. A. Roberts, P. E. Dimotakis and A. Roshko recorded by Van Dyke (1982) An Album of Fluid Motion. Stanford, CA:
Parabolic Press. (B) Photograph of instability of a stratified shear layer. The long rectangular tube is filled with water above dyed brine.
After the fluid came to rest the tube was suddenly tilted to create the shear layer with downward acceleration of the brine and upward
acceleration of the water. The upper stream of water is moving rightwards, and the lower stream of brine leftwards. Photograph of S. A.
Thorpe recorded by Van Dyke (1982) An Album of Fluid Motion. Stanford, CA: Parabolic Press.
KELVIN-HELMHOLTZ INSTABILITY 1071
Expressing s = 0 - iw in real and imaginary parts, we Kelvin’s model to deal with basic velocity and density
identify o as the frequency of the mode, and 0 as its profiles varying smoothly with height. Their results for
relative growth rate. It follows that if various shear layers seemed to support Richardson’s
criterion, but John Miles and Louis Howard con-
firmed Richardson’s criterion mathematically in the
then this mode may grow exponentially with 1960s. Typical relative growth rates of a shear layer
are found to be 0 x 1 U2 - U11/1OL when the Richard-
son number is appreciably less than a quarter (note
that, by Richardson’s criterion, 0 = 0 if Ri(z) 2
everywhere), where L is the thickness of the layer.
and the x-component of its phase velocity is the mass- Taking 1 U2 - U11 = 10 m s - l and L = 100 m, as order
mean basic velocity of magnitude estimates for billow-cloud formation,
(3 - P1Ul+P2U2
-
we see that the linear instability breaks up the shear
layer with an e-folding time of about 0-l x 100s.
k - PI+P2
Thus Kelvin-Helmholtz instability is a transient
Now the flow is unstable if any one mode grows process in the atmosphere. The transience presents
exponentially, and so if 0 > 0 for any pair of real one way to distinguish billow clouds from lee-wave
values k and 1. But the formula above shows that 0 > 0 clouds, which are forced orographically.
for some modes with large values of k, provided that In 1880 Kelvin himself examined the streamlines
U2 # U1 and p l , p2 > 0, and therefore that all such due to instability of a shear layer of an unstratified
two-layer flows are unstable to short waves. However, fluid, and found what is now called the Kelvin’s cat’s-
Kelvin himself showed that surface tension, as well as eye pattern, shown in Figure 2. Taylor showed in 1931
buoyancy, could in fact stabilize the flow. (In passing, that for a stratified fluid the regions of closed stream-
one may note that the above formula for s gives the lines alternate at slightly different levels.
velocity of internal gravity waves at the interface of As the sinusoidal waves of the linear instability
two fluids when UZ = VI.) grow, nonlinearity will moderate the growth. Avortex
In 1931, Sydney Goldstein, Bernhard Haurwitz, sheet and a shear layer will then begin to roll up,
and Geoffrey Taylor independently generalized as shown in Figure 3. Of course, instability in the
Figure 4 A row of billow clouds photographed by Paul E. Branstine. For the meteorological data see Drazin PG and Reid WH (1981)
Hydrodynamic Stability, p. 22. Cambridge: Cambridge University Press.
1072 KINEMATICS
atmosphere is not so neat as in the careful laboratory Further, it has been conjectured that shear instabil-
experiments of Figure 3. ity in the absence of buoyancy plays a fundamental
role in turbulence itself.
Atmospheric Phenomena
Kelvin-Helmholtz instability occurs in the atmos- See also
phere as a sporadic, but widespread, phenomenon. It is Buoyancy and Buoyancy Waves: Theory. Clear Air
usually invisible, and so can be dangerous as clear air Turbulence. Clouds: Climatology. Dynamic Meteorol-
turbulence. But it can be detected by radar or seen by ogy: Waves. Instability: Inertial Instability; Symmetric
chance as billow clouds when the humidity is such that Stability. Lee Waves and Mountain Waves. Turbu-
the rising air in a vortex leads to condensation as cloud lence, Two Dimensional.
and the falling air leads to evaporation (see Figure 4).
When seen at an angle, the lines of billow clouds are
often called a ‘mackerel sky’; this is because of their Further Reading
resemblance to the pattern on the back of a mackerel, Drazin PG and Reid WH (1981) Hydrodynamic Stability,
the North Atlantic fish. 9$4,44, Chap. 4. Cambridge: Cambridge University Press.
Kelvin-Helmholtz instability is not only of local FaberTE (1995)FluidDynamics forPhysicists,$$8.11,8.12.
significance. It, and the turbulence into which it Cambridge: Cambridge University Press.
usually develops, play an important role in the energy Green J (1999) Atmospheric Dynamics, chapter 6. Cam-
budget of the atmosphere by transferring energy from bridge: Cambridge University Press.
the larger to smaller scales of motion until it is finally Van Dyke M (1982)A n Album of Fluid Motion, photos 145-
dissipated as heat by viscosity. 147. Stanford, CA: Parabolic Press.
D D Houghton,University of Wisconsin-Madison, History

Madison, WI, USA
Studies of the kinematics of air motion date back to the
advent of observing system networks for wind which
first became available in the nineteenth century over
parts of Europe and the United States. Kinematics was
Introduction applied to early studies of the extratropical cyclone.
Kinematics, when applied to the atmosphere, refers to Historical presentations by Kutzbach in 1979 and
the description of both air motion and the motion of Eliassen in 1999 document how the analysis and
patterns describing other properties of air, such as development of theory for these cyclones depended on
moisture content, temperature, and pressure. This the kinematics of the surface wind field. In particular,
description is without regard to forces and other conflicting theories were resolved by noting whether
physical processes that cause the motions. Air motion the dominant surface wind flow was primarily con-
itself is an important causal factor for many of the vergent, rotational, or confluent - three of the impor-
pattern changes of the other properties of the air. Daily tant basic air motion descriptors of atmospheric
sequences of weather maps showing horizontal wind kinematics to be discussed later.
flow and the movement of patterns in pressure and Weather map depictions of the atmosphere became
temperature dramatize the kinematic perspective of common in the twentieth century. Kinematics became
the atmosphere. a basic analysis perspective not only for wind fields but
This article presents an overview of atmospheric also for describing the motions of weather map
kinematics as follows. First, we look at a brief patterns, such as the movement of pressure isobars,
historical perspective. Then, basic characteristics and and weather fronts. The determination of large-scale
descriptors for air motion are reviewed. Finally, vertical motion (a quantity that could not be measured
examples of the wide range of applications of kine- directly) was an important application of kinematics.
matic analysis for the atmosphere are presented. Using a kinematic method, vertical motion was
1072 KINEMATICS
atmosphere is not so neat as in the careful laboratory Further, it has been conjectured that shear instabil-
experiments of Figure 3. ity in the absence of buoyancy plays a fundamental
role in turbulence itself.
Atmospheric Phenomena
Kelvin-Helmholtz instability occurs in the atmos- See also
phere as a sporadic, but widespread, phenomenon. It is Buoyancy and Buoyancy Waves: Theory. Clear Air
usually invisible, and so can be dangerous as clear air Turbulence. Clouds: Climatology. Dynamic Meteorol-
turbulence. But it can be detected by radar or seen by ogy: Waves. Instability: Inertial Instability; Symmetric
chance as billow clouds when the humidity is such that Stability. Lee Waves and Mountain Waves. Turbu-
the rising air in a vortex leads to condensation as cloud lence, Two Dimensional.
and the falling air leads to evaporation (see Figure 4).
When seen at an angle, the lines of billow clouds are
often called a ‘mackerel sky’; this is because of their Further Reading
resemblance to the pattern on the back of a mackerel, Drazin PG and Reid WH (1981) Hydrodynamic Stability,
the North Atlantic fish. 9$4,44, Chap. 4. Cambridge: Cambridge University Press.
Kelvin-Helmholtz instability is not only of local FaberTE (1995)FluidDynamics forPhysicists,$$8.11,8.12.
significance. It, and the turbulence into which it Cambridge: Cambridge University Press.
usually develops, play an important role in the energy Green J (1999) Atmospheric Dynamics, chapter 6. Cam-
budget of the atmosphere by transferring energy from bridge: Cambridge University Press.
the larger to smaller scales of motion until it is finally Van Dyke M (1982)A n Album of Fluid Motion, photos 145-
dissipated as heat by viscosity. 147. Stanford, CA: Parabolic Press.
D D Houghton,University of Wisconsin-Madison, History

Madison, WI, USA
Studies of the kinematics of air motion date back to the
advent of observing system networks for wind which
first became available in the nineteenth century over
parts of Europe and the United States. Kinematics was
Introduction applied to early studies of the extratropical cyclone.
Kinematics, when applied to the atmosphere, refers to Historical presentations by Kutzbach in 1979 and
the description of both air motion and the motion of Eliassen in 1999 document how the analysis and
patterns describing other properties of air, such as development of theory for these cyclones depended on
moisture content, temperature, and pressure. This the kinematics of the surface wind field. In particular,
description is without regard to forces and other conflicting theories were resolved by noting whether
physical processes that cause the motions. Air motion the dominant surface wind flow was primarily con-
itself is an important causal factor for many of the vergent, rotational, or confluent - three of the impor-
pattern changes of the other properties of the air. Daily tant basic air motion descriptors of atmospheric
sequences of weather maps showing horizontal wind kinematics to be discussed later.
flow and the movement of patterns in pressure and Weather map depictions of the atmosphere became
temperature dramatize the kinematic perspective of common in the twentieth century. Kinematics became
the atmosphere. a basic analysis perspective not only for wind fields but
This article presents an overview of atmospheric also for describing the motions of weather map
kinematics as follows. First, we look at a brief patterns, such as the movement of pressure isobars,
historical perspective. Then, basic characteristics and and weather fronts. The determination of large-scale
descriptors for air motion are reviewed. Finally, vertical motion (a quantity that could not be measured
examples of the wide range of applications of kine- directly) was an important application of kinematics.
matic analysis for the atmosphere are presented. Using a kinematic method, vertical motion was
KINEMATICS 1073
determined from the horizontal divergence in the The differential advection capabilities of the hori-
observed wind field using a relationship from the zontal wind field can change the horizontal distribu-
continuity of mass equation. tion patterns of properties such as temperature,
By the middle of the twentieth century, kinematic pressure, moisture, or the wind field itself. Spatial
analysis for the atmosphere had become fully devel- variation in a velocity field can be described as the
oped. Authors such as Saucier in 1955 and Pettersen in summation of four distinct properties of the wind field.
1956 described a wide range of atmospheric kinematic These properties and their effects on an individual
principles and applications. element, for example a specified region of water vapor,
are summarized below and in Figure 2:
Basic Air Motion Properties 1. uniform translation:movement with no change in
shape, orientation, or horizontal extent;
Motion Characteristics
2. rotation: change in orientation without change in
The advection (transport) of atmospheric properties shape or horizontal extent;
by air motion is a fundamental means by which the air 3. divergence: change in horizontal extent or area
motion affects the state of the atmosphere. This without change of shape and orientation;
advection alone can change conditions at a specific 4. deformation: change in shape without rotation or
point in space as patterns of fluctuation in the change in horizontal extent.
atmosphere are moved past this position. For instance,
a west wind in an area where the air temperature is Mathematical definitions of these four properties
warmer to the west than to the east will tend to for the two-dimensional wind field can be derived
increase temperatures in the middle area (Figure 1). using a linear Taylor series expansion of the velocity
The mathematical description of this effect can be field with reference to the center position of the
expressed as specified area of interest. Let uo and vo be the values of
the x component (eastward) and y-component (north-
a T p t = --u a T / a X PI ward) of the horizontal velocity at the center point and
u(x, y) and v(x, y) be the velocity component values at
where T is the temperature, t the time, u the air distances x and y from the center point. Then, the wind
motion speed in the west-east direction, x the field characteristics relevant for each of the four
distance in the west-east direction, and a the effects, respectively, are:
partial derivative operator. The right-hand-side
quantity is generally referred to as thermal advection. 1. uniform translation: movement with speed uo in
The temperature change, aT/at, is positive as u the x direction and vo in the y direction;
is positive and aT/ax negative for the conditions 2. rotation: turning defined by the vertical component
described above. of vorticity, { = av/ax - au/ay;
The fact that the air motion, itself, varies in space 3. divergence:expansion defined by horizontal diver-
adds an important characteristic to the advection gence, 6 = au/ax + avlay;
effect. Namely, the differential in advection effects can 4. deformation: shape distortion at a rate defined by
change the shape and pattern of the quantity of air the horizontal component magnitude,
being moved at the same time it is being displaced.
These effects have been examined most thoroughly for
the two-dimensional case, i.e., the horizontal wind
field and horizontal distribution of air properties. Note that ‘vorticity’ referenced in the rotation
Significant distortions in horizontal patterns can arise definition may be described as ‘relative vorticity’ in
with differential advection acting over sufficiently atmospheric applications where velocities are meas-
long durations of time. ured ‘relative’ to the Earth surface which itself is
turning. Also note that the deformation expression is
more complex than the vorticity and divergence
expression in that it depends on the position of the
coordinates, Le., the details of deformation effects
- - - -
depend on the orientation of the deformation field.
Warm .A Cold
The deformation has two components described in the
definition above by the two quantities that are squared
Figure 1 Temperature and wind pattern (arrows) that lead to
inside the square root function.
warming at Point A by the advection (transport) process due to the The basic differential advection characteristics can
air motion. be generalized for the full three-dimensional wind
1074 KINEMATICS
field. This results in many more scalar descriptors for

the motion properties compared to the two-dimen-
sional case presented above. The two-dimensional
(horizontal) perspective is sufficient for many atmos-
pheric applications, especially for large-scale motions,
such as those depicted in weather maps, where the flow
field is predominantly horizontal. This would not be
true for small-scale systems such as turbulence near the
Earth’s surface, cumulus clouds, and tornadoes. In
these cases, vertical motion fluctuations can be as large
as those for the horizontal motions.
Motion Descriptors
There are a number of descriptors for air motion that

are frequently used in kinematic analysis of the
atmosphere. It is useful to define them and to indicate
why they are useful and how they relate to the
fundamental definition of motion. Air motion is a
vector with a direction and magnitude (speed) that
alternatively may be described by the components of
motion in the directions of the coordinate system. In
A the three-dimensional atmosphere, an example would
be the u, u, and w components in the eastward ( x ) ,
northward ( y ) ,and vertical ( z )directions, respectively.
The four distinct properties of the air motion
described above (uniform translation, vorticity, diver-
- gence, and deformation) are not only important

components for the kinematic description but also
serve as basic variables for analyzing the dynamics of
T-
I
- the system. For large-scale horizontal motions where
the vertical vorticity component is generally larger
\ than the horizontal divergence and deformation com-

ponents, the governing equations of motion can be
approximated by equations for the rate of change of
i the vertical component of vorticity. Divergence has a
direct relationship to vertical motion and is often used
in descriptions for large-scale airflow to represent
vertical motion conditions. Deformation has more
limited use as a separate variable. Its most basic
application is in relation to analysis for atmospheric
frontogenesis, the situation where horizontal varia-
tions (gradients)of temperature, or other atmospheric
descriptors, are increasing.
Six other descriptors used to characterize air
motions in the atmosphere are discussed below.
1. Streamline: line parallel to and following along the

flow direction in space at a given time. Streamlines
Figure 2 Four basic kinematic properties of air motion and the give information only on direction of flow, not the
effect of each respectively on an element being advected (trans- speed of flow. The direction taken by smoke
ported) by the air. Panels A-D show uniform translation, rotation, coming out of a smokestack will be in the same
divergence, and deformation, respectively, for horizontal motion.
The solid-line enclosed area shows the advected element at the direction as the streamlines at the time when it
initial time and the dashed-line enclosed area shows the element at comes out. Often streamlines are shown for two-
a slightly later time. dimensional space.
KINEMATICS 1075
2. Isotach: contour surface connecting positions with

the same wind speed. On a two-dimensional chart,
isotachs are contour lines which together with
streamlines together give a complete description of
the two-dimensional flow field.
3. Trajectory: line following the course of an air
parcel element which is moving with the air
motion. In general, the trajectory is three dimen-
sional. Often trajectories are shown for a nearly
horizontal two-dimensional space such as the
Earth surface. Trajectories will follow streamlines
if flow conditions are steady state. The track taken
by a single puff of smoke emitted from a smoke-
stack defines its trajectory.
4. Streakline: line connecting the air parcel elements Figure 4 Example of a lateral wind shear field for straight-line
whose trajectories all emanate from a single posi- wind flow. The arrows show the (uniform) direction of flow and the
tion in space. The pattern produced by a smoke- arrow length is proportional to the speed.
stack that is continually emitting smoke is one
example of a streakline.
5. Streamfunction: a function in two-dimensional streamfunction defines the associated vertical com-
space (e.g., $(x, y ) ) whose spatial gradients define ponent of vorticity, 5,according to the relationship
the wind flow for nondivergent flow conditions in
that space in the following way: u = -a$/ay and
u = a$/&. The contours of constant streamfunc-
tion, like the streamlines, are parallel to the wind 6. Velocitypotential: a function (e.g., ~ ( xy ,,z ) )whose
flow. In addition, the gradients of streamfunction gradients in space define the divergent components
value perpendicular to the lines of constant stream- of motion. In atmospheric sciences where a two-
function define the wind speed. Thus, in a map dimensional formulation is commonly used, the
showing streamfunction contoured at a regular horizontal divergence is defined from ~ ( xy , ) by the
interval, the distance between adjacent contours relationships ux = -ax/& and ux = -BX/By,
will be inversely related to the wind speed where uxand ux refer to the divergent component
(Figure 3). For many quasi-horizontal, large-scale of velocity in the x and y directions, respectively.
atmospheric flows, which have negligible horizon- 7. Wind shear: variations of the wind vector or its
tal divergence, the streamfunction is sufficient to components in a given direction. Commonly, this
define the horizontal wind field. The Laplacian of term is used to refer to a lateral wind shear, i.e., a
change in speed in a direction perpendicular to
the wind direction (Figure 4).Thus for west-to-east
wind flow defined by the u component of motion in
the x direction, lateral wind shear involves varia-
tions in the y direction (north) and z direction
(vertical).The presence of wind shear is a factor in
many types of flow instability. Wind shears are also
representative of both vorticity (rotational) and
deformation flow components in the wind.
Applications
Kinematic analysis is used in a wide number of
applications to describe conditions in the atmosphere.
\------
Several of these are briefly described here.
Figure 3 Contours for streamfunction (solid lines) and the Describing Pressure Field Patterns
associated wind field (arrows) for a two-dimensional and nondi-
vergent flow. The arrows show the direction of flow and have
As mentioned earlier, kinematic analysis has been
lengths proportional to speed. Streamfunction values of the applied to the motion of patterns of pressure, temper-
contours decrease going from the bottom to the top of the chart. ature, humidity, etc. describing the atmospheric
1076 KINEMATICS
condition, such as seen in weather map sequences. A locations commonly used to indicate pending weather
common depiction is made for pressure distributions, changes. Such pressure tendencies correspond directly
which relate directly to the horizontal wind for large- to the motions of isobars and the spacing between
scale conditions. Pressure at a constant height (often isobars (magnitude of horizontal pressure gradient).
mean sea level) is used to describe the surface weather For large-scale motions, the isobars and height
systems and winds using lines connecting regions with contours themselves have a well-defined relationship
the same pressure (called isobars). At higher levels in to the wind flow in extratropical latitudes according to
the atmosphere, the horizontal pressure variations the ‘geostrophic law’ approximation. Specifically, the
are commonly represented by variations of the horizontal flow tends to be parallel to the isobars and
height where a given pressure is found. In this case, height contours in the direction where, in the Northern
we use lines of constant height on a surface of fixed Hemisphere, the lower pressure (or height) is to the left
pressure (called height contours) rather than lines facing in the direction of the flow and the higher
of constant pressure (called isobars) on a surface of pressure (or height) is to the right. (Directions are
fixed height to define the horizontal pressure varia- opposite in the Southern Hemisphere.) In addition, the
tions. In the discussion below these terms are used flow speed is inversely proportional to the horizontal
interchangeably. pressure gradient, Le., the distance between adjacent
The motions of the surface pressure isobars and isobars or height contours. This means that the
height contours themselves are due to both movement pressure contour patterns have a direct relationship
and amplitude changes of weather systems. Such to flow streamlines so that kinematic properties of the
motion is witnessed in the propagation of high- and airflow itself, such as vorticity and deformation, are
low-pressure system centers and their associated suggested by the pressure contour patterns. Near the
isobar patterns in a sequence of daily weather map Earth’s surface frictional effects have a systematic
depictions (Figure 5 ) . These motions relate to the effect on the large-scale wind flow, causing air to have
pressure tendencies observed by barometers at fixed a flow component across isobars from high to low
pressure. This leads to horizontal wind divergence and
associated vertical motion when the isobars are curved
as around low- and high-pressure centers (Figure 6 ) .
Considerable specialized and descriptive terminol-
ogy is used for features and their time changes in the
large-scale horizontal pressure field patterns not only
by scientists but also by weather presenters on radio
and television. Commonly used terms include trough,
ridge, digging trough, inverted trough, cut-off low,
filling low, building ridge, etc. Some of these terms
characterize pressure patterns and their changes which
are quite useful for describing dynamic as well as
kinematic conditions. For instance, in the situation
where isobars or height contours run generally east-
west with the lower pressures to the north, a ‘trough’
refers to a region of relative minimum in pressure
;;I
where isobars for various pressure values are displaced
southward compared to positions to the east and west.
A ‘tilted trough’ means that the longitude of minimum
pressure differs according to latitude, so that a line
connecting positions for the location of lowest pres-
Y i Y H I sure at each latitude ‘tilts’ eastward or westward with
latitude (Figure 7). Such tilts for the associated large-
(B) scale motion fields imply systematic correlations
Figure 5 Surface pressure isobars (solid lines) for large-scale involving temperature, momentum, and transport
weather systems for two consecutive times demonstrating the processes. These correlations would be expected to
movement and changes of pressure field patterns, a focus of result in extratropical storm system development
atmospheric kinematics: (A) earlier time and (B)later time. The H
and L denote local regions of highest and lowest pressure,
located in the trough area (in the Northern Hemi-
respectively. The contour interval is a fixed value so changes in sphere) if the trough is tilted from south-west to north-
the number of isobars go along with changes in pressure east (a ‘positive’ tilt) with the reverse for a tilt from
magnitudes at the H and L positions. south-east to north-west, a ‘negative’ tilt.
KINEMATICS 1077
Determining Vertical Motion
Vertical motion is a key factor for many aspects of

weather, such as cloud development, precipitation,
development of weather systems, etc. Nevertheless, as
mentioned before, few direct measurements of vertical
motion are available. However, a number of methods
exist for estimating large-scale vertical motion. The
kinematic method is based on the principle of conser-
vation of mass which provides a direct relationship
between the horizontal divergence field and the
vertical motion. Recall that horizontal divergence is
one of the basic components of the horizontal motion
field as discussed earlier.
The atmosphere is a compressible gas so that
variations in density must be included in the relation-
ship. For the incompressible case (e.g., like water), a
very simple relationship between horizontal diver-
gence and vertical motion exists:
aw/az = -[au/ax + av/ayl [31
where w is vertical motion, z is height, the other
variables are as defined earlier. Integration of this
relationship with respect to height provides magnitu-
des of vertical motion at all levels as long as it is known
at one level. To account for the compressibility aspects
important for atmospheric weather systems, a density-
Figure 6 Large-scale horizontal wind field (arrows) in the weighted formula can be used:
Northern Hemisphere associated with surface pressure isobars
(solid lines): (A) without and (8) with surface friction effects, a(pw)/az = - m w X +~ W P Y N [41
respectively. The H and L denote local regions of highest and
lowest pressure, respectively. Note that in panel B with the curved where p is density and the variation of p in time is
isobars, the horizontal divergence of the wind is more evident. assumed to be negligible. This formulation is less easy
According to the conservation of mass relationship, vertical motion to use because of the spatial variations in density. In
above the surface would be expected in the low- and high-pressure some applications density variations on the right-hand
center areas in panel B.
side of the equation can be neglected.
An equivalent formulation of the continuity of mass
relationship using pressure instead of height for the
vertical coordinate system is often used for determin-
ing large-scale vertical motion. In this coordinate
-w
- - \
iw
-e
North -.L
--t-
G
,7 - 7
J
v
--H
A system, the vertical motion is represented by the
variable o representing the pressure change following
an element of air. In this case the equation becomes
aw/ap = -[aulax + av/ayl 151
(A) (B)
where the derivatives on the right-hand side are
Figure 7 Examples of large-scale pressure field patterns (height evaluated on constant pressure surfaces.
contours) and associated motion flow direction (arrows) typical in
the upper troposphere (Northern Hemisphere) which demonstrate Frontogenesis
'tilted trough' terminology. Side A: negative tilt trough (a more
westward position of the minimum pressure on a latitude line as one Weather 'fronts' relate directly to stormy activity and
goes further north); side B: positive tilt trough (a more eastward rapid changes. Fronts commonly refer to regions in the
position of the minimum pressure on a latitude line as one goes large-scale setting where there are maxima in hori-
further north. Heights (pressures) decrease to the north. The
motions associated with the positive tilted trough would be zontal variations (gradients) for descriptors such as
expected to produce development of an associated extratropical moisture, temperature, and winds in the atmosphere.
cyclone. Horizontal temperature gradients are commonly used
1078 KINEMATICS
provide a more systematic propagation of the pollu-

tion. Kinematic analysis can provide useful informa-
tion on where to expect pollution conditions,
especially for the larger-scale motions, and also to
determine the source regions for pollution conditions.
Flow trajectories define the path of systematic
propagation by air motion. Since most pollution cases
involve sources that act over sustained periods of time
or even continuously, the distinction between trajec-
Figure 8 Horizontal flow deformation field (arrows) and temper- tory and streamline is important for understanding the
ature contour (dashed lines) relationships where the temperature
variation (gradient) would be expected to increase in time (A) and
overall lateral spread of pollution from a source area.
decrease in time (B) because the deformation field axis of Namely, the wind flow at any one time is not sufficient
confluence (shown by a dotted line) is perpendicular or parallel, information. It is the cumulative effect over time
respectively, to the temperature contours moving them closer measured by the ensemble of trajectories originating in
together and further apart, respectively. Temperatures are colder the pollution source area. The availability of computer
at the top side of the diagrams as labeled.
models for atmospheric motion makes it possible to
determine atmospheric trajectories routinely. It is even
possible to determine the source regions of trajectories
to define the fronts and are shown on weather maps as reaching a given region by looking back in time. This
cold, warm, and stationary fronts. Understanding the allows for definition of the source regions for air
movement and changes in such fronts is important for pollution observed in a given area.
describing, understanding, and predicting rapid
changes in atmospheric conditions. Current Weather and Short-Range Prediction
An important factor in the formation of large
Kinematic analysis of the wind flow is important for
horizontal gradients is the kinematic deformation in
understanding current weather conditions and mak-
the horizontal wind field. Such an increase in gradients
ing predictions for the near future, i.e., within 12 h.
occurs when the horizontal deformation flow in the
Dynamical weather prediction models are particularly
atmosphere is oriented so that the axis of dilatation
limited in the first 1 2 h for predicting clouds and
(expansion) is parallel to the contours of temperature
precipitation since these variables are poorly repre-
and the axis of confluence (contraction) is perpendic-
sented in the initial conditions, and it takes a while for
ular to the contours of temperature (Figure 8). This
results in the contours of different temperature values the prediction model to recreate existing fields in the
forecast. Extrapolation of initial conditions based on
being brought closer together, signifying an increase in
motion transports alone determined by kinematic
the horizontal temperature gradient. An examination
analysis can provide important information on the
of the existing horizontal temperature distribution and
atmospheric conditions, especially for moisture,
the horizontal deformation distribution in the wind
clouds, and precipitation.
field or as implied from the pressure field can give
Kinematic analysis for horizontal wind flow also
considerable insight on how existing temperature
provides a direct measure of vertical motion through
gradient regions might change.
the kinematic relationship between horizontal diver-
gence and vertical motion (discussed before). Such
Air Pollution analysis applied on the small scale (horizontal dimen-
sions of 1 0 0 k m or less) is used by forecasters to
Air pollution arises from the transport of pollutants
identify regions of upward vertical motions and thus
from source regions often at or near the Earth’s
probably more likely for development of convective
surface. Sources may be obvious point sources like a
precipitation systems in the near future, i.e., several
smokestack or more dispersed sources like automo-
hours. Such analysis is also useful for describing heavy
biles. Temperature differences between the source
rain situations that have occurred.
material and the atmosphere near the source region or
density differences due to the pollutants themselves
Measurement of Winds
may influence how the pollution propagates. Howev-
er, much of the propagation will be due to transports A variety of methods are used to measure the
by the atmospheric motion. Such transports can be horizontal wind flow. Some operate from fixed posi-
separated into small-scale turbulent motion effects tions such as surface weather station anemometers and
which cause a general dispersion of the pollution over weather vanes that give measurements in a fixed
time in all directions and larger-scale motions that (Eulerian) reference system. Others such as weather
KINEMATICS 1079
balloons, clouds tracked from satellites, and raindrops Conclusions

tracked with Doppler radar provide estimates for
Atmospheric kinematics embodies the description of
tracers presumed to be following the wind flow (a
movement of air and its descriptors such as moisture
Lagrangian system reference). Cloud tracking tech-
content, pressure, and temperature with a focus on the
niques for this purpose have the additional challenge
air motion itself. The importance of kinematics has
of determining the height of the cloud and sorting out
been shown by a sampling of its use in a wide range of
situations where the clouds do not move with the
applications. In the discussion here, the focus has been
horizontal wind such as for the stationary wave clouds
on the conceptual aspects of kinematics without
seen around mountains. Combining such observations
dwelling on the mathematics. It should be noted that
into a single description of the wind flow requires
since air motion is a vector, the complete mathematics
taking into account the distinction between trajecto-
for the concepts presented here can involve compli-
ries and streamlines and the differing impacts of space
cated mathematical expressions.
and time averaging.
See also
Analysis and Dynamical Process Studies Aerosols: Observations and Measurements. Beaufort
Wind Scale. Cyclones, Extra Tropical. Fronts. Syn-
Published descriptive summaries of atmospheric cir- optic Meteorology: Weather Maps. Tornados. Trac-
culation features include routinely many of the ers. Turbulent Diffusion. Vorticity.
kinematic horizontal flow descriptors discussed here.
For example, the publication Climate Diagnostics
Bulletin published by the US National Weather Further Reading
Service National Centers for Environmental Predic- Bluestein HB (1992) Synoptic-Dynamic Meteorology in
tion provides monthly summaries of many global- Midlatitudes - Vol. I: Principles of Kinematics and
scale circulation features. Quantities presented in- Dynamics. New York: Oxford University Press.
clude zonal wind flow (the west-to-east ‘u’ component Bluestein HB ( 1993) Synoptic-Dynamic Meteorology in
of motion), horizontal flow vectors (streamline direc- Midlatitudes - Vol. I l : Observations and Theory of
tion indicators) and isotachs (speed)for both the lower Weather Systems. New York: Oxford University Press.
and upper troposphere along with upper tropospheric Climate Prediction Center. Climate Diagnostics Bulletin.
level streamfunction, divergent wind and velocity National Centers for Environmental Prediction, Nation-
potential fields. The latter two quantities give an al Weather ServiceNational Oceanic and Atmospheric
indication of large-scale vertical motion in the atmos- Administration. Camp Springs, MD: US Department of
Commerce (publishedmonthly since 1983).
phere through the kinematic relationships discussed
Eliassen A (1999) Vilhelm Bjerknes’s early studies of
before. atmospheric motions and their connection with the
The kinematic descriptors also are used in studies cyclone model of the Bergen school. In: Shapiro M and
of the dynamical processes in the atmosphere. Grmnds S (eds)T h e Life Cycles of Extratropical Cyclones.
The kinematic variables include some more complex Chapter 1, pp. 5-13. Boston, MA: American Meteoro-
than the basic kinematic variables described here. logical Society.
As examples, dynamical studies of large-scale weather Hess SL (1959) Introduction to Theoretical Meteorology.
systems utilize potential vorticity (a combination New York: Holt, Rinehart and Winston.
of rate of temperature decrease with height and Kutzbach G (1979) T h e Thermal Theory of Cyclones - A
vorticity measures) and Q-vectors (diagnostic varia- History of Meteorological Thought in the Nineteenth
bles that relate to vertical motion derived from the Century. Boston, MA: American Meteorological Society.
Petterssen S (1956)Weather Analysis and Forecasting - Vol.
equations of motion and thermodynamics as well as
I: Motions and Motion Systems. New York: McGraw-
the continuity of mass relationships). The study of Hill.
vortices such as dust devils, tornadoes, and rotating Petterssen S (1956)Weather Analysis and Forecasting - Vol.
thunderstorms is facilitated by a kinematic variable Il: Weather and Weather Systems. New York: McGraw-
called helicity. This variable combines the velocity and Hill.
three-dimensional vorticity (rotational aspects) of the Saucier WJ (1955) Principles of Meteorological Analysis.
velocity field. Chicago, IL: University of Chicago Press.
LABORATORY GEOPHYSICAL FLUID DYNAMICS 1081
R L Pfeffer, Florida State University, Tallahassee, FL, experiment, and for which regime diagrams have been
USA constructed that map the fluid behavior over a broad
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. range of parameters.
Introduction Similarity
Laboratory experiments represent one of several In developing a laboratory prototype, one strives to
approaches employed to gain insight into the processes attain similarity in as many ways as possible. An exact
responsible for natural phenomena. Other approaches model would be one that had the same geometry as the
include diagnostic studies of observational data, system to be modeled, but on a smaller scale, and the
analytic theory, and numerical solutions of the gov- same governing equations, with the same relative
erning equations. Each has strengths and limitations. magnitudes of the different terms in these equations.
In concert, they can provide good insight into the The relative magnitudes are measured by dimension-
mechanisms involved. less numbers, which are ratios of different terms in the
Diagnostic studies are useful in the characterization governing equations. For example, the ratio of the
of dynamical properties of observed phenomena such relative acceleration of a fluid parcel to the Coriolis
as convection; turbulence; the propagation, growth acceleration in a rotating system leads to the definition
and damping of waves; and the time-mean or space- of the Rossby number Ro I(dV/dt)j/12R X VI
mean circulation. They provide a framework in which where V is the fluid velocity, t is time and R is the
to assess theories and laboratory simulations, which angular velocity of the system. If the acceleration is
-
must exhibit essentially the same properties in order to dominated by advection V VV, and if V is a charac-
be considered valid. They cannot, however, isolate the teristic speed of the flow and L a characteristic length
roles of controlling parameters (e.g., rotation and scale, the Rossby number can be expressed as
differential heating) in bringing about observed phe- Ro = V / f L , where f is the Coriolis parameter.
nomena, or provide insight into how things would Some dimensionless numbers are arrived at after
change if these parameters were altered. combining all or several of the governing equations
Analytic theory, based on simplified models, is not (usually in linearized form) to obtain equations that
designed to deal with the full complexity of different reveal the relative influences of thermodynamic and
phenomena, but rather to determine the roles of the mechanical processes. One example is the Rayleigh
major controlling parameters. number Ra ga(-aT*/az)H4/vq where -aT*/az is
Numerical simulation, using discretized approxi- the excess of the undisturbed vertical temperature
mations to the full nonlinear governing equations, can lapse rate over the adiabatic lapse rate in a character-
treat much more complex systems but is limited in istic depth H of fluid, z is height, g is the acceleration of
accuracy by the fact that important sub-grid-scale gravity, and v, IC, and c( are respectively the kinematic
phenomena must be parameterized, and by the reality viscosity, thermometric conductivity, and coefficient
that the solution of the discretized (algebraic) equa- of volume expansion of the fluid due to temperature T.
tions departs progressively with time from that of the The Rayleigh number measures the ratio of the
governing differential equations. destabilizing effect of buoyancy to the stabilizing
Laboratory experiments, under carefully controlled effect of viscous and thermal diffusion.
and reproducible conditions, can be used to verify In geophysical problems, it is usually impossible to
predictions based on analytic theory and numerical construct laboratory analogues with the same geom-
simulation and to discover phenomena that require etry as the systems to be modeled. Although some
new theory to explain them. In order to gain insight, it experiments have been done in space vehicles in zero
is not desirable to deal with the full complexity of each gravity, using electromagnetic forces to simulate
phenomenon. Moreover, full similarity cannot gener- gravity normal to a sphere, laboratory experiments
ally be achieved. on Earth typically replace the Earth's spherical geom-
In this article, we select for discussion a few of the etry with cylindrical geometry, in order that gravity be
many types of experiments that have been employed to normal to the bottom of the fluid. Moreover, it is not
understand phenomena in the Earth's atmosphere, generally possible to achieve the same balances among
oceans, and fluid interior. Examples are chosen that all terms in the governing equations in the model as in
illustrate the interplay of diagnostics, theory, and nature. Experimentalists attempt, instead, to model
1082 LABORATORYGEOPHYSICAL FLUID DYNAMICS
only the balances that are essential to the problem and

to explore how the flow changes as the most important
parameters are varied.
Vertical Convection between

Horizontal Planes (Rayleigh-Benard
Convection)
Convection due to heating at the bottom and cooling
at the top of a fluid layer is ubiquitous in the Earth’s
atmosphere, oceans, and interior. It is also found in the
Sun, in other stars, and in the atmospheres and
interiors of other planets. The simplest experiment
illustrating this phenomenon is accomplished by
maintaining the bottom boundary of a liquid at a
higher temperature than the top boundary. In the
absence of dissipation, the resulting density increase
with height would clearly be unstable, leading t o a
massive overturning until the more dense fluid is found
below the less dense fluid. In the presence of heat
conduction and viscosity, however, there is a critical
value of the Rayleigh number Ra, below which
convective overturning cannot occur. This value
depends on the details of the boundary conditions
and the heating arrangement (e.g., constant tempera-
tures or constant heat fluxes at the boundaries), but is
always of the order lo3. Henri Bknard, who first
performed such experiments in 1900, found that,
when convection set in, it took the form of polygonal
cells (which ultimately became regular hexagons).
Later experiments showed that the preferred plan Figure 1 (A) Plan view of convection cells generated by lowering
form at instability with most imposed conditions is a cylinder containing a high Prandlt number fluid into a larger
steady two-dimensional rolls. Today, it is generally container of warm fluid. The irregular polygonal plan form is due to
the changing mean temperature of the convecting fluid. (6) Vertical
accepted that Bknard’s cells were produced by surface
cross-section depicting broad-scale downward motion along the
tension gradients due to temperature variations on the cell boundaries and upward and outward motion in the interiors of
free upper surfaces of very shallow fluids. the cells. Experiment by the author, Geophysical Fluid Dynamics
At successively larger Rayleigh numbers, the plan Institute, Florida State University.
form of the convection changes, the details depending
on Prandtl number (PY v / K ) . At low Pr, the transi- between the lower and upper portions of the fluid by
tion is from steady rolls to three dimensional periodic means of Reynolds stresses, thereby creating and
oscillations. At high PY, there is an intermediate range maintaining a steady large-scale circulation of the
of Ra in which steady three-dimensional cells are fluid. Incorporation of this mechanism into the cloud
observed. At sufficiently large Ra, the periodic oscil- parameterization in one global numerical prediction
lations give way t o turbulent convection. model led to the correction of systematic errors in the
Figure 1 shows patterns and the associated circula- predicted tropical wind field. At very high Ra, the
tion that can arise in cellular convection. Figure 2 is a plumes form clusters that exhibit systematic horizon-
regime diagram covering a broad range of Ra and PY. tal drift.
Below Ra, the vertical heat flux 2 is accomplished by Much theoretical work has been done on the
heat conduction. With increasing Ra above this convection problem, beginning with the linear theory
transition, R (accomplished primarily by convection) of Lord Rayleigh in 1916. The theory establishes that
increases with Ra at fixed PY,with jumps in the rate of the only parameter controlling the onset of convection
increase at each successive transition. At high Rayleigh is Ra, and that the cell diameters at the onset of
numbers, within the turbulence regime, convective convection, which are controlled by the buoyancy and
plumes tilt with height and transfer momentum dissipative scales, are about twice the depth of the
LL
I
107 - I
I
LL
I
Figure 2 Regime diagram depicting the dependence of plan form and time dependence on Ra and Pr when the top and bottom
boundaries are rigid and maintained at different temperatures. Symbols: 0 steady cellular flow; 0 , time-dependent cellular flow; , cells
transient bubbles; --, large-scale flow with tilted plumes; r,heat flux transitions. The dashed curve
and transient bubbles (transitional); 0 ,
represents the transition from one to two hot spots circulating in each convection cell. (Reproduced with permission from Krishnamurti R
and Howard L (1981) Large-scale flow generation in turbulent convection. Proceedings ofthe National Academy of Sciences of the USA
78: 1981-1 985.)
convecting layer. Laboratory experiments confirm the downward or upward, respectively. Laboratory
predicted values of Ra, and the cell size. experiments confirm the existence of these plan
Weakly nonlinear theory has been employed to forms under the same circumstances. The theory and
investigate the stability of each of the possible plan
forms allowed by the linear theory. Consistent with the
results of laboratory experiments, the solutions show
that only two-dimensional rolls are stable to small
disturbances just above the critical Rayleigh number
Ra, in the absence of vertical asymmetries. Hexagonal
cells become the preferred plan form in the presence of
such asymmetries.
It is known from both theory and experiment that if
horizontal flow is superimposed, the convection
becomes aligned in rolls in the direction of the shear.
If rotation is superimposed, the horizontal dimensions
of the convection cells shrink.
In the Earth’s atmosphere, cloud patterns give
evidence of turbulent convection, clusters, two-di-
mensional roll convection, and three-dimensional
cellular convection under different conditions. Behind
maritime cold fronts, convection has been observed in
the form of both open and closed cells (Figure 3), the
Figure 3 Section of Air Force DAPP satellite photograph (2230
former having upward motion at the cell boundaries GMT 4 April 1973) of convective clouds, centered near 30” N and
and downward motion in the centers, and the latter the 61‘ W, showing open cells in a region of large-scale downward
reverse circulation. These observations have sparked motion and closed cells in a region of large-scale upward motion.
further experiments and theory that, although highly The region is roughly 500km on a side. (Reproduced with
permission from Shaugnessy JE and Wann TC (1973) Picture of
idealized, seem to have relevance to the observed cloud
the month-frontal rope in North Pacific. Monthly Weather Review
patterns. Weakly nonlinear theory predicts that three- 101: 774-776; Krishnamurti R (1975) On cellular cloud patterns,
dimensional open and closed hexagons, rather than Part 3: Applicability of laboratory and mathematical models.
rolls, are stable when the mean vertical motion is Journal of the Atmospheric Sciences 32: 1373-1 383.)
experiments do not, however, exhibit the aspect ratio been found below an Arctic ice island, as well as in
observed in the atmosphere, where the horizontal Antarctic lake Vanda and in other lakes, in which
scales are from 10 to 50 times the depth of the warm sea water from old intrusions is found below
convecting layer. Clouds are formed by the release of cold, fresh lake water.
latent heat of condensation, a complication not dealt In the Earth’s mantle, convection is shaped by the
with in Rayleigh-BCnard convection. More recent temperature dependence of the viscosity of the con-
experiments demonstrate that large cell sizes can be vecting material. The mantle behaves like an extreme-
realized by simulating the release of latent heat of ly viscous, low thermal conductivity fluid. Internal
condensation in clouds that form in an otherwise heating, accompanied by cooling at the surface,
stably stratified atmosphere. creates global-scale convection, resulting in the mo-
In the ocean, the water density is controlled by both tions of continents. Moreover, the core heats the
temperature and salinity, and thermal diffusion is lowest layer of the mantle, providing buoyancy and
much more rapid than salt diffusion. Convection lowering the viscosity locally. Evidence suggests that
under these circumstances is controlled by four convection from this source, under very high Rayleigh
dimensionless parameters, Ra, Pr (both defined earli- numbers, takes the form of narrow plumes which,
er), the ratio of the diffusivities of salt and temperature upon reaching the surface, form volcanic chain islands
z = I C ~ / I1)
C ( and
< the salinity Rayleigh number (e.g., the Hawaiian islands), among other features. An
Ras = gr(aS/az)H4/v~.Here y is the coefficient of example of plume convection modeled in the labora-
volume expansion of the fluid due to salt and S is the tory is shown in Figure 4.
salt concentration. When as/&<0 and aT/az > 0, In the Earth’s fluid core, rotation and electromag-
the density gradients due to salt and temperature are netic (Lorentz) forces play important roles, and the
both stably stratified, so no instability is possible. four dimensionless parameters needed to study con-
When aS/az > 0 and aT/az<O, both gradients are vection under such influences are the Rayleigh
unstably stratified, and convective instability is ob- number, the Prandtl number, the Taylor number (the
served. When either aS/az > 0 and i3T/az > 0, or square of the ratio of the Coriolis to the viscous force),
W/az < 0 and aT/az< 0, double diffusive instabilities and the Elsasser number (the ratio of the magnetic
are possible. Such instabilities occur when the net body force to the Coriolis force). Rotation renders the
density stratification is stable, but either the temper- motions nearly two-dimensional in planes normal to
ature is stably stratified and the salt is unstably the angular velocity Q which is not parallel to gravity,
stratified, or vice versa. and tends to make the cells narrower than they would
Theory and experiment confirm that when warm, be in its absence, Electromagnetic forces tend to make
salty water lies over cold, fresh water, with a net stable the cells wider. Buoyancy is created in part by thermal
density stratification, thin ‘salt fingers’ develop for effects and in part by compositional changes (solidi-
sufficiently large positive Ras (unstable) and small fication of iron and nickel, leaving a buoyant residual
negative Ra (stable), with rising fingers of cold, fresh liquid at the boundary with the inner core). Convec-
water and sinking fingers of warm, salty water. For tion takes place in a geometry of variable depth owing
very small negative Ra, the salt fingers become to the alignment of the convection cells within the
unstable and, below a comparatively thin layer, are spherical annulus. Modeling of all these effects in one
replaced by a well-mixed convection layer. If the fluid laboratory experiment would be very complicated and
is sufficiently deep, another salt finger layer forms is probably not possible. Separate experiments mode-
below the well-mixed layer. Within a range of values of ling thermal convection in a self-gravitating sphere or
the controlling parameters, reflecting, in particular, annulus, thermal convection under the simultaneous
small negative Ra and deep water, the instability takes influence of rotation and a magnetic field, and com-
the form of a number of convecting layers bounded positional convection show qualitative agreement
above and below by comparatively thin salt finger with theory and numerical simulation, but much of
layers. Evidence of such phenomena has been found in this work has not approached the parameter range of
the ocean (e.g., in the Mediterranean outflow). the core.
When cold, fresh water lies above warm, salty
water, with a net stable density stratification, theory
and experiment show that oscillatory instability takes Sloping Convection in Rotating,
place for sufficiently large positive Ra and sufficiently
small negative Ras. Once again, layering occurs in
Laterally Heated Fluids
sufficiently deep fluids, with well-mixed convective Pioneering experiments by Dave Fultz and Raymond
layers bounded above and below by stably stratified Hide beginning around 1950 demonstrated that it is
diffusive layers. Evidence of this phenomenon has possible to simulate basic features of the general
Figure 4 Plume convection produced by maintaining a reservoir of low-density fluid (water) below a more dense, highly viscous fluid
(corn syrup). Seen here are two thermals and one ‘starting plume’. The thermals entrain more dense fluid and decelerate. The plume
entrains buoyant fluid from the thin trailing filament leading back to the source and accelerates upward. (Courtesy of D. Loper, Geophysical
Fluid Dynamics Institute, Florida State University.)
atmospheric circulation, including the jet stream and predictions of analytic theory. Baroclinic growth is
-
the growth and propagation of mid-latitude, synoptic- achieved by the rising of warm fluid downstream of the
scale ( L 1000 km) baroclinic waves. In these exper- troughs and the sinking of cold fluid downstream of
iments, a liquid contained in an upright cylinder or
cylindrical annulus is rotated about the axis of the
cylinder, heated at the outer wall and cooled either at
the center or at the inner wall. Differential heating
establishes a radial temperature gradient, analogous
to the meridional temperature gradient in the atmos-
phere. Overturnings create a stable vertical density
stratification with Brunt-Vaisala (buoyancy) fre-
quency N. Two main dimensionless parameters con-
trol the instability and the scale of the baroclinic
waves: the Burger number B 3 N 2 H 2 / f 2 L 2(which
measures the competition between stratification and
rotation in determining the baroclinic wave scale) and
the Taylor number Ta = 12S2 x V/2/vV2V12.Here, f is
the Coriolis parameter and N 2 G ga/3Taz > 0. The
growth rate of the disturbances is proportional to the
ratio of the destabilizing effect of vertical shear to
the stabilizing effect of stratification laV/azl/N,
the square of which is the inverse of the Richardson
number. While fluid motions in the cylinder
(or ‘dishpan’) experiments look more like those
5 Plan view of streamlines (solid curves), isotherms
seen on upper tropospheric weather maps, the more Figure (dashed curves), and vertical velocity contours (dotted curves with
regular wave patterns in the annulus experiments shading for upward motion) at mid-depth in an annulusexperiment.
possess the same essential ingredients and are easier to The inner and outer cylinders are held at uniformly cold and warm
interpret. temperatures, respectively. The annulus is rotated counterclock-
Figure 5 shows the streamlines, isotherms, and wise. The pattern travels counterclockwise relative to the annulus.
A meandering jet stream, following the most closely spaced
vertical velocity contours corresponding to a 4-wave streamlines, flows counterclockwise through the pattern. (Repro-
pattern in an annulus experiment. The relationships duced with permission from Pfeffer RL, Buzyna G and Fowlis WW
among these variables are similar to those in mid- (1974) Synoptic features and energetics of wave amplitude
latitude developing baroclinic waves and confirm the vacillation. Journat of the Atmospheric Sciences 31 : 622-645.)
the ridges in the stream field. Away from the bound- thermal Rossby number ROT gaHAT/4a2AR (in
aries, the motions are quasigeostrophic and hydro- which the annular gap-width AR is taken as the length
static, implying thermal wind balance and the scale and the imposed thermal wind as the velocity
existence of a jet stream above the strongest temper- scale) reflects the fact that the imposed temperature
ature gradients. contrast AT controls both the Burger and Richardson
The slope of the azimuthal mean potential temper- numbers.
ature surfaces in the radial plane is such that the Outside the curve containing the baroclinic wave
coldest fluid is found at the bottom near the cold regime, the motions are axisymmetric, resembling a
source and the warmest fluid at the top near the heat Hadley cell with rising and sinking motions near the
source. A parcel stability argument reveals that insta- warm and cold walls, respectively, and fluid spiraling
bility can take place only if fluid parcels are displaced inward in the upper layer and outward in the lower
within the angle between the geopotentials and the layer. The cutoff of wave instability at large ROTis due
potential temperature surfaces (called the Eady angle, to the fact that the preferred wave scale for instability
after the scientist who elucidated this in his 1949 is larger than that which can fit within the annulus.
theory of baroclinic instability). According to linear The cutoffs at small Ta and ROT are due to the
theory, this can take place only at horizontal scales dominance of dissipation for small wave scales and/or
larger than a critical value that depends on B. Shorter rotation rates. Within the regular wave regime, the
waves have parcel motions outside the Eady angle. The observed wave scales correspond to the predicted
fastest-growing waves are those for which the parcel value A, becoming smaller as AT is decreased and/or 0
-
motions follow the bisector of this angle. The scale
of these waves is L 3. = N H / f (called the Rossby
radius of deformation), corresponding to B 1. -
is increased. At lower AT (i.e., smaller N 2 )and higher
!2, where nonlinearity becomes important in the more
irregular flow, the observed scales are generally larger
Longer and shorter waves have slower growth rates than predicted. Two kinds of time dependence are
because their parcel paths have steeper or less steep observed in these experiments: amplitude and struc-
slopes. tural vacillation.
Figure 6 is a sketch of the regime diagram showing Amplitude vacillation is a nearly periodic growth
the various wave scales and forms of time dependence and decay of waves (Figure 7A-D). Axisymmetric
observed in annulus experiments. The use of the overturnings near the side walls set up large temper-
ature gradients across the middle of the annular gap.
Baroclinic instability sets in, leading to explosive wave
I Free
symmetric
growth. The waves transport heat across the annular
gap and induce a thermally indirect Ferrel cell between
the thermally direct meridional cells near the walls.
The heat transport is so great that it reduces the
horizontal temperature gradient below the critical
value needed to overcome dissipative effects, and the
waves die out. When the waves become sufficiently
weak, axisymmetric overturnings reestablish the ra-
dial temperature gradient and the cycle begins anew.
Although not periodic, striking cases of a similar
process of hemispheric baroclinic wave growth and
decay in middle latitudes in winter have been observed
in the Earth’s atmosphere (Figure 7E, F).
Structural vacillation is characterized by time-
dependent changes, in the wave shapes with negligible
0.01 ‘
107
I
108
I
109
I
10’0
changes in wave amplitude. As ROTdecreased and Ta
is increased, this form of vacillation first appears as a
weak, almost periodic, oscillation. In some experi-
Ta
ments, it manifests itself as a periodic tilting of the
Figure 0 Sketch of the regime diagram for baroclinic flow in an troughs and ridges such that they transport angular
annulus with gap width twice the fluid depth, depicting the most momentum first inward and then outward (Figure 8).
probable wave scales and time dependence as a function of R q
and Ta at f r = 21. The upper shaded region shows where
In others it manifests itself as an energy oscillation in
amplitude vacillation is most prevalent. The unshaded region the radial direction. With further decreases of ROT and
immediately below it indicates where structural vacillation is most increases of Ta, the oscillations become modu-
common. The details differ with aspect ratio and Prandtl number. lated and intermittent in time, finally giving way to
Figure 7 (A, B) Upper-level wave patterns and (C, D) mid-depth temperature fields in a thermally driven rotating laboratory experiment, depicting minimum and maximum wave amplitudes at
opposite phases of an amplitude vacillation cycle. (E,F) 500 hPa height charts depicting opposite phases of an atmospheric ‘index cycle’. (Reproduced with permission from Pfeffer RL and
ChangY (1967)Two kindsof vacillation in rotating laboratoryexperiments.Month& WeatherReview95:75-82;Pfeffer RL, BuzynaG and FowlisWW (1974)Synoptic features and energeticsof
wave amplitude vacillation. Journal of the Atmospheric Sciences 31 : 622-645.
Figure 8 A sequence of flow patterns depicting tilted trough vacillation in a thermally driven rotating annulus of fluid. (Courtesy of
Professor D. Fultz, University of Chicago.)
fully-developed geostrophic turbulence. There is good varies linearly with scale rather than quadratically as in
evidence to suggest that structural vacillation repre- the formula for Rossby waves in the atmosphere.
sents a form of low-order deterministic chaotic Variations of the annulus experiment have been
behavior, with the order increasing as one moves used to study other influences that affect atmospheric
toward the region of geostrophic turbulence. behavior. Among these are bottom topography, land-
In addition to the two forms of vacillation, wave sea temperature variations, cyclic variations of the
dispersion has been observed near the boundaries imposed temperature contrast (simulating seasonal
between wavenumbers. In such experiments, two, and forcing) and the effect of internal heating in the
sometimes three, adjacent wavenumbers are observed, atmospheres of the major planets on the formation
with phase speeds that increase with decreasing scale. It and maintenance of large detached vortices (e.g.,
has been speculated that this phenomenon is associated Jupiter’s Great Red Spot and related features on Saturn
with the slope of the free surface due to rotation, which and Neptune). Additional studies of baroclinic fluid
can act in part like the variation of the Coriolis behavior have been performed with two homoge-
parameter in the atmosphere, although the wave speed neous, incompressible fluids driven by a differentially
rotating lid. Here the baroclinicity is concentrated at

the interface between the two fluids. Other experi-
ments designed to account for detached vortices have
been performed with mechanically driven, rotating
barotropic fluids at high Reynolds numbers (the ratio
of advection to viscous forces), high Taylor numbers,
and low Rossby numbers. Such experiments have also
been used to study the properties and stability of the
Earth's stratospheric polar night jet.
One important property of synoptic-scale waves in
the atmosphere that has not yet been successfully
modeled is the meridional and vertical propagation of
such waves, made possible by the presence of a
background potential vorticity gradient, the absence
of lateral walls, and the unlimited upward extent of the
atmosphere. Unlike atmospheric waves, annulus
waves are trapped radially and vertically. Innovative
experiments attempting to address some of these issues
have not yet borne fruit.
Other Experiments
Numerous other geophysical fluid dynamics experi-
ments have been conducted to illustrate fundamental
theorems or explain observed phenomena: for exam-
ple, the Taylor-Proudman theorem (which states that
rapid rotation characterized by small Ro and large Tu
makes the flow invariant in the direction of a);
Kelvin-Helmholtz instability (instability generated by
a sufficiently large velocity shear across the interface Figure 9 Taylor vortices in a tall cylinder of fluid produced by
between a lower density fluid and an overlying higher bringing the fluid into solid rotation and then stopping the rotation.
The fluid near the wall slows down while that away from the wall
density fluid); axisymmetric inertial instability (the continues to spin, creating an unstable velocity profile. Courtesy of
instability of a circular flow when the square of the R Kung and the author, Geophysical Fluid Dynamics Institute,
angular momentum decreases outward and Ta ex- Florida State University.
ceeds a critical value: Figure 9);various routes to chaos
as a control parameter, such as rotation, is increased; with a block of ice, and the top is maintained at a
and critical layer absorption (the absorption of neutral temperature well above 4"C, the layer below the 4°C
linear waves near a surface along which their phase isotherm will have an unstable density stratification,
speeds in the direction of the background flow equal which will lead to convection, and the layer above will
the background flow speed). Kelvin-Helmholtz insta- have a stable density stratification.
bility is responsible for the formation of billow clouds Western boundary currents have been modeled in a
and wind-driven water waves. Critical layer absorp- rotating cylinder of fluid in which the base is a sloping
tion plays a crucial role in the quasi-biennial oscilla- plane that creates variable depth. It is readily shown
tion in the equatorial stratosphere. that, in an unstratified fluid, the variable depth has the
Other experiments model phenomena such as top- same effect on the flow as the latitudinal variation of
ographic waves in stratified fluids, penetrative con- the Coriolis parameter has on the atmosphere and the
vection in the atmosphere (where tropospheric ocean. Experiments and theory confirm that this is the
convection penetrates into the stably stratified strat- crucial ingredient in the formation of western bound-
osphere), and the strong boundary currents found ary currents.
along the western shores of most oceans (e.g., the Gulf Although the present article has by no means
Stream, the Kuroshio Current, the Brazil Current). covered all types of geophysical fluid dynamics exper-
Penetrative convection has been modeled by making iments that have been performed, it should give the
use of the property that the maximum density of water reader sufficient familiarity with the type of research
occurs around 4°C. If the bottom of a container of that is done and the factors that are considered when
water is maintained at O'C, by placing it in contact designing such experiments.
1090 LABORATORY KINETICS
See also (eds) Dynamics of Atmospheres and Oceans, Special

Issue 27(3): 243-256.
Baroclinic Instability. Cyclogenesis.Dynamic Mete-
Krishnamurti R (1997) Convection induced by selective
orology: Overview. Instability: Inertial Instability; Sym-
absorption of radiation. A laboratory model of condi-
metric Stability; Wave-CISK. Jet Streaks. Kelvin-
tional instability. In: Navon IM, Kalnay E and Stone PH
Helmholtz Instability. Middle Atmosphere: Quasi-
(eds) Dynamics of Atmospheres and Oceans, Special
Biennial Oscillation. Ocean Circulation: Surface-Wind
Issue 27(3): 367-382.
Driven Circulation. Planetary Atmospheres: Jupiter and
Loper DE (1997) Mantle plumes and their effect on the
the Outer Planets. Predictability and Chaos. Quasi- Earth’s surface: a review and synthesis. In: Navon IM,
geostrophic Theory. Wave Mean-Flow Interaction.
Kalnay E and Stone PH (eds) Dynamics of Atmospheres
and Oceans, Special Issue 27(3):35-54.
Pfeffer RL, Kung R, Applequist S, Long C and Buzyna G
Further Reading (1998)Progress in characterizing the route to geostrophic
Busse FH, Hartung G, Jaletzky M and Sommermann G turbulence and redesigning thermally-driven rotating
(1997) Experiments on thermal convections in rotating annulus experiments. Theoretical and Computational
systems motivated by planetary problems. In: Navon IM, Fluid Dynamics 9: 253-57.
Kalnay E and Stone PH (eds)Dynamics of Atmospheres Read PL (1993) Coherent baroclinic waves in a
and Oceans, Special Issue 27 (3): 161-174. rotating, stably-stratified fluid and transitions to
Chandrasekhar S (1961)Hydrodynamic and Hydromagnet- disordered flow. In: Mobbs SD and King JC (eds) Waves
ic Stability, London: Oxford University Press. and Turbulence in Stably Stratified Flows. Oxford:
Cushman-Roisin B (1994) Introduction to Geophysical Clarendon Press.
Fluid Dynamics. Englewood Cliffs, NJ: Prentice-Hall. Sommeria J, Meyers SD and Swinney HL (1991) Experi-
Drazin PG and Reid WH (1981) Hydrodynamic Stability. ments on vortices and Rossby waves in eastward and
Cambridge: Cambridge University Press. westward jets. In: Osborne AR (ed.) Nonlinear Topics in
Fultz D (1962) An experimental view of some atmospheric Ocean Physics, pp. 227-269. Amsterdam: North-Hol-
and oceanic behavioral problems. Transactions of the land.
New York Academy of Sciences, Ser. II. 24(4):421-446. Stern ME (1975) Ocean Circulation Physics. New York:
Greenspan HP (1968) The Theory of Rotating Fluids. Academic Press.
Cambridge: Cambridge University Press. Stern ME and Radko T (1997)Maintaining the inshore shear
Hart JE, Adler B and Leben R (1997) Cyclonic/anticyclonic of continental boundary currents. In: Navon IM, Kalnay
gyre asymmetries: laboratory and intermediate-model E and Stone PH (eds) Dynamics of Atmospheres and
experiments. In: Navon IM, Kalnay E and Stone PH (eds) Oceans, Special Issue 27(3): 662-678.
Dynamics of Atmospheres and Oceans, Special Issue Swinney HL and Gollub JP (eds) (1981) Hydrodynamic
27(3):219-232. Instabilities and the Transition to Turbulence. Berlin:
Hide R (1997)On the effects of rotation on fluid motions in Springer-Verlag.
cylindrical containers of various shapes and topographic Turner JS (1973) Buoyancy Effects in Fluids. Cambridge:
characteristics. In: Navon IM, Kalnay E and Stone PH Cambridge University Press.
D J Donaldson,University of Toronto, Toronto, Ontario, neutral fragments of molecules containing a n un-

Canada paired electron a n d are generally very reactive. Some
Copyright 2003 Elsevier Science Ltd. All Rights Reserved important atmospheric examples include OH, HOz,
H C O , CH3, a n d CH30. Stable molecules such as 03,
NO a n d NO2 also contain unpaired electrons, which
Introduction can influence their reactivity.
Chemical kinetics is the study of the rates of transfor-
mation of chemical compounds from reactant species
Principles of Chemical Kinetics
into products. T h e rate of a reaction is defined to be the
rate of decrease with time of the reactant concentra-
Overall and Elementary Reactions
tion (in number of moles o r molecules per unit volume)
due to chemical reaction(s),or, equivalently, the rate of The rate is expressed in terms of one mole of reaction,
increase of the product concentration. In the atmos- so rates of concentration change are normalized to the
phere, chemical reactions typically involve free radi- reaction stoichiometric coefficients, which give
cals, as reactants, products o r both. Free radicals are the number of moles of each compound appearing in
See also (eds) Dynamics of Atmospheres and Oceans, Special

Issue 27(3): 243-256.
Baroclinic Instability. Cyclogenesis.Dynamic Mete-
Krishnamurti R (1997) Convection induced by selective
orology: Overview. Instability: Inertial Instability; Sym-
absorption of radiation. A laboratory model of condi-
metric Stability; Wave-CISK. Jet Streaks. Kelvin-
tional instability. In: Navon IM, Kalnay E and Stone PH
Helmholtz Instability. Middle Atmosphere: Quasi-
(eds) Dynamics of Atmospheres and Oceans, Special
Biennial Oscillation. Ocean Circulation: Surface-Wind
Issue 27(3): 367-382.
Driven Circulation. Planetary Atmospheres: Jupiter and
Loper DE (1997) Mantle plumes and their effect on the
the Outer Planets. Predictability and Chaos. Quasi- Earth’s surface: a review and synthesis. In: Navon IM,
geostrophic Theory. Wave Mean-Flow Interaction.
Kalnay E and Stone PH (eds) Dynamics of Atmospheres
and Oceans, Special Issue 27(3):35-54.
Pfeffer RL, Kung R, Applequist S, Long C and Buzyna G
Further Reading (1998)Progress in characterizing the route to geostrophic
Busse FH, Hartung G, Jaletzky M and Sommermann G turbulence and redesigning thermally-driven rotating
(1997) Experiments on thermal convections in rotating annulus experiments. Theoretical and Computational
systems motivated by planetary problems. In: Navon IM, Fluid Dynamics 9: 253-57.
Kalnay E and Stone PH (eds)Dynamics of Atmospheres Read PL (1993) Coherent baroclinic waves in a
and Oceans, Special Issue 27 (3): 161-174. rotating, stably-stratified fluid and transitions to
Chandrasekhar S (1961)Hydrodynamic and Hydromagnet- disordered flow. In: Mobbs SD and King JC (eds) Waves
ic Stability, London: Oxford University Press. and Turbulence in Stably Stratified Flows. Oxford:
Cushman-Roisin B (1994) Introduction to Geophysical Clarendon Press.
Fluid Dynamics. Englewood Cliffs, NJ: Prentice-Hall. Sommeria J, Meyers SD and Swinney HL (1991) Experi-
Drazin PG and Reid WH (1981) Hydrodynamic Stability. ments on vortices and Rossby waves in eastward and
Cambridge: Cambridge University Press. westward jets. In: Osborne AR (ed.) Nonlinear Topics in
Fultz D (1962) An experimental view of some atmospheric Ocean Physics, pp. 227-269. Amsterdam: North-Hol-
and oceanic behavioral problems. Transactions of the land.
New York Academy of Sciences, Ser. II. 24(4):421-446. Stern ME (1975) Ocean Circulation Physics. New York:
Greenspan HP (1968) The Theory of Rotating Fluids. Academic Press.
Cambridge: Cambridge University Press. Stern ME and Radko T (1997)Maintaining the inshore shear
Hart JE, Adler B and Leben R (1997) Cyclonic/anticyclonic of continental boundary currents. In: Navon IM, Kalnay
gyre asymmetries: laboratory and intermediate-model E and Stone PH (eds) Dynamics of Atmospheres and
experiments. In: Navon IM, Kalnay E and Stone PH (eds) Oceans, Special Issue 27(3): 662-678.
Dynamics of Atmospheres and Oceans, Special Issue Swinney HL and Gollub JP (eds) (1981) Hydrodynamic
27(3):219-232. Instabilities and the Transition to Turbulence. Berlin:
Hide R (1997)On the effects of rotation on fluid motions in Springer-Verlag.
cylindrical containers of various shapes and topographic Turner JS (1973) Buoyancy Effects in Fluids. Cambridge:
characteristics. In: Navon IM, Kalnay E and Stone PH Cambridge University Press.
D J Donaldson,University of Toronto, Toronto, Ontario, neutral fragments of molecules containing a n un-

Canada paired electron a n d are generally very reactive. Some
Copyright 2003 Elsevier Science Ltd. All Rights Reserved important atmospheric examples include OH, HOz,
H C O , CH3, a n d CH30. Stable molecules such as 03,
NO a n d NO2 also contain unpaired electrons, which
Introduction can influence their reactivity.
Chemical kinetics is the study of the rates of transfor-
mation of chemical compounds from reactant species
Principles of Chemical Kinetics
into products. T h e rate of a reaction is defined to be the
rate of decrease with time of the reactant concentra-
Overall and Elementary Reactions
tion (in number of moles o r molecules per unit volume)
due to chemical reaction(s),or, equivalently, the rate of The rate is expressed in terms of one mole of reaction,
increase of the product concentration. In the atmos- so rates of concentration change are normalized to the
phere, chemical reactions typically involve free radi- reaction stoichiometric coefficients, which give
cals, as reactants, products o r both. Free radicals are the number of moles of each compound appearing in
LABORATORYKINETICS 1091
the balanced chemical equation. Thus for the reaction The reaction mechanism is a sequence of elementary
[I], in which a, b, c, and d represent the stoichiometric chemical reactions, each of which occurs by a single
coefficients for their respective compounds in the interaction between reagents. This sequence of reac-
balanced chemical equation, the rate is defined by tions, taken together, must give rise to the observed
eqn H I . rate expression. The rate expressions for elementary
reactions are easily written from their balanced
aA+bB+cC+dD 111 chemical equations; for example, the elementary
reaction [III], which is the first step in the atmospheric
oxidation of methane, has a rate given by eqn [4], since
the reaction takes place by the interaction of one O H
molecule with one methane molecule.
OH + CH4 -+ H20 + CH3, [I111
This expression gives the phenomenological rate at
which reactants A and B are transformed to products C d[CH41
dt = -k[CH4] [OH] 141
and D. It contains no information concerning the
mechanism of the reaction, nor does it have any Elementary reactions generally involve atoms, free
predictive utility. radicals or other highly reactive species.
Since most chemical reactions involve an exchange
of atom(s) among the reagents, it is intuitive that such Elementary Chemical Reactions: General Features
processes should proceed via collisions, or at any rate, If a reaction is known to be elementary, the explicit
close approaches of the reagents to one another. Since time dependence of the concentration of all chemical
the likelihood of such collisions increases with number species involved may be obtained through integration
density (concentration), the rate is often expressed in of the rate expression. A few instances are of particular
terms of the concentrations of the compounds in- interest in atmospheric chemistry.
volved (eqn [2]), where k represents a concentration-
independent factor (which may depend on other First-order reactions These are reactions in which
parameters; see below). the exponents in the rate expression sum to unity. This
class of systems includes photochemical transforma-
Rate = k[A]W[B]X[C]y[D]Z 121 tions and unimolecular reactions (in the high-pressure
The exact form of this expression must be determined limit). For example, the rate expression for photode-
experimentally; any of the species present in the composition of formaldehyde in the near ultraviolet,
reaction system may appear and the exponents w-z shown in reaction [IV], is given by eqn [5].
may or may not bear any relationship to the stoichio- H2CO + hv -+ H + HCO [IVI
mentric coefficients a-d. For example, the low tem-
perature oxidation of methane to carbon dioxide and
water, whose balanced equation is given by [11], has a d[H2C01 = -k(T,
hv)[H2CO]
dt 151
rate expression that may be approximated by eqn [3],
where [MI represents the total concentration of all In eqn [5] the rate constant depends on the photon
species present, and k and k’ are two different energy, hv, as well as temperature, T . This is easily
constants. integrated to yield eqn [6].
CH4 +202 C02 +2H20 [H2CO],= [H2CO]oe-kt
This represents a simple exponential decay from the

d[CH41- -k[CH4][M]{5(e4k’io2!i- 1))
-- [3] initial concentration, [H2C0I0.
dt
This rate expression is clearly not intuitively derived Second-order reactions These include bimolecular
from the balanced reaction and indicates that the reactions between different reagents, as well as self-
reaction does not proceed, at a molecular level, by the reactions. The rate expression depends upon the
simultaneous interaction of a methane molecule with instantaneous concentrations of both species involved
two oxygen molecules. A more complex process, in the reaction. For example, the reaction of O H
involving many individual reaction steps, is respon- with methane (reaction [111] above) is second order.
sible for the observed rate expression. The time dependence of reactant concentrations is
somewhat more complicated than that for first-order Sequences of Elementary Reactions
reactions, as shown by eqn [ 7 ] . As mentioned earlier, the overall reaction mechanism
is constructed from a sequence of elementary reac-
tions. These elementary reactions must combine to
yield the observed reaction rate expression. Often in
In laboratory studies of bimolecular reactions, the such a sequence, a species X is formed as the product of
concentration of one reagent is typically forced to be in one reaction, and consumed in a subsequent reaction.
great excess, and so remains essentially constant It can be shown that, if the reaction consuming X
concentration during the reaction. For example, if occurs much more rapidly than its formation, the
[CH4] >> [OH],, [CH4], = [CH4I0, and the rate of concentration of X will be small and almost time
change of the O H concentration can be expressed as in invariant. Here, the rate of formation of X is said to be
eqn 181. the ‘rate-limiting step’ in the reaction sequence, since
X is consumed as soon as it is formed. Under these
d[oH1 = - k(T)[CH4Io[OH] circumstances, the concentration of X is said to be in
dt ‘steady state’. Since the majority of atmospheric
= k’(T)[OH] PI radicals are highly reactive, the analysis of their steady
Here k’(T) = k(T)[CH4],. In this case, the reaction is state concentrations becomes important.
said to exhibit pseudo-first-order behavior, and the Consider the sequence of reactions given below
time dependence of the O H concentration is given ([VIII-[XI), which represent a simplified version of the
‘oxygen only’ chemistry of ozone.
by ~91.
[OH],= [OH],, e-k’(T)t [91
0 2 + hv -+ 0 + 0 [VI11
The true bimolecular rate coefficient is then obtained

by performing measurements at many different values
of [CH4I0, then plotting the observed k’(T) as a
function of JCH41,. The slope of such a plot yields
k(T)* 0 3 + hv 0 2 +0
-+ [XI
Third-order reactions In the atmosphere, many The rate of change of the ozone concentration is the
elementary reactions require the participation of difference between the rates of its formation, due to
three molecules, and are thus strictly third order. reaction [VIII], and its consumption, through [IX] and
Important reactions such as [VI and [VI] require the [XI. The rate of change of concentration the oxygen
participation of a ‘third body’ (usually designated as species is likewise given by the difference between the
M ) to remove the excess energy in the newly formed rates of formation and destruction (eqns [l11and [12]).
product.
O H + NO2 +M -+ HON02 +M [VI
d[031 -
--
dt
kVIII[OI 1021[MI - kIX[OI [ 0 3 1
- kx[03] I111
CH3 + 0 2 +M -+ CH300+M [VI1
In the atmosphere, M is generally N2 and 0 2 , since
d[Ol = 2kv11[02]+ kx[03] - kv111[0][02][Ml
these gases constitute approximately 99% of the dt
atmosphere. The rate expressions of such reactions -~ I 1
011031
X 1121
are properly written as eqn [lo], in which the
participation of the third body is made explicit. If the concentrations of these reactive species are time
invariant, the rates of formation and destruction are
equal and the steady state concentrations of each may
be calculated by eqns [13] and [14].
At the higher pressures present. in the lower atmo-
sphere, some such reactions are often at their high-
pressure limits: collisions with M are frequent enough
that the reaction exhibits pseudo-second-order kinet-
ics, with the concentration of M incorporated into the
rate constant.
LABORATORYKINETICS 1093
The measured atmospheric concentrations of oxygen- reactants as an unsuccessful collision. If the activated
atoms and ozone are fairly constant over a time scale complexes are held to be in rapid quasi-equilibrium
of several hours, under constant illumination from the with reagent species, with product formation being
Sun. The analysis given above, although crude, pro- rate limiting, a statistical mechanical analysis yields
vides a useful picture for how these species’ concen- the rate constant in terms of the ‘partition functions’,
trations depend upon altitude (through the total Q , of the reactants (QAand Q B )and the activated
pressure, related to [MI and the altitude dependence complex (designated Q # ) .This can be expressed by
of solar ultraviolet intensities, through kvII and k x ) . eqn [16], where kg is the Boltzmann constant (equal to
the gas constant divided by the Avogadro number) and
Temperature Dependence of Elementary Chemical h is the Plancks constant.
Reactions
The rates of elementary reactions depend upon the
reagent concentrations, as discussed above, but also
on other parameters, most importantly on tempera-
ture. The T dependence of reaction rates is incorpo- The partition functions depend only upon molecular
rated into k , the rate constant (more properly, the rate properties, such as bond lengths and angles and
coefficient). Empirically, it is often found that over the vibrational energies, which may be determined spec-
temperature range important in the lower and middle troscopically or calculated theoretically, allowing
atmosphere k depends on T exponentially, as shown in calculation of the rate constant for the reaction using
eqn [15]. eqn [16].
This dependence of the rate constant on molecular
properties provides an explanation as well for ob-
served isotope effects in reaction rates. Often it is
In this expression, E , represents the activation energy, found that reactions of chemically identical, but
R is the gas constant (8.314JK-lmol-l), T is the isotopically different, species will exhibit different
temperature in Kelvin, and A is unimaginatively rates, with the heavier isotope displaying the smaller
named the pre-exponential factor. Careful measure- rate constant. The smaller vibrational frequencies, and
ments over wide enough temperature ranges reveal thus lower zero-point energy, of the heavier isotope
that the pre-exponential factor depends weakly on give rise to somewhat larger reagent partition func-
temperature, so is strictly written as A ( T ) . The tions and higher values of Eo, and hence a smaller rate
dependence is generally much weaker than exponen- constant.
tial, however, and a temperature-independent value
for A is often used. This empirical expression may be
interpreted simply as follows: For a chemical reaction Methods for Measuring Atmospheric
to occur, the reagents must collide (i) with sufficient Rate Parameters
energy to overcome any energy barrier(s) along the
Extracting Useful Parameters from Experiments
reaction path, and (ii) in an appropriate geometry to
facilitate product formation. The pre-exponential A host of methods is in current use for measuring gas-
factor contains information concerning the collision phase reaction rates. The specific technique employed
rate (dependent upon the reagent velocities, and hence in any particular case depends somewhat on the
temperature) and any geometric constraints on the particulars of the reaction being studied, and also on
reaction. The exponential term arises from consider- the available apparatus in any given laboratory.
ation of the fraction of collisions, under conditions of As may be inferred from the foregoing discussion,
thermal equilibrium, which possess energy in excess of the object of almost all kinetics experiments is to
some threshold value, Eo, This energetic threshold is determine the reaction rate coefficient, preferably as a
assumed to arise from the presence of energetic function of temperature, and also, in the case of
‘barriers’ along the reaction path, due to the energetic termolecular and unimolecular reactions, of pressure.
cost(s) of rearranging the chemical bonds. The activa- Very often this is done by following the concentration
tion energy is closely related to threshold energy; for of some reactant or product as a function of time,
our purposes they may be taken as equivalent. generally under pseudo-first-order conditions. The
An alternate interpretation considers the elementa- inverse of the (single exponential) time constants
ry reaction to occur in two steps: In the first, the crest of obtained from fits to plots of concentration against
the energetic barrier is achieved; an activated complex time are then plotted as a function of the concentration
(or transition state) is reached. From this point, the of the reagent held in excess. This yields the true
reaction may continue to products, or return to second-order rate coefficient at a given temperature.
Figure 1 illustrates this situation for a hypothetical decays for successively higher RH concentrations. All
reaction between O H and an unspecified hydrocarbon three exhibit single exponential decays, as required for
molecule, RH, at room temperature. The reaction of straightforward pseudo-first-order analysis. Figure 1B
O H with propane, given as reaction [111] above, is an shows the exponential decay constants obtained from
example of such an O H + RH reaction. Figure 1A data like those in Figure lA, plotted as a function of
shows three plots of the decay of O H as a function of RH concentration; the slope of this linear fit gives the
time, following its creation at time t = 0. The O H bimolecular rate coefficient at room temperature.
concentration is at all times much smaller than that of Note the existence of a finite (positive) intercept,
RH, ensuring that pseudo-first-order conditions are indicating a finite rate of disappearance of O H in the
maintained. Curve (i) displays the decay for a low absence of RH. This loss of reagent may be due to a
concentration of RH; curves (ii) and (iii) display combination of wall reactions, reactions of O H with
impurities, and self-reaction of OH.
Modern experiments yield measurement uncertain-
ties of 5-10% on the rate coefficient at any given
temperature. Uncertainties are generally larger at
lower temperatures, where the reaction rate is smaller
for most reactions. The agreement between different
laboratories is usually within 10-20% for uncompli-
cated reactions.
Experimental Techniques
Clearly, the technical issues in measuring reaction
rates are: (a)to generate reactive species at some well-
defined time t = 0; (b) to measure accurately the
reactant (or product) concentrations as a function of
time, over the course of the reaction; and (c) to
minimize losses of reagent due to processes other than
the reaction of interest. Implicit in (b) is the absolute
identification of the species whose concentration is
being followed. A later section will briefly discuss
spectroscopic methods of detection.
Flow tubes There are two techniques in common use

for fairly fast reactions (i.e., those with rate coeffi-
cients greater than 10-15-10-16 cm3 mol-' s-'). The
flow tube method has several variants, all of which
share the same basic principles. Reacting gases are
entrained in a fast flow of an inert gas (He or Nz) at low
pressure (a few mbar) passing down a cylindrical tube
at constant velocity. One of the reagents is introduced
into the flow through a concentric inner tube (the
injector), whose longitudinal position may be varied.
The zero-of-time is established when the reagents first
come into contact with one another; therefore, varying
the position of the injector also varies the exposure of
(6) RH concentration the reagents to one another along the length of the flow
Figure 1 Extraction of a bimolecular rate coefficient from tube. Since the reacting mixture moves down the flow
experimental pseudo-first-order reaction conditions, using the tube at constant (known) velocity, contact length is
+ +
OH RH --t H 2 0 R reaction as an example. (A) Decay of OH related to exposure time. The disappearance of
as afunction of exposure time to RH. Curves (i), (ii) and (iii) display reagent(s) and/or growth of product is measured, for
the result for increasing RH concentration. (6)The pseudo-first-
example, by optical spectroscopy or mass spectrome-
order rate coefficient obtained from fits to data such as that shown
in (A), plotted as a function of the (constant) RH concentration. The try as a function of injector position (and hence
slope of the linear fit to these data gives the true second-order rate exposure time) in a volume near the end of the
coefficient. flow tube, before the gas flow enters the pump.
LABORATORY KINETICS 1095
Figure 2 illustrates diagramatically a typical flow tube be measured over a wider pressure regime than
apparatus. hitherto.
In such methods, reactive atomic or radical species
are generated continuously, most often by passing a Flash photolysis Another commonly used method
precursor gas through a plasma generated by a also has several variants; all are based on the early
microwave or radiofrequency discharge. This gener- flash photolysis work of Norrish and Porter.
ates atomic reagents directly (i.e., 0 or H atoms fromReactive species are generated in a reaction volume
0 2 or H2, respectively); these may then react to formby photodissociating a suitable precursor molecule
radicals, via processes such as reactions [XI], [XII] or
using a short ( I1ps) pulse from a laser or flash-
[XIII], when additional coreactants are introduced lamp (almost always a laser nowadays). For example,
into the flow tube. methyl radicals are generated efficiently by the
photolysis of acetone using the 1011s pulse from a
+
H 0 2 + M + HO2 + M 1x11 193 nm wavelength ArF excimer laser (reaction
[XIVl).
+
H + 0 3 ( 0 r N O z ) t O H 02(0r NO) [XI]
(CHj),CO hv + 2CH3 CO + [XIVI +
+
F CH4 + H F CH3 + [XIII] Similarly, O H may be formed via the 248nm KrF
The radical concentrations are generally between one excimer laser photolysis of nitric acid (reaction [XV]).
and three orders of magnitude smaller that those of the
molecular reagents, so that a pseudo-first-order kinetic
HN03 + hv + O H + NO2 [XVI
analysis is often possible with ‘normal’ concentrations The precursor species is present in a well-mixed gas
of the coreactant. mixture, which also contains the second reagent and
Recent developments in this technique include the generally a buffer gas as well. The buffer gas is present
ability to measure simultaneously the concentrations to thermalize the photodissociation products prior to
of two radical reagents (and thus to determine radical- reaction. Unless radical-radical reactions are being
radical reaction rates) and the use of higher pressure studied, the photolysis conditions are maintained such
flow tubes (up to several hundreds of mbar), which that the concentration of radicals remains much smaller
allow recombination and other termolecular rates to than that of molecular reagents. The zero-of-time
Figure 2 Schematic of a flow tube apparatus. Reagents and a carrier gas are introducedfrom the gas reservoirs (GR) into the flow tube.
One reagent is introducedthrough the movable injector tube (Inj). The other gas flow passes through a microwavedischargecavity (MW),
where an atomic or radical reagent is formed from its precursor compound. The concentration of radical reactant (or of product) may be
monitored using the tunable diode laser (TDL)-detector (D) combination. The total pressure in the flow tube is monitored at point P.
is established by the pulse of light that creates the Spectroscopic Detection Methods
reactive species, and the decay of the radical concen-
Here we present a brief overview of the various
tration (or the growth of product species concentra-
methods in common use at the time of writing. All
tion) is followed as a function of time following that
spectroscopic methods rely upon the resonant absorp-
pulse. Measurements are typically made using time-
tion or emission of radiation in a wavelength region
resolved optical spectroscopic methods such as tran-
characteristic of the species being detected. The
sient absorption or laser-induced fluorescence (LIF)
absorption and emission of radiation by atoms and
spectroscopy. These methods are discussed below.
molecules only occurs in particular spectral regions,
Experiments using LIF detection (often called the
corresponding to the energy differences between
pump and probe method), there is a variable delay
quantum levels. The amount of radiation absorbed
time between the firing of the radical-generating laser
by a sample of molecules or atoms at a particular
and the firing of the probe LIF laser, allowing a
wavelength 2 depends upon an absorption intensity
variable exposure time of the reagents to one another
factor &(A),which is specific to the atom or molecule
prior to their interrogation. Following each pulse-
and is somewhat dependent upon the temperature, the
probe cycle, the reaction volume may be replenished
path length of interaction between the light and the
with fresh reactants.
sample, 1, and the concentration of absorbing species,
Recent developments in this method include the
c. Under appropriate experimental conditions the
application of time-resolved Fourier-transform tech-
inverse of the fraction of incident light transmitted at a
niques, especially Fourier transform infrared spectro-
particular wavelength, (1/10),can be related to the
scopy (FTIR),with increasingly higher time resolution
concentration of absorbers, by the Beer-Lambert law
and sensitivity, and the use of cavity ring-down
detection in absorption measurements, with the po- (eqn E181).
tential to increase detection sensitivity by many orders
ln(Io/I) = ~ ( 2 ) k [I81
of magnitude.
As long as the fraction of incident light which is
Relative rate determinations Often, it is more con- absorbed remains small (i.e., less than about 20%),
venient to measure the relative rate of reaction of two this expression may be manipulated to relate the
reagents with a third reactant, rather than perform an fraction of light absorbed, A(),) to the absorber
absolute rate determination. For instance, slower concentration, as shown in eqn [19].
reactions may suffer from significant artifacts such
as wall reactions, but their rate constants may be A(;") = ~ ( 2 ) l ~ P 91
measured with quite reasonable precision (although
not necessarily with great accuracy) in this manner. The relationship between the amount of light
For a simple set of two competing reactions [XVI] and absorbed and concentration, if known, may be
[XVII], it is possible to derive a simple relationship exploited to determine the absolute concentrations
between the concentrations of the reactants and their of reacting species as they change in time during a
rate constants, as shown by eqn [17]. reaction. Several implementations are in common use.
In direct absorption spectroscopy the absorption,
A+X+Pl [XV? A(A), is recorded, either at a specific wavelength, or
as a function of wavelength, yielding an absorption
A+Z+P2 [XVII] spectrum. In flow tube methods, this may take place in
a sample cell through which the reacting flow passes
downstream of the mixing region; in flash photolysis
methods, the absorption is recorded as a function of
Here XO and ZOrepresent the respective concentra- time after the initial pulse. This variation is commonly
tions at t = 0, X , and Zt give the concentrations at known as transient absorption spectroscopy. In its
time t, and kx and kz represent the two rate coeffi- simplest form, a single wavelength, corresponding to a
cients. If one of the rate coefficients is known particular absorption feature, is transmitted through
independently, the other may be determined in this the sample, using resonance lamps or lasers. The time-
way. Note that the concentration of reactant A need resolved change in the intensity of this source after it
not be measured at all; a similar relationship may be has passed through the sample is measured following
derived for the two product concentrations, meaning the initiation of reaction. Depending upon the exact
that the experimenter may monitor either of the ratios experimental configuration used, the time resolution
[X]/[Z]or[Pl]/[P2],depending on convenience, detec- can be very good (up to 10 - I 2 s); more typically it lies
tion sensitivity, or other experimental factors. in the range 10-6-104s. This technique is readily
LABORATORY KINETICS 1097
coupled to long path-length absorption cells to yield formed in the reaction. For example, the time depend-
very sensitive concentration measurements. Modern ence of the concentration of ground state oxygen
variations involve the use of tunable diode lasers atoms may be followed by detecting emission (see
(TDLs) and cavity ring-down (CRD) systems. Under reaction [XIX]) from electronically excited NOz,
some fortunate circumstances, a multiwavelength formed in the chemiluminescent reaction [XVIII].
technique such as FTIR may be used to identify all
reactants and products simultaneously as the reaction 0 +NO + M -+NO; + M [XVIII]
proceeds. However, the sensitivity is not particularly
high, nor is the time resolution, so this method is
generally limited to slower reactions, often those being
studied in environmental chambers. Other spectroscopic detection methods, which are
A more sensitive, but less general spectroscopic no longer in such common use, include laser magnetic
probe of concentrations involves the measurement of resonance (LMR) and electron spin resonance (ESR)
some quantity related to the amount of light absorbed. spectroscopy. Both rely upon the ability of an external
Many atoms and molecules re-emit some fraction of magnetic field to perturb the energy levels of an atom
the initially absorbed light, often at a wavelength or molecule. By varying the magnetic field strength,
shifted away (to the red) of the absorbing wavelength. optical transitions may be brought into resonance with
This re-emission forms the basis of resonance fluores- a fixed-wavelength light source, allowing light ab-
cence (RF) spectroscopy and of laser-induced fluores- sorption (and hence detection) to occur.
cence (LIF) spectroscopy. The fraction of the light
absorbed by a sample that is re-emitted depends upon
many variables, but remains constant if experimental See also
conditions do not change. The intensity of emitted Chemistry of the Atmosphere: Chemical Kinetics; Gas
light is thus a very sensitive proxy for the amount of Phase Reactions; Ion Chemistry. Observations for
absorption. Although it is extremely difficult to relate Chemistry (In Situ): Chemiluminescent Techniques;
the emission intensity to absolute concentrations, the Gas Chromatography; Resonance Fluorescence.
time dependence of emission intensity does track that
of concentration, so that rates may be determined.
In RF spectroscopy, an atomic resonance lamp Further Reading
optimized for atom X is used to illuminate the sample;
resonance emission is observed from any atomic X Finlayson-Pitts BJ and Pitts JN Jr (2000) Chemistry of the
Upper and Lower Atmosphere. San Diego: Academic
which is present. Reactions of atomic species such as Press.
C1 and ground state 0 have been studied using such Gierczak T, Talukdar RK, Herndon SC, Vaghjiani GL and
lamps. LIF spectroscopy generally utilizes a tunable Ravishankara AR (1997) Rate coefficients for the reac-
laser source to scan over the absorption spectrum of a tions of hydroxyl radicals with methane and deuterated
molecule of interest. When the laser wavelength is methanes. Journal of Physical Chemistry, A 101(17):
resonant with a molecular absorption transition, 3 125-3 134.
emission from the excited molecules may be observed. Howard CJ (1979)Kinetic measurements using flow tubes.
The O H radical is especially well suited to LIF Journal of Physical Chemistry 83(1):3-9.
detection, and this is the method of choice in studying Pilling MJ and Seakins PW (1995) Reaction Kinetics.
its kinetics. Oxford: Oxford University Press.
Another detection method that is based on measur- Steinfeld JI, Francisco JS and Hase WL (1999) Chemical
Kinetics and Dynamics, 2nd edn. Upper Saddle River, NJ:
ing emission intensity relies upon the formation of Prentice Hall.
electronically excited products from a chemical reac- Thorn RP, Cronkhite JM, Nicovich JM and Wine PH (1995)
tion. The emission of light from such excited products Laser flash photolysis studies of radical-radical kinetics:
is known as chemiluminescence; its intensity is related the O ( 3 P ~+) BrO reaction. Journal ofChemica1 Physics
to the concentration of electronically excited products 102(10):4131-4142.
1098 LAGRANGIAN DYNAMICS
I Roulstone, Met Office, Reading, UK mechanics is the generalization of the discrete label i.)
Lagrangian mechanics is the basis for many important
conceptual models of fluid flow. In the second method,
known as the Eulerian method, we fix our attention on
Introduction a region of space and study the motion of fluid relative
to that region. That is, the independent variables are
In meteorology and oceanography we regard our fluid the space coordinates x = ( x ,y, z ) and time t , and the
as a continuum - a continuous distribution of mass in dependent variables are most commonly taken to be
space. In so doing, the atomic or molecular nature of the velocity v(x, t), the mass density p(x, t ) , and the
the fluid is neglected and this implies that any small pressure p(x,t). From the density and pressure all
volume element is always supposed to be sufficiently other thermodynamic quantities can be determined,
large that it still contains a huge number of molecules. provided the equation of state is known.
Accordingly, when we speak of an infinitesimal The time derivative of an arbitrary quantity
volume element - or a fluid ‘parcel’ - we mean that it A(a, t ) = A(x(a, t ) ,t ) (where we have used the same
is small compared with the volume of the system but symbol to denote two different functions) measured in
still sufficiently large to contain very many molecules. the two descriptions is related by the chain rule
The adoption of the continuum description is the
ax aA
first step towards specifying how we solve the funda-
mental problem of the science of kinematics for our
fluid; i.e., the specification of suitable methods for
aA
-
at la =
aA +-.-
t
i at ax
describing and analyzing fluid motions. The problem
of kinematics, for any given physical system, should be
distinguished from that of dynamics, which is con-
- aA
at I ly,z.l I L lx,y,t-
- - +u-
aA
ax
+v-
aA
aY
aA
+w-
az
DA
=-
Dt
121
cerned with determining the state and/or motion of the
system at any instant. It is common practice to refer to which illustrates the relationship between the familiar
the combined science of kinematics and dynamics as material or substantial derivative, D/Dt, and (slat),.
mechanics. The relationship ( a / a t ) , = D/Dt tells us why the
There are two common descriptions of continuum material derivative is often referred to as the ‘deriva-
motion (both due originally to Leonhard Euler (1707- tive following a parcel’.
83)).In the first method, known as the Lagrangian The Lagrangian description of fluid kinematics and
method, we fix our attention directly on the fluid dynamics, and related topics, are the subject of this
parcels and study their motion through space. Unless review. The Eulerian description is adopted by most
otherwise stated, we shall consider motions in a three- textbooks, but there are many fundamental principles
dimensional space. The independent variables are a set in fluid dynamics that are inherently Lagrangian in
of particle labels a = ( a , b, c) and time t. The depend- nature. To illustrate this point, consider the continuity
ent variables are the coordinates equation in Eulerian form
of the fluid parcel identified by ( a ,b, c). At any given

time there is a one-to-one correspondence between the where, unless otherwise stated, V is the usual gradient
coordinate systems a and x(a, t ) .We assume that each operator in three-dimensional space. This familiar
fluid parcel, which really refers to an imaginary piece equation arises from the requirement that fixed
of the continuum, is uniquely defined by values of the volumes in particle label space always contain the
labels a for all time. If x is a suitably differentiable same mass. That is, if we assign labeling coordinates so
function of its arguments, then the velocity of a parcel that x(a, 0) a at some initial time t = 0,then the
is given by v = (ax/at), = ( u ,v , w ) and the accelera- mass containedin the small volume d3x(a,0) of fluid at
tion by (a2x/at2)a,where the subscript a means that x(a, 0) is given by p(x(a, 0), 0) d3x(a, 0) = p(a, 0) d 3 a
the time derivative is evaluated on a parcel (a constant This defines the density, p, which we may consider as
value of a). (In conventional point-particle mechanics, either a function of Lagrangian labels, a, or Eulerian
we are accustomed to assigning a position vector, ri positions, x. At a later time t, we must have
say, to the ith particle; the continuous label a in fluid p(x(a, t ) ,t )d3x(a, t ) = p(a, 0) d3a. If we define the
LAGRANGIANDYNAMICS 1099
specific volume, a, as the Jacobian for all i, where the overdot denotes d/dt. Newton’s law
follows as a consequence of the arbitrariness of Gxi(t),
and the specific form [5]follows from the functional
[41
dependence of the potential energy V on xi alone, and
the usual expression for the kinetic energy
then the conservation of mass is given by
p(x(a, t), t)x = p ( a , 0). For an incompressible fluid
-
we have a = 1 (which corresponds to V v = 0 in the
Eulerian description). It is now possible to show that
.
(aalat), = aV v, which in turn can be expressed in the There is a fundamental connection between certain
form [ 3 ] . invariant properties of Lagrangians and the conserva-
tion laws for dynamical systems. This connection is
Variational Principles and embodied in Noether’s theorem, which requires con-
siderable expertise in group theory to understand
Conservation Laws completely. We shall therefore restrict ourselves to a
In writing down variational principles for fluid didactic description. Noether’s theorem applies to the
systems, one is immediately faced with the choice of equations that arise from variational principle like
adopting either Lagrangian or Eulerian kinematics. Hamilton’s principle. According to Noether’s theorem
Occasionally, it is convenient to change between these (1918): If a variational principle is invariant to a
representations when performing calculations. In this continuous transformation of its dependent and inde-
section we shall review one of the most important pendent variables, then the equations arising from the
features of the Lagrangian description: a connection variational principle possess a conservation law.
between an invariant property of this representation
and the conservation of a meteorologically significant Example. Let us assume that a system has kinetic
quantity, potential vorticity. energy of the form [8], but the potential is independent
of x (but still dependent on y and z, say). Therefore the
Hamilton’s Principle and Noether’s Theorem Lagrangian is invariant with respect to small varia-
tions in x . The integrand of eqn [7],the so-called
We begin by discussing a finite-dimensional problem.
Euler-Lagrange equation,
Consider a system of n point-particles, with masses
mi (i = 1, . . . , n ) and locations xf(t), moving under
the influence of some potential V(xi). Newton’s $&)-”=O ax [91
second law of motion for such a system is
then yields, assuming independence of x,
151
Hamilton’s principle is a variational principle equiv-

alent to Newton’s second law, and it states that the This is the statement that the x-component of
action momentum, mk,is conserved.
tl Application to Fluid Mechanics: Theorems
S = i Ldt of Ertel and Kelvin
The fundamental meteorological-cum-oceanographic
is stationary, where the Lagrangian principle of the conservation of potential vorticity can
be traced to the invariance of variational principles for
L(x,?) E T-V
hydrodynamical flows under certain variations of the
is the difference between the kinetic energy T and particle labels, 6a.
the potential energy V of the system. Hamilton’s Let x ( a , t ) be the location of the fluid particle
principle thus states that the variation of the action [6] identified by labeling coordinates a = (a, b, c) at time
vanishes for arbitrary, independent variations 6xi (t) t. The Lagrangian for a perfect fluid is
that vanish at tl and to. Since 6xi(tl) = 6xl(to) = 0, we
have C=
s d3a [$k2 - E ( a ,S(a))]
where the internal energy E(x,S) is a prescribed

thermodynamic function of the specific volume, x,
and the specific entropy, S(a). The entropy is a expresses the conservation of potential vorticity -
Lagrangian conserved quantity. The integration over Ertel’s theorem.
a measure in particle-label space, Jd3a, replaces the Now consider any closed loop in a-space. By eqn
discrete sum Xi, and the overdot is the time derivative [14], it follows that
(a/at)a.
Hamilton’s principle states that a/atjA.da=O [I81
6
/ Cdt=O But A . da = v. dx by eqn [13], and the statement
a/atjv.dx =0 1191
where 6 stands for arbitrary independent variations
6x(a, t), and implies
is Kelvin’s theorem.
x + avp = 0 1111
Reduction of Order
where p = -aE/aa is the usual thermodynamic equa-
tion relating pressure to internal energy and it may be The state of a perfect fluid at a fixed time t corresponds
considered the equation of state. Equation [11] is the to a point in an infinite-dimensional phase space in
momentum balance for a perfect fluid. which each dimension represents the value of one
The Lagrangian is unaffected by particle-label component of v(a) or x(a) at a fixed value of a. The six
variations 6a(x, t) that leave the density and entropy Lagrangian fields
unchanged. For the purposes of the following calcu-
lation, we shall assume that the particle labels enter
only through the specific volume, which is equivalent
to the assertion that the fluid is homentropic:
E = E ( a ) . (The results can easily be generalized to
nonhomentropic fluids where E = E ( a , S).) Consider However, each choice of { V ( X ) ~~ ( xS(x)} ) ~ corre-
variations 6a(x, t ) , of eqn [lo], that leave the density sponds to infinitely many choices of {v(a), .(a)}.
and entropy unchanged. Then Thus the Eulerian description is a reduced phase space
for the fluid. A group-theoretic picture of the reduction
6
S &dt=
=
/J 5 a(&)
d t d a--.-
-//dtd’aA.-
from Lagrangian to Eulerian variables gives a more
rigorous description, but this fundamental topic is a
1121 subject for further reading.
at
where
Numerical Techniques
Lagrangian thinking manifests itself in the design of
V, is the gradient operator in particle-label space, and numerical methods. Here we give a brief review of
we used the chain rule for partial derivatives t o go two schemes: semi-Lagrangian techniques which are
from the second term to the third term in eqn [12]. widely used and a novel Lagrangian approach to the
After integration by parts, noting that Ga(x,t) is semi-geostrophic frontogenesis model - the so-called
arbitrary, one can show that geometric method.
Semi-Lagrangian Schemes
a/at(V,xA) =0 1141
Equation [14] is a general statement of vorticity Semi-Lagrangian schemes are important because they
conservation. Let e(a) be any quantity that is con- offer the promise of allowing longer time steps, with
served on fluid particles. Then no loss of accuracy, than Eulerian-based advection
schemes whose time step is limited by stability criteria.
a/at[(v,xA) .v,e] = 0 1151 These issues are of practical importance in numerical
weather prediciton.
Using eqns [4] and [ 131 we have To illustrate the essential ideas, let us consider
(v, XA) .v,e =~(vXV). ve 1161
a simple one-dimensional advection equation in
Eulerian form
and, from eqns [15] and [16], the statement
aA aA
-+u-=O
a / at [ a (vxv) .vel = o 1171 at ax
LAGRANGIAN DYNAMICS 1101
where A ( x , t ) is the transported quantity. The appro- Geometric Method

priate Lagrangian form of eqn [20] is the pair of
equations The geometric method was born out of an attempt to
demonstrate the value of a Lagrangian approach to
dA understanding the dynamics of rotating, stratified
-
dt
=o fluids. It is a novel technique for integrating the semi-
geostrophic equations that involves dividing the
atmospheric domain into elements, each characterized
dx
- by a certain mass, potential temperature, and absolute
=u
dt momentum. Applications of the geometric method
include modeling frontogenesis, embedded convec-
Equation [21a], states that A is conserved along the tion, sea breezes, and flow over orography. Some of the
trajectory, which is in turn given by the eqn [21b]. advantages of the geometric method include the ability
We seek to integrate eqn [21a,b] by gridpoint to handle a frontal discontinuity, no eddy diffusion
techniques. (This is the origin of the term ‘semi’ in requirement and, in principle at least, mountain
semi-Lagrangian - the combination of Lagrangian barrier effects (e.g., drag) can be represented without
mechanics with Eulerian gridpoint techniques.) Let A t parameterization. An example of a ‘tropopause fold’,
denote a time step and let n label the time steps. For constructed via the geometric method with l o 5
each point on the grid we approximate the trajectory elements, is illustrated in Figure 1A.
+
that would arrive at that point at time ( n 1)At using The geometric method may be summarized as
the wind u at time n. We refer to the point at which the follows. Consider a domain D with coordinates
forecast is made as the arrival point (A) and the point ( x ,y , z ) , taken to represent a region of the Northern
from which the trajectory departed at time ( n - 1)A t Hemisphere (y pointing polewards). Let C be a cross-
the departure point (D).The point in the middle of the section
trajectory at time n A t is called the midpoint (M).The
departure point is determined by approximating eqn
[21b] with (where typical values of L and H may be 1000 km and
10 km, respectively). The geometric method dis-
cretizes the fluid in C into ‘parcels’ of equal area c,,
so that C = E, C,. To the ith parcel is associated a
The midpoint, at which u is needed in eqn [22], is value of ‘absolute +
momentum’ M, = (up f x ) , and a
obtained by a similar expression value of potential temperature 0,. Here, ug is the
y-component of the geostrophic wind ug =
( l / f ) a $ / a x , $ ( x , z ) is the geopotential function, and
f is the constant Coriolis parameter. One can show
that a convectively and inertially stable arrangement
which is implicit since u at X M cannot be determined of the fluid parcels can be represented uniquely by a
until X M is known. This requires an iterative process in convex function P ( x ,z ) = if2,, + $ (sometimes re-
which the next guess of X M on the left is based on u at ferred to as the ‘modified geopotential’ or ‘modified
X M from the previous iteration. Once XM is found, the pressure’), such that each element C , can be associated
departure point calculated from eqns [22] and [21a] with a plane whose height ‘above’ the (x,s)-plane at
yields any point corresponds to the value of P at that point.
The intersection of these planes defines a piecewise
1241 planar solution P ( x ,z ) (Figure 1B).The momentum of
each parcel, M,, and potential temperature of each
Typically the departure points and midpoints will not parcel, Q,, are proportional to the gradients of P ( x ,z )
coincide with grid points. The values of A and u at with respect to x , and z,, respectively.
these points must be found by interpolation from The constraint that P be a convex function is crucial
neighboring points and the scheme is stable provided to the solution procedure. It allows us to write the
the interpolation is based on data points surrounding physical solution as the supremum of a family of
the departure point (or midpoint). There is essentially generating surfaces
no time step restriction and thus, near the poles of a
global model where grid lines converge, the departure
+ +
P ( x , z ) = max,{xMj z6’j S i } = max P,
point may be many grid intervals away from the where Pi denotes the modified pressure plane of the ith
arrival point without being unstable. element and Si the point of intersection on the P-axis of
Solution-surface P(x, z)
(6)
Figure 1 (A) Projection of a polyhedral surface P ( x .z) onto the ( x ,z)-plane. The front is modeled as a discontinuity in the gradients of
the faces that make up the piecewise planar surface. (6)Construction of a convex polyhedral surface P ( x ,z) from faces with given
gradients (01.Mi) and areas.
a plane in (x,z,P)-space. In a physical problem Lagrangian Analysis

involving time integration, we begin with a set of
The description of weather systems and ocean eddies
values Ci, Mi,Bi, and Si. The construction algorithm
has always utilized the Lagrangian description of the
requires the determination of the set of Si values such
flow. Examples include synoptic developments in
that the element areas are correct - a process that must
be carried out iteratively. If the domain boundary is terms of air masses (dating back to the Bergen School
in the early twentieth century), the parcel theory of
fixed or a known function of time, one may regard Si as
convection, and the description of the dynamics of
a function of Ci, Mi,and 8i. In the classical fronto-
precipitation systems in terms of conveyor belts. The
genesis problem where a pure barotropic deformation
power of the Lagrangian description in these contexts
field is imposed the governing Lagrangian equations
is the conceptual simplification achieved by burying
take the form
the nonlinearity of the material derivative D/Dt in the
Jacobian of the map between particle labels and
DMi
-- - -qMi, -- D Ci Dei Lagrangian positions: GI = a(x)/a(a) (cf. eqn [4]).
- -yci, -= Fi ~ 5 1
Dt Dt Dt However, whenever we wish to carry out a theoretical
analysis of hydrodynamical flows the Eulerian frame-
where y is the deformation rate and Fi is a forcing work is the most commonly used vehicle. In this
function (which is zero in the frontogenesis problem section we give a brief resumi: of a Lagrangian method
under consideration). Therefore, at any future time Ci, of analysis that, although hitherto not widely used,
Mi, and 8i are known and the solution can be found by offers a powerful technique for analyzing, for
determining Si. example, the conceptually simple (yet analytically
LAGRANGIANDYNAMICS 1103
difficult in the Eulerian setting) problems of transpor- of rearrangement that can be reached from given
tation along particle trajectories. initial data.
The precise definitions of a rearrangement of
functions (both scalar-valued and vector-valued) re-
Atmospheric Dynamics and Rearrangements quire concepts from measure theory, and as such have
A concept that arises naturally when considering the very precise technical definitions. However, the tech-
Lagrangian description of fluid motion is that of the nical nature of the definitions is extremely important,
rearrangement of the fluid by the material derivative. since it gives a firm basis to the mathematical analysis.
The action of the D/Dt operator on an arbitrary Bearing this in mind, we will give an intuitive
quantity A is to advect, or rearrange, it. Consider for definition. Consider a region, D , spanned by the
example a conserved quantity, such as potential coordinates ( x ,y, z ) and let f(x) be a function defined
vorticity q = p - l c . VO,that retains its value following on this region (e.g., moisture or potential vorticity).
a fluid particle. Here, p is the fluid density, 0 the Within the region we have our fluid with Lagrangian
potential temperature, and 5 the total vorticity. As time coordinates ( a , t ) and let us assume we can attach
advances the fluid particles are permuted or rear- values of the function f to each fluid particle, thereby
ranged, but each particle retains its original value of q. giving us a function F(x, to), at some reference time to.
Atmospheric cyclones and anticyclones, and ocean As the state of the fluid evolves according to a
eddies, can be idealized as the stratified, rotating dynamical model, at a later time tl we have a new
coherent structures that correspond to circular vor- function G(x, t l ) ,which we call a rearrangement of F
tices in ordinary two-dimensional Euler flow. Their if the two functions satisfy a certain equivalence
interaction and evolution, which play a major role in relationship between the ‘sizes’ or ‘volumes’ of the sets
weather developments and in the behavior of ocean on which F and G take values greater than or equal to a
eddies, have been much studied using approximations datum value, for all real positive datum values. An
to Newton’s second law of motion. These approxi- example of two functions F and G that are rearrange-
mate models seek to describe flows in which there is a ments according to this definition is given in Figure 2.
dominant balance between the Coriolis, buoyancy, Possible applications of rearrangements include
and pressure-gradient forces. Such approximations to numerical methods in which the goal is to model the
Newton’s second law are commonly referred to as evolution of a quantity such as potential vorticity as
balanced models. Many such models can be described accurately as possible. In variational data assimilation
in terms of the Lagrangian conservation of potential it may be useful to work with Lagrangian increments,
vorticity by an equation instead of Eulerian perturbations, in which minimiza-
tion is carried out over certain classes of ‘dynamically
_
Dq -
-0
accessible’ rearrangements. Rearrangements can be
Dt used to compare two functions of spatial variables and
as such there may be applications in the future to
together with a so-called invertibility principle forecast verification techniques.
A significant achievement of rearrangement theory
to date is its application to the study of the stability of
that relates the wind field v, pressure p , and potential steady states. Steady states can be characterized as
temperature 0 to the potential vorticity. Typically, E stationary points of the energy with respect to
will be an elliptic operator, which may be nonlinear,
and certain boundary conditions must be specified. A
relationship between the wind, temperature, and
pressure fields is known as a balance condition, which
is required to define the relationships implicit in eqn
[26]. Examples of models that can be formulated in
this way are the barotropic vorticity equation, quasi-
geostrophic theory, and semi-geostrophic theory. The
solutions to all these models can be described as
rearrangements of the initial potential vorticity distri-
bution, and because the advecting velocity is con-
strained by the invertibility procedure, progress in
understanding features such as existence and unique-
ness and the topological properties of solutions can be Figure 2 Two rearrangements F and G; the area of the shaded
made because the elliptic operator governs the type regions is the same for every value of the datum point E .
1104 LAKE-EFFECT STORMS
rearrangement perturbations, which is not possible if Lamb H (1932) Hydrodynamics, 6th edn. Cambridge:
Eulerian perturbations are used. Cambridge University Press.
Marsden JE and Ratiu T (1994)Introduction to Mechanics
and Symmetry. Texts in Applied Mathematics, vol. 17.
Berlin: Springer-Verlag.
See also Norbury J and Roulstone I (eds.) (2002) Large-Scale
Atmosphere-Ocean Dynamics: vol. 1 Analytical Meth-
Dynamic Meteorology: Balanced Flows; Potential Vorti- ods and Numerical Models; vol. 2 Geometric Methods
city. Hamiltonian Dynamics. Kinematics. Numerical and Models. Cambridge: Cambridge University Press.
Models: Methods. Tracers. Wave Mean-Flow Inter- Salmon R (1998)Lectures on Geophysical Fluid Dynamics.
action. Oxford: Oxford University Press.
Shepherd TG (1990) Symmetries, conservation laws and
hamiltonian structure in geophysical fluid dynamics.
Further Reading Advances in Geophysics 32: 287-338.
Hoskins BJ (1982) The mathematical theory of frontogen- Shutts GJ, Cullen MJP, and Chynoweth S (1988)Geometric
esis. Annual Review of Fluid Mechanics 14: 131-151. models of balanced semi-geostrophic flow. Annales
Hoskins BJ, McIntyre ME, and Robertson AW (1985) On Geophysicae 6: 493-500.
the use and significance of isentropic potential vorticity Staniforth A and Cot6 J (1991)Semi-lagrangian integration
maps. Quarterly Journal of the Royal Meteorological schemes for atmospheric models: a review. Monthly
Society 111: 877-946. Weather Review 119: 2206-2223.
P J Sousounis, Michigan State University, Ann Arbor, Great Lakes are the largest single source of fresh water
MI, USA in the world (except for the polar ice caps), the fact that
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. the Great Lakes are situated approximately halfway
between the Equator and the North Pole, the fact that
Great Lakes are located in the interior of a large
Introduction continent, the fact that each of the lakes is approxi-
Each winter, lake-effect storms develop on the down- mately the size of a small inland sea, and the fact that
wind shores of the North American Great Lakes, as there are several lakes - separated from each other by
arctic winds blow across the relatively warm water. distances less than their own size, make for some very
The associated clouds and snow ( o r rain) showers tend unique weather in the region. These characteristics
to organize in narrow bands, usually only a few suggest that the lakes rarely freeze over completely,
kilometers wide but sometimes over 200 km long. even in the coldest of winters, and thus remain a nearly
There may be one band, or there may be as many as 10 continuous and very large source of heat and moisture
or 20, each separated from the next by only a few for the atmosphere. Lake-effect storms continue to be
kilometers of clear sky. These bands may remain a forecast challenge despite improvements in numer-
stationary over a region or they may oscillate in ical mesoscale models because of their meso-ylmeso-P
snakelike fashion. They may produce nothing more scale size.
than one or two centimeters of snow, or they may
dump over 120cm of snow in a single storm. These
lake-effect storms are primarily a product of relatively CIimatology
simple air mass modification by warm water, compli-
cated lakeshore geometry, and the prevailing synoptic Lake-effect snow accounts for 25-50% of the total
situation. annual snowfall in many lakeshore regions (Figure 1).
Lake-effect storms develop in other parts of the The snowbelts (areas of heavier snow) that shoulder
United States, Canada, and the world, but nowhere the southern and eastern shores of the Great Lakes
else do they occur as frequently or with such intensity reflect the direction of the prevailing north-westerly
as they do in the Great Lakes region. The reasons for flow relative t o the orientation of the lakes, the sharp
the unique weather in the Great Lakes region can be contrast in surface friction between the relatively
traced to several geographic aspects. The fact that the smooth lake surface and the rough land, and terrain
rearrangement perturbations, which is not possible if Lamb H (1932) Hydrodynamics, 6th edn. Cambridge:
Eulerian perturbations are used. Cambridge University Press.
Marsden JE and Ratiu T (1994)Introduction to Mechanics
and Symmetry. Texts in Applied Mathematics, vol. 17.
Berlin: Springer-Verlag.
See also Norbury J and Roulstone I (eds.) (2002) Large-Scale
Atmosphere-Ocean Dynamics: vol. 1 Analytical Meth-
Dynamic Meteorology: Balanced Flows; Potential Vorti- ods and Numerical Models; vol. 2 Geometric Methods
city. Hamiltonian Dynamics. Kinematics. Numerical and Models. Cambridge: Cambridge University Press.
Models: Methods. Tracers. Wave Mean-Flow Inter- Salmon R (1998)Lectures on Geophysical Fluid Dynamics.
action. Oxford: Oxford University Press.
Shepherd TG (1990) Symmetries, conservation laws and
hamiltonian structure in geophysical fluid dynamics.
Further Reading Advances in Geophysics 32: 287-338.
Hoskins BJ (1982) The mathematical theory of frontogen- Shutts GJ, Cullen MJP, and Chynoweth S (1988)Geometric
esis. Annual Review of Fluid Mechanics 14: 131-151. models of balanced semi-geostrophic flow. Annales
Hoskins BJ, McIntyre ME, and Robertson AW (1985) On Geophysicae 6: 493-500.
the use and significance of isentropic potential vorticity Staniforth A and Cot6 J (1991)Semi-lagrangian integration
maps. Quarterly Journal of the Royal Meteorological schemes for atmospheric models: a review. Monthly
Society 111: 877-946. Weather Review 119: 2206-2223.
P J Sousounis, Michigan State University, Ann Arbor, Great Lakes are the largest single source of fresh water
MI, USA in the world (except for the polar ice caps), the fact that
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. the Great Lakes are situated approximately halfway
between the Equator and the North Pole, the fact that
Great Lakes are located in the interior of a large
Introduction continent, the fact that each of the lakes is approxi-
Each winter, lake-effect storms develop on the down- mately the size of a small inland sea, and the fact that
wind shores of the North American Great Lakes, as there are several lakes - separated from each other by
arctic winds blow across the relatively warm water. distances less than their own size, make for some very
The associated clouds and snow ( o r rain) showers tend unique weather in the region. These characteristics
to organize in narrow bands, usually only a few suggest that the lakes rarely freeze over completely,
kilometers wide but sometimes over 200 km long. even in the coldest of winters, and thus remain a nearly
There may be one band, or there may be as many as 10 continuous and very large source of heat and moisture
or 20, each separated from the next by only a few for the atmosphere. Lake-effect storms continue to be
kilometers of clear sky. These bands may remain a forecast challenge despite improvements in numer-
stationary over a region or they may oscillate in ical mesoscale models because of their meso-ylmeso-P
snakelike fashion. They may produce nothing more scale size.
than one or two centimeters of snow, or they may
dump over 120cm of snow in a single storm. These
lake-effect storms are primarily a product of relatively CIimatology
simple air mass modification by warm water, compli-
cated lakeshore geometry, and the prevailing synoptic Lake-effect snow accounts for 25-50% of the total
situation. annual snowfall in many lakeshore regions (Figure 1).
Lake-effect storms develop in other parts of the The snowbelts (areas of heavier snow) that shoulder
United States, Canada, and the world, but nowhere the southern and eastern shores of the Great Lakes
else do they occur as frequently or with such intensity reflect the direction of the prevailing north-westerly
as they do in the Great Lakes region. The reasons for flow relative t o the orientation of the lakes, the sharp
the unique weather in the Great Lakes region can be contrast in surface friction between the relatively
traced to several geographic aspects. The fact that the smooth lake surface and the rough land, and terrain
LAKE-EFFECT STORMS 1105
effects. The largest snowfall totals exist across Lake-effect snow falls almost exclusively during the
the upper peninsula of Michigan, where north-west- unstable seasou - that portion of the year when the
erly flow across Lake Superior is forced upward lakes are climatologically warmer than the ambient air
abruptly over steep terrain upon reaching the northern and thus provide heat and moisture to the lower
coast of the upper peninsula - especially around atmosphere to destabilize it. Enhanced cloudiness and
the Keewanaw Peninsula, and across the Tug Hill precipitation exist across much of the lake shore
Plateau in western New York, where west-south- regions and far inland as well. The percentage of
westerly flow across the lower lakes provides a long cloudy days peaks in November for many places of the
fetch and ample opportunity for the air to be Great Lakes region - owing in part to significant lake-
moistened and destabilized. In both of these locations, enhanced cloudiness. Precipitation during the unsta-
long fetches and orography are key aspects. Terrain ble season begins typically with episodes of nocturnal
can enhance individual snowstorm totals by about rain showers during cool nights in late August. As the
5 cm for every 100 m of rise. Additionally, portions of mean air temperature drops through the fall months,
the lakeshore with enhanced concavity promote con- lake-effect rain showers change to lake-effect snow
vergence zones that can further enhance snowfall showers. Much of the lake-effect snow falls typically
totals. Heavier lake-effect amounts fall typically between November and February, which constitutes
during cold winters, when the lake-air temperature the heart of the unstable season, when lake-air
differences are enhanced. temperature differences tend to be greatest.
Figure 1 Average 1951-1980 Great Lakes seasonal snowfall total. (From Figure 2 in Norton DC and Bolsenga SJ (1993)
Spatiotemporal trends in lake effect and continental snowfall in the Laurentian Great Lakes, 1951-1 980. Journalof Climate6: 1943-1956.
Adapted with permission from the American Meteorological Society.)
1106 LAKE-EFFECTSTORMS
Climatological lake-air temperature differences the lake. These bands (Type 111) actually develop as
may be around 7-8"C, but may exceed 30°C during midlake bands in north-westerly flow over Lake
intense cold-air outbreaks. Coupled with winds Huron, farther upwind (Figure 3, lower panel). They
sometimes in excess of 2 0 m s - l , combined surface may lose their visible cloud characteristics over
sensible and latent heat fluxes can typically exceed southern Ontario, but maintain the convergence
1000Wm-2 - comparable to that found in a cat- zone so that they redevelop once they reach the lower
egory-1 hurricane (see Hurricanes). lakes. If the wind is very light against areas of
Lake-effect clouds and snow can occur locally on enhanced concavity, then lake vortices (Type V) may
-3040% of the days in winter under a variety of develop (Figure 3, upper panel). Phenomena thought
synoptic patterns -whenever there is an onshore fetch to play roles in this type of lake-effect storm are
and the lake-air temperature difference allows the stretching and tilting of vorticity from low-level
lowest layers of the air to destabilize. However, certain convergence and vertical wind shear; differential
synoptic patterns are more favorable than others for diabatic heating; and synoptic scale vorticity and
allowing lake-effect snow to develop. A typical temperature advection. Table 1 summarizes some of
sequence of events begins with a synoptic-scale low the differences between various aspects of multiple-
moving across the Great Lakes region from south-west and single-band lake-effect snowstorms in western
to north-east (Figure 2). Additionally, in late autumn Michigan (lower peninsula).
especially, these lows are deepening as they cross the The most common type of lake-effect event over the
region because of baroclinic forcing and aggregate northern lakes (Superior, Huron, and Michigan) is the
heating from all the lakes. Strong north-westerly multiple-band variety, which occurs 50% of the time.
winds on the back side of the low bring progressively The next most common is the shore-parallel or
colder polar or arctic air across the warm lakes. midlake variety, which occurs 25% of the time. Effects
Subfreezing temperatures may reach as far south as the occurring during the remaining 25 % include meso-
Gulf Coast and northern Florida, with - 20°C read- scale vortices, hybrid combinations, and undetermi-
ings just north of the lakes. The strong winds and cold nable forms. The most common type of lake-effect
air generate strong surface fluxes over the lakes that event over the southern lakes (Erie and Ontario) is the
moisten and destabilize the air, leading to snow shore-parallel or midlake variety, because of the
showers along the downwind lakeshores of the Great different orientation of these lakes (Figure 4). In
Lakes. The deepening of the low and the destabiliza- general, across the region, the frequency of multiple
tion both allow stronger winds from above to mix bands decreases from west to east and the frequency of
down to the surface and further increase the heat and shore-parallel and midlake bands increases from west
moisture fluxes. to east.
Depending on the wind speed, the orientation of the
wind flow relative to the long lake axis, stability,
moisture, and upper-level forcing, different types of
lake-effect storms can develop. Basically, when the
Boundary Layer Dynamics
prevailing flow is more parallel to the short axis than to When cold air flows across the warm waters of the
the long axis of a lake (i.e., there are strong short-axis Great Lakes, strong sensible and latent heat fluxes
winds), multiple wind parallel bands (Type 11)develop warm and moisten the air closest to the surface first,
(Figure 3, middle panel). These bands are typically causing the lowest levels of the atmosphere to desta-
2-4 km wide and spaced 5-8 km apart. Snowfall is bilize. Strong turbulent motions mix upward the
usually spread over a large area of the downwind warmed and moistened air in convective fashion.
lakeshore, and amounts are usually light (<4 mm Steam fog typically develops and steam devils may be
liquid precipitation per day). When the prevailing visible near the surface, especially within a few tens
short-axis winds are weak, midlake (Type I) or shore of kilometers of fetch (Figure 5 ) . A very unstable
parallel (Type IV) bands can develop - even when the convective internal boundary layer (CIBL)forms near
long-axis prevailing winds are strong (Figure 3 upper the surface and grows rapidly upward in the down-
panel). These bands can be 10-20km wide and wind direction (Figure 6 ) . The upwind temperature,
generate copious amounts of snow. If the short-axis humidity, and flow characteristics of the air before it
wind is essentially not present then the band will be reaches the lake determine how the air will be
located near the middle of the lake. If the short axis modified. As the air crosses the downwind lakeshore,
prevailing wind is present but weak, then the band will frictional convergence enhances ascent.
be located closer to the downwind shore. Sometimes, Convective updrafts can exceed 4-5 m s - in nar-
especially over Lake Erie or Lake Ontario, a midlake row cores 100 m wide. After only a few kilometers of
band will develop at an obtuse angle to the long axis of fetch the depth of the internal boundary layer may
W
700 hPa
Figure 2 Surface and upper air analysesdepictingtypicalsynoptic setting for lake-effect snow in the Great Lakes. (A) 700 hPa heights (solid, dm) and temperatures (dashed, 'C);(B) 850 hPa
heights (solid, dm) and temperatures (dashed, "C); (C) sea-level pressure (solid, hPa) and locationsof highs, lows, and frontsf or times shown. ((A) From Figure 6 and (B,C) from Figure5 in Niziol
TA, Snyder WR and Waldstreicher JS (1995) Winter weather forecasting throughout the eastern United States. Part IV: lake effect snow. WeatherandForecasting10: 61-77. Adapted with
permissionfrom the American Meteorological Society.)
two-dimensional bands, but in time, with sufficient

fetch and heating, these bands can develop into chains
of three-dimensional cells and may eventually
evolve into a regular array of three-dimensional
mesoscale cellular convection. Maximum cloud drop
concentration and liquid water content (e.g.,
0.25 g m - 3 ) occur near cloud-top and increase over
the first two-thirds of fetch. Near the downwind
lakeshore, snow particle production increases (e.g.,
5 L - I) to reduce drop concentration and increase the
height of cloud base.
The most common type of convection associated
with cold air outbreaks over the Great Lakes is
the longitudinal roll. This occurs in other parts of the
world - wherever cold air crosses warm water. Over
the Great Lakes, the rolls are usually oriented parallel
to the wind at the base of the inversion and also to
the low-level wind shear. Classical rolls are typically
2-5km wide and 0.5-2km deep, so the
aspect ratio is roughly 2 to 5. Rolls over the
Great Lakes are 10-20 km wide and 1-2 km deep, so
the aspect ratio is about 10. Some roll circulations
exhibit a multiscale configuration so that the cloud
streets are best developed when the wavelengths of
the convection and rolls are in phase. Such multiscale
phasing can generate cloud streets with aspect ratios
of 10-20.
Linear analytic studies indicate that three different
types of instability mechanisms may be responsible for
rolls. First, inflection point instability theory suggests
that a dynamical instability can develop in a (rotating)
Ekman layer that is neutrally stratified if the cross-
geostrophic wind component exhibits an inflection
point. The most unstable wind profiles lead to rolls
with an aspect ratio of 3 . Rolls are best developed
when they are oriented about 14" to the left of the
Figure3 Lake-effect snowband structures. Upper panel shows a geostrophic wind (in neutral conditions). Second,
lake vortex (Type V) over Lake Huron (C); midlake band (Type I) parallel instability theory involves the curvature of
over Lake Erie (A); shore parallel band (Type IV) over Lake Ontario wind speed profiles parallel to the roll axes and the
(6).Middle panel: multiple bands (Type II)over Lake Huron (A) and boundary layer mean wind shear. This type of insta-
shore parallel band (Type IV) over Lake Ontario (B). Lower panel:
midlake band (Type I) over Lake Huron (A); hybrid band (Type Ill)
bility mechanism generates rolls with aspect ratios
over Lake Erie (Band C). (From Figure 2 in Braham, RR Jr (1983) that are twice as large as those from inflection point
The midwest snow storm of 8-11 December 1977. Monthly instability and closer to those observed during lake-
Weather Review 111: 253-272. Adapted with permission from effect conditions. Third, convective instability (i.e., an
the American Meteorological Society.) unstable boundary layer) in the presence of wind shear
also leads to the formation of rolls that are parallel to
the mean wind shear but with aspect ratios smaller
reach or exceed the depth of the planetary boundary than those from inflection point instability (e.g.,
layer. At some point the air within the thermal internal around 2).
boundary layer becomes moist enough and deep Understanding roll development during lake-effect
enough for the lifting condensation level to be below situations is further complicated by the fact that large-
the top of the boundary layer and cloud to form (Type I eddy numerical simulations suggest that only convec-
boundary layer). Subsequent latent heat release and tive instability is capable of generating rolls. Gravity
radiative heat transfer become important and increase waves generated within the stable air above the
the rate of entrainment. The clouds develop initially as convective boundary layer by wind shear near the
Table 1 Typical values for multiple and shore-parallel band lake-effect storms in western Michigan (lower peninsula)
Characteristics Multiple Single
Mesoscale features
Location Inland western shore Along western shore

Band orientation NW-SE bands N-S band
Band length 30 to 90 km 90 to 180 km
Band width 2to5km 10 to 20 km
Band (boundary layer) depth 1 to2km 2to3km
Aspect ratio 5to10km 10to20
Lake temperature 0 to 10°C 0 to 2°C
1000 air temperature Oto -10°C -10to -20°C
850 air temperature - 5 to- 15°C - 10 to - 30°C
700 air temperature - 15 to - 25°C - 20 to - 40°C
1000 winds W-NWabove 1Oms-' NW-N at 0 to 10m s - '
1000-850 shear dirkpeed 30"/0-10 m s ' 10~/0-5ms-'
' '
~
1000-700 shear dirkpeed 60"/0-20 m s - 20~10-1o m s -
Cloud/precip characteristics
Precipitation Moderate snow ( < 10 cm) Heavy snow ( > 10 cm)

Cell movement With prevailing wind With prevailing wind
Echo persistence Short(<l h) L o n g ( > l h)
Cloud base 1 km <1 km
Cloud top 2km 3km
Updrafts <I m s - ' i2rns-l
Liquid water content 0.2to 1 g m - 3 0.5 to 2 g m - 3
Cloud droplet radius 2 vm 5 Pm
Cloud droplet concentration 100 to 3 0 0 c m - ~ 300 to 500cr1--~
Natural ice nucleus 1 to 10L-1 1 to 10 L - '
Ice-to-water ratio to 1 0 - ~
Ice crystal size 1 to4mm
Ice crystal concentration 1 to l O L - '
Figure 4 Percentages of all days categorized as lake-effect (dashed), wind parallel bands (solid), and shore parallel bands (dotted).
Results based on visible satellite imagery from 1988-93 (October-March). (From Figure 2 in Kristovich DAR, Steve Ill, RA (1995)
A satellite study of cloud-band frequencies over the Great Lakes. JournalofAppliedMeteorology34: 2083-2090. Adapted with permission
from the American MeteorologicalSociety.)
1110 LAKE-EFFECTSTORMS
Figure 5 Example of steam fog and roll development during cold air outbreak over Lake Michigan. Photo taken at 14.25 UTC on
13 January 1998 during the Lake-Induced Convection Experiment. (Photo from David C. Rogers at NCAR taken over Lake Michigan
during the Lake-ICE Experiment conducted during December 1997 to January 1998.)
boundary layer top have also been suggested as a combination of mechanisms may be responsible for
possible mechanism. If the wind shear at the top of the different aspects of roll development or during certain
boundary layer is roughly perpendicular to the bound- conditions.
ary layer wind direction, then bands of convection
parallel to the mean boundary layer wind can be
induced by gravity waves. Latent heat release, cloud
microphysical processes and low-level wind shear
Sensitivity to Synoptic Conditions
(e.g., below 2 0 0 m ) may also influence the develop- The thermal modification of air over relatively cold
ment of rolls. The partial agreement between obser- land as it crosses a relatively warm lake results in a
vations and theoretical studies thus far suggests that a horizontal temperature gradient across the lakeshore,
Figure 6 Vertical cross-section of equivalent potential temperature (solid 1 K contour interval) and cloud frequency (dashed 10%
contour interval) across southern Lake Michigan (from Sheboygan, WI, to Benton Harbor, MI) during a cold air outbreak (on 20 January
1984). Presence of cloud determined by cloud droplet concentrations greater than 10 cm-3. Wisconsin and Michigan land surfaces are
indicated by thick horizontal gray bars. (From Figure 5 in Chang SS, Braham RR Jr (1991) Observational study of a convective internal
boundary layer over Lake Michigan. Journal of the Atmospheric Sciences 48: 2265-2279. Adapted with permission from the American
which can generate a thermally direct solenoidal geometry enhances radial convergence, or if the
circulation similar to that of a sea or land breeze. In nearby orography enhances lifting. These observed
that sense, it can be seen that the lake-air temperature impacts of fetch have been confirmed only recently
difference, speed and direction of the prevailing wind, using analytic and numerical models. For example, it
and height and strength of the capping inversion are has been shown analytically for Lake Michigan that
perhaps the most crucial parameters for determining three convergence centers develop near the eastern
not only how much lake-effect snow will fall, but how shore when a westerly wind prevails, two cells or
it will fall. For example, for conditions character- snowbands develop when a north-westerly wind
ized by moderate lake-air temperature differences prevails, and one midlake band develops when a
(>lO"C), and strong winds (>10 m s - ') blowing northerly wind prevails.
across the short axis of a lake, multiple snowbands The vertical structure of the environmental wind
will usually develop. If the winds across the short axis also affects lake-effect storms. For example, when the
are lighter and/or the temperature difference is greater prevailing wind is parallel to the long axis of Lake Erie,
then some form of shore-parallel band will develop. If moderate directional shear (e.g., between 30" and 60")
the wind is very light against areas of enhanced from the surface to 700 hPa causes weakly to-moder-
concavity then lake vortices can develop. ately precipitating multiple snowbands rather than a
The impact of the wind speed on the lake-effect single intensely precipitating snowband to occur.
response characteristics can be analyzed two-dimen- Stronger shear (e.g., greater than 60") over Lake Erie
sionally in terms of the Froude number Fr = n U / N H , causes the breakdown of precipitating snowbands
where U is the mean wind speed across the short axis of altogether - allowing only instead the development of
the lake, and N and H are the Brunt-Vaisala frequency a non-precipitating stratocumulus deck. While wind
and depth of the planetary boundary layer respective- shear can understandably inhibit the organization of
ly. The values for N and H depend on the lake-air rolls or wider bands, it is not clear whether some of the
temperature difference and stability of the pre-lake- observed effects from shear are simply a manifestation
modified air. The Froude number may be interpreted of a shallow boundary layer.
as the ratio of the mean wind speed U to the gravity The height and strength of the (capping) inversion
wave speed cg = N H / n for a boundary layer of depth are significant limiting factors to cloud depth and
H . Three regimes are important to consider. When therefore to precipitation. Typically, the boundary
FY< 1 then the gravity wave speed exceeds the mean layer must have a depth greater than 1km in order for
flow speed (cg > U). Opposing sea-breeze type circu- lake-effect snow to develop. The most convectively
lations develop with respect to the short axis of the active lake-effect storms have inversion heights ex-
lake and the heaviest precipitation falls over the lake ceeding 3km. Sometimes the capping inversion is
(Figure 7A). This regime corresponds to the midlake entirely absent. During such cases, thunder and
band type of event. When Fr >1, the gravity wave lightning typically accompany copious snowfall rates
speed is less than the mean flow speed (cg < U).The e.g., these exceeding 10cm h - l .
response is characterized by alternating regions of The vertical temperature and moisture distributions
ascent and descent that propagate downwind from the within the boundary layer also play a role. For
leeward shore, and precipitation is diffuse and weak. example, it has long been known that a minimum
This regime corresponds to the multiple-band type of temperature difference of 13°C between the lake
event (Figure 7B). When Fr = 1, the gravity wave surface and the upstream airflow at 85OhPa is
speed equals the mean flow speed (cg = U ) . A required for lake-effect storms to develop. This tem-
resonance condition develops where gravity waves perature difference criterion means that the lapse rate
generated at the downwind shore cannot propagate should be unstable with respect to unsaturated ascent.
upwind. This regime corresponds to the shore-parallel A dry boundary layer is less conducive for lake-effect
band type of event that can generate significant snow than a moist one, although a long fetch can
precipitation at the downwind lakeshore (Figure 7C). compensate for very dry boundary layers. The impact
Setting Fr = 1, and using typical values of N = of moisture is greater for low-stability profiles than for
s-l and H = 2 km, suggests that a value of U M high-stability ones.
6 m s-' maximizes heavy snow along the downwind The presence of large-scale forcing can also influ-
lakeshore, which is consistent with observations. ence lake-effect storm development. Typically, the
The impact of fetch on lake-effect storm develop- coldest air passes over the lakes as high pressure at the
ment has been known since the early 1900s. Long surface moves eastward across them, accompanied by
fetches usually result in heavy snowfalls. Short fetches negative vorticity advection at upper levels and cold
also may produce significant snowfalls if the pre-lake- advection near the surface. Thus, the impacts
modified air is relatively unstable, if the lake shore of synoptic-scale forcing typically act to suppress
Figure 7 Gravity wave interpretation of lake-effect morphology dependence on wind speed. (A) Weak windspeeds create subcritical
(Fr< 1) regime, which allows gravity waves to propagate upwind and downwind and a midlake band and moderate snow to develop over
the lake: (B) strong windspeeds create supercritical (Fr > 1) regime, which allows gravity waves to propagate only downwind and multiple
bands and light snow to develop beyond the downwind lakeshore; (C) moderate windspeeds create near-critical (Fr 1) regime, which
allows gravity waves to travel downwind but traps gravity waves trying to propagate upwind - resulting in an intense shore-parallel band
and heavy snow to develop along the downwind lakeshore. Heavy arrows indicate windspeed and wavy arrows indicate gravity wave
propagation. Plus (+) signs and shaded columns indicate ascent; minus (-) signs and open ovals indicate descent. Asterisks indicate
snowfall.
lake-effect storm development (Figure 2). There are across the region. The synoptic forcing coupled with
however instances when cold air, positive vorticity the cold air that is already established over the region
advection, and even warm advection exist simultane- can combine to generate intense snowfall.
ously over the region. Such situations usually come in Sensible and latent heating from all the Great Lakes
the form of Alberta Clippers (short waves) that (e.g., the lake aggregate) can also influence lake-effect
develop in cold air masses and move south-eastward storms over individual lakes. Basically, if warming
(and moistening) occurs over all the Great Lakes for at In contrast, surface winds across Lake Erie (south of
least a day then surface pressures and stability can the aggregate-induced plume) can become more
drop over a broad region and cause a perturbation westerly. The aggregate-altered winds can cause a
aggregate-scale, low-level cyclonic circulation to de- longer fetch across Lake Erie and a shorter fetch across
velop. The position, the size, and the warmth and Lake Ontario, and can shift the regions of lake-effect
moisture from this aggregate circulation can modify convective bands, so that less (intense) lake-effect
lake-effect precipitation throughout the region. Spe- precipitation can fall along the lakeshores downwind
cifically, when the synoptic-scale flow is north-west- (east) of Lake Ontario and more lake-effect precipi-
erly, aggregate effects can augment snowfall along the tation can fall along the eastern shores of Lake Erie
north-western shores of lower Michigan, and reduce (Figure 8).
snowfall along the south-western shores (Figure 8).
Shore-parallel bands located offshore can migrate
eastward (e.g., onshore) or evolve into multiple bands.
These aggregate affects over Lake Michigan include Forecasting
enhanced westerly flow, increased heat and moisture, The mesoscale nature of lake-effect storms, their
and lower stability. The lake aggregate can also intensity, and the short development times continue
influence lake-effect precipitation in the lower lakes to challenge forecasters. Highly variable snow-to-
region. For example, as the lake-aggregate-induced liquid ratios ( 1 O : l to 50:l) and terrain effects, espe-
plume of heat and moisture extends south-eastward, cially near Lakes Erie and Ontario, can enhance the
surface winds across Lake Ontario (north of the inherently large spatial variability of lake-effect snow
aggregate induced plume) can become more northerly. and hence the forecast challenge. While the problems
Figure 8 Illustrationof lake aggregate effect on prevailing winds and lake-effect snowstorms. On the south side of the developingwarm
plume (shaded oval), north-west winds respond in sea-breeze fashion to become south-west winds with increased fetch and heavy snow
across Lake Erie. Lake effect snows across portions of western Michigan (lower peninsula) may or may not change characteristics.On the
north side, north-westwinds respondto become north-north-westerlywinds with reducedfetch and light snow across Lakes Superior and
Ontario and increased fetch heavy snow across Lake Huron.
of forecasting when lake-effect snow is going to occur this model because of its ability to forecast the large-
have essentially been solved, the equally significant scale conditions to which lake-effect snowstorm
problems of exactly where lake-effect snow will occur, development is very sensitive. The LFM model, in
what form(s) it will take, how intense it will be, and conjunction with forecaster decision trees based on
how long it will last remain outstanding forecast key large-scale parameters, and experience, allowed
issues. A combination of high-resolution numerical forecasters to at least be able to issue general forecasts
weather prediction models, statistical methods, Dop- of when lake-effect snow was going to occur.
pler radar, and forecaster savvy are the basic forecast Two operational models currently being used in-
tools. clude the nested grid model (NGM), with 4Skm
Numerical models have come a long way since the horizontal grid spacing, and the Eta model. Several
use of the limited fine mesh (LFM) model. The different versions of the Eta model are run at several
horizontal grid spacing (1SO km) and the exclusion different resolutions and times including one run at
of the lakes in terms of their heat, moisture, and 12 km horizontal grid spacing four times daily. The
momentum characteristics in that model precluded increased resolution in the Eta model has been
any explicit model development of lake-effect precip- especially helpful for identifying areas where lake-
itation. Regardless, operational forecasters relied on shore enhanced snowbands may develop. Recently,
e 40
Lake-effect guidance: Ontario Output from NGM, 122 18 Feb 93:
Forecast parameters
a Conditional Moderate Extreme
a Variable 12h 24 h 36 h 48 h
Wind direction (degrees)
700 hPa 260 310 290 270
850 hPa 290 300 290 240
B.L. 290 290 290 180

lo15 20 25 30 35 40
+ Change In wind direction with height
Temp (lake) -Temp (700 hPa) ("C)
8501700 030 010 000 030
12 h extreme instability ...030 degree shear from B.L.1700 030 020 000 090
700 hPa to SFC ...and 090-mile fetch at 850 hPa.
Fetch (miles)
24 h moderate instability...020 degree shear from
700 hPa to SFC ...and 070-mile fetch at 850 hPa. 850 hPa 090 070 090 150
36 h moderate instability...000 degree shear from
700 hPa to SFC...and 090-mile fetch at 850 hPa. Temp ("C)
48 h moderate instability ...090 degree shear from
700 hPa -30 -28 -21 -1 8
700 hPa to SFC...and 150-mile fetch at 850 hPa.
850 hPa -2 1 -23 -1 7 -1 4
B.L. -1 0 -1 4 -1 0 -09
Lake 02 02 02 02
TSm3 layer inversion intensity
TS-T3 -08 -08 -03 -05
Vertical velocity (microbarls)
700 hPa +02 -02 -02 eo0
Figure 9 Lake-effect snow guidance product for 1200 UTC on 18 February 1993 generated for Lake Ontario at WSFO Buffalo. TS and
T3 correspond to the surface and 900 hPa temperatures respectively and BL corresponds to the boundary layer. (From Figure 10 in Niziol
TA, Snyder WR and Waldstreicher JS (1995) Winter weather forecasting throughout the eastern United States. Part IV: Lake effect snow.
Weatherand Forecasting 10: 61-77. Adapted with permission from the American Meteorological Society.)
several forecast offices have experimented with run- heat, momentum, and moisture fluxes - and simula-
ning locally high-resolution mesoscale models. The ting more accurately the convective precipitation. A
Forecast Office in Buffalo, New York, has been fourth challenge is initializing more accurately the lake
running a 10 km version of the PSUNCAR model surface temperatures, which are specified currently
MM5 since 1996. The Forecast Office in Detroit, using AVHRR satellite data that represent a multiday
Michigan, has been running a 6 km version of the Eta average and may have gaps because of persi-
model since 1998. Both of these offices have reported stent cloudiness. Finally, the simulation of sub-
the ability to provide more specific and more accurate sequent changes in lake surface temperatures, which
forecasts. may also improve forecast accuracy, has yet to be
Despite significant and continuing improvements in included.
numerical weather prediction, lake-effect snow con-
tinues to challenge the abilities of even the most
sophisticated numerical models, because of several See also
inadequacies. These inadequacies include horizontal Air-Sea Interaction: Storm Surges. Boundary Layers:
resolution that is still too coarse for resolving the 2- Convective Boundary Layer; Modeling and Parameterizat-
4 km wide bands, convective schemes tuned originally ion; Overview. Climate: Overview. Convective Storms:
for deep (tropical) convection that are inappropriate Convective Initiation; Overview. Hurricanes.Mesoscale
to simulate intense shallow precipitating convection, Meteorology: Mesoscale Convective Systems; Models.
and boundary layer schemes that are too simplistic to Numerical Models: Methods. Synoptic Meteorology:
develop the low-level temperature, moisture, and Forecasting. Weather Prediction: Regional Prediction
Models.
cloud-microphysical structures that exist within
lake-effect snow environments near the surface.
To address some of these inadequacies, forecasters Further Reading
currently use various statistical methods. These meth-
ods were used almost exclusively prior to the existence Braham RR (1995) The midwest snow storm of 8-11
December 1977. Monthly Weather Review 111:
of numerical models. As early as the middle of last 253-271.
century, various investigators had outlined conditions Chang SS and Braham RR (1991) Observational study of a
necessary for prolonged lake-effect storms to occur at convective internal boundary layer over Lake Michigan.
the eastern end of Lake Erie. Afterwards, more Monthly Weather Review 48: 2265-2279.
sophisticated statistical models, based on multiple Kristovich DAR and Steve RA (1995) A satellite study of
discriminant analysis, the perfect prog (PP) method, cloud band frequencies over the Great Lakes. Journal of
model output statistics (MOS), and classification and Applied Meteorology 34: 2083-2090.
regression trees (CART), were developed for many of Niziol TA, Snyder WR and Waldstreicher JS (1995)Winter
the lake-effect snow belts. Currently, the use of weather forecasting throughout the United States,
numerical model output in terms of larger-scale part IV. Lake-effect snow. Weather and Forecasting 10:
61-77.
features, coupled with highly tuned, sophisticated
Norton DC and Bolsenga SJ (1993) Spatiotemporal trends
statistical models, has proven a very effective forecast in lake-effect and continental snowfall in the Laurentian
method (Figure 9). Great Lakes, 1951-1980. Journal of Climate 6:
Remaining challenges for numerical lake-effect 1943-1956.
snow forecasting include resolving the lakeshore Sousounis PJ and Mann GE (2000) Lake-aggregate meso-
geometry and nearby terrain, simulating more accu- scale disturbances, part V.Impacts on lake-effect precip-
rately the evolution of the boundary layer - including itation. Monthly Weather Review 128: 728-743.
1 116 LAND-ATMOSPHERE INTERACTIONS / Overview
Contents
Overview
Canopy Processes
Trace Gas Exchange
of downward thermal radiation. When cloud droplets

Overview are able to join into larger drops, or water vapor
deposits to form the crystals that grow large enough to
R E Dickinson, Georgia Institute of Technology, fall, the resulting precipitation supplies water to the
Atlanta, GA, USA soil. This water is used by plants to maintain their
assimilation of carbon dioxide and hence to grow, and
the surplus that is not so used fills streams, rivers,
ground water, and reservoirs, and is either captured for
human requirements or delivered at the mouths of
Introduction rivers to the ocean. The transpiration of plants in using
soil water is the major contributor to the evaporative
The atmosphere forms with land and the oceans a
cooling of the land surface by the atmosphere in moist
closely coupled system. This system is driven by the
regions.
absorption of solar radiation. Much of this radiation is
Besides the radiative and hydrological inputs, the
absorbed by the underlying land and ocean surfaces.
winds, temperature, and moisture of the overlying
Land differs from the ocean in having a much lower
atmosphere also strongly affect the land surface, and
capacity to store the thermal energy received from the
Sun. Rather, solar energy raises its temperature until it in turn are affected by the land surface. Describing this
adjusts its properties to lose as much energy as it has influence requires knowledge of the mechanisms by
received. This continual tendency of the land surface which the land surface achieves energy balance.
to achieve a balance with the solar energy it receives,
together with the influence of the overlying atmos-
phere on this balance, is an important determinant of What Determines the Land Surface
the day-to-day weather we experience and is the prime
cause of climates over land. Because the properties of
Absorption of Solar Radiation?
the overlying atmosphere are strongly controlled by Solar radiation is the primary source of warmth for
the influences of underlying surfaces that it has land, and the driver of plant photosynthesis. Designs
previously been in contact with, the climate at a of homes take advantage of this by siting their walls,
particular location acts to average influences from windows, and landscaping to capture the levels of light
elsewhere. Thus, the climate of near-coastal regions needed for optimum growth of indoor and outdoor
appears similar to that of adjacent ocean areas, plants, to reduce the solar heating at times of year and
whereas that of continental interiors combines the day when outside temperatures exceed human com-
direct influence of local solar radiation with how such fort levels, and to maximize solar heating when
radiation has influenced the atmosphere in adjacent outside temperatures are uncomfortably cold. We
regions. like to wear light-colored clothing in the summer to
The atmosphere influences physical, biogeochemi- maximize the reflection of solar radiation and dark
cal, and biological properties of the land surface. clothing in the winter to absorb solar radiation. These
Incident solar radiation, besides its heating of land, patterns of human behavior are natural responses to
also provides the energy required to drive the photo- our understanding of the varying supply of solar
synthesis of plants and other organisms with chloro- radiation.
plasts. The clouds that form when atmospheric water The amount of solar radiation absorbed by any
vapor condenses into water droplets substantially surface is simply the product of how much solar energy
reduce the available solar radiation, but this cooling is incident on that surface, and the fraction of that
effect is compensated to some extent by their increase incident radiation that is absorbed. The flux of solar
LAND-ATMOSPHERE INTERACTIONS/ Overview 1117
radiation at any location is a vector; that is, it has a albedo is not simply determined by the reflection from
magnitude, which is the energy it carries, and a a single flat surface but also depends on how much of
direction, which is the direction from the Sun to that reflected light is absorbed by other surfaces. For
location. The key parameter determining how much example, snow at the bottom of a deep canyon may
energy is incident on a given surface is the angle reflect most of the solar radiation it receives, but the
between the direction of the Sun and that surface. If the canyon walls may in turn mostly absorb this reflected
surface is turned to face the Sun, i.e., directly overhead light so that the albedo of the snow canyon system is
for a flat surface on the ground, it receives its much less than that of the snow alone. This reduction
maximum solar energy, the same as the magnitude of of surface albedos by light trapping is perhaps most
solar flux at that level. Otherwise, the surface receives widespread for vegetated surfaces, where the plants’
an amount reduced by the cosine of the angle between requirements for photosynthesis dictate that the
the direction of the Sun and the line that the Sun would absorption of solar radiation be maximized. Leaves
take for maximum receipt of solar energy. are so arranged that plants as a whole reflect less than
This geometrical reasoning for the determination of half as much solar radiation as do the individual
the amount of incident solar radiation enters in many leaves.
ways our descriptions of climate over land and climate The albedo of land surfaces, and especially the
in general. On average for the Earth, it is in nighttime vegetated components, depends strongly on the wave-
half the time, and during daytime, on average, the length of the solar radiation. The wavelengths of solar
angle formed between a vertical line from the surface radiation that plants use are essentially the same as
of the Earth and the Sun is 60”. Consequently, the those of human vision, that is ‘visible light’. Approx-
daytime average sunlight received at the top of the imately half of solar radiation occurs at the longer
atmosphere is half of that which a point receives when wavelengths of the near infrared. These longer wave-
the Sun is directly overhead, and the day-night average lengths are not used for photosynthesis and may
is a quarter of that received from an overhead Sun. overheat plants, so that plant leaves have much higher
Solar radiation at the surface is additionally affected albedos at these wavelengths.
by greater atmospheric reflection and absorption of
radiation when the Sun is closer to the horizon. These
geometric factors reduce the solar radiation received What Determines the Fluxes of Water
in high latitudes and in winter, and enhance the solar and Dry Atmospheric Energy from the
radiation received in the tropics and in summer, hence
Land Surface?
largely determining the seasonal and geographic
variations of climate. However, quantitative details For given amount of solar absorption and net radia-
are also highly dependent on how the atmosphere tion, the land surface can still reach a wide range of
responds to the seasonal and geographical variations states and interact with the atmosphere over a wide
of solar energy delivery and in turn affects the land range of possibilities. Over the course of a day, some of
surface. the daytime heating can conduct downward into the
Clouds are the primary mechanism by which the soil and be released again at night. Averaged over day
atmosphere alters the receipt by the land surface of and night, the net absorption of radiation is largely
solar radiation. Also important are other smaller balanced by fluxes of energy that the land surface
atmospheric particles referred to as aerosols. The delivers to the atmosphere. The turbulent motions of
amounts of atmospheric gaseous absorbers of solar air near the land surface determine these fluxes.
radiation, such as water vapor and ozone can also be Convection and mechanical mixing in turn determine
significant. The total net radiation that heats the land the intensity of the turbulence. The intensity of
surface includes that of downward thermal emission mechanical mixing is determined by the strength of
from the atmosphere, which also depends on clouds surface winds and by the roughness of the surface.
and water vapor, and is reduced by the upward Positive net land surface heating normally occurs only
emission of thermal radiation. This ‘Earth’ radiation during the day, and it usually has small negative values
depends on surface temperature but more weakly than at night.
do the fluxes of energy from surface evaporation and The energy carried from the surface by water vapor
dry sensible heat, so in discussing the atmospheric is simply the energy that was required to evaporate the
radiative fluxes to the surface in the next section, we water from its liquid state at the land surface or
use the term ‘total net radiation’. equivalently the energy that will be released when this
The fraction of solar radiation that is absorbed by a water vapor is converted back to liquid form through
surface is referred to as its albedo. Because land the formation of clouds and precipitation. The trans-
consists of surfaces oriented in all directions, the port of dry atmospheric energy is the energy carried
1118 LAND-ATMOSPHERE INTERACTIONSI Overview
upward by relatively warm air rising and cold air Hence, the fluxes of water vapor from the land surface
sinking. This transport of dry atmospheric energy to the atmosphere largely occur through extraction of
provides the intense daytime convection that generates soil water by roots and transport through plants and
the boundary layer turbulence needed to remove the out their leaves. This is called transpiration. Normally,
energy supplied to the surface by daytime radiation. with adequate soil moisture, the plant leaves are the
The flux of dry energy is proportional to the difference main obstacle to movement of water to the atmos-
between land surface temperatures (Le., that of leaves phere and hence an important control on the Bowen
and soil surface) and that of the overlying air. The flux ratio. The leaves lose water primarily through tiny
of water vapor is proportional to the difference holes called ‘stomates’. A leaf normally has to contain
between water vapor concentrations at the surface water to maintain its structure, and if the roots cannot
and that of the overlying air. Where materials at the supply water as fast as it loses it, it wilts. Hence, its
surface are supplied with water, such as inside leaves water loss can be a threat to its survival. Why do leaves
and in moist soil, the consequent water vapor concen- have this stomatal loss mechanism for water and what
tration depends only on temperature. How the land determines the extent by which the stomates lose
surface responds to a given amount of net radiation water? There may be some benefit from the resulting
depends on how it divides the removal this energy evaporative cooling in warm regions where leaves may
between evaporation and dry atmospheric energy flux. be threatened by thermal damage if their leaves reach
The ratio of dry atmospheric energy flux to energy high enough temperatures. However, the primary
carried by water vapor is called the ‘Bowen ratio’. reason is an even more basic element of plant
The Bowen ratio depends on surface temperatures, requirements for growth and survival. The photosyn-
on the relative humidity of the overlying air, and on thesis of plants uses light from the Sun to convert
limitations to removal of water from the soil. In the carbon dioxide to the carbohydrates and proteins
extreme, if the surface is very dry, there is very little or needed for plant growth (some of which eventually
no water available to be moved into the atmosphere become our own food). This can only work if the
and the Bowen ratio becomes very large. The smallest plants receive, besides light, an adequate supply of
Bowen ratios result in warm areas when there are no carbon dioxide. The stomates pass carbon dioxide
limitations on the removal of water from the surface from atmosphere to the inside of leaves where it can
but the atmospheric relative humidity is extremely reach the chloroplasts and be used. This function of the
low. This can only happen when such conditions are stomates, however, is in a somewhat delicate balance
local and differ from those of the surrounding land with the controls of the stomatal opening as to how
surface, because then the consequent flux of energy much water is lost. If the net stomatal passageway to
will exceed that provided by net radiation and will the atmosphere is too large, the leaves may lose more
depend on air energy carried to that point from water than the soil can provide, whereas if it is too
elsewhere. This is referred to as the ‘oasis’ effect. small, the leaf may be starved of carbon dioxide, at
The role of temperature in determining the Bowen least relative to the light energy available to convert it
ratio is isolated by consideration of its value when to carbohydrate.
surface air is at 100% relative humidity. Daytime Plants and related aspects of the land surface
evapotranspiration will still occur because the heating exchange not only carbon dioxide but other important
of the surface will raise the concentrations of water in gases with the atmosphere, as addressed in the next
the near surface soil and inside plant leaves to larger section.
values of atmospheric humidity than that of the
overlying air. However, because these concentrations
increase substantially with larger temperatures, the What Determines the Fluxes of Carbon
Bowen ratio will be less at larger temperatures. Dioxide and Other Gases to and from
In sum, for moist surfaces the Bowen ratio will be
greater for colder temperatures and drier air. The
the Land Surface?
Bowen ratio is further increased by various surface As concluded in the previous section, the flux of
resistances that affect only the movement of water carbon dioxide into plants is closely linked to the
vapor or affect it more than the movement of dry heat. transpiration by vegetation and hence to much of the
Perhaps most obvious is that when the surface soil is flux of water from the land into the atmosphere.
dry, water vapor must diffuse upward from deeper soil However, if the removal of carbon dioxide were
layers, and the rate at which this diffusion occurs may simply one-way or unbalanced in the direction of
limit transport of water into the air. removal, we might expect the atmosphere to become
The presence of precipitation and hence soil mois- exhausted and no longer provide an adequate supply
ture is normally accompanied by growth of plants. of this essential plant nutrient. This may not be such a
LAND-ATMOSPHERE INTERACTIONS/ Overview 1119
far-fetched possibility, because the amount of carbon The plants themselves use about half of the carbo-
from carbon dioxide currently contained in land hydrate energy stores for their own needs. They
plants is as large as that held in the atmosphere, require this energy both to convert the carbohydrates
including the atmosphere over the oceans; and that to more complex molecules such as proteins and fat
stored in the soil that came from plants is considerably compounds and to continuously repair complex mol-
larger. The possibility that large amounts of carbon ecules such as enzymes that tend to ‘wear out’. Some
dioxide can be removed from the atmosphere is energy is also needed in roots to facilitate the acqui-
demonstrated by the large amounts that have been sition of soil nutrients either by directly powering the
converted to fossil fuels at earlier times in Earth’s movement of soil ions or by indirectly by feeding other
history. Indeed, currently accessible coal deposits have soil organisms such as mycorrhizae that facilitate the
been estimated to hold over ten times the carbon movement of insoluble soil nutrients to the roots.
currently contained in the atmosphere, and at the time The carbon that the plants do not use themselves,
these deposits were formed the atmosphere may have and that is not harvested by humans or other animals,
held several times as much carbon dioxide as it does is delivered to the land surface and soils as dead plant
now. materials. This dead plant material feeds many small
The oceans hold even more carbon than does the to microscopic organisms, at the bottom of the food
land, dissolved as bicarbonates and carbonates in sea chain mostly bacteria and fungi, the ultimate decom-
water. This carbon eventually makes its way into posers of plant materials that are responsible for
limestone sediments, which through movements of the return of carbon dioxide to the atmosphere. All these
Earth’s crust are returned to land and the atmosphere. biological processes are strongly controlled by the
Fortunately, the natural exchanges of carbon dioxide physical environment, the overlying atmosphere, the
between land, atmosphere, and ocean are normally water provided by precipitation, and the supply of
very close to being in balance. The land in particular energy and warmth by the Sun.
acts to move as much carbon dioxide back into the The only major exception to the conversion of
atmosphere as it removes. Currently, land does carbon compounds to carbon dioxide by living
appears to taking up more carbon dioxide than it organisms is that fraction that is oxidized directly by
gives back to the atmosphere. However, this is in high-temperature combustion: that is, fires. Human
response to the extra carbon dioxide being supplied by use for energy, human land management practices,
the human conversions of fossil fuels back into carbon and natural or accidently started fires release compa-
dioxide, which still causes atmospheric carbon diox- rable amounts of carbon dioxide, in total perhaps as
ide to increase. Hence the net land removal of carbon much as 20% of that captured by plants on average.
dioxide must be regarded as a very valuable service. It Natural fires are themselves a land process that
reduces the effort we otherwise might have to make to strongly interacts with the atmosphere.
limit our usage of fossil fuels out of concern for the Besides carbon dioxide, numerous other important
consequent climate change. Although we know much carbon compounds are exchanged between the land
about how the land exchanges carbon with the surface and the atmosphere. Especially important is
atmosphere, our current understanding of the details methane, which is emitted in places where there is very
of the net removal into land is sufficiently poor that we little oxygen, such as swamps and rice paddies. In
cannot determine under what conditions this removal addition, complex organic compounds are given off in
could cease or possibly reverse. copious amounts by leaves and contribute to enhanc-
How is the natural cycle of carbon dioxide removal ing regional levels of ozone and photochemical smog.
from the atmosphere into plants balanced by land Forest fires are an especially important mechanism for
supplies back to the atmosphere? The carbohydrates the supply of other complex organic materials to the
that the plants produce from photosynthesis are atmosphere, including important aerosols.
largely eaten. What this means, at the cellular level, Without going into as much detail was done for
is that the mitochondria present in all eukaroyate cells carbon, other elements important for living cells that
metabolize (that is oxidize) the carbohydrates into are exchanged between land and the atmosphere
energy and carbon dioxide. The energy supplies the include nitrogen and sulfur. The molecular nitrogen
needs of the cell and the more complete organisms, and in the atmosphere is very inert to chemical change, but
the carbon dioxide is returned to the atmosphere. The enough must be supplied to the land surface, in the
final step of this process in our bodies is our exhaling of form of nitrate or ammonium compounds, to maintain
carbon dioxide from our lungs. Only about 1%of this this element in living cells. Ammonium compounds
carbohydrate energy is metabolized by humans to are generated naturally by nitrogen-fixing organisms
return the carbon dioxide to the atmosphere. Where is that live freely in the soil or are attached to the root
the rest? systems of some plants, such as the legume family.
1120 LAND-ATMOSPHERE INTERACTIONSI Overview
Humans now add even larger amounts of ammonium are especially important. Likewise, aspects of the
and nitrate to the soils directly as fertilizers or vegetation that determine its albedo and its control of
indirectly as wet and dry deposition of these com- movement of water between the soil and the atmos-
pounds from atmospheric pollution. Although some phere can be characterized from observation. The
sulfur moves from the land into the atmosphere, much latter include the roughness of the vegetation for
more is initially put into the atmosphere by the producing turbulence in the air that flows over it, the
combustion of fossil fuels and this is then deposited degree of resistance the leaves produce to water
back to the land. movement through them, and the depth into the soil
Land processes supply ammonia and oxides of to which the roots are able to mine water.
nitrogen to the atmosphere, which are major elements These required properties can be estimated by
in determining aerosols and atmospheric chemistry. mapping of the world’s land surfaces into plant
Most of the ammonia comes from areas where it has functional types or major biomes. These two ap-
been concentrated by human management practices, proaches can overlap where biomes consist largely of
such as from feed lots and heavily fertilized fields. single plant functional types. For example, the tropical
Nitrous and nitric oxides escape from the soil as by- evergreen biome consists largely of the tropical ever-
products of the incomplete nitrification of soil ammo- green tree plant functional type. On the other hand,
nium, which is its bacterially mediated conversion to the savanna class biome consists of a mixture of trees
nitrate. and grasses. For either of these or more detailed land
cover classifications and for determination of albedos,
global mapping requires the use of satellite imagery of
the land surface. Such measurements are not as simple
How Do We Describe the Geographical as determining the presence or absence of clouds but
Variations of Land Properties in Ways can be done with the current generation of satellite
That Are Meaningful for Determining instruments that NASA has had in orbit since the year
Their Interaction with the 2000.
Atmosphere?
Models of the Earth system that describe the interac- See also
tions of the land with the atmosphere require not only
the best efforts to describe the processes reviewed Boundary Layers: Overview; Stably Stratified Boundary
above with equations but also knowledge of the Layer; Surface Layer. Global Change: Biospheric Im-
pacts and Feedbacks. Land-Atmosphere Interactions:
geographically varying properties of the land surface
Canopy Processes; Trace Gas Exchange. Solar Terres-
that determine these interactions. Many of the needed trial Interactions.
properties are generated by the models themselves,
such as those involving the absorption of solar
radiation, temperatures, and aspects of the hydrolog-
ical cycle such precipitation and soil moisture. How-
Further Reading
ever, some necessary features are more accurately Dickinson RE (1983) Land-surface processes and climate:
observed than modeled, or are best treated by a surface albedos and energy balance. Advances in Geo-
combination of modeling and observations. physics 25: 305-353.
Some particularly important parameters that need Hartmann DL (1994) Global Physical Climatology. Or-
to be constrained by observations are the land albedos, lando, FL: Academic Press.
Rosenburg NJ, Blad BL and Verma SB (1983)Microclimate,
characteristics of the soil, and characteristics of the
The Biological Environment. New York: Wiley.
vegetation. Features of the soil and other aspects of the Schlesinger WH (1997) Biogeochemistry: A n Analysis of
land surface that influence the movement of water and Global Change. Orlando, FL: Academic Press.
help determine how much ends up in streams and Sellers PJ, et al. (1997) Modeling the exchange of energy,
rivers rather than being returned directly to the water, and carbon between continents and the atmos-
atmosphere through evaporation and transpiration phere. Science 275: 502-509.
LAND-ATMOSPHERE INTERACTIONSI Canopy Processes 1121
Canopy Processes
P D Blanken, University of Colorado at Boulder, through individual leaf stomata. A measure of the
Boulder, CO, USA openness of leaf stomata at the canopy scale, canopy
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. conductance, responds to the air temperature, humid-
ity, wind speed (turbulence),and CO2 concentrations,
all of which are influenced to various degrees by the
Introduction canopy. As a result, a canopy often creates its own
The arrangement of leaves and supporting structures microclimate.
(branches and stems), integrated with height between A description of canopy processes requires integra-
the ground and top of the vegetation, forms our tion or aggregation of processes across spatial scales
definition of a vegetation canopy. There is almost ranging from the cellular to canopy scales. Such
infinite variety in the morphology of canopies in the integration remains an issue, as an accurate represen-
world’s vegetation. Canopies may have several well- tation of canopy processes requires the ability to ‘scale
defined layers, such as in rainforests, or may have one up’ from the leaf to the canopy level measurements,
well-defined overstory, with or without a vegetated and ‘scale down’ back to the leaf from canopy level
groundcover (the understory). A typical way to measurements. In this article, canopy processes will be
quantify the canopy is in terms of canopy closure (or described following a scaling-up approach, covering
gaps where sky is visible) or the leaf area index, L , the following topics: photosynthesis, canopy conduct-
equal t o half of the total leaf area per unit ground area. ance, light penetration, canopy productivity, wind and
The latter takes into account dense, multilayered turbulence, and remote sensing.
canopies where L can exceed one. In general, L varies
zonally across the globe, varying roughly with precip-
itation and temperature, reaching a maximum near
Photosynthesis
the Equator and a minimum near the poles. Leaves are surfaces remarkably well adapted for
Superimposed on the world’s spatial variability in intercepting light. The conversion of light into carbo-
canopy structure, there is often a great deal of hydrates and starches occurs through the process of
temporal variability for a given canopy. In the long
term, the canopy changes through the ecological
processes of succession, as a disturbance such as fire
destroys the canopy and L generally increases over
photosynthesis:
CO2+H20 -
light. plant (CH20)+02
111
perhaps hundreds of years as different species colonize Only short-wave (solar) radiation between wave-
the area. Annually, L can vary from a maximum lengths of 400 and 700 nm (photosynthetically active
during the summer to a minimum during the winter as radiation: PAR), and especially 680 and 700nm, are
deciduous canopies seasonally lose their leaves, effec- absorbed by the plant for photosynthesis. The en-
tively shedding most of their canopy. On a short-term trance way for COZ into the leaf and subsequent loss of
basis (hours) some species can change leaf orientation water vapor is through the stoma (plural, stomata)
in response to water stress (e.g., soybeans), track the (Figure 1).Stomata are small openings (typical length
sun (e.g., alfalfa, cotton, soybeans), or shed leaves in 15 pm) typically on the underside of leaves flanked by
response to water stress (e.g., cotton). Even within a two guard cells. These guard cells regulate the aperture
single species at any given time, leaf structure can vary of the stomata by changing their turgor (internalwater
from being thick near the top of a light-rich canopy to pressure). The plant actively regulates the stomata
thin within a shaded canopy. Oak leaves, for example, aperture (often expressed as the stomatal conduct-
grown in the upper canopy tend to be smaller, more ance) by changing the guard cell turgor through
deeply lobed, and inclined at steeper angles than those changing the potassium ion concentration inside the
grown beneath in the shade. guard cells. If, for example, the tension of water inside
Regardless of this spatial and temporal variability, it the water-conducting vessel in the plant (the xylem)
can be argued that the motivation driving all atmos- became too great, then, to prevent cavitation inside the
pheric-related canopy processes is to achieve an xylem (irreparably damaging the xylem), potassium
optimum L that maximizes light absorption for would flow out of the guard cells. In response to the
photosynthesis while minimizing transpiration water developing water potential gradient, water would flow
loss. The quantity of light absorbed as it passes out of the guard cells and into the adjacent epidermal
through a canopy drives the process of photosynthesis, cells. The turgor in the guard cells would decrease,
which is connected to the transpiration water loss and the stomata would close, thus reducing water loss
1122 LAND-ATMOSPHERE INTERACTIONS/ Canopy Processes
Figure 1 Plan (A and C) and cross-sectional (Band D) views of an open (A and B) and closed stomata (C and D). The transfer of water
vapor out of the leaf, and COPinto the leaf, are regulated by the aperture of the stomatal pore, quantified as the leaf stomatal conductance
or its reciprocal, the leaf stomatal resistance.
(also carbon gain). The signal to transfer potassium in (PCR) cycle with the C4 photosynthetic carbon
or out of the guard cells is thought to be controlled by assimilation (PCA) or crassulacean acid metabolism
the hormone abscisic acid (ABA)produced at the root (CAM)pathways.
hairs and carried through the xylem to the guard cells All photosynthetic eukaryotes reduce COZ to car-
in the leaf. bohydrates using the PCR (Calvin) cycle. In this cycle,
atmospheric C 0 2 and water are combined with a five-
carbon acceptor molecule (ribulose l,.S-bisphosphate)
to generate two molecules of a three-carbon interme-
Biochemical Pathways
diate, phosphoglycerate. This carboxylation reaction
As eqn [ 11 shows, water is required for photosynthesis is catalyzed by the enzyme ribulose bisphosphate
as it provides the solution in which carbon carboxylase/oxygenase (rubisco), the most abundant
can dissolve. In addition to modifying canopy enzyme in the world (estimated at lo7 tonnes). Using
and leaf morphology to optimize light absorption, the photochemically derived energy adenosine tri-
plants have evolved a number of strategies to phosphate (ATP), phosphoglycerate is then reduced
maximize photosynthesis when water is the to form carbohydrates (sucrose, starch). Next, the COZ
limiting resource. Biochemically, this includes supple- acceptor molecule is regenerated, again using ATP. At
menting the C3 photosynthetic carbon reduction a thermodynamic efficiency of about 90%, six turns of
LAND-ATMOSPHERE INTERACTIONS / Canopy Processes 1123
this cycle are required t o produce the equivalent of 1 under water deficits tend to be smaller than those
glucose molecule from 6 carbon atoms. grown under nonstressed conditions. Existing leaves
The enzyme rubisco discriminates poorly between often senesce and fall off during periods of water
C 0 2 and 02,and will oxygenize ribulose 1,5-bis- stress. These leaf area limitations and adjustments are
phosphate and release C02 if combined with 0 2 . This usually the first responses of plants to slow, long-term
process, known as photorespiration (the C2 photores- dehydration. The structure of the canopy, and of
piratory carbon oxidation or PCO cycle), and diamet- individual leaves (e.g., edge shapes, leaf hairs) can also
rically opposed to photosynthesis, results in a loss of affect water loss. For example, a canopy that is
up to 50% of the C02 gained by photosynthesis. To aerodynamically rough (e.g., forests) tends to enhance
recover some of the photorespired C02, species found water and heat loss by having a low aerodynamic
in somewhat dry and high-energy (light) tropical and resistance (see ‘Wind and Turbulence’ below), whereas
subtropical climates (e.g., tropical grasses, sugarcane, aerodynamically smooth canopies (e.g., crops) tend to
corn, sorghum) have a particularly well-developed have a large aerodynamic resistance which suppresses
ability t o form C4 acids malate and/or aspartate. These water loss and heat exchange. Prolonged drought can
acids are broken down near the site of carboxylation also lead to an expansion of the root system into
(the mesophyll) to regenerate an additional COZ deeper soils where water may be more plentiful than at
acceptor molecule. Hence, C4 species typically have the surface.
high photosynthesis rates, and a low stomatal con- In response to short-term water stress, or after the
ductance (high water use efficiency; see below), yet plant has reached its maximum leaf area, stomatal
require high temperatures and ample light, and are closure can effectively reduce transpiration water loss.
therefore seldom found in cool or shady locations such Stomatal closure occurs either passively when rapid
as those often found beneath a canopy. water loss from the guard cells cannot be replenished
In especially arid environments, CAM species such by water from adjacent epidermal cells or else meta-
as cacti have the ability to reduce water loss by closing bolically when solute transport from the guard cells
their stomata during the day, and opening them at results in water loss, decreased turgor, hence closure. It
night. This high water use efficiency is achieved by is thought that abscisic acid (ABA)delivered to the leaf
obtaining C02 at night and fixing it in the form of the from the roots, in addition to playing a role in leaf
acid malate, which is stored in vacuoles. During the abscission, also plays a key role in initiating stomatal
day, carboxylation of malate releases C02, which closure, as it is at the roots where drying is first
cannot escape back to the atmosphere since the detected.
stomata are closed, is reduced to carbohydrates via
the C3 PCR cycle.
Canopy Conductance
The transpirational water loss from a leaf is con-
Water Use Efficiency trolled not only by processes internal to the leaf,
The success of these various processes of maximizing but also external processes, and there are con-
carbon uptake while minimizing water loss is ex- siderable feedbacks between the two. The aggregation
pressed by the water use efficiency, W E : of leaves into a canopy alters not only the light regime,
but also temperature, wind, and the concentration of
Moles of C02 fixed gases such as water vapor and C02. There are also
WUE =
Moles of H20 transpired PI concerns with scaling between the leaf and canopy;
that is, does a canopy behave like a ‘big leaf‘, so that
This ratio captures what is often referred to as processes observed at the leaf level can be simply
the photosynthesis-transpiration dilemma; how multiplied to the canopy level as a function of leaf
to maximize carbon intake while minimizing water area?
loss? In regions where water is nonlimiting, canopies An Ohm’s law electrical analogy ( I = V / Y where
, V
usually maximize leaf area to maximize light is voltage, I current, and Y resistance) has been
interception without regard for water loss. In successfully applied to describe the transfer of heat,
regions where water is a limiting factor (sometimes water vapor, and C02 between the leaf and the
seasonally or even diurnally), various strategies in atmosphere, and between the canopy and the atmos-
addition to the various biochemical pathways just phere (Figure 2),and hence forms a basis for predicting
discussed may be used to minimize transpirational either the leaf stomatal ( g L ) or canopy (gc)conduct-
water loss. ance. Note that conductance is the reciprocal of
Leaf and canopy morphology are often modified to resistance (i.e., g = I /r ). Simply stated, the flux or
minimize water loss. New leaves and stems grown exchange per unit area over a given time (i.e., I ) is
Heat Water
vapor
TR eR
Ta ea Ca
Heat Water co2

vapor
Ta ea Ca
fbC rbC rbC
0
rb lb
f CC
rC
eS
Figure 2 Schematic of transfer of heat, watervapor, and C 0 2 between a leaf (A) or canopy (B) and the atmosphere. At the leaf level, heat
transfer depends on the difference between the leaf surface (Ts)and air temperature (Ta)divided by the leaf's boundary layer resistance
( f b ) , created by the transfer across the layer of still air adjacent to the leaf surface. The transfer of water vapor from the leaf is described by
the difference in the saturation vapor pressure calculated at Ts (e*(Ts))and the vapor pressure at the leaf surface (e) divided by the leaf
stomatal resistance (rL),and the difference bewteen es and the atmospheric vapor pressure (ea)divided by rb.Similarly, COPtransfer into
the leaf depends on the difference in COPconcentration in the atmosphere (ca)and the leaf surface (cs) divided fb, and the difference
between cs and the leaf's internal COPconcentration (q)divided by rL. At the canopy level, leaf-level values must be replacedwith canopy-
level values (subscript C), and fb must be replaced with a canopy boundary layer resistance, fbc. In addition, the aerodynamic resistance
(fa) between the atmospheric values and those measured at some reference height above the canopy (subscript R) must be included.
equal to the difference in concentration (potential A popular approach to calculating gc from gener-
difference) between the leaf or canopy and the ally available canopy-level meteorological measure-
atmosphere (i.e., V), divided by the resistance to this ments is by solving the Penman-Monteith combi-
transfer (i.e., Y): nation equation for gc:
Flux =
Potential difference [31 YC = - =
1 ra[S(Rn- G) - %E(S- r)] PC,D + 141
Resistance gc yl,E
LAND-ATMOSPHERE INTERACTIONS/ Canopy Processes 1125
where YC and Y, are the canopy and aerodynamic example PAR, temperature (leaf or air), humidity
resistances, respectively, S is the slope of the saturation (relative or the saturation deficit), and the C 0 2
vapor pressure versus temperature curve, R, the net concentration, [CO,] (leaf or air). The response of
radiation, G and AE the soil and latent heat fluxes, individual species to these or other variables varies
respectively, y the psychrometric constant, p the air greatly, but examples of the general responses in the
density, cp the specific heat of dry air, and D the absence of any other environmental stresses are shown
saturation deficit. This energy balance approach in Figure 4.
works well in an analytical sense when LE has already The approach taken by researchers to develop
been measured for example, by eddy covariance species-specific relationships has been either to make
(Figure 3), and when soil water evaporation is negli- measurements on individual leaves in a laboratory
gible. When measurements of 3,E are not available, the under controlled conditions, or to make measure-
use of eqn [4]in a predictive fashion requires that gc be ments on individual leaves in the natural environment
modeled, not an easy task given the complexity of a either under controlled or a wide variety of conditions.
plant’s response to the environment and vice versa. Individual leaf measurements are obtained by placing
At both the leaf and canopy level, conductance has a leaf into a cuvette (Figure S), and then either using
been found to respond to several variables, for multiple regression (an additive model; eqn [5],where
a through d are regression coefficients, and XI through
x3 are the various independent variables influencing
gL) or a boundary line analysis to determine the
stomatal response to the measured environmental
variables. The latter analysis involves measuring gL
under as many ambient conditions as possible (thus
requires large amounts of data), then fitting and
defining curves to the upper data points (e.g., f ( x l ) ) ,
where points below these curves represent times when
gL was limited by some other variable. The data are
then standardized by the maximum observed gL
(gLmnx), to form a multiplicative model (eqn [6]).
Once the relationships between gL and the environ-
ment are quantified, scaling from leaf to canopy is then
accomplished with a knowledge of leaf area index (eqn
[ 7 ] )If. a single-species canopy has several layers, then
gL and L should be measured for each individual layer,
multiplied, and then summed. If there are several
species in the canopy, then gL and L for each species
should be measured, multiplied, and then summed to
scale up to the canopy.
gL = u + bxl+ C X +~ dx3.. . 151
gc = gLL 171
It has also been recognized, however, in addition to
g L responding to relative humidity at the leaf surface
Figure 3 Eddy covariance instruments used to directly measure
the fluxes of heat, water vapor, C o n , and momentum. A sonic
(hs),gL in some species also responds to the [CO,] by
anemometer measures the vertical, horizontal, and lateral wind varying gL to maintain a constant [CO,] at the leaf
velocity components by measuring the speed of sound between surface (cs). These two driving variables, hs and cs, are
pairs of transducers. An open-path gas analyzer measures the combined in an empirical model widely recognized as
concentration of water vapor and Conby measuring the attenu- the Collatz model, which is based on the Ball-Berry-
ation of infrared radiation between a source and detector. The
correlation or covariance between fluctuations in simultaneous
Woodrow index:
high-frequency measurements of the vertical wind speed, and air
A
temperature, water vapor, COP or the horizontal wind speed are
used to calculate each of the fluxes.
gL = m-hs
CS +b 181
Next Page
u
a Photosynthetically active radiation --+ Leaf temperature -
Saturation deficit - Atmosphere COPconcentration -

Figure 4 Examples of the response of leaf stomatal or canopy conductance to photosynthetically active radiation (A), leaf temperature
(E), saturation deficit (C), and atmospheric COPconcentrations (D) when all other environmental stresses are absent.
or, at the canopy scale: reflected light created as light passes through the upper
canopy.
[91 To quantify the fraction of the top of the canopy
(incident)radiation ( f )that penetrates the canopy to a
where m and b are empirically derived coefficients depth z, a version of Bouger’s or Beer’s law is often
based on cuvette gas exchange studies, A is the net used:
carbon assimilation rate, and the subscript C refers to
canopy-level values. In eqns [SI and [9], conductance is
now a function of the net assimilation rate and vice
versa; hence eqns [SI or [9] must be solved iteratively where K is the extinction coefficient and LT the
using a series of equations that describes both cumulative leaf area index between the top of the
conductance and photosynthesis. canopy and z. The fraction of the incident beam
radiation intercepted by the canopy is 1 - fi. The
extinction coefficient K varies with the solar zenith
angle (thus K varies both diurnally and seasonally), the
optical properties of the leaves, and the canopy’s leaf
Light Penetration
angle distribution, and thus can vary largely with the
A canopy develops because of competition for sun- canopy architecture. Equation [lo] can also be used to
light, required for photosynthesis and hence growth. estimate the attenuation of other streams of radiation
Trees, through their intricate structure of leaves with canopy depth (e.g., net all-wave radiation or
supported by stems and branches, are remarkably PAR), as long as the appropriate radiation-specific K is
well adapted for light interception, and hence outgrow used.
nonstemmed species. Whereas an individual leaf When L is measured either optically with commer-
typically absorbs roughly 50% of the incident short- cially available instruments, or by leaf harvesting or
wave radiation, a canopy typically absorbs roughly litter collection, eqn [lo] allows for profiles of light
80%. The canopy’s advantage stems from the absorp- penetration to be calculated (Figure 6 ) . This in turn
tion by leaves lower in the canopy of the scattered and allows canopy processes driven by light absorption to
1136 LANDSEA BREEZE
determined by the relative proportion of wet and dry Mixing ratio of gas s (dimensionless)
areas. (volume fraction)
Transfer resistance for pathway n (s m - ')
Transfer conductance (m s- ')
Model Parameterizationfor Mapping Deposition velocity ( = fludconcentration)
Trace Gas Fluxes , ( ms - l )
One of the main objectives of measuring trace gas Height above surface (m)
exchange is to be able to estimate fluxes at the regional, Displacement height of canopy (m)
national, or global scale. The complexity of the Roughness length of surface (m)
processes involved, for all but a few gases, means Air density (g m - 3 ,
that simplifications must be introduced into models Eddy diffusivity of gas s (m2s -
designed to estimate large-scale fluxes and budgets. Vertical wind speed (m s -
Initially, the modeling of surface exchange as part of
schemes to represent the long-range transport and fate
of trace gases used single parameters for the deposition See also
velocity of each gas. More complex models might vary
the deposition velocity by night and day. As under- Biogeochemical Cycles: Nitrogen Cycle: Sulfur Cycle.
standing improved, models were modified to calculate Global Change: Biospheric Impacts and Feedbacks.
explicitly the atmospheric resistance terms, as a Land-Atmosphere Interactions: Canopy Processes:
Overview. Methane. Ozone: Surface Ozone Effects on
function of mean wind speeds and vegetation type,
Vegetation.
using climatological maps of wind speed and vegeta-
tion classifications based on survey or remote sensing.
The most recent models include factors to account for
surface wetness, and may model stomatal opening Further Reading
from parameterizations based on a range of different
Bouwman AF (ed.) (1999) Approaches to Scaling of Trace
vegetation types and responses, with explicit depend-
Gas Fluxes in Ecosystems, Developments in Atmospheric
ence on temperature and light levels. The introduction Science 24. Amsterdam: Elsevier.
of detailed surface chemical processes (e.g., the Jarvis PG and Fowler D (2000).Forests and the atmosphere.
dependence of SO2 deposition rates on the availability In: Evans J (ed.) The Forests Handbook. Oxford: Black-
of NH3) is still beyond the scope of regional and well Scientific.
national modeling, not least because the required data Matson PA and Harriss RC (eds) (1995) Biogenic Trace
on air concentrations are not available at the spatial Gases: Measuring Emissions from Soil and Water.
scales necessary. The evaluation of the uncertainties in Oxford: Blackwell.
such models, arising from the inevitable use of Monteith JL (ed.) (1975-1976) Vegetation and the Atmos-
spatially and temporally averaged data (for meteoro- phere, vols. 1,2. London: Academic Press.
logical as well as chemical parameters) is an active area Monteith JL and Unsworth M H (1990) Principles of
Environmental Physics, 2nd edn. London: Edward
of current research.
Arnold.
Oliver HR, Bell BG and Clymo RS (eds) (1998) Terrestrial
Nomenclature Initiative in Global Environmental Research - the
T I G E R Trace Gas Programme. Atmospheric Environ-
Fs Flux of trace gas s (gm-2s-1) ment Special Issue, vol. 32(19),pp. 3205-3370. Oxford:
xs Air concentration of gas s (gm - 3 , Pergamon.
R A Pielke Sr, Colorado State University, Fort Collins, theoretically. This is undoubtedly a result of the
CO, USA geographically fixed nature of the phenomenon (the
Copyright 2003 Elsevier Science Ltd. All Rights Resewed. location of land-water boundaries), as well as the
repetitive nature of the event. The sea breeze is defined
to occur when the wind is onshore, i.e. blowing
towards the land, whereas the land breeze occurs when
Of all the mesoscale phenomena, sea and land breezes the wind is offshore and the opposite flow exists.
have been the most studied, both observationally and Detailed discussion of sea and land breezes is given in
LAND-SEA BREEZE 1137
p3
p2
Pl
Po6.00a.m.
p3
p2
Pl
Po - 6.00 p m
Radiational cooling becomes dominant
over solar heating; sea breeze winds
remove pressure gradient
PQ p3
p; p2
Pl
Po 9.00 a.m. P
O
p1 9.00 p.m.
Mass mixed upwards Sinking as air cools by radiative
flux divergence; downward mass flux
-
p3
p2
'
PO Midnight
p3
p2
Pl
Po
--
7 +
c
3.00 p.m.
Inland penetration of the sea breeze;
penetration distance controlled by
latitude ( f )
p3
p2
Pl
Po - ~~~~*+i
3.00 a.m.
Shallower land breeze-more stable
at night
Figure 1 Schematic of the diurnal evolution of the sea and land breeze in the absence of synoptic flow. (From Pielke, 1984.)
Simpson (1994), with a briefer discussion in Pielke

(1984, 2002) and Atkinson (1981). Sea and land
breezes which occur associated with larger lakes are
called lake and land breezes (e.g., Neumann and
Mahrer 1975). The leading edge of the sea breeze
winds is called the sea breeze front.
During the case of calm large-scale winds and in
flat terrain, it is comparatively easy to describe
the diurnal variations of the coastal wind circulations.
Defant (1951) presented an excellent quali-
tative description for this condition, which is illustrat-
ed in Figure 1.The idealized sequence of events is as
follows:
1. At some time in the early morning the pres-

sure surfaces become flat and no winds
occur (e.g., at 0600 LST - perhaps an hour after
sunrise).
2. Later in the morning, mass is mixed upward over
land by turbulent mixing in the unstably stratified
boundary layer and also because of the expansion
of the volume of air due to its heating creating, an
offshore pressure gradient at some distance above
the ground (Tijm and von Delden 1999; Nicholls
and Pielke 1994). Over water, the penetration of
sunlight and resultant distribution of radiative
Figure 2 Schematic of the influence of coastline configuration
heating with depth and the ability of water on the sea breeze in the absence of large-scale flow. (From
to mix minimizes significant heating of 'the Pielke RA (1984) Mesoscale Meteorological Modeling: Academic
surface (e.g., at 0900 LST). The temperature of Press.)
1138 LANDSEA BREEZE
the water is not important in determining the

strength of the sea breeze, as long as the air above is
warmer than the water.
3. The resultant offshore movement of air above the
ground near the coast creates a low-pressure region
at the ground, and onshore winds (the sea breeze)
develop (e.g., at 1200 LST).
4. The onshore winds transport cooler marine air over
the land, thereby advecting the horizontal temper-
ature gradient and, hence, the sea breeze inland.
The distance the sea breeze travels inland depends
most directly on the intensity of the total heat input
to the air (Pearson 1973; Tijm et al. 1999;
. # i Neumann 1977) and the latitude (Rotunno 1983)
(e.g., at 1500 LST).
5. As the sun sets, longwave radiational cooling
Figure 3 Radar echo coverage at 1501 EST on 19 August 1971 as becomes dominant over solar heating, and the local
seen by the Miami WSR-57 10 cm radar. The sites with weather ob-
servations are noted by using the standard airport identifiers. (From
wind field removes the horizontal temperature
Pielke RA (1974) A three-dimensional numerical model of the sea gradient. The pressure surfaces again become hor-
breezes over south Florida. Monthly Weather Review1 02: 1 15-1 39.) izontal (e.g., at 1800 LST).
HORIZONTAL VELOCITY HOUR = 3.01 HORIZONTAL VELOCITY HOUR = 5.00

USYNOP = 6.0M/SEC ANGLE = 135
HORIZONTAL VELOCITY HOUR 8.00 HORIZONTAL VELOCITY HOUR = 10.01
Figure 4 Horizontal wind at 50m level, 3, 5, 8, and 10 hours after simulated sunrise for a uniform synoptic south-east wind case over
south Florida. Note how the wind convergence pattern in Figure 3 closely corresponds to the wind convergence pattern in Figured. (From
Pielke RA (1 974) A three-dimensional numerical model of the sea breezes over south Florida. Monthly Weather Review 102: 115-139.)
LAND-SEA BREEZE 1139
HORIZONTAL VELOCITY HOUR = 3.00 HORIZONTAL VELOCITY HOUR = 5 01

USYNOP = 6.0 M/SEC ANGLE = 225. USYNOP I6.0 MiSEC ANGLE = 225.
I FVFl = 0 0SKM LEVEL = 0.05KM
HORIZONTAL VELOCITY HOUR = 8.00

USYNOP = 6.0 M/SEC ANGLE = 225
LEVEL = 0.05KM
Figure5 As Figure 4, except for a uniform synoptic south-west wind. (From Pielke RA (1 974) A three-dimensional numerical model of
the sea breezes over south Florida. Monthly Weather Review 102: 115-139.)
6 . As longwave cooling continues and compresses, the op, as illustrated for the daytime portion of the cycle in
air near the ground becomes denser and sinks. The Figure 2. (Such zones of preferential convergence help
resultant lowering of the pressure surfaces a short explain the preference for showers and thunderstorms
distance above the ground creates an onshore wind in certain locations in south Florida during the
at that level (e.g., at 2100 LST). summer, as seen, for example, in Figure 3 and
7. In response to the loss of mass above the surface discussed in Pielke et al. 1991.)
over the water a pressure minimum develops at the The evolution of the sea breeze is somewhat more
ocean interface immediately off the coast. The complicated when a weak or moderate (i.e., 6 m s - ’)
offshore wind that then develops near the surface is prevailing synoptic flow is included. For the two
called the land breeze (e.g., at 2400 LST). distinct situations of comparatively cold water and
8. The distance of offshore penetration of the land comparatively warm water relative to land, a synoptic
breeze depends on the amount of cooling over the wind direction from the colder to the warmer surface
land. Because the planetary boundary layer over weakens the intensity of the local wind by diminishing
land is stably stratified at night and, therefore, the horizontal temperature gradient. By contrast,
vertical mixing is weaker and closer to the ground, when a prevailing larger-scale flow of the same
the land breeze is a shallower and weaker phenom- strength is from the warmer to the colder surface, if
enon than the daytime sea breeze (0300 LST). the synoptic wind speed is not too strong, the
temperature gradient is strengthened and the subse-
There may even be a higher, third layer of flow quent local wind flow is stronger. An example of this
associated with these local winds, which Tijm et al. effect is shown in Figures 4 and 5 where the sea breeze
(1999b) refer to as a ‘return-return current’. wind convergence is more clearly evident when the
When the coastline is irregular, local regions of large-scale wind is in the opposing direction from the
enhanced or weakened low-level convergence devel- sea-breeze-altered flow.
1140 LARGE EDDY SIMULATION
Examples of water that is warm relative t o the land See also

include the eastern sides of continents in the tropics Coastal Meteorology. Mesoscale Meteorology:
and midlatitudes at night and over coastal waters Mesoscale Convective Systems; Models; Overview.
during a polar outbreak. Situations with water that is
cold relative t o the adjacent land occur on the eastern
sides of continents in tropical and midlatitudes during Further Reading
sunny days, and also along the western sides of Atkinson BW (1981)Mesoscale Atmospheric Circulations.
continents in which upwelling is occurring, as well as Academic Press.
along polar coastal areas in the summer. Fog and low Defant F (1951)Local winds. Compendium of Meteorology,
stratus often form over the relatively cold water in American MeteorologicalSociety, Boston, MA, 655-672.
polar and upwelling ocean areas and move onshore Neumann J (1977)On the rotation rate of the direction of sea
with the sea breeze. and land breezes. Journal of the Atmospheric Sciences 34:
The magnitude of the effect of a particular horizon- 1913-1917.
tal temperature gradient can be estimated from exist- Neumann J and Mahrer Y (1975)A theoretical study of the
ing observational and numerical studies. It has been lake and land breezes of circular lakes. Monthly Weather
Review 130: 474-485.
found that, in the tropics and midlatitudes, a horizon-
Nicholls ME and Pielke RA (1994) Thermal compression
tal gradient of less than about 1 0 W m P 2per 30 km has
waves. 11. Mass adjustment and vertical transfer of total
only a minor influence on local wind patterns. With a energy. Quarterly Journal of the Royal Meteorological
gradient of 100 W m P 2 per 30 km, however, signifi- Society 120: 333-359.
cant effects are discernible from the statistical evalu- Pearson RA (1973) Properties of the sea breeze front as
ation of observational data, whereas at 1 0 0 0 W m - 2 shown by a numerical model. Journal of Atmospheric
per 30 km, the influence on local wind patterns is very Science 30: 1050-1060.
pronounced in case-by-case studies. With a nonzero Pielke RA (1974) A three-dimensional numerical model of
large-scale wind, the heating must be greater in order the sea breezes over south Florida. Monthly Weather
for a sea breeze t o develop. Using observational data it Review 102: 115-139.
has been shown that a sea breeze does not develop Pielke RA (1984,2nd edn. 2002) Mesoscale Meteorological
Modeling. Academic Press.
when the horizontal pressure gradient generated by
Pielke RA, Song A, Michaels PJ, Lyons WA and Arritt RW
the differential heating between land and adjacent
(1991) The predictability of sea breeze generated thun-
water is insufficient t o overcome the kinetic energy of derstorms. Atmosfera 4: 65-78.
the large-scale flow. Rotunno R (1983)On the linear theory of the land- and sea-
When the coastal terrain is hilly or mountainous, breeze. Journal of the Atmospheric Sciences 40: 1999-2005.
sea and land breezes interact with local winds that Segal M, Mahrer Y and Pielke RA (1983) A study of
are created as a result of the heating and cooling of meteorological patterns associated with a lake confined
this elevated terrain relative t o the adjacent atmos- by mountains -the Dead Sea case. Quarterly Journal of
phere at the same altitude. The sea breeze and upslope the Royal Meteorological Society 109: 549-564.
mountain flow that are created as the terrain is Simpson JE (1994) Sea Breeze and Local Wind. New York:
heated during the day, for example, can generate Cambridge University Press.
particularly strong onshore winds. However, Tijm ABC and vonDelden AJ (1999)The role of the sound
waves in sea-breeze initiation. Quarterly Journal of the
the subsidence in the adjacent atmosphere caused
by the up-slope flow can inhibit the development of Tijm ABC, AAM Holtslag and AJ Van Delden (1999)
the sea breeze, resulting in an onshore wind that is Observations and modeling of the sea breeze with the
less then the sum of the two winds. In addition, return current. Monthly Weather Review 27: 625-640.
the intensity of combined local wind circulation Tijm ABC, vanDelden AJ and Holtslag AAM (1999) The
tends to be less when the terrain slope is larger inland penetration of sea breezes. Contributions to
(Segal et al. 1983). Atmospheric Physics 72: 317-328.
C-H Moeng and P Sullivan, National Center Introduction

for Atmospheric Research, Boulder, CO,
USA Turbulence consists of three-dimensional, chaotic, or
Copyright 2003 Elsevier Science Ltd. All Rights Resewed. random motion that spans a range of scales that
Examples of water that is warm relative t o the land See also

include the eastern sides of continents in the tropics Coastal Meteorology. Mesoscale Meteorology:
and midlatitudes at night and over coastal waters Mesoscale Convective Systems; Models; Overview.
during a polar outbreak. Situations with water that is
cold relative t o the adjacent land occur on the eastern
sides of continents in tropical and midlatitudes during Further Reading
sunny days, and also along the western sides of Atkinson BW (1981)Mesoscale Atmospheric Circulations.
continents in which upwelling is occurring, as well as Academic Press.
along polar coastal areas in the summer. Fog and low Defant F (1951)Local winds. Compendium of Meteorology,
stratus often form over the relatively cold water in American MeteorologicalSociety, Boston, MA, 655-672.
polar and upwelling ocean areas and move onshore Neumann J (1977)On the rotation rate of the direction of sea
with the sea breeze. and land breezes. Journal of the Atmospheric Sciences 34:
The magnitude of the effect of a particular horizon- 1913-1917.
tal temperature gradient can be estimated from exist- Neumann J and Mahrer Y (1975)A theoretical study of the
ing observational and numerical studies. It has been lake and land breezes of circular lakes. Monthly Weather
Review 130: 474-485.
found that, in the tropics and midlatitudes, a horizon-
Nicholls ME and Pielke RA (1994) Thermal compression
tal gradient of less than about 1 0 W m P 2per 30 km has
waves. 11. Mass adjustment and vertical transfer of total
only a minor influence on local wind patterns. With a energy. Quarterly Journal of the Royal Meteorological
gradient of 100 W m P 2 per 30 km, however, signifi- Society 120: 333-359.
cant effects are discernible from the statistical evalu- Pearson RA (1973) Properties of the sea breeze front as
ation of observational data, whereas at 1 0 0 0 W m - 2 shown by a numerical model. Journal of Atmospheric
per 30 km, the influence on local wind patterns is very Science 30: 1050-1060.
pronounced in case-by-case studies. With a nonzero Pielke RA (1974) A three-dimensional numerical model of
large-scale wind, the heating must be greater in order the sea breezes over south Florida. Monthly Weather
for a sea breeze t o develop. Using observational data it Review 102: 115-139.
has been shown that a sea breeze does not develop Pielke RA (1984,2nd edn. 2002) Mesoscale Meteorological
Modeling. Academic Press.
when the horizontal pressure gradient generated by
Pielke RA, Song A, Michaels PJ, Lyons WA and Arritt RW
the differential heating between land and adjacent
(1991) The predictability of sea breeze generated thun-
water is insufficient t o overcome the kinetic energy of derstorms. Atmosfera 4: 65-78.
the large-scale flow. Rotunno R (1983)On the linear theory of the land- and sea-
When the coastal terrain is hilly or mountainous, breeze. Journal of the Atmospheric Sciences 40: 1999-2005.
sea and land breezes interact with local winds that Segal M, Mahrer Y and Pielke RA (1983) A study of
are created as a result of the heating and cooling of meteorological patterns associated with a lake confined
this elevated terrain relative t o the adjacent atmos- by mountains -the Dead Sea case. Quarterly Journal of
phere at the same altitude. The sea breeze and upslope the Royal Meteorological Society 109: 549-564.
mountain flow that are created as the terrain is Simpson JE (1994) Sea Breeze and Local Wind. New York:
heated during the day, for example, can generate Cambridge University Press.
particularly strong onshore winds. However, Tijm ABC and vonDelden AJ (1999)The role of the sound
waves in sea-breeze initiation. Quarterly Journal of the
the subsidence in the adjacent atmosphere caused
by the up-slope flow can inhibit the development of Tijm ABC, AAM Holtslag and AJ Van Delden (1999)
the sea breeze, resulting in an onshore wind that is Observations and modeling of the sea breeze with the
less then the sum of the two winds. In addition, return current. Monthly Weather Review 27: 625-640.
the intensity of combined local wind circulation Tijm ABC, vanDelden AJ and Holtslag AAM (1999) The
tends to be less when the terrain slope is larger inland penetration of sea breezes. Contributions to
(Segal et al. 1983). Atmospheric Physics 72: 317-328.
C-H Moeng and P Sullivan, National Center Introduction

for Atmospheric Research, Boulder, CO,
USA Turbulence consists of three-dimensional, chaotic, or
Copyright 2003 Elsevier Science Ltd. All Rights Resewed. random motion that spans a range of scales that
LARGE EDDY SIMULATION 1141
increases rapidly with Reynolds number (see Turbu- fluid dynamics community, but in this article we
lence and Mixing). The complete numerical integra- restrict our attention to LES of PBL flows. Most LES
tion of the exact equations governing the turbulent research in the PBL community focuses on applica-
velocity field (the Navier-Stokes equations) is known tions of the technique in studying various PBL regimes
as direct numerical simulation (DNS). Because of including the effects of convection, rotation, rough
today’s limited computing power or resolution, DNS is surfaces, entrainment, radiation, and/or condensa-
restricted to low-Reynolds-number turbulence, which tion; so far only a small fraction of the total effort is
exists in laboratory flows, e.g., those in wind tunnels. towards the development of improved SGS models or
At present, the largest DNS utilizes l o 9 gridpoints, numerical techniques.
which is still insufficient to simulate high-Reynolds-
number flows like geophysical turbulence. As an
illustration of the computer memory required for The LES Technique
DNS, consider the atmospheric planetary boundary Governing Equations and Filtering Procedures
layer (PBL; see Boundary Layers: Neutrally Stratified
Boundary Layer) where the largest turbulent eddies The Navier-Stokes equations for an incompressible
are on the order of kilometers and the smallest on the fluid
order of millimeters; thus, the entire spectrum of
turbulent motion spans more than six orders of
magnitude. To numerically integrate the Navier-
Stokes equations for this turbulent flow would require
at least 10l8 numerical gridpoints; this is far beyond form the basis for an LES of the PBL, where ui satisfy
today’s computing capacity or that in the foreseeable the continuity equation:
future.
When the scale range exceeds that allowed by
computer capability, which is typical in flows of
meteorological interest, only a portion of the scale In eqns [I] and [2], u, are flow velocities in the three
range can be explicitly resolved, presumably the larger spatial directions (i.e., i = 1 and 2 for the horizontal
eddies or most important scales of the flow; the directions and i = 3 for the vertical direction), X I are
remaining scales must be roughly represented or the ith-component of body forces, p is the air density, p
parameterized in terms of the resolved portion. This is the pressure fluctuation, v is the kinematic viscosity
partly resolved simulation is the philosophy behind of the fluid, tis time, and x , are the spatial coordinates.
large eddy simulation (LES). For PBL turbulence, the For PBL applications, the major body forces are
large eddies (LE)contain most of the turbulent kinetic gravity and Coriolis forces and hence X , can be
energy (TKE),and hence are called energy-containing approximated as g$/TO - fc,,3u,, where the gravita-
eddies; they are responsible for most of the turbulent tional acceleration g, is nonzero only in the x3 (or z )
transport. A simulation that explicitly calculates these direction, 0 is the virtual potential temperature,’ TOis
large eddies while approximately representing the the temperature of some reference state, and f is the
effects of smaller ones is potentially a good surrogate Coriolis parameter. This X , is obtained by expanding
for the entire flow field. As defined, the accuracy of eqn [I] over a reference state of hydrostatic equilib-
LES increases as the grid resolution becomes finer and rium and also using the Boussinesq approximation
finer. LES is a compromise between DNS, in which all (see Dynamic Meteorology: Primitive Equations). The
turbulent fluctuations are resolved, and the tradition- numerical integration of eqns [ l ] and [2] is DNS; for
al, Reynolds-averaging approach, in which all fluctu- LES eqns [l]and [2] need to be spatially filtered.
ations are parameterized and only ensemble-averaged The volume-filtered Navier-Stokes equations are
statistics are calculated. derived by first decomposing all dependent variables,
The first true LES calculation was performed by Jim e.g., u,, into a volume average, Ut,and a subgrid-scale
Deardorff at NCARin the late 1960s using a computer (SGS) (or subfilter) component, uy, i.e., u, = U, u;. +
that allowed for 32 x 32 x 32 (32 768) gridpoints. Here the volume-averaged or resolved-scale variable is
On today’s computing machines, calculations with defined as
N 106-107 gridpoints are common practice and l o 8
computations are possible on massively parallel

machines. As computer power increases, we anticipate
a much broader application of LES to more compli-
cated geophysical turbulence problems. LES has been ‘An additional transport equation is required for 0 if buoyancy is
studied and used intensively also in the engineering considered.
2b
4 their counterparts in eqn [l], except that they depend
on filtered (resolved-scale) fields. The nonlinear re-
solved advective terms embody vortex stretching and
generate an energy cascade, which are crucial dynam-
ics in 3D turbuleEe. ThLSGS g e s s (or flux) tensor is
defined as zij = Uiuy + uyiij + uyuy. Alternatively, eqn
0 [4] can be written as
aiii - aiiiiii gi Q- - f ~ i j 3 i i j
azii + -
at axi axi T~
1 ap + v - a2iii
PI
po ax, ax;
where now t i j = uiij - ii,iij.
Equations [4] or [5] equally describes the evolution
of the LE field; they differ in their forms of the resolved
advection and SGS terms. In eqn [4], the SGS term
Figure 1 A sketch illustrating spatial fluctuations of (A) a total consists of two kinds of influences: cross-products of
turbulent field (solid curve)and its filtered field (dashedcurve), and
resolved-SGS components (i.e., iiiuy + ut!iij) and a
N N
(B) its subgrid-scalefield.

n s l i n e a r product of SGS-SGS components (i.e.,
ui!ur). The SGS term in eqn [5] includes all of these
where G is a three-dimensional (low-pass) filter influences plus a rzolved scale contribution or Leo-
function, e.g., Gaussian, top-hat or sharp wave cutoff nard term,3 Lij = iiiiij - iiiiij. Thus in principle differ-
filter. A one-dimensional random signal can be used to ent SGS models should be used in eqns [4] and [5]. In a
illustrate the details of the filtering process. In Figure spectral-based LES code, which applies filter explicit-
lA, the solid curve denotes the total signal (fluctuating ly, the advection term in eqn [4] can be calculated
in x ) . Application of the filter operator in eqn [ 3 ]to the directly. A finite-differencing based LES code can only
total signal yields a smoother field (indicated by the solve eqn [5]. For geophysical turbulence, the molec-
dashed curve), which is the so-called filtered field or ular viscosity term is negligibly small compared with
resolved-scale motion. The difference between total the advection terms in eqn [4]or eqn [5], and hence can
and resolved signals (Figure 1B) represents the SGS be neglected.
fluctuations. The partitioning between resolved and So far in deriving the above LE equations, [4] or
SGS components depends on the filter; i.e., cutoff scale [5], no approximations have been made. Because of
and sharpness. the spatial filtering procedure, the LE equations
Applying the filtering procedure, term-by-term, to contain SGS terms that are unknown and must be
eqn [ l ] leads to equations that govern large (resolved- modeled in terms of the resolved fields. Because the
scale) eddies magnitudes of SGS terms depend on the filter, its
modeling in principle should depend on the filter size
and shape.
Subgrid-Scale Parameterization
141
To solve eqn [4] or eqn [5], the SGS terms need to be
parameterized, which results in some uncertainty in
where the decomposition ui = iii + ui' is used along LES. This uncertainty contaminates the LES solution,
with the assumption that the filtering operation is particularly in regions where small eddies dominate,
commutative with differentiation.2 In eqn [4], the first i.e., near a wall boundary and perhaps in the entrain-
term on the right-hand side is advection of H i by the ment zone of the PBL. However, in regions where
resolved-scale motion i i j , the second term is the SGS energy-containing eddies are well resolved, LES flow
contribution, and the remaining terms are identical to fields have been shown to be rather insensitive to SGS
'Though this assumption is strictly valid only for unbounded 3Note that the conversion between eqn [4] a k e q n [5]c a n b e
flows, one can however show that the error introduced is of the same o b g n e d &sing the identity - ii,iil = (&iil - iitiil)+ ii,u:
UT
order as the truncation error of the numerical method. +U:'iiI + u:'u;.
models. This results from the fact that, in the interior and
of PBL, the SGS motions serve mainly as net energy
sinks that drain energy from the resolved motions, a +
KH = [1 ( ~ ~ / A s ) ] K M [I21
process that can be modeled adequately with a simple where CK is a diffusion coefficient to be determined and
SGS parameterization. C is another SGS length scale, which is often taken as
Most PBL-LESS adopt a similar scheme for the SGS the minimum of two length scales
terms, regardless of whether eqn [4] or eqn [5] is used
or whatever filter function is applied. The Smagorin-
sky-Lilly (S-L) model is the most widely used SGS
closure scheme and relates SGS stresses to resolved-
scale strain tensors by
Equation [13] assumes a direct effect of local stability
Zjj = - 2 K ~ S i j [61 on the local SGS length scale. The SGS TKE e evolves
from the following equation:
+
where the strain tensor Sjj = (ai&/axj aiij/axi)/2. - -
SGS heat fluxes are similarly related to local gradients ae - aiije a(uj’e+uj’~’’) x, ahj
- uiu . -
in the resolved temperature field by the relation at axi axj I axj
171
The terms on the right-hand side of eqn [14] re-
Without buoyancy effects the SGS eddy viscosity K M
present, in order, advection of e by the resolved-
and diffusivity KH are expressed as
scale motion, turbulent and pressure transports,
KM = (CSAS)2 S 181 local shear production (nonlinear scrambling), local
buoyancy production, a n d molecuhr dissipa-
KM tion. In solving eqn [14], uyu/ and w’/# are app-
K H =- 191 roximated by eqns [6] and [7],respectively, the
Pr
transport terms by
where the Smagorinsky constant cs remains to be
determined, As is a filtered length scale often taken to
be proportional to the grid size, the magnitude of the
strain tensor, S, is (2SijSij)1’2,and Pr(- 3) is the SGS
Prandtl number. One of the most important features of and the molecular dissipation rate by
the S-L model is that the SGS fluxes are nonlinear cEe3f2
functions of the resolved strain rate, a crucial differ- E=-----
ence from the viscous (molecular) stress-strain rela- !
tionship. where cEis a dissipation coefficient.
To include local buoyancy effects, the KM expres- The SGS model parameters cs, CK, and cEare chosen,
sion in the original Smagorinsky model is modified to in most of the existing PBL-LESS, to be consistent
depend on local Richardson number Ri (the ratio of with Kolmogorov inertial-subrange theory, i.e.,
buoyancy to shear production terms of TKE the constants are derived assuming that the SGS
budget), motions are isotropic with a k-5/3 spectral slope.4
-
Commonly used values are: cs 0.18, CK N 0.10, and
N
c, 0.19 +0.74k/As. With these model parameters,

LESSare in a way forced - in an ensemble-mean sense -
to drain energy at a rate sufficient to produce a k - s / 3
where Ri, is the critical Richardson number often set spectral slope near the filter cutoff scale.
between 0.2-0.4, and n = is often used. When the The above SGS models are based on ensemble
local Richardson number reaches the critical value, average concepts but are used inside LES on an
turbulence within that grid cell vanishes and the eddy instantaneous basis, i.e., to represent SGS effects at
viscosity is shut off. every gridpoint and time step. Laboratory studies and
A further extension of the S-L eddy viscosity model DNS provide evidence that small-scale turbulent
explicitly calculates the SGS-TKE e and relates K M motions are anisotropic and intermittent, and that
and KH to e via
41n the inertial subrange the equilibrium 3D energy spectrum
KM = CK~& 1111 varies as 4 ( k ) M C k - S / 3where k is the wavenumber.
100x100 gridpoints would cover 3 to 5 large domi-

locally the energy transfer can either be forwardscatter
(from large to small scales) or backscatter (from smallnant eddies in each horizontal direction and at the
to large scales), which causes deviations from the same time resolve small eddies down to about
equilibrium kP5l3law. Eddy viscosity SGS models also 100 m x 100 m x 40 m in size, assuming model resolu-
assume that SGS stresses and strains are perfectly tion is twice the grid size. For the stable PBL where
aligned (see eqn [6]), and hence the local dissipation dominant eddies are smaller, a smaller domain (and
rate E = -z&j is always positive, thus preventing consequently a finer grid) is preferred.
backscatter of energy. Numerical truncation errors and specification
These deficiencies of eddy viscosity models have of boundary conditions add uncertainties to all
motivated continued development of new SGS mod- numerical models including LES. Most PBL-LES
els, including (1)stochastic models where a random codes use finite difference methods in all three direc-
field is imposed at the SGS level, thus permitting tions to compute derivatives, although some
a backscatter of energy, ( 2 ) dynamic models where LESs employ a spectral (Fourier) representation in
the Smagorinsky coefficient is dynamically predicted x-y planes taking advantage of the horizontally
homogeneous nature of the PBL. Sharp gradients in
using a resolved field filtered at two different scales,
and ( 3 ) velocity estimation models that attempt to flow variables can exist at the top of the PBL because of
model the SGS velocity fluctuations uy instead of SGS the presence of a strong, stably stratified overlying
stresses zjj. layer, which leads to oscillations (dispersion errors)
The deficiency of S-L SGS models is most evident in when finite differencing methods are used. To over-
the surface layer of the PBL, where the vertical come this flaw, sign-preserving (monotone) schemes
gradients of the mean fields are known from a large are frequently used for scalar transport to maintain
body of measurements and scaling arguments to physical realizability,' often at the expense of intro-
follow Monin-Obukhov (M-0) similarity theory5 ducing more numerical diffusion (see Numerical
(see Boundary Layers: Surface Layer). However, the Models: Methods).
M-0 vertical distribution of mean velocity and The surface boundary condition in LES borrows
temperature fields cannot be reproduced by LESs heavily from a wealth of field observations. Note
using standard S-L eddy viscosity prescriptions, par- that for PBL applications, LES cannot possibly resolve
ticularly for shear-driven and stable PBLs. One reason the viscous layer close to the surface; its lowest
for this shortcoming is that near the surface the SGS grid level lies in the inertial sublayer (i.e., the surface
motions dominate the whole flow field. In other layer). Hence, as a surface boundary condition M-0
words, very near the surface, PBL-LES is no longer similarity theory is used to relate surface fluxes to
an LES; almost all of the turbulent eddies are subgrid-resolved-scale fields at each grid point just above
scale. This deficiency has been improved somewhat the surface. The primary empirical input parameter to
using SGS models that contain either backscatter or an these formulas is the surface roughness. This
explicit contribution from a Reynolds-averaged roughness length can vary from less than 0.001 m for
model. a smooth sea surface to more than O.lm for
heavily wooded terrain. This rough-wall boundary
condition is different from the smooth-wall condition
Numerical Setup, Methods, and Boundary used in engineering flows. Caution should be used,
Conditions however, because M-0 theory describes ensemble-
The choice of LES grid and domain sizes depend on the mean flux-gradient relationships in the surface layer
physical flow of interest and the computer capability. (see Boundary Layers: Surface Layer) and may not
LES differs from other meteorological models in that apply well at the local LES grid scale. This problem
its grid-scale motion is nearly isotropic, and hence becomes more acute when the LES horizontal grid size
requires a grid mesh close to isotropic. Most comput- is comparable to or smaller than the height of the first
ers today can fit a code of about 1OOx1OOx100 grid level.
gridpoints. From these gridpoints, an LES domain is The upper boundary of a typical LES domain is
then chosen to resolve several largest (dominant) usually set to be well above the PBL top, in order to
turbulent eddies and at the same time resolve avoid influences on simulated PBL flows from artificial
eddies as small as possible into the inertial-subrange upper boundary conditions. At the top of the domain,
scales. For example, for a convective PBL with 1km turbulence is negligible and a no-stress condition is
depth, a 5 km x 5 km x 2 km domain of LES with 100 x applicable. Because turbulent motions in the PBL may
'In a shear-driven PBL where the buoyancy forcing is zero, M-0 6Physical realizability implies that the concentration of a scalar
theory reduces to the logarithmic law of the wall. can never be lower than the background concentration.
LARGE EDDY SIMULATION 1 145
Figure2 Contour plots of vertical velocity (top) and temperature (bottom) fluctuations in a vertical cross-section from an LES of the free
convective PBL: w contours (-2.5. -2. -1.5. -1. -0.6.0.6.1.1.5,2.2.5), dark (light) shadlng values larger (smaller) than 0.6 (-0.6); 8’
contours (-1. -0.7. -0.5. -0.35. -0.08,0.08.0.35.0.5.0.7.1 .O),dark (light) shading values larger (smaller) than 0.08 (-0.08).
excite gravity waves in the stably stratified inversion positively skewed vertical velocity field, a unique
layer, a means of handling gravity waves is often feature of convectively driven turbulent flows.
applied. Typically, a radiation condition, which allows The vertical velocity and temperature fields are corre-
for an upward escape of gravity waves, or a wave- lated strongly in the lower half of the PBL, which
absorbing sponge layer is used at the top of the results in a large positive heat flux there. Strong
simulation domain. updrafts penetrate into the capping inversion, which
For the lateral boundary conditions, almost all PBL can engulf wisps of warm inversion air into the PBL.
LESS today use periodic boundary condition^.^ These wisps of air may subsequently be entrained and
Periodic boundary conditions are appropriate mixed into the PBL. This penetration-lead-to-entrain-
for PBLs with homogeneous terrain but also are a ment phenomenon has also been documented with
computational convenience in that no explicit state- radar and sodar observations and convection tank
ment of the sidewall boundary (turbulence)conditions experiments (see Boundary Layers: Convective
needs to be imposed. The use of periodic boundary Boundary Layer).
conditions, however, prevents us from simulating Plan views of both temperature and vertical velocity
realistic meteorological flows with inhomogeneous fields given in Figure 3 reveal spokelike, irregular
surface. polygonal structures near the surface, similar to those
observed in Rayleigh-Btnard convection experiments.
LES Flow Fields This spokelike feature is most evident in the free
convective PBL, i.e., with a zero mean wind. Intersec-
The solution of eqn [4] or eqn [5] consists of tions between these near-surface polygons are local
three-dimensional, time-evolving flow fields of large horizontal convergence regions and hence are sites to
turbulent eddies. An example of such a flow field form strong updrafts which can penetrate into the
is shown in Figure 2, where a vertical cross-section of capping inversion. These coherent updrafts or
an LES calculation of a free convective PBL thermals, which are also well documented by obser-
using 96 x 96 x 96 gridpoints is presented. The figure vations, are believed to be responsible for significant
shows that updrafts, which are associated with turbulent transport in the PBL.
warm thermals, are more intense and occupy a For shear-driven PBLs, LES reveals elongated high-
narrower area than downdrafts. This is known as a low-speed streaks along the mean shear direction in
the surface layer (Figure 4),similar to those found in
’Periodicity implies that the inflow ar each gridpoint on a sidewall the wind tunnel and also observed in the near neutral
is equal to the outflow on opposite sidewall. PBL.
Figure 3 Contour plots of vertical velocity (top) and temperature (bottom) fluctuations in two horizonal cross-sections (left: near the
surface; right: nearthe PBLtop)fromanLESofthefreeconvective PBL: wcontoursare (-0.5,0.5) while8'contoursare (-0.12,0.12)for
z/zi = 0.16 and (-0.08,0.08) for z / z i = 0.84.
Statistics Derived from LES Flows One should be cautious about using statistics
constructed from LES flows, however. Some statistics,
Moment statistics can be calculated readily from 3D
especially higher moments, may be sensitive to the LES
LES data volumes by correlating the fluctuations
grid resolution, domain size, and SGS models. A
among variables. Then the vertical profiles or distri-
necessary but not sufficient rule of thumb is to rely only
butions of these statistics can be systematically doc-
on the statistics that are insensitive to the LES grid
umented for various PBL regimes generated under
resolution or SGS modeling.
different large-scale forcing. For example, the TKE
budgets calculated from LESS show the expected
differences between the shear- and buoyancy-driven Applications to PBL Research
PBLs (Figure 5 ) . In a shear-driven PBL, shear produc-
Early Work
tion nearly balances molecular dissipation, with all the
other terms remaining small, while in the convective LES has become a prominent research tool in advanc-
PBL the TKE budget is dominated not only by the ing our understanding of the structure and physics of
buoyancy production and molecular dissipation but PBL turbulence. Before Deardorff's LES calculations
also by the turbulent and pressure transports. in the early 1970s, the scaling parameters for the PBL
A significant contribution from LES is the ability to statistics above the surface layer were the friction
obtain the pressure statistics. Pressure fluctuations are velocity u* and the length scale u,/f, where f is the
difficult, if not impossible, to measure in the field, yet Coriolis parameter. From LES calculations, Deardorff
they play an important role in determining moment discovered that the convective PBL turbulence can be
statistics, such as pressure transport in the TKE budget better described using the convective velocity scale
and the return-to-isotropy behavior for velocity var- ' ~ the PBL depth zi, where
w* 5 [ ( g / T o ) ~ i a ] 'and
iances. The LES-generated pressure field, which re- is the surface buoyancy flux.* This new descrip-
mains to be verified from observation when available,
provides a unique tool to estimate important pressure- 'Here overbars (-) denote an ensemble average formed by spatial
related statistics. and temporal averages in LES.
tion, now known as mixed-layer scaling, makes it

possible to collapse data collected from convective
PBLs with different surface heating conditions onto
universal profiles. For example, measurements of the
scaled vertical flux of T K E l a / w ; ) and the scaled
vertical-velocity variance ( w 2 / w i ) ,obtained by air-
craft for different surface heat fluxes, collapse to
universal profiles when shown as functions of the
dimensionless height z / z I (Figure 6 ) (see Boundary
Layers: Convective Boundary Layer).
LES has also provided a revolutionary discovery
about plume dispersion in the convective PBL. The
release of a tracer from an elevated source within an
LES generated convective PBL shows that the maxi-
mum mean concentration in the plume first descends
until the plume intercepts the ground, then rises
(Figure 7 ) .The descent of the elevated plume maxi-
mum is due to the greater areal coverage of down-
drafts, i.e., the positively skewed vertical velocity field.
This finding, also observed at about the same time in
the Willis and Deardorff tank experiments, has an
important application to air pollution; i.e., it can be
used to predict the location and magnitude of the
maximum surface concentration of emissions. The
above results provided the basis for the revision of
short-range dispersion models in the 1980s (see
Turbulent Diffusion).
Another breakthrough from LES is the discovery
of the asymmetry of turbulent diffusion from
area sources at the surface and top of the convective
PBL. Any passive, conservative scalar can be linearly
decomposed into two conceptual scalar fields:
top-down (which is emitted at the PBL top and has
zero flux at the surface) and bottom-up (which
is emitted at the surface and has zero flux at the PBL
top). Under a quasi-steady state, the fluxes of the top-
down and bottom-up scalars are both linear in height
and hence, after normalization by their respective
boundary flux, are symmetric about the mid-PBL. LES
shows that the mean gradients of the top-down and
bottom-up concentrations, after normalization by w*,
z t , and the appropriate boundary flux, are not
symmetric about the mid-PBL. While the top-down
gradient function remains positive throughout the
Figure 4 Contour plots of fluctuations of streamwise velocity u whole PBL, the bottom-up mean gradient is positive in
(top), vertical velocity w (middle), and local flux uw (bottom) in a the lower half of the PBL but becomes negative in the
horizontal plane near the surface from an LES of the shear-driven
upper part. The negative gradient in bottom-up scalar
PBL: u contours (-2,-1.6, -1.2, -0.8.-0.4. -0.2, -0.1.0.1.
0.2,0.4,0.8,1.2), dark (light) shading values larger (smaller) than indicates the countergradient transport feature, where
0.2 (-0.2); w contours (-1.2. -0.9. -0.6, -0.3,-0.1.0.1.0.3. flux and mean concentration gradient have the same
0.6,0.9,1.2), dark (light) shading values larger (smaller) than 0.3 sign. This asymmetric feature of the gradient functions
(-0.3); uw contours (-3,-2.5,-2,-1.5.-1.-0.5,-0.1, results in different top-down and bottom-up eddy
0.1,0.5.1,1.5), dark (light) shading values larger (smaller) than
diffusivities, where the latter becomes ill-defined in the
0.5 (-0.5). (Reproduced with permission from Moeng CH and
Sullivan PP (1994) A comparison of shear- and buoyancy-driven mid-PBL. Thus, in the convective PBL, a scalar emitted
planetary boundary layer flows. Journal of the Atmospheric from the surface diffuses differently from one emitted
Sciences 51 : 999-1 022.) from the top.
I I I I I I I I
-20 0 20 -1 0 1
(A) TKE budget (B) TKE budget
(normalized by u2/zi) (normalized by w?/zi)
Figure5 Vertical distributions of the terms in the TKE budget from LESS of (A) a shear dominant PBLand (B) a buoyancy dominant PBL.
Notation: B, buoyancy production; S, shear production; T , turbulent transport; P , pressure transport; E , molecular dissipation rate.
(Reproduced with permission from Moeng CH and Sullivan PP (1994) A comparison of shear- and buoyancy-driven planetary boundary
layer flows. Journal of the Atmospheric Sciences 51 : 999-1 022.)
Current Research regimes that play important roles in weather or

climate.
Earlier LES work focused mainly on idealized cloud- One climatologically important PBL regime is the
free, flat-terrain convective PBL. That flow regime is stratocumulus-topped PBL (STBL) where the cloud
most suited to LES because of the presence of large amount can significantly alter the solar radiation input
thermal plumes and no other physical processes (e.g., to the Earth's surface. To simulate this PBL regime,
radiation and latent heating) involved. Recently, we include parameterizations of latent heating and
however, LES has been intensively used for more radiation processes, which unfortunately introduce
complicated and difficult PBL regimes because of the more uncertainties into the LES. The focus has been
community need to better understand other PBL studying entrainment processes and estimating the
t Lenschow 1968
A Lenschow 1970
o'8 11 %1 -9- Willis-Deardorff 1
0.6
t+ Adrian-Ferreira
i
$-
0.8 (1-0.9')'
0.4 4 3,
0.2
0.0
0 0.2
' 0.4
-
WE/w 3
Figure 6 Observed profiles of TKE flux and vertical-velocity variance, both scaled with zi and w,. (Adapted with permission from
Lenschow DH, Wyngaard JC,and Pennel WT (1980) Mean-field and second-moment budgets in a baroclinic, convective boundary layer.
Journal of the Atmospheric Sciences 37: 1313-1 326.)
complication arises from the heterogeneous nature

(nonuniformity)of the underlying surface. The Earth’s
land surface is characterized by spatially varying
patches, undulating terrain, and urban development,
which can induce circulations that interact with, and
hence change, the usual turbulence dynamics in the
PBL. Over the ocean, surface currents move at
0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4
different speeds and waves break, leading to complex
air-sea interactions. These complex surface couplings
can significantly influence turbulent transport in many
Figure 7 Contours of nondimensional crosswind-integrated meteorological applications, e.g., air pollution, vege-
concentration as a function of dimensionless height and downwind tation growth, and cloud formation.
distance from an elevated source in a convective PBL as predicted
from LES. Here h is the PBL depth, U the mean wind, and x
The next generation LES for PBL applications faces
downwind distance from the source. (Adapted with permission the challenge of including more realistic surface
from Lamb RG (1982) Diffusion in the convective boundaty layer. conditions for more realistic PBL simulations. For
In: Nieuwstadt FTM and van Dop H (eds.) Atmospheric Turbulence these PBLs, current LESSneed not only improved SGS
and Air Pollution Modeling, pp. 159-229. Dordrecht: Reidel.) models but also better prescriptions for surface and
lateral boundary conditions. Periodic lateral bound-
ary conditions, although convenient computationally,
entrainment rate of the STBL, which remains one of are clearly not applicable for many applications: PBLs
the most difficult problems in PBL turbulence. near coasts and with sloping terrain are good exam-
LES has also been applied to many other compli- ples. Complex surface conditions in particular affect
cated geophysical turbulence problems. They include: turbulence dynamics of the very stable PBL, where
the oceanic mixed layer, which is driven by wind turbulence is no longer continuous but becomes
stress with strong effects from surface gravity intermittent in space and time, and the ability to
waves (see Boundary Layers: Ocean Mixed Layer); simulate local transition to turbulence becomes cru-
turbulence within vegetative canopies, which has cial. There are also attempts to use LES for studying
numerous small-scale influences due to leaves (see interactions between cloud microphysics, biochemis-
Land-Atmosphere Interactions: Canopy Processes); try, and aerosols in the cloudy PBL, which are key
PBLs with inhomogeneous land surfaces, which can players in the climate change process. These applica-
induce local circulations that interact with PBL tions require a thorough reexamination of small-scale
turbulence (see Land-Atmosphere Interactions: Over- effects in LES.
view); PBLs with shallow cumulus clouds; and chem- Field observations can provide a good database for
istry transport and reaction within the PBL, with or examining small-scale effects in PBL-LES. Traditional
without clouds. measurements of turbulence have focused on ensemble
The most practical application of LES to the averages deduced from time series of a single sensor. To
meteorological community so far is the development measure resolved and SGS eddies separately requires
or calibration of PBL parameterizations (i.e., PBL spatial filtering in at least two dimensions. There are
models; see Boundary Layers: Modeling and Param- attempts to measure SGS motions using an array of
eterization) for use in meteorological forecast models. multiple sonic anemometers in the field. These new
Various PBL parameterizations have been proposed, measurements provide insights into the relationship
but few of them have been evaluated or verified between resolved and SGS fields and thus can be used
because field observations are often incomplete for this to improve SGS models for LES of the PBL turbulence.
application. As an alternative, LES results have been
used to examine, for example, closure assumptions of
the terms in the TKE budget; the mass flux and lateral
entrainmenddetrainment closures for mass flux mode-
Nomenclature
ling; entrainment-rate closure assumptions in mixed- SGS diffusion coefficient
layer modeling; and the countergradient effect in eddy Smagorinsky constant
diffusivity models. SGS dissipation coefficient
SGS turbulent kinetic energy (m2 s - ~ )
Coriolis parameter (s-l)
Future Challenges
gravitational acceleration (m sp2)
In reality, the PBL is much more complicated than filter function
what is currently simulated by LES. Much of the PBL depth (m)
1150 LEE VORTICES
k wave number zij SGS stress tensor (m2 s - ~ )

KH SGS eddy diffusivity (m2 s-’) zei SGS heat flux in xi direction (K m s-l)
KM SGS eddy viscosity (m2 s-l)
e SGS length scale (m)
P total pressure fluctuations (hPa) See also
h filtered pressure fluctuations (hPa)
Boundary Layers: Convective Boundary Layer; Modeling
P” SGS pressure fluctuations (hPa)
Pr SGS Prandtl number and Parameterization;Neutrally Stratified Boundary Layer;
Ocean Mixed Layer; Surface Layer. Coriolis Force.
Ri Richardson number
Dynamic Meteorology: Primitive Equations. Land-At-
R i, critical Richardson number mosphere Interactions: Canopy Processes; Overview.
S magnitude of strain tensor (s-l) Numerical Models: Methods. Turbulence and Mixing.
Sij strain tensor (s-l) Turbulent Diffusion.
t time (s)
TO reference temperature (K)
U mean wind speed (m s-’)
flow velocities in xi direction (m s-l)
Further Reading
ui
iif filtered flow velocities in xi direction (m s-l) DeardorffJW (1972)Numerical investigation of neutral and
U!’ SGS flow velocity fields in xi direction (m s-’) unstable planetary boundary layers. Journal of the
u* friction velocity (m s-l) Atmospheric Sciences 29: 91-115.
convective velocity scale (m s-l) Galperin B and Orszag SA (1993)Large Eddy Simulation of
w*
-
ensemble-mean turbulent flux of TKE
Complex Engineering and Geophysical Flows.
WE
New York: Cambridge University Press.
- (m3 s - ~ )
Lumley JL (1990) Whither Turbulence? Turbulence at the
W2 ensemble-mean vertical velocity variance Crossroads:Proceedings of a Workshop Held at Cornel1
- (m2 sp2) University, Ithaca, NY, 22-24 March 1989. New York:
WQO ensemble-mean surface buoyancy flux Springer.
(K m s-l) Mason PJ (1994)Large-eddy simulation: a critical review of
Xi spatial directions for horizontal ( i = 1,2) and the technique. Quarterly Journal of the Royal Meteoro-
vertical (i = 3 ) ( m ) logical Society 120: 1-26.
Xi body forces in xi direction (m s - ~ ) Nieuwstadt FTM, Mason PJ, Moeng C-H and Schumann U
z height (Ex3) (m) (1993) Large-eddy simulation of the convective bound-
PBL height (m) ary layer: a comparison of four computer codes. In: Durst
Zi
F, Launder BE, and Friedrich R (eds) Turbulent Shear
As SGS length scale ( m )
Flows 8 , Berlin: Springer.
E molecular dissipation rate (m2 s - ~ ) Wyngaard JC (1984)Large-Eddy Simulation: Guidelines for
8 virtual potential temperature (K) its Application to Planetary Boundary Layer Research.
e filtered virtual potential temperature (K) US Army Research Office Contract 0804.
8’’ SGS virtual potential temperature (K) Wyngaard JC and Peltier LJ (1996)Experimental microme-
V molecular viscosity of air (m2 s-l) teorology in an era of turbulence simulation. Boundary-
P density of air (kg mP3) Layer Meteorology 78: 71-86.
C C Epifanio, National Center for Atmospheric from above, such a wake is manifest as a pair of
Research, Boulder, CO, USA counterrotating vortices circulating about vertical
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. axes. Observations suggest that this pattern of coun-
terrotating lee vortices can sometimes persist over a
time scale of days. Flow past the island of Hawaii often
Introduction provides examples of such quasi-steady recirculating
Flow of the atmosphere past a high mountain barrier wakes (Figure 1A). In other cases the counterrotating
often results in a low-level wake of reversed flow wake pattern is unstable and a transition occurs to a
immediately downstream of the obstacle. Viewed state in which vortices of alternating sign are period-
1150 LEE VORTICES
k wave number zij SGS stress tensor (m2 s - ~ )

KH SGS eddy diffusivity (m2 s-’) zei SGS heat flux in xi direction (K m s-l)
KM SGS eddy viscosity (m2 s-l)
e SGS length scale (m)
P total pressure fluctuations (hPa) See also
h filtered pressure fluctuations (hPa)
Boundary Layers: Convective Boundary Layer; Modeling
P” SGS pressure fluctuations (hPa)
Pr SGS Prandtl number and Parameterization;Neutrally Stratified Boundary Layer;
Ocean Mixed Layer; Surface Layer. Coriolis Force.
Ri Richardson number
Dynamic Meteorology: Primitive Equations. Land-At-
R i, critical Richardson number mosphere Interactions: Canopy Processes; Overview.
S magnitude of strain tensor (s-l) Numerical Models: Methods. Turbulence and Mixing.
Sij strain tensor (s-l) Turbulent Diffusion.
t time (s)
TO reference temperature (K)
U mean wind speed (m s-’)
flow velocities in xi direction (m s-l)
Further Reading
ui
iif filtered flow velocities in xi direction (m s-l) DeardorffJW (1972)Numerical investigation of neutral and
U!’ SGS flow velocity fields in xi direction (m s-’) unstable planetary boundary layers. Journal of the
u* friction velocity (m s-l) Atmospheric Sciences 29: 91-115.
convective velocity scale (m s-l) Galperin B and Orszag SA (1993)Large Eddy Simulation of
w*
-
ensemble-mean turbulent flux of TKE
Complex Engineering and Geophysical Flows.
WE
New York: Cambridge University Press.
- (m3 s - ~ )
Lumley JL (1990) Whither Turbulence? Turbulence at the
W2 ensemble-mean vertical velocity variance Crossroads:Proceedings of a Workshop Held at Cornel1
- (m2 sp2) University, Ithaca, NY, 22-24 March 1989. New York:
WQO ensemble-mean surface buoyancy flux Springer.
(K m s-l) Mason PJ (1994)Large-eddy simulation: a critical review of
Xi spatial directions for horizontal ( i = 1,2) and the technique. Quarterly Journal of the Royal Meteoro-
vertical (i = 3 ) ( m ) logical Society 120: 1-26.
Xi body forces in xi direction (m s - ~ ) Nieuwstadt FTM, Mason PJ, Moeng C-H and Schumann U
z height (Ex3) (m) (1993) Large-eddy simulation of the convective bound-
PBL height (m) ary layer: a comparison of four computer codes. In: Durst
Zi
F, Launder BE, and Friedrich R (eds) Turbulent Shear
As SGS length scale ( m )
Flows 8 , Berlin: Springer.
E molecular dissipation rate (m2 s - ~ ) Wyngaard JC (1984)Large-Eddy Simulation: Guidelines for
8 virtual potential temperature (K) its Application to Planetary Boundary Layer Research.
e filtered virtual potential temperature (K) US Army Research Office Contract 0804.
8’’ SGS virtual potential temperature (K) Wyngaard JC and Peltier LJ (1996)Experimental microme-
V molecular viscosity of air (m2 s-l) teorology in an era of turbulence simulation. Boundary-
P density of air (kg mP3) Layer Meteorology 78: 71-86.
C C Epifanio, National Center for Atmospheric from above, such a wake is manifest as a pair of
Research, Boulder, CO, USA counterrotating vortices circulating about vertical
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. axes. Observations suggest that this pattern of coun-
terrotating lee vortices can sometimes persist over a
time scale of days. Flow past the island of Hawaii often
Introduction provides examples of such quasi-steady recirculating
Flow of the atmosphere past a high mountain barrier wakes (Figure 1A). In other cases the counterrotating
often results in a low-level wake of reversed flow wake pattern is unstable and a transition occurs to a
immediately downstream of the obstacle. Viewed state in which vortices of alternating sign are period-
LEE VORTICES 1151
(A) (B)
Figure 1 (A) Aerial photograph of cloud layer in flow past the island of Hawaii. The cores of lee vortices feature warmer air than the
surrounding flow and are typically manifest in aerial and satellite photographs as holes in the cloud layer. The two lobes of clear air
extending downstream of the island in (A) are the signature of a pair of counterrotating vortices. Arrows suggest the flow field as inferred
from the cloud pattern (Photo courtesy of Vanda Grubiiic). (B)Vortex street downstream of Alejandro Selkirk Island off the coast of Chile
as seen by the Landsat 7 Satellite. The island is in the bottom left part of the figure.
ically shed downstream to form a vortex street. The formation lee vortices; but in general the effect of
imprint of vortex streets in cloud layers downstream of planetary rotation on lee-vortex formation is a topic in
mountainous islands is occasionally captured in sat- need of further study.
ellite images. A particularly striking example is shown
in Figure 1B.
Observational studies suggest that wakes and vor- Stratified Flow Past Topography:
tices are a common feature of atmospheric flow in Basic Phenomenology
mountainous regions. Lee vortices that form in the
Overview
vicinity of cities tend to recirculate pollutants and thus
have important consequences for local air quality. Here we briefly review the basic phenomenology of
Examples include the Melbourne Eddy (near Mel- nonrotating stratified flow past an isolated ridge in
bourne, Australia), the Santa Barbara Eddy (near three dimensions (3D) as revealed by theoretical and
Santa Barbara, CA, USA), and the Denver Cyclone numerical investigations and laboratory experiments.
(near Denver, CO, USA). In some cases orographic Most idealized studies of 3D flow over orography have
vortices are linked with the initiation and intensifica- assumed a uniform upstream flow speed U and
tion of severe weather. A well-studied example is the constant upstream buoyancy frequency (or Brunt-
Denver Cyclone, which is often associated with the Vaisala frequency) N,and we make the same restric-
development of severe storms producing hail, flood- tions here. Such a model gives a rough first approx-
ing, and tornados. Studies of flow past the Alps suggest imation to many atmospheric flows but excludes
that low-level mountain wakes may also interact with phenomena that depend on vertical variations in N
upper-level troughs to produce larger, synoptic-scale and U such as trapped lee waves (or so-called ship
lee cyclones. waves). We suppose that the ridge shape may be
Lee vortices generally develop on time scales that characterized by a streamwise length scale a, a cross-
are short compared to a day and have length scales on stream length scale b, and a maximum height h. The
the order of 10-100 km. As a result, the rotation of the nondimensional control parameters governing the
Earth has only a secondary effect on the motion. Most behavior of the flow for constant N and U are then:
theoretical studies of mountain wakes have neglected (a) the nondimensional mountain height E = N h / U ,
the Coriolis force (i.e., they have considered nonro- which measures the amplitude of the disturbance; (b)
tating flow) and we focus on such studies in the the vertical aspect ratio 6 = U / N a , which measures
following sections. There is some evidence to suggest the importance of nonhydrostatic effects; and (c) the
that on longer time scales and larger spatial scales the horizontal aspect ratio p = b/a. For most atmos-
influence of the Earth’s rotation tends to suppress the pheric flows of interest the vertical aspect ratio 6 is
1152 LEE VORTICES
small (less than O . l ) , so that the flow is essentially

hydrostatic; the set of control parameters then reduces
to E and /3.
Laboratory, numerical, and theoretical studies sug-
gest roughly four classes of phenomena of importance
in hydrostatic flow over topography with uniform
upstream N and U : (i) small-amplitude waves; (ii)
wave breaking; (iii) upstream stagnation and flow-
splitting; and (iv) lee vortices. The schematic regime
diagram in Figure 2 summarizes the occurrence of
these phenomena as a function of E and p. Detailed
descriptions of the flow classes are as follows.
Small-amplitude waves When E << 1 the mountain-

induced disturbance for all /3 takes the form of a small-
amplitude mountain wave (see Lee Waves and Moun-
tain Waves for examples of small-amplitude waves).
The flow in this regime is well described by theoretical
approaches valid in the limit of small E . As E increases,
the streamlines in the wave pattern above the lee slope
steepen somewhat in the vertical as a result of
nonlinear (or finite-amplitude) effects.
Wave breaking For /la 1 (i.e., for elongated ridges),

streamlines above the lee slope overturn when E
exceeds a critical value E ~ ( P ) usually
, in the range
0.7-1.2 depending on obstacle shape. Overturning of
the streamlines places more-dense fluid over less-dense
fluid, causing the wave to break and become locally
turbulent. An example of overturning streamlines and
wave breaking in a laboratory flow with E = 2.5 and
p = 3 is shown in Figure 3A. Wave breaking produces Figure 3 (A) Dye lines in the centerline plane for flow over a 3D
a well-mixed region above the lee slope that tends to ridge with E = 2.5 and /3 = 3. The streamline above the lee slope
decouple the low-level lee-slope flow from the flow has steepened to the point of overturning, causing the wave to
break and the flow to become turbulent. (From Castro IP and
farther aloft. The resulting low-level flow is strongly Snyder WH (1 993) Experiments on wave-breaking in stratified flow
accelerated and is similar to supercritical shallow- over obstacles. JournalofFluidMechanics255: 195-21 1.) (B) Dye
lines in the centerline plane and on the obstacle surface for flow
over a ridge with E = 2.5 and p = 2. The basic flow is from left to
right. Flow splitting is apparent on the windward slope below the
41 ,.*-
second height contour. Also apparent is a wake in the lee with a hint
Flow splitting,
Lee vortices
,/'
I of possible reversed flow along the centerline. (From Baines PG
,,' (1995) Topographic Effectsin Stratified Flows. Cambridge: Cam-
,/' Wave breaking,
bridge University Press.) (C) Surface streamlines from a numerical
Flow splitting,
computation using an obstacle identical to that shown in (B). Note
Lee vortices
that the dye lines in (B) are suggestive of the streamlines in (C).
(Figures 3A and 38 reprinted with the permission of Cambridge
University Press.)
.....................
1 Wave breaking
Small-amplitude waves water (or single-layer hydraulic) flow (see Downslope

0 I I I I I I / I I I / Winds for a discussion of shallow-water flow over a
0 1 2 3 4 5 6 ridge; see Hydraulic Flow for a basic discussion of
/3 shallow-water dynamics). Downstream of the obsta-
Figure 2 Schematic regime diagram for stratified flow past an cle the depressed lee-slope streamlines abruptly return
isolated ridge as a function of E and p. Note that the actual shapes to their upstream heights in a structure resembling a
and positions of the curves will depend on the obstacle shape. shallow-water hydraulic jump (Figure 3A).
LEE VORTICES 1153
Flow splitting For all fi, there exists a critical

mountain height ~ ~ ( atf i which
) the flow stagnates on
the upstream face of the ridge. This flow stagnation is
due to the positive upstream pressure anomaly
associated with the disturbance and occurs at a finite
height zs on the upstream face. Below zs the flow splits
and passes primarily around the obstacle rather than
over, while above zs the flow ascends the crest as for
(4 (B)
E <E,. Figure 3B illustrates upstream stagnation and
flow splitting in a laboratory flow with E = 2.5 and Figure 4 (A) Schematic close-up of the boundary layer at the
fi = 2; Figure 3C shows surface streamlines from a surface of a solid obstacle. (B) Separation of the boundary layer.
numerical computation with the same E and fi.
Upstream splitting reduces vertical displacements in
the flow and thus tends to suppress wave breaking. For complex than in laboratory experiments, but the basic
fis I, upstream stagnation occurs at smaller E than premise of a thin shear zone still applies. For our
wave breaking and the latter is suppressed for all E . By purposes we might define the thickness of the bound-
contrast, for elongated ridges &b<Es and the pheno- ary layer to be the depth over which shear effects
mena are observed to coexist over some range in E dominate the effects of stratification. In the stably
(cf. Figure 2 ) . stratified atmosphere this depth is typically several
tens of meters.
Above the boundary layer the flow speed varies
Lee vortices A counterrotating pair of eddies
relatively little with height and viscous effects are
appears in the lee of the obstacle when E exceeds
negligible. In principle, the flow in this region could be
a value (Figures 3B, C ) . In laboratory
determined by displacing the effective obstacle surface
studies extensive lee vortices are typically observed
upward to a flow surface (or material surface) at the
only when upstream stagnation is also present, as in
top of the boundary layer and assuming that the fluid
Figure 3B. However, numerical simulations indicate
passes freely over this surface with no frictional
that for fi< 1 vortices form at mountain heights
coupling. If the boundary layer is thin and remains
smaller than E ~ ,suggesting that vortex formation is
everywhere attached to the obstacle, a reasonable
essentially independent of flow splitting. In any case, it
approximation is to neglect the boundary layer
seems to be true that for most obstacle shapes E, x
altogether and consider the obstacle surface to be
(fi) M ~ , ( f i for
) fi > 1and the two curves are taken to be frictionless. This approximation is referred to as a free-
coincident over this range of fi in Figure 2 . Counter-
slip condition.
rotating vortex pairs with weak flow reversal are likely
When atmospheric vortex streets were first observed
to be stable and quasi-steady. Vortex pairs with
in satellite imagery in the early 1960s, it was com-
extensive regions of vigorous reversed flow are
monly assumed that the vorticity of the wake derives
expected to be unstable and transition to a vortex-
from the shear in the boundary layer. The process by
shedding state.
which the vorticity of the boundary layer enters the
interior of the flow is well established (if not com-
pletely understood) for high-Re homogenous flows. If
Boundary Layers and Free-Slip Models
the boundary layer fluid traversing the obstacle surface
Fluid flowing past a stationary solid topographic at some point encounters an adverse pressure gradient,
obstacle is constrained to be motionless at the obstacle the boundary layer tends to separate from the obstacle
surface owing to frictional coupling between the fluid and shed its vorticity into the fluid interior (Figure 4B).
and the solid boundary. The character of the flow Boundary-layer separation is responsible for the
immediately above the boundary depends on the production of wakes in most flows of interest in
influence of viscous effects, as measured by the engineering applications (flow past airplanes, cars,
Reynold’s number Re = Uu/v, where v is the kinematic etc.).
viscosity. In laboratory models of atmospheric flows, Laboratory studies refined this view by demonstrat-
Re is usually quite large (>lo3).The flow immediately ing that in stratified flow over topography-boundary
above the obstacle surface then takes the form of a thin layer separation is strongly influenced by the stratified
boundary layer over which the flow speed increases wave dynamics above the boundary layer. In
from zero at the surface to roughly the free-stream particular, it was found that the mountain-wave
speed U away from the surface (Figure 4A). The pattern of depressed streamlines and accelerated
boundary layer in the atmosphere is generally more flow above the lee slope tends to suppress lee-slope
1154 LEE VORTICES
boundary-layer separation. For obstacles of gentle occur without breaking waves when the upstream
slope ( h / a = ES 5 0 . 2 ) ,as is typical in the atmospheric conditions include a strong low-level inversion (or
context, boundary-layer separation does not occur at layer of large N), as is often the case in Trade Wind
all for small-amplitude hydrostatic waves (unless lee flow. To a first approximation the inversion acts as a
waves are present owing to reflection from the tank free surface when large-amplitude waves are present.
top). However, separation may occur in connection Indeed, observations of lee wakes in Trade Wind flow
with hydraulic jumps (as in Figure 3A) or similar past the island of Hawaii are suggestive of hydraulic-
structures when wave steepening and breaking are type behavior including weak hydraulic jumps. Con-
present. Flow splitting is also believed to promote siderations such as these have led to attempts to model
boundary-layer separation. In this case the fluid below lee vortex formation in terms of shallow-water theory.
the stagnation height zs passes nearly horizontally Note that the shallow-water results described here
around the obstacle and separates from the boun- have close analogues for stratified flow as discussed in
dary in essentially two-dimensional fashion (i.e. as in the following section.
flow past a vertically oriented cylinder). This separa- We consider flow of a thin layer of homogenous fluid
tion of the boundary layer in nearly horizontal past a submerged obstacle of gentle slope. Such a flow
planes has often been invoked to explain the observed is well described by the shallow-water equations (see
eddies circulating about vertical axes in stra- Hydraulic Flow for details) (eqns [ l ] and [2]).
tified wakes.
By the late 1980s computing power had increased to aU
-+
the extent that numerical computations of 3D flow at
(u.V)u = PI
over topography were relatively common. Research-
ers soon discovered that numerical models with free-
slip boundary conditions reproduce the observed
3
at
+v * (u(d + =0
features of topographic wakes in stratified laboratory Here d(x, y) is the resting depth of the fluid, y(x, y) is
experiments with surprising accuracy, at least for the displacement of the free surface from the rest
small 6 and moderate E (say, 10). As discussed position, u = (u(x, y ) , u(x, y)) is the horizontal flow
above, free-slip models completely neglect the bound- field, and V = (a/ax, a/ay). The geometry of the fluid
ary layer, so that the numerically simulated lee vortices layer is indicated in Figure 5.From eqns [l]and [2] are
are clearly not due to boundary-layer separation. derived the useful relations [3] and [4], where B =
Similarly, researchers modeling observed atmospheric us u/2 + gy is the Bernoulli function and i is the
wakes have found that reducing or neglecting surface vertical vorticity (eqn [5]).
friction often results in intensification of the vortices.
On the basis of these results it is reasonable to suppose
- + u * v B=-g- aY
that for hydrostatic flow at moderate E , the formation
of lee wakes and vortices is not intrinsically dependent
t:( ) at [31
on boundary-layer separation. The vorticity of the

wake is instead generated in the interior of the fluid
through either buoyancy gradients (i.e., baroclinicity)
(a+ u . v) d+vl=
i 0 [41
or turbulent stresses. At steady state both B and i / ( d + y ) are unchanged
Most of the theoretical work on topographic following the flow. The vorticity eqn [4] is often
wakes in the past decade has focused on wakes in
free-slip flows. The following sections review some of
the basic results of this effort. However, it should be
kept in mind that for 6 2 1 or e 2 10 boundary-layer
separation may in fact dominate the behavior of the
wake.
Shallow-Water Theory
As discussed in the overview above, analogues to
supercritical shallow-water flow and hydraulic jumps
Figure 5 Side view of shallow-water flow through a hydraulic
may occur in stratified flows with uniform N and U jump. In the limit of ideal flow the width of the jump shrinks to zero,
when the obstacle is sufficiently high to force over- resulting in a discontinuity. This is indicated by the dotted vertical
turning waves. Similarities to hydraulic flow also line in the figure.
LEE VORTICES 1155
usefully considered in the equivalent conservation form

with J = u[ denoting the flux of vorticity (eqn [.SI).
From eAqn[ l ] it is straightforward to show eqn [6],

where k is the vertical unit vector.
The Bernoulli function thus serves as the effective Figure 6 Schematic illustration of vorticity production in steady
streamfunction for the vorticity flux at steady state. flow past a hydraulic jump. The heavy line over the lee slope
indicates the position of the jump. Solid lines are streamlines and
Let the upstream flow be given by constant speed U shading represents regions of reduced Bernoulli function, with
and depth D and let the maximum obstacle height be h. darker shading indicating greater reduction. (Here 6, is the value
The nondimensional control parameters for the flow of the Bernoulli function upstream.) Heavy solid arrows show the
are then the upstream Froude number FO = U/(gD)1’2 advective vorticity flux u[ as inferred from the Bernoulli gradient
and the nondimensional mountain height M = h / D . using eqn [6]. The open arrow indicates the flux of [ in the jump.
(Reproduced with permission of the American Meteorological
Numerical calculations suggest that vigorous wakes Society, from Schar C and Smith RB (1993) Shallow-water flow
with recirculating vortices are possible only when the past isolated topography. Part I: Vorticity production and wake
upstream flow is subcritical (i.e., Fo < 1)and we restrict formation. Journal of the Atmospheric Sciences 50: 1373-1 400.)
attention to this case. For simplicity we assume that the
flow has adjusted to be steady everywhere except
possibly far downstream of the obstacle. e. The sense of the vorticity advection implies [ > 0 on
Both theory and numerical computations show that the right side of the flow (facing downstream) and < 0
for obstacle heights M less than a critical value M, on the left.
(depending on F o ) the flow remains subcritical and Note that the conservation equation [5] can in fact
continuous everywhere. In this case eqn [4] shows that be generalized to hold for arbitrary (Le., dissipative
the vorticity is everywhere zero (sinceit is zero upstream) and nonhydrostatic) flow conditions. It is then con-
and no wake forms. However, when M exceeds M,the sistent to suppose that in addition to conserving mass
flow becomes supercritical at the crest and over the and momentum the flow in the hydraulic jump also
lee slope of the obstacle and continuous solutions no conserves [ (in the sense given by eqn [SI). It can be
longer exist. A hydraulic jump then forms in the lee shown that for steady flow the conservation of [
as shown in Figure 5 . As discussed below, the formation implies a flux of vorticity in the jump directed tangent
of a hydraulic jump in general implies vertical vorticity to the jump with the net flux given by the local
production and consequent wake development. Bernoulli difference VB across the jump. This flux is
A shallow-water hydraulic jump is essentially a denoted by the open arrow in Figure 6. Since VB varies
thin transition zone between two regions of along the jump there is in general nonzero divergence
fluid where eqns [l] and [2] apply. In the jump (or convergence) of the vorticity flux in the jump. This
zone the dynamics differs from that described by eqns divergence/convergence is exactly balanced by the
[ l ] and [2] and some sort of dissipation occurs. Note local advective flux of [ into/away from the jump as
that details of the flow in the jump need not be given by eqn [6].
specified. All that is required to determine the flow We thus find that in a hydraulic jump of finite extent,
outside the jump is that mass and momentum be vertical vorticity is produced and then advected away
conserved and energy dissipated in the jump. Since downstream. If the vorticity production is sufficiently
eqns [l]and [2] do not hold in the jump, the material strong, the resulting wake will feature reversed flow
relations [3] and [4] are violated in the jump as well. In along the centerline and an associated pair of recircu-
particular, the presence of dissipation produces a lating lee vortices. An example of such a flow is shown
decrease in the Bernoulli function for particles cross- in Figures 7A, B.
ing the jump. This results in a wake of fluid with The structure of the wake is modified somewhat
decreased Bernoulli function extending downstream when the height of the obstacle is sufficient to
of the jump, as shown in Figure 6. For a steady jump of penetrate the fluid surface. Instead of a single hydrau-
finite length, such a wake entails gradients of the lic jump in the lee of the mountain, a pair of jumps
Bernoulli function which, according to eqn [6], are form, one on each side of the obstacle. Figures 7C, D
necessarily associated with nonzero advective fluxes of give an example of this case. The detailed time
1156 LEE VORTICES
Figure 7 Lee vortices in steady shallow-water flow past an isolated obstacle. (A) Streamlines and (B) nondimensional vorticity
a [ / ( g D ) ’ / * (contour interval 0.3 with negative contoursdashed and zero contour suppressed) for M = 0.8and Fo = 0.5. (C) and (D) as in
(A) and (B) but for the case M = 2 and Fo = 0.5 in which the obstacle protrudes through the fluid surface. Heavy lines in all panels indicate
the positions of hydraulic jumps. (Reproduced with permission of the American Meteorological Society, modified from Schar C and Smith
RB (1993) Shallow-water flow past isolated topography. Part I: Vorticity production and wake formation. Journal of the Atmospheric
Sciences 50: 1373-1 400.)
evolution leading to the steady state shown in Figures of the previous section can be based on analogues
7 C , D is somewhat complex and beyond the scope of of eqns [5] and [6] for stratified flow. The strati-
the present discussion. However, at steady state at fied generalization of the conservation equation [5]
least, the basic principles of Bernoulli reduction and -
takes the form [ 7 ] ,where Q = 5 V8 is the potential
vorticity generation in the jumps still apply. v0rticity.l
Shallow-water theory allows some basic insight into -aQ
+V.J=O
the nonlinear dynamics of orographic flows when at [71
hydraulic-like conditions are present. Such flows are
Here [is the three-dimensional vorticity vector and 8 is
commonly observed to form lee vortices. The drawback
the potential temperature. The potential vorticity flux
of this approach is that the processes responsible for
J may be divided into an advective part UQ and a
wake formation occur in hydraulic jump regions where
dissipative part JD the latter resulting from viscous and
the details of the flow are not specified. In particular,
diabatic effects. It can then be shown that the stratified
shallow-water theory does not explicitly address the
version of eqn [6] is eqn [8], which is clearly analogous
dynamics of vorticity generation in the wake. A deeper
to eqn [ 6 ]at steady state.
understanding of wake formation then necessitates the
consideration of more complete models.
Stratified Theory ‘The potential vorticity is usually defined as II = 5 . V % / p where

,
p is the density. Then Q = p I I is the potential vorticity per unit
The dynamics of stratified fluids is significantly more volume in the same sense that p is the x momentum per unit volume.
complex than that of a shallow layer of homogenous As this is cumbersome terminology, we refer to Q simply as the
fluid. Nonetheless, an analysis similar in spirit to that potential vorticity.
LEE VORTICES 1157
+
Here B = u.u/2 c,T + gz (with T the sensible be noted that this steady-state relationship does not
temperature, cp the specific heat at constant pressure, imply a sense of causality. At present it is not clear
and g the gravitational constant) is the Bernoulli whether viscous and diabatic effects play a primary
function, which at steady state is constant following role in vortex development or are instead the by-
the flow (cf. eqn [ 3 ] )except where modified through product of an essentially inviscid and adiabatic
dissipation. The relation [8] shows that for steady tendency to form vortices (or whether such a distinc-
flow, gradients of B on an isentropic surface (or surface tion can even be made). Note also that while the
of constant e) imply fluxes of potential vorticity. steady-state Bernoulli analysis predicts the production
Figure 8 depicts the flow fields on an isentropic of potential vorticity in the wake due to dissipative
surface passing over an obstacle with a thin dissipative processes, it does not address in detail the dynamics of
region over the lee slope. The dissipation may be due to vorticity and potential vorticity generation. As such,
wave breaking, a hydraulic jump, or a similar distur- the approach has the same limitation as the shallow-
bance. As in the shallow-water case, the localized water analysis, namely, that it does not explain how
dissipative region produces a wake of fluid with the individual air parcels in the wake acquire vorticity
reduced Bernoulli function extending downstream. or potential vorticity.
At steady state the Bernoulli gradients implicit in such The most fundamental attempt to explicitly account
a wake are necessarily associated with nonzero fluxes for vorticity generation in orographic wakes is based
of potential vorticity as given by eqn [SI. Since JD = 0 essentially on the vorticity dynamics of inviscid three-
away from the dissipative region, the potential vorti- dimensional mountain waves. In incompressible (as is
city fluxes downstream are advective and imply approximately the case for most atmospheric flows of
nonzero Q. We thus find that potential vorticity is interest) and inviscid stratified flow, the vorticity is
generated in the dissipative region and advected governed by eqn [ 9 ] ,where p is the density and p is the
downstream at the lateral edges of the wake. Down- pressure.
stream of the dissipative region the isentropic surface
is essentially horizontal, so that the presence of at; . VPXVP p,
nonzero Q also implies nonzero vertical vorticity. If -+
at
(u * V)t;= (t; V)u+- P’
the vertical vorticity is sufficiently strong, the associ-
ated wake will possess reversed flow and a pair of lee The second term on the right gives the generation of
vortices. vorticity by baroclinicity, while the first term describes
The above analysis nicely demonstrates the stretching and tilting of the vorticity by the flow field.
close relationship between dissipative processes In most orographic disturbances the baroclinic term is
and wake structure at steady state. However, it should well represented by eqn [lo], where k is again the
vertical unit vector and b is the buoyancy (see
Buoyancy and Buoyancy Waves: Theory).
In this approximation, vorticity is generated through

horizontal gradients in the buoyancy and the associ-
ated tendency toward vertical motion. Note that the
vorticity generation is entirely horizontal. To produce
vertical vorticity the baroclinically generated horizon-
tal vorticity must be tilted by the flow field so as to have
Figure 8 Schematic depiction of potential vorticity production in a vertical component.
steady stratified flow past a topographic obstacle. Flow fields are
For inviscid and adiabatic flow the potential vorticity
shown on a low-level isentropic surface passing over the obstacle.
Thin lines are streamlines. Dark shading over the lee slope Q remains everywhere zero (if it is zero upstream) and
indicates a localized region of dissipation due to wave breaking or a the vorticity is thus tangent to isentropic surfaces.
hydraulic jump. The shaded area extending downstream repre- Vertical vorticity can then be diagnosed by tracing
sents reduced values of the Bernoulli function, with darker shading vortex lines’ along a constant4 surface while account-
indicating greater reduction. Open arrows show the potential
ing for the vertical deflections of the surface. If the
vorticity flux J associated with the Bernoulli gradient on the
isentropic surface. (Reprinted with permission of the American vortex lines cross contours of height on the surface, then
Meteorological Society, from Schar C and Durran DR (1997)
Vortex formation and vortex shedding in continuously stratified ’A vortex line is a curve in space along which the vorticity is
flows past isolated topography. Journal of the Atmospheric everywhere tangent. Vortex lines are thus the analogues for the
Sciences 54: 534-554.) vorticity field of streamlines for the velocity field.
1158 LEE VORTICES
vertical vorticity is necessarily present. Note that for connect. Figure 9B shows a vortex line on the
adiabatic flow the potential temperature 8 is unchanged isentropic surface corresponding to the streamline in
following the motion of fluid particles; in this case Figure 9A. Downstream of the obstacle both the
isentropic surfaces are also stream-surfaces in the flow. surface and the vortex line are displaced downward by
Figure 9 summarizes the basic dynamics leading to the descending flow, while upstream the displacement
vertical vorticity production in inviscid and adiabatic is upward. As a result, there are regions at the ends of
mountain waves. The heavy solid line in Figure 9A the obstacle where the vortex line ascends or descends
represents a streamline in the wave field above the the sloping isentropic surface and the vorticity has a
mountain at a height N z / U = 7112. Above the wind- nonzero vertical component. Note that the sign of the
ward slope, fluid particles in the wave ascend and vorticity is of the appropriate sense to describe the flow
become negatively buoyant, leading to a negative x - in lee vortices.
gradient of buoyancy upstream of the obstacle. We thus find that baroclinically generated horizon-
According to eqn [lo] this generates vorticity of the tal vorticity inherent in the mountain wave is tilted by
sense pointing into the plane of the figure (i.e., in the the ascending and descending flow in the wave to
positive y direction). As the flow descends above the produce vertical vorticity at the lateral ends of the
lee slope, the opposite occurs and a negative y - obstacle. This vorticity may then contribute to the
component of vorticity is generated. In three dimen- formation of a counterrotating vortex pair. Further
sions these vortex lines running into and out of the analysis suggests that dissipation may act to extend the
figure loop around the lateral ends of the obstacle and vorticity pattern downstream in a pair of potential
vorticity streamers as depicted in Figure 8.
Recent work suggests a possible link between the
0.375
mountain-wave analysis described above and the
shallow-water approach of the previous section.
0.25
Figure 10 shows results of a numerical simulation of
stratified flow past an isolated ridge with E = 1.8 and
2cu
. fi = 5. A hydraulic jump similar to that in Figure 3A
9 0.125 forms downstream of the obstacle; the shading in
Figure 10 shows the position of the jump. Behind the
jump the flow is weakly reversed, indicating the early
" stages of vortex formation. From a macroscopic
-4 -2 0 2 4 perspective the flow in Figure 10 is similar to the
(A) xla
shallow-water calculation of Figures 7A, B, with
streams of vertical vorticity extending downstream of
the lateral ends of the jump. However, in the case of
Figure 10 it can be shown that the vertical vorticity of
the wake originates in the mountain wave upstream of
the jump through a mechanism similar to that
described in Figure 9 . Upon reaching the jump, the
vorticity is amplified severalfold through vertical
stretching to produce the pronounced vorticity anom-
alies at the edges of the wake.
In the flow of Figure 10 the viscosity is sufficiently
high to suppress the onset of small-scale turbulence.
Real flows are likely to be turbulent in both the
steepened mountain wave and the hydraulic jump,
raising the possibility of vorticity generation by
Figure 9 Vertical vorticity production in a 3D mountain wave. (A)
turbulent stresses. It remains to be seen how well the
Streamline with upstream height NZIU= 7d2 in the centerline plane
for flow over an obstacle with E = 0.5 and p = 2.4. Circular arrows mechanisms of vorticity generation in the viscous
indicate horizontal vorticity generated through buoyancy gradients. laminar model extend to more realistic flows with
(B) Schematic vortex line on the isentropic surface corresponding turbulence. The details of vorticity generation in some
to the streamline in (A). Circular arrows indicate vertical vorticity in large-s ( E k 3, say) cases with prominent flow-splitting
regions where the vortex line ascends or descends the sloping are also uncertain at present. Numerical simulations
isentropic surface. (After Smolarkiewicz PK and Rotunno R (1989)
Low Froude number flow past three-dimensional obstacles. Part I: of flow at large E often show jump-like features on the
Baroclinically generated lee vortices. Journal of the Atmospheric lateral slopes of the obstacle as in the shallow-water
Sciences 46: 1154-1 164.) calculations of Figures 7C, D. The tilting and stretching
LEE VORTICES 1159
property of the shear flow in the wake and has little

dependence on either the mechanisms of wake forma-
tion or the details of the flow near the obstacle. As
such, the transition to vortex shedding in geophysical
flows is expected to be somewhat similar to that found
in homogenous flows past submerged obstacles, for
which there is an extensive literature. Some of the
more illustrative results on vortex shedding in 2D
homogenous flows have been extended in detail to
shallow-water flow over an isolated obstacle. Subse-
quent studies of wake stability in stratified flow have
revealed instability structures similar to those found in
the shallow-water case. However, it should be noted
that the literature on wake stability and vortex
shedding in stratified flows is rather limited.
Figure 11 describes the instability and transition to
vortex shedding of an elongated wake in shallow-water
flow. The wake in Figures 11A, B supports essentially
two distinct types of instability. The shear lines at the
lateral edges of the wake each in isolation support
unstable modes associated with the local vorticity
Figure 10 Incipient vortex formation in viscous stratified flow
extremum. However, the energy of these modes prop-
past a long ridge with E = 1.8, 3 / = 5 and Re = 1000. Fields are agates rapidly downstream away from the obstacle
shown on a low-level terrain-following surface. Vectors show the before the disturbance has an opportunity to grow
horizontal velocity and contours give the nondimensional vertical significantly. As a result, these modes have little impact
vorticity [ / E ~(contour
~N interval 0.21 with negative contours on the overall (or global) stability of the wake. A
dashed and zero contour suppressed). Shaded area indicates the
position of a hydraulic jump similar to that shown in Figure 3A.
second type of unstable mode is antisymmetric about
(From Epifanio CC and Durran DR (2002) Lee-vortex formation in the wake centerline and depends on a coupling of the
free-slip stratified flow over ridges. Part I: Comparision of weakly two shear lines with the reversed flow in between.
nonlinear inviscid theory and fully nonlinear viscous simulations. Modes of this type grow in place and disrupt the
Journal of the Atmospheric Sciences 59: 1153-1 165.) counterrotating wake pattern. The disturbance is
typically first manifest as a wavelike oscillation near
mechanism described above likely plays an important the downstream stagnation point (Figure 11C).At later
role in creating the vertical vorticity of the wake in times (Figure 11D) isolated patches of vorticity are
such cases. However, in other large-s flows the jumps shed from the downstream end of the wake. Eventually
on the lateral slopes are weak or apparently absent (at the disturbance becomes evident in the immediate lee
least at steady state). One possibility is that the of the mountain as well and vortices of alternating sign
Bernoulli dificit of the wake in such cases is determined are shed from either side of the obstacle with nearly
not by jump-like features but rather by weak dissipa- perfect periodicity. In this final (or saturation) state the
tion distributed throughout the length of the wake. A flow is similar to that shown in Figure 1B.
Bernoulli gradient and associated potential vorticity Detailed stability analyses of wake flows show
flux are then produced where the recirculating (low- that the growing disturbance tends to emanate from
Bernoulli) wake flow joins the incident flow on the the part of the wake featuring reversed flow. Wakes
flanks of the obstacle. Further work is needed to without reversed flow are expected to be stable.
understand the nature of vorticity generation in this Similarly, if the wake features only a limited region
latter type of flow. of weak reversed flow, then the dispersion of distur-
bance energy away from this region may be sufficient
to suppress the growth of the mode. The transition (or
Vortex Shedding bifurcation) from a stable wake to an unstable wake
Vortex shedding and vortex streets are thought to with increasing obstacle height has yet to be explored
result in most cases from the instability of an attached for shallow-water flow. It seems likely that in the free-
(i.e., nonshedding) counterrotating wake flow. Any slip inviscid case virtually any wake with flow reversal
noise or impulsive disturbance in the flow tends to will eventually become unstable as the wake elongates
trigger the instability and leads to growing modes that in the downstream direction. However, the addition of
disrupt the attached wake pattern. The instability is a viscosity, and/or bottom friction tends to stabilize
1160 LEE VORTICES
wakes with weak reversed flow. This stabilization is

effected primarily through a decrease in the down-
stream length of the wake (thus reducing the extent of
the flow reversal) and t o a lesser degree through direct
reduction in the growth rates of the disturbance
modes. Numerical computations with realistic param-
eters suggest that bottom friction may indeed exert an
important stabilizing influence on atmospheric wakes,
such as that behind the island of Hawaii (Figure 1A).
See also
Boundary Layers: Overview. Buoyancy and Buoyancy
Waves: Theory. Downslope Winds. Hydraulic Flow.
Lee Waves and Mountain Waves. Mountain Meteor-
ology. Orographic Effects: Lee Cyclogenesis. Vorticity.
Further Reading
Baines PG (1995) Topographic Effects in Stratified Flows.
Cambridge: Cambridge University Press.
Chopra KP (1973) Atmospheric and oceanic flow problems
introduced by islands. Advances in Geophysics 16:
297-42 1.
Epifanio CC and Durran DR (2002)Lee-vortex formation in
free-slip stratified flow over ridges. Part I: Comparison of
weakly nonlinear inviscid theory and fully nonlinear
viscous simulations. Journal of the Atmospheric Sciences
59: 1153-1165.
Huerre P and Monkewitz AM (1990) Local and global
instabilities of spatially developing flows. Annual Review
Figure 11 Wake instability and vortex shedding in shallow-water of Fluid Mechanics 22: 473-537.
flow past an isolated obstacle. (A) Streamlines and (B) nondimen- Schar C and Smith RB (1993) Shallow-water flow past
sional vorticity al/(gD)’/’ (contour interval 0.25 with negative isolated topography. Part I: Vorticity production and
contours dashed and zero contour suppressed) of the initial state. wake formation. Journal ofthe Atmospheric Sciences 50:
The initial state was obtained through a numerical computation in 1401-1412.
which symmetry about the wake centerline was explicitly enforced, Smith RB (1989) Hydrostatic airflow over mountains.
thus inhibitingthe growth of antisymmetric disturbance modes. The Advances in Geophysics 31: 1-41.
symmetry condition was then removed, allowing the wake to
Smith RB and GrubiSi? V (1993) Aerial observations of
become unstable. (C) and (D) showthe subsequent evolution of the
wake in terms of the vorticity distribution at times (gD)‘/‘t/a = 36
Hawaii’s wake. ]ournu1 of the Atmospheric Sciences 50:
and (gD)”’t/a = 72. (Reprinted with permission of the American 3728-3750.
Meteorological Society, adapted from Schar C and Smith RB Smolarkiewicz PK and Rotunno R (1989) Low Froude
(1993) Shallow-water flow past isolated topography. I Part II: number flow past three-dimensional obstacles. Part I:
Transition to vortex shedding. Journalof the Atmospheric Sciences Baroclinically generated lee vortices. Journal of the
50: 1401-1412.) Atmospheric Sciences 46: 1154-1164.
LEE WAVES AND MOUNTAIN WAVES 1161
D R Durran, University of Washington, Seattle, WA, mountain wave, and such theory will be the subject
USA of the next section. Nevertheless, nonlinear effects do
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. exert an significant influence on the wave amplitude
and are essential to the dynamics of mountain wave
dissipation in regions of wave-breaking; such effects
Introduction will be considered later in this article.
Buoyancy perturbations develop when stably strati-
fied air ascends a mountain barrier. These perturbat- Linear Mountain Wave Theory
ions often trigger disturbances that propagate away
The strongest mountain waves are forced by long
from the mountain as gravity (or buoyancy) waves.
quasi-two-dimensional ridges that are sufficiently
Gravity waves triggered by the flow over a mountain
narrow that the dynamical influence of the Coriolis
are referred to as mountain waves or lee waves.
force can be neglected. The basic dynamics of these
Mountain waves sometimes reveal their presence
waves are largely captured by the linear theory for
through dramatic cloud formations, such as smooth
steady two-dimensional Boussinesq flow over an
lenticular clouds (see Figures 4 and 5 ) and ragged rotor
obstacle. The linearized momentum, thermodynamic,
clouds. Large-amplitude mountain waves can gener-
and continuity equations may be reduced to eqn [ 11for
ate regions of clear air turbulence that pose a hazard to
the vertical velocity w.
aviation. Large-amplitude mountain waves may also
produce very strong winds that blow down the lee
a2w a2w
slope of ridge-like topographic barriers (see Down- -ax2
+ - + e waz2= o PI
slope Winds).
What happens to mountain waves after they are Here x is the horizontal coordinate perpendicular to
generated? If the wave amplitude becomes large in the ridge-line, z is the vertical coordinate, and e is the
comparison to the vertical wavelength, the streamlines Scorer parameter, given by eqn [2], in which U ( z )is the
in a vertically propagating mountain wave steepen and speed of the basic-state flow and N ( z ) is the Brunt-
overturn in a manner roughly analogous to a breaking Vaisala frequency (or alternatively, the buoyancy
wave in the ocean. Such ‘convective’ overturning often frequency).
occurs as the waves enter the lower stratosphere,
where they encounter increased static stability and 1 d2U
decreasing horizontal wind speeds. The convective 1 N2 121
U2 U dz2
overturning of vertically propagating waves is also
promoted by the systematic decrease in atmospheric In the Boussinesq limit, the Brunt-Vaisala frequency
density with height. Those waves that do not break may be defined in terms of the basic state potential
down due to convective overturning before reaching temperature e ( z ) ,a constant reference potential tem-
the mesosphere are ultimately dissipated by the perature 0 0 , and the gravitational acceleration g, such
vertical transfer of infrared radiation between the that N2 = (g/Bo)dG/dz.
warm and cool regions within the wave and surround- Neglecting the effects of surface friction, the veloc-
ing atmosphere (radiative damping). ity perpendicular to the topography must vanish at the
Horizontal momentum is transported by mountain surface of the topography z = h ( x ) . This constraint
waves from the regions of wave dissipation to the provides a lower boundary condition for eqn [ 11, and
surface, where a net pressure force is exerted on the can be approximated to the same accuracy as the
topography. A decelerative force is exerted on the linearized governing equations as w ( x ,0) = U a h / a x .
large-scale atmospheric circulation in those regions The atmosphere has no distinct upper boundary, so the
where the wave undergoes dissipation. upper boundary condition is imposed in the limit
The basic structure of a mountain wave is deter- z + m. In order to assure the physical relevance of
mined by the size and shape of the mountain and by the mathematical solutions to eqn [ 11in the infinitely deep
$.a
vertical profiles of temperature, wind speed, and atmosphere, those solutions must satisfy one of two
moisture in the impinging flow. The overall character possible conditions: either (1)the perturbation energy
of the wave can often be predicted on the basis of linear density must approach zero as z -+ m, or ( 2 ) if
theory, in which the mountain is assumed to be small in the perturbation energy density is finite as z + 30,
comparison with the vertical wavelength of the then the perturbation energy flux associated with each
1162 LEE WAVES AND MOUNTAIN WAVES
individual vertically propagating mode must be up- frequency and k and rn are the horizontal and vertical
ward. The second condition allows the representation wavenumbers in an arbitrary wave of the form
of disturbances generated within the domain that R[ei(kx+mz-wt)I.
propagate energy upward to arbitrarily great heights,
but it prohibits downward-propagating modes from Nk
w=Uk& 171
radiating energy into the domain from infinity. (k2 + rn2)1/2
Constant Wind Speed and Stability, Sinusoidal Since by assumption U > 0, all steady waves (for
Ridges which w = 0) are associated with the negative root in
As a first example, consider flow in a horizontally eqn [7],and their vertical group velocities are given by
periodic domain in which h ( x ) = ho sin ( k x ) . The eqn 181.
lower boundary condition becomes w ( x ,0) =
Uhok cos kx, and solutions to eqn [ l ] subject to
this lower boundary condition may be written in
the form of eqn [3]. Upward group velocity and upward energy transport
w ( x , z ) = tZll(z)coskx+tZl2(z)sinkx [3] are obtained when k and rn have the same sign. Thus
when t > k , the solution to eqn [ 11satisfying the upper
Substituting [3] into [l], one obtains eqn [4]. and lower boundary conditions may be expressed as
eqn [91.
W(X,Z) = Uhok COS (kx + VZ) [91
Consider the simplest possible atmospheric structure The difference between these two wave structures is
in which N and U are constant with height. Without illustrated in Figure 1,which shows streamlines over a
loss of generality we will focus on the case in which series of sinusoidal ridges in a steady flow with N =
U > 0 and k > 0. Since N and U are constant t2 = 0.01 s-l and U = 15 m s-l. In the case in Figure 1A
N 2 / U 2is also constant. Defining v = (e2 - k2)'j2 and the topographic wavelength is 8 km and e2 < k2 (or
,u2 = -v2, the solution to eqn [4] may be written as [5], equivalently, Uk > N); the waves decay exponentially
where A, B , C, and D are constants to be determined with height, and the wave crests are aligned vertically.
by the upper and lower boundary conditions. In the case in Figure 1B the topographic wavelength is
40 km and e2 > k2(or U k N ) ; the waves propagate
vertically without loss of amplitude, and the wave
crests tilt upstream with height. The waves decay away
from the forcing when the intrinsic frequency exceeds
Note that the fundamental character of the solution the Brunt-Vaisala frequency (Uk > N ) because there
depends on the relative magnitudes of the Scorer is no way for buoyancy restoring forces to support
parameter and the horizontal wavenumber. oscillations at such high frequencies (see Buoyancy
If e < k , or equivalently, if the intrinsic frequency of and Buoyancy Waves: Optical Observations; Theory).
the wave Uk is greater than N, solutions to [4] either On the other hand, when the intrinsic frequency is less
grow or decay exponentially with height. Only the than the Brunt-Vaisala frequency, vertical propaga-
solution that decays with height is admitted by the tion occurs because buoyancy restoring forces can
upper boundary condition that the perturbation ener- support air-parcel oscillations along a path slanted off
gy density must approach zero as z --f co.The vertical the vertical at an angle 4 = cos-' ( U k / N ) .In steady
velocity satisfying [ l ] and the upper and lower mountain waves, 4 is the angle at which lines of
boundary conditions is given by eqn [6]. constant phase tilt off the vertical.
w ( x ,z ) = Uh0ke-P' cos kx [GI Isolated Mountain, Vertical Variations in N or U
On the other hand, if e > k, the solutions to [4] are The mountain wave solutions [6] and [9] are only valid
sinusoidal functions of z that neither amplify nor for air streams with constant basic-state wind speed
decay as z --t cm.The upper boundary condition then and stability flowing across an endless series of
requires that the perturbation energy flux in the wave sinusoidal ridges. If more realistic terrain profiles
be upward, or equivalently, that the group velocity (see and atmospheric structures are considered, other
Dynamic Meteorology: Waves) in the wave be directed linear solutions can be obtained that more strongly
upward. The dispersion relation for the time-depend- resemble observed mountain waves. In this section, we
ent generalization of [ l ] is given by [ 7 ] ,where w is the will describe how the wave response is influenced by
9, I I I
0
0 25 50 75 100 0 25 50 75 100
(A) Cross-ridge distance (km) (B) Cross-ridge distance (km)
Figure 1 Streamlines in steady airflow over an infiniteseriesof sinusoidal ridgeswhen N = 0.01 si,U = 15 m ssi, and the wavelength
of the topography is (A) 8 km (case Uk > N)or (B) 40 km (case U k i N).The flow is from left to right. The lowest streamlinecoincides with
the topography.
isolated topography and vertical variations in atmos- determined, the total vertical velocity w ( x ,z ) must
pheric wind speed and stability. be obtained by computing an inverse Fourier trans-
Suppose that the mountain profile consists of a form. The relative weight attached to each individual
single ridge from which the terrain elevation drops to wavenumber in the composite solution is determined
some reference level at all distances sufficiently far by the Fourier transform of the mountain.
upstream and downstream. Just as Fourier series can Streamlines for steady linear flow over an isolated
be used to represent a wide variety of periodic ridge of the form given by eqn [ 121 are shown in Figure
functions with an infinite sum of sines and cosines, 2A for the case N = 0 . 0 1 0 4 7 ~ - ~U, = lOms-l, and
the isolated mountain can, under rather general Nho/U = 0.6.
conditions, be constructed from periodic functions
by the use of Fourier transforms. Let &(k, z ) denote the [I21
Fourier transform p f w(x,z ) with respect to the x
coordinate, and let h ( k )be the Fourier transform of the In this case N u / U x 10 and the dominant horizontal
topography h(x). wavenumbers in the Fourier transform of the topo-
The kth component of the Fourier-transformed graphy satisfy k2 << .E2, which eliminates the depend-
vertical velocity &(k, z ) must satisfy the Fourier ence of the vertical structure on the horizontal
transform of the governing equation [l], expressed wavenumber in [ l l ] . As a result, all modes associated
by eqn [lo], which has the same form as [4]. with these dominant wavenumbers have approxi-
mately the same vertical wavelength (2n:U/N =
a2w ( e 2 - k 2 ) w = 0
-+
6 km), so the streamline at 6 km approximately
a22
reproduces the mountain profile while those at 3 and
9 k m are roughly the mirror image of the topo-
The lower boundary condition transforms to graphy. The solution shown in Figure 2A is com-
w ( k ,0) = iUkhoh. When N and U are constant, puted numerically without making the hydrostatic
the solution to eqn [IO], subject to the appropriate assumption and is very similar to that which would be
upper and lower boundary conditions, is given by eqn obtained in the hydrostatic limit, in which all hori-
[Ill. zontal wavenumbers have exactly the same vertical
wavelength and the mountain profile is exactly repro-
t ~ ( k , z=) i k U L ( k ) exp [ i ( t 2- k2)l”z], K > o [111 duced by the streamline originating at the 6 km level
upstream.
Equation [ll]is just the complex analog of [5]; each As suggested by Figure 2A, when an infinitely long
Fourier component &(k, z ) of the transformed vertical ridge is sufficiently wide that the flow is approximately
velocity is identical to the tzti forced by an infinite series hydrostatic ( N u / U >> l),but still narrow enough that
of sinusoidal ridges having wavenumber k and ampli- Coriolis forces can be neglected (Ifia/U<< 1,where f is
tude h ( k ) . The solutions obtained in the preceding the Coriolis parameter), energetic mountain waves are
section are therefore also applicable to the case of found only in the region directly above the mountain.
isolated topography. The only complication arises In the general case, the absence of wave energy in the
from the requirement that after the k ( k , z ) are region downstream can be deduced from the horizon-
0 25 50 75 100 0 25 50 75 100
Cross-ridge distance (km) (B) Cross-ridge distance (km)
Figure 2 Streamlines in steady air flow over an isolated mountain as predicted by linear theory when (A) a = 10 km, N is constant, and
Nho/U = 0.6;(B)a = 5 km, Nisconstantthroughouteachoftwolayerssuchthatbetweenthesurfaceand3kmN~holU = 0.6,andabove
3 km Nuho/U = 0.24.
tal group velocity for a steady two-dimensional wave Inequality [14] states that the difference in wave
in a mean flow with U > 0. This is expressed by eqn propagation characteristics in the two layers must
[13], which approaches zero in the hydrostatic limit exceed a certain threshold before the waves can be
(for which k 2 / m 2 -+ 0). trapped. The horizontal wavenumber of any resonant
lee wave in the two-layer system satisfies t~> k > tu,
a a - Nk2 implying that the wave propagates vertically in the
ak - ( k 2 + m2)3/2 1131
lower layer and decays exponentially with height in
the upper layer. As shown in Figure 2B, trapped waves
A sufficient decrease in the width of the mountain will have no tilt, even though they can propagate vertically
lead to the generation of nonhydrostatic waves with in the lower layer. The reason for this is that wave
downstream group velocities, but large-amplitude lee energy is repeatedly reflected, without loss of ampli-
wave trains, such as those shown in Figure 2B, do not tude, from the upper layer and the flat ground
occur unless there are significant vertical variations in downstream from the mountain. As a result, the
the wind speed and static stability. downstream disturbance is the superposition of equal-
If the vertical variations in U and N are such that the amplitude upward and downward propagating
Scorer parameter decreases significantly with height, a waves, a combination which has no tilt.
cross-topographic flow may generate a qualitatively
different type of wave, the trapped lee wave. A series of
trapped lee waves (also known as resonant lee waves) Nonlinear Mountain Waves
are apparent extending downstream from the ridge
Now suppose that the mountain height is not small
throughout the layer 0 5 z 5 4 k m in Figure 2B; a
compared to the vertical wavelength of the mountain
vertically propagating wave is also visible directly
wave. If N and U are constant, the streamline
above the mountain. The streamlines shown in Figure
displacement 6(x,z ) in steady two-dimensional Bous-
2B are for the linear solution to the same problem
sinesq flow over such a ridge is still governed by a
considered in Figure 2A, except that a = 5 km and
relatively simple mathematical model known as
the static stability above 3 km is reduced by a factor of
Long’s equation (eqn [IS]).
0.4. (The Brunt-Vaisala frequencies in the upper
and lower layers are thus NU = 0.004 188 s-l and
NL = 0 . 0 1 0 4 7 ~ - respectively.)
~,
A necessary condition for the existence of trapped
waves in the two-layer problem is expressed by the
inequality [14], where tu and t~ are the Scorer Although Long’s equation is a linear partial differen-
parameters in the upper and lower layers, and H is tial equation, it may be derived from the fully
the depth of the lower layer. nonlinear equations without making any linearization
or small-amplitude assumptions. Nevertheless, eqn
[15] may also be derived by assuming the mountain is
infinitesimally high and linearizing the governing
equations in the usual manner. When N and U are shortcoming of linear theory is that it cannot capture
constant, the only difference between the linear and the tendency of the nonlinear dynamics to enhance the
nonlinear solutions arises from the lower boundary short-wavelength Fourier components in the low-level
condition, which requires 6[x,h ( x ) ] = h ( x ) in the wave field over the lee slope. The nonlinear enhance-
exact finite-amplitude case and is approximated by ment of these short-wavelength flow perturbations
6(x,0) = h ( x ) in the small-amplitude limit. often produces more forcing at the wavelength of the
As one might guess from the similarities in the resonant lee waves than does the direct forcing by the
governing equations, when N and U are constant the topographic profile itself.
influence of nonlinear dynamics on the wave structure Clouds that form in regions of net upward displace-
is often relatively minor. This similarity can be ment in vertically propagating hydrostatic waves may
appreciated by comparing the linear solution in Figure appear like the cloud in Figure 4. The large single
2Awith the corresponding nonlinear solution in Figure region of cloudiness parallel to the mountain crest is
3A, both of which show streamlines in a Boussinesq probably formed by air parcel displacements qualita-
flow for which Nho/U = 0.6. Nonlinear processes tively similar to those in the streamline originating
steepen the streamlines around z = 4.5 km, which is 3/ near the 6 km level in Figure 3A. Clouds that form in
4 of a vertical wavelength (3&/4) above the topogra- trapped lee waves may appear as a series of long bands
phy. Conversely, the nonlinear waves are less steep parallel to the generating ridge. Such bands are often
than their linear counterparts near z = 1.5 km, which visible in satellite photographs and are formed by
is &/4 above the mean height of the topography. streamline patterns qualitatively similar to those
Despite these modest differences in the shape of the originating in the layer between 2 and 4 km in Figure
streamlines in the linear and nonlinear waves, the wave 3B. Nevertheless, three-dimensional variations in the
amplitude is almost identical in both cases. Nonlinear upstream topography often break these bands into the
processes do not have a dramatic impact on the waves superposition of many lens-shaped cloud masses, such
forced by flow over a infinitely long ridge unless either those shown in Figure 5.
(1)there are vertical variations in N and U or (2) the Returning to the discussion of how nonlinear
mountain is high enough to force wave overturning. dynamics modify the structure of mountain waves,
The influence of nonlinear wave dynamics on the consider the influence of wave breaking on the flow.
flow in the two-layer atmosphere previously consid- Two examples in which the wave amplitude becomes
ered in connection with Figure 2B is shown in Figure large enough to overturn are shown in Figure 6 . The
3B. The amplitude of the lee waves in the nonlinear case shown Figure 6A is one with constant N and U
solution is much larger than that in the linear solution, identical to that in Figure 3A, except that the mountain
and in the nonlinear case some spatial variation is height is increased so that Nho/ U = 1.2. (The vertical
visible among the individual troughs and crests in the scale also extends to z = 15 km.) Wave overturning
region 65 5 x 5 100 km. As suggested by this exam- first begins at the 3&/4 level, which is the same level at
ple, and demonstrated in several observational cam- which the wave faces appear to be steepened in Figure
paigns and numerical studies, linear theory does not 3A. As the wave begins to overturn, a &/2 deep region
reliably predict the amplitude of trapped lee waves of well-mixed stagnant fluid develops over the lee
generated by finite-amplitude mountains. The main slope and begins to extend downstream. A second
- 6 - -
E
Y
Em -
.-a
= 3 -
.
0
0 25 50 75 100 0 25 50 75 100
(A) Cross-ridge distance (km) (B) Cross-ridge distance (km)
Figure 3 As in Figure 2, except that the streamlines are for a fully nonlinear flow as computed using a numerical model. The trapped
waves in panel (6)are not completely steady; the solution is shown a nondimensionaltime U t / a = 20 after starting the flow from rest.
Figure 4 Single lenticular cloud over Laguna Verde, Bolivia. This cloud was probably formed by avertically propagating mountain wave.
(Copyright Bernhard Muhr, www.wolkenatlas.de.)
region of wave overturning eventually develops at a speed (10 m s - '), and the low-level stability
height of 7&/4, although the perturbations are weak- ( 0 . 0 1 0 4 7 ~ - ~are) identical to those for the case in
er at this level due to the dissipation experienced by Figure 6A, but the wind speed increases linearly to
the wave as it propagates through the first wave- 2 5 m s - 1 at a height of 9km. The presence of a
breaking level. Figure 6 shows the solution at a stratosphere is modeled by increasing N to 0.02 s - '
nondimensional time ( U t / a ) of 30, by which time above 9km and a linearly decreasing U back to
the near-mountain solution is quasi-steady, but the 10 m s- at z = 1 3 km. The wind speed is a constant
layers of well-mixed fluid continue to expand further lOms-' above 13km. The increase in the cross-
downstream. Also shown are contours of the sub-grid- mountain wind with height throughout the tropo-
scale eddy diffusivity. Regions in which the sub-grid- sphere decreases the local value of the nonlinearity
scale diffusivity is large are regions in which the parameter N ( z ) h o / U ( z )to a minimum just below the
numerical model has diagnosed the present of vigor- tropopause at z = 9 km. Above the tropopause N x
ous small-scale turbulence such as that which occurs ( z ) h o / U ( z )increases rapidly with height due to the
due to wave breaking. factor of two increase in N and the reversal of the wind
Although the breaking of mountain waves in an shear. As evident in Figure 6B, these more realistic
atmosphere with constant N and U has received a vertical variations in the upstream flow are sufficient
great deal of theoretical attention, the morphology of to focus the primary region of wave-breaking in the
such flows is not representative of most real-world lower stratosphere (around z = 12 km) and to prevent
wave-breaking events, in which the wave structure is wave-breaking in the troposphere.
significantly modified by vertical wind shear in the The influence of wave-breaking is highly nonlocal.
upstream flow. Those ridges that run north-south in In the case with constant N and U , the entire lee-side
the middle latitudes are oriented perpendicular to the flow in the wave-breaking regime (Figure 6A) is
climatological westerly flow and are frequent gener- dramatically different from that in the nonbreaking
ators of large-amplitude mountain waves. A proto- regime (Figure 3A). In particular, the surface winds
typical example of the mountain waves generated above the lee slope are significantly enhanced in the
by such ridges in a deep westerly flow is shown in wave-breaking regime (see Downslope Winds). The
Figure 6B. The mountain profile, the surface wind breaking waves in Figure 6B also exert a nontrivial
Figure 5 Multiple lenticularclouds over Mhatin, Icelandformed by trapped lee waves. (CopyrightGeorg Muller,www.wolkenatlas.de.)
influence on the low-level flow, although this influence upstream of the ridge crest and a region of low pressure
is considerably less dramatic than that which develops appears in the lee. The distribution of these pressure
as a consequence of wave-breaking in Figure 6A. perturbations is revealed by the along-flow variation
in the spacing between the two lowest streamlines in
Figures l B , 2A, 3, and 6 . The asymmetry in the
Vertical Momentum Transport pressure distribution across the ridge gives rise to a net
When air flowing over a mountain generates vertically pressure force on the topography that tends to
propagating waves, a region of high pressure develops accelerate the topography in the direction of the
15
12
E 9
E
P)
'5 6
I
0
0 25 50 75 100 0 25 50 75 100
(A) Cross-ridge distance (km) (6) Cross-ridge distance (km)
Figure 6 Streamlines (solid) and contours of the sub-grid-scaleeddy diffusivity (dot-dashed,at intervals of 20 m2s-') for (A) the case,
shown in Figure 3A, except that Nho/U = 1.2 and the vertical scale extends to 15 km. (6) As in (A), except with westerly wind shear
throughout the tropopause and a realistic stratosphere (see text).
mean flow. An equal and opposite force is exerted on one obtains eqn [19].
the mean flow by the topography.
To see how the topographically induced decelerative a (pou’w’)
apoc - - - -
forcing is distributed throughout the fluid, consider at az
the horizontal momentum equation [16], in which v is
the total velocity vector, p is the pressure, p is the A decelerative forcing will therefore be exerted on the
density, and i is the unit vector along the x coordinate, flow in those regions in which the mountain-wave-
and u = v ’ i. induced momentum flux is divergent, i.e., where
a(poU”)/az > 0.
-a +
PUV.(puv+pi)=O The vertical profile of the momentum flux is
at particularly easy to describe for steady, inviscid,
Integrate the preceding throughout the volume be- small-amplitude waves in a periodic domain (or in
tween the surface h(x) and an arbitrary level zt; use the an unbounded domain in which the waves decay as
divergence theorem; note that there is no advective x -+ hm).The cross-mountain pressure drag in such
momentum flux through the lower boundary; and waves is identical to the vertical momentum flux at
assume that the domain is periodic in the horizontal z = 0, as may been seen from the steady state version
direction. Then eqn [17] is obtained. of [17] in the limit zt -+ 0. Furthermore, a classic
theorem due to Eliassen and Palm states that under the
aat JJJ pu d v =- JJ puw dx dy 1 Z=Zt
preceding assumptions pou/w’is constant with height
except at a ‘critical level’ at which ii = 0. Mountain
waves are dissipated at the mean-state critical layers
found in real atmospheric flows. Mountain waves are
also dissipated through breaking and overturning if
they attain sufficiently large amplitude due to the
When vertically propagating mountain waves are
decrease in density with height or, as in Figure 6, if they
present, the cross-mountain pressure drag (given by
propagate into a region in which the local value of
the last term in eqn [17] must decelerate the volume-
averaged flow in the layer between the surface and zt
N/U increases significantly. Small-amplitude moun-
tain waves that propagate all the way to the meso-
unless the pressure drag is balanced by a downward
sphere without experiencing overturning are damped
transfer of momentum through level zt. This same
by infrared radiation.
result can be obtained for flow in nonperiodic domains
The Eliassen and Palm theorem implies that small-
under the assumption that the perturbation quantities
amplitude mountain waves transport a fraction of the
vanish at the lateral boundaries, although caution is
momentum of the cross-mountain flow downward to
advised when trying to apply eqn [ 171in a nonperiodic
the surface from those elevations at which the waves
domain because non-negligible mountain-wave-
undergo dissipation. There will be no vertical mo-
induced perturbations may extend far upstream and
mentum flux divergence and no forcing of the mean
downstream from a very long ridge.
flow within those layers of the atmosphere in which
The interaction between the mean flow and the
the waves are steady and nondissipative. The momen-
mountain-wave-induced momentum fluxes can be
tum fluxed downward by the waves is transferred to
described more precisely by separating the dynamical the topography by the cross-mountain pressure drag.
variables into an average over the domain (denoted by
Similar distributions of the vertical momentum flux
an overbar and taken as representative of the synoptic-
are obtained even when the waves are nonsteady and
scale flow impinging on the mountain) and the nonlinear. For example, the vertical momentum flux
perturbation about that average (denoted by a prime
profile associated with the finite amplitude waves
and assumed to represent the contributions from
shown in Figure 6B is approximately nondivergent
mountain waves generated by the flow over the ridge).
between the ground and the region of wave-breaking
The horizontal momentum equation for two-dimen-
in the layer 11 5 z 5 13 krn In contrast, the momen-
sional inviscid Boussinesq flow can be expressed by
tum flux profile is strongly divergent in the wave
eqn [18].
breaking region, and the mean flow is subject to a
significant decelerative forcing throughout this layer
(see Wave Mean-Flow Interaction). Unlike surface
friction, the drag associated with mountain waves is
If this is averaged over a periodic domain (or if it is typically exerted on the flow well above the lower
assumed that the perturbations vanish at the lateral boundary. Numerical experiments with general circu-
boundaries of a nonperiodic domain) and if w = 0, lation models suggest that mountain-wave-induced
LIDAR / Atmospheric Sounding Introduction 1169
drag plays a nontrivial role in the total momentum Durran DR (1986) Mountain waves. In: Ray PS (ed.)
budget of the atmosphere. Mesoscale Meteorology and Forecasting, pp. 472-492.
Boston: American Meteorological Society.
Eliassen A and Palm E (1960)On the transfer of energy in
See also stationary mountain waves. Geof. Publikasjoner 22:
1-23.
Buoyancy and BuoyancyWaves: Optical Observations; Gill AE (1982)Atmosphere-Ocean Dynamics. New York:
Theory. Downslope Winds. Dynamic Meteorology: Academic Press.
Waves. Lee Vortices. Wave Mean-Flow Interaction. Holton JR (1992) A n Introduction to Dynamic Meteoro-
logy, 3rd edn. San Diego: Academic Press.
Smith RB (1979) The influence of the mountains on
Further Reading the atmosphere. In: Saltzman B (ed.) Advances in
Baines PG (1995) Topographic Effects in Stratified Flows. Geophysics, vol. 21, pp. 87-230. New York: Academic
Cambridge: Cambridge University Press. Press.
Contents
Atmospheric Sounding Introduction

Backscatter
DIAL
Doppler
Raman
Resonance
light in the ultraviolet, visible, and infrared, which in

Atmospheric Sounding modern lidar systems is generated by lasers. The
different wavelengths used by radar and lidar leads to
Introduction the very different forms the actual instruments take.
The range of atmospheric parameters measurable
P S Argall and R J Sica, The University of Western with lidar includes temperature, wind velocity, atomic
Ontario, London, Ontario, Canada and molecular species concentration, and aerosol and
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. cloud properties.
In addition to its atmospheric applications, lidar is
also used in ocean research and military applications,
Introduction
including the detection of chemical and biological
Lidar (an acronym for light detection and ranging) is a agents and the remote identification and tracking of
remote sensing technique used predominately for vehicles. Lidar-equipped binoculars are used by
measuring atmospheric parameters, such as tempera- hunters and golfers as they provide accurate range
ture, composition and wind. Lidar operates on the measurements.
same principle as radar; in fact, it is sometimes called
laser-radar. Both these techniques operate by trans-
mitting a beam of electromagnetic radiation and
subsequently detecting any radiation scattered back
Evolution
to the instrument. The scattered radiation is analyzed The principle of lidar was first proposed in 1930. The
in order to determine some property or properties of original proposal suggested the measurement of
the medium through which the radiation traveled. atmospheric density profiles by the detection of
Lidar and radar differ in the wavelength of the scattering from a beam of light projected into the
radiation utilized. Radar uses wavelengths longer atmosphere. This proposed scheme suggested an
than about 1cm, in the radio band, while lidar uses antiaircraft searchlight as the source of the beam and
LIDAR / Atmospheric Sounding Introduction 1169
drag plays a nontrivial role in the total momentum Durran DR (1986) Mountain waves. In: Ray PS (ed.)
budget of the atmosphere. Mesoscale Meteorology and Forecasting, pp. 472-492.
Eliassen A and Palm E (1960)On the transfer of energy in
See also stationary mountain waves. Geof. Publikasjoner 22:
1-23.
Buoyancy and BuoyancyWaves: Optical Observations; Gill AE (1982)Atmosphere-Ocean Dynamics. New York:
Theory. Downslope Winds. Dynamic Meteorology: Academic Press.
Waves. Lee Vortices. Wave Mean-Flow Interaction. Holton JR (1992) A n Introduction to Dynamic Meteoro-
logy, 3rd edn. San Diego: Academic Press.
Smith RB (1979) The influence of the mountains on
Further Reading the atmosphere. In: Saltzman B (ed.) Advances in
Baines PG (1995) Topographic Effects in Stratified Flows. Geophysics, vol. 21, pp. 87-230. New York: Academic
Cambridge: Cambridge University Press. Press.
Contents
Atmospheric Sounding Introduction

Backscatter
DIAL
Doppler
Raman
Resonance
light in the ultraviolet, visible, and infrared, which in

Atmospheric Sounding modern lidar systems is generated by lasers. The
different wavelengths used by radar and lidar leads to
Introduction the very different forms the actual instruments take.
The range of atmospheric parameters measurable
P S Argall and R J Sica, The University of Western with lidar includes temperature, wind velocity, atomic
Ontario, London, Ontario, Canada and molecular species concentration, and aerosol and
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. cloud properties.
In addition to its atmospheric applications, lidar is
also used in ocean research and military applications,
Introduction
including the detection of chemical and biological
Lidar (an acronym for light detection and ranging) is a agents and the remote identification and tracking of
remote sensing technique used predominately for vehicles. Lidar-equipped binoculars are used by
measuring atmospheric parameters, such as tempera- hunters and golfers as they provide accurate range
ture, composition and wind. Lidar operates on the measurements.
same principle as radar; in fact, it is sometimes called
laser-radar. Both these techniques operate by trans-
mitting a beam of electromagnetic radiation and
subsequently detecting any radiation scattered back
Evolution
to the instrument. The scattered radiation is analyzed The principle of lidar was first proposed in 1930. The
in order to determine some property or properties of original proposal suggested the measurement of
the medium through which the radiation traveled. atmospheric density profiles by the detection of
Lidar and radar differ in the wavelength of the scattering from a beam of light projected into the
radiation utilized. Radar uses wavelengths longer atmosphere. This proposed scheme suggested an
than about 1cm, in the radio band, while lidar uses antiaircraft searchlight as the source of the beam and
1170 LIDAR / Atmospheric Sounding Introduction
a distant large telescope for the receiver. In this

configuration, now known as bistatic, the range of
the scattering can be determined by geometry. In the
bistatic configuration, shown in Figure 1, the field of
view of the receiver is scanned along the transmitted
beam in order to obtain an altitude profile of the
scattered light.
The first results obtained using this principle were
reported in the late 1930s when photographic record-
ings of light scattered from a searchlight beam were
made.
Typically, modern lidar systems are monostatic in
configuration, with the transmitter and receiver co-
located. Monostatic systems can be subdivided into
two categories: coaxial systems, where the laser beam
is transmitted coaxially with the receiver’s field of
view, and biaxial systems, where the transmitter and
receiver are located adjacent to each other. Monostatic Figure 2 Schematic illustrating the process of ranging based on
timing the returned signal.
lidar systems use pulsed light sources, thereby enabling
the range at which scattering occurs to be deter-
mined from the round-trip time of the scattered light detection and recording systems. Figure 3 is a block
(Figure 2). diagram of a generic lidar system, which shows how
By the early 1950s, refinements in technique and these subsystems work together to form a complete
improved instrumentation, including electrical re- lidar.
cording of the intensity of the backscattered light,
allowed the measurement of atmospheric density Transmitter
profiles up to altitudes of around 67km. These The transmitter generates light pulses with the re-
measured density profiles were then used to derive quired properties and directs them into the atmos-
temperature profiles using the Rayleigh-lidar tech- phere. Pulsed lasers, with their inherently low
nique, which is described later. divergence, narrow spectral width, and short, intense
The invention of the laser in 1960 and the giant pulses are ideal as the light sources for lidar systems.
pulse, or Q-switched, laser in 1962 provided a In addition to a laser, the transmitter of a lidar often
powerful new light source for lidar. The first use of a includes a beam expander, whose purpose is to reduce
laser in a lidar system was reported in the early 1960s the divergence of the beam being transmitted into the
and since then developments in lidar have been linked atmosphere. This allows a reduction in the back-
closely to advances in laser technology. ground measured by the lidar. At night, the back-
ground is due to light from the Moon, stars, airglow,
and artificial lights. During the day, background is
Instrument Basics predominately due to the Sun. Background can enter
Lidar hardware can be conveniently divided into three the lidar receiver either directly or after scattering in
subsystems: the transmitter, the receiver, and the the atmosphere. A reduction in the divergence of the
Backscanered
expander
Transmitter
! Electrical
: recording 6
Optical to electrical i
transducer j
: system
Detector
.......................................................
Figure 1 Three possible alignment arrangements of a lidar’s
transmitted beam and receiver field of view. Figure 3 Schematic of a generic lidar.
LIDAR /Atmospheric Sounding Introduction 1171
transmitted beam allows the field of view of the schemes are employed in Doppler and high-spectral-
receiver to be reduced, resulting in a lower back- resolution lidar systems.
ground. Signal separation based on polarization is a tech-
The narrow spectral width of the laser has been used nique often used in the study of atmospheric aerosols.
to advantage in a variety of ways in lidar systems. It Information on aerosol properties can be obtained
allows the spectral filtering of light by the lidar from the degree to which light scattered from a
receiver. A bandpass filter tuned to the laser wave- polarized laser beam is depolarized.
length selectively transmits photons backscattered Processing of the backscattered light based on range
from the laser beam, while rejecting photons at other can be performed in order to protect the detector from
wavelengths, thereby enabling a reduction in the the intense near-field returns of high-power lidar
background by several orders of magnitude. The pulse systems. This protection is achieved by using a fast
properties of pulsed lasers allow ranging to be shutter that closes the optical path to the detector
achieved by timing the backscattered signal, thus while the laser is firing and for a short time afterward.
allowing the simpler monostatic configuration. The shutter opens again in time to allow transmission
The major influence on the type of laser used in a of light backscattered from the altitude range being
lidar is the parameters the lidar is being designed to studied.
measure. Some measurements require a very specific
wavelength and/or tunability, i.e. resonance-fluores-
Detection and Recording
cence and differential-absorption lidar (DIAL).These
types of lidars can require complex laser systems to The signal detection and recording section of a lidar
produce the required wavelengths, while other simpler takes light from the receiver and produces a permanent
lidars, such as Rayleigh, Raman, and aerosol lidars, record of the measured intensity as a function of
can operate over a wide wavelength range. Although it altitude. In the first lidar systems the detection and
may be possible to specify the exact performance recording system comprised a camera and photo-
characteristics of the laser required of a particular lidar graphic film. Today detection and recording is
measurement, these characteristics often need to be achieved electronically. The detector is a device that
compromised in order to select from the types of lasers converts light into an electrical signal and the recorder
available. is an electronic device, often involving a microcom-
puter, which processes and records this electrical
signal.
Photomultiplier tubes (PMTs) are devices used as
Receiver
detectors for incoherent lidar systems working in the
The receiver of a lidar collects and processes the visible and ultraviolet. PMTs convert an incident
scattered laser light before directing it onto the photon into an electrical current pulse large enough to
detector. The first optical component, the primary be detected by sensitive electronics. Other devices that
optic in the receiver usually has a large diameter, are less commonly used as detectors in lidar systems
enabling it to collect a large amount of the scattered include multianode PMTs, micro-channel-plates
laser light. (MCPs), and avalanche photodiodes.
Lidar systems typically utilize primary optics with There are two ways the output of a PMT can be
diameters ranging from about 10cm up to a few recorded electronically; the pulses can be counted
meters in diameter. Optics at the smaller end of this individually (photon counting) or the average current
scale are used in lidar systems that are designed to due to the pulses can be measured and recorded
work at close range - a few hundred meters - and may (analog recording). Which method is the more appro-
be lenses or mirrors. Optics at the larger end of this priate depends on the rate at which the PMT produces
range are used in systems designed to probe the middle output pulses, which is proportional to the intensity of
and upper atmosphere and are typically mirrors. the light incident on the PMT. If the average time
After collection by the primary optic, light is usually between PMT output pulses is much less that the
processed in some way before being directed to the average pulse width, then individual pulses can be
detector system. Processing can be based on wave- easily identified and photon counting is the more
length, polarization, and/or range, depending on the appropriate recording method. However, if the aver-
purpose for which the lidar has been designed. age time between PMT output pulses is close to, or
As described previously, the simplest form of greater than, the average pulse width, then it becomes
processing based on wavelength is the use of a impossible to distinguish overlapping pulses, and so
narrow-band interference filter to reduce the back- analog recording becomes the more appropriate
ground. Much more sophisticated spectral filtering method.
1 172 LIDAR /Atmospheric Sounding Introduction
Coherent Detection transmission of the atmosphere along the laser path;

(doi/dR) (i is)the backscatter cross-section for scat-
There is a class of lidar systems that determine wind
tering of type i; and Ni(r) is the number density of
speed by measuring the Doppler shift of backscattered
scattering centers, which cause scattering of type i.
light. There are two ways these measurements can be
The general form of the lidar equation, as expressed
achieved, namely incoherent and coherent detection.
in eqn [l], can usually be greatly simplified when
Incoherent systems measure the wavelength of the
applied to a particular lidar system.
transmitted and received light independently, using
a spectrometer, and determine the Doppler shift
from these two measurements. Coherent detection Rayleigh Lidar
systems use a local oscillator, a narrow-band contin-
uous-wave laser, to set the frequency of the transmitted Rayleigh lidar is the name given to the class of lidar
pulses. Systems incorporating coherent detection use a systems that measure the intensity of light backscatter
local oscillator on a photomixer. This arrangement by molecules from altitudes between about 30 and
results in the output of the photomixer being a radio- 100 km. The intensity profiles measured by Rayleigh
frequency (RF) signal whose frequency is the differ- lidars are used to calculate relative density profiles,
ence of the frequencies of the local oscillator and the which are in turn used to calculate absolute temper-
backscattered light. Standard RF techniques are then ature profiles. The terms Rayleigh scattering and
used to measure and record this RF signal. The molecular scattering are often used interchangeably,
measured RF signal is used to determine the Doppler as are the terms Mie scattering and aerosol scattering.
shift of the backscattered light and thus the wind Rayleigh theory named after its founder, Lord Ray-
speed. leigh, describes the scattering of light by molecules
that are small compared with the wavelength of the
incident radiation; Mie theory describes scattering by
The Lidar Equation aerosols that are not small compared with the wave-
length, so there is a strong connection between these
The lidar equation is used to determine the number of
two pairs of terms.
photons detected by a lidar system. The lidar equation
Rayleigh scattering explains the color, intensity
takes into account both instrumental parameters and
distribution, and polarization of the blue sky in terms
geophysical variables. The general form of the lidar
of scattering by atmospheric molecules. For objects
equation includes all forms of scattering and it can be
with dimensions greater than about 0.003 times the
used to calculate the signal strength for any lidar.
incident wavelength, the more general Mie theory
The number of photons detected as pulses at the
must be used to calculate scattering effects.
photomultiplier output, per laser pulse, is
The Rayleigh backscatter (e = T C ) cross-section for
the atmosphere below 90 km can be expressed as
where the value of C is between about 4.75 x

and 5.00 x depending on the value used for
In eqn [ l ] A is the area of the telescope; PS(3,) is the index of refraction of air. Above 90 km altitude, the
convolution of P(A) andS(A),where P ( L ) is the number concentration of atomic oxygen becomes significant,
of photons emitted by the laser in a single laser pulse causing the refractive index of air to change, resulting
and S ( L ) is a function which takes into account any in eqn [2] becoming less accurate with increasing
wavelength shift during scattering, including Doppler altitude. The Rayleigh backscatter cross-section, eqn
and Raman shifts; AIb is the wavelength range for [2], can be used in conjunction with the lidar eqn [ 11 to
which P S ( 2 ) is nonzero; q ( A ) and q(A) are the optical determine the intensity of the backscatter that can be
transmission coefficients of the transmitter and re- expected for a particular Rayleigh lidar system.
ceiver optics respectively; Q(A) is the quantum effi- The Rayleigh lidar technique relies on the measured
ciency of the photomultiplier; r is the range and R1 and signal being proportional to the atmospheric density.
R2 are the minimum and maximum ranges for a range This is not the case in any region that contains
bin; ( ( 2 ) is the overlap factor which takes into account aerosols. From the surface to the top of the strato-
the intensity distribution across the laser beam and the spheric aerosol layer, about 25-30 km, the atmosphere
physical overlap of the transmitted laser beam and the contains a significant concentration of aerosols, thus
field of view of the receiver optics; z,( Y, 2) is the optical the Rayleigh technique cannot be directly applied to
this region. However, the atmosphere above this

altitude contains very few aerosols, allowing the
application of the Rayleigh technique.
The principle of operation of a Rayleigh lidar system -
E
is quite simple. A pulse of laser light is fired up into the Y
v
c
atmosphere, any photons backscattered and collected .-0
I
by the receiving system are counted as a function of 9

0
W
range. The lidar eqn [ l ] can be applied directly to a I
.-
S
Rayleigh lidar system to calculate the expected signal c
0
strength. For Rayleigh lidar a number of simplifica- r
m
I
tions can be made to eqn [l], allowing it to be v)
expressed as -2
W
n
W
U
1 .-c
Signal = K -N,( R )6R 131
I
3
R2
where K is a constant that includes all constant terms

from eqn [l],R is the range, 6R is the length of a range
bin, and N,(R) is the number density of air. Equation
Temperature error (K)
[3] shows that after correction for range a Rayleigh
lidar’s signal will be proportional to the atmospheric Figure 4 Propagation of the temperature error caused by a (A)
number density profile. 2%, (B) 5%, and (C) 10% error in the initial estimate of the pressure
for the Rayleigh temperature retrieval algorithm.
Due to the uncertainties in atmospheric transmis-
sion and instrumental parameters it is not possible to
determine the value of the constant K in eqn [3]
precisely enough to enable the determination of an duced by the seed pressure estimate are not easily
absolute density profile. The measured relative density quantified.
profile can be scaled to a model density profile to Above about 90 km, changes in composition of the
obtain a density profile that is well scaled. atmosphere cause the Rayleigh backscatter cross-
The relative density profile is integrated, using the section and the mean molecular mass to change with
hydrostatic equation, to determine a relative pressure altitude. These changes lead to errors in the temper-
profile. This integration requires an initial or seed atures derived using the Rayleigh lidar technique. For
pressure, usually chosen from a model atmosphere, to the current generation of Rayleigh lidar systems other
initiate the integration at the maximum altitude of the sources of error, statistical fluctuations and seeding
density profile. The pressure profile calculated in this error are generally larger than errors due to compo-
way has the same ratio to the actual pressure as the sition changes above 90 km. However, more powerful
relative density profile has to the actual density, i.e. Rayleigh lidar systems may ultimately be limited
their scaling factors are the same. in their maximum altitude extent by composition
An absolute temperature profile can be calculated changes.
by applying the ideal gas law to the relative density and While even the most technically advanced, ground-
pressure profiles. The application of the ideal gas law based, middle-atmosphere lidar systems need clear
divides the relative pressure by the relative density so skies to operate, the addition of Fabry-Perot etalons in
that their scaling factors, which are the same, cancel the receiver allows daytime measurements. This day-
out, resulting in an absolute temperature profile. time capability is technically complex and has been
The selection of the seed for the pressure integration implemented on only very few Rayleigh lidar systems.
may introduce an error into the calculated tempera-
ture profile. The magnitude of this error is propor-
Doppler Effects
tional to the difference between the actual pressure
and the seed pressure used. As the actual pressure is not The motion of air molecules has components due to
known, the resulting error in temperature is unknown. both random thermal motions and wind. When light is
However, the magnitude of this error reduces as scattered by a molecule it suffers a change in frequency
the calculation of temperature proceeds downward due to the Doppler effect. The magnitude and direc-
(Figure 4).Users of this technique are well advised to tion of the Doppler shift is determined by the compo-
ignore temperatures from at least the uppermost nent of the molecule’s velocity along the direction of
10 km of the retrieval, since the uncertainties intro- the lidar beam.
1174 LIDAR I Atmospheric Sounding Introduction
The random thermal motions of air cause backscat- and lower stratosphere. Aerosols and clouds are easily
tered laser light to be spectrally broadened. Using detected by elastic backscatter lidar; however, instru-
Maxwell’s velocity distribution function and the ments using multiple-wavelength transmitters and
Doppler equation, it can be shown that the broadening receivers and polarization techniques provide signif-
function is a Gaussian and has a temperature-depend- icantly more information on their properties.
ent width. In September of 1994, NASA flew a space shuttle
Wind, the average motion of air molecules, causes mission, STS-64, which included the Lidar In-Space
backscattered laser light to suffer a frequency shift Technology Experiment (LITE) instrument, the first
while maintaining its shape. The frequency shift is successful space-based lidar. LITE was used to meas-
directly proportional to the component of the wind ure tropospheric and stratospheric aerosols, clouds,
velocity in the direction of scattering, the radial wind and surface reflectance on a global scale.
velocity. Figure 5 shows how the spectrum of a Lidar systems can utilize the backscatter from
narrow-bandwidth laser is modified due to scattering aerosols to measure wind velocity. Light backscattered
by atmospheric molecules. from aerosols undergoes the same Doppler shift due to
Middle atmospheric winds can be determined by wind as light scattered back from molecules. However,
measuring the spectrum of backscattered light; how- the spectral broadening of the light backscattered from
ever, Rayleigh-Doppler temperature measurements aerosols is much narrower than that backscattered
are quite difficult, as the signal-to-noise requirements from molecules, owing to the difference between the
are much greater than those for wind velocity meas- masses of the two types of scatterers. The high signal
urement using this technique. level offered by scattering from aerosols in the lower
atmosphere allows the use of coherent detection for
the determination of wind velocity. Steerable lidars
Aerosol Lidar based using this technique are capable of making high-
The theory of scattering developed by Mie early in the resolution wind field maps.
last century is a general solution to the scattering of
electromagnetic radiation by a homogeneous sphere. Differentia I-Absorption Lida r (DIAL)
This early work has been extended to cover numerous The differential-absorption lidar (DIAL) technique is
other geometries and so provides a useful approxima- used for measuring the concentration of trace species
tion for scattering from atmospheric aerosols. in the atmosphere. The DIAL method relies on sharp
The influence of clouds and aerosols on the atmos- variations in optical transmission near an absorption
pheric energy budget is complex, as they scatter and line of an atmospheric constituent. A DIAL transmits
absorb both incoming solar and outgoing terrestrial two closely spaced wavelengths, one coinciding with
radiation. Since the early 1960s many lidar systems an absorption line of the constituent of interest and the
have been operated at various stations around the other in the wing of this absorption line. During the
world to study aerosols and clouds in the troposphere transmission of the two wavelengths through the
atmosphere, the emission tuned to the absorption line
will suffer greater attenuation than the emission in the
Intensity wing of the absorption line. The intensities of the two
Spectrum after wavelengths backscattered to the DIAL instrument
scattering by can then be used to determine the optical attenuation
owing to the constituent, and thus the concentration of
that constituent.
The DIAL technique has proven to be useful in
providing tropospheric measurements with good time
L a s r s i o n
and spatial resolution for a number of trace species,
including NO, H20, 0 3 , SOz, and CH4, as well as
stratospheric ozone measurements. DIAL allows
mapping and wide-area monitoring of industrial
effluents and pollution.
Wavelength
Figure 5 Doppler shift effects on Rayleigh scattering a narrow- Raman Lidar

line-width laser from atmospheric molecules. The broadening is
due to thermal motion and the shift is due to wind. The intensity of If monochromatic light, or light of sufficiently narrow
the two spectra are not to scale. spectral width, is scattered by gas or liquid then the
spectrum of the scattered light can be observed to have low abundance but very high resonant scattering
contain lines at wavelengths different from that of the cross-sections. Resonant scattering occurs when the
incident radiation. This effect was first observed by energy of an incident photon is equal to the energy of
Raman; it is due to the interaction of the radiation with an allowed transition within an atom. In this elastic
the quantized vibrational and rotational energy levels process, the atom absorbs a photon and instantly emits
of the scattering molecule. Raman scattering involves another photon at the same frequency. As resonant
a transfer of energy between the scattered light and the scattering involves an atomic transition between
molecule; it is therefore an inelastic process. As the allowed energy levels, the probability of this process
energy levels for each type of molecule are unique, so occurring is much greater than that for Rayleigh
the Raman spectrum is unique and provides a method scattering, leading to the much higher scattering cross-
of sensing a particular molecular species. sections. The resonant-scattering cross-section for
The term Raman lidar refers generally to a lidar that sodium at 589 nm is about times larger than the
utilizes light scattered by molecules that undergo a cross-section for Rayleigh scattering by air at the same
change in their vibrational quantum number. Meas- wavelength. As each species of alkali metal has a
urement of the intensity of the scattered Raman light unique absorption and, hence, resonant-scatter and
allows the calculation of the abundance of the molec- fluorescence spectrum, these may be used to identify
ular species. The selection of vibrational Raman lines and measure the concentration of each individual
can be achieved with high-quality narrow-band inter- species. Although most commonly applied to sodium,
ference filters. However, the blocking of such a filter resonance-fluorescence lidar has been applied to
must be made high enough for elastic backscatter from calcium (Ca and Ca+),potassium, lithium, and iron.
molecules and aerosols to be attenuated effectively. Sodium lidar systems are used to measure the
Owing to the small cross-sections for Raman scatter- abundance profiles of sodium at between 85 and
ing, Raman lidar is limited to molecules with a 105 km, with time resolution of tens of seconds and
relatively high abundance, such as water vapor and altitude resolution of a few hundred meters. Density
molecular nitrogen. Raman lidar is generally simpler perturbations due to wave motions are present in the
to implement than DIAL. sodium density profiles, enabling the determination of
Raman lidar is used predominately for the meas- wave parameters in this dynamically active region of
urement of atmospheric water vapor and temperature. the atmosphere to be determined. Spectral resolution
Raman molecular nitrogen profiles can be used to of resonance-fluorescence scattering from sodium
determine atmospheric temperature profiles, using the allows the determination of the temperature and
Rayleigh technique described above, even in regions wind. This technique, narrow-band resonance-fluo-
containing aerosols. Elastic scattering from aerosols rescence lidar, allows accurate, high-resolution tem-
can be separated effectively from the Raman nitrogen perature and wind measurements in the mesopause
backscatter by spectral filtering. The Raman nitrogen region.
signal is therefore approximately proportional to the
number density profile, although a correction must be
made for the optical attenuation of the atmosphere See also
due to both aerosols and molecules. Aerosols: Observations and Measurements. Lidar: Ra-
The pure rotational Raman spectrum (PRRS), man; Resonance.Optics, Atmospheric: Optical Remote
which is due to scattering involving a change in the Sensing Instruments. Radar: Incoherent Scatter Radar.
rotational quantum state only, is difficult to measure as
the spectral shift of the lines is quite small. The
separation of lines in the PRRS of Nz is about Further Reading
',
16 cm - while the first vibrational transition causes Frehlich R (1996) Coherent doppler lidar measurements of
a shift of about 2331 c m - l . The shape of the PPRS is winds. In Consortini A (ed.) Trends in Optics: Research,
temperature-dependent, allowing pure rotational Ra- Development and Applications, pp. 351-370. London:
man lidar to make atmospheric temperature measure- Academic Press.
ments. Grant WB (1995) Lidar for atmospheric and hydrospheric
studies. In Duarte JF (ed.) Tunable Laser Applications,
pp. 213-305. New York: Marcel Dekker.
Killinger DM and Mooradian A (eds) (1983) Optical and
Resonance-Fluorescence Lidar Laser Remote Sensing. New York: Springer.
Measures RM (1984)Laser Remote Sensing: Fundamentals
The constant ablation of meteors in the Earth's upper and Applications. New York: Wiley.
atmosphere leads to the existence of extended layers of Schreiber U and Werner C (eds) (1999) Laser Radar
alkali metals at altitudes around 90 km. These metals Ranging and Atmospheric Lidar Techniques 11 (Europto
1176 LIDAR I Backscatter
Series). Society of Photo-optical Instrumentation Engi- Application, pp. 369-396. Norwood, MA: Artech
neers. Bellingham, Washington, USA. House.
Sedlacek AJ and Fischer KW (eds.) (1999) Application Thomas L (1995)Lidar methods and applications. In Clark
of Lidar to Current Atmospheric Topics I I I (Pro- RJH and Hester RE (eds)Spectroscopy in Environmental
ceedings of SPIE, Vol. 3757). Society of Photo-optical Science, pp. 1-47. Chichester: Wiley.
Instrumentation Engineers. Bellingham, Washington, Weitkamp C (1996) Lidar measurements: atmospheric
USA. constituents, clouds, and ground reflection. In Raschke
Singh UN (1997) Lidar for atmospheric remote sensing. In E (ed.) Radiation and Water in the Climate System, pp.
Rastogi RK (ed.) Optical Measurement Techniques and 217-247. Berlin: Springer-Verlag.
Backscatter
C M R Platt, Colorado State University, Colorado, USA scattered in any direction forms a pattern that is
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. described by the single scattering phase function P(9)
where 9 is the angle between the scattered light and the
forward direction as shown in Figure 1. The amount
Introduction scattered by a particle is dependent on the diameter of
This article covers several aspects of lidar backscatter. the particle and its size compared with the wavelength
After a general introduction, various definitions per- of light. It is also dependent on whether the particle or
taining to lidar backscatter are described. This is molecule is absorbing as well as scattering. Thus,
followed by a description and explanation of the particles that are small compared with the wavelength
scattering and backscattering phase functions, includ- scatter less than if the scatter were determined solely
ing presentation of typical scattering phase functions by the particle cross-section, and the amount is
of molecules, water drops, and clouds. The polarized described by the scattering efficiency.
nature of the radiation and its treatment by a scatter- The efficiency of backscatter is very important for
ing matrix are then described briefly. The lidar lidar systems and is also related to the backscatter
equation is presented, together with a simple solution phase function P(n),the phase function at a scattering
in order to point out the importance of the backscatter angle of 180" from the forward direction.
phase function and its relation to the extinction to
backscatter ratio and its use in solving the equation. Backscatter Efficiency and
Examples of extinction to backscatter ratios of various
atmospheric constituents are presented. Several ex- Backscatter Coefficient
amples of profiles of measured atmospheric backscat- The scattering efficiency Q,,(%,Y) of an atmospheric
ter are described, including stratospheric aerosols, particle (molecule, aerosol, water drop, or ice crystal)
cirrus clouds, and depolarizing effects in midlevel ice, determines how much radiation is scattered in all
water, and mixed-phase clouds. The article does not directions by the particle. Here 2 is wavelength and r
cover inelastic backscatter such as Raman scattering particle dimension. Consider a uniform light beam of
and fluorescence. intensity I ( W m P 2 )incident on a particle of area of
Lidar is used to detect and profile certain constitu- cross-section A. If I,, is scattered, I , is absorbed and IO
ents in the atmosphere, such as molecules, aerosols, + +
passes straight through, then I = I,, I , IO.For
and clouds. The backscatter from such entities is visible lidar scatter on spherical water drops and
important in lidar because most lidar (laser radar) typical ice crystals the last two terms are close to zero.
systems are monostatic, that is, there is a telescope The scattering efficiency is defined as
receiver placed close to, or coaxial with, a laser pulse
transmitter. Pulses of light sent into the atmosphere are
scattered in all directions by molecules, aerosols, and
clouds, and a small amount scattered into the back Because of the nature of electromagnetic scattering
direction is returned to the receiver. The time taken for Q,,(i,,Y) can approach a value of 2 for non-absorbing
the laser pulse to return gives the range of the particles large compared with the wavelength A. This is
atmospheric volume being studied and the amplitude because diffraction occurs around and outside the
of the return is proportional to the volume density of edges of the particle, causing the effective cross-section
the atmospheric particles or molecules. The amount to be about 2A for large particles and less than 1A for
LIDAR I Backscatter 1177
ciency, as shown under ‘Scattering Phase Function’

below.
When lidar is observing the atmosphere, it is
observing the molecular atmosphere into which are
mixed aerosols and clouds in different proportions
and varying in both time and space. Of course,
aerosols and clouds - particularly clouds - tend to
form in well-defined layers, whereas gas molecules are
distributed throughout the atmosphere, the molecular
density depending on the local pressure and temper-
ature. Aerosols and clouds generally contain a range of
particle sizes with different number densities, defining
a smooth particle number size distribution. Of course,
the backscatter efficiency will vary also with particle
size. Lidar measures the returns within a volume of the
atmosphere and the volume backscatter coefficient
p(i) is defined as
p(2) = Sm
0
Qn(2,r)n(r)n$ dr 121
where n (r ) is the number of particles of dimension Y,

and in the case illustrated above the particles are
spheres (a water cloud, for example) of radius r.
For ice crystals, the dimensions and area depend on
the crystal type or habit. Consider hexagonal ice
crystals with length L and diameter D. Then an area
can be defined as LD and the backscatter coefficient as
p(2) = Qn(i,r)LD dL 131

0
We assume in eqn [3] that ice crystals are falling with

their long dimensions horizontal, which can be the
case if they are large enough, and that the lidar is
pointing vertically. If the ice crystals are tumbling then
an effective dimension has to be determined. The area
Figure 1 Schematic of backscattering at an atmospheric of cross-section of hexagonal crystals is actually a little
particle. more complicated than shown in eqn [3], which is
shown only as an example of how specifically the
cross-sectional area must be defined.
particles smaller than the wavelength, where scatter-

ing becomes less efficient. Scattering Phase Function
We can define backscatter efficiency Qn(J.,r ) as the The phase function P(9) defines the intensity of the
fraction of the total scattered radiation I,, that is radiation scattered by a particle at an angle 9 to the
scattered in a cone of unit solid angle around the back forward direction. The total scattered radiation nor-
direction of scattering. Because the integrated solid malized to 4n is then
angle over the scattering sphere is 471, if there is
uniform scattering over the whole sphere (isotropic 180
scattering), the backscatter efficiency is 1/4n, or P(9) cos9 sin 9 d 9 = 471 [41
0.0796. An alternative quantity is the radar backscat-
ter efficiency, which is 471 times the lidar efficiency, so The value of the backscatter phase function, together
that for isotropic scatter it has a value of 1. The with the scattering efficiency, will determine the
backscatter phase function, as well as the total backscatter efficiency. This can be either larger or
scattering efficiency, determines the backscatter effi- smaller than the isotropic backscatter efficiency of
0.0796, depending on whether the scattering phase

function is peaked or depressed near the backscatter-
ing direction.
We now consider some phase functions for a range
of atmospheric constituents and the corresponding
backscatter phase function and efficiency. The single
scattering phase function for an atmospheric molecule
is shown in Figure 2. As noted above, the phase
function denotes the fraction of the total radiation that
is scattered into a given angle to the incident direction.
Note that the scattering differs according to the
polarization of the radiation. For scattered radiation
in a plane containing the direction of polarization and
the vector of the incident radiation, denoted by 1, the
phase function falls to zero at scattering angle of 90".
For radiation perpendicular to the plane containing
the incident vector, denoted by Y, scattering is inde-
pendent of direction. The scattering direction in Figure
2 is depicted in a polar diagram where incident
radiation is entering from the left of the diagram. x = 17.14
The normalized scattering phase function for unpo- r = 1.5p-n
larized light is given by Figure 3 Scattering phase function for a water drop. The size
parameter and water droplet size are shown below. (Adapted with
3
P(9) = - (1+ cos2 9) [51
permission from Twomey S (1977) Atmospheric Aerosols. Amser-
16n dam: Elsevier.)
The backscatter phase function P(n) is then 6/16n which is not normalized, is seen to be small compared
(=0.119 per steradian). with the forward scattering. This is a characteristic of
For particles of molecular size, or of dimensions scattering on large particles, with some exceptions, as
small compared with the wavelength, Rayleigh scat- we will see. The phase function depicted here is the
tering theory is appropriate. For particle dimensions sum of the two directions of polarization (see under
approaching, or larger than, the wavelength of the 'Polarization Effects' below).
incident radiation, Mie scattering theory is appropri- Figure 4 shows typical scatter on an ice crystal that
ate. Figure 3 shows the polar scattering diagram for a has dimensions large compared with the wavelength.
water sphere, as might be found in a water cloud. The The presentation is different, with scattering angle on
diagram shows a succession of high and low values of the x-axis and phase function on the y-axis. Again one
scatter, caused by complex interference effects be- can see that the backscatter phase function is quite low,
tween penetrating and surface waves on the sphere. and much lower than that from a molecule. Of course,
The quantity r is the radius of the sphere and x is the the cross-sections of ice clouds are much greater per
size parameter, 2nr/A. The backscatter phase function, unit volume than for molecules, so that backscatter
intensity from ice clouds tends to be greater, although
this is not always the case. The crystal phase function is
calculated either from a complex geometrical theory
or by geometrical ray tracing.
What is relevant here to lidar backscatter is the large
differences in the backscatter phase function between
the various atmospheric components in Figures 2 to 4.
If p is the backscatter coefficient and is the volume
extinction coefficient, then
Figure 2 The scattering phase function for scattering from a

molecule - Rayleigh scattering. Labels are described in the text.
(Adapted with permission from Twomey S (1977) Atmospheric where Q, is the effective scattering efficiency of the
Aerosols. Amserdam: Elsevier.) volume-of particles and similarly for the backscatter
1 04
103
1 02
0 0.2 0.4 0.60.81.0 1.2 1.r
Q (degrees)
0 30 60 90 120 150 180

Angle of scatter, 0 (degrees)
Figure 4 Scattering phase function for an ice crystal; schematic representation from a number of experimental phase functions.
(Adapted with permission from Platt CMR et a/. (1981)Journal of the Atmospheric Sciences 38:156-167.)
efficiency. The quantity k is known as the backscatter- lasers used in lidar are usually linearly polarized to a
to-extinction ratio and it is clear that it is numerically high degree, and we have to consider the known
equal to the normalized backscatter phase function. A depolarizing effects in the backscatter from some
value of k is needed to solve the backscatter lidar atmospheric components. Molecules have no depo-
equation for volume attenuation of the lidar beam and larization and water spheres only very weak depolari-
is therefore a crucial quantity to lidar backscatter zation. Ice crystals, however, are found to depolarize
measurements. A popular quantity is S , defined as the by various amounts.
lidar ratio, which is simply the reciprocal of the Scattering of radiation is conveniently described by
backscatter-to-extinction ratio and therefore is meas- a scattering matrix, whose components can be meas-
ured in steradians. Typical ranges of values of S and k ured by optical instruments.
are shown in Table 1.Note that aerosol populations
also possess a range of values of these quantities. A
value k representing the isotropic, or radar, ratio has Table 1 Range of values of backscatter parameters k and S for
also been used. It has the value of 47cP(n). various atmospheric scatterers
Atmospheric constituent k (sr- ') s (sr)

Molecular 0.119 8.4
Polarization Effects Water cloud 0.047-0.050 20.0-21.3
Thus far we have considered the total radiation Ice cloud theoretical 0.008-0.072 14-1 25
Ice cloud experimental 0.011-0.036 28-91
backscattered by a particle without consideration of Aerosol 0.01 5-0.05 20-60
whether the light is polarized or not. In fact, pulsed
Here we consider a backscattering matrix for sum of the two polarization components in the
nonspherical particles as a general example (see Sassen parenthesis and o ( r ) is the volume extinction coeffi-
2000): cient of the layer within ranges r1 and r2. The power is
F ( 180) = diag[F11( n o ) , F22(l8O), calibrated against that from a molecular layer at a
different range. The lidar (eqn [9])can then be written:
F33 ( 18 01, F44 ( 18 011 171
This matrix is simplified as it applies to particles that P’(4 =P(4G2 1101
are randomly oriented in space and which contain a where
plane of symmetry, such as typical hexagonal crystals.
The medium is called macroscopically isotropic and T:2 = exp -2 llrio(r)dr
symmetric. If there is horizontal orientation of ice
crystals, which can occur, then more scattering
elements in the matrix need to be considered. The
= exp -2s ll 12
P(r) dr 1111
present treatment represents the present state of
progress in the field. and we consider P(r) as the sum of the two polarization
components.
In terms of the above matrix a depolarization ratio
The solution for P(r) in eqns [lo] and [ll]is then
A, defined as the ratio of the perpendicular to parallel
components, is given by
The crucial importance of the quantity S in the

solution is clear.
Now, for spheres, Fll(l8O) = F22(180), so that for
that case, A = 0, a well-known result, although there is
some residual depolarization.
\
Hexagonal crystals depolarize because the laser
radiation undergoes several internal reflections before 9
ending up in the back direction, so that (except for Experimental points
specular reflections at perpendicular surfaces) some
Theoretical molecular
rotation of the plane of the polarization vector occurs. scattering profile
Measured values of depolarization ratio are quite 1011
large in cirrus clouds, which allows some distinction in
cloud phase using a depolarization lidar. The use of
such a lidar in such observations is thus very desirable.
N
-
E
The Lidar Equation v
N
I
0
Lidar is used to measure the properties of atmospheric
layers through their backscattering and depolarization
properties. As the layers scatter the laser beam, then 10’0
they also attenuate the beam during passage through

that layer. Thus we must understand the retrieval of
the backscatter coefficient of a particular layer. Meas- I\
ured backscatter must be corrected for this attenua-
tion. It is apparent then as to how the extinction-to-
backscatter ratio S of a volume of particles is impor-
tant to the retrieval process. The lidar equation is
written in the first place in terms of the power P ( r ) 109 I I I I I
measured from a particular range r: 10 20 30 40 50
Height (km)
Figure 5 Typical backscatter profile (in units of range squared)

for the stratospheric aerosol layer. The backscatter for a model
molecular atmosphere is also shown. (Adapted with permission
where E is the pulse energy, A the telescope area, and c from Kent G et a/. (1967) Journal of Atmospheric and Terrestrial
the velocity of light. The backscatter coefficient P is the Physics 29: 169-1 81 .)
The solution is generally more complex, as there is eruptions. Such eruptions send a cloud of enhanced
the molecular atmosphere to consider. The above is an aerosol mass around the globe that is easily observed
example of how a volume backscatter coefficient can by lidar. Series of observations now exist covering
be recovered if we know the extinction-to-backscatter thirty years that show how peak backscatter in the
ratio S. Junge layer has varied over the years and how it
Table 1 shows values of k and S for various responds to volcanic activity. An example is shown in
atmospheric constituents from both experimental Figure 5 . The total backscatter is a mixture of aerosols
and theoretical data. and molecules. A model of molecular backscatter from
Backscatter lidar has also been used at infrared local aerological information is shown, and the
wavelengths where the absorption by both water and enhanced backscatter in the stratosphere is indicated
ice can be quite strong. In that case, the scattering clearly. The aerosol backscatter amplitude waxes and
efficiency and backscatter efficiency will be corre- wanes in strength over the years as various volcanic
spondingly less than for transparent particles. The clouds appear and then dissipate.
same is true for some aerosols, such as those composed The atmospheric boundary layer also contains
of soot where strong absorption of visible radiation copious amounts of aerosol as a capping temperature
can occur. Values of S will be correspondingly larger inversion impedes upward movement out of the layer.
than for the transparent-particle case. Such aerosol layers can be dense in large urban areas,
in regions of desert dust, and from forest fires, as
Measurements of Atmospheric examples.
Cirrus was also an early target of investigation
Backscatter because of the semitransparent nature of such clouds.
Early lidar observations were made of the strato- The Sun's disk is often seen hazily through even quite
spheric aerosol, which responds to large volcanic deep layers of cirrus. Lidar pulses of radiation can
1.o I ~ " " ~ " 1 '
Wavelength: 0.532 pm
0.8
0.6 Cirrus
z
az
8
m
Y
V
2 0.4
'0
Q
z
c
3 0.2
Mt Pinatubo
0.0
-0.2 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
10 15 20
Altitude (km)
Figure 6 Typical backscatter profile from a cirrus cloud, showing the strong backscatter and atmospheric attenuation. (Adapted with
permission from Platt CMR et a/. (1998) Journal of the Atmospheric Sciences 55: 1977-1 996.)
Depolarization ratio
0.0 0.2 0.4 0.6 0.8 1.o
- - - - - - - -- - -- ---==--
4.5 -
h
E a,
Y
_- - -
4.0 -- =---e*
-- -- 1614 LT
E, = 0.997
6, = 5.74
0 2 4 6 8
Attenuated backscatter coefficient (km-')
Figure 7 Examples of returns from a mixed phase cloud, showing various patterns of backscatter (full line) and depolarization ratio
(broken line). (Adapted with permission from Young SA et a/. (2000) Journal ofAppliedMeteorology39: 135-1 53.)
often penetrate through such layers with sufficient rapidly, indicating the close angle to horizontal at
photons returning to the receiver to be detectable. which the crystals fall. Such unusual and dramatic
Figure 6 shows a typical return from cirrus, the depth returns from the atmosphere show the strength of lidar
being rather typical for such clouds. Attenuation is backscatter in picking out details in cloud layers. The
fairly weak for this cloud. The figure also reveals layer at about 4.5 km is probably composed of ice,
typical structure with variations in backscatter with some supercooled water drops possibly also
with altitude. The Mt Pinatubo volcanic cloud present. The top layer is interesting because there
was strong at the time this profile was taken and appears to be strong attenuation and a pattern of
shows through the cirrus. When several profiles depolarization ratio that commences at cloud base
taken at successive time intervals are investigated, with a low value but increases steadily upwards. This
cirrus ice crystals falling out and being swept sideways happens to be characteristic of a layer of water drops.
by the wind are revealed. There is often also an Because these are relatively small and numerous,
indication of cirrus cloud base becoming progressively attenuation is quite strong. The depolarization pattern
lower with time. is a consequence of strong multiple scattering in the
Examples of linear depolarization ratio are shown beam, where the scattered photons near the back
in Figure 7. Here, layers of midlevel cloud exhibit very direction, or even at larger angles, can be scattered
variable characteristics. This is because layers of ice several times more, thus finding their way back into
crystals, water drops, and mixed-phase cloud can exist the telescope receiver beam. This process can, in a
separately in the atmosphere. The lidar backscatter is manner equivalent to crystal internal reflections,
shown as the full line and the depolarization ratio A as rotate the plane of polarization to give depolarization
the broken line. The bottom layer just above 4 km has in the back direction.
a value of A that is typical of ice clouds. The next layer The above examples are given to show the power of
has intense backscatter but very low values of depo- lidar backscatter to distinguish various layers of
larization. This is an interesting case that is fairly aerosols and clouds in the atmosphere.
common at these altitudes and temperatures. Between
atmospheric temperatures of - 10°C and -2O"C, ice
crystals are often hexagonal plates that float horizon- See also
tally through the air. They thus present a large area of Aerosols: Climatology of Tropospheric Aerosols; Obser-
cross-section but also mirror-like surfaces, which do vations and Measurements; Physics and Chemistry of
not depolarize in the back direction. If the lidar is tilted Aerosols; Role in Cloud Physics; Role in Radiative Trans-
a few degrees off the horizontal such returns disappear fer. Cloud Microphysics. Lidar: Atmospheric Sounding
LIDAR I DIAL 1183
Introduction; DIAL; Doppler; Raman; Resonance. Obser- Measures RM (1984)Laser Remote Sensing. Fundamentals
vations for Chemistry (Remote Sensing): Lidar. Op- and Applications. New York: Wiley.
tics, Atmospheric: Airglow Instrumentation; Optical Mischenko MI, Hovenier JW and Travis LD (eds) (2000)
Remote Sensing Instruments. Radiative Transfer: Light Scattering b y Nonspherical Particles. Theory,
Cloud-radiative Processes. Measurements, and Applications. London: Academic
Press.
Sassen K (1991) The polarization lidar technique for cloud
research. Bulletin of the American Meteorological Soci-
Further Reading ety 72: 1848-1866.
Bohren CF and Huffman DR (1983) Absorption and Sassen K (2000)Lidar backscatter depolarization technique
Scattering of Light by Small Particles. New York: Wiley. for cloud and aerosol research. In: Mischenko MIJ,
Deirmendjian D (1969) Electromagnetic Scattering on Hovenier JW and Travis LD (eds) Light Scattering by
Spherical Polydispersions. New York: Elsevier. Nonspherical Particles: Theory, Measurements, and Ap-
Goody RM and Yung YL (1989) Atmospheric plications, pp. 393-416. London: Academic Press.
Radiation: Theoretical Basis. New York: Oxford Uni- Twomey S (1977) Atmospheric aerosols. Developments in
versity Press. Atmospheric Science 7 .Amsterdam: Elsevier.
Lynch DE, Sassen K, Starr DO and Stephans GL (eds)(2002) Van de Hulst HC (1957)Light Scattering b y Small Particles.
Cirrus. Oxford: Oxford University Press. New York: Wiley.
E V Browell, S lsmail and W B Grant, NASA Langley ( CH4)are being explored. The DIAL technique is also
Research Center, Hampton, VA, USA being proposed for global measurements of H20,03 ,
Copyright 2003 Elsevier Science Ltd. All Rights Resewed and C02 from space. All of these topics are discussed
in this article.
Introduction
Differential absorption lidar (DIAL) is a laser remote
DIAL Technique
sensing technique that is used for range-resolved
(profile)measurements of atmospheric gas concentra- DIAL is a remote sensing technique that uses two lidar
tions. This technique was first applied in 1966 for returns to determine the distribution of a selected gas
remote measurements of water vapor (H20)and since along the direction of the lidar beams. A simplified
then it has been used to measure other naturally version of the DIAL concept is shown in Figure 1. The
occurring atmospheric gases such as ozone (03) and molecules and aerosols in the atmosphere provide the
many pollutant gases, such as sulfur dioxide (SOZ), backscattering media for the laser light via Rayleigh
nitrogen dioxide (NOZ), ammonia (NH3), mercury and Mie scattering, respectively. Two laser wave-
(Hg), carbon monoxide (CO), and hydrocarbons. lengths are employed, one tuned t o a strong absorp-
While the initial DIAL technique development focused tion feature of the gas of interest, generally called the
on H20, the main thrust of the DIAL applications in ‘on’ wavelength (Aon) and the other tuned to a nearby
the 1970s and early 1980s was on pollution monitor- wavelength with weak absorption by the gas, gener-
ing. The first airborne measurements with DIAL were ally called the ‘off’ wavelength (ioff).The value of the
aimed at studying tropospheric O3 in large-scale average gas concentration, N A , in the range interval
pollution studies over the east coast of the US in from R1 to R2, can be determined from the ratio of the
1980. Subsequently, airborne DIAL measurements of backscattered lidar signals at A,, and ioff, as shown
H20 were demonstrated in 1982. Technological in Figure 1. In that equation, Ac = c o n - coff,is the
advancements in airborne DIAL systems have greatly difference between the absorption cross-sections at the
increased the measurement capabilities of ground- on and off wavelengths, and Pr,on(R1)and Pr.off(RZ)
based and airborne DIAL systems for measurements of are the signal powers received from range R at the on
tropospheric and stratospheric 0 3 and tropospheric and off wavelengths, respectively. This is essentially an
H20, along with pollution measurements of many application of the familiar Beer-Lambert law for an
other gases in range-resolved and column measure- absorbing medium. The ioff lidar return also provides
ments. Even the possibility of high-precision DIAL important information on the molecular and aerosol
measurements of carbon dioxide (C02) and methane scattering properties of the atmosphere, and this
1 184 LIDAR I DIAL
Figure 1 Differential Absorption Lidar (DIAL) concept.
contributes greatly to the science interpretation of the to account for wavelength differences in at-
gas profile measurement. mospheric scattering and attenuation must be
Some of the key considerations for being able to use applied.
the DIAL technique for measuring range-resolved gas 3 . The DIAL measurement wavelength region should
concentrations are listed below: be selected such that the atmospheric scattering is
sufficient to provide adequate backscattered sig-
1. Pulsed laser sources are required that can generate nals from the atmosphere. Molecular scattering
sufficient pulse energies at the DIAL wavelengths drops off as ,Ie4and is very weak for wavelengths
on and near a suitable absorption feature of a gas of longer than about 1pm. Thus, for longer wave-
interest, with the ,Ionand oonoptimized so that the lengths, typically longer than 1pm, aerosols,
integrated absorption by the gas at the maximum clouds, or surfaces, including topographic targets,
measurement range has a one-way optical depth of have to provide the backscattered signal. For
about one. This ensures an optimum compromise species that are measured in the thermal infrared
between having a large Ao and having enough spectral region (3-12 pm), gas profile measure-
signal from the most distant range to be able to ments are generally constrained to the planetary
make a DIAL measurement. boundary layer (mixing layer), although long-path
2. The laser wavelengths must be kept as close measurements can be made using topographic
together as possible in order to minimize DIAL targets.
measurement errors that result from differences 4. Unknown absorption interference from other gas
in the atmospheric scattering and attenuation species should be minimized.
at these two wavelengths. When the laser wave- 5. Lasers with short pulse lengths compared with the
length separation is unavoidably large, a correction desired measurement range resolution should be
LIDAR I DIAL 1185
used. For example, a 100 ns. laser pulse length has a Closely spaced DIAL wavelengths will help to reduce
15m folded scattering length. any biases due to this effect.
6. The detectors used must have the necessary sensi-
tivity and low noise level for making the DIAL
Application Areas
measurements.
The primary applications of the DIAL technique have
Note that the DIAL technique generally performs been in the areas of O3and H20 measurements. These
better in the nadir direction than in the zenith gases are of great importance in such areas as
direction, especially for wavelengths shorter than atmospheric chemistry, health, and weather, and they
1pm. The reason is that molecular density and, in are discussed in detail in this article. A number of other
general, aerosol scattering decrease with increasing gases have been or can be studied using the DIAL
altitude, so that the backscattered lidar signal falls offtechnique. While H 2 0 was the first gas measured with
more rapidly in the zenith than in the nadir. Thus the the DIAL technique (temperature-tuned ruby laser
DIAL technique is well suited for operation from lidar system in 1966), NO2 was the first pollutant gas
airborne platforms, especially for relatively well- measured using this technique in the early 1970s.
mixed gases such as O3 and H20. For longer wave- Nitrogen dioxide is the only gas of interest with a
lengths, the DIAL measurements depend almost strong absorption band in the visible spectral region
entirely on aerosol scattering, and the nadir direction (absorption features are in the blue region). However,
is still slightly more preferred above the planetary since NO2 occurs in low concentrations in most
boundary layer (PBL). Within the PBL the zenith situations, appearing at high concentrations only in
measurements would be slightly preferred because the major pollutant plumes, there has not been significant
relative humidity (RH) generally increases with alti- activity in measuring it using DIAL. Sulfur dioxide has
tude and the aerosol scattering also increases with been measured using dye lasers operating in the W
altitude due to the growth in aerosol size with spectral region near 300 nm, where it has a strongly
increasing RH. modulated absorption spectrum. The interest in SO2
In designing and operating DIAL systems, one must was primarily related to emissions from power plants,
be aware of a number of atmospheric and instrumental but there has also been some interest in SO2 emitted
effects that can cause bias in the measurements, from volcanoes. Mercury is another gas of interest.
including: scattering and extinction differences be- Mercury has a strong absorption line near 254.3 nm,
tween on and off wavelengths; sensitivity of absorp- and it is emitted from a number of sources including
tion and scattering to atmospheric temperature and chlorine-alkali plants, geothermal fields, volcanoes,
pressure; interference due to other absorbing species; cinnabar mining areas, and from coal-burning power
laser spectral characteristics; background radiation; plants. Hydrocarbons have seen some interest as well.
detection system noise, etc. A well-designed DIAL The absorption bands vary from near 300nm for
system reduces these errors to a manageable, if not benzene to 3.4pm for methane and 10.5pm for
insignificant, level. In some cases, these biases are ethylene. Other hydrocarbons can also be measured
reduced or removed during data processing. in the 3.2-3.7pm and 9.3-10.7pm spectral regions.
Hydrocarbons have various urbadindustrial sources,
Use of Topographic Targets
and it can often be cost effective to use laser systems to
When the DIAL measurement does not have sufficient detect hydrocarbon gas leaks in industrial plants so
atmospheric backscatter or range resolution for a gas they can be eliminated. Ammonia is commonly found
profile measurement, such as for a continuous-wave near feedlots and other agricultural operations, and it
laser or a low-pulse-energy laser in the infrared, a has absorption lines in the 9.3-10.7pm spectral
topographic target may be employed to provide the region. Various laser sources can be used for the above
backscattered laser radiation. This results in a long- IR DIAL measurements including optical parametric
path or column measurement of the gas. By using a oscillators in the 3.2-3.7 pm spectral region and C 0 2
series of targets at different ranges, it may be possible lasers in the 9.3-10.7 pm spectral region.
to obtain some range-resolved information. There are There is also increasing interest in the measurement
several problems that have to be faced when using of C 0 2 due to its importance in global climate change
topographic targets. One is that unless the target is and the carbon cycle. The sources and sinks of
moving or being scanned, the measurement accuracy atmospheric C 0 2 are not well understood on a global
will not increase rapidly with the number of pulses scale, and there is a need to map the large-scale sources
averaged. Another concern is that there are sometimes and sinks of C 0 2 , preferably from space. There are
very sharp spectral changes in the reflection features of candidate absorption lines in the 1.6 and 2 pm spectral
the materials contained in the topographic targets. regions, and solid state laser technology is being
1186 LIDAR I DIAL
developed in these regions for possible space-based accuracy for O3measurements of better than 10% or 2
DIAL systems. Carbon dioxide is difficult to measure ppb by volume, whichever is larger, and a measure-
in part because it is a long-lived gas and it has a ment precision of 5% or 1 ppb by volume with a
relatively high ‘background’ concentration, which vertical resolution of 300 m and an averaging time of
does not vary much around the Earth. As a result, 5 min (about 70 km horizontal resolution at typical
the measurement accuracy and precision must be aircraft ground speeds). An example of the O3
extremely high for meaningful measurements. Also, it measurements made with this system is shown in
is important to measure the atmospheric mixing ratio Figure 3. This figure shows a wide range of atmos-
of COZYand thus it will be important to measure pheric processes associated with the production and
atmospheric number density to convert DIAL-meas- transport of tropospheric O3 from the tropics to high
ured C02 concentrations to mixing ratios. Both active latitudes. The large latitudinal variation of the tropo-
and passive remote sensing techniques are being pause height at about 100ppb by volume is clearly
considered to derive accurate atmospheric number seen. The low O3 at low latitudes near the surface is
densities. caused by photochemical loss, and the result of
convective cloud pumping of the low-O3 air into the
upper troposphere can also be seen in the tropics. At
Airborne DIAL Systems and mid-latitudes, the higher O3 is caused by a combina-
Applications tion of photochemical production, horizontal trans-
Determining the large-scale variations of O3 and H 2 0 port, and stratospheric-tropospheric exchange.
is important to our understanding of a broad range of The NASA Langley airborne UV DIAL systems have
atmospheric processes. For example, measurements of made significant contributions to the understanding
O3 and H 2 0 distributions can lead to an improved of O3 in both the troposphere and stratosphere.
understanding of the relative role of transport versus They have been used in 19 international and three
photochemistry in the tropospheric O3budget. Ozone national field experiments during the past 22 years,
and H2O are important radiatively and contribute to and during these field experiments, measurements
the radiation budget and climate change. Water vapor were made over, or near, all of the oceans and
is influential in many different meteorological proc- continents of the world. A map of the regions where
esses and in the transport of energy on large scales. For these airborne field experiments were conducted is
a better understanding of the atmosphere, it is impor- shown in Figure 4.
tant to study the spatial and temporal variations of
these gases over many regions of the Earth and Airborne H20 Measurements
ultimately be able to make measurements of them
from space. The first DIAL measurements of H 2 0 with a contin-
uously tunable laser was demonstrated in the late
1970s. In an initial step towards the development of a
Airborne O3 measurements
space-based H2 0 DIAL system, the first airborne H20
The first airborne DIAL system was flown for O3 and DIAL system was developed and demonstrated in
aerosol investigations in conjunction with a large- 1982. This system was based on Nd:YAG pumped dye
scale pollution field experiment conducted over the laser technology, and it was used in the first airborne
east coast of the US in the summer of 1980. This initial H 2 0 DIAL investigation of the marine boundary layer
system evolved into the advanced UV DIAL system over the Gulf Stream. This laser was later replaced
that has been used in many field experiments over the with a flashlamp pumped solid state alexandrite laser,
last two decades. A schematic of the UV DIAL system which had high spectral purity ( > 99% of laser energy
is shown in Figure 2. This system uses two 30Hz, contained within a narrow, 1pm, spectral region), and
frequency-doubled Nd:YAG lasers to sequentially this system was used to make accurate H2 0 profile
pump two dye lasers that are frequency-doubled into measurements across the lower troposphere under a
the UV to produce on-line/off-line wavelength pairs of variety of atmospheric conditions. A third H 2 0 DIAL
288.2nm 299.6nm or 301 nm 310nm for DIAL 0 3 system called LASE (Lidar Atmospheric Sensing Ex-
measurements in the troposphere and stratosphere, periment) was developed as a prototype for a space-
respectively. The residual 1064 nm and 600 nm beams based H2 0 DIAL system. This system was designed to
from the frequency-doubling processes of the Nd:YAG operate autonomously from a high-altitude ER-2
and dye lasers, respectively, are also transmitted for aircraft, and it uses a Tisapphire laser and from one
aerosol and cloud measurements. The parameters for to three different H20 absorption cross-sections to
the NASA Langley airborne DIAL system are given in make H2 0 measurements across the entire tropo-
Table 1. This system has a demonstrated absolute sphere. While the LASE system was initially designed
Next Page
Figure 2 Configuration of the NASA Langley airborne UV DIALsystem. Four beams are simultaneouslytransmittedin the nadirand zenithdirectionsformeasurementsofO3profiles with DIAL
wavelengths of 288 and 300 nm and for aerosol and cloud profiles at 600 and 1064 nm.
121 6 LIGHTNING / Overview
~ ~ ~ _ _ _ _
Brinksma EJ, Meijer Y, McDermid S, et al. (1998)First lidar and South Poles. Geophysical Research Letters 28(7):
observations of mesospheric hydroxyl. Geophysical Re- 1199-1202.
search Letters 23(1):51-54. She CY, Yu J, Latifi J and Bills R (1992) High-spectral
Gardner CS (1989) Sodium resonance fluorescence lidar resolution fluorescence light detection and ranging for
applications in atmospheric science and astronomy. mesospheric sodium temperature measurements. Ap-
Proceedings ofthe IEEE 77(3):408-418. plied Optics 31(12):2095-2106.
Gardner CS, Papen GC, Chu X and Pan W (2001)First lidar von Zahn U and Hoffner J (1996)Mesopause temperature
observations of middle atmosphere temperatures, Fe profiling by potassium lidar. Geophysical Research
densities, and polar mesospheric clouds over the North Letters 23: 141-144.
Contents
Overview
Production of Nitric Oxide
electrification were widely studied, if not explained.

Overview Improvements in remote sensing of the relevant
variables have further advanced our understanding
M B Baker, University of Washington, Seattle, WA, USA of lightning as an atmospheric phenomenon, and
lightning observations now provide information on
cloud and atmospheric processes that complement
traditional meteorological measurements.
Introduction
Data and Models
This article outlines the current understanding of
lightning from tropospheric clouds. It begins with a Our understanding of lightning is derived from data
brief history of research in this area and a summary of from many sources. In-situ field mills, current meters,
the tools used in the present-day studies. It then and induction rings mounted on meteorological bal-
presents a short description of a lightning flash and loons and/or aircraft yield small-scale information on
discusses the major physical processes that lead to the in-cloud electric fields and charge distributions. Sur-
production of lightning. This discussion provides the face sensors located beneath storms and/or on mobile
basis for the next topics: the links between lightning laboratories record electric field changes and currents
and other meteorological parameters, and the clima- associated with lightning along with colocated
tology of lightning. cloud physical variables. Sophisticated instruments
measuring various components of the radiation asso-
ciated with lightning can pinpoint the in-cloud loca-
Background tions of lightning strokes, and follow the lightning
channel trajectories. Finally, lightning can be detected
History of Lightning Research
and its location mapped from great distances by
Lightning is arguably the most dramatic naturally optical detectors placed on space-based platforms
occurring atmospheric phenomenon, but the first and by detection at the surface of low-frequency
understanding of its electrical nature came with electromagnetic signals propagating in the global
Benjamin Franklin’s famous experiments in the waveguide.
1750s. These were followed by rapid advances in the These data can be used in numerical thunderstorm
theory of electromagnetic phenomena, with practical models of varying complexity to simulate the evolu-
advances in lightning protection devices. In the 1920s tion of cloud and electrical properties over a cloud life
C.T.R. Wilson suggested the role of thunderstorms in cycle. These provide useful means for testing new
the atmospheric electrical system. In the 1950s and ideas, although at present incorporation of elec-
1960s, as the field of cloud physics came into its own, trical phenomena in models requires semiempirical
the microphysical processes involved in thunderstorm parameterizations.
LIGHTNING I Overview 1217
Lightning: the Process

Lightning is an electrical discharge; i.e., a rapid
redistribution of electrical charge within thunder-
clouds or between a cloud and its surroundings. IC, or
intra-cloud lightning, is a discharge between two
points in the same cloud; C-G, or cloud-to-ground
lightning, transfers charge between a point in cloud
and a point on the ground below. (Intercloud light-
ning, connecting points in different clouds, will not be
treated here.)
A lightning flash consists of several components. In
C-G flashes (which are those most widely and closely
studied) a low-current ( w 1kA), low-luminosity chan-
nel is initiated in the cloud. The channel is a fully
ionized plasma tube several centimeters in diameter.
Joule heating by electrical currents in the plasma raises
the channel temperature to above 20 000 K. The
heated gas expands, giving rise to shock waves
producing the sound called thunder. The current
carried by the channel discharges tens of millions of
volts. The high temperatures inside the channel favor
certain chemical reactions of atmospheric signifi-
cance; in particular, the production of NO,, trea- Figure 1 Lightning flash photographed with a streak camera.
ted elsewhere in this encyclopedia (see Stratospheric The camera moves during the lightning flash, enabling the
Chemistry and Composition: Reactive Nitrogen (NO, resolution of a number of individual strokes. (Reproduced with
and NO,)). permission from Uman M (1987) The Lightning Discharge. Inter-
This first channel propagates downward in dis- national Geophysics Series, Vol. 39. San Diego, CA: Academic
Press.)
crete steps 2-50m in length separated by pauses of
-5Ops, and is commonly referred to as a stepped
leader. This is followed by a return stroke (i.e., a
The Global Circuit
return current) from ground to cloud. There may be
several leader-return-stroke pairs but the time The surface of the Earth behaves like a negatively
between strokes is usually too short for our eyes to charged conductor. This conductor and the electro-
resolve and thus what we call a lightning flash usually sphere (a region of net positive charge whose base is
consists of several individual strokes. The return around 65 km altitude) constitute the ‘plates’ of a
stroke is the brightest phase of the flash with typical concentric spherical condenser. There is a fairly con-
peak currents -40 kA; the average total current is stant, spatially uniform fair-weather electric field in
several amps. Figure 1 shows a sequence of strokes the atmosphere between the plates: EfairM 100 V m-’
recorded on film. at the Earth’s surface. This field is associated with a fair
The entire discharge lasts about 0.1 s. C-G flashes weather current of positive charge to the ground:
typically lower 20-30 C to ground but values in excess ]fair = 2 x A mP2.Without ‘batteries’ the current
of 200 C have been recorded. would discharge the field in a few minutes. The major
The high temperature inside the channel increases batteries, or generators, are thunderstorms, which
the pressure inside it and gives rise to a shock wave that deposit predominantly negative charge on the Earth’s
expands into the surrounding air. Air resistance damps surface through lightning and precipitation. Thus
the wave and the pressure of the disturbance decreases lightning at any point perturbs the local electric charge
rapidly, so that the far field effect is that of sound distribution and the electromagnetic waves travelling
waves, or thunder. in the global waveguide.
Thunderstorm Electrification
Thunderstorms and Lightning
The electrical conductivity inside a cloud is much
To understand the electrification of thunderstorms lower than that in the free air, since ions become
we review the electrical context in which they attached to the hydrometeors. Thus insertion of a
develop. completely passive cloud into the fair-weather field
1218 LIGHTNING/ Overview
results in the attachment of ions from the environment The fact that the negative charge center is con-
on hydrometeors at the cloud edges. As these move, fined in temperature suggests that microphysical
the electric field distribution in the cloud changes. interactions control the electrical charge distributions
Fields up to 100-lOOOVm-l are often found in in this type of storm, although dynamic pro-
“nonelectrified” clouds. In thunderstorms, on the cesses (transport in updrafts and downdrafts and
other hand, internal charging mechanisms produce turbulent motions) are also important. The major
electric fields of up to 100 kVm-l. microphysical charging mechanism in convective
The information we have on in-cloud charge distri- storms appears to be charge separation between
butions is very limited, and we know only some colliding ice particles.
general features, shown schematically in an idealized Laboratory studies show that electric charge is
isolated convective thunderstorm in Figure 2 . In separated during rebounding collisions between ice
simple storms like this one, a vertical dipolar or particles, so that the rebounding particles carry equal
tripolar charge distribution develops as the cloud and opposite charge. The sign and magnitude of the
grows, with typical charges Q x 10-1OOC in each charge are reproducible functions of temperature,
center at maximum, carried by the cloud water and ice particle types and sizes, and the cloud water content
particles. The positive charge is distributed in a diffuse and distribution. At temperatures above a ‘reversal’
region aloft; typically, the charge density in this region threshold, around -15”C, the hail generally receives
is several C km-3. The negative charge center can be positive charge, and at lower temperatures it receives
more compact, and is always located in the part of a negative charge during a collision. Typically the
cloud (called the ‘charging zone’) in which the charge separated is around 10-14C per collision
temperature lies between -10°C and about -20 to when one of the partners is hail; far more charge is
-25°C. This is the only region inside clouds in which separated when both are hail and less when both are
vapor-grown ice, supercooled water, and hail particles vapor-grown ice crystals. In general, the large (precip-
coexist. There are often other important pockets of itating) hail particles receive net negative charge as
charge, including a secondary positive charge center; they fall through the cloud of smaller ice crystals.
these become larger and more complex as the cloud Subsequently, gravitational separation of the large,
ages. negatively charged precipitation particles and the
Figure 2 Schematic picture of an idealized thunderstorm, showing the ascent of hydrometeors (round circles) in the updraft (velocity w )
and sedimentation of the larger particles, which collide with smaller ones below. Reboundingcollisions between hail particlesand small ice
crystals result in charge transfer between them. Gravitational separation of the lighter (positively charged) ice particles and the negatively
charged graupel (Le., soft hail) particles leads to the electric charge distribution shown. This precipitation-based charging is thought to
dominate early thunderstorm electrification. (Reproduced with permission from Schroeder V (2000) How Does Lightning lnitiate and What
Controls Lightning Frequency? PhD thesis, University of Washington.)
LIGHTNING / Overview 1219
lighter, positively charged ice crystals constitutes a Microphysical Processes Leading

‘generator current’Jgen x 0.1 A kmP2. to Lightning Production
The collision-based process is the most important,
but not the only one responsible for creating the In order to make further progress in understanding
observed in-cloud charge distributions. Other pro- lightning generation we must solve two outstanding
cesses become more important as electrification pro- puzzles; namely, (1)what is the mechanism for the
ceeds; for example, cloud particles in the existing charge transfer between colliding ice particles, and ( 2 )
electric field become charged inductively; melting how is lightning initiated? In this section we briefly
appears to be associated with charging in stratiform discuss our current understanding of these processes
clouds; ions become attached to aerosol particles and and the hypotheses now being tested to explain them.
cloud particles near cloud edges; finally, motions of Charge Transfer via Ice-Ice Collisions
charged particles redistribute them throughout the
cloud system. The mechanism by which the rebounding ice particles
This inhomogeneous electric charge distribution become charged remains somewhat unclear. Both sign
produced by these processes is associated with and magnitude of the charge depend on the rate of
an electric field. The field grows as the cloud develops growth of the ice particle from the vapor. The charge
and eventually conditions are met in which lightning transfer appears to be due to the fact that positive
is triggered. The first flash is typically initiated just water ions in ice have much higher mobilities than do
above the lower, negative charge center and the first negative ions. Ions are created at high rates near the ice
lightning is usually IC. The lightning flash transfers surface, and the positive ions migrate inside, so that
charge between the upper, positive charge center and the growth creates net negative charge at the ice-vapor
points below. The net charge density aloft is thereby interface. During a collision, material from the two
decreased and positive charge builds up somewhat particles is mixed and net negative charge is trans-
below the negative charge center. Continued IC ferred to the particle with the slower growth rate. The
lightning as the convective cell reaches full vigor various characteristics of the hail - i.e., its smoothness
continues this charge redistribution process. At some or roughness, the density of grain boundaries and
point the electric field between the lower charge dislocations, and the chemical composition of the
centers and ground becomes strong enough to trigger frozen water - all modify the charge transfer.
C-G flashes.
Lightning Initiation
This classical picture does not describe all thunder-
storms. Recent measurements (for example, those Lightning initiation is thought to involve the following
made during the Severe Thunderstorm Electrification stages. (1)The in-cloud electric fields intensify via
and Precipitation Study (STEPS) project in 2000) microphysical and dynamic processes. (2) Free elec-
reveal that the charge structure in clouds is often trons in some region of the field begin to accelerate,
inverted, with negative charge aloft. The exact mech- creating more electrons via ionization, in so-called
anisms for this inversion are not known. Moreover, in electron avalanches. If these continue they result in
complex cloud systems containing large stratiform propagating corona streamers, or small-scale currents.
regions and/or multicellular structures the charge ( 3 ) Heating associated with the propagation of high
distribution can be highly three-dimensional and electric current through and beyond the region of very
complex. high local fields produces the hot, completely ionized
Most C-G flashes from convective storms lightning channel, or leader.
carry negative charge to ground, indicating that the This process is known as ‘dielectric breakdown’. In
in-cloud point of origin of the flash is associated with a the laboratory, dielectric breakdown can occur only if
negative charge pocket, but in midlatitude storms the ambient electric field reaches a (pressure-depend-
around 10% deposit positive charge to ground. ent) threshold value, denoted E , h ( p ) . At surface
Positive flashes tend to be associated with low precip- pressure (1000 hPa) Eth x 2600 kVm-l. This value,
itation regions in mesoscale storm systems, and they and the physics of breakdown in the laboratory
are often very intense. In general, reversed lightning context, are well understood. However, measured
polarity results from (1)wind shear, displacing the electric fields inside thunderclouds never reach the
upper positive charge so that it overlies cloudfree air threshold value and are usually more than an order of
and is electrically coupled to ground; ( 2 ) inverted magnitude less than this value.
charge structure in cloud, possibly due to charging at Two classes of hypotheses are typically invoked to
temperatures higher than the reversal temperature; resolve this puzzle. While both involve acceleration of
and ( 3 ) absence of large concentrated pockets of an initial electron and subsequent production of
negative charge. electron avalanches, everything else about the two
1220 LIGHTNING I Overview
mechanisms is different. The first is the ‘conventional Therefore

breakdown’ hypothesis, which operates on very small
spatial scales. It is based on the fact that in the vicinity PQH
FN-
of conducting hydrometeors the electric field magni- Ebe
tude can reach the breakdown threshold value.
According to this lightning initiation hypothesis, free Flash Rate and Radar Reflectivity in
electrons near cloud ice and water particles are Charging Zone
accelerated by the local field and can produce local
For simplicity we neglect all contributions to bQ
streamers.
except that due to collisions between hail particles and
The second class of triggering mechanism hypoth-
vapor grown crystals. Consider a situation in which
eses depends on the fact that in-cloud electric fields are
Nhail [mP3]hail particles of diameter Dhail collide with
fairly large, even if not of breakdown magnitude, over
Nice vapor-grown ice particles of diameter dice. The
several kilometers. An electron moving in such a field
laboratory charging results can be parameterized in
is accelerated by the field and decelerated by its
terms of these variables, leading to an expression of the
interactions with neutral molecules. If the electric field
form
magnitude is sufficiently great, it can compensate for
the deceleration. The balancing electric field is called
the ‘breakeven’ field; Ebe(p) [kVm-l] M 200p [atm],
much smaller than the breakdown threshold field.
The ‘runaway breakdown’ hypothesis of lightning x ZMi,, [31
initiation suggests that if E b e is exceeded, even if where Z is the radar backscatter and Micethe mass of
slightly, over sufficient distance, then high-energy vapor-grown ice in the region of the collisions. Thus,
electrons (resulting from cosmic ray showers or we can write
atmospheric radioactivity, or short-lived, small-scale
intense electric fields) travelling in the field can create
sufficient daughter electrons by ionization to initiate a FN- ZMiceH [41
Ebe
leader, or ionized channel. The more the field exceeds
the breakeven value the shorter the distance needed to Therefore we expect the measured lightning fre-
create the leader. Measured electric fields in clouds are quency to increase with radar reflectivity (i.e., the large
generally much smaller than the breakeven field, and hail concentration) and with the mass of vapor grown
just reach it or barely exceed it immediately prior to a ice. Both of these predictions have been borne out in
lightning stroke, lending credence to the runaway preliminary analyses of lightning frequency vs. radar
breakdown hypothesis. reflectivity and 85 GHz signal from thunderstorms. As
Neither the conventional nor the runaway mecha- a rule of thumb, a minimum radar reflectivity of
nisms are completely satisfactory; neither can explain around Z = 40 dBZ for temperatures about -7°C
all the observations. Unfortunately, balloon measure- seems to be required for rapid electrification.
ments cannot solve the lightning initiation puzzle
because they are Lagrangian in nature, they sample Flash Rate and Updraft Velocity
very small volumes of cloud, and their spatial and
temporal resolution is too coarse. Therefore consider- We can rewrite eqn [4] to relate F to updraft velocity in
able effort will have to be devoted to interpretation of the charging zone. The fall velocity of the hail particles
complementary measurements in order to clarify the is very roughly proportional to the hail particle
nature of the lightning trigger(s). diameter. Therefore, if we assume that most of the
collisions take place at the balance point, where the fall
velocity of the hail particles is equal to the updraft
Lightning and Cloud Properties velocity w, then
Although much remains to be learned about lightning
generation, we can examine some basic relationships
linking lightning frequency F [s-l] to other important
Thus
parameters. A simple dimensional argument illus-
trates the important links. F is a function of the charge w ~ M ~ ~ ~ H
generation rate, ,be, the depth H of the electrically FN
Eth
active region and the threshold field for producing
lightning (which we assume here is a function of &e). Since updraft velocity in the mixed-phase zone is of
The charge-generating current density Jgen N PQH. importance in weather prediction, the strong depend-
LIGHTNING / Overview 1221
ence of lightning on this factor may provide a useful Lightning and Atmospheric Aerosols
remote indicator of storm strength. More
Forest fires and urban pollution have been shown
careful studies show that as a general rule the very
to produce anomalies in C-G lightning flash
existence of lightning implies vertical velocities of at
rate, lightning intensity (peak current, and/or
least 7-8 m s-' in the charging zone. Updraft velocities
radiance) and in the sign of the charge brought to
are often tied to buoyancy, or CAPE (con-
ground by lightning. These intriguing observations
vective available potential energy), so that it is
might be due to modifications of the hydrometeor
not surprising that observations tie F to CAPE in
populations by variations in the distributions of
regions where sounding shape is not highly variable,
aerosol particles, to modification of the soundings by
such as over oceans; however, this relationship is
anomalous surface heating, to chemical modifications
regional over land and attempts to identify a single
of the charge transfer process or to some combination
CAPE 'threshold' with the onset of lightning have not
of these.
been very fruitful.
Flash Rate and Vertical Water Fluxes Lightning Climatology

It has long been hoped that, because of the role of hail The advent of routine lightning monitoring programs
in generating lightning, reliable and useful relation- from satellites and important advances in ground
ships could be found by which lightning frequency sensing capability have allowed great advances in our
could be used as a surrogate for precipitation at the understanding of the distribution and frequency of
ground. Unfortunately, although it is possible to derive lightning. Here we present an overview of lightning
dimensional relationships between flash rate and in- climatology.
cloud hail concentrations (see eqn [3]), universal The global, diurnal average C-G flashrate is FC-G x
quantitative relationships of sufficient accuracy for 12-16 s-l; F,, x 55 s-l in N H summer, over land.
use in prediction are far from our reach. Many high As shown in Figure 3, about 70% of all lightning
precipitation storms produce no lightning at all occurs between 30" S and 3 0 " N , echoing the
(particularly if most of the precipitation comes from distribution of convective activity. In the tropics
warm cloud processes), and the complex processes there is no seasonal cycle in the lightning frequency.
involved in producing precipitation (particularly pre- The seasonal cycle in midlatitude lightning flash rate
cipitation at the ground, where predictions are most (with more lightning in the summer months) is due
needed) are so variable on small spatial and temporal mostly to differences in the number of storms (i.e.,
scales that precipitation cannot be accurately predict- presumably, in CAPE), rather than in differences in the
ed from simple macroscopic measures like lightning flash rate per storm.
frequency. Estimates of precipitation 'yield' y for F is maximum at 1500-1800 local time, with much
example, range from 10'1 y l 1O1O [kgH,O flash-'] smaller amplitude variation over the oceans than over
in one given location. land. The diurnal cycle at any fixed location is due in
Since lightning frequency increases with increasing part to a cycle in flash rate per storm and in part to the
rate of supply of water to the charging zone and diurnal cycle in the number of storms.
upward flux of small ice particles, lightning can be Lightning production requires both high vertical
used in certain predictive schemes. Remote sensing of velocity and high upward water flux in the charging or
lightning has been successfully incorporated into mixed phase zone. Water fluxes in this region are
predictions of hurricane landfall time and location, relatively weak in many oceanic storms, which may
and more such applications are likely to follow. On provide a clue to the large land-ocean differences in F
larger scales, flash rate can also be considered a indicated in Figure 3: F,,,,, (on land) x 2 x Fsto,, (over
surrogate for the rate of water lofting to the upper ocean), but Ftotal(land)M 10xFtotai(ocean).(Probably
troposphere in convective towers. A simple argu- the difference is enhanced by large continental
ment suggests that globally around MiceM storms.)
lo6 [kg ice flash-'] is lofted to the upper troposhpere. The electric field distribution in and below a
At present, attempts to make more quantitative thunderstorm determines whether the lightning pro-
estimates of regional condensate lofting from light- duced is IC or C-G. Globally, the ratio IC/C-G ranges
ning observations have proved frustrating because from about 1 to 5, increasing with increasing latitude.
the relationship is highly variable, but global light- This latitudinal variation probably occurs because
ning activity as inferred from ELF (extremely low- the base of the charging zone is higher and the depth
frequency) waves in the global circuit appears to be of the charging zone larger in the tropics than in
strongly correlated with tropical water lofting. midlatitudes.
1222 LIGHTNING /Overview
Figure 3 Lightning incidence in the months of December 1997, January 1998 and February 1998, as measured by Lightning Imaging
Sensor, in tropical orbit aboard the Tropical Rainfall Measuring Mission (TRMM) satellite. Note preponderanceof lightning over land and
surrounding land masses. (Reproduced with permission from Christian HJ, Blakeslee RJ, Goodman SG, et a/. (1999) The lightning
imagingsensor. In: Proceedings of the 7 7th lnternationalConferenceon Atmospheric €lectriciw, pp. 746-749. Guntersville,Alabama, 7-
11 June 1999.)
Summary charging zone The region in a cloud in which

vapor-grown ice particles, graupel
Lightning is an electrical discharge that arises in the
particles and supercooled liquid
high electric fields inside a thunderstorm. While
drops coexist and encounter one
neither the field generation mechanisms nor the details
another in frequent collisions.
of the lightning initiation are well understood, we have
This region is confined to a tem-
enough information to begin to relate lightning to the
perature band between approxi-
atmospheric conditions in which it originates. Light-
mately -10 and -20 to -25°C.
ning frequency, polarity, intensity, and the spatial
distribution of lightning flashes in principle carry electric charge Q [C] 1nC = 10-9C.
information about spatial distributions of updraft electric field
velocities and hydrometeors, particularly in the Ebe ( P ) [kV m-ll pressure-dependent magnitude of
mixed-phase regions of storms. Lightning can be the ‘breakeven’ electric field, in
sensed at great distance and it yields a measure of the which an energetic electron of
state of a storm that is independent of those measures energy 1MeV can propagate at
derived from meteorological instruments. Thus, as our constant speed without accelera-
understanding of the links of lightning to other tion or deceleration.
atmospheric phenomena improves, we can design pressure-dependent dielectric
ways to utilize lightning data to improve storm breakdown magnitude of the elec-
prediction. Moreover, distributions of lightning can tric field, at which clear air breaks
provide long-term information on the atmospheric down and massive current flow
portion of the hydrological cycle. Incorporation of ensues.
lightning-related variables into the suite of measures F [s-l] lightning frequency: the number
used routinely to diagnose surface temperature and of lightning flashes per second in a
atmospheric stability may eventually aid in monitor- defined region.
ing global and regional climate change.
global the electrical system consisting of
circuit the Earth’s surface and the elect-
Acknowledgements rosphere, which act like two con-
centric conductors, coupled by the
This article was prepared with the help of V. Schroeder dielectric atmosphere between
and R. Solomon, and was supported by NASA them.
NAG1 8 19.
H [ml depth of charging zone.
J [Am-21 electrical current density.
Nomenclature lightning region of space in which high
CAPE [J kg-l] convective available potential en- channel current produces a fully ionized
ergy; a measure of convective vigor. plasma.
LIGHTNING / Production of Nitric Oxide 1223
lightning An electrical discharge comprising w [m s-l] updraft velocity in charging zone.

flash several individual strokes. Z [dBZ] radar reflectivity from region of
cloud-to-ground A discharge between a point in a charging zone.
(C-G) flash cloud and the surface of the Earth PQ [Cm-31 charge density.
below.
intra-cloud A discharge between two points in
(IC)flash the same cloud.
lightning an individual event within a light- See also
stroke ning flash, in which current prop-
agates between the two points in Aviation Weather Hazards. Electricity, Atmospheric:
one direction. Global Electrical Circuit; Ions in the Atmosphere; Sprites.
microphysical Lightning: Production of Nitric Oxide. Radar: Cloud
Radar. Stratospheric Chemistry and Composition:
parameters
Reactive Nitrogen (NO, and NO,).
Mice[kg flash-'] mass of ice rising from top of
charging zone in a given time
interval divided by number of Further Reading
lightning flashes during the same
period. Christian HJ, Blakeslee RJ, Goodman SJ, et al. (1999)
The lightning imaging sensor. In: Proceedings of the
Nhail w31, concentration and average diam-
eter of hail particles at top of
1 1 th International Conference on Atmospheric Electric-
Dhail [m] ity, pp. 746-749. Guntersville, Alabama, 7-1 1 June
charging zone. 1999.
Nice [mP3], concentration and average diam- Houze RA (1993) Cloud Dynamics. San Diego, CA:
dice [m] eter of small ice crystals at top of Academic Press.
charging zone. Latham J (1981) The electrification of thunderstorms.
stepped the channel in the first stroke of a Quarterly Journal of the Royal Meteorological Society
leader flash, usually seen in C-G flashes. 107: 277-298.
So-called because the propagation MacGormanDRand Rust WD (1998)The ElectricalNature
occurs in a series of discrete steps. of Storms. Oxford: Oxford University Press.
Schroeder V (2000)How Does Lightning Initiate and What
thunder the sound (pressure disturbance at Controls Lightning Frequency? PhD thesis, University of
audible frequencies) that accom- Washington.
panies the shock wave emanating Uman M (1987) The Lightning Discharge. Vol. 39 in the
from the channel into the sur- International Geophysics Series. San Diego, CA: Aca-
rounding air. demic Press.
Production of Nitric Oxide

C Price, Tel Aviv University, Tel Aviv, Israel polluted boundary layer, observations show that up to
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. 95% of the NO, in the anvils of thunderstorms is
produced by lightning, with maximum measured
concentrations reaching 25-50 ppbv. Furthermore,
enhanced NO, concentrations in the upper tropo-
Introduction sphere have been measured hundreds of kilometers
Besides being one of the most powerful and beautiful downwind of thunderstorms. Although the major
atmospheric phenomena, lightning also plays an source of atmospheric NO, is currently the anthro-
important role in the chemistry of our atmosphere. pogenic burning of fossil fuels, it is believed that
Lightning is a major source of nitrogen oxides lightning may be the largest natural source of NO
(NO, = NO (nitric oxide) + NO1 (nitrogen dioxide)) in the troposphere (Table l), with influences on
in the troposphere. Field measurements in thunder- the biosphere, tropospheric ozone, and hence the
storms show high concentrations of NO,, particularly Earth's climate.
in the anvil regions where most of the NO, is The lifetime of NO, in the atmosphere increases
transported by the strong updrafts in thunderstorms. with altitude, with NO, in the boundary layer having a
Although some NO, can be transported aloft from the lifetime of a few hours, while NO, deposited in the
1224 LIGHTNING I Production of Nitric Oxide
Table 1 Sources of tropospheric NO, (10’2gyr-’ = Tgyr-’) 9.45 MJ mol - are needed. In a lightning discharge
Fossil fuel burning
Biomass burning
-
-8
24
the energy can reach lo2-103MJ, with temperatures
rising to 30000K in a few microseconds (p). This is
Soil emissions -12 five times hotter than the surface of the Sun. All
NH3oxidation
Aircraft
Transport from stratosphere
-
-3
0.4
< 0.4
this energy is deposited into a very narrow channel
that at its maximum is only a few centimeters in
Lightning 5-20
diameter but is many kilometers long. Although the
distance from the cloud to the ground may be 5-10 km,
recent observations indicate that if we include all the
branching of the channel within the cloud and below,
upper troposphere has a lifetime of a few days. the total length of these narrow channels may be
Lightning-produced NO, plays a key role in the hundreds of kilometers.
photochemical reactions that determine tropospheric NO, is formed by the thermal dissociation of
and stratospheric ozone ( 0 3 )concentrations. In the molecular 0 2 in the hot channel and the subsequent
troposphere, regions of high NO, concentrations oxidation of N2 to form NO,. Approximately 75-95%
-
( 10 ppbv) result in the production of 0 3 while , in of the NO, is in the form of nitric oxide (NO).
-
those regions with low NO, concentrations
( 10 pptv) O3 is destroyed. The photochemical
destruction of O3 is closely linked to the presence of
0 2 @ 0 + 0
odd hydrogen (HO, = O H + HOz), while the pro- 0 +N2 H NO +N

duction of O3is related to the concentrations of NO,.
The overall production or destruction of O3 therefore
depends on the relative abundances of HO, and NO,.
N+ 0 2 @ NO +0
Because 0 3 is a strong greenhouse gas, NO, is an
important factor when considering global climate The final yield of N O is determined primarily by the
change. Radiative forcing by 0 3 is particularly sensitive rate of cooling of the lightning channel. A slow decay
to changes in the vertical distribution of O3 in the upper of the hot channel would allow the N O to return to N2
troposphere, where lightning has a key role. Further- and 0 2 . However, owing to the rapid cooling of the
more, the reaction of NO, with the hydroxyl radical channel as it mixes turbulently with the surrounding
(OH)influences the ability of the atmosphere to oxidize air, there is not enough time for the reverse reaction to
and remove pollutants from the atmosphere. Less O H occur, and the NO, remains ‘fixed’ or ‘frozen’ in the
implies more pollutants remaining in the atmosphere. atmosphere. The freezing out of NO occurs when the
Nitrogen is also an essential element for life on channel temperature drops below 2500 K. Nitric
Earth, and is often the limiting nutrient in ecosystems. oxide produced by this mechanism is known as the
The reaction between NO2 and O H leads to the Zel’dovich mechanism and is expected to have volume
formation of relatively stable nitric acid ( H N 0 3 ) , mixing ratios of 1 4 %when the air cools.
which can be removed from the atmosphere by Various studies have shown that the production
precipitation, and therefore provides an important yield of N O in the hot channel is 1017m~lecule~
N J-’.
source of nitrate for the biosphere. Although biolog- There has been discussion as to whether the N O is
ical nitrogen fixation is today a common means of produced primarily in the hot channel, or perhaps in
producing nitrogen in the biosphere, lightning was the the shock front that propagates outward at speeds
major source of nitrogen for the biosphere on the early of 50-60 km s-’. Recent laboratory experiments
Earth, and may have had a major influence in the have determined that the shock front produces three
orders of magnitude less N O than the hot channel
evolution of life on Earth.
-
( loT4molecules J-’). In addition, there are regions
around the lightning channel with high electric
Production of NO, fields that produce coronae and streamers. Although
Lightning is continuously occurring around the Earth at recent laboratory studies conclude that the coronae
a rate of 30-100 flashes s-’. Each flash results in the and streamers (cold discharges) produce two orders
dissociation of the air molecules within the lightning of magnitude less NO, than the hot channel produc-
channel, and the recombination into new compounds. tion, the volume of air processed by coronae and
In order to produce NO, in the atmosphere, the streamers is much larger than the volume of the hot
molecular bonds of air molecules (N2 and 0 2 ) need to channel.
be broken, for which a great deal of energy is needed. To It is also necessary to know the amount of energy
split the N-.N bonds in molecular nitrogen, in a typical lightning discharge. Different types of
LIGHTNING/ Productionof Nitric Oxide 1225
lightning discharges (negative cloud-to-ground, posi- stroke having less current than the first stroke,
tive cloud-to-ground, intracloud, intercloud, sprites) however often more charge. Return strokes are
have different physical characteristics. There are also preceded by stepped leaders and dart leaders. Intra-
different ways to calculate energy (optical, electrical, cloud discharges have leaders, but no return stroke.
based on current, based on charge removed). Cloud- Therefore, many different processes need to be con-
to-ground flashes normally contain a number of return sidered regarding the energy of lightning and its
strokes within each flash, with each subsequent return capability for producing NO,.
Figure 1 The 1" x 1" global distributions of lightning-produced NO, in January and July. (From Price C, Penner J and Prather M (1997)
NO, from lightning, 1. (Global distribution based on lightning physics. Journal of Geophysical Research 102(D5):5929-5941 .)
1226 LIGHTNING / Production of Nitric Oxide
For the cloud-to-ground discharge, it has recently being in the Northern Hemisphere in July, and vice
been shown that the energy of the discharge appears to versa in January. Using the best estimates available
be related to the peak current of the lightning today, the monthly global production rates of NO,
discharge (- 30 kA). Ground-based lightning detec- from lightning are shown in Table 2, with July being
tion networks around the globe typically record peak the month with the largest lightning production of
current as one of their parameters, so it may be NO, (1.44Tg of nitrogen). Since there is more
relatively easy to estimate NO, production rates on a landmass in the Northern Hemisphere, the Northern
routine basis. In addition, the first return stroke results Hemisphere summer has a lot more lightning than the
in the dissociation of the 0 2 and N2 in the channel, and Southern Hemisphere summer, so that nearly twice as
any additional strokes down the same channel will not much NO, from lightning is produced annually in the
result in any additional dissociation, especially since Northern Hemisphere than the Southern Hemisphere.
the subsequent strokes always have much smaller peak This implies that even prior to anthropogenic influ-
currents ( ~ 5 0 of
% the first return stroke). So it appears ences on the Earth's climate, there existed a natural
that the peak current of the first return stroke is the imbalance in NO, and hence tropospheric 0 3 between
most important factor to determine the energy of the the hemispheres. On an annual mean basis the amount
flash. It has been estimated that an average cloud-to- of NO, produced by lightning is approximately
ground discharge (with three return strokes) has an 12 Tg yr - with a range from 5 to 20 Tg yr - * owing
energy of 6.7 x l o 9J. Intracloud flashes, which make to the above uncertainties.
up -70% of global lightning, are different from cloud- As mentioned above, the effect of lightning-pro-
to-ground flashes since they do not have a return duced NO, on the climate system depends strongly on
stroke. However, there is very little information where this NO, is deposited in the troposphere. At
regarding intracloud flashes, since the ground-based higher altitudes, NO, has a lifetime of a few days, with
lightning networks were designed to primarily detect a large impact on 0 3 production and hence the Earth's
cloud-to-ground flashes. Nevertheless, we know that longwave radiative forcing. It is therefore also impor-
the largest currents in the intracloud flashes (- 3 kA) tant to know at what altitudes NO, from lightning is
are about an order of magnitude less than in the injected into the troposphere. NO, is relatively insol-
ground flashes (-30 kA). Furthermore, there is evi- uble in cloud water, and therefore can be redistributed
dence that lower ambient pressures in the clouds may within clouds by updrafts and downdrafts, and is not
reduce the amount of NO, produced. It is therefore easily removed from the atmosphere, due to scaveng-
believed that the intracloud discharges have an order ing by precipitation particles. Although the deposition
of magnitude less energy than cloud-to-ground flashes of NO, along the lightning channel may have a
for producing NO,. uniform distribution with height, recent observations
and model simulations imply that after the thunder-
storm dissipates the lightning NO, profile has a C-
Global Contribution shaped distribution with large concentrations in the
In order to understand the importance of lightning upper and lower troposphere. This results from strong
to atmospheric chemistry and climate, the above updrafts and downdrafts transporting the NO, to the
estimates need to be translated into global numbers. anvils and the boundary layer. Owing to the long
Large uncertainties are introduced in extrapolating
from an individual lightning flash to global scales to
determine the global production rate of NO, from
Table 2 Global estimates of monthly and annual production of
lightning. nitrogen by lightning (Tg)
We have recently gained great insight into the global
distribution of lightning from satellites, but we still do January 0.91
February 0.82
not know accurately the frequency of global lightning March 0.95
activity, since the satellites only sample a small fraction April 0.91
of the total lightning. The estimates range from 30 to May 1.03
100 flashes per second. From the spatial distributions June 1.21
of lightning we know that globally lightning occurs July 1.44
August 1.27
mostly over the tropical landmasses, with 75% September 1.01
between 30" N and 30" S. The three main regions of October 0.91
NO, production from lightning are tropical South November 0.85
America, tropical Africa, and tropical South-East Asia December 0.88
(Figure 1).These centers of NO, production migrate Annual 12.2
with the seasons, with 90% of the NO, production
LIGHTNING / Productionof Nitric Oxide 1227
lifetime of NO, relative to the thunderstorm lifetime, PPbV parts per billion by volume
it is not critical where the NO, is produced in the PPtV parts per trillion by volume
storm, but rather where it ends up after the storm has S second
dissipated. Using the same NO, concentrations and Ti3 terragram gram)
spatial distributions (Figure l),but different vertical Yr year
profiles, will result in significantly different tropo-
spheric O3concentrations, and hence climate forcing.
Nomenclature See also

nitric acid Chemistry of the Atmosphere: Chemical Kinetics.
Climate: Overview. Convective Storms: Overview.
joule
Lightning: Overview. Ozone: Ozone as a UV Filter;
Kelvin Photochemistryof Ozone; Role in Climate. Stratospheric
kiloampere ( lo3 ampere) Chemistry and Composition: Reactive Nitrogen (NO,
kilometer and NO,).
megajoule ( l o 6joule)
mole
molecular nitrogen
nitric oxide Further Reading
nitrogen dioxide MacGorman DR and Rust WD (1998)The Electrical Nature
NO and NO2 of Storms. Oxford University Press, Oxford.
molecular oxygen Price CyPenner J and Prather M (1997)NO, from lightning,
ozone 1.Global distribution based on lightning physics.Journa1
hydroxyl radical of Geophysical Research 102(D5):5929-5941.
MAGNETOSPHERE 1229
G K Parks, University of Washington, Seattle, WA, USA immediately verified that the solar wind carries with it
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. solar magnetic fields into space. Space is therefore
permeated with magnetized plasma.
The solar wind is different from winds in the lower
Introduction atmosphere because it is always blowing. All of the
planets immersed in the solar coronal atmosphere are
Magnetospheres are new magnetic structures discov- interacting with it all the time. The electromagnetic
ered during the space age by satellite-borne instru- (EM) interaction induces large-scale currents and
ments that made possible physical measurements in forms magnetic cavities around magnetized planets.
distant regions previously not accessible. The first These cavities are called magnetospheres. Except for
magnetosphere discovered was Earth‘s. Soon after- Mars and Venus, which do not have intrinsic magnetic
ward, another discovery showed that space is not fields, the planets in our solar system all have magne-
empty as once thought but is filled with ionized gases tospheres. This article will focus on planetary magne-
emanating from the Sun, stars, and other celestial tospheres and emphasize features that are associated
bodies with high temperatures. It then became evident with Earth’s magnetosphere, which has all of the
that magnetospheres are ubiquitous in space. What is a elements to characterize a planetary magnetosphere
magnetosphere, how is it formed, and what are some (Figure 1).
of the important internal dynamics? The lower boundary of a planetary magnetosphere
Let us first describe the environment in which begins from that part of the atmosphere where ionized
magnetospheres are found. In our solar system, for
example, the Sun’s coronal atmosphere is hot, lo6 K, - constituents play an important role in the dynamics of
and dynamic, so it expands into space. The expanding
solar coronal atmosphere is called solar wind and
-
the upper atmosphere. For Earth, this boundary is
located at 100 km where the ionosphere begins
(ionospheres are formed by the Sun’s ultraviolet
consists mostly of hydrogen (-95% H+)and helium
(-5% He2+) ions and an equal number of electrons. radiation). The ionosphere is therefore part of the
Matter in the ionized state is called plasma and much magnetosphere. The outer boundary of the magneto-
of known matter in the Universe exists as plasmas. sphere is called the magnetopause, and it separates the
Since ionized matter is a good electrical conductor and domains of the planetary magnetic field and the solar
magnetic fields decay slowly in conductors, it was wind that blows outside it. Its location is determined
Van Allen belts
Figure 1 A sketch of Earth’s magnetosphere in the noon-midnight plane. The dashed lines are the original dipole field. The solid lines
are magneticfields modifiedby external currents. IMF stands for interplanetarymagneticfield, which is of solar origin. Major features of the
magnetosphere are shown. (RE earth radius.)
1230 MAGNETOSPHERE
by the pressure balance between the solar wind and the that produce flares and coronal mass ejections (CME).
planetary magnetic field. On an average day Earth’s Spectacular auroral displays and intense radio emis-
--
magnetopause at local noon (subsolar point) crosses sions that occur in the polar regions of the planet are
the equatorial plane at 1 0 R ~ (average Earth radius, manifestations of space storms. The dancing lights of
-6367 km), and at 20 RE in the dawn and dusk aurora are atmospheric emissions excited by precip-
sectors. In the antisunward direction, the magneto- itating energetic electrons that bombard the Earth’s
sphere has a magnetic tail. The geomagnetic tail outer atmosphere. The radio emissions are generated
extends beyond 100 RE. by the unstable auroral particles.
As the Sun’s coronal atmosphere expands into space Particles with millions of electron volt (MeV)
the wind speed increases, and near Earth’s position it is energies are frequently produced during large space
-400 km s-’. This is faster than the speed of Alfven storms. These penetrating particles can impact on
waves in the solar wind medium. Alfven waves, named mankind as they can disrupt communication, impair
after the Swedish Nobel laureate Hannes Alfven, are satellite instrumentation and even cause damage to
transverse magnetohydrodynamic waves travelling in spacecraft. Another havoc is that currents of several
the direction of the ambient magnetic field. As with million amps flow in the ionosphere during these
objects that travel faster than the sound speed in the storms. These ionospheric currents induce strong
terrestrial atmosphere, a shock wave forms in front of currents on the ground and have caused power outages
Earth’s magnetosphere. The Alfven wave steepens in cities located in the auroral zone. A new practical
nonlinearly and a shock forms as the magnetosphere goal of magnetospheric research is to learn to forecast
plows through the super-Alfvenic solar wind. The space weather so as to forewarn when disruptive
Alfven Mach number M A is about 8, which makes the storms will occur and to predict which storms accel-
Earth’s shock wave a strong shock. The shock wave erate particles to MeV energies.
is detached and separated from the magnetopause by Particles in space rarely collide, because the density
-1 RE. is so low and the mean free path so long. For example,
The region behind the shock wave is called magneto- in the outer magnetosphere there are ten thousand or
sheath and it extends to the magnetopause. The so particles per cubic meter and in the solar wind a few
magnetosheath is a turbulent region permeated by million. The mean free path is of the order of an
large amplitude waves and hot particles that have been astronomical unit, 1.5 x 10l1m. This unique feature
created in the shock formation. As in ordinary shocks, of the space system makes it different from other
the solar wind stream energy is converted to the particle systems where collisions are prevalent.
thermal energy. The solar wind speed in the down- Many space phenomena are driven by ‘collisionless’
stream region just behind the shock is much reduced, processes that involve collective interactions through
-50kms-l. The solar wind speed picks up again the long-range electromotive force. Very little is
further downstream. known about these collective processes. Even though
Inside the magnetosphere, Van Allen radiation belts, it is known that the disturbed solar wind fuels auroral
named after their discoverer, James Van Allen, are and magnetic storms, the actual mechanism of how
found. These are divided into inner (ionosphere to the solar wind mass, momentum, and energy are
-4-5 RE) and outer (-4-5 RE to the magnetopause) transported across the magnetopause is not yet
radiation belts. The inner radiation belt energetic understood. The Earth’s shock wave is a collision-
particles come from neutrons produced by cosmic rays less shock and, like ordinary shocks, it dissipates
that bombard the planet’s atmosphere. Neutrons are energy. But the collisionless dissipation mechanism
unstable and have short lifetimes and they decay in is very different from the classical dissipation
flight into protons, electrons, and antineutrinos. The mechanism where viscosity is produced by the
charged particles are captured by the planetary mag- colliding particles. How viscosity is produced in a
netic field. The primary source of Earth’s energetic collisionless process is a fundamental problem yet to
population in the inner radiation belt comes from be solved.
these cosmic ray albedo neutron decay (CRAND)
particles.
The source of the outer radiation belt particles is tied
Basic Equations
to solar wind and auroral disturbances which are
dynamic. The outer magnetosphere is sometimes Understanding magnetospheres requires knowledge
quiet, sometimes stormy, like the weather in the lower of how EM fields interact with charged particles
atmosphere of Earth. But unlike terrestrial weather, and how large-scale currents are generated. The
‘space weather’ is driven by electrical forces powered fundamental equations that describe the physics
by the disturbed solar wind connected to solar storms of magnetospheres are the Maxwell equations of
MAGNETOSPHERE 1231
electrodynamics, particles are interacting through the long-range elec-

tromagnetic field.
V.B =0 PI Our present knowledge of the magnetosphere is
based primarily on the synthesis of various pieces of
aD
VXH=J+- magnetospheric elements that have been studied. A
at quiet-time picture of the magnetosphere based on
V-D=p [31 time-independent formulation is given below. This
simple model is a reasonable starting point for
aB describing the more real magnetospheres.
VXE=-- i41
at
Magnetic Field
B is magnetic induction and is related to the magnetic The equations that govern the time independent
field intensity H by the constitutive relationship magnetic field are
B = poH, where ,uo is the magnetic permeability of
free space and equal to 4 ~ x l O - ’ H m - ~ ,D is the V-B=0 [sal
electric displacement vector related to the electric field
E by the constitutive relationship D = EOEwhere EO is VXH=J [8bI
the dielectric constant of free space equal to 8 . 8 5 ~
F m-l, p is the charge density and J is the current The solution of these equations is
density. The equations for p and J are given by
P=Cg AV where use was made of B = poH. Equation [9] is Biot-

Savart’s law and it states that given the current density
J=CS J at r, we can obtain the magnetic field B everywhere,
AV or, given B, we can invert the equation and obtain the
source current J responsible for the magnetic field.
where q k and vb are charge and velocity of the kth
Equation [9] is difficult to use for the magneto-
particle. The summation is carried over a suitably
sphere as a whole, because the particles contributing to
chosen small volume AV.
the source term J are not completely understood and
The velocity of the kth particle is obtained from the
they have not been measured for all regions. Also,
Lorentz equation of motion,
spacecraft-borne magnetometers have measured B
over a large region of the magnetosphere but the
[71 magnetosphere is dynamic and single-point measure-
ments made at different times are not easily related.
There are as many equations of motions as there are Information on J can be obtained if V x B can be
particles and they are coupled through the electro- measured. This requires simultaneous measurements
magnetic fields. Equations [ 11-[7] plus the constitutive with identical instruments from multiple spacecraft
relationships define a system of charged particles and that can measure gradients of the magnetic field. This
EM fields. The physics of magnetosphere studies the is one of the primary goals of the Cluster mission to be
science of large-scale electromagnetic dynamics with launched in the summer of 2000 by the European
an extremely large number of particles, or alternative- Space Agency (ESA). It is anticipated that Cluster
ly it can be viewed as a branch of statistical physics of experiments will definitely improve our knowledge of
charged particles driven by electromagnetic forces. magnetospheric currents.
The Dipole Field
Steady-State Magnetosphere We begin with the dynamo current interior to the solid
For an understanding of how magnetospheres are Earth, which is the source of the geomagnetic field.
formed, and of the physics of the interaction that The dipole field is a good approximation for describ-
induces external currents strong enough to deform ing the relatively stable magnetic field in the lower
and modify the planetary dipole field and excite the region of the magnetosphere. The dipole configuration
dynamic activity inside the magnetosphere, eqns [11- also provides a standard of reference for many other
[7] need to be solved self-consistently. But this is not planets and celestial bodies.
yet possible. Even though the density in space is very Let the dipole moment of the planet be given by M.
small, a magnetosphere occupies a large volume and In regions outside the planet, there are no currents,
1232 MAGNETOSPHERE
hence J = 0. We can then let B = -VY, where Y is the because the centered dipole is symmetric about 4.
magnetic scalar potential of a dipole given by The magnitude of the field at ( r ,A) is obtained from
[ 111 and is given by
PO
=-M.V-
1
471 r [lo1
Assume that Earth has a centered dipole moment M
(Figure 2 ) . In spherical coordinate system with M = which shows r P 3 dependence of the dipole field
-Mi ( - for Earth),the three components of the dipole strength. A dipole has the strongest field at the pole
magnetic field are (A = n/2) and the smallest on the Equator (A = 0).
Another useful relation is the equation of the locus of a
Br po M sin A dipole field given by
-
2 IT r3
r = ro cos2Aand4 = do ~ 3 1
/.LO M COS 2 Here ro and $o are distance and longitude of the line of
A -
4 .n r3 force at A = 0.
The dipole approximation is fairly good until about
B4 = 0 PI1 4-5 RE from the Earth. Further out, the magnetic field
where for Earth M = 8 x A m2, 1. is the latitude begins to depart as the other current sources become
( A = 0 at the Equator) and r is the radial distance. B$ = 0 important.
Magnetopause
Consider now the dayside boundary of the magneto-

sphere. For simplicity, consider the boundary to be a
plane and assume the solar wind consists of electrons
and protons. The planetary magnetic field is in vacuum
(no particles in the magnetosphere) and the solar wind
is not magnetized (no interplanetary magnetic field).
The solar wind particles that impinge on this boundary
are deflected by the Lorentz force. This creates a
boundary current running from east to west and
modifies the planetary field.
In Figure 3, the total magnetic field B , produced by
the current just inside the boundary can be estimated
as
where Bd is the dipole field of the planet at that point

and B s is the surface magnetic field produced by the
magnetopause current. Just outside the boundary,
Therefore,
This result states that the current at the magnetopause

will produce a magnetic field whose intensity is twice
the value of the undistorted dipole field at that point. It
is based on an ideal model which stipulates the
planetary field to totally vanish inside the solar wind.
X
J The magnetopause boundary separates the two
domains completely and there is no normal compo-
Figure 2 The top diagram shows the contours of the planetary
dipole magnetic field. The magnetic moment for Earth points from nent of the magnetic field in this model. This magne-
north to south. The bottom diagram shows the Earth-centered tosphere is essentially the model proposed by S.
coordinate system in which the dipole field is defined. Chapman and V. C. A. Ferraro in 1931 to explain
MAGNETOSPHERE 1233
total energy density of the particles perpendicular to

Solar wind Magnetosphere the surface is 2NmV?, cos2 5. At the magnetopause
(vacuum) boundary, we require a balance between the solar
wind and planetary magnetic energy densities. Hence,
2NmVzw cos2 5 = B2/2po
Typical values are V,, = 400 kms-l and N = S X
lo6 mP3. Equation [17]then yields B E 7 0 n T for the
subsolar point, [ = 0. From equation [16], we see that
in the absence of the solar wind, the undistorted dipole
X ,f field has a value approximately 35 nT. The surface
N
field of Earth is - 0 . 3 2 ~lop4T. Scaling this as l/r3, we

deduce the location for the magnetopause to be
Boundary -9.2 RE. Observations show the subsolar magneto-
pause position varies typically between N ~ R and E
0
Bd 11RE.
Geomagnetic Tail
Solar wind
particles The solar wind imparts momentum on the planetary
magnetic field and creates a current such that when
superposed on the dipole field, the resulting field on the
x7 Y
antisunward direction has a tail-like geometry (Figure
1).The equation V x H = J shows that a line or sheet
current in the x-direction with the current flowing
from dawn to dusk in the noon-midnight plane is
Boundary
needed. Several functional models have been proposed
Figure3 The top sketch shows a boundary in the noon-midnight to account for the tail geometry. One model of this
plane that separates the solar wind and the magnetosphere. The current is given by
current at the boundary is out of the page.The bottom sketch shows
how the current is set up by the turning around of the solar wind
particles due to the Lorentz force.
where Bo is the magnitude of the magnetic field at the

terrestrial magnetic storms. This magnetosphere is outer boundary and L the half-thickness in the z-
‘closed’ and particles cannot enter the magnetosphere. direction. This current is uniform in the y-direction but
If the boundary supports a normal component of the has a z-dependence. The magnetic field deduced from
magnetic field, it can lead to an ‘open’ magnetosphere eqn [ 181 yields
which allows the particles to enter the magnetosphere z
from the solar wind (see ‘Dynamic Magnetosphere’ B, = Bo tanh - 1191
L
below).
If the boundary is curved then the same procedure The magnetic field is directed along the x-direction and
can be used, but now currents that give curved field increases with z. The magnetic field vanishes at z = 0.
configurations need to be taken into account. Another This is the magnetic field free line (neutral line) that
feature not incorporated in this simple pedagogic results from the fields above and below that point in
model is that the real magnetopause includes a opposite directions. While functional models provide
boundary layer. How boundary layers can be pro- a useful picture of how the tail might be formed, the
duced in collisionless plasmas is still not understood. structure of the real tail current is very complicated.
We can estimate where the outer boundary is Although it is attributed to the solar-wind-geomag-
located. Let the solar wind be specularly reflected at netic-field interaction, how the current is produced
the boundary. The transfer of momentum per particle and maintained is still unknown.
per collision is 2mV,, cos 4, where m is the mass of
Electric Field
H+, V,, the velocity of the solar wind, and 5 the angle
of incidence. The number of particles striking a unit Understanding the origin of electric fields in magne-
area of the boundary per second is NV,, cos 5, where tospheres is important because they can change
N is the number density of the solar wind. Thus, the particle energies and also alter their trajectories. For
1234 MAGNETOSPHERE
example, in the presence of an electric field perpen- (unprimed) and c is the speed of light. Magnetic fields
dicular to the direction of the magnetic field, the measured in the two frames are nearly equal, but
particles drift across the magnetic field (further electric fields are different. Thus, it is necessary to
discussion below). specify the coordinate frame in which the measure-
The two Maxwell equations governing the electric ments are made.
field for static magnetospheres are An important result of the Lorentz transformation is
that if the velocity of the moving frame is given by
V-D=p POaI
ExB
V=-
VXE=O PObI B2
The electric field can be defined in terms of the scalar then E’ = 0 and no work will be done on the particles
potential $ , E = -V$, which inserted in eqns [20] in that moving frame. This has important application
yields Poisson’s equation, for plasma systems in motion. Consider, for example,
the motion of an arbitrarily shaped system of charged
02$ =p PI1 particles in the magnetosphere. Lorentz’s result states
that if the motion of this plasma is given by eqn [24],
whose solution is then E’ = 0. Thus, the electric field vanishes in the
moving frame which is also the plasma frame of
=
1
GG 1-p(r’)d3r reference. Equation [23a] then states
E=-VxB PI
where r’ is the location of the charge density and the
potential is evaluated at r. This is the Coulomb The electric field in the rest frame is related to the
potential that results from the charge density p . vector product of the velocity of the plasma system and
Electric field is obtained by taking the gradient of $. the magnetic field.
Although magnetospheres are populated by charged Equations [24] and [25] are equivalent expressions.
particles, free charges p are not maintained (charges do These equations state that if there is an electric field,
not accumulate in good conductors). Thus, eqn [22] is the plasma will move. Equivalently, if the plasma is
not useful when considering large-scale electric fields moving, there is an electric field. In a manner analo-
in space. gous to the convective motion in fluids which arises to
Large-scale electric fields in magnetospheres are equalize the nonuniform temperature, the plasma
produced mainly by inductive effects. Michael Fara- motion arises to transform away the electric field
day in 1831 showed that an electromotive force (charges) in the moving frame (plasma frame) because
(e.m.f.) is induced when magnetic flux changes in free charges cannot be maintained in good conductors.
time or in space. In the magnetosphere, the inductive In that sense, the plasma motion in eqn [24] is referred
field comes from motions of a plasma across a to as a convective motion.
magnetic field or rotation of a magnetized planet If the coordinate frame is spinning, this frame is not
through a plasma medium. These motions induce a an inertial frame and the Lorentz transformation
motional e.m.f., which is the primary source of large- theory is not valid. General relativity effects must be
scale electric fields in space. taken into account. We state only the results here. The
relationships of EM fields are given by
Motional Electric Fields
E” = E + V X B [26aI
Electric and magnetic fields are measured on moving
platforms through a plasma medium which itself may
B” = B P6bI
be in motion. The relationship of electric and magnetic
fields in the rest and moving frames of references is where the double prime (”) denotes the spinning frame
given by the Lorentz transformation. For linear of reference, V = o x r and o is the angular frequency
motions and nonrelativistic case (V/c << l),the rela- vector. Equations [23] and [26] have the same form,
tions are except that eqns [26] are exact to all orders in V/c,
whereas eqns [23] are only approximate.
E’=E+VxB ~ 3 4
Electric Field at the Magnetopause
B’xB 1 2 3 ~In closed magnetospheres, Maxwell equations require
Here the prime (’) denotes a moving frame and V is the tangential component of the electric field to vanish
the velocity of that frame relative to the rest frame at the boundary, in addition to the requirement of the
MAGNETOSPHERE 1235
absence of a magnetic normal component. The pres- Van Allen Radiation Belts
ence of a tangential electric field will allow particles to
drift across the boundary which is not permitted in The motion of a charged particle is governed by the
closed models. This contrasts the open model which Lorentz equation of motion given in eqn [7].We can
invokes a magnetic field merging process (see ‘Dy- study the behavior of a single particle using this
namic Magnetosphere’ below). In closed models, equation, neglecting the presence of other particles.
particles can cross the boundary, for example, by This test particle approach will give us a sense of how a
diffusive mechanisms. charged particle moves around in magnetospheres and
how Van Allen radiation belts are formed. Since we are
Electric Field in the Plasma Sheet ignoring the collective effects of interacting particles,
information on the dynamic behavior of magneto-
The geomagnetic tail is populated by plasmas of the
spheric particles in aurorae is not revealed.
solar wind and the ionosphere. When these plasmas
Consider first the motion of a particle in a time-
move, we can apply the above results to study the
independent magnetic field and assume there is no
behavior of the electric field. Although the motions are
electric field (E = 0). The equation of motion is then
generally turbulent, we will assume the motion is
mdv/dt = qv x B. In inhomogeneous magnetic fields,
laminar. Consider an observer on the equatorial plane
B = B(r), this differential equation yields three types
at rest observing the plasma moving toward the Earth.
of motion. The first is the cyclotron motion around the
Equations [23] can be used to estimate the magnitude
magnetic field. The cyclotron frequency of this circular
and direction of the electric field. For example, in the
motion is o,= qB/m and the cyclotron radius is
noon-midnight meriodinal plane, V is earthward and
r, = mv,/qB. Here u l is the magnitude of the particle
B is upward, and thus the electric field E points from
velocity perpendicular (I)to the magnetic field
dawn to dusk. The dawn-dusk electric field measured
direction. In terms of the pitch a of the particle, the
during quiet solar wind and geomagnetic field is
angle between the particle velocity v and the magnetic
N 0.3x lop3V m-l. The origin of this large-scale
~ z, sin ci.
field B, z , =
dawn-dusk field is thought to be the solar wind, but
The second type of motion comes from particles
the mechanism for establishing this field across the tail
with ci # n/2. These particles can move along the
of the magnetosphere is not known.
direction of the magnetic with V I ( = z, cos a. Here (1
denotes parallel to B. In an inhomogeneous magnetic
Electric Field in the Plasmasphere
field such as the dipole field, the field becomes stronger
The Earth’s rotation is important for particles in the as the particle gets closer to the planet. The particle
lower region of the magnetosphere. This region, orbit winds tighter as it approaches the stronger field
dominated by the dipole field and populated mostly region and it also encounters a force that pushes the
by ionospheric plasmas, is called plasmasphere. The particle back in the direction from which it arrives.
plasmasphere starts from the ionosphere and extends This results in ‘mirroring’ of the particles, which
out to about 4 - 5 R ~in the equatorial plane. It thus bounce back and forth between the Northern and
includes the inner radiation belt. The boundary of the Southern Hemispheres.
plasmasphere is called plasmapause. Outside this The third type of motion comes from particles
boundary is the outer radiation belt. travelling on magnetic fields that are curved and not
The rotation of the planet induces an electric field in uniform in the radial direction. Particles travelling on
the plasmasphere. Consider an observer at rest at curved magnetic fields experience a centrifugal force.
position r on the equatorial plane. Noting the plasma Particles in fields that are not uniform experience a
is rotating with the planetary angular frequency they continually changing cyclotron radius. Both of these
apply eqns [26] to calculate the induced electric field. effects result in particle drifts in the azimuthal direc-
Since the electric field in the rotating frame (plasma tion, westward for positively charged particles and
frame) vanishes, they use eastward for negatively charged particles. The curva-
ture and gradient drift velocities are energy-depend-
E = -(a X r ) x B ~ 7 1 ent, with higher energy particles drifting faster.
In the presence of electric fields, the motions
to find the direction and magnitude of E. The result described above must now be augmented with the
shows the electric field is radially outward. Induced effects of electric fields. Consider a time-independent
electric field due to the rotation of the planet is also electric field given by E = Eli + EL, where the /I and I
known as the corotational electric field. The Earth’s are directions relative to the direction of the magnetic
corotational electric field is N 0 . 5 ~ 1 0 - ~ V m at
- ~ field. The effect of Ell is to accelerate or decelarate
ionospheric heights. particles travelling along the magnetic field. EL gives
1236 MAGNETOSPHERE
rise to a drift in the azimuthal direction given by eqn -4& strong enough to affect the terrestrial magnetic
[24]. A peculiar feature here is that the drift is the same field measured on the ground. In the polar ionosphere,
for + and - particles and also independent of the brilliant and wild auroral displays luminate the night
mass, charge, and energy of the particles. sky accompanied by the roaring of natural electro-
In summary, the total motion of particles in mag- magnetic radio waves that are emitted over millihertz
netospheres consists of a superposition of the cyclo- to megahertz frequencies. The geomagnetic tail wags
tron motion, bounce motion and the drift motion. In and flaps like a wind sock on a windy day. An observer
the absence of Ell these are energy-conserving motions. in the interplanetary space sees the magnetosphere
-
In the outer radiation belt at synchronous altitudes
--
( 6 . 6 R ~geocentric) for example, the electron cyclo-
tron frequency is 1kHz, the bounce period for
soaring through the heliosphere with bright flickering
lights, resembling a comet.
Unable to solve eqns [I]-[7] in a self-consistent way,
40 keV electrons is 1s, and the drift period is -2 space researchers have thus far obtained only approx-
hours. For the ions, the cyclotron frequency is smaller imate solutions. For this reason, the picture of the
by the ratio of electron to ion mass, the bounce magnetosphere is incomplete. The approach most
frequency by the square root of the mass ratio, and the commonly used has been to treat the collisionless
drift times are same for the same energies. space plasmas as a magnetohydrodynamic (MHD)
The lifetime of the magnetospheric particles is fluid and then solve the mass, momentum, and energy
determined by how close they approach the planetary conservation equations using the variables of density,
atmosphere. If they approach close to the planet where velocity, and temperature. MHD theory has provided
the atmospheric densities are sufficiently high, so that insight into the ways the solar wind flows and how
they collide with them, then these particles will be lost magnetospheres respond to the solar wind. An impor-
into the atmosphere. These are called precipitated tant paradigm of the MHD theory is that interplan-
particles and they are the source of atmospheric etary and Earth’s dipole magnetic fields can ‘merge’,
emissions responsible for aurorae at high magnetic which opens the magnetosphere so the solar wind can
latitudes. If the particles bounce at sufficiently high enter. This innovative concept was introduced by
altitudes where collisions are infrequent then these James W. Dungey nearly 40 years ago. The physics of
particles can persist for a long time and drift around merging requires a kinetic treatment which has not yet
the magnetosphere many times. These particles are been solved.
trapped particles and they form the Van Allen radia- A limited number of dynamical processes have
tion belts. been studied from the kinetic point of view, assuming
that an ensemble of collisionless particles can be
defined in terms of the distribution function. One then
solves the Boltzmann transport equation coupled to
Dynamic Magnetosphere the EM equations. This approach has been most
Contrary to the static magnetosphere described fruitful in the study of the microphysics of wave-
above, the real magnetosphere is time-dependent and particle interactions and instabilities. Progress has
very dynamic. As an example, we describe phen- been made towards understanding the microphysics of
omenologically what happens when the solar wind is auroras observed at ionospheric heights, and much has
moderately disturbed. The magnetopause boundary been learned about the structure of aurorae, how
moves in and out in response to the solar wind particles are accelerated in the ionosphere and how
variations. The boundary is no longer smooth but certain types of radio emissions are excited. The full
is modulated with surface waves, reminiscent of kinetic formulation has not however been applied to
atmospheric and ocean waves in stormy weather. problems of large-scale global spatial structure and
These magnetopause waves could be excited by the dynamics, and this will remain at the forefront of
Kelvin-Helmholtz instability mechanism owing to magnetospheric studies.
the presence of the large solar wind velocity shear Among the important dynamic problems to be
across the boundary. Inside the magnetosphere, studied in magnetospheres include how the solar wind
particles are injected from the plasma sheet into gets into the magnetosphere, how the particles are
the outer radiation belt, and trapped Van Allen accelerated in the magnetosphere, how electric fields
particle intensity and energy increase by orders of are set up in the geomagnetic tail, how the magneto-
magnitude. This happens when the tail current sphere and the auroral ionosphere are coupled, how
abruptly disrupts and the geomagnetic field returns global currents can be generated during auroral storms
for a short time to the dipole shape. The trapped to reconfigure the entire magnetosphere and how some
particle intensity increases considerably during mag- of the solar wind energy is captured to produce the
netic storms and the particles form a ‘ring current’ at global aurora.
MESOSCALE METEOROLOGY I Overview 1237
See also Hess WN (1968) The Radiation Belt and Magnetosphere.

Waltham, MA: Blaisdell.
Ionosphere.Solar Winds. Kamide Y (1988) Electrodynamic Processes i n the Earth’s
Ionosphere and Magnetosphere. Kyoto: Kyoto Sangyo
Further Reading University Press.
Kivelson MG and Russell CT (eds) (1995)An Introduction
Alfven H and Falthammar CG (1963)Cosmical Electrody- to Space Physics. Cambridge: Cambridge University Press.
namics, Fundamental Principles, 2nd edn. Oxford: Parker EN, Kennel CF and Lanzerotti LJ (eds) (1979)Solar
Oxford University Press. System Plasma Physics. Amsterdam: North Holland.
Baumjohann W and Treumann RA (1996) Basic Space Parks GK (1991)Physics ofspace Plasmas, An Intvoduction.
Plasma Physics. London: Imperial College Press. Reading, MA: Addison-Wesley.
Hargreaves JK (1992) T h e Solar-Terrestrial Environment. Uberoi C (2000) Earth’s Proximal Space. Hyderabad:
Cambridge: Cambridge University Press. Universities Press.
Contents
Overview
Cloud and Precipitation Bands
Mesoscale Convective Systems
Models
Since it spans a range of phenomena and processes,

Overview the mesoscale is often subdivided into the rather
unmemorable categories of
D J Parker, University of Leeds, Leeds, UK
0 meso-a: 200-2000 km (e.g., secondary frontal cy-
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. clones),
0 meso-b: 20-200 km (e.g., mesoscale convective
Introduction systems),
0 meso-y: 2-20 km (e.g., cumulonimbus cells).
-
The mesoscale is an intermediate regime between the
larger, synoptic scale and the smaller scales of bound- Since these categories are in order of roughly increas-
ary layers, turbulence and micrometeorology. The fact ing Rossby number (defined as Ro = UlfL, with U and
that the mesoscale is an intermediate regime is L the velocity and length scales and f the Coriolis
reflected in the physical processes that influence parameter), they correspond to a decreasing impor-
mesoscale systems. While synoptic-scale systems can tance of planetary rotation and balanced dynamics.
be regarded as being close to geostrophic balance, and The relegation of mesoscale systems to a transition
smaller-scale systems are variously independent of zone between larger and smaller scales of the atmos-
Coriolis force, compressibility or nonhydrostatic phere is to belie the fact that the mesoscale exhibits a
effects, at the mesoscale we cannot immediately majority of the interesting, observable weather sys-
neglect any components of the equations of motion, tems. These include cumulonimbus cells and com-
except for effects associated with the Earth’s curva- plexes, a variety of frontal structures, and various
ture. In a sense, the mesoscale could be defined as coherent flows relating to topography or coastal
the regime in which all the components of the regions. These systems are fascinating as remarkably
dynamic and thermodynamic equations may be im- coherent and repeatable structures in a nonlinear and
portant. By necessity, this article refers in only super- complex dynamical regime of the atmosphere, but
ficial detail to a wide range of phenomena and are also important practically, being responsible for
processes, each of which is discussed in more detail damaging storms (such as squall lines, supercells,
in separate articles. mesoscale convective complexes, secondary frontal
MESOSCALE METEOROLOGY I Overview 1237
See also Hess WN (1968) The Radiation Belt and Magnetosphere.

Waltham, MA: Blaisdell.
Ionosphere.Solar Winds. Kamide Y (1988) Electrodynamic Processes i n the Earth’s
Ionosphere and Magnetosphere. Kyoto: Kyoto Sangyo
Further Reading University Press.
Kivelson MG and Russell CT (eds) (1995)An Introduction
Alfven H and Falthammar CG (1963)Cosmical Electrody- to Space Physics. Cambridge: Cambridge University Press.
namics, Fundamental Principles, 2nd edn. Oxford: Parker EN, Kennel CF and Lanzerotti LJ (eds) (1979)Solar
Oxford University Press. System Plasma Physics. Amsterdam: North Holland.
Baumjohann W and Treumann RA (1996) Basic Space Parks GK (1991)Physics ofspace Plasmas, An Intvoduction.
Plasma Physics. London: Imperial College Press. Reading, MA: Addison-Wesley.
Hargreaves JK (1992) T h e Solar-Terrestrial Environment. Uberoi C (2000) Earth’s Proximal Space. Hyderabad:
Cambridge: Cambridge University Press. Universities Press.
Contents
Overview
Mesoscale Convective Systems
Models
Since it spans a range of phenomena and processes,

Overview the mesoscale is often subdivided into the rather
unmemorable categories of
D J Parker, University of Leeds, Leeds, UK
0 meso-a: 200-2000 km (e.g., secondary frontal cy-
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. clones),
0 meso-b: 20-200 km (e.g., mesoscale convective
Introduction systems),
0 meso-y: 2-20 km (e.g., cumulonimbus cells).
-
The mesoscale is an intermediate regime between the
larger, synoptic scale and the smaller scales of bound- Since these categories are in order of roughly increas-
ary layers, turbulence and micrometeorology. The fact ing Rossby number (defined as Ro = UlfL, with U and
that the mesoscale is an intermediate regime is L the velocity and length scales and f the Coriolis
reflected in the physical processes that influence parameter), they correspond to a decreasing impor-
mesoscale systems. While synoptic-scale systems can tance of planetary rotation and balanced dynamics.
be regarded as being close to geostrophic balance, and The relegation of mesoscale systems to a transition
smaller-scale systems are variously independent of zone between larger and smaller scales of the atmos-
Coriolis force, compressibility or nonhydrostatic phere is to belie the fact that the mesoscale exhibits a
effects, at the mesoscale we cannot immediately majority of the interesting, observable weather sys-
neglect any components of the equations of motion, tems. These include cumulonimbus cells and com-
except for effects associated with the Earth’s curva- plexes, a variety of frontal structures, and various
ture. In a sense, the mesoscale could be defined as coherent flows relating to topography or coastal
the regime in which all the components of the regions. These systems are fascinating as remarkably
dynamic and thermodynamic equations may be im- coherent and repeatable structures in a nonlinear and
portant. By necessity, this article refers in only super- complex dynamical regime of the atmosphere, but
ficial detail to a wide range of phenomena and are also important practically, being responsible for
processes, each of which is discussed in more detail damaging storms (such as squall lines, supercells,
in separate articles. mesoscale convective complexes, secondary frontal
1238 MESOSCALEMETEOROLOGY/ Overview
cyclones, polar lows, and downslope windstorms), of motion can be cast in different forms, depending on
rainfall distribution, and significant inhomogeneity in the choice of vertical coordinate: here they will be
local weather conditions. In addition, cumulonimbus stated in physical height coordinates. The vector
convection and active fronts are the two families of momentum equation is
systems responsible for rapid vertical transport of
trace gases from the boundary layer to the upper
troposphere and lower stratosphere: although such
systems are poorly resolved or parameterized in
global-scale models, they are critical to global-scale where u is the vector velocity, k the upward unit vector,
chemical transport. g the acceleration due to gravity, p the density, p the
Increasing computer power over recent decades has pressure, and ti, the Reynolds stress tensor, with the
led to a rapid increase in the possibilities for numerical summation convention assumed for this last term. The
simulation of mesoscale phenomena. In the area of system also requires a thermodynamic equation,
cumulonimbus convection, for example, simulations
of evolving cloud systems have been able to explore the -DO= s 121
sensitivity of such systems to environmental parame- Dt
ters, with useful benefit to forecasting. As remarked where O is the potential temperature and S is a
above, these kinds of simulations tend to contain all thermodynamic source term (say due to radiative
the components of the dynamic and thermodynamic flux convergence or phase changes of water and
equations, since they relate to a scale where nothing including a term according to the divergence of the
can be wholly neglected, and are often termed ‘full turbulent flux of potential temperature) and conser-
physics’ simulations. Indeed, simulations of mesoscale vation of mass, or continuity,
convective systems are performed using large eddy
simulations incorporating explicit cloud microphysics -DP
+pv.u=o
as well as the influence of the Coriolis acceleration. Dt [31
Until recently, the resolution of global-scale models
has meant that representation of mesoscale systems Turbulence Closure
has by necessity involved subgrid parameterization.
Currently, however, the horizontal resolution of oper- The momentum and thermodynamic equations in-
ational models is intruding into the mesoscale, and we volve turbulent fluxes on the right-hand side, which
are left with a delicate balance between what is need to be obtained by some form of turbulence
resolved and what is parameterized: if a model has a closure. Although a great deal of effort can be
resolution of 40km and the coherent convective expended on dealing with the turbulent terms, there
structures are of a similar scale, they are neither well are no schemes which are entirely satisfactory. The
resolved nor well parameterized. most reliable approach for numerical modeling is
Despite the increasing possibilities of mesoscale regarded to be large eddy simulation, in which, crudely
numerical modeling, the observed coherence of meso- speaking, a spatial resolution is chosen to resolve the
scale structures points to the importance of idealized anisotropic scales of eddies, with the assumption that
and intuitive interpretations of the atmospheric dy- smaller (isotropic) scales can be dealt with effectively
namics, and this remains an active area of research. by a relatively simple turbulent scheme. In practice,
The simple model of the density current, for example, the resolutions required by this method, on the order
is very efficient at obtaining robust estimates of the of meters, mean that computational demands are
propagation characteristics of small-scale fronts, is extremely high.
intuitively easy to understand, and requires some For some flows, such as small-amplitude buoyancy
considerable effort to better with a numerical model. waves, the turbulence terms will be very small, but
most mesoscale systems involve turbulent mixing to an
important degree.
Mathematical Description of Different
Mesoscale Regimes Simplifications to the Continuity Equation
The only universal simplification appropriate to all In most instances the continuity equation may be
mesoscale systems is to neglect effects of the Earth’s simplified using the assumption of subsonic flow, to
curvature and consider an f-plane version of the the anelastic form,
‘primitive equations’ (which represent a shallow layer
on the quasi-spherical Earth). The resulting equations v . (p,u) = 0 141
MESOSCALEMETEOROLOGYI Overview 1239
in which p,(z) is a reference density which depends tions of the form

only on height, z. If, as in the case of a sea breeze
current, the vertical scale of the motion is small relative Du
- + f k x u = g’k - Vq5 + -
azij
to the scale height, Dt axj PI
RT
H=--8km
g
PI where q5 = p’/pr is the modified pressure and g’ is the
buoyancy.
where R is the gas constant for air and T is a mean Semi-geostrophic Theory
temperature for the atmosphere, it is possible to use
the incompressible form, The semi-geostrophic (SG) model is a refinement of
quasi-geostrophic (QG) dynamics, to allow for short
v.u=o [61 lengthscales in the region of a synoptic front. This
There are few instances where one of these approxi- approach, although based on balanced dynamics, is
mations will not be valid, but it should be recalled that thought to work effectively in quasi-two-dimensional
deep waves in the atmosphere have speeds which scale frontal zones down to scales of tens of kilometers, and
approximately with the vertical wavelength. In a is therefore one of the few areas where analytical
quiescent atmosphere of constant static stability, theory has been able to bridge the mesoscale regime.
the maximum group and phase speeds of buoyancy Broadly, SG frontal dynamics involves an along-front
waves are component of the wind, vg, in geostrophic balance
with the cross-frontal pressure structure, but also
N
c=- 171 accommodates advection by the ageostrophic cross-
m L ,
frontal and vertical winds (a feature that is absent from

where N is the Brunt-Vaisala frequency and m Q G dynamics). This is made analytically more trac-
is the vertical wavenumber: for a wave with a single table by transformation to a geostrophic momentum
maximum in the troposphere this can be of the order coordinate,
of 50 m s - I, while for a wave with a first maximum x = x + v-g [91
at the tropopause it can be 1 0 0 r n ~ - ~Further . f
refinements to the continuity equation exist, and
have been proposed to take into account such very where x is the cross-front direction. The transforma-
deep motions. tion between coordinate systems can develop a singu-
larity in finite time, and mathematically it is this which
Hydrostatic Balance leads to the formation of frontal singularities in the
model; physically, it is the process of advection by the
It is often possible to assume a state of hydrostatic
ageostrophic winds which leads to the enhanced
balance, provided the vertical scale of the motion is
frontogenesis.
significantly smaller than the horizontal scale, which
SG models are based on a thermal wind balance, and
tends to imply the meso-cr regime. This is, for example,
can conveniently be solved by inversion of a conserved
applicable to most scales of frontal dynamics, but not
(in the absence of diabatic processes) potential vorti-
to the circulations in a cumulonimbus storm. For
city function. This procedure fails when the potential
inertia-gravity waves, the applicability of hydrostatic
vorticity becomes negative, a condition which implies
balance depends on the aspect ratio of the waves, being
slantwise instability of the atmosphere.
suitable when the ratio of horizontal to vertical
Higher-order balance models have also been applied
wavenumbers, klm, is small.
to mesoscale phenomena such as organized convective
The Boussinesq Approximation systems. However, the prevalence of latent heating and
turbulent processes, as well as the difficulty of apply-
This approximation amounts to neglecting density ing a balance condition at high Rossby number, have
variations in inertia terms but retaining them from the meant that the use of potential vorticity concepts is less
point of view of buoyancy, which is valid when the common to mesoscale systems than it is in synoptic
vertical scale of the motion is much less than the scale analysis.
height, H. A basic state density field pr(z), dependent
only on height, is then incorporated into a modified
pressure function. This is very convenient in that it Tour of Mesoscale Phenomena
retains the simplicity of uniform, incompressible
Buoyancy Waves
dynamics, while accommodating buoyancy varia-
tions. There are several versions of the Boussinesq Often termed ‘gravity waves’, these patterns occur as a
approximation: typically it leads to momentum equa- universal response of the stably stratified atmosphere
1240 MESOSCALEMETEOROLOGY / Overview
to mesoscale perturbations. They propagate with waves act to modify the environment of the convec-
components in the vertical and horizontal, and are tion, through adiabatic warming and cooling, and are
modified strongly by the vertical structure of both one mechanism whereby the convection communi-
stratification and horizontal wind. For shallow cates its thermodynamic forcing into its environment.
waves, for which klm becomes small, the waves are The primary mode of response to the convective heat
increasingly influenced by the Coriolis terms, and are source is downward motion, warming the environ-
known as inertia-gravity waves. Two characteristic ment, but the vertical structure of the heating (includ-
regimes of solution are upward-propagating waves, ing forcing due to downdraught cooling) forces modes
which carry energy and momentum upwards away of different vertical structure. As in the case of
from the wave source, and trapped waves, where a orographic waves, there is an important possibility
wavelike region is bounded by an evanescent layer, of ducting if waves are trapped in the troposphere.
leading to ducting of the waves in the horizontal. Generation of buoyancy waves by orography and by
Although simple, linear buoyancy wave solutions for a convection, and the subsequent momentum flux con-
quiescent, Boussinesq atmosphere are relatively easy vergence when the waves break, are thought to be
to obtain, for real profiles it seems that a detailed significant processes in the global circulation, yet
consideration of the full vertical profile is necessary in neither source is well represented in global models.
order to diagnose the correct buoyancy wave charac- Diagnosis of the relevant wave regime is not easy, and
teristics. waves may propagate long vertical or horizontal
In regions of orography, vertically propagating distances, equivalent to many model grid lengths,
waves tilt backwards against the mean flow before breaking.
(Figure l A ) , and the resulting shift in the streamlines
contributes to downslope acceleration of the wind. In
Cumulonimbus Convection
conditions of trapping (Figure l B ) , lee waves can
propagate significant distances downstream and can Cumulonimbus storms are the most dramatic of cloud
often be seen in low-level cloud patterns. Such features, and are a vital component in the atmospheric
lee waves give rise to local regions of low-level circulation. Individual cumulonimbus cells occur on
convergence, which may occasionally initiate moist horizontal scales of around 10km and commonly
convection. extend to the tropopause, over a time scale of an hour
Deep convection forces waves propagating both or so. However, it is common for cumulonimbus
upward into the stratosphere and horizontally in the systems to self-organize into a mesoscale convective
troposphere: deep modes in the troposphere propagate system (MCS) with a significantly longer life cycle:
at speeds upward of 50 m s - and evidence of wave squall lines in west Africa can persist for 48 h and
trains forced by convection has been observed in propagate for thousands of kilometers across the
microbarograph arrays. The horizontally propagating continent.
Cumulonimbus storms involve rapid and active
microphysical transitions, on time scales of a few
minutes, yet can organize into a significant degree of
*-\-
geostrophic balance, so these systems can truly
encompass the whole mesoscale atmospheric regime.
In terms of mathematical representation, the strong
up- and downdraughts in these storms, combined with
intense turbulence and their great depth, mean that
none of the mathematical approximations outlined
(A) above can comfortably be applied (although the
anelastic approximation may be used if the fast
buoyancy wave response is not critical to the solution).
Cumulonimbus systems also involve other distinct
mesoscale flows as components, or responses to the
forcing. The buoyancy wave response has been
discussed above; other component flows include the
cold pool and gust front. Evaporation in the precip-
(B) itation-driven downdraught leads to the cold pool of
Figure 1 Schematics of streamlines over a hill in the cases of (A) air at the surface, and this propagates like a density
upward wave propagation and (B)wave trapping. In (A) the wave tilt current, of around 1km depth, into the environmental
with height contributes to a wind speed maximum in the lee. boundary layer. Forced lifting at the gust front acts to
MESOSCALEMETEOROLOGY/ Ovenriew 1241
Adiabats
/
Sea breeze
----------
(A) Sea Land
. Adiabats
Figure2 Plan view of an idealizedcumulonimbussystem. A new

cumulonimbus cell is often formed at the gust front, in a direction
determined by the low-level wind and shear structure.
- Barolinic overturning
\
Katabatic flow
trigger new convective cells (Figure 2). Although the
gust front spreads in all directions outward from its
source, the lifting is optimized in a given direction in
relation to the ambient wind, and this in turn feeds Figure 3 Baroclinic tendencies in horizontal vorticity lead to
back on the morphology of the cumulonimbus system. thermally driven flows: (A) the sea breeze and (B) a katabatic,
When two gust fronts from adjacent storm cells downslope wind.
collide, they produce enhanced lifting at the region
of collision. The regeneration of cumulonimbus
storms at the gust front is an intriguing way in which be relatively strong convective turbulence inland,
the most turbulent, nonlinear and irreducible of which tends to dissipate the sea breeze front and
atmospheric flows can be understood in terms of render it a more diffuse baroclinic zone. However, as
relatively simple, coherent components. the surface heating diminishes in the evening, the
In some cases, extremely intense downdraughts lead convective turbulence decays and the front may
to downbursts and microbursts, which produce ex- intensify and propagate further inland.
tremely intense gusts at the surface, and are a severe A density current in the atmosphere is in a state of
hazard for aircraft. The interaction between boundary balance between the pressure gradient force due to the
layer shear and differential lifting at the gust front is density change across the front, and drag on the
also thought to lead to precursor vortices for tornado current due to turbulent stresses (principally Kelvin-
development. Helmholtz instability at the head). However, the sea
breeze is also influenced by the Coriolis acceleration,
Sea Breezes and Other Thermally Generated Winds
over a time scale llf, and can be expected to turn with
the Coriolis acceleration as the day progresses. Sea
The sea breeze is an example of a flow generated by the breezes are quite sensitive to the larger-scale flow, and
relatively rapid generation of a baroclinic zone (den- will not develop if the ambient winds are strong.
sity gradient). Warming of the land surface after After sunset it is possible for a land breeze to
sunrise is rapid, while the sea surface temperature develop, as the land surface cools more rapidly than
remains almost constant, so the boundary layer inland the sea surface. However, the ensuing surface inver-
becomes relatively warm over the space of a few hours. sion, which suppresses turbulence, does not develop as
This leads to a baroclinic overturning (Figure3A) and deeply as the daytime convective boundary layer, so
the development of a flow resembling a density the land breeze tends to be less active than the sea
current, as a cold sea breeze front pushes inland. The breeze.
front may be perceived in cloud formed due to the O n sloping terrain, the diurnal cycle of surface
forced ascent, or in visibility changes. Continued heating leads to baroclinicity relative to the back-
heating of the land surface means that there is likely to ground air, and tends to cause upslope, or anabatic,
1242 MESOSCALE METEOROLOGY I Ovenriew
flow in the daytime and downslope, katabatic, flows at Mesoscale Cyclones

night (Figure 3B). Again, these flow regimes only
Vortices occur on all scales in the atmosphere, and
develop under conditions of light ambient wind, but
those on the mesoscale are particularly important in
when they do occur they may dominate the local
terms of observable weather. As ever in this regime,
meteorology. Accurate representation of these flows in
mesoscale vortices are formed and influenced by the
numerical models requires close attention to surface
whole spectrum of physical and thermodynamic
and boundary layer conditions.
processes. Secondary cyclones commonly form on
the synoptic fronts of parent cyclones. These second-
Synoptic Fronts aries can develop by baroclinic instability of the parent
front, but may also have a significant or even domi-
The fronts that form within synoptic cyclones exhibit
nant component of barotropic instability, arising from
length scales far shorter than those of the parent
the horizontal shears at the front. Such systems may
system, and are controlled by a spectrum of dynamical
develop explosively and cause significant damage. The
processes, so that synoptic fronts are really mesoscale
horizontal scales of these secondary cyclones range
features. Their genesis is well represented in primitive
from the synoptic scales right down to scales of tens of
equation simulations, but is best understood in SG
kilometers or less.
theory, which predicts the collapse of a synoptic
In addition to the formation of cyclones through
thermal gradient to a singularity in finite time, and
dynamical energy conversions, cloud processes can
thereby can diagnose the point at which the frontal
contribute significantly to the energetics of cyclogen-
surface becomes unstable in terms of low Richardson
esis. Equivalent static stability is generally lower than
number. Subsequent to this point it is assumed that a
dry stability, so the growth rate of baroclinic waves in a
real front is in a state of balance between Coriolis
saturated atmosphere is faster than the dry growth
force, thermally derived pressure gradient forces, and
rate, and the length scale of the resulting cyclone is
turbulent stresses.
shorter. Tropical cyclones, including hurricanes and
In addition to the interplay between balanced and
typhoons, also lie within the mesoscale regime. These
turbulent forces at the front, most synoptic fronts are
extreme weather systems have been the subject of a
characterized by significant cloud and precipitation
great deal of research and forecasting effort over many
features, which can have a first-order influence on the
years, yet remain difficult to predict. Their dynamics
dynamics. Cold fronts, in particular, tend to trigger
are dependent on many processes in the atmosphere,
cumulonimbus convection, which in turn generates
including balance, moist convection and radiative
strong downdraughts and gust front structures at the
fluxes, as well as complex ocean-atmosphere interac-
surface: a number of observations of the low-level
tions. Polar lows occur in high latitudes, usually
structure of active cold fronts has shown them to
associated with the rapid release of latent heat where
resemble a density current quite closely. Sometimes, a
cold air moves from the ice sheets over a relatively
frontal zone is composed of a number of rain bands,
warm ocean, but also may derive energy from baro-
aligned almost parallel to the front, whose nature can
clinic conversions. Mesoscale convective systems,
resemble that of a squall line. Rain bands have been
which appear through convective self-organization,
attributed to a number of processes, such as condi-
develop mesoscale cyclones in their lower levels over a
tional symmetric instability. Synoptic fronts, then, can
period of a few hours. On the meso-y scale, the strong
span the range of mesoscale phenomena.
ascent which develops in cumulonimbus storms can
One of the important features of the SG model of a
lead to mesocyclones though tilting, stretching, and
synoptic front is that the system remains a continuous
baroclinic generation of vorticity. It appears that
feature - a zone of smooth changes in the atmospheric
mesoscale vortices can occur on all scales, but in
properties - up to the time at which a singularity
many cases there have as yet been too few observations
occurs. Since these fronts form as components of a
of such vortices to categorize them properly, and the
baroclinically unstable wave, the front may propagate balance between dynamic and thermodynamic energy
as a wavelike phenomenon, and there is an important
sources for cyclogenesis across the spectrum of scales
possibility of material transport through the frontal is still not well understood.
zone. This behavior is observed and has important
consequences for chemical transport: a front is not
a barrier to airflow. This, in conjunction with the
rapid vertical motion due to the cross-frontal circula- See also
tions and the frontal cloud features, means that Baroclinic Instability. Boundary Layers: Modeling
synoptic fronts are important agents of transport in and Parameterization; Overview. Buoyancy and Buoy-
the atmosphere. ancy Waves: Optical Observations; Theory. Coastal
MESOSCALE METEOROLOGYI Cloud and PrecipitationBands 1243
Meteorology. Cold Air Damming. Convective Cloud for Interpreting Satellite And Radar Imagery. Cam-
Systems Modelling. Convective Storms: Convective bridge: Cambridge University Press.
Initiation: Overview. Coriolis Force. Cyclogenesis. Carlson TN (1991) Mid-Latitude Weather Systems. Lon-
Density Currents. Downslope Winds. Dynamic Me- don: Harper Collins.
teorology: Balanced Flows; Overview. Fronts. Gust Cotton WR and Anthes RA (1992) Storm and Cloud
Fronts. Instability: Inertial Instability: Symmetric Stabil- Dynamics. San Diego: Academic Press.
ity: Wave-CISK. Katabatic Winds. Large Eddy Simu- Emanuel KA (1994) Atmospheric Convection. Oxford:
lation. Lee Waves and Mountain Waves. Mesoscale Oxford University Press.
Meteorology: Cloud and Precipitation Bands: Mesoscale Gill AE (1982) Atmosphere-Ocean Dynamics. New York:
Convective Systems; Models. Microbursts. Polar Lows. Academic Press.
Tornados. Valley Winds. Holton JR (1979) A n Introduction to Dynamic Meteorol-
ogy. San Diego: Academic Press.
Pielke RA and Pearce RP (eds)(1994)Mesoscale Modeling of
Further Reading
the Atmosphere. Boston, MA: American Meteorological
Atkinson BW (1981)Meso-Scale Atmospheric Circulations. Society.
New York: Academic Press. Ray PS (ed.) (1986) Mesoscale Meteorology and
Bader MJ, Forbes GS, Grant JR, Lilley RBE and Waters AJ Forecasting. Boston, MA: American Meteorological
(1995)Images In Weather Forecasting: A Practical Guide Society.

R M Rauber and M K Ramamurthy, University of gravity waves. Precipitation bands also exist far from
Illinois - Urbana-Champaign,Urbana, IL, USA cyclones. For example, precipitation bands can form
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. and sometimes align along topographical features,
such as mountains and islands, or can be forced by
variations in surface heating, such as along coastlines
Introduction or over the US Great Lakes in winter. Precipitation
When viewed from satellites, large weather systems bands organize on the mesoscale, and range in width
such as extratropical and tropical cyclones are often from about 5 to 250 km, and in length from tens to
composed of clouds that organize on the mesoscale in over a thousand kilometers. Precipitation bands have
linear features called bands. From a radar perspective, time scales that range from less than an hour to more
precipitation falling from these clouds also organizes than a day. These scales are usually determined by
along lines. Meteorologists use the term ‘squall line’ to scales of the forcing mechanisms that trigger and
describe a line where the precipitation originates from maintain them.
thunderstorms produced by strong convective up-
drafts. The term ‘precipitation band’ is used to
describe an area of precipitation that typically origi-
Band Classification
nates from updrafts that are either nonconvective or
Extratropical Cyclones
weakly convective and is sufficiently elongated that an
orientation can be assigned. Precipitation bands are The taxonomy used to classify precipitation bands in
called ‘rainbands’ and ‘snowbands’ depending on the extratropical cyclones arose primarily from studies of
type of precipitation they produce. When precipita- cyclones in coastal locations of the United States, the
tion echoes appear on radar in several long lines or British Isles, and Japan. This classification, shown in
bands, the weather system is said to exhibit banded Figure 1, consists of six major band groups and is
structure. If bands of heavier precipitation appear based on the physical relationship of the bands to
within a larger field of widespread precipitation, they frontal boundaries. The narrow cold-frontal rain-
are often called ‘embedded bands’. band, typically about 5 km wide, is aligned with the
The structure, intensity and orientation of precip- surface position of the cold front. O n satellite images,
itation bands in extratropical cyclones are primarily the band often appears as a narrow ropelike cloud that
related t o forcing occurring along frontal zones. In can extend hundreds or even thousands of kilometers
tropical cyclones, rainbands outside the eyewall are along the surface cold front. Updrafts associated with
related to forcing associated with the relative motion forced convection can exceed 5 m s - within the band,
of the vortex through its environment and to inertia- and locally heavy rain can occur as the band passes.
1244 MESOSCALE METEOROLOGYI Cloud and Precipitation Bands
red satellite images as a sharp line demarking high and

shallow or sometimes no clouds (e.g., Figure 2),
typically occurs ahead (east) of the surface cold front
and has been described in meteorological literature as
an overrunning upper-level cold front, a pre-frontal
cold surge, a cold front aloft, or as a split front. The dry
air behind this boundary is called the dry slot. The
passage of the upper level front is characterized by a
sharp drop in equivalent or wet-bulb potential tem-
perature, which is always associated with a sharp
reduction in relative humidity and sometimes with
a reduction in temperature. The organization of
Figure 1 Idealization of the cloud and precipitation pattern

associated with a mature extratropical cyclone. (From Houze
(1993), originally adapted from Matejka et a/. (1980) and Houze
(1981), reprinted with permission from Academic Press, American
Geophysical Union and the Royal Meteorological Society.)
From a radar perspective, the narrow cold frontal

rainband often consists of small ellipsoidal cores of
heavier rainfall oriented about 30-35" to the cold
front, separated by gaps of lighter rainfall. These
regularly spaced precipitation cores and gaps are
believed to develop as a result of horizontal shearing
instability associated with strong wind shear across
the frontal interface. Occasionally, the updrafts in the
precipitation cores have been shown to produce weak
tornadoes. Wide cold-frontal rainbands develop be-
hind the surface cold front. These bands, which form
within the warm air rising along and over the frontal
surface, are about 50 km wide and are parallel to the
front. The updrafts in these bands are generally much
weaker than the narrow cold frontal rainband, on the
order of 0.5 m s - or less. Warm frontal rainbands
have similar physical characteristics to wide cold
frontal rainbands, except that they develop north of
the surface warm front, as warm air from the south
overruns and rises over the cold air. They are typically
parallel to the surface warm front, are often embedded
in lighter precipitation, and have updrafts with mag-
nitudes similar to those of wide cold-frontal rain-
bands.
Mature oceanic and continental cyclones often
appear as well-defined comma-shaped cloud patterns Figure 2 Isentropic analyses of pressure and wind on (A) the
on satellite pictures. These comma-shaped cloud 304 K surface and (6)the 310 K surface overlaid on an infrared
satellite image and radar echoes. Areas of cold (warm) advection
patterns are characterized by a sharp boundary aloft
are indicated as cross-isobaric flow toward high (low) pressure. On
between dry air subsiding from the upper each panel, a precipitation band appears at the leading edge of the
troposphere and moist air rising from the lower cold air advection marking the position of the upper-level cold front
troposphere. This boundary, which appears on infra- (see arrows).
MESOSCALE METEOROLOGY / Cloud and PrecipitationBands 1245
precipitation ahead of the upper level front (at the enhancements of stratiform precipitation that charac-
leading edge of the dryslot) depends on the stability terizes the comma-cloud pattern. Bands also occur in
and vertical wind shear in the moist air ahead of the the cyclone’s warm sector, ahead of the primary
front. In oceanic and cold-season continental cy- cyclone cloud pattern. When they occur, these bands
clones, a rainband, called the pre-frontal cold surge are called warm sector rainbands. Bands well behind
rainband or cold-front aloft rainband, typically de- the surface cold front are called post-cold frontal
velops along or slightly ahead of this boundary (Figure rainbands.
2A).This rainband can extend hundreds of kilometers.
In warm-season continental cyclones, the moist air
Tropical Cyclones
ahead of the front is often unstable and the vertical
shear may be significant. In these cases, a squall line or Concentric and spiral bands of clouds and precipita-
a line of supercell thunderstorms may be triggered tion are among the most striking features of a mature
ahead of the upper level cold front (Figure 2B). tropical cyclone. The classification of precipitation
Convection, organized in bands, can also develop bands in tropical cyclones, shown in Figure 4,has been
behind the upper-level cold front, where enhanced developed primarily from studies of hurricanes over
radiation (due to the dry, cloud free air aloft within the the North Atlantic Ocean. In some tropical cyclones,
dry slot), along with moist low-level surface air can particularly intense hurricanes, precipitation inside a
create extreme thermodynamic instability. radius of about 100 km from the vortex center tends to
Awide band of precipitation typically wraps around be axisymmetric, with convection organized in one or
the northwest quadrant of the low-pressure center of more concentric rings. These rings, called the outer
mature cyclones (Figure 3). In continental winter and inner eyewalls, contract toward the center of the
cyclones, this band often produces moderate to heavy vortex as a hurricane intensifies. With time, the outer
snowfall and is responsible for major winter storms ring replaces the inner ring, leading to cyclic changes in
and blizzard conditions. Smaller bands of heavier central pressure. Precipitation organizes in spiral
snowfall, often aligned along the mid-tropospheric rainbands outside the radius of the outer eyewall in
thermal wind, are sometimes embedded within the stronger hurricanes, and sometimes into the eyewall
broader wraparound precipitation region. As a cy- radius in weaker tropical cyclones. Unlike the con-
clone decays, the band typically elongates and nar- centric eyewall rainbands, precipitation in the outer
rows under the influence of background deformation spiral bands can be either stratiform or convective in
flow. nature. A large rainband, called the principal band,
All of the bands described above occur within the often extends from near the radius of the outer eyewall
general envelope of clouds that compose the cyclone’s to the outer radius of the hurricane cloud pattern. This
comma cloud pattern. With the exception of the band normally occurs on the east side of a cyclone. The
narrow cold frontal rainband, these bands are general principal band typically has cellular convection along
Figure 3 Radar image of a precipitation band in the north-west quadrant of a cyclone on 8 January 1998. This band produced heavy
snow across the state of Michigan.
1246 MESOSCALE METEOROLOGY/ Cloud and Precipitation Bands
and/or frontogenetic circulations, moist symmetric

instability, boundary-layer convergence, gravity
waves, topographic effects, and seeder-feeder pro-
cesses. While it is difficult to generalize which of these
mechanisms will explain the formation of a specific
precipitation band, there is considerable evidence that
the aforementioned mechanisms are among the most
important ones.
Frontogenesis Banded features in extratropical cy-

clones are associated with vertical circulations organ-
ized along lines. Frontogenesis, a primary forcing
mechanism for these vertical motions, is the process by
which the thermal and moisture gradients between air
masses are concentrated into narrow zones called
fronts. During frontogenesis, cloud and precipitation
producing transverse vertical circulations are gener-
ated, with ascending motion, concentrated into a
quasi-linear narrow zone parallel to a front, on the
warm side of the front. Frontogenetic circulations
have been studied using quasi-geostrophic and semi-
geostrophic theory, two of a hierarchy of approxima-
tions to the fundamental governing equations. The
most realistic of these theories uses the semi-geo-
Figure 4 Schematic illustration of the radar reflectivity in a strophic system of equations, with inclusion of latent
Northern Hemisphere tropical cyclone with a double eyewall and heat release to simulate the effects of condensation.
outer bands. (Reproduced with permission from Willoughby From a quasi-geostrophic perspective, the intensifica-
(1988). Copyright Commonwealth of Australia.) tion of the cross-front thermal gradient by background
geostrophic deformation, leading to warm and cold
advection on opposite sides of the front, destroys the
its axis, a clear region along its concave side, and thermal wind equilibrium. The atmosphere tries to
stratiform precipitation falling from an anvil-like restore the thermal wind equilibrium by generating a
feature on its outer side. Other bands, called secondary secondary circulation that produces vertical and
bands, are found on either side of the principal band. ageostrophic motions that counter the effects of
These typically contain shallow convection and vary frontogenesis. During this process, adiabatic cooling
significantly in intensity. Sometimes short rainbands, is generated by the rising motion on the warm side, and
called connecting bands, are present between the adiabatic warming is produced by the sinking motion
principal and secondary bands. A distinct connecting on the cold side of the front, partially negating the
band often joins the principal band to the eyewall. This effects of frontogenesis. Semi-geostrophic theory takes
band contains either stratiform precipitation or weak into account the characteristically different length and
convection. These bands together are termed the velocity scales in the along-front and cross-front
stationary band complex, since they move slowly, if directions by including the effects of thermal advect-
at all, relative to the vortex. Tropical storms also ion by the ageostrophic flow, effects that are ignored
contain smaller-scale banded precipitation features by quasi-geostrophic theory. In addition, the inclusion
that propagate outward relative to the vortex. of moisture in the semi-geostrophic equations ac-
counts for effects of latent heating associated with
condensation on the warm side of the front. The
Mechanisms Leading to the Formation resulting semigeostrophic frontal circulations include
of Banded Features a more intense, narrow, sloped updraft on the warm
side of the front (Figure 5), a frontal surface that tilts
Extratropical Cyclones
in the vertical toward the cold air, and a stronger
Several mechanisms have been advanced to explain thermal gradient near the surface than in quasi-
the existence of banded precipitation regions within geostrophic theory. The orientation, scale, and vertical
extratropical cyclones. These include frontal lifting motions associated with certain rainbands found in
MESOSCALE METEOROLOGY / Cloud and Precipitation Bands 1247
produce squall lines or other deep convective

phenomena.
0.8
Ducted gravity waves Gravity waves are buoyancy
0.6 oscillations in a stably stratified fluid where the
hl restoring force is gravity. Gravity waves are ubiquitous
0.4 in the atmosphere, but most are not large enough in
amplitude to trigger precipitation bands. This is
0.2
because atmospheric gravity waves are highly disper-
sive, propagating vertically and rapidly losing their
energy to the upper atmosphere. As a result, gravity
-1.4 -1.2 -1.0 -0.8 -0.6 -0.4-0.2 0 0.2 0.4 waves normally cannot travel very far from their
X source before they no longer have sufficient energy to
lift air to the lifting condensation level and trigger
Figure 5 Ageostrophic stream function about a front obtained by
solving the semi-geostrophic Sawyer-Eliassen equation for a case
clouds and precipitation. A particular class of gravity
of deformation forcing under moist conditions. Height ( z ) and waves, termed ducted gravity waves because their
distance ( x ) are expressed in nondimensional units. (Reproduced energy is trapped in an atmospheric duct, has been
with permission of the American Meteorological Society from associated with precipitation bands. Ducted gravity
Emanuel (1985).) waves require very specific conditions that include (1)
a sufficiently deep low-level stable layer that can
the vicinity of fronts qualitatively match these accommodate a quarter vertical wavelength of the
theoretical predictions. wave; (2)a deep moist neutral layer above the stable
layer; and ( 3 ) a critical level above the stable layer,
Boundary layer convergence The most prominent where the mean wind speed in the direction of
example of rainbands associated with boundary layer propagation equals the wave phase speed. These
convergence is the narrow cold frontal rainband. conditions most commonly occur north of warm
Under the influence of deformation flow and the fronts associated with cyclones, but have been ob-
accompanying frontogenesis, the thermal gradient at a served in other environments. When these conditions
cold front sometimes collapses to a near-discontinuity exist, theory predicts that wave energy will be trapped
at the surface. When this collapse occurs, the air within the duct, thus allowing the wave to propagate
behind the cold front takes on characteristics of a long distances. If the wave amplitude is sufficiently
density current, and strong localized convergence large, the vertical motion induced by the wave can be
occurs at the front. A sharp, narrow updraft develops sufficient to trigger a precipitation band. For rain-
in the warm air ahead of the front as the warm air rises bands to occur, the atmosphere above the stable layer
over the advancing density current. The resulting must be near saturation, a common condition over
rainband, the narrow cold frontal rainband discussed warm frontal surfaces. The structure of a propagating
above, often contains updrafts exceeding several ducted gravity wave is shown in Figure 6 . The location
meters per second and produces locally heavy rain. of greatest rising motion, where a precipitation band
Studies of the narrow, cold frontal rainband in oceanic would be triggered, is a quarter of a wavelength
regions have shown that the warm air lifted by the upstream from the location of lowest pressure.
density current is often stable or very slightly unstable, Observations of precipitation bands in conditions
indicating that the updrafts are forced rather than due where ducting is favorable often show this relationship
to free convection. The depth of the updrafts is limited in surface precipitation and pressure traces, suggesting
to the depth of the forced ascent, usually no more than that these bands may be forced by ducted gravity
3-Skm. The updrafts in the narrow cold frontal waves.
rainband are often complicated by small vortices that
form along the wind-shift line associated with the Moist symmetric instability Moist symmetric insta-
advancing density current. Shearing instability is bility is a two-dimensional, semi-geostrophic meso-
believed to be the principal mechanism for the scale instability in which both gravitational and
formation of such vortices. The updrafts then organize inertial body forces determine the stability of a
in discrete precipitation cores, separated by gaps. displaced air parcel. The term symmetric refers to a
Narrow, cold frontal rainbands are not as common basic state and resulting circulations that do not vary
in continental cyclones as they are in oceanic cyclo- in a particular horizontal direction, e.g., along a
nes, probably because the air ahead of cold fronts is baroclinic zone. In the Northern Hemisphere, the
more likely to be potentially unstable and, when lifted, condition for inviscid, inertial instability is am,/ax < 0
1248 MESOSCALE METEOROLOGY/ Cloud and PrecipitationBands
C Z
L
H
//////////////////////////////////I
-X
Figure 6 Idealized vertical cross-section of a linear plane gravity

wave with no base current, propagating to the right at speed c. The
heavy sinusoidal line is a representative isentropic surface or a Ov + AOv
temperature inversion. Surface pressure extrema are labeled H
X 4
and L, while cold and warm temperature anomalies are denoted C
and W, respectively. (Reprinted with permission of the American
Figure 7 Idealized example of a vertical cross section in the
Meteorological Society from Bosart and Sanders (1986).)
Northern Hemisphere, normal to the thermal-wind vector, showing
surfaces of constant m, (solid lines) and saturation virtual equiv-
alent potential temperature, O:v In this example, O:, increases with
+
where mg = vg fx is the geostrophic absolute mo- height (conditional stability), m, increases with increasing x
mentum, vg is the geostrophic wind in the direction (inertial stability). The lifting condensation level, level of free
perpendicular to the temperature gradient, f is the slantwise convection, and level of neutral buoyancy are denoted
LCL, LFS, and LNB, respectively.
Coriolis parameter, and x is the cross front distance,
increasing toward warmer air. The condition for
conditional instability is that d@&/dz< 0 at a level
where rm < - dT,/dz< r d . Here 0;" is the saturation d8:"/dzlmg M 0, suggesting that the atmosphere may
virtual equivalent potential temperature, T, is the have undergone symmetric overturning, reaching a
virtual temperature, and Tm and r d are, respectively, state of moist slantwise neutrality.
the moist and dry adiabatic lapse rates. In an atmos-
phere where the lapse rate of 8;" is negative when
Tropical Cyclones
evaluated along a surface of constant mg,a saturated
air parcel can be inertially stable to horizontal Tropical cyclones can be considered dynamically to
displacements (amg/ax> 0) and gravitationally sta- consist of two distinct regions, an inner core where
ble to vertical displacements (dQ:,/dz > 0), but accelerations due to relative rotation dominate over
unstable with respect to slantwise displacements. the Coriolis force, and an outer region where the
The release of this instability, termed conditional Coriolis force is comparable to rotation in influencing
symmetric instability, has been suggested as a mech- storm dynamics. As a result, in the inner gyre the air
anism for the development of banded structure in trajectories form closed paths, whereas in the outer
frontal precipitation. In situations where the atmos- region they do not. The boundary between these
phere is unsaturated, but may be brought to saturation regimes appears to be the boundary between bands
by lifting (for example, during frontogenesis), the that appear as concentric rings, such as the eyewall,
potential for slantwise instability to occur can be and outer bands, composed of a stationary band
evaluated by identifying regions where the gradient of complex and moving convective spiral bands. The
the virtual equivalent potential temperature along an location of the boundary between these two re-
m g surface, dO,,/dz/M, is 5 0. This condition is iden- gimes varies with hurricane strength, so the stationary
tical to the equivalent potential vorticity being nega- band complex tends to occur nearer the center of
tive. Potential and conditional symmetric instability weaker hurricanes and on the peripheries of stronger
are both types of moist symmetric instability (MSI). hurricanes.
Bands associated with MSI are aligned along the Analyses of hurricane structure obtained from
thermal wind, move with the environmental flow, and aircraft penetrations suggest that the principal band
have spacing that is related to both the depth of the lies along the flow streamline separating two distinct
unstable layer and the slope of the moist isentropes. regions of the storm. On the concave, or inner, side of
Figure 7 shows a schematic cross-section through an the principal band, moist high 8, air lies within the
environment susceptible to MSI. The slantwise ascent closed vortex circulation of the hurricane and orbits
within the bands occurs between the slope of the mg the center several times during its residence time
and ":6' surfaces. Observations in regions of banded within the vortex (seeFigure 8). On the convex outside
precipitation in frontal systems have often found that of the principal band, somewhat drier lower 0,
MESOSCALE METEOROLOGY / Cloud and Precipitation Bands 1249
develop outside these circulations, primarily forced

by variations in topography and surface conditions.
Flow over or around topographic features, such as a
mountain range or an island, can create elongated
zones of convergence and vertical motion that create
precipitation bands. For example, the Island of
Hawaii, which consists of two volcanic mountains
exceeding 4000 m elevation, lies in the north-easterly
Trade Wind regime of the North Pacific Ocean. As the
Trade Winds encounter Hawaii, they are unable t o
flow over the island because of an inversion at the top
of the Trade Wind layer and must, instead, flow
around the island. At night, cool drainage flow
originating in the higher elevations of the island often
flows offshore on the east (upwind) side of the island
and meets the incoming Trade Winds. The boundary
between the drainage flow and the Trade Winds leads
to a persistent zone of convergence, which forces
vertical motion, clouds, and precipitation. The defor-
Figure 8 Schematic representation of the stationary band mation of the Trade Wind flow around the island
complex of a tropical cyclone, the bands that compose the organizes the precipitation into rainbands that align
complex, and the flow in which the complex is embedded. approximately parallel the shoreline, as shown in
(Reproduced with permission of the American Meteorological
Society from Willoughby et a/. (1984))
Figure 9.
Variations in surface properties can also lead to the
formation of precipitation bands. In fall and winter
months, when cold, arctic airmasses move across
environmental air remains in the vortex circulation for warmer water bodies such as the Great Lakes,
less time than that required to orbit the center once. lake-effect snowstorms may occur. Under suitable
The principal band develops as a result of the relative
motion of the vortex through its environment. In the
environment of low-level easterly winds, the relative
motion of the vortex leads to a region of concentrated
low-level convergence along the streamline where
environmental flow encounters the rotating flow
within the vortex. Connecting bands generally have
stratiform characteristics, although they may contain
shallow convection. A stratiform band will result if a
plume of ice particles originating from the eyewall falls
through the 0°C level after spiraling outward at upper
levels of the storm. The connecting bands in hurricanes
generally sharply cross low-level streamlines, suggest-
ing that they arise from this process.
Propagating bands also exist in tropical storms.
These bands appear to be associated with inertia-
gravity waves propagating outward from the core
vortex. Radiating gravity waves are believed to be
generated during oscillations of the storm track. These
track oscillations are thought to be excited both when
tropical storms encounter land and by normal-mode
oscillations of the hurricane vortex.
Other Precipitation Bands

Figure 9 Radar reflectivity on 3 August 1990 measured by a
Although precipitation bands most commonly appear radar located along the north-east shore of the island of Hawaii,
in extratropical and tropical cyclones, they also showing a rainband located just offshore of the island.
Next Page
1250 MESOSCALEMETEOROLOGY/ Cloud and Precipitation Bands
Other local topographic and geographic features,

such as mountains, shorelines, islands, and bays,
frequently induce vertical motions organized along
lines. The precipitation bands that form often have
characteristics unique t o the specific region, but can
have a significant local impact on rain or snowfall.
See also
Buoyancy and Buoyancy Waves: Theory. Clouds:
Classification. Convection: Convection in the Ocean.
Convective Storms: Convective Initiation. Cyclogene-
sis. Frontogenesis. Fronts. Hurricanes. Instability:
Symmetric Stability. Lake Effect Storms. LandSea
Breeze. Mesoscale Meteorology: Mesoscale Convec-
tive Systems; Models; Overview. Radar: Precipitation
Radar.
Figure 10 Radar reflectivity on 7 March 1996 measured by the

Further Reading
NEXRAD radar located in Romeoville,IL. A lake-effectsnowband Bluestein HB (1993) Synoptic-Dynamic Meteorology, vol.
extends down the center of Lake Michigan and onto the southern II: Observation and Theory of Weather Systems. Oxford:
shoreline. Oxford University Press.
Bosart LF and Sanders F (1986)Mesoscale structure in the
megalopolitan snowstorm of 11-12 February 1983. Part
conditions of lake-air temperature difference, wind 111: A large amplitude gravity wave. Journal of the
speed and direction, extent of fetch over the lake, and Atmospheric Sciences 43: 924-939.
environmental stability upstream of the lake, strong Cotton WR and Anthes RA (1989) Storm and Cloud
organized convective precipitation bands may devel- Dynamics. International Geophysical Series, 44, San
op. Specifically, three morphological types of precip- Diego: Academic Press.
itation bands have been identified t o occur over the Emanuel KA (1985) Frontal circulations in the presence of
Great Lakes in wintertime. They are wind-parallel small, moist symmetric stability. Journal of the Atmos-
bands, shoreline bands, and mid-lake bands. For pheric Sciences 42: 1062-1071.
example, when air over the lake is very cold, and Houze RA Jr (1981) Structures of atmospheric precipita-
tion systems - a global survey. Radio Science 16:
winds are either weak or parallel to the long axis of the
671-689.
lake, a long precipitation band will sometimes develop Houze RA Jr (1993) Cloud Dynamics. International Geo-
near the center of a lake approximately parallel t o the physical Series, 53. San Diego: Academic Press.
lake’s shores (Figure 10).This band develops as a result Kelly RD (1986) Mesoscale frequencies and seasonal snow-
of a land breeze circulation between the lake and its falls for different types of Lake Michigan snowstorms.
shorelines. Air residing over the lake, heated by the Journal of Climate and Applied Meteorology 25:
warm lake surface, becomes unstable and rises. Cooler 308-3 12.
air flows inward from both shorelines t o replace the Matejka TJ, Houze RA Jr and Hobbs PV (1980) Micro-
rising air. The air from shore, in turn, is heated as it physics and dynamics of clouds associated with meso-
flows over the lake. A circulation develops, with air scale rainbands in extratropical cyclones. Quarterly
flowing inward over the lake, rising in a narrow zone Journal of the Royal Meteorological Society 106: 29-56.
Ray P (ed.) (1986)Mesoscale Meteorology and Forecasting.
at the lake center, and returning shoreward aloft while
slowly descending. O n the other hand, shoreline Willoughby HE (1988) The dynamics of the tropical
bands, which are usually weaker than mid-lake bands, hurricane core. Australian Meteorological Magazine
develop along or close t o the shoreline of a lake, when 36: 183-191.
winds have a cross-lake component. Such lake-effect Willoughby HE, Marks FD Jr and Feinberg RJ (1984)
snowbands have been shown t o result in locally heavy Stationary and moving convective bands in hurricanes.
snowstorms in the lee of the lakes. Journal of the Atmospheric Sciences 41: 3189-3211.
MESOSPHERE/ Metal Layers 1265
Further Reading Mesinger F (1998)Comparison of quantitative precipitation

forecasts by the 48- and by the 29-km Eta model: an
Anthes RA and Warner TT (1978) Development of hydro-
update and possible implications. Preprint, 22th Confer-
dynamic models suitable for air pollution and other
ence on Numerical Weather Prediction, Phoenix, AZ,
mesometeorological studies. Monthly Weather Review
106: 1045-1078. American Meteorological Society, pp. J22-J23.
Benoit R, Desgagne M, Pellerin P, et al. (1997)The Canadian Mesinger F, Janjic ZI, Nickovic S, Gavrilov D and
MC2: a semi-Lagrangian, semi-implicit wide band at- Deaven DG (1988) The step mountain coordinate:
mospheric model suited for fine-scale process studies and model description and perfromance for cases of
simulation. Monthly Weather Review 125: 2382-2415. alpine cyclogenesis and for a case of an Appalachian
Black TL (1994) The new NMC mesoscale Eta model: redevelopment. Monthly Weather Review 116:
description and forecast examples. Weather and Fore- 1493-1518.
casting 9: 265-277. Park SK and Droegemeier KK (2000)Sensitivity analysis of a
Daggupaty SM, Tangirala RS and Sahota H (1994) 3-D convective storm: implications for variational data
BLFMESO - A 3-dimensional mesoscale meteorological assimilation and forecast error. Monthly Weather Review
model for microcomputers. Boundary-Layer Meteoro- 128: 140-159.
logy 71: 81-107. Pielke RA (1984) Mesoscale Meteorological Modeling.
Dudhia J (1993)A nonhydrostatic version of the Penn State/ New York: Academic Press.
NCAR mesoscale model: validation tests and simulations Pielke RA (2002)Mesoscale Meteorological Modeling, 2nd
of an Atlantic cyclone and cold front. Monthly Weather edn. New York: Academic Press.
Review 121: 1493-1513. Pielke RA, Cotton WR, Walko RL, et al. (1992) A compre-
Grasso LD (2000) The differentiation between grid spacing hensive meteorological modeling system - RAMS. Mete-
and resolution and their application to numerical model- orology and Atmospheric Physics 49: 69-91.
ing. Bulletin of the American Meteorological Society 81: Saito K (1997) Semi-implicit fully compressible version of
5 79-5 80. the MRI mesoscale nonhydrostatic model: forecast
Gross G (1992) Results of supercomputer simulations of experiment of the 6 August 1993 Kagoshima torrential
meteorological mesoscale phenomena. Fluid Dynamics rain. Geophysics Magazine 21: 109-137.
Research 10: 483-490. Thunis P and Clappier A (2000)Formulation and evaluation
Hodur RM (1997) The Naval Research Laboratory’s Cou- of a nonhydrostatic Mesoscale Vorticity Model (TVM).
pled OceadAtmosphere Mesoscale Prediction System Monthly Weather Review 128: 3236-3251.
(COAMPS).Monthly Weather Review 125: 1414-1430. Weidman S and Pielke RA (1983)A more accurate method
Laprise R (1992) The resolution of global spectral models. for the numerical solution of nonlinear partial differential
Bulletin of the American Meteorological Society 73: equations. Journal of Computational Physics 49:
1453-1454. 342-348.
Laprise R, Caya D, Bergeron G and Giguere M (1997) The Xue M, Droegemeier KK and Wong V (1995) Advanced
formulation of the Andre Robert MC2 (Mesoscale Regional Prediction System (ARPS) and real-time storm
Compressible Community) Model. Atmosphere-Ocean prediction. Preprint, International Workshop on Limit-
35: 195-220. ed-Area and Variable Resolution Models. World Meteo-
Mathur MB (1974) A multiple grid primitive equation rological Organization, Beijing, China.
model to simulate the development of an asymmetric Yamada T (2000) Numerical simulations of airflows and
hurricane (Isbell, 1964). Journal of the Atmospheric tracer transport in the southwestern United States.
Sciences 31: 371-393. Journal of Applied Meteorology 39: 3 9 9 4 1 1 .
Contents
Metal Layers
Polar Summer Mesopause
Introduction
Metal Layers
The ablation of interplanetary dust particles generates
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. the thin layers of neutral metal atoms that occur
1266 MESOSPHERE/ Metal Layers
globally at altitudes between 80 and 100 km. Several noctilucent cloud formation. In the stratosphere,
of these metals (sodium, iron, potassium, calcium, and metallic species may influence the chlorine-catalyzed
lithium) possess suitable optical transitions and can be removal of ozone, and provide nuclei for the forma-
observed from the ground by the spectroscopic tech- tion of sulfate particles in the Junge layer.
niques of photometry and lidar. These metals occur as
free atoms because above 80 km the concentration of
atomic oxygen (0)exceeds that of ozone ( 0 3 ) : while Meteoric Ablation as a Source
0 3 oxidizes metal atoms to metal oxides that then go
on to form a variety of compounds such as hydroxides,
of Mesospheric Metals
carbonates, and bicarbonates, atomic oxygen and Up until the 1990s, the daily input of meteoric material
associated atomic hydrogen reduce these compounds into the atmosphere had been estimated to be about 44
back to metal atoms. Above lOOkm the metals tonnes. This figure was derived by estimating the mass
become ionized by charge transfer with the increasing flux passing through the beam of a meteor radar.
levels of E region ions such as NO' and 0;. Several approximations are involved in making such
Surprisingly, the relative abundances of the metal an estimate, particularly in relating the mass of the
atoms are quite different from their relative abun- meteor to the ion trail that is actually observed by the
dances in chondritic meteorites. For example, atomic radar. Furthermore, the wavelength band of the radar
calcium is depleted by more than two orders of will sample only a subset of the masshelocity distri-
magnitude with respect to sodium. The metals also bution of the meteoroids. Revised meteor radar
exhibit different seasonal behavior: the integrated estimates now indicate that the daily input is higher
column densities of all the metals peak in early winter, by a factor of 2-3. An alternative estimate obtained by
and sodium and iron have a marked mid-summer measuring the accumulation of iridium in sediments
minimum, whereas calcium and potassium have a has indicated that the meteoric influx could be up to 6
secondary mid-summer maximum and hence little times greater. Most recently, the Long Duration
seasonal variation. The explanation for these differ- Exposure Facility, an impact detector placed on a
ences appears to be a combination of differential spacecraft for several years, has yielded an estimate of
ablation (for example, the least volatile metal, calci- 110 tonnes per day; this is probably the most reliable
um, is ablated about 10 km lower in the atmosphere estimate.
than sodium), and differences in the gas-phase chem- It has been postulated that most of the incoming
istries controlling the layers. meteoric mass is in the form of interplanetary dust
Recently, sporadic (or sudden) metallic layers particles in the mass range 1-1000 pg (median mass
have been observed. These are very thin layers of 10 pg), with a radius range of 50-500 pm (median
neutral metal atoms that appear explosively at about radius 100pm). The average entry velocity of these
95 km. They are most probably linked to sporadic particles is about 18 kms-l, so that frictional heating
E layers, which are concentrated layers of metallic to over 2000K causes about half of the meteoric
ions and electrons that occur between 90 and 130 km material entering the atmosphere to ablate in the 80-
and that have an important influence on radio 1OOkm region, providing a direct source of metal
communications. atoms and ions. The major metallic constituents of
Lidar enables the metallic layers to be observed with meteorites by weight are Mg 12.5%0, Fe 11.5%0,A1
excellent spatial and temporal resolution. Hence, the 1.7%, Ni 1.5%, Ca 1.0%, Na 0.6%. Thus, the average
metal atoms can be used as tracers of dynamical influx of sodium from meteoric ablation is about
processes such as gravity waves and tides. Narrow- 3 x l o 4 cm-2 s-l, based on the Long Duration Expo-
linewidth lidar observations of the hyperfine structure sure Facility.
of the Na and K D lines have also been used to measure In addition to meteoric ablation as a direct source of
temperature and wind speed with excellent precision. metals, serious consideration has also been given to a
Temperature profiles can also be obtained with a new layer of dust at about 90 km, from which metal atoms
technique employing a two-colour lidar to observe the are released during the day by thermal evaporation
relative populations of the lowest spin-orbit states of and photosputtering. The postulated origins of this
atomic iron. dust have included sea-salt aerosols, volcanic debris,
Finally, meteoric metals affect the atmosphere in a and particles formed from the condensation of refrac-
number of ways. Sporadic E layers have a significant tory species produced during meteoric ablation.
effect on the electrical conductivity of the lower However, there is now strong evidence that direct
thermosphere. In the mesosphere, metallic com- ablation of incoming meteoroids is the major source of
pounds in the form of individual molecules or dust these metals. This evidence includes the reasonably
particles are probably the major source of nuclei for good correlation between the relative abundances of
metallic ions (though not neutral atoms, see below) made rapidly (typically every 60 s for the Na layer), so
observed in the lower thermosphere and their elemen- that the metal layers can act as tracers of atmospheric
tal abundances in chondritic meteorites; sporadic motions such as tides and gravity waves. Third, in the
enhancements in the total column abundance of case of Na and K a narrow-linewidth laser can be used
sodium during meteor showers; and the direct obser- in the lidar transmitter to measure temperature and
vation by high-resolution lidar of metal deposition in wind profiles in the upper mesosphere. This is possible
short-lived meteor trails. because the absorption spectrum of each D line
contains hyperfine structure. For example, the D2
absorption spectrum consists of six hyperfine reso-
Techniques for Observing Metals nance lines, which become blended at temperatures
above 6 K. The degree of blending is very sensitive to
in the Mesosphere temperature, so, employing a laser with a tuning
The first quantitative observations of metal atoms accuracy and frequency stability of at least 50 MHz to
were made in the 1950s using ground-based photom- scan across the DZ absorption spectrum, the temper-
eters that measured the resonance fluorescence from ature can be measured with an uncertainty x 1K. The
spectroscopic transitions of the metal atoms excited by laser can also be scanned to the wings of the D2
solar radiation. Emission lines from Na, K, Fe and Ca' absorption peak in order to measure the Doppler
were successfully observed because these metals have width of the peak. From this the radial wind (i.e., along
extremely large resonant scattering cross-sections. A the line-of-sight of the lidar) can be calculated with an
large scattering cross-section is essential because their error of less than 3 m s-l, and can then be resolved into
concentrations relative to the general atmosphere are the vertical and zonal wind components. A recent
less than 100 parts per trillion (1OOx Photom- development has been to use a two-laser lidar to
eters are generally pointed to near zenith during measure the populations of the spin-orbit multiplets
twilight, when the geometrical shadow height of the of ground state Fe(5D), which are related to temper-
Earth (the terminator) is close to mesospheric alti- ature through the Boltzmann equilibrium.
tudes. Radiative transfer theory is used to derive the There have also been a number of measurements by
vertical concentration profile from the variation of the rocket-borne mass spectrometers of the concentra-
emission signal as the terminator passes up through tions of positive metallic ions in the upper atmosphere.
the metal layer. These flights have been motivated by a need to
Photometry was superseded in the 1970s when the establish the link between meteor showers and the
discovery of tuneable laser sources allowed the devel- abundance of metallic ions, and to study the role of
opment of the resonance lidar (laser radar) technique. metallic ions in forming sporadic E layers (see below).
In this technique, a pulsed laser beam is tuned to a Metallic ions such as Mg' have also been observed by
strongly allowed spectroscopic transition of the metal resonant scattering of sunlight, using spectrometers on
atom of interest, and transmitted up through the satellites and the space shuttle.
atmosphere. The laser pulse is Mie- and Rayleigh-
scattered, particularly in the lower atmosphere where
there are aerosol layers and the pressure is greater. In Observations of Metallic Species
the mesosphere, the pulse is resonantly scattered by the
metal atoms. A small fraction of the scattered light
in the Atmosphere
returns to the ground, where it is collected by a Figure 1 shows profiles of the annual mean layers of
telescope and measured by photon-counting. The Na, Fe, K and Ca observed by lidar at several mid-
return signal is electronically binned to provide the latitude locations. Note that although magnesium is
range and hence height resolution of the scattering the most abundant meteoric metal, atomic magnesium
layer, typically to within 40m. The absolute metal cannot be observed from the ground because its
density is calibrated from the Rayleigh-scattered optical transition at 285.2nm is obscured by the
cross-section at a lower altitude of known atmospheric stratospheric ozone layer. The Na layer has been
temperature and density. Lidar has so far been used to studied in far greater detail than those of the other
observe Na, K, Li, Ca, Ca' and Fe. metals because it is the easiest metal to observe
The technique has a number of important advan- spectroscopically. The column density of Na is about
tages over photometry. The first is that observations 5 x lo9 atoms cmP2,although this can vary by a factor
can be made continuously over a complete diurnal of 10 depending on time and location. The layer
cycle, provided an astronomical-quality telescope and exhibits a seasonal variation with a wintertime max-
narrowband optical filter are employed for daytime imum, which is also latitude dependent. For instance,
measurements. The second is that observations can be at low latitudes the winter enhancement is only about
110 g
10 100 1000 10000 10-4 10-3 10-2 10-1 1 10 102

Concentration ( ~ m - ~ ) Chondritic abundance relative to Na
Figure 1 Vertical profiles of the annual mean concentrationsof Figure 2 Relative annual averageabundances of meteoric metal
Fe, Na, K, and Ca measured by lidar at a number of mid-latitude atoms (at mid-latitudes), compared to their relative chondritic
locations in the United States and Europe. abundances.
1.3:1,whereas at mid-latitudes this variation increases sometimes appearing within a matter of minutes and
to about 3:l and to more than 1 O : l in the polar then surviving for perhaps a few hours. The average
regions. The height of the peak of the Na layer varies FWHM of these sporadic layers is only about 2 km,
between 88 and 92 km, with the highest peak heights and their peak concentrations can be as much as 40
occurring during summer. The full width at half- times the peak of the background metal layer. They
maximum (FWHM)of the layer is about 10 km, and it have also been observed with a horizontal extent of
is usually characterized by strikingly small scale- more than 1000km. In explaining this intriguing
heights of 2-3 km on the top and bottom sides of the phenomenon, it has been noted that sporadic layers
layer (the scale height is the distance over which the commonly occur together with sporadic E layers.
concentration changes by a factor of e (2.18,..).) These are thin layers of metallic ions that can be
There is a significant increase in the layer column formed by horizontal wind transport across magnetic
density during the day (by 30-loo%), which field lines, which produces convergence of the ions into
N
appears to be caused by a combination of photochem- layers at null points in the wind shear. Sporadic neutral
istry and the semidiurnal atmospheric tide. layers can then result from metal ions forming ion
The other neutral metal layers that have been clusters that undergo dissociative recombination with
observed in the upper atmosphere are those of FeyK, electrons (see below). However, there may well be
Ca and Li. Figure 2 illustrates the relative column other mechanisms for sporadic layer formation,
abundances of these metals to that of Nay plotted including auroral precipitation acting on meteoric
against their relative abundances in chondritic mete- dust particles.
orites recovered from the Earth's surface. Since there is
strong evidence that meteoric ablation is the major
source of the metals, a good correlation might be Metallic Species in the Earth's Airglow
expected. Apart from lithium for which there is a It was first reported in 1929 that radiation at 589 nm is
rather small observational database, the other metals present in the nightglow spectrum. A decade later it
are all depleted with respect to sodium. In particular, had been established that this radiation is due to the
calcium is depleted by the enormous factor of 120-360 transition Na(32P3/2,1/2-32S1/2) from a source located
depending on season. There are other unexpected within the Earth's atmosphere. Sydney Chapman then
differences in the layers. Compared with the peak of postulated the sequence of reactions [I] and [11] to
the sodium layer, the lithium peak is about 4 k m account for the emission.
higher, while the potassium, calcium, and iron peaks
are several kilometers lower and have considerably Na+O3 -+NaO+03 [I1
smaller scale heights, as shown in Figure 1.
Lidar observations have also revealed the curious NaO + 0 + Na(2P, 2S) + 0 2 1111
phenomenon of sporadic (or sudden) metal layers.
These are very thin, concentrated layers of metal Historically, there have been two significant problems
atoms that occur at altitudes between 90 and 110 km, associated with validating the Chapman mechanism.
The first was that the rate coefficients for reactions [I]
and [11] have to be fast enough to generate the
measured D-line emission intensity of 50-200 R
(1 Rayleigh = l o 6 photons cmP2s-' emitted in all
directions). However, recent laboratory measure-
ments (see below) have now confirmed that both
reactions are extremely fast. Reaction [I], which is rate
determining, proceeds via the electron jump (or
harpoon) mechanism. The second problem was the
size of the branching ratio, f , for production of Na(2P)
in reaction [II]. Geophysical models of the Na layer
show that f has to be E 0.1, in agreement with a field
experiment in which a rocket carrying a sodium
photometer was launched through the Na layer while
a ground-based lidar observed the sodium atom
concentration. However, an early laboratory meas-
urement appeared to show that f was less than 0.01.
This apparent disagreement has now been resolved
through a series of elegant laboratory experiments.
First, it was shown that reaction [I] produces N a O
almost entirely in the low-lying NaO(A2X+) excited orH2
/ CO,, M
electronic state, rather than the NaO(X211) ground

state. The NaO(A) state has a long radiative lifetime Figure 3 Schematic drawing of the significant chemical cycles of
and is not quenched efficiently, so that in the meso- sodium in the upper mesosphere/lower thermosphere region.
sphere reaction [11] involves NaO(A) rather than Major sodium species are shown in bold in boxes. The less
important chemistry of Na03,NaC03, and ions such as Na' . H20
NaO(X).A second very recent experiment has shown has been omitted for the sake of clarity.
that f for NaO(A)+O is 0.14 f 0.04, thereby recon-
ciling the geophysical observations with the underly-
ing chemical physics.
models. Almost all of these reactions have now been
studied in isolation under conditions appropriate to
the upper mesosphere. Figure 4 illustrates vertical
Modeling of Metallic Layers profiles of the major sodium compounds, predicted
Since the only metal species that can be observed from a one-dimensional model incorporating the
directly in the mesosphere are the atomic neutrals and chemistry in Figure 3. Above the atomic sodium layer
ions, our understanding of the chemistry that forms at around 90 km, ion chemistry predominates. Sodium
the metal layers has come from a combination of atoms are ionized mostly by charge transfer with the
laboratory studies and modeling. Up until the 1980s, ambient NO+ and 0; ions, with a small contribution
laboratory measurements of the rate coefficients for from solar photoionization. Dielectric recombination
reactions involving metallic species were available (i.e., Na' + electron + Na) is a very inefficient proc-
only for ion-molecule reactions, which were studied in ess. Instead, Na' is neutralized by forming clusters,
ion drift tubes. However, since then the two classical which then undergo dissociative recombination with
techniques of flash photolysis and the fast flow tube an electron. The rate of neutralization is governed by
have been applied with great success to the challenging the initial formation of N a . N l , followed by a
task of studying reactions of neutral metallic species in competition between atomic 0 and C02, the former
the gas phase at the low temperatures characteristic of converting the cluster back to Na+, and the latter
the upper atmosphere. In addition, photoelectron forming a stable cluster that will subsequently react
spectroscopy and molecular beams have been em- with an electron.
ployed to investigate the production of excited states Below 90 km, N a becomes converted to the stable
in exothermic reactions. While most laboratory work reservoir NaHC03, via a series of steps beginning with
has concentrated on sodium chemistry, there is now a the oxidation of Na by 0 3 to NaO. As shown in Figure 3,
growing database on reactions of iron, magnesium, species such as NaO, NaOH, and N a H C 0 3 are
calcium, potassium, and lithium. converted back to N a by reaction with 0 and H. The
Figure 3 is a schematic diagram of the gas-phase reaction between atomic hydrogen and NaHC03 has
chemistry of sodium that is employed in current recently been shown to be extremely slow, so that
Na D emission intensity (photon c m 3 s-') chondritic ratios (Figure 2) are not primarily due to
0 50 100 150 200 250 300 differences in chemistry but are due to differential
I I I I I meteor ablation. Indeed, current models require iron,
105 - potassium, and calcium ablation efficiencies of about
l6%, 13%, and 0.3%, respectively. Models of iron
100
Na D emission - and calcium are able to explain quite satisfactorily
h
E most of the seasonal differences between these metals

5
-
s.-
95 and sodium. However, the very surprising differences
between sodium and potassium, which are from the
2 90 same group in the Periodic Table, remain to be fully
explained.
85 These one-dimensional models have also been
modified to explore gravity wave perturbations on
the metal layers. Gravity waves in this region can cause
0 2000 4000 6000
temperature variations of over 1 0 K and vertical
Concentration
displacements of several kilometers in less than an
Figure 4 Vertical profiles of the major sodium species and the hour. The motivation for studying the coupling of
sodium D nightglow intensity predicted by a one-dimensional metal chemistry and dynamics has been to examine
model for January, 40" N (nighttime). whether the metals are suitable inert tracers of
atmospheric motion, or whether the apparent dynam-
ical perturbation is amplified by a fast chemical
NaHC03 should be the major reservoir below 88 km, response to changes in temperature and the concen-
as shown in Figure 4.Note, however, that there have trations of minor species such as 0 3 , 0, and H.
not yet been direct observations of this compound in However, several studies now show that the sodium
the atmosphere. It is only when 0 and H become layer is a conservative tracer of short-period gravity
abundant above 85 km that atomic sodium becomes waves. In contrast, the underside of the iron layer
the dominant form. Hence, the small scale-height on appears to be controlled by fast chemistry, which may
the underside of the Na layer mirrors the fall-off in lead to significant chemical amplification.
atomic 0 and H.
The chemistries of iron, magnesium and calcium are
somewhat different from that of sodium. Unlike Na', Impact of Meteoric Metals in the
the ions Fe+,Mg+, and Ca' are chemically active,
reacting with 0 3 to form oxides and forming strongly
Mesosphere and Stratosphere
bound dioxides with 0 2 . The rate of neutralization of It has been proposed that metallic species play a
these metal ions is governed by competition between significant role as nuclei for noctilucent clouds. These
atomic oxygen and electrons for the metal oxide ions. are thin ice clouds that occur between 82 and 85 km, at
The removal of metal atoms on the underside of the high latitudes during summer. They are currently
layer involves oxidation by 0 3 to form neutral metal being studied intensively as indicators of climate
oxides, followed by recombination with 0 2 , C02, or change in the upper atmosphere because their occur-
H 2 0 to form the trioxide, carbonate, or dihydroxide, rence, frequency, and latitudinal coverage appear to be
respectively. Although the dihydroxide is thermody- increasing, probably as a result of increasing water
namically most stable, it is not formed very rapidly vapor and decreasing temperatures. Because the upper
because of the dryness of the upper mesosphere. mesosphere is a very dry region, the condensation
Hence, it is not clear at present which of these nuclei for noctilucent clouds must be particularly
compounds form the major reservoir species for these effective. Two proposed sources of nuclei are hydrated
three metals. metallic ions, and neutral molecules such as sodium
Models using the sodium chemistry in Figure 3 are bicarbonate with very large dipole moments. In both
able to reproduce very satisfactorily the observed Na cases, strong electrostatic forces promote the binding
layer as a function of season and latitude. Models for of water molecules.
the Fey K and Ca layers have also been published In the middle and lower mesosphere it is presumed
recently. As Figure 1 shows, although there are that metallic compounds polymerize together to form
differences in the heights and shapes of these layers a metal-rich dust. This probably occurs quite rapidly
with respect to the Na layer, these differences are of because several of these species, particularly metal
second order. This implies that the enormous devia- carbonates and bicarbonates, have very large dipole
tions in the metal abundances from their expected moments that will induce long-range attractive forces
MESOSPHERE I Polar Summer Mesopause 1271
between them. However, if the metal compounds measurements with an aerosol mass spectrometer
remain in the gas phase as they enter the stratosphere, have shown that Junge layer sulfate particles have a
then, being alkaline species, they will react with substantial meteoric content.
hydrogen chloride produced from destruction of
chlorofluorocarbons. Indeed, laboratory studies have
shown that several sodium compounds react at every
collision with HCl to yield NaC1. See also
It has been proposed that formation of metal Atmospheric Tides. Chemistry of the Atmosphere:
chlorides could have two impacts on the chlorine- Chemical Kinetics; Ion Chemistry; Laboratory Studies.
catalyzed removal of stratospheric ozone. First, a Lidar: Resonance. Mesosphere: Polar Summer Meso-
typical metal chloride will readily be photolyzed to pause. Meteors. Noctilucent Clouds. Observation
yield atomic chlorine (Cl) and the metal atom, which Platforms: Rockets. Ozone: Ozone Depletion.
will rapidly reform a metallic compound so that this
sequence of reactions can occur once again. The
overall effect is thus t o convert HC1 catalytically t o C1, Further Reading
the active form of chlorine that destroys ozone. This
cycle will be most effective in the upper stratosphere Clemesha BR, Batista PP and Simonich DM (1996)Forma-
between 40 and 50 km, since this is where C1 has the tion of sporadic sodium layers. Journal of Geophysical
greatest impact on ozone depletion. Lower in the Research 101: 19701-1 9706.
stratosphere, photolysis of these metal compounds Gardner CS and Taylor MJ (1998) Observational limits for
lidar, radar, and airglow imager measurements of gravity
will become slower and they will polymerize and wave parameters. Journal of Geophysical Research 103:
eventually diffuse out of the stratosphere. These 6427-6437.
polymers could remove a significant fraction of Plane JMC (1991)The chemistry of meteoritic metals in the
atmospheric chlorine. upper atmosphere. International Reviews of Physical
Metallic species could also play a role in the Chemistry 10: 55-106.
heterogeneous chemistry that leads to severe ozone Plane JMC and Helmer M (1994)Laboratory studies of the
depletion during springtime at high latitudes. Atmos- chemistry of meteoric metals. In: Hancock G and
pheric subsidence in the polar vortex is likely to cause a Compton RG (eds). Research in Chemical Kinetics, pp.
significant accumulation of meteoric debris in the 313-367. Amsterdam: Elsevier.
lower polar stratosphere by the beginning of spring. Plane JMC, Cox RM and Rollason RJ (1999)Metallic layers
These metallic compounds possibly act as nucleation in the mesopause and lower thermosphere region.
Advances in Space Research 24: 1559-1570.
centers for nitric acid and water in the formation of von Zahn U, Gerding M, Hoffner J, McNeil WJ and Murad
polar stratospheric clouds. Furthermore, the alkaline E (1999)Iron, calcium, and potassium atom densities in
metallic species may facilitate the removal of nitric the trails of Leonids and other meteors: strong evidence
acid from the gas phase by forming metal nitrates in for differential ablation. Meteoritics and Planetary Sci-
the ice particles. Even at mid-latitudes, recent airborne ence 34: 1017-1027.
Polar Summer Mesopause

M M Huaman and M C Kelley, Cornell University, lower boundary of the Earth’s partially ionized plasma
Ithaca, NY, USA blanket, called the ionosphere. The latter is collocated
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. with the thermosphere, which has temperatures up to
ten times higher than in the mesopause.
The polar summer mesosphere is particularly fasci-
Introduction nating. Although the mesosphere has very little water
The mesosphere is the region of the atmosphere vapor, the highest clouds on Earth are found in this
located between the stratosphere and the thermo- region. Remarkably, in the fwll polar summer sunlight
sphere, between 50 and 90 km, in which temperature the temperature often reaches values as low as 110 K,
decreases with height. The transition between the with one measurement as low as 90 K. This is clearly
mesosphere and the thermosphere is called the meso- the coldest natural temperature found on or near the
pause and is the altitude at which the temperature Earth. Additionally, we find an intense but as yet
reaches a minimum before increasing with height in unexplained radar scattering layer together with
the thermosphere. The mesopause is also near the atmospheric motions that are dominated by poorly
1272 MESOSPHERE/ Polar Summer Mesopause
understood gravity waves and tides. These properties An important discovery was made during the
are reason enough for scientific interest in the polar International Geophysical Year (ICY) of 1957-1958.
mesosphere, but there are important global change Rocket grenades launched from many locations re-
aspects as well. vealed that, against all expectations, the temperature
in the polar summer mesosphere is colder than in the
winter polar zone. In fact, a temperature difference of
Noctilucent Clouds (NLC) and the about 100K is found between summer and winter.
temperature anomaly More recent data are presented in Figure 2 and reveal
these hemispheric temperature differences quite well,
The interest in the polar summer mesosphere started along with the level of temperature fluctuation in the
on 18 June 1885, two years after Krakatoa's eruption region, which is quite high in the winter hemisphere.
in 1883. Silvery-blue clouds were observed from the
This result supports speculations that the NLC were
ground under twilight conditions when the Sun was
composed by ice that forms at extremely low temper-
below the horizon but was still illuminating the atures even at the low water vapor pressure of the
mesosphere (i.e., just before dawn or just after sunset). mesopause zone (1-2 parts per million).
Using photographic triangulation, it was found that
The gravity or buoyancy waves that create the
the height of these clouds was about 82 km, the highest
interesting structure in Figure 1 are of more than pass-
clouds ever seen on Earth. Owing to their unusual ing importance to understanding the low temperature
nighttime brightness, they were named noctilucent of the polar summer mesosphere. Such waves are also
clouds, or NLC. It is believed that the major volcanic called internal waves since they can propagate easily
eruption introduced a considerable amount of water
through the atmosphere from one height range to another.
vapor into the stratosphere that took 2 years to be Typical sources include surface wind flow over
transported to the mesosphere, eventually contribut-
orographic features, frontal systems, and severe
ing to the ice particles that formed these clouds. The ice
storms; even earthquakes and nuclear explosions
particles were large enough to scatter sunlight and be
have created substantial waves in the atmosphere.
seen by the naked eye.
What makes these waves unique and important to the
These clouds are only observed during the summer
upper atmosphere is that they increase in amplitude as
months, usually presenting a wavy pattern, an effect
they propagate upward. This seems counterintuitive
that is attributed to their interaction with passing
but is actually firmly rooted in the principle of
gravity waves. The ideal zone of viewing is between
conservation of energy. The kinetic energy per unit
53" and 57" latitude because of the long twilight and
volume in a wave packet is given by eqn [l].
the polar location of the clouds themselves. The
example presented in Figure 1 displays some of the
characteristics of noctilucent clouds.
Figure 1 Noctilucentclouds as observed at 2255 UTC on 19 July 1997 from Glengarnork,Ayshire, Scotland. (Figure courtesy of Tom
McEwan.)
MESOSPHERE I Polar Summer Mesopause 1273
V
V
V
90 90 -
80 - 80 -
r
Y
w
70-
70 70 -
-
al
n
3
.-
2 60-
60 60 -
50 - 50 -
Summer 1987 Wiinter 19831

40 - 40 -
30
100 150 200 250 300 100 150 200 250 300
Temperature (K) Temperature (K)
Figure2 Temperature profile measurements taken using radar tracking of falling spheres during summer 1987 (left) and winter 1983-
84 (right) over Andpiya, Norway. (Adaptedfrom von Zahn U and Meyer W (1989) Mesopausetemperatures in polar summer. Journalof
Geophysical Research 94: 1647.)
In eqn [I], 6u is the perturbation wind velocity in the bation velocity exceeds the wave velocity, the wave
wave (m s - ’) and po is the background atmospheric breaks.
mass density (kgm - ’). (Since the kinetic energy of a What does this have to do with the cold summer
simple molecule of mass M is $Mu2, measured in mesopause? When waves break on a beach or in the
joules, exchanging po for M yields the kinetic energy clear air, they deposit their energy and momentum
density in JmP3.) Similar expressions hold for other back into the local medium (swimmers on the surface
forms of energy such as heat and potential energy in a know this very well). The mesosphere is so tenuous
wave packet. The important thing is that as the wave that the input of momentum from waves generated in
propagates upward, po decreases drastically, expo- the dense lower atmosphere is very significant. Cur-
nentially in fact, with the form for a uniform atmos- rent theories of the mesosphere argue that the waves
pheric temperature being po(z) = po(0)e-Z/H,where that reach these heights come from preferred direc-
H = kT/Mgis the atmospheric scale height ( % 7 km at tions that are different in the two hemispheres. In the
surface temperatures) and k is Boltzmann’s constant. summer hemisphere the waves preferentially come
So if energy is to be conserved as a wave propagates
upward, each time the density po decreases by a factor
of 2 the wave velocity must increase by a factor of fi,
or by about 40%. Eventually, any given wave will
reach a height where the amplitude is so great that it
breaks. A good rule of thumb is that a wave will break
when its internal wave perturbation velocity exceeds
its propagation speed - it catches up with itself,
steepens, and breaks. Figure 3 shows an analogous
situation for water waves. The wave speed slows as the Figure 3 Analogy for gravity wave breaking similar to that of
water becomes shallow and when the internal pertur- water waves on a beach.
from the west, depositing a net eastward momentum model is the TIME-GCM (Thermosphere-Ionosphere-
into the medium. This spins up the atmosphere Mesosphere-Electrodynamics General Circulation
somewhat and it moves away from the pole. To Model). The calculated yearly variation of tempera-
conserve mass, there is a net upflow at high latitudes, ture at 85 km is shown in Figure 5 . We can observe an
resulting in adiabatic cooling. The opposite effect asymmetry between hemispheres for solstice condi-
occurs at the winter pole and the temperature rises. tions at high latitudes. The southern polar summer
The preferential direction could arise in a variety of hemisphere seems to be warmer than the northern
ways and we discuss only one here, the so-called polar summer hemisphere by 100 K.
critical layer effect. Gravity or buoyancy wave velo- New ways to study NLC have been developed that
cities are small enough that jet stream winds can be should improve our understanding experimentally.
larger than the wave propagation speed. Suppose a For instance, in the early 1980s satellite measurements
wave propagates upward to a height where its hori- detected NLC, though, owing to the different method
zontal phase speed equals that of the background of detection, they were called polar mesospheric
wind, height ho in Figure 4.At this height the wave is clouds or PMC. They are believed to be the same as
not a wave at all, just some eddies in the flow, and it NLC, with the only difference being that, from orbit,
ceases to exist. In fact, only waves with horizontal they could be observed 24 hours a day. Another
velocities greater than u,,, get through the jet stream observation method that is not hampered by lighting
at all. But waves propagating in the other direction, conditions takes advantage of the unexpectedly high
against the flow, are never subject to this effect and radar cross-section in the polar summer mesosphere.
pass through easily. Since the jet stream is to the west in This scattering process is of considerable interest in its
the summer and to the east in the winter, gravity wave own right and is discussed at length in the next section.
filtering might explain the mechanism described above The latest way to monitor NLC involves lidar, a
and the observed temperature asymmetry. method analogous to radar but using light waves
Modern global circulation models can include such instead of radio waves.
effects by parameterizing momentum fluxes. One such NLC are usually observed only at high latitudes
between 50" and 60", but on 22 June 1999 they were
observed in Boulder, Colorado (40"N), an indication
that NLC are moving south. This could be due to
Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Month (mtimes 010 0O:OO to 360 0O:OO)
urnax
Wind speed
Figure 5 TIME-GCM variation of zonal average neutral gas
Figure 4 Internal gravity waves have phase velocities of the temperature in kelvin at 85 km over a year. (Figure courtesy of
same order as wind speeds in the jet stream. Here we illustrate a Roble RG (2000) On the feasibility of developing a global
wind profile whose peak value exceeds the phase velocity of a atmospheric model extending from ground to the exosphere.
particular wave. In this case an upward-propagating wave will Atmospheric Science Across the Stratopause. Geophysical Mon-
reach the critical layer, where it is absorbed by the fluid. ograph 123, American Geophysical Union, 53-67.)
MESOSPHERE/ Polar Summer Mesopause 1275
cooling and/or increased water vapor caused by rising physics, interplanetary dust cloud studies, meteor
levels of methane and carbon dioxide due to human ablation, and recoagulation science all have something
activity, a topic we take up later. in common with the polar summer mesopause region.
Polar Mesosphere Summer Echoes

Proposed Mechanisms of PMSE
(PMSE)
Generation
As noted above, an intriguing feature of the polar
summer mesosphere is the ability of the region to Many mechanisms have been proposed as being
strongly scatter radio waves. Early studies of VHF responsible for, or at least partly responsible for,
radar echoes from the high-latitude summer meso- PMSE generation. An example showing six hours
sphere and lower thermosphere using the Poker Flat of radar echo data from this region is presented in
MST radar in Alaska (65"N) showed a relatively Figure 6 . This can be viewed much like a Doppler
narrow and surprisingly intense echoing layer cen- weather radar plot except that the beam is fixed in the
tered at about 86 km. The echoes were characterized vertical direction. The time evolution is similar to
by their strong VHF (50 MHz) radar backscattering that provided by a slit camera focused on a spot at
cross-section, with backscattered powers 2-3 orders 80-90 km altitude. The radar frequency is 51.5 MHz,
of magnitude greater than typical values observed at corresponding to backscatter from structures of
low or middle latitudes (in any season) or at high 5.83 m wavelength. Radio waves scatter most easily
latitudes (in nonsummer periods). The echoes are now from electrons and, when these are present, usually
referred to as Polar Mesosphere Summer Echoes, or dominate the physics. Indeed, since the summer polar
PMSE. PMSE are both intriguing and surprising zone is in full sunlight, the ionosphere actually extends
because the 3 m irregularities responsible for Bragg downward to the altitudes of these echoes. But this is
backscatter at 6 m radar wavelengths (i.e., the irreg- true at other latitudes in daytime, so simply having
ularities responsible for VHF backscattering) should electrons is not the answer. Another possible answer
lie within the viscous subrange of turbulence at 86 km lies in the ice particles themselves: could they scatter
the waves just as rain and snow enhance weather radar
-
and, as a consequence, should be strongly damped.
As observed at VHF ( 50 MHz), northern-hemi-
spheric PMSE exhibit the following characteristics:
signals? Again the answer is no. Such particles are too
few and too small to create a huge signal. But, as we
shall see, collocation of the electron gas and the finite-
-
They display a thin but intense echoing region near
the summer mesopause ( 85 km) with a margin of
f3 km.
sized (nonmolecular) particles seems to hold clues to
the phenomena.
Rocket data of measurements of the electron struc-
While some relatively strong, albeit sporadic, ture in the medium provided the first clear indication
echoes have been reported at latitudes as low as that the echoes were related to the coupling of
52", the strongest, most continuous echoes are electrons and small particles. Figure 7 shows simulta-
observed at latitudes poleward of about 65". neous VHF scatter echo profile and the electron
The echoes appear around mid-May, last until mid- density during a strong event. A severe, sharp biteout
August, and are relatively continuous. in the latter is coincident with the echoing region. It
Both the height range and seasonal variations of seems curious that removing electrons can increase the
PMSE correlate reasonably well with those of the
cold temperature mesopause (i.e., the coldest
known atmospheric region).
Joint observations using VHF radar and sounding
rockets show that intense PMSE can be often
associated with sharp 'biteouts' in the ambient
electron density.
Subsequent to their discovery in 1981, many observa-

tions related to PMSE have been made using radar,
lidar, and rockets. These observations have helped
formulate a number of theories proposed to explain
the generation of the intense radar echoes and the
remarkable physical conditions associated with them. Figure 6 Contour plot of the signal-to-noise ratio at Resolute
Subfields in research as disparate as dusty (icy)plasma Bay, Canada (75" N, 95" W). The time axis is in UTC.
1276 MESOSPHERE I Polar Summer Mesopause
spectrum implied that this explanation could not stand

alone. Furthermore, it is difficult to explain the PMSE
by traditional theories of backscatter from turbulent
structures in ionized gases. This is because the spatial
extent of the density inhomogeneity 'seen' by a radar
must be on the order of the radar half-wavelength
( 3 m, 0.67 myand 0.12 m for 50 MHz, 224 MHz, and
1.29 GHz, respectively). Such short scales are much
smaller than the viscous cutoff scale and should not
exist in the neutral gas, which supports the turbulence.
Traditional theories therefore suggest that, based on
turbulence arguments alone, PMSE cannot exist. The
fact that the PMSE are detected at meter wavelengths,
however, is consistent with a more complex turbulence
theory that the allowed scales for the passive scalar
electron gas can be much smaller than those for the
neutral gas. This idea invokes the possibility that
PMSE are caused by fluctuations in the electron gas,
which, in the presence of a mixture of charged heavy
ice particles, can be characterized by an inner scale
80 smaller than the Kolmogorov inner scale of neutral
102 103 104 turbulence.
Electron density ( t r 3 ) The occurrence of PMSE in the summer polar
Figure 7 Comparison of rocket measurements of the electron
mesosphere is undoubtedly linked to very cold tem-
density profile (solid line) with the simultaneous 53.5 MHz SOUSY peratures. Simultaneous measurements of PMSE with
radar observations of radar reflectivity (circles) during the MAC/ radar and of temperature with rockets have shown
SINE campaign (14 July 1987; 0929 UT). (Courtesy of Alcala CM that a sufficiently cold mesopause temperature is a
and Kelley MC (2001) Nonturbulent layers in polar mesosphere necessary but not a sufficient condition for the
summer mesosphere, 2. Application of wavelet analysis to VHF
scattering. Radio Science 36: 891.
existence of PMSE. A study of PMSE occurrence and
low seasonal mesopause temperatures, whose results
are presented in Figure 8, has shown that the proba-
bility curve of seasonal occurrence of PMSE in the
signal, but if the resulting coupled electron-ice gas has Northern Hemisphere was similar to, but delayed by
very sharp edges or small eddies, the crucial require- about 10 days from, a mean seasonal curve of low
ment that there be structure at half the radar wave- mesopause temperatures deduced from a number of
length may be accomplished. Spectral analysis of the measurements.
rocket data indeed shows that for this VHF The cold mesopausePMSE occurrence is further
(53.5 MHz) study the electron gas was structured at supported by recent measurements of PMSE in the
this small scale. Southern Hemisphere at Machu Picchu Station
Unfortunately, there is not yet an established theory (62" S), which showed that PMSE are much weaker
that adequately explains the entirety of this phenom- and more sporadic in the Southern Hemisphere. One
enon. In the beginning, many of these studies were reasonable inference emerging from this observation is
based on observations at 50 MHz. When later studies that the mean mesopause temperature at southern
showed that PMSE were observed at higher frequen- high latitudes is warmer than its northern counterpart.
cies (UHF),newer processes were proposed. Now that Indeed, preliminary examination of daily-averaged
echoes have also been observed using radars at summertime temperature using satellite data (HRDI,
lower frequencies (HF), this phenomenon becomes High Resolution Doppler Imager) at 84 f 1.5km
even more intriguing and open to new theoretical suggests a few degrees of temperature difference
interpretations. between comparable summertime months found at
One of the earliest studies suggested that PMSE 65" latitude, with the Southern Hemispheric temper-
could be explained by enhanced turbulence arising atures being warmer. These results support previous
from the breakdown of tides and low-frequency indications of a warmer Southern Hemisphere from
gravity waves. However, the lack of evidence for other satellite observations (SME, Solar Mesosphere
mrbalmt breadming of the repreczted signal in pro- Explurer).w-hen &e tempi-atures were wamei- in &e
portion to the scattered power in the radar return Southern Hemisphere by about 4 K.
MESOSPHERE / Polar Summer Mesopause 1277
Figure 8 Comparison of percentage of occurrence of polar mesospheric summer echoes (PMSE) at Poker Flat, Alaska, and various
estimates of high-latitude seasonal mesospheric temperature fluctuations. (Adapted from Balsley BB and Huaman MM (1997),On the
relationship between seasonal occurrence in northern hemispheric polar mesosphere summer echoes and mean mesopause
temperatures. Journal of Geophysical Research 102: 2021 .)
Another study supporting north-south asymmetries presence of a large number of charged ice particles or
suggests that PMSE can be detected at lower latitudes very large water-cluster ions, making a favorable
in the Northern Hemisphere than in the Southern environment for the occurrence of PMSE. These ice
Hemisphere, and it seems that PMSE start occurring in particles, on occasion, are large enough that we can see
an earlier stage in the south than in the north with them from the ground as clouds (NLC) or detect them
respect to the solstice. These results could help in by lidar scatter. Figure 9 is a simultaneous measure-
understanding why PMSE were not detected late in the ment of lidar and radar backscatter from the polar
season at Machu Picchu Station. Collection of data at summer mesosphere. Notice that the lidar signal
Machu Picchu Station has not been possible earlier (which detects the largest ice particles) is at the lower
than late December for logistic reasons, thereby edge of the radar signal. This seems a clear indication
missing late November and early December observa- that the radar detects small particles as they fall and
tions when the occurrence of PMSE is supposed to grow larger to be detected by the lidar just before they
start. begin to sublimate as the temperature rises.
Additional support for a causal temperature mech- All this discussion requires that PMSE generation is
anism is that it has been shown that a significant somehow associated with charged aerosols and, pri-
reduction in plasma diffusivity would occur in the marily, ice particles. This suggests again that, in
for NLC generation, although not the only one. About

half of the mesospheric water vapor is believed to
come from the photodissociation and oxidation of
upwardly transported CH4 with the chemically active
radical (OH). Colder temperatures and more water
vapor can produce more NLC events, so they can be
used as indicators of global change.
The necessary conditions for the occurrence of
PMSE and NLC or PMC appear to be similar: they
require low temperatures and are apparently related to
water vapor. The seasonal PMSE occurrence corre-
sponds well with the high-latitude seasonal occurrence
of NLC. Recent studies have shown a correlation of
Figure9 Relationship between PMSE (light contours) detected PMC with PMSE in the Northern Hemisphere where
by radar and NLC particles (shaded regions) detected by lidar. the mean long-term PMC occurrence ratio curve fits
(Adaptedfrom von Zahn U and Bremer J (1999) Simultaneousand symmetrically inside the PMSE occurrence ratio
common-volume observations of noctilucent clouds and polar (Figure 8). A close correlation between NLC and
mesosphere summer echoes. Geophysical Research Letters 26: PMSE has also been observed using lidar and radar
1521.)
data, respectively. These studies were made using a
common volume and they agree most of the time. Thus
addition to cold temperatures, water vapor is also a PMSE are of particular interest in view of their
necessary ingredient in the equation for PMSE gener- frequent coincident occurrence with NLC and the
ation. The differences between the seasonal variation possible association of recent increased detection of
of PMSE and temperature were speculated to be NLC with global warming trends. If we could monitor
caused by water vapor. PMSE for long periods and observe increaseddecreas-
Deep electron density biteouts have been observed es in these events (and their relative strength) over
with rockets and have also been seen in electron time, we could use such information as a possible
density profiles obtained using incoherent-scatter indicator of global change.
radar techniques. How does this scavenging of elec-
trons occur and how can the edges be so steep? We do
not know yet, but anomalously low diffusivity in the See also
presence of charged ice particles has been proposed. Carbon Dioxide. Global Change: Upper Atmospheric
The coupling of these particles with ambient D-region Change. Middle Atmosphere: Gravity Waves. Noctilu-
electrons allows electron density fluctuations of much cent Clouds. Parameterization of Physical Process-
smaller spatial scales than expected from viscosity es: Gravity Wave Fluxes. Radar: MST and ST Radars and
arguments. The fact that PMSE could be detected by Wind Profilers.
higher-frequency radars was confirmed by a 224 MHz
radar and at still higher-frequencies (933MHz,
1.29GHz). The low diffusivity ideas might well Further Reading
explain echoes at 224 MHz, but echoes near or above Cho JYN and Kelley MC (1993)Polar mesosphere summer
1GHz are truly remarkable and may require a whole radar echoes: observations and current theories. Reviews
new theory. of Geophysics 31: 243-265.
Cho JYN and Rottger J (1997)An updated review of polar
On the Possible Relationship between mesosphere summer echoes: observation, theory and
their relationship to noctilucent clouds and subvisible
PMSE, NLC, and Global Change aerosols. Journal of Geophysical Research 102: 2001-
As time has passed, more observations of NLC have 2020.
been reported and an increasing trend has been Gadsden M and Schroder W (1989) Noctilucent Clouds.
New York: Springer-Verlag.
detected. This increment in NLC is explained by the Rottger J (1994) Middle atmosphere and lower thermo-
observed increment in atmospheric methane (CH4) sphere processes at high latitudes studied with the
and carbon dioxide (COS) due to human activity. A EISCAT radars. Journal of Atmospheric and Terrestrial
doubling increment in either of these components will Physics 56: 1173-1196.
produce a cooling of the thermosphere and meso- Thomas GE (1991) Mesospheric clouds and the physics
sphere by about 50 K and 10 K, respectively. Remem- of the mesopause region. Reviews of Geophysics 29:
ber that cold temperatures are a necessary condition 553-5 75.
METEORS 1279
P Jenniskens, SET1 Institute, Moffett Field, CA, USA If the meteoroids are sufficiently small (<50 pm)
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. and come in slow enough (<20kms-I), and if they
are large enough to effectively radiate the heat
( >10 pm), then they may survive the heating process
almost intact. These particles are collected in the
Nomenclature of Sources and Sinks stratosphere and are called Brownlee particles, inter-
The Earth’s upper atmosphere is continuously being planetary dust particles, or micrometeorites. They are
bombarded by solid objects from interplanetary space. also collected in sediments on Earth’s surface, where
Their constituents may have provided prebiotic or- they are called exclusively micrometeorites. Micro-
ganic molecules for the origin of life, and even today meteorites tend to have an asteroidal origin, like the
this influx of solid particles is responsible for the larger meteorites, because cometary material tends to
continuous replenishment of a layer of metal atoms at be more fragile and enter Earth’s atmosphere at higher
an altitude of between 80 and 1lOkm. Even the speed and therefore tends to not survive the ablation
smallest grains are precious momenta of the origin of process.
the solar system and trace the dynamical processes of
formation and disintegration of the first kilometer-
sized planetesimals called comets and asteroids, from Annual and Daily Variation of
which they originate. The daily influx of meteoroids is Meteoroid Influx
the more relevant for most studies of atmospheric
chemistry and physics. The much more dramatic, but The daily influx of extraterrestrial matter (Figure 1) is
rare, impacts of comets and asteroids will not be dominated by particles of size 100-200pm, to the
discussed here.
Meteoroids, also called interplanetary dust parti-
cles, carry much kinetic energy, about 1 % of which
is converted to visible light in collisions with air
molecules, creating the transient phenomenon called a
meteor. This energy and momentum transfer results in
atmospheric chemistry and meteoroid ablation, and
leaves behind a range of atomic, molecular, and solid
particle products.
The following products have been identified in
Earth’s atmosphere. Meteor trains of enhanced den-
sities of electrons, oxygen atoms, and meteoric metal
atoms are routinely observed by radar, lidar, and other
remote sensing techniques. Meteoric neutral metal
atoms are the source of the sodium airglow emission
and form a layer between about 80 and 110 km, called
the neutral atom debris layer. These metal atoms can
chemically react and condense into nanometer-sized
solid particles called recondensed meteoric vapor.
There is indirect evidence that nanometer-sized solid
particles reside at altitudes of 70-85 km, where they
act as nucleation sites for water vapor to form
noctilucent clouds and are a source of weakly bound
electrons. Larger solid particles, of size 10-100 pm,
that survive the ablation process are called meteoric
debris and have been collected in the stratosphere and
at Earth’s surface. Small, nanometer-sized meteoric
matter with chondritic abundances has been detected Figure I Overview of the influx of extraterrestrial matter in
as a contaminant in (or nucleation site of) larger Earth’s atmosphere and the nomenclature for sources, phenom-
sulfuric aerosols in the stratosphere. ena, and products. Strength of 4 m peak is uncertain.
1280 METEORS
amount of 4 x IO'kgyr-' (for sizes up to 300pm). that is, radiant in the zenith and the star limiting
These meteoroids have a mean impact velocity of magnitude = 6.5):
25kms-', but with a wide range from 11 to
72 km s - '. Depending on the latitude of the observing ZHR = ZHR,,, x 10-~B(io-Ao~max)~[TI
site, most cometary matter in retrograde orbits and the
dust particles of the zodiacal cloud that have nearly This defines the duration parameter B (degrees-I). 20
circular orbits are accreted in the morning hours, (JZOOO) is the solar longitude, a measure of time and
between 3.00 and 6.00 a.m. local time (the apex related to the Earth's position in its orbit.
source), when the direction of Earth's motion is Meteor outbursts represent even larger flux varia-
highest in the sky. For the same reason, the annual tions, that are observed when Earth crosses a recently
variation of meteor rates tends to peak in autumn on formed comet dust trail. In that case, rates can increase
the Northern Hemisphere. On the other hand, the fall up to 5 orders of magnitude over a period of 1-2 hours.
of meteorites and micrometeorites peaks in the after- If the meteor rate increases to above 1 per second (or
noon hours because of the predominantly prograde ZHR > 1000 h-'), the event is called a meteor storm.
orbit of the asteroidal matter and the low encounter More important than such arbitrary nomenclature are
speed needed for survival of this material (Figure 2). the dynamical processes that underlie these manifes-
The daily sporadic influx is interrupted by meteor tations. Comet dust trails are a manifestation of
showers, which are significant especially for relatively ejected dust grains returning at different times from
large meteoroids that cause visible meteors (10 - g to their orbit around the Sun because of slightly different
l o 3g). Most well-known examples are the Perseid orbital periods. The encounter conditions with fresh
shower in summer and the Geminid and Quadrantid comet dust trails vary from year to year because the
showers in winter (Table 1). Meteor showers are position of the trails relative to Earth's orbit is a
caused by dust grains released from comets, which all function of the combined influence of planetary
still approach Earth on nearly parallel orbits from perturbations, which move the trails in a pattern that
a direction called the shower radiant (RA, DEC in mirrors the Sun's reflex motion around the barycenter
Table 1).Unlike the sporadic background, the meteor of the solar system. Over time, the natural waving
shower size distribution is not collisionally equilibrat- motion of the trails fades into a broader stream called
ed. Most mass is in the larger grains, which replenish the filament, which is due to small perturbations of the
the population of smaller meteoroids. The number orbital period of grains so that different sections of a
distribution index r = N ( m + l ) / N ( m ) tends to be dust trail catch up on each other. The filament causes
lower than those of the sporadic background. r relates meteor outbursts with a duration of about a day. Such
to the size distribution index: s = 1 + 2.5 lg r. structures persist only when the grains become
During annual meteor showers, rates vary over time trapped in mean-motion resonances that prevent close
scales of days to weeks. The variation of rates is encounters with the planets, because such close
usually expressed in units of number influx called the encounters will put the grains in the annual shower
zenith hourly rate (ZHR; the rate of visible meteors component.
seen by a standard observer under ideal conditions,
Meteor Luminosity
-
-
80 -
-1 week Perseids 1
When a meteoroid encounters the Earth's atmosphere,
an initial phase of V-shaped luminosity is seen,
possibly caused by the interaction of released electrons
a
4- with the ambient ionosphere. Leonids 1kg in size have
e
-h
L
60 - been detected as high as 196 km.
3
r
0
Most meteors are not seen until around 135-
5 40 -
120 km, where massive evaporation of meteoric sili-
c cates starts to occur. The process is one of sputtering,
Direction of
it where each impact of an air molecule releases a cloud
of atoms and molecules from the meteoroid. That
builds a small vapor cloud traveling along with the
meteoroid. Subsequent air collisions are with the
100 110 120 130 140 150 160
vapor cloud, creating a warm ( T = 4400 K) plasma in
Solar longitude (J2000) the immediate wake of the meteoroid.
Figure 2 Typical variation of annual meteor shower activity. The When the pressure of the vapor cloud exceeds that of
asymmetry is the result of a gradually precessing orbit. the ambient air, it expands as into a vacuum. The
METEORS 1281
Table 1 Meteor showers

Shower n y ZHRmax B RA, DEC Vlnt rb
Quadrantids 283.32 130 2.5230 +49 43 2.2

y-Velids 285.7 2.4 0.12 124 - 47 35 3.0
(-Aurigids 285.8 (5) 0.20 077 +58 16 -_
Corona-Borealids 291.1 (15) 0.20 230 + 37 37 --
6-Cancrids 298.9 (11) 0.20 132 +20 28 --
cr-Leonids 309.0 (7) 0.10 157 +06 29 --
a-carinids 311.2 2.3 0.16 091 - 54 25 2.5
a-Centaurids' 319.4 7.3 0.18 209 - 58 57 2.3
o-Centaurids 323.4 2.2 0.15 175 - 57 51 2.8
&Leonids 334.7 1.1 0.049 154 +19 23 3.0
n-Virginids 343.9 (2) 0.10 184 +01 32 3.0
S-Virginids 359.7 (4) 0.05 196 - 01 28 3.0
N-Virginids 013.7 (3) 0.05 210 - 08 30 3.0
y-Normid 353.0 5.8 0.19 249 - 51 56 2.4
6-Pavonids 011.1 5.3 0.075 309 - 63 60 2.6
April Lyrids' 032.4 12.8 0.22 272 +33 49 2.7
p-Virginids 039.7 2.2 0.045 229 - 06 30 3.0
q-Aquarids 046.5 36.7 0.080 339 - 01 66 2.7
cr-Scorpiids 055.9 3.2 0.13 251 - 23 35 2.5
May Arietids 054.2 (7) 0.10 036 +18 27 --
o-Cetids 064.7 (7) 0.08 028 +01 35 --
o-Scorpiids 072.6 5.2 0.15 240 - 20 21 3.0
Daytime Arietids 075.9 54 0.10 038 +24 38 2.7
(-Perseids 078.8 (17) 0.05 070 +27 27 -_
y-Saggitarids 089.2 2.4 0.037 285 - 27 29 2.9
r-Cetiids 095.7 3.6 0.18 096 +23 66 2.5
P-Taurids 096.7 (20) 0.08 086 +19 30 --
0-Ophiuchids 097.7 (2.3) (0.037) 251 - 15 27 2.8
.r-Aquarids 098.0 7.1 0.24 342 - 12 63 2.5
u-Phoenicids 111.2 5.0 0.25 026 - 41 48 3.0
o-Cygnids 116.7 2.5 0.13 305 + 47 37 2.7
a-Capricornids 122.4 2.2 0.041 301 - 10 25 2.0
6-Aquardids N 124.1 1 .o 0.063 324 - 12 42 3.3
Pisc. Australids 124.4 (2.9) (0.26) 337 - 33 42 3.2
&Aquarids S 125.6 11.4 0.091 339 - 17 43 3.3
i-Aquaris Z 131.7 1.5 0.070 334 - 15 36 3.3
Perseids' 140.19 84 0.20 046 +58 61 2.5
K-Cygnids 146.7 2.3 0.069 290 +52 27 2.2
?-Dorids 155.7 4.8 0.18 060 - 50 41 2.8
Aurigids 158.2 (9) 0.19 072 + 43 69 2.7
K-Aquarids 177.2 2.7 0.1 1 338 - 05 19 2.8
Sextantids 186.7 (9) 0.08 154 + 00 32 --
Draconids' 197.2 2 0.20 262 +54 20 3.0
S-Arietids 201.7 (3) 0.03 035 +10 28 --
&-Geminids 206.7 2.9 0.082 103 +28 71 3.0
Orionids' 208.6 25 0.12 095 +16 67 3.1
Leo-Minorids 209.7 1.9 0.14 159 +38 61 2.7
Taurids 223.6 7.3 0.026 049 +18 30 2.3
(-Puppids 232.2 3.2 0.13 117 - 42 41 3.4
Leonids' 235.6 13 0.20 153 +22 71 3.0
PuppidsNelids 251.7 4.5 0.034 128 - 42 40 2.9
Phoenicids 252.4 2.8 0.30 018 - 58 18 2.8
Monocerotids 260.9 2.0 0.25 100 +14 43 3.5
Geminids 262.1 88 0.39 113 + 32 36 2.6
cr-Hydrudids 265.5 2.5 0.10 133 - 01 59 3.0
Ursids' 271 .O 4.0 0.20 220 + 75 35 3.0
'Occasional outbursts.
3.4 for sporadic meteors.
1282 METEORS
density falls off rapidly with increasing distance from Observed dependencies are less steep (q x p:.4-0.8),
the meteoroid, while the temperature increases. This perhaps due to fragmentation. At high elevations, the
region may be the source of an emitting gas attenuation due to this dispersion leads to low electron
( T x 10 000 K) detected in bright fast meteors. The densities and an ‘echo height ceiling’ effect in back-
size of the vapor cloud is a function of particle mass scatter radar. For a radar at any given wavelength there
and speed, and about 1-10 orders of magnitude larger is a height beyound which no underdense echoes can
than the size of the meteoroid (Figure 3). be seen.
The wake of the meteoroid is formed by the ‘Overdense echoes’ occur when the electron density
impacting air molecules and by the vapor cloud remains high enough to cause mirror-like reflection.
products that receive any amount of momentum Sometimes, there are also ‘head echo’ reflections,
transfer, that is after a single collision. This is the which are the signature of a hard (unresolved) target
main source of light emission at visible wavelengths crossing the radar beam: a spherical region of ioniza-
and reflections. Head echoes are the signature of a tion that travels along with the meteoroid and may be
hard (unresolved)target crossing the radar beam. The related to the spherical region of optical luminosity in
decelerated atoms will expand into the ambient air the bottom part of Figure 3.
until stopped by subsequent collisions. The mean free The observed light intensity (I)and electron line
path determines the initial radius ( q )of the wake, density ( 4 )are proportional to the rate of loss of kinetic
which is of the order of a few meters and has the energy from the ablated atoms:
following expected dependence on velocity and at-
mospheric mass density (pa): dm
I = 0.5~-V’ 131
dt
with z the dimensionless luminous-efficiency factor
z M 0.01, depending on the particulars of the meteor
emission spectrum (changing with speed, size, etc.)
and the instrument response.
The blue and green part of the optical spectrum
consists mainly of ablated metal atom emission lines
with a T M 4400 K, mostly from iron and magnesium
lines. Fast and bright meteors show the violet Cat
doublet. The orange and red part of the spectrum are
dominated by neutral oxygen and nitrogen lines and
by the first positive bands of molecular nitrogen.
Those atmospheric lines and bands are also well
described by T M 4000 K vibrational temperature,
electronic excitation temperature, and chemical equi-
librium. Many deviations from local thermodynamic
equilibrium are observed, however, and are the focus
of study. Emissions from small organic molecules such
as CN have been reported, but with insufficient
certainty to derive quantitative information. The
thermal infrared spectrum remains unexplored, while
the near-UV spectrum is dominated by lines of Mg and
Mg+ at 280 and 285 nm (Figure 4).
For the typical response of the photographic plates
used in the Harvard Super Schmidt program (mph),the
brightness as a function of mass is given by
lg M(g) =5.15 - 0.44*mph

- 3.89 lglo Vinf (kms-l)
- 0.67 lglo [sin(h,)] [41
Figure 3 Meteor anatomy. Meteor vapor cloud in a model by
Olga Popova, whereby gas densities are shown. Meteor wake in a
model by lain Boyd, showing translational temperatures. And a
where h, is the angle of incidence. The magnitude mph
high frame-rate still image of a - 3 magn. Leonid meteor by Hans refers to an ‘absolute brightness’ at a distance of
Stenbaek-Nielsen. 100km. On this scale, a zero-magnitude Leonid
METEORS 1283
where max denotes the altitude ( H ) of the peak

brightness, which varies with entry velocity, zenith
angle, and meteor brightness:
H,,, (km) = 21.2 + 44 Ig,, V (kms-l)

+ 1.1Mmax 181
(and less for higher than typical meteoroid density).
Lightcurves from cometary meteoroids tend to be
flatter than this and are better described by a sum of
such classical lightcurves, the assumption being that a
matrix material is evaporated or bonding sufficiently
weakened for the meteoroid to break apart. Because
400 500 600 700 800 there is not much momentum transfer to the grains,
Wavelength (nm) these fragments move along as a cloud of solid bodies.
Because the larger fragments penetrate deeper than do
Figure 4 Optical spectrum of a Leonid meteor from data by
Shinsuke Abe.
the smaller ones, the size distribution determines the
shape of the composite lightcurve. In the calculations,
it is important to take into account the higher
deceleration of the smaller fragments, with the resist-
would weight O.O7g, and would emit visible light ance or drag equation:
(400-700 nm) at a rate of about 500 W.
The relationship between visual meteor magnitude
(m,) and the number of electrons formed per meter of [91
meteor path (the electron line density) is
with r ( = 0.5-1.0) the drag coefficient. Evidence for
m, = 38 - 2.5 lglo q (m-') + 2.5 lg V(kms-l) [5] such fragmentation comes from the early release of
sodium relative to magnesium in optical spectroscopy,
This density pertains to the wake shortly after it has the detection of jets of debris fragments ejected by
formed. The density decays rapidly by ambipolar rapidly spinning meteoroids, and larger-than-ex-
diffusion, with a coefficient D x 4.2 m2 at 93 km. pected emission volumes, among others.
The peak mass loss is also a function of the
Altitude of Deposition sublimation temperature of the mineral components
of the meteoroid. Differential ablation is the term used
How is the ablated matter (and amount of ionization/ to indicate the deposition of matter in successive order
chemistry) distributed as a function of altitude? The of volatility. The most volatile organic compounds
traditional approach is to assume the meteoroid mass with T x 500 K have their peak loss as high as 117km
loss to be proportional to the kinetic energy trans- in a fast 71 km s - Leonid, while the most refractory
ferred to the intercepted air mass: compounds with T = 2400K are expected to have
peak mass loss at 105 km altitude. For slower T =
25kms-1 meteors, the organics are not lost until
90 km altitude in such models (Figure 5 ) .
Evidence for differential ablation is much less
where ( is the heat of ablation of the meteoroid abundant than expected. In general, optical emission
material (0.2 - 1.0 x lO3Jg-'); A is the dimension- spectra do not show evidence for differential ablation.
less heat transfer coefficient that specifies the efficiency Only in unusually fragile meteoroids do we see an
of the collision process in converting kinetic energy to earlier onset of the sodium than magnesium atoms,
heat (0.6, down to O . l ) , A the effective surface area, thought to be because the minerals involved are more
and p the density of the meteoroid (0.1-1.5 g c m P 3 ) . efficiently exposed to the meteoric vapor plasma.
From this, the classical expression for the rise and Lidar observations of neutral atom debris trails,
fall of meteoric luminosity (or ionization) is derived. however, tend to show only one of several metal
For given atmosphere density profile p,(H): atoms at a given altitude, which has been interpreted
as evidence for differential ablation. Such trails are
detected only minutes to hours after deposition, when
they drift by the lidar beam (Figure 6).
1284 METEORS
200 emission at 557nm from the 0 ('S-ID) transition is

200 seen in fast meteors, even relatively faint ones. The
180
h emission is thought to be produced by the Barth
3
v 160
9 mechanism. Intensity estimates show that about 15%
E
.-CI) 140
Begin (photo)
.. .... .
b 8..
150
Q of the 0 2 in the initial train is dissociated in a -12
.-. 2 magnitude Leonid.
Q)
r ln
p 120 z Radar reflections have detected an extended wake
.-
9.
-3
h
.-E0) 100 out to 6 km and several seconds behind a meteoroid

100
that contributes to nonspecular reflections at UHF and
m" 80
.. VHF frequencies. Plasma instabilities and turbulence
60 are responsible for an anomalous cross-field diffusion
-15 -10 -5 0 5 of meteor trails in the Earth's magnetic field that can be
Magnitude up to an order of magnitude faster than the rate
expected from ambipolar diffusion.
Figure 5 Beginning and end heights of Leonid meteors.
In bright fireballs, this wake is accompanied by an
afterglow rich in metal atom line emissions, without
the atmospheric emission lines that are typical for the
meteor spectrum. The afterglow has been interpreted
as due to secondary ablation from debris particles.
Evidently, even a fast meteor can deposit solid debris in
vapor cools from -4400 K to -

its path, if conditions are favorable. The ablation
1200 K in a few
seconds. This afterglow is followed by a recombina-
tion emission phase that lasts tens of seconds. Mid-
infrared emission remains detected for longer,
even when the gas and dust has cooled to just 5 0 K
over the ambient T M 250 K a few minutes after the
fireball.
A persistent chemiluminescence remains for Leo-
nids < -3 magnitude, which is called the persistent
train (Figure 7).Those persistent trains can be visible
-
at 1 3 magn/arcsec2 for many (tens of) minutes. The
dynamical mechanism is not yet fully understood.
Typically, two bright bands of light are seen with
various levels of billowing. The train spectrum
Figure 6 Lidar response to a neutral iron atom debris layer, shows sodium emission lines and a broad molecular
measured by University of Illinois Fe boltzmann lidar. band continuum identified as the FeO orange arc
band. The luminous mechanism is thought to be
the catalytic recombination of ambient ozone with
Conditions in the Meteor Path after
Deposition
Once the air is heated and meteoric matter has been
deposited, the resulting pressure increase by more than
a factor of 10 creates a shock wave which expands
radially. Assuming that this expansion occurs adia-
batically, then the pressure will equilibrate with the
background atmosphere when the radius is about
70 m (example for a -12-magnitude Leonid). Even at
the very low pressures of the upper mesosphere/lower
thermosphere (<10 - hPa), the size and velocity
(Mach 270) of such a meteoroid would create a
turbulent wake (Reynolds number > 2000).
At this stage, the meteor is detected in several Figure 7 Persistent train of a Leonid fireball. (Photo 0R. Haas,
manners. First, a wake of the forbidden 'green line' Dutch Meteor Society.)
METEORS 1285
oxygen atoms in the trail through the Chapman nonequilibrium processes and have not yet been
airglow mechanism: calibrated by observations. The chemical changes of
organic compounds in meteors are of particular
Na+03 +NaO+02 1101 interest, because meteors may have contributed more
than two-thirds of all infalling organic matter on the
+
N a O 0 + Na(32P,32S) 0 2 + [I11
early Earth, in a form determined uniquely by the
chemistry in the meteor wake and subsequent terres-
trial evolution.
where the branching ratio of reaction [ 111 to produce Meteoric metals and ions contribute to a steady-
the Na (32P)state (which then emits an orange photon state condition in the upper atmosphere. The influence
at 589 nm) is about 10%. The FeO molecular emission of meteors may be measured directly during unusual
band probably arises from disturbances of meteoric influx. Indeed, variations of
O H airglow intensity with meteor shower rates have
+ +
Fe 0 3 + F e O ( 3 etc.) 0 2 1121 been reported in conjunction with the 1999 Leonid
meteor shower. Tentative associations have also been
+
FeO 0 + Fe 0 2 + ~ 3 1
made between meteor ionization and the occurrence
of sprites and elves during lightning storms. Lightning
has been observed to travel along the ionized path of a
where reaction [I21 is sufficiently exothermic to meteor. Much progress in these fields is expected in the
produce FeO in excited electronic states, leading to near future.
emission between 570 and 630 nm with about a 2%
efficiency. Thus persistent trains serve as a model for
natural airglow emission, in extreme conditions and See also
with the reactive components separated. The trains
also probe upper atmosphere winds, wind sheer, and Aerosols: Role in Cloud Physics. Carbon Dioxide.
diffusion rates. Lightning: Overview. Noctilucent Clouds. Tropo-
O n the relatively short time scale of the train spheric Chemistry and Composition: Carbon
Monoxide.
(minutes rather than hours), and particularly in the
presence of elevated concentrations of atomic oxygen,
it is very unlikely that the metallic species would be
able to form more stable reservoir compounds such Further Reading
as N a H C 0 3 or Fe(0H)Z. Indeed, between 85 and Bronshten VA (1983) Physics of Meteoric Phenomena.
100 km the meteoric metals are overwhelmingly in the Dordrecht: Reidel.
atomic form in the background atmosphere. The Ceplecha Z, Borovicka J, Elford WG, et al. (1998)Meteor
postulated formation of nanometer-sized recondensed phenomena and bodies. Space Science Reviews 84:
meteoric vapor particles in the warm meteor wake has 327-471.
not yet been demonstrated. However, when the metal Jenniskens P (2001)Meteors as a vehicle for the delivery of
atoms settle to lower elevations, they quickly react organic matter to the early Earth. ESA-SP 495. In:
chemically and can condense to form particles. Most Warmbein B (ed.) Proceedings of the Meteoroids 2001
Conference, Swedish Institute of Space Physics, Kiruna,
of this fine-grained material is expected to be trans- Sweden, 6-10 August 2001, pp. 247-254. Noordwijk:
ported to one of the poles, following seasonal winds in ESA, ESTEC.
the upper mesosphere. Levin (Lewin) BY (1961) Physikalische Theorie der
Meteore und die meteoritische Substanz im Sonnensys-
tem, vol 11. Scientia Astronomica 4 , Berlin: Akademie-
Impact of Meteors on the Atmosphere Verlag.
Much remains unknown about the chemistry in the McKinley DWR (1961) Meteor Science and Engineering.
meteor wake plasma. Molecular abundances for New York: McGraw-Hill.
equilibrium air plasmas show that the T E 4400K Rietmeijer FJM (2002) Interrelationships among meteoric
metals, meteors, interplanetary dust, micrometeorites,
temperature enables interesting organic chemistry in and meteorites. Meteoritics and Planetary Science 35:
C02-rich atmospheres, because not only C 0 2 but also 1025-1 04 1.
CO is dissociated. At slightly lower temperatures, Warmbein B (ed.) (2001) Proceedings of the Meteoroids
much of the electron charge is balanced by N O + , while 2001 Conference. ESA-SP 495. Swedish Institute of Space
at higher temperatures much of the charge is carried by Physics, Kiruna, Sweden 6-10 August 2001. Noordwijk:
Cf or O+. NO production rates may be affected by ESA, ESTEC.
1286 METHANE
E Dlugokencky,NOAA Climate Monitoring and Since many of the sources that emit CH4 to the
Diagnostics Laboratory, Boulder, CO, USA atmosphere are diffuse and highly variable in space
S Houweling,Space Research Organization and time, estimating total emissions is difficult. (This is
Netherlands, Utrecht, The Netherlands in contrast to anthropogenic compounds such as the
chlorofluorocompounds, whose emissions are deter-
mined from production.) Constraining the global CH4
budget requires a range of studies including back-
ground atmospheric measurements, laboratory and
Introduction field studies of CH4 emission rates and the factors
In the year AD 2000, less than 2 parts in l o 6 (on a affecting these emission rates, and computer modeling
molar basis) of the Earth’s atmosphere was methane studies.
(CH4),but this is a factor of 2.5 more than was present
in 1750, prior to the industrial revolution. Even
Atmospheric Methane Measurements
though this abundance is relatively small, CH4 is the
most abundant organic compound present in the Ice Core Measurements
atmosphere, and it significantly affects Earth’s climate
and atmospheric chemistry. Systematic modern measurements of atmospheric
Methane is a greenhouse gas because it absorbs CH4 abundance began in 1978. Our knowledge of
infrared radiation in a region of the terrestrial IR CH4 abundance prior to that comes from analysis of
spectrum that is unaffected by water vapor and carbon bubbles trapped in cores drilled and extracted from
dioxide (COZ),the two most important greenhouse polar ice. It is assumed that, for methane (and many
gases. Increasing amounts of atmospheric methane other stable, long-lived species), the air trapped in
will potentially contribute to climate change. Methane polar ice bubbles accurately represents the atmosphere
-
currently contributes 0.5 W m-’ to the total direct
radiative forcing (estimated to be 2.5 W m P 2 )caused
at the approximate time period when the ice was
formed. The longest record of atmospheric CH4
by increasing atmospheric burdens of long-lived abundance is for the past 420 000 years, based on an
ice core from Greenland. This record shows that prior
-
greenhouse gases since 1750. Only C 0 2 contributes
more to this total ( 1.5 W m - 2 ) .
Methane is removed from the atmosphere predom-
to industrialization, atmospheric methane varied from
about 350 parts per billion (ppb) during glacial times
inantly by gas-phase oxidation initiated by hydroxyl to 700 ppb during interglacial times. (Note: most
radical (HO). Changes in the abundance of atmos- methane measurements are reported in a dry-air, mole
pheric CH4 affect the oxidizing capacity of the fraction scale; ppb nmol mol - I.) There is excellent
atmosphere. It is predicted that increasing CH4 will correlation between CH4 abundance and indicators of
decrease the concentration of HO in the atmosphere, temperature extracted from the ice. In Figure 1,
so it will have an indirect effect on climate by affecting globally averaged CH4 mole fractions estimated
the atmospheric residence times of other reduced, from CH4 measurements in Antarctic ice cores,
long-lived greenhouse gases that are removed by Greenland ice cores, Antarctic firn, archived air from
reaction with HO. Under conditions where the con- Tasmania, and a globally distributed network of air
centrations of compounds called nitrogen oxides sampling sites (since 1984) are plotted for the past
(NO, = NO+NO;?) exceed a specific value, oxidation 1000 years. Detailed comparisons of measurements
of CH4 produces 0 3 , itself an oxidant and greenhouse from Arctic ice cores with those from Antarctic cores
gas. About 1 0 % of CH4 emitted to the atmosphere is indicate greater emissions in the Northern Hemi-
oxidized in the stratosphere, where it produces H20 sphere than in the Southern. A large increase in
vapor, again potentially affecting climate. methane begins -200 years ago, and it is related to
Since CH4 has the potential to impact climate increased CH4 emissions associated with the increased
significantly, it has been targeted by the Kyoto food and energy demands of a rapidly growing human
Protocol for reduced emissions. Strategies designed population.
to mitigate the potential impact of CH4 on climate
Modern Measurements
must rely on a detailed understanding of methane’s
atmospheric budget (i.e., the balance of sources that The current global distribution of atmospheric CH4 is
emit CH4 to the atmosphere, and sinks that remove it). based on measurements from sites that regularly
METHANE 1287
1750 1 0.5
n
1500 1 0.4 ‘E
-
N
s.
-
Q
Q
1250
0.3 .-
2
P
0
m-
0.2 .-$
loo0 1 0.1
e
2
500
1000
I 1
1200
I I
1400
I I
1600
I I
1800
1 1
2000
Year
Figure 1 Globally averaged CH4 mole fractions for the past thousand years determined from ice cores, firn, and whole air samples are
plotted with the left y-axis. (Values based on ice core and firn air are courtesy of D. Etheridge, CSIRO Division of Atmospheric Research,
Australia; modern record from CMDL cooperative air sampling network (available at www.cmdl.noaa.gov).) Increased radiative forcing
due to CHI since the preindustrial era, approximated as a linear scale, is plotted on the right y-axis.
receive ‘background’ air. We define as ‘background’ air amplitude, and CH4 values in general, are larger than
that is well mixed and representative of a large volume at the South Pole, and the data are more variable at
of the atmosphere. These observations generally fall Barrow. Since the samples from both sites are collected
into two categories. The first is measurements from with comparable equipment and methods, and both
discrete samples collected under predetermined mete- sets of samples were analyzed on the same analytical
orological conditions (wind speed and direction) at system, the larger variability at Barrow is due to real
low frequency (e.g., weekly) in vacuum-tight contain- variability in atmospheric CH4 and not to measure-
ers and shipped to a central laboratory for analysis. ment error. This is because there are many CH4 sources
The second method is measurements in situ at in the northern latitudes, and, depending on the
relatively high frequency with an analytical instru- trajectory of an air parcel, it can contain highly
ment maintained at the sampling site. Gas chroma- variable amounts of CH4. Methane mole fractions are
tography with flame ionization detection is the
analytical method most often used for CH4 measure-
- 150 ppb larger at Barrow than at the South Pole,
because -70% of total CH4 emissions are in the
ments. There are advantages and disadvantages to Northern Hemisphere and atmospheric mixing be-
both methods. Measurement in situ can be costly if tween hemispheres (or north-south mixing within a
multiple measurement sites are desired, and requires hemisphere) is not rapid enough to homogenize CH4.
highly skilled personnel to maintain the analytical Measurements at sites from comparable latitudes
instruments and standard gases. While discrete sam- indicate that in addition to N-S gradients, longitudi-
pling can achieve extensive geographical coverage at nal (or E-W) gradients are also observed despite the
relatively low expense and with greater consistency short time scales of zonal mixing. For example, CH4
(since all samples are analyzed on the same analytical values at Mace Head, Ireland (53” N in the eastern
system), it cannot match the sampling frequency of Atlantic Ocean) are about 10 ppb lower than at Cold
measurement in situ. Bay, Alaska (55’ N) and Shemya Island (53”N) in the
The most cost-effective method for determining Pacific Ocean. Qualitatively, this makes sense because
large-scale features in the global CH4 distribution is Cold Bay and Shemya are much closer to strong
discrete sampling. In Figure 2, discrete sample meas- Siberian sources of CH4 than Mace Head. Another
urements from Barrow, Alaska, and the South Pole are important constraint on the global CH4 budget is the
plotted aganist time. Some features in the data are vertical gradient. Some measure of the vertical gradi-
evident. At both sites, methane has been increasing ent is observable in Figure 3, where CH4 measure-
since the measurements started in 1984, and there is a ments are plotted for air samples collected above Carr,
strong seasonality, with smaller values during summer a site in northern Colorado at 1740 m above sea level
than during winter. At Barrow, the seasonal cycle (triangles) and Cape Grim, a site on the north-west
1288 METHANE
I I I I I I I I I I I I I I I I I
1850
1800
1750
(I
(I
v
d
5 1700
1650
1600
84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99
Year
Figure2 Methane molefractions(ppb = nmol mol-') determined from weeklydiscretesamplesplottedassymbolsfor Barrow,Alaska

(A) and South Pole (B). In both cases, smooth curves and deseasonalized trend lines fitted to the data are also plotted.
coast of Tasmania in Australia (circles). At Carr, CH4 Since CH4 at many sites is variable, particularly in
is, on average, greater at the surface, close to sources, the Northern Hemisphere, it is useful to smooth the
than it is aloft. The low values above 6500 m are due to measurements into zonally (i.e., longitudinally) aver-
transport from the stratosphere. Above Cape Grim, aged values. To do this, data from the sampling sites
CH4 values are greater aloft, since there are few (Figure 2 ) are smoothed temporally and as a function
sources of CH4 in the high Southern Hemisphere, of latitude to define a surface of atmospheric methane
and most CH4 found in the Southern Hemisphere has as a function of latitude and time (Figure 4). The
been transported from the Northern in the upper surface is used to calculate zonal averages; examples
troposphere. for the Northern and Southern hemispheres and global
loooo
t
-
-
E
v
6000 1 0 A
A
A
A
-
A
A -
A
A
A -
t
A
A
2000
0
* 0
. A
A
A
A
0
0 1 ~ ~ ~ ' " ~ 1 1 1 ' ~ 1 1 I
' I' I ~ I '1 I I I
METHANE 1289
Figure 4 Smoothed, zonally averaged representationof the global distribution of methane in the marine boundary layer for the period
1989-1999. The grid spacing is 10" latitude by 1 week. Data from 42 sampling sites were used to construct the surface.
averages are plotted in Figure 5A. The global averages constrains the fraction of emissions in the Northern
can be used to calculate the 'burden' (mass) of CH4 in
-
the atmosphere ( 4825 Tg CH4; where 1Tg = 10I2
Hemisphere to 70%. -
Further tests of our understanding of the methane
g). When the burden is combined with an estimate of budget are obtained when measurements are com-
over the late 1 9 9 0 ~total

-
the globally averaged methane atmospheric lifetime
(8.5yr) and rate of increase ( 14 Tg yr- averaged
) ~ emissions are calculated as:
pared with models that include atmospheric transport
and chemistry. For example, the observed seasonal
cycle of CH4 constrains seasonality of loss, emissions,
and transport. Model studies, described in more detail
below, have helped constrain methane's budget, but
future progress is limited by weaknesses in model
transport schemes and limited measurements.
where t is time CH4 atmospheric lifetime (years), Q
emissions (Tg CH4), and (CH4) the atmospheric
Isotope Measurements
methane burden (also in Tg CH4).
Rearranging eqn [l] and solving for Q, the total Each source of atmospheric CH4 has a characteristic
source of CH4 to the atmosphere is -580 Tg CH4 range of values for the ratios of 14Cand 13Crelative to
yr-'. Also plotted in Figure 5A are deseasonalized 12C, and D (Le., 2H or deuterium) relative to H, and
trend curves for each time series. The time derivative of the ratios depend on the mechanistic details of CH4
the trend gives the instantaneous growth rate (shown production and destruction prior to release to the
for Northern and Southern hemispheres only in Figure atmosphere. The ratios are measured relative to a
5B). From these curves, it is clear that the growth rate standard and expressed in parts-per-thousand devia-
of methane has been decreasing over the period of tions from that standard (in units called 'per mil',
these observations, and that significant interannual abbreviated %o). Isotopic measurements of CH4 pro-
variations in growth rate are present. Based on vide a constraint on its budget; the mass-weighted
measurements of CH4 from ice cores and firn discussed isotopic composition of CH4 emissions must equal the
earlier, this decrease in growth rate likely started in the value found in the background atmosphere, after
late 1970s or early 1980s. Some useful constraints on accounting for fractionation during loss processes
the global methane budget are also obtained from the (chemical oxidation and destruction by soil bacteria).
hemispheric averages. The interhemispheric differ- For example, CH4 from biological sources contain
ence, i.e., the average difference between the N H and about 1part in 10I2 14C, while the CH4 in natural gas
SH, is 90 ppb. Assuming an interhemispheric, ex- has no 14C. (Note: 14Cis a radioactive isotope of C that
change time of 1year, a methane lifetime of 8.5 years, has a half-life of 5730 years, therefore the CH4 in fossil
and a rate of increase of 5 ppb yr-', this difference fuels contains no 14C.) To first order, comparing a
1290 METHANE
1800
1750
h
D
-I"
Q
Q
1700
0
1650
1600
84 85 86 87
v '\
\ /
'
88 89 90 91 92 93 94
:! I
I
\
95 96 97 98 99
Year
Figure 5 (A) Methane values determined from the surface plotted in Figure 5 as function of time, where cosine (latitude) was used to
weight for surface area. Plotted are Northern Hemisphere, global, and Southern Hemisphere mole fractions, from top to bottom.
Deseasonalizedtrend lines are also plotted for each. (B) Instantaneous CH4growth rates determined as the derivatives of the trend lines
for the Northern (solid line) and Southern (dashed line) Hemispheres.
measure of 14C in background atmospheric methane to the diversity of sources, CH4 is produced by only a
with I4C in CH4 from biological sources will give an few different processes, which are described below.
indication of the amount of CH4 emitted during fossil
fuel exploitation. Further examples of the use of
Biological Production
isotopes are given later.
Decomposition of organic matter by bacteria under
anaerobic conditions in, for example, wetlands,
flooded soils, sediments of lakes and oceans, sewage,
Sources of Atmospheric Methane and digestive tracts of ruminant animals, involves
There are 10 major known methane sources at the
N complex simultaneous processes that can produce
Earth's surface. Emission rates from many of these are methane as a byproduct. Three steps, each the
small and vary by orders of magnitude over a few responsibility of different types of organisms, are
meters spatial scale, and they are distributed over required: fermentation degrades organic matter into
enormous geographical regions. As a result, total simple fatty acids and C02; other organisms convert
emissions from some sources must be extrapolated the fatty acids into C02 and H,; and methane-
from relatively few direct flux measurements, so generating bacteria (methanogens) metabolize these
uncertainties in these emissions are large. In contrast substrates ( e g , by hydrogen reduction of acetate,
METHANE 1291
formate, and CO2) and CH4 is produced as a bypro- relatively small amounts of CH4 compared with the
duct. In wetland environments, methane that diffuses smoldering phase of forest fires. Since the timing and
through the water column must pass through an the extent of tropical biomass burning is closely
oxic (oxygen-rich) layer at the surface. Most of related to climatic factors (precipitation and temper-
the methane that diffuses to the surface is destroyed ature), these sources vary strongly with season and
by methane-oxidizing bacteria (methanotrophs) contribute to interannual variations in atmospheric
at the water-atmosphere interface. Generally, the methane abundance and its isotopic ratios.
amount of methane released to the atmosphere
depends on the production rate and the efficiency of
Emission Estimates
transport through the oxic layer. In natural wetlands
and rice paddies, methane can also reach the surface Table 1 summarizes the most important known
through bubbles and transport through the stems sources of methane. Recent estimates of their annual
of plants; in contrast to diffusion, these modes of source strengths are given for preindustrial times, the
transport to the atmosphere are relatively efficient. present, and the near future. As can be inferred from
Methanogens are also found in the digestive tracts these estimates, the 2.5-fold increase of atmospheric
of ruminant herbivores. The animals do not have methane since preindustrial times is explained by the
the enzymes necessary to digest cellulose; rather, increases of anthropogenic sources such as fossil
they have a symbiotic relationship with fermenting fuel use, rice cultivation, and cattle production. At
bacteria that produce short-chain fatty acids, vita- present anthropogenic emissions are estimated to
mins, and protein used by the animal for growth. As a contribute -60% of total atmospheric CH4 emis-
result, species such as cows, sheep, buffaloes, and sions. Studies of future emission scenarios suggest
termites are efficient methane producers, and they that this fraction will increase to 80% in 2050. At
contribute -20% of total CH4 emissions to the present, natural wetlands, such as tropical swamps,
atmosphere. temperate and boreal peat land, and tundra still
constitute the largest source class of methane. During
Thermogenic Production preindustrial times, about 75% of atmospheric meth-
ane was emitted from these ecosystems. In Figure 6 ,
At a few kilometers’ depth in soils, pressure and
model calculated distributions of methane are plotted
temperature become great enough for bacteria
for the past (A), present (B), and future (C) based on
to be unable to survive. Beyond this depth, organic
the global budgets in Table 1. In addition to an
matter is decomposed by condensation and cracking
increase of the global mean abundance with time, it
processes, eventually resulting in the formation of can also be seen that regions with high CH4 values shift
coal, oil, and natural gas. Geological processes can
from wetland and biomass burning regions in the
move layers containing natural gas close to the
tropics to the industrialized centers of the Northern
surface and leakage of methane to the atmosphere
Hemisphere.
may occur through cracks, volcanic activity, or seep- Methane sources can be characterized by source-
age from hydrate reservoirs in the ocean sediments.
specific isotopic signatures. Since these characteristics
However, the largest emission of thermogenic meth-
stem from the process of methane formation, they
ane results from human exploitation of fossil fuel
follow roughly the source classification given
resources. At every step along the line of mining, above. In fact, they have been listed from relatively
distribution, and use, small fractions of methane are
depleted to relatively enriched in 13C and deuterium.
released to the atmosphere. Globally integrated,
While methanogenesis results in emissions that
the contribution of these sources is estimated to be
are depleted of these isotopes, methane consumption
15-20% of total emissions.
by methanotrophs results in enrichment of these
isotopes in the methane left behind. As a result,
Pyrogenic Production
the net effect depends on the balance between
The incomplete combustion of organic material yields bacterial production and oxidation of methane.
many compounds, including CO, formaldehyde, ace- Methane derived from biomass burning retains the
tonitrile, and methane. Large amounts of biomass are isotopic characteristics of the fuel, which is slightly
burned in the tropics in mostly human induced fires different for C-3 (6I3C -27%0) and C-4 plants
related to shifting cultivation, deforestation, burning (6I3C z -18%0), but it is highly enriched in 13C
of agricultural wastes, and the use of biofuels. The relative to the isotopic composition of background
fraction of carbon that is released as methane depends atmospheric methane (613C x -47%0).As mentioned
on the fuel type and the burning conditions. For earlier, fossil sources can be distinguished from other
example, the efficient burning of dry savanna releases sources by the absence of the radioisotope I4C.
1292 METHANE
Figure 6 Contours of CH4 mole fraction (ppb) for 1800, 1993, and 2050 calculated from a forward run of a model of atmospheric
chemistry and transport and the emission scenarios in Table 1.
METHANE 1293
Table 1 Estimates of annual CH4 emission rates (in Tg CH4, The methyl radical (CH3) adds 0 2 to form CH3O2.
where 1 Tg = 1O'* g) for years 1800, 1993, and 2050 The subsequent CH4 oxidation reactions affect 0 3 ,
Sourcea Pre-industrial Present Future but the sign of the effect depends on the abundance of
(1800) (1 993) (2050) oxides of nitrogen (NO,). In high-NO, environments
(>30 ppt) (ppt = parts per trillion = 10l2), O3 is pro-
Fossil fuel use (B, T) 90 175
Waste treatment (B) 5 75 120
duced, but with NO,<30 ppt, O3 is normally
Biomass burning (P) 10 35 50 destroyed. The abundance of NO, also affects H O
Domestic ruminants (B) 5 95 225 directly. In NO,-poor environments H O is destroyed,
Rice agriculture (B) 10 80 95 but with sufficient NO, present CH4 oxidation
Natural wetlands (B) 160 145 145 regenerates HO. On average, photochemistry of the
Termites (B) 20 20 20
Oceans (B)
troposphere is limited by the amount of NO,, which
15 15 15
Wild ruminants (B) 15 5 5 means that the oxidation of methane is a net sink of
Wild fires (P) 5 5 5 radicals. As a consequence, an increase of methane is
Other sources (B) 5 25 10 expected to decrease HO. Ultimately, most of the CH4
that reacts with H O produces C02.
Total 250 590 865
Soil oxidation 12 30 45
Oxidation of CH4 also influences the isotopic
Tropospheric oxidation 222 510 750 composition of methane. The rate coefficient for
Stratospheric oxidation 16 40 60 reaction of H O with 12CH4 is -0.5% greater than
that for reaction with 13CH4. This means that CH4
Total 250 580 855
found in the background atmosphere is enriched in 13C
aCodes in parentheses refer to process responsible for CH4 relative to the mass-weighted sum of its sources.
production: B = biogenic, T = thermogenic, P = pyrogenic. In the stratosphere, reactions with chlorine and
oxygen atoms (O'D) become significant sinks of
methane, in addition to HO reaction. In the meso-
sphere, photolysis also contributes to CH4 destruc-
tion, although this process remains negligible relative
Atmospheric Sinks to the total sink. Stratospheric methane is relatively
About 90% of atmospheric methane is removed from enriched in 13C, both because it has been in contact
the atmosphere by reactions initiated by hydroxyl with I2C-depleting H O radicals for a relatively long
radical (HO). The amount of H O in the atmosphere time and because the C1 reaction leads to a larger
depends on the concentrations of O3and H20 (vapor) fractionation of 13C than the H O radical reaction.
and the UV actinic flux as follows: Oxidation of methane in the stratosphere is an
important source of stratospheric water vapor.
O3 + hv(A 5 330nm) -+ 0 2 + O(lD) [I]
Most of the O(lD)is quenched by reaction with N2 or
0 2 to produce ground-state O(3P),but in the tropics Modeling as a Tool to Constrain
(where up to 3 % of air is water vapor) up to 25% the CH4 Budget
reacts as follows: Various types of numerical models have been devel-
oped to improve our understanding of the atmospheric
O(lD) + H20 + 2 H 0 1111 methane budget. As discussed earlier, the global
(Note: the rate coefficient for reaction of O('D) with burden of methane is relatively well constrained by
H20 is 5 to 10 times faster than the quenching measurements. Highly simplified models that repre-
reactions with N2 and 0 2 , so despite the large atmos- sent the atmosphere with one or a few boxes (called
pheric concentrations of N2 and 0 2 , a significant box models) have been used to interpret observations
fraction of O(lD) will result in H O production in the of atmospheric CH4 over large spatial scales, and to
tropics.) The photolysis of O3 to form O(lD)depends calculate global sources and sinks. To interpret real-
on sunlight, so H O production varies diurnally. Much istically the measurements on subhemispheric scales
of tropospheric O3 is produced in situ. The chem- or at specific air sampling locations, models with finer
istry that produces O3 is complex, and the concentra- spatial and temporal resolutions are needed, including
tion of O3 is a balance between production and loss. detailed parameterizations of atmospheric transport
Oxidation of CH4 is initiated by the following and chemistry.
reaction: Box models and three-dimensional (latitude, longi-
tude, and altitude) chemistry transport models typi-
CH4 + H O + H20 + CH3 PI1 cally calculate atmospheric abundances based on a set
1294 MICROBURSTS
of hypothetical emissions that are prescribed as ‘Delta notation’ 6 = (Rsample/Rstandard- 1)x 1000,
boundary conditions to the model. The simulated expressed in per mil (%o), where R is the ratio of
and observed abundances are then compared to rare to abundant isotopes; e.g., to get B13 CH4,
R = 13c 12
test how realistic these boundary conditions were. I c*
It is also possible to compute the sources and sinks Sink a process that removes CH4 from the atmos-
that result in the best possible agreement between phere; e.g., chemical oxidation.
model and measurements. This technique, called Burden total mass of CH4 in the atmosphere, usually
inverse modeling, is an attempt to reconstruct expressed in Tg CH4.
the sources and sinks from atmospheric measure- Methanotroph methane-eating bacterium.
ments. Here it is assumed that differences between Methanogen methane-producing bacterium.
measurements and the ‘forward’ computed abun-
dances can be attributed to sources and sinks. This
requires that inaccuracies in the representation See also
of transport and the errors caused by comparing Biogeochemical Cycles: Carbon Cycle. Chemistry of
measurements at a fixed site with model ‘grid- the Atmosphere: Chemical Kinetics; Gas Phase Reac-
box’ averages (where a grid box may include tions; Ion Chemistry; Principlesof Chemical Change. Data
200 000 km2) are small compared with the uncertain- Analysis: Time Series Analysis. Isotopes, Stable. Nu-
ties in the sources. Further, there are currently too few merical Models: Chemistry Models. Observations for
measurements to properly constrain this type of Chemistry (In Situ): Gas Chromatography. Ozone:
Photochemistry of Ozone; Role in Climate. Paleoclima-
inverse problem; the influence of sources on atmos-
tology: Ice Cores. Tropospheric Chemistry and
pheric abundance is quickly attenuated by atmospher- Composition: Hydroxyl Radical; Oxidizing Capacity.
ic mixing as the distance of the measurements from the
sources increases.
Further Reading
Cicerone RJ and Oremland RS (1988) Biogeochemical
aspects of atmospheric methane. Global Biogeochemical
Glossary Cycles 2: 299-327.
Mole fraction (moles of substance of interest)/(moles Fung I, John J, Mathews E, et al. (1991)Three-demensional
of air) in a sample; e.g., CH4 measurements are model synthesis of the global methane cycle. Journal of
reported in nmol mol - abbreviated ppb (partsper Geophysical Research 96: 13033-13065.
billion by moles). Mole fractions normally reported Khalil MAK (ed.) (1993) Atmospheric Methane: Sources,
for ‘dry air’. Sinks, and Role in Global Change Berlin: Springer.
R M Wakimoto, University of California, Los Angeles, unstable. The leading edge of this outflow is referred to
CA, USA as the gust front.
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. Many convective downdrafts do not produce strong
winds a t the surface; however, in some instances the
downward velocities and subsequent outflow winds
Introduction can result in severe damage. Indeed, these strong winds
can result in considerable crop, tree and structural
The atmosphere undergoes a dramatic overturning damage and have been identified as a causal factor in a
when a convective storm forms. Warm, moist air in the number of aircraft accidents. These intense wind
boundary layer is transported aloft, while relatively events are called downbursts and are defined as an area
cool and dry air is brought down to the lowest levels. of strong winds produced by a downdraft over an area
The primary mechanism that transports the latter air from < 1km to 10 km in horizontal dimensions.
mass to the boundary layer is the downdraft. The Downbursts can be further subdivided into macro-
thermodynamic conditions of the downdraft are bursts and microbursts with the following definitions:
familiar to the general public once it reaches the
surface by the respite from hot, humid conditions that 0 Microburst. Small downburst, less than 4 k m in
are prevalent in the ambient air that is convectively outflow diameter at the ground, with peak winds
1294 MICROBURSTS
of hypothetical emissions that are prescribed as ‘Delta notation’ 6 = (Rsample/Rstandard- 1)x 1000,
boundary conditions to the model. The simulated expressed in per mil (%o), where R is the ratio of
and observed abundances are then compared to rare to abundant isotopes; e.g., to get B13 CH4,
R = 13c 12
test how realistic these boundary conditions were. I c*
It is also possible to compute the sources and sinks Sink a process that removes CH4 from the atmos-
that result in the best possible agreement between phere; e.g., chemical oxidation.
model and measurements. This technique, called Burden total mass of CH4 in the atmosphere, usually
inverse modeling, is an attempt to reconstruct expressed in Tg CH4.
the sources and sinks from atmospheric measure- Methanotroph methane-eating bacterium.
ments. Here it is assumed that differences between Methanogen methane-producing bacterium.
measurements and the ‘forward’ computed abun-
dances can be attributed to sources and sinks. This
requires that inaccuracies in the representation See also
of transport and the errors caused by comparing Biogeochemical Cycles: Carbon Cycle. Chemistry of
measurements at a fixed site with model ‘grid- the Atmosphere: Chemical Kinetics; Gas Phase Reac-
box’ averages (where a grid box may include tions; Ion Chemistry; Principlesof Chemical Change. Data
200 000 km2) are small compared with the uncertain- Analysis: Time Series Analysis. Isotopes, Stable. Nu-
ties in the sources. Further, there are currently too few merical Models: Chemistry Models. Observations for
measurements to properly constrain this type of Chemistry (In Situ): Gas Chromatography. Ozone:
Photochemistry of Ozone; Role in Climate. Paleoclima-
inverse problem; the influence of sources on atmos-
tology: Ice Cores. Tropospheric Chemistry and
pheric abundance is quickly attenuated by atmospher- Composition: Hydroxyl Radical; Oxidizing Capacity.
ic mixing as the distance of the measurements from the
sources increases.
Further Reading
Cicerone RJ and Oremland RS (1988) Biogeochemical
aspects of atmospheric methane. Global Biogeochemical
Glossary Cycles 2: 299-327.
Mole fraction (moles of substance of interest)/(moles Fung I, John J, Mathews E, et al. (1991)Three-demensional
of air) in a sample; e.g., CH4 measurements are model synthesis of the global methane cycle. Journal of
reported in nmol mol - abbreviated ppb (partsper Geophysical Research 96: 13033-13065.
billion by moles). Mole fractions normally reported Khalil MAK (ed.) (1993) Atmospheric Methane: Sources,
for ‘dry air’. Sinks, and Role in Global Change Berlin: Springer.
R M Wakimoto, University of California, Los Angeles, unstable. The leading edge of this outflow is referred to
CA, USA as the gust front.
Copyright 2003 Elsevier Science Ltd. All Rights Reserved. Many convective downdrafts do not produce strong
winds a t the surface; however, in some instances the
downward velocities and subsequent outflow winds
Introduction can result in severe damage. Indeed, these strong winds
can result in considerable crop, tree and structural
The atmosphere undergoes a dramatic overturning damage and have been identified as a causal factor in a
when a convective storm forms. Warm, moist air in the number of aircraft accidents. These intense wind
boundary layer is transported aloft, while relatively events are called downbursts and are defined as an area
cool and dry air is brought down to the lowest levels. of strong winds produced by a downdraft over an area
The primary mechanism that transports the latter air from < 1km to 10 km in horizontal dimensions.
mass to the boundary layer is the downdraft. The Downbursts can be further subdivided into macro-
thermodynamic conditions of the downdraft are bursts and microbursts with the following definitions:
familiar to the general public once it reaches the
surface by the respite from hot, humid conditions that 0 Microburst. Small downburst, less than 4 k m in
are prevalent in the ambient air that is convectively outflow diameter at the ground, with peak winds
MICROBURSTS 1295
lasting only 2 to 5 min. They may induce dangerous equilibrium vapor pressure and, hence, lower relative
tailwind and downflow wind shears which can humidity. Both effects increase the potential for
reduce aircraft performance. evaporation, resulting in latent cooling. Finally, indi-
0 Macroburst. Large downburst, with 4 km or larger vidual parcel vertical excursions tend to be less than
outflow diameter at the ground; damaging wind 4 k m for downdrafts and often greater than 1 0 k m
lasts 5-20 min. An intense macroburst causes tor- (i.e., the depth of the troposphere) for updrafts. The
nado-force damage up to F3 intensity (on the Fujita primary reason for this difference in length scale is that
scale of damage intensity).' the positive buoyancy in the updraft is much greater
than the negative buoyancy in the downdraft.
It is the former phenomenon that has garnered the
The forcing mechanisms of the microburst can be
most interest, owing to its small temporal and spatial
understood by considering the inviscid vertical
scales. Indeed, during the period 1974-1985, micro-
momentum equation
burst winds were a factor in at least 11civil transport
accidents and incidents in the United States, involving
over 400 fatalities and 145 injuries. Extensive research
has been focused on documenting the reasons why this
small subset of downdrafts produces damaging winds.
Fundamentals of the Downdraft (4)

In order to understand the microburst phenomena, it is
instructive to describe the fundamental characteristics where i~ is the mean vertical velocity, p the pressure, 0:
of the downdraft and how it differs from the updraft. A the virtual potential temperature, c p the specific heat
naive view of the downdraft is that it is an upside- at constant pressure, c, the specific heat at constant
down version of the updraft; however, this is far from volume, rc the mixing ratio of cloud water, rr the
an accurate description. While the updraft is typically mixing ratio of rain water, and ri the mixing ratio of ice
slightly supersaturated, the downdraft is often appre- water.
ciably subsaturated, owing to the inability for The primes denote departures from a basic state
condensate cooling by evaporation, melting, or sub- (subscript 0), which varies only in height.
limation to offset completely the warming from Term 1 is the vertical gradient of perturbation
adiabatic compression. For a number of years this pressure. Term 2 represents thermal buoyancy ac-
was, perhaps, one of the greatest misconceptions of the counted for in parcel theory and term 3 is the
thermodynamic characteristics within the downdraft. perturbation pressure buoyancy. Condensate loading
It was often assumed that downdraft air parcels of cloud, rain and ice water are represented by term 4.
descended moist adiabatically (i.e., assumed saturated In addition, entrainment (or mixing) of environmental
descent). The wet-bulb temperature was also believed air with cloudy air or precipitation has been shown to
to be representative of the temperature within a be an important factor in downdraft dynamics.
descending air parcel. It is now known that the The vertical gradient of perturbation pressure is
saturated wet adiabat can only be approached by the generally small for most downdrafts; however, its
downdraft if it is weak (this allows sufficient time for effect becomes significant in intense cumulonimbi and
latent cooling to offset adiabatic compression during mesoscale convective systems associated with strong
descent), the mean drop size is small (see discussion vertical wind shear. Strong downdrafts can develop
below), or the rainfall is heavy. owing to rapid pressure falls generated in response to
It is more important to document the microphysical an intensifying low-level mesocyclone within a super-
details of the liquid or solid condensate within the cell thunderstorm.
downdraft than within the updraft. For example, for a The effect of thermal buoyancy is the best-known
prescribed water content, small raindrops are more term in the vertical momentum equation. Relatively
conducive to producing stronger downdrafts than cold (hot) air promotes negative (positive) vertical
large drops for two reasons: (1)there is greater surface velocities. The importance of using virtual potential
area exposed to the environment; and (2) smaller temperature in the vertical momentum equation has
drops have greater curvature which results in a larger been shown. Increasing the environmental relative
humidity at low levels increases the virtual tempera-
'The F-scale rates the damage intensity of tornadoes and ranges ture difference and will lead to stronger downdrafts
from F1 to F5. A rating of F5 has been associated with the most even if the ambient air temperature remains the same.
intense tornadoes. The pressure buoyancy term suggests that a downdraft
1296 MICROBURSTS
will be produced if an air parcel has a higher accident is shown in Figure 1. When an aircraft flies
perturbation pressure than its surroundings. This through a microburst during takeoff it first encounters
effect is ignored in classical parcel theory, which a headwind component from the microburst outflow.
assumes that the environmental and parcel pressures This headwind increases lift by increasing the relative
are equal. In general, this effect is relatively small in airflow over the wing. The plane may then pitch up,
comparison to thermal buoyancy and the pressure and the pilot may attempt to compensate by leveling
gradient effects. off. But only a matter of seconds later the plane
Some of the early hypotheses on downdrafts sug- encounters a decreasing headwind, downdraft (within
gested that the air was initially dragged downward by the center of the microburst), and then a strong
the weight of precipitation particles (term 4 in eqn [ l]), tailwind. The plane has lost lift and could find itself
then cooled by evaporation. This has been reinforced flying too low and with insufficient air speed to avoid a
by numerous studies that have examined the forcing crash. A similar scenario could be created when an
mechanisms of downdrafts. While precipitation drag aircraft is on approach for a landing at an airport.
can be important in initiating a downdraft, the critical The conceptual model of the descending microburst
role of latent cooling produced via evaporation can be is shown in Figure 2. The microburst is characterized
shown by considering a case where the water mixing by a shaft of strong downward velocity. It is also
ratio is evaporated completely. A water content of associated with strong divergent flow at its center
1g kg - is approximately equivalent to a temperature when it reaches the ground, and is followed by an
deficit of 0.30"C in the buoyancy term in eqn [l].If accelerating outburst of strong winds in an overturn-
this water mixing ratio evaporates then the resulting ing rotor propagating away from the center of the
temperature deficit is given by microburst. The highest wind speeds are usually
associated with these rotors with the peak speeds
occurring in the lower portion of the ring vortex,
where outflow speeds are augmented by the circula-
tion of the ring. The mechanism for intensifying the
where L is the latent heat of evaporation. outflow winds is the stretching of the vortex as the ring
Thus, the temperature deficit has increased by a expands. Horizontal pressure gradients can also con-
factor of 8.3 or nearly an order of magnitude by tribute to high-outflow wind speeds, as discussed
evaporating the water. Accordingly, the evaporation of below. The vertical gradient of perturbation pressure
raindrops rather than water loading would be more (refer to eqn [I])appears to play a relatively minor role
effective in accelerating a downdraft. in most microbursts. The exception appears to be the
The effect of entrainment is complex. At higher microburst associated with the supercell storm.
levels, the entrainment of dry air promotes downdrafts Microburst damage has been frequently documented
by evaporation or sublimation of cloudy air and
precipitation, especially when the entrained region
corresponds to the level of minimum equivalent or
wet-bulb potential temperature. The effect of entrain- Microburst
ment is different at lower levels once a downdraft i '4'' 1
forms. If the downdraft is driven primarily by negative
thermal buoyancy then its intensity will be determined
by the virtual temperature difference between the
descending air parcel and the environment as shown in
J
eqn [l].Mixing of environmental air at this stage will
deplete the negative buoyancy by decreasing this
difference and will lead to reduced downdraft speeds.
The Microburst Figure 1 Schematic diagram illustrating the impact of a micro-

burst on aircraft performance during takeoff. The airplane first
The discovery of the microburst can be traced back to a encounters a headwind and first experiences added lift. This is
series of aircraft accidents and the existence of small- followed in short succession by a decreasing headwind compo-
scale divergent damage patterns in crops and forests nent, a downdraft, and finally a strong tailwind which may lead to an
discovered after the passage of a thunderstorm. Both impact with the ground. Composite drawing based on numerous
studies of aircraft accidents. (See Fujita TT (1985) The Downburst
events were shown to be a result of a transient but - Microburst and Macroburst. SMRP Research Paper No.
intense downdraft accompanied by strong outflow. 210, NTlS PB-148880. Chicago: University of Chicago Printing
The typical scenario for a microburst-related aircraft Department.)
MICROBURSTS 1297
Wet/High reflectivity microburst. A microburst

associated with 0.25mm of rain or a radar echo
> 35 dBZ in intensity.
Low-Reflectivity or Dry Microburst

Figure 2 Conceptual model of a microburst hypothesized to Numerical simulations have shown the sensitivity of
explain ground-damage patterns. Three stages of development downdraft intensity as a function of drop size, rain
are shown. The regions of high wind within the rotor are indicated.
(Adapted from Fujita TT (1985) The Downburst - Microburst and
intensity, and subcloud lapse rate. One of their
Macroburst. SMRP Research Paper No. 210, NTlS PB-148880. conclusions is that when the environmental lapse
Chicago: University of Chicago Printing Department.) rate is approximately equal to the dry-adiabatic lapse
rate then the rates of evaporation place little restriction
on the downdraft magnitude, and even in light
very near tornado damage tracks. Minimum pressures precipitation there may be strong downdrafts gener-
within the low-level mesocyclones are believed to ated. In the absence of pressure effects and in light rain
enhance the downward velocities in these downdrafts. situations (i.e., dry microbursts) it is often convenient
The microburst in three dimensions is shown to view the maintenance of the downdraft as the
schematically in Figure 3. A key feature in the figure competing forces of cooling due to evaporation and
is that a number of microbursts are associated with sublimation versus dry-adiabatic warming due to
small-scale circulations aloft. In fact, it is common for compression. When the subcloud lapse rate is dry-
many microbursts to rotate, with the strongest ones adiabatic, any cooling by condensate results in a
associated with stronger rotation. The magnitude of negative temperature perturbation that will maintain
the vertical vorticity within these circulations can be the downdraft. This effect increases with the depth of
comparable to the vorticity present in the parent dry-adiabatic lapse rate. Compressional warming in a
mesocyclone associated with a tornado. descending parcel can counteract this cooling when
Results from numerous observational and numeri- the subcloud lapse rate is less than dry-adiabatic.
cal studies suggest that microburst winds are associ- Condensate loading plays a relatively minor role
ated with a continuum of rainrates that range from in producing dry microbursts because only light
heavy precipitation from thunderstorms to virga precipitation is present.
shafts from either altocumuli, or clouds that have These results have been confirmed by observations
been referred to as shallow, high-based cumulonimbi. of virga shafts from weakly precipitating cloud
New definitions were created to account for these two systems producing low-reflectivity microbursts. These
extreme microphysical situations: microbursts are particularly hazardous to aircraft
because the parent cloud and pendant virga shafts
Dry/Low-reflectivity microburst. A microburst as- appear innocuous. The weak echoes and low precip-
sociated with <0.25mm of rain or a radar echo itation rates often result in little or no temperature
< 35 dBZ in intensity. change at the surface. In some cases, the air temper-
ature has been shown to rise although the virtual
temperature typically falls. Figure 4 is a plot of
microburst occurrence as a function of radar reflec-
tivity and sub-cloud lapse rate. Each point on the
figure represents a microburst event identified by
Doppler radar. Microbursts can be seen to occur most
frequently at lapse rates of temperature exceeding
- '.
8.5 K km - Practically no microbursts occurred
for lapse rates less than -8.OKkm-l. The axial
profiles for cooling owing to phase change of conden-
sate for the dry microburst is shown in Figure 5. The
dry microburst is associated with small cooling but it
occurs through a deep column. Thus the integrated
negative buoyancy is large. The prevalence of dry
Figure 3 Three-dimensional visulization of a microburst. (Adap-
ted from Fujita TT (1985) The Downburst - Microburst and
microburst over the high plains of the United States is
Macroburst. SMRP Research Paper No. 210, NTlS PB-148880. attributed largely to the deep, dry-adiabatic layer
Chicago: University of Chicago Printing Department.) characteristic of this geographic region.
1298 MICROBURSTS
draft speeds nearly twice as strong as those generated

by hail. The sublimation process associated with
snowflakes is important for three apparent reasons:
( 1) the numerous low-density snow particles readily
sublimate, with much of the snow content depleted
before melting into rain; (2) the latent heat of
sublimation is greater than the latent heat of either
evaporation or melting; and ( 3 ) the cooling from
sublimation takes place at a relatively high altitude
within the deep adiabatic layer, allowing the down-
draft to accelerate through a deep column.
High-Reflectivity or Wet Microburst

In more stable lapse rates, the downdraft tends to be
Figure 4 Plot of microburst occurrence as a function of radar warmer than the environment. In such a case, the drag
reflectivity and environmental lapse rate. (Adapted with permission
from the American Meteorological Society from Srivastava RC
of the condensed water may become important in
(1985) A simple model of evaporatively driven downdraft: Applica- accelerating the downdraft, especially at high levels.
tion to microburst downdraft. Journal of the Atmospheric Sciences Accordingly, higher rainwater mixing ratios (i.e.,
42: 1004-1 023.) radar reflectivities) are required for microbursts to
form. This effect is well illustrated inFigure 4.The few
The sensitivity of the dry microburst to the micro-

physics of the condensate has also been shown.
-
microbursts that occurred for lapse rates less than
8.0 K km were associated with radar reflectivities
in excess of 45dBZ, suggesting that higher water
Numerical studies initiated with different precipita- contents are needed to produce strong downdrafts.
tion particles reveal that snowflakes produce down- This often manifests itself in radar reflectivity images
as a prominent descending precipitation core. It has
been shown that melting also plays an important role
in producing a wet microburst. Narrow shafts of hail,
coincident with the microburst downdraft, embedded
within a heavy shower of large raindrops have been
noted.
The comparison of wet and dry microburst simula-
tions is shown in Figure 5. The dry and wet micro-
bursts were driven by snow and hail, respectively.
Note that the temperature deviations for the two cases
are different. The dry microburst is associated with
smaller cooling, but it occurs through a deep column.
The temperature deviation for the wet microburst is
largest at the ground but diminishes rapidly with
height until it becomes warmer than the ambient air
above 1km. The warming aloft for the hail case
indicates the importance of precipitation loading in
the early stages of the wet microburst which over-
comes the positive buoyancy. It is hypothesized that
hail produces stronger microbursts in more stable
environments, since the downdraft is maintained only
at lower elevations where it is less likely to be depleted
Figure 5 Comparison of the axial profiles for cooling owing to of negative buoyancy because of compressional
phase changes of condensate for the dry microburst with snow and heating. Note that strong negative buoyancy is con-
the wet microburst with hail. Location of the melting level based on fined to the lowest levels of the wet microburst in
the environmental sounding is shown. (Adapted with permission Figure 5. The dominance of negative thermal
from the American Meteorological Society from Proctor FH
(1989) Numerical simulations of an isolated microburst. Part 11:
buoyancy near the ground combined with the
Sensitivity experiments. Journal of the Atmospheric Sciences 46: decreasing effect of precipitation loading can result
2143-21 65.) in a displacement between the location of the strongest
MICROBURSTS 1299
downdraft velocities and the maximum in radar winds are likely to form. The two types of profiles
reflectivity. that have been associated with strong outflow are
An important observation was revealed in the the 'inverted V' profile conducive to dry or low-
simulations shown in Figure 5. Although the down- reflectivity microbursts and the weakly capped wet or
draft associated with the dry microburst was much high-reflectivity microburst. The former is shown in
deeper and almost twice as intense as that of the wet Figure 6 . Its main characteristics are the deep, dry-
microburst, both produced identical outflow speeds. adiabatic, subcloud layer from near the surface to the
This points to the immense difficulty in estimating midlevels, a dry lower layer, and a moist midtropo-
peak outflow speeds based on the expected downdraft spheric layer. The convective instability is small
velocities. The cold air for the wet microburst is or marginal, therefore the convection is often
situated primarily at low levels in Figure 5 . Although weak. The latter characteristic is consistent with
this cooling near the ground is not able to translate into numerous observations of dry microbursts associ-
strong enhancement of the downdraft, it may strength- ated with virga shafts pendant from innocuous
en the outflow speeds through enhanced horizontal clouds. Forecasting schemes are usually based on the
pressure gradient forces produced by the formation of 1200 UTC sounding and expectations for solar
a mesohigh. Dry microbursts are typically accompa- heating and maximum surface temperatures later in
nied by small, cold pools (Le., weak mesohighs). Thus, the day.
strong outflow winds for the dry microburst are Wet microbursts profiles typically display high
possible only when the downdraft speeds are also moisture values through a deep, surface-based layer,
intense. This comparison of these two cases empha- with the top of the moist layer sometimes extending
sizes the nonlinear relationship between vertical beyond 4-5 km AGL. Relative humidities above
velocities and outflow speeds. the moist layer are low. The dry-adiabatic subcloud
layer may only be 1.5km deep and the capping
inversion is weak. It has been suggested that the
Forecasting and Detection vertical profile of equivalent potential temperature
Although intense microbursts are often associated can be useful in identifying environmental conditions
with the supercell storm, most microbursts develop capable of supporting wet microbursts (Figure 7).
within environments characterized by weak vertical The difference in the equivalent potential tempera-
wind shear. The thermodynamic profile determined by ture between the surface and midlevels equal to or
nearby upper-air soundings are particularly useful greater than 20°C appears to be a characteristic
for identifying when strong, convectively induced profile during wet microburst events. In addition
Figure 6 Schematic of the characteristics of the thermodynamic profile of the morning and evening soundings favorable for dry
microburst activity over the high plains. (Reproduced with permission from the American Meteorological Society from Wakimoto RM
(1985) Forecasting dry microburst activity over the high plains. Monthly Weather Review 113: 1131-1 143.)
1300 MICROBURSTS
Figure 7 Thermodynamic model summarizing the environment conducive for wet-microburst occurrence. (Reproduced with
permission from the American Meteorological Society from Atkins NT and Wakimoto RM (1991) Wet microburst activity over the
southeastern United States. Weather and forecasting 6:470-4132,)
to the morning sounding, predicted moisture See also

and thermal advection patterns must be taken into
Convective Storms: Convective Initiation. Gust Fronts.
consideration. MesoscaleMeteorology:MesoscaleConvectiveSystems.
Microburst detection in real time has focused on the Tornados.
use of single-Doppler radar. A number of radar signa-
tures have been identified that occur typically 2-6 min
prior to the initial surface outflow. Descending reflec-
tivity cores, increasing radial convergence within the Further Reading
cloud (in response to an accelerating downdraft), Fujita TT (1981) Tornadoes and downbursts in the context
intense small-scale circulations, and weak-echo reflec- of generalized planetary scales. Journal of the Atmos-
tivity notches (in response to entrainment of low- pheric Sciences 38: 1511-1534.
equivalent-potential-temperature air) have all been Fujita TT (1985) The Downburst - Microburst and
found to be important microburst precursors. Macroburst. SMRP Research Paper No. 210, NTIS
Basic research on microbursts has been quickly PB-148880. Chicago: University of Chicago Printing
Department.
and successfully transferred into the operational Proctor FH (1989) Numerical simulations of an isolated
community. Many of the airports around the United microburst. Part 11: Sensitivity experiments. Journal of
States have now implemented systems that provide the Atmospheric Sciences 46: 2143-2165.
timely warning to air traffic controllers and pilots Srivastava RC (1985) A simple model of evaporatively
of impending wind shear events associated with driven downdraft: application to microburst downdraft.
microbursts. Journal of Atmospheric Sciences 42: 1004-1023.
MICROCLIMATE 1301
M W Rotach, Swiss Federal Institute of Technology, Variables Describing and Determining

Zurich, Switzerland
Microclimate
P Calanca, Swiss Federal Research Station for
Agroecology and Agriculture, Zurich, Switzerland As usual in climatology, the microclimate of a partic-
ular location is characterized through a number of
climate variables. Traditionally, these variables are
those describing the thermodynamic and dynamic
state of the atmosphere, i.e. radiation, temperature,
Introduction humidity, wind speed, and pressure (density).Depend-
ing on the research focus, other variables, such as trace
On the global scale the climate is mainly determined gas concentrations, should be added to the list. For
by the incoming solar radiation. The word climate example, the health and the comfort of the ever
itself stems from the Greek ~ i i y a ,which means increasing number of people living in cities is directly
inclination and is appropriate for describing the effects related to the concentration and distribution of air
of latitude on the availability and distribution of solar pollutants, which are additionally required to charac-
energy. On very small scales, however, global factors terize the microclimatic state of this particular
set only the boundary conditions, while the actual environment.
climate at any particular location is defined by the As for the surface properties determining the
local surface properties. An outstanding example is microclimate, the following groups of variables are
that of an oasis surrounded by a desert. Obviously, it is relevant in the case of simple, flat surfaces: radiative
not the Earth’s orbital parameters nor the composition properties (albedo, emissivity); aerodynamic charac-
of the atmosphere that determine the climatic state teristics (roughness length, zero plane displacement);
within the oasis but rather the local soil moisture thermal properties (heat capacity and conductivity),
conditions. and properties affecting the moisture status (hydraulic
The microclimate of a particular location can characteristics of the soil, type of surface cover). At
hence be defined as the statistical state of the sites with a more complex geometry, slope and
atmosphere in the layer being affected directly by exposition have an important impact, in particular
the characteristics of the underlying surface. In on the components of the radiation balance. These
this sense, microclimatology is the study of the surface properties are not static, but reflect for instance
long-term average and typical variability of climate forcing by the large-scale wind field, diurnal changes in
variables in this lowest layer of the atmosphere, the soil moisture conditions, seasonal changes in the
and physical microclimatology can be thought of as vegetation, or the presence or not of snow.
the study of the processes by which the lowest layer The temporal and spatial distributions of the state
of the atmosphere responds to surface boundary variables are governed by the conservation equations
conditions. for energy, momentum, and mass (dry air, water vapor,
The processes controlling the microclimate at a tract gases). Owing to the generally turbulent nature of
particular location involve time scales from a few the flow in the vicinity of the Earth’s surface, the terms
seconds to several years. For example, evapotrans- representing turbulent transport in these equations are
piration from a vegetated surface is regulated on the particularly important and receive much attention in
short term by the physiological processes of the research of microclimate.
plants, whereas decadal changes in the species
composition of that surface determine its long-term
trends.
Like the temporal scales, the spatial scales usually
The Lowest Layers of the Atmosphere
associated with the term microclimate encompass The atmospheric layer affected by the local surface
several orders of magnitude. The microclimate of properties is called the planetary boundary layer. It has
animals or single plants deals with scales ranging a characteristic depth of the order of 1000 m and can
from centimeters to meters. On the other hand, the be divided into the upper or outer layer (the uppermost
characteristic scales associated with the microclimate 90%) and the inner or surface layer (that is, the lowest
of an extended downtown area with high-rise build- lo%), respectively. However, a direct influence of the
ings are on the order of hundreds of meters, or surface characteristics on the atmospheric state is
kilometers. observed only in the lowest part of the surface layer, in
1302 MICROCLIMATE
average height of roughness elements. In this layer,

form (or pressure) drag and viscous drag on any
individual roughness element are both significant and
lead to a retardation of the flow. In addition, the
material and orientation of the obstacles give rise to
large variation in the energy balance and primarily the
radiation balance. Moreover, within the canopy layer
there is a substantial variability in the sourceknk
distribution of sensible heat, water vapor, or trace
gases.
Figure 2 shows the height ranges covered by the
roughness sublayer and its lower part, the canopy
layer, in a dimensionless form. Arrows indicate the
Figure 1 Conceptual sketch and terminology for the lowest typical situation for a surface covered with crop, trees,
layersof theatmosphereovera rough surface. Note the logarithmic
height scale. The level 4 refers to the boundary layer height and h
and houses, respectively. We observe that the vertical
denotes the (average) height of roughness elements. range, to which the study of microclimate is confined,
can extend up to several tens of meters or more in the
the immediate vicinity of the roughness elements. If case of an urban surface (if we use typical values for h ) .
this layer of influence has any discernible thickness In the case of a shallow boundary layer (small
then it is because of a nonnegligible vertical extension boundary layer height z i and large roughness ele-
of the so-called roughness elements (stones, vegeta- ments), no inertial sublayer may be present at all.
tion, trees, buildings). This layer is therefore usually So far we have dealt only with rough but horizon-
called the roughness sublayer, and the upper part of the tally homogenous surfaces like an extended cornfield,
surface layer is then referred to as the inertial sublayer. a large forest, or a uniform city. In the case of small-
Note that over a relatively smooth surface such as scale variability of surface properties, a so-called
short grass or sand the roughness sublayer becomes internal boundary layer forms downwind of each
very thin, and the inertial sublayer is often associated major change in the surface characteristics. This
with the entire surface layer (Figure 1).
On the basis of these considerations, microclima-
tology can be thought of as the study of the climatic
100
state of the roughness sublayer or specific entities
therein, whereas ‘micrometeorology’ is more properly Z/Zi Crop
related to the study of the dynamics and thermody-
namics of the surface layer or even the entire planetary
boundary layer.
lo-’ .\
..
--
‘t.
..
I I
I
..
..
The Roughness Sublayer 10-2 - ... ...
.. .. . . . . . . . . . . . . . Y
In the roughness sublayer, the flow is affected by the .. .. . . . . . . . . . .Hougnness
individual roughness elements, and is hence fully .......... ul
10-3 - :. .: . . . . . . . . . . . . . . . . . . . .>,
-tahl-t,-v.
three-dimensional in nature. The upper boundary of ....

the roughness sublayer, z*, is the level at which the ............................
.
horizontal variability vanishes and the flow becomes ............................. .‘
. . . . . . . . . .I.. . . . . . . . . .I.. . . . . . .
-
10-~
horizontally homogeneous. Properly scaled profiles of 10’ 102 1 03 104
either mean flow characteristics or turbulence statis-
Zi/h
tics will then merge to one curve that is characteristic
BL height/Roughness element height
for the underlying surface. The depth of the roughness
sublayer depends on the height and distribution of the Figure 2 Sketch of the vertical extension of the various layers
roughness elements. For most surfaces 2h <z* < 5h over rough surfaces and their variation with the nondimensional
quantities z/q and q/h. The height 4 refers to the boundary layer
covers the range of estimates, where h is the average
height and h denotes the (average) height of roughness elements.
height of roughness elements. A value of z*/h = 3 is assumed for the height of the roughness
sublayer. The arrows ‘city’, ‘forest’, and ‘crop’ are based on typical
The Canopy Layer valuesforthe height of theroughnesselements handthe boundary
layer height 4. (Reprinted from Rotach MW (1999)On the urban
The lowest part of the roughness sublayer is the so- roughness sublayer. Afmospheric Environment 33: 4001-4008,
called canopy layer. It ranges from the surface up to the with permission from Elsevier Science.)
MICROCLIMATE 1303
internal boundary layer may be thermal (thermal The turbulent fluxes of sensible and latent heat arise
internal boundary layer) if there is primarily a change from the fact that the atmospheric flow in the vicinity
in the thermal properties of the surface, as for example of the surface is generally turbulent. They constitute
on a shoreline. Alternatively, if mainly the roughness the major mechanism of nonradiative vertical energy
changes then the internal boundary layer is called transport in the lowest p o m n of the atmosphere,
mechanical. Most common, of course is a combined and are defined as H = pcpw’O’and L,E = pL,ur’q’,
thermal and mechanical internal boundary layer. The respectively, where p is the air density, cp the specific
concept of a blending height has been introduced to heat at constant pressure, 0 the potential temperature,
denote the level, above which turbulent mixing has L, the latent heat of vaporization, and q the specific
levelled out horizontal gradients of climate variables. humidity. The primed variables refer to the turbulent
In the strictest sense, the study of microclimate is thenfluctuations with respect to a temporal average (rep-
confined to the airspace below the blending height. resented by an overbar).
The net radiation exhibits a strong seasonal and
diurnal cycle; typically, it assumes its maximum
The Surface Energy Balance around noon and is slightly negative during the night.
The way the surplus or deficit of radiative energy is In complex terrain, owing to orographic effects, the
redistributed or compensated for at the Earth’s surface amplitude and phase of the net radiation’s daily cycle
is described through the energy balance equation, can vary substantially over short distances, as can be
which in a simplified form reads seen from Figure 3. In general, net radiation is
dependent on the latitude (solar radiation), the surface
N R = H + L,E G + A S + 111 properties (reflectivity or albedo and emissivity), the
atmospheric conditions (clouds) as well as the topo-
Here, NR denotes the net radiation, H a n d L,E are the graphy and surface cover of the surroundings (shading
turbulent fluxes of sensible and latent heat, respec- effects, longwave emission).
tively, and G is the ground heat flux. If the energy The partitioning of radiative energy surplus or
balance is set up for a volume instead of an infinites- deficit into the other terms of eqn [l]again depends on
imally thin surface such as the entire canopy layer, a the properties of the underlying layer (vegetation, bare
storage term AS needs to be included in eqn [l].This soil, rock, water, snow cover). In particular, the
latter term is the result of discernible divergence of availability of moisture at the surface or in the layer
the other energy fluxes involved and leads to warm- considered determines which of the latent heat flux or
ing or cooling (drying or moistening) of the volume sensible turbulent heat transport is dominant. This is
considered. most easily seen by considering the ratio between the
700
600
500
-I
400
E
3
Y
300
cc
2 200
100
-1 00
0 4 8 12 16 20 24
Time (LST)
Figure 3 Daily cycles of net radiation at four sites in an approximately north-south-oriented alpine valley in southern Switzerland.
Shown are average data over 16 fine days with little or no cloud in late summer-fall 1999. Open triangles: vineyard site near the bottom of
the east-facing slope. Full dots: valley floor. Squares: meadow on the west-facing slope. Full triangles: alpine rubble site near the east-
facing ridge top.
1304 MICROCLIMATE
turbulent energy fluxes, the so-called Bowen ratio, layer can be solved analytically to yield an exponential
expressed by p = H/L,E. wind profile:
Microclimate of a Vegetated Surface

2 ( z ) = u ( h )exp -a
[ (1 - 31 - 131
where a is a parameter dependent on the density and

Mean Vertical Structure
character of the roughness elements and found to be
In the roughness sublayer, the state variables and the in the range of 1 to 4 by fitting observed profiles to
fluxes of energy, momentum, and mass all show a eqn ~31.
pronounced vertical variability. At a sufficient distance Roughness sublayer profiles of the other climate
from the surface and under neutral conditions the variables are also presented in Figure 4. The vertical
profile of mean wind speed can be shown to follow a distribution of net radiation depends on the albedo of
logarithmic law: the canopy, as well as on the extinction of solar
radiation (during daytime) and the absorption and
emission of thermal radiation at different levels within
the canopy. It is related to the distribution of biomass
as given by the so-called leaf-area index. Owing to the
where uI is the friction velocity as determined by the greater absorptance in the visible than in the near-
Reynolds stress at the canopy top, k = 0.4 the von infrared range, there is a change in the spectral
Karman constant, zo the roughness length, and d the composition of the solar radiation flux from the top
zero-plane displacement. A stability dependent mod- to the bottom of a stand.
ification may be added to eqn [2] for nonneutral The profiles of temperature and humidity and
conditions. Both the zeroplane displacement and the similarly COZ (not shown) are complex. In forest
roughness length are characteristic for the type and stands they typically show an inflection point within
status of the underlying surface. They can be deter- the canopy. These profiles can be explained by the
mined from fitting observed wind speed profiles to eqn divergence of the relevant energy and mass fluxes and
[2] or, alternatively, estimated from morphometric by the distribution of the sources and sinks (including
properties of the surface. As a rule of thumb, d = 0.7 h those at ground level), but do also reflect advective
and zo = 0.1 h can be used. Equation [2] is, strictly effects. All profiles exhibit strong diurnal variations.
speaking, valid only in the inertial sublayer above the Turbulence statistics such as velocity variances or
roughness sublayer. Within the roughness sublayer the turbulent fluxes of sensible heat and momentum
mean wind speed departs from the logarithmic exhibit a strong reduction from the canopy top down
behavior, with a strong retardation at the mean height to the ground where very small or vanishing values are
of the roughness elements (Figure 4). observed. The most striking feature of canopy turbu-
This retardation is due to the form and viscous lence is probably the fact that it is governed by so-
drags. Under certain simplifying assumptions the called coherent structures with spatial scales on the
momentum balance equation for the roughness sub- order of the canopy height. As a consequence, the
r
I
-
I.
0.0 0.5 1.0 1.5 2.0 -2 -1 0 1 -0.0015 0.0000 0.0015 0 100 200 300 400
u/uh @-Oh (K) q - q h (kgkg-') NR (W m-2)
Figure 4 Typical profiles of mean wind speed u , daytime potential temperature 0, specific humidify q, and net radiation NR within and
abovevegetated canopies. The lower portion of the specific humidity profile is typical for moist surfaces only. Forwind speed the dashed line
corresponds to eqn [3] and the full line sketches a typical profile close to and within the roughness elements. Note that the profile for wind
speed qualitatively applies to both vegetated and urban surfaces.
MICROCLIMATE 1305
turbulent exchange of heat, moisture, or trace gases spatial variation in the thermal conditions can be
within the canopy is often characterized by counter- found down to the level of the smallest entities, like the
gradient transport. This means that the so-called walls of an urban street canyon. Further, the scale of an
K-theory, which is based on the assumption of individual roughness element is large enough for it to
small dominating eddies and is a well-established be accessible for detailed experimental investigation.
concept for the inertial sublayer, is not a useful The best-investigated entity of an urban surface is
description of turbulent transport in the roughness therefore the (generic) urban street canyon and its
sublayer. immediate surroundings.
Owing to the rough nature of the surface, turbulent The best-known characteristic of urban climate is
mixing is stronger in the upper part of the roughness the so-called urban heat island. This refers to a
sublayer than in the inertial sublayer. As for the fluxes pronounced increase in near-surface temperature in a
of sensible heat and specific humidity (latent heat), this city as compared with its rural surroundings. Differ-
increase in most pronounced under near-neutral con- ences are found to be strongest (several degrees
ditions. In this case the turbulent transport just above Celsius) some hours after sunset and are attributed
the canopy can be up to three times larger than in the to, among others, the specific radiation conditions
inertial sublayer, whereas for momentum the enhance- within block-like structures (street canyons), the
ment is of the order of 10%. differences in the turbulent diffusion of heat over
urban and rural surfaces, respectively, and (at least in
Water and Carbon Budgets winter) anthropogenic heat release.
There is an evident feedback between the microclimate Effects of Building Geometry on Radiation,
of a vegetated surface and vegetation itself. On the one Temperature, and Flow Characteristics
hand, growth and composition of the canopy reflect The daily cycle for the components of the radiation
the boundary conditions set by the microclimate. On balance differs for the various active surfaces within a
the other hand, vegetation exerts an active control on street canyon (sunlit and shadowed walls, street
the microclimatic conditions in its immediate sur- canyon floor). Long-wave incoming radiation is dom-
rounding, as seen when considering the water and inated by the emission from nearby walls, while short-
carbon budgets. Precipitation, interception, evapo- wave (solar) incoming radiation is highly dependent
transpiration, and surface runoff, as well as infiltra- on the orientation of the surface and may be several
tion, drainage, and capillary rise, determine the water times reflected on the various surfaces. Thereby, all
storage in the soil. Transpiration, however, is a radiation components can be shown to be related to
function of the water content and water potential the so-called sky-view factor (i.e., the ratio of the sky
not only in the soil but also in the roots, the stand and ‘seen’ at a particular point within the urban canopy to
the leaves, and depends therefore on the particular that potentially available) and to the aspect ratio
state of the plants. Furthermore, photosynthesis and (width/height)of the canyon.
photorespiration not only are regulated by physiolog- Within a street canyon the temperature may be
ical processes at the level of leaves and cells, but also slightly higher than aloft, owing to the trapping of
are controlled by the overall radiation conditions and solar radiation, but little spatial variation is observed
CO2 concentration within the canopy. except very close to the walls.
In the case of agricultural surfaces, human activities Profiles of mean wind speed over urban surfaces are
must be considered an important factor determining usually described using the same concepts as for
the microclimate of the stand. In fact, management vegetated canopies (eqn [2]).As for the latter, the two
options are explicitly chosen to provide the best roughness parameters d and zo are properties of the
growing and reproductive conditions for each partic- inertial sublayer and eqn [ 2 ]and, strictly speaking, are
ular type of crop. valid only there.
The transition from the inertial to the roughness
sublayer, and the actual profile within the roughness
Microclimate of an Urban Surface sublayer, are strongly dependent on the orientation of
The microclimate of an urban surface differs from that

Encyclopedia of Atmospheric Sciences - 1st Edition (Volume 1-6) (2003) PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Encyclopedia of Atmospheric Sciences - 1st Edition (Volume 1-6) (2003) PDF

Uploaded by

Copyright:

Available Formats

ENCYCLOPEDIA

Copyright 2003 by Elsevier Science Ltd. All Rights Reserved

BOUNDARY LAYERS: Overview

The following articles are US government works in the

DYNAMIC METEOROLOGY: Balanced Flows; Potential Vorticity

The following article has a CSIRO author

N o part of this publication may be reproduced or transmitted in any form or by any

Library of Congress Control Number: 20021 14120

William Blumen' Stanley Q. Kidder

Stefan L. Hastenrath Mankin Mak

Julian C. R. Hunt, FRS Mitchell W. Moncrieff

Richard H. Johnson Roger A. Pielke, Sr

Michael Kelley William J. Randel

A. R. Ravishankara John Thuburn

Hanwant B. Singh Peter J. Webster

Peter A. Taylor Morris Weisman

James R Holton, Judith A Curry, and John A Pyle

Structure of the Encyclopedia

ii. To draw the reader’s attention to parallel discussions in other articles.

iii. To indicate material that broadens the discussion.

The appendices are located in volume 6, before the index.

Cortinas Jr, J V Donaldson, D J

Croft, P J Doswell 111, C A

Gayet, J-F Grant, W B

Koch, O W Lau, K-M

Complex Terrain JJ Finnigan 234

DATA ANALYSIS See also WEATHER PREDICTION: Data Assimilation

EL NIRO AND THE SOUTHERN OSCILLATION

FLOODING C A Doswell I I I 769

GAIA HYPOTHESIS T Lenton 815

GRAVITY WAVES See BUOYANCY AND BUOYANCY WAVES: Optical Observations;

HADLEY CIRCULATION IN James 919

KATABATIC WINDS T R Parish 1057

LABORATORY GEOPHYSICAL FLUID DYNAMICS R L Pfeffer 1081

SATELLITE REMOTE SENSING

SOLITARY WAVES ] P Boyd 2086

Inter Tropical Convergence Zones D E Waliser 2325

ULTRAVIOLET RADIATION K Stamnes 2467

VALLEY WINDS J Egger 2481

WEATHER FORECASTING See SYNOPTIC METEOROLOGY: Forecasting SYNOPTIC

‘phase angle’ and gives the phase of the wave at x = 0

For a symmetrical source, such as a small explosion

10-0 point source in an atmosphere with a constant

Diffraction of Acoustic Waves wave equation for a 500Hz point source in an

Figure 14 Geometry for a monostatic sodar. An array of

‘phase angle’ and gives the phase of the wave at x = 0

For a symmetrical source, such as a small explosion

10-0 point source in an atmosphere with a constant

Diffraction of Acoustic Waves wave equation for a 500Hz point source in an

Figure 14 Geometry for a monostatic sodar. An array of

Climatology of Tropospheric Aerosols

plant debris. Most particles include some water,

afford the most extensive geographic mapping of those

Further Reading Hobbs PV (ed.) (1993) Aerosol-Cloud-Climate Interac-

Observations and Measurements

-+- ............. Sample Collection and Analysis:

Interception work is underway to extend this methodology to

H+, NH,: so:-, NO;

NaCl HNO,, SO2

polyurethane foam is used to collect certain organic Aerosol inlet

Impacts on Ozone and Methane HO2 + NO2 + M + HO2N02 + M [VII]

summarized in the next two sections along with

100 101 102