Remote Sensing of Geomorphology,

Volume 23 1st Edition - eBook PDF

Go to download the full and correct content document:
https://ebooksecure.com/download/remote-sensing-of-geomorphology-volume-23-eb
ook-pdf/
More products digital (pdf, epub, mobi) instant
download maybe you interests ...

Atmospheric Remote Sensing 1st Edition - eBook PDF

https://ebooksecure.com/download/atmospheric-remote-sensing-
ebook-pdf/

Remote Sensing of Ocean and Coastal Environments 1st

Edition - eBook PDF

https://ebooksecure.com/download/remote-sensing-of-ocean-and-
coastal-environments-ebook-pdf/

Hyperspectral Remote Sensing: Theory and Applications

1st edition - eBook PDF

https://ebooksecure.com/download/hyperspectral-remote-sensing-
theory-and-applications-ebook-pdf/

Atmospheric Remote Sensing: Principles and Applications

1st edition - eBook PDF

https://ebooksecure.com/download/atmospheric-remote-sensing-
principles-and-applications-ebook-pdf/
Radar Remote Sensing: Applications and Challenges 1st
Edition- eBook PDF

https://ebooksecure.com/download/radar-remote-sensing-
applications-and-challenges-ebook-pdf/

(eBook PDF) Comprehensive Remote Sensing by Shunlin

Liang

http://ebooksecure.com/product/ebook-pdf-comprehensive-remote-
sensing-by-shunlin-liang/

Bio-optical Modeling and Remote Sensing of Inland

Waters 1st Edition Edition D.R. Mishra - eBook PDF

https://ebooksecure.com/download/bio-optical-modeling-and-remote-
sensing-of-inland-waters-ebook-pdf/

Nanotechnology-Based Smart Remote Sensing Networks for

Disaster Prevention 1st Edition - eBook PDF

https://ebooksecure.com/download/nanotechnology-based-smart-
remote-sensing-networks-for-disaster-prevention-ebook-pdf/

Precipitation Science: Measurement, Remote Sensing,

Microphysics and Modeling 1st Edition Silas Michaelides
(Editor) - eBook PDF

https://ebooksecure.com/download/precipitation-science-
measurement-remote-sensing-microphysics-and-modeling-ebook-pdf/
 Developments in Earth Surface Processes
REMOTE SENSING OF GEOMORPHOLOGY
VOLUME 23
Developments in Earth Surface Processes, 23
Series Editor – J.F. Shroder, Jr.

For previous volumes refer http://www.sciencedirect.com/science/bookseries/09282025

Developments in Earth Surface
Processes
REMOTE SENSING OF
GEOMORPHOLOGY

VOLUME 23
Volume Editors

PAOLO TAROLLI
Department of Land, Environment, Agriculture and Forestry,
University of Padova, Legnaro (PD), Italy

SIMON M. MUDD
University of Edinburgh, School of GeoSciences, Edinburgh, United Kingdom
Elsevier
Radarweg 29, PO Box 211, 1000 AE Amsterdam, Netherlands
The Boulevard, Langford Lane, Kidlington, Oxford, OX5 1GB, United Kingdom
50 Hampshire Street, 5th Floor, Cambridge, MA 02139, United States
First edition 2020
© 2020 Elsevier B.V. All rights reserved.
No 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.
Details on how to seek permission, further information about the Publisher’s permissions policies and our arrangements with
organizations such as the Copyright Clearance Center and the Copyright Licensing Agency, can be found at our website: www.
elsevier.com/permissions.
This book and the individual contributions contained in it are protected under copyright by the Publisher (other than as may be
noted herein).
Notices
Knowledge and best practice in this field are constantly changing. As new research and experience broaden our understanding,
changes in research methods, professional practices, or medical treatment may become necessary.
Practitioners and researchers must always rely on their own experience and knowledge in evaluating and using any information,
methods, compounds, or experiments described herein. In using such information or methods they should be mindful of their
own safety and the safety of others, including parties for whom they have a professional responsibility.
To the fullest extent of the law, neither the Publisher nor the authors, contributors, or editors, assume any liability for any injury
and/or damage to persons or property as a matter of products liability, negligence or otherwise, or from any use or operation of
any methods, products, instructions, or ideas contained in the material herein.
Library of Congress Cataloging-in-Publication Data
A catalog record for this book is available from the Library of Congress
British Library Cataloguing-in-Publication Data
A catalogue record for this book is available from the British Library
ISBN: 978-0-444-64177-9
ISSN: 0928-2025

For information on all Elsevier publications

visit our website at https://www.elsevier.com/books-and-journals

Publisher: Candice Janco

Acquisitions Editor: Amy Shapiro
Editorial Project Manager: Kelsey Connors
Production Project Manager: Bharatwaj Varatharajan
Cover Designer: Greg Harris
Typeset by SPi Global, India
 Contents

Contributors ix 3. Ground-based remote sensing

Foreword xi of the shallow subsurface: Geophysical
Introduction to remote sensing methods for environmental applications
Giorgio Cassiani, Jacopo Boaga, Ilaria Barone, Maria Teresa Perri,
of geomorphology xiii Gian Piero Deidda, Giulio Vignoli, Claudio Strobbia, Laura Busato,
Rita Deiana, Matteo Rossi, Maria Clementina Caputo, and
Lorenzo De Carlo
1. Structure from motion
1 Introduction 56
photogrammetric technique
2 Methods 56
Anette Eltner and Giulia Sofia
3 Application examples 67
4 Future challenges and conclusions 80
1 Introduction 1
Acknowledgments 83
2 Method 5
References 83
3 Reconstructing processes across
Further reading 89
space 12
4 Reconstructing processes in time 14
5 Final remarks 18 4. Topographic data from satellites
References 18
Simon M. Mudd
Further reading 24
1 The importance of topography 91
2 Collection of topographic data from satellites 92
2. Topo-bathymetric airborne 3 Global and large regional datasets 97
4 Accuracy of global datasets 105
LiDAR for fluvial-geomorphology
5 Implications of increasing resolution
analysis on geomorphic studies 113
Dimitri Lague and Baptiste Feldmann 6 Future developments 118
7 Conclusions 119
1 High-resolution topography: Where is the References 120
bathymetry? 26
2 Synoptic fluvial bathymetry survey
techniques 27 5. Linking life and landscape with
3 Controls on depth penetration and remote sensing
surveyable rivers 32
David T. Milodowski, Steven Hancock, Sonia Silvestri, and
4 Data processing 41 Simon M. Mudd
5 Applications in fluvial geomorphology 44
6 Conclusions and remaining challenges 50 1 Introduction 130
Acknowledgments 52 2 Linking remote sensed data to life and
References 52 landscapes 131

v
vi Contents

3 Passive remote sensing methods 137 5 Science from point clouds in fluvial
4 Radar 141 geomorphology 247
5 Lidar 143 6 Conclusion and outlook 251
6 Airborne electromagnetics 161 Acknowledgments 251
7 Conclusions 164 References 251
Acknowledgments 165
References 166
9. Remote sensing for the analysis
of anthropogenic geomorphology:
6. SfM photogrammetry Potential responses to sediment dynamics
for GeoArchaeology in the agricultural landscapes
Sara Cucchiaro, Daniel J. Fallu, Pengzhi Zhao, Clive Waddington, Paolo Tarolli and Giulia Sofia
David Cockcroft, Paolo Tarolli, and Antony G. Brown
1 Introduction 255
1 Remote sensing 183 2 Materials and methods 257
2 SfM photogrammetry 185 3 Study area 259
3 SfM in geoarchaeology: Agricultural terraces 4 Results 260
in Europe 187 5 A holistic view of land planning 264
4 Final remarks 200 6 Conclusions 267
Acknowledgments 200 Acknowledgments 267
References 201 References 267
Further reading 269

7. Landslide analysis using laser scanners

Michel Jaboyedoff and Marc-Henri Derron 10. Using UAV and LiDAR data for gully
geomorphic changes monitoring
1 Introduction 207 Mihai Niculiță, Mihai Ciprian Mărgărint, and Paolo Tarolli
2 A short history 210
3 Basics of laser scanners 211 1 Introduction 271
4 LiDAR uses 214 2 Study area: The reservoir bottom gullies from Jijia
5 Characterization of landslides 216 Hills (Romania) 274
6 Monitoring 219 3 Materials and methods 276
7 Modeling based on LDTM 222 4 Results 286
8 Discussion and perspectives 223 5 Discussions 297
Acknowledgments 225 6 Conclusions 305
References 226 Acknowledgments 305
Further reading 230 References 305
Further reading 315

8. Terrestrial laser scanner applied

to fluvial geomorphology 11. Zero to a trillion: Advancing Earth
Dimitri Lague surface process studies with open access
to high-resolution topography
1 Challenges in using terrestrial laser scanner to Christopher J. Crosby, J Ramón Arrowsmith, and
understand river dynamics 232 Viswanath Nandigam
2 Data acquisition 233
3 3D point cloud postprocessing operations 237 1 Introduction 318
4 Topographic change measurement and volume 2 Scientific motivations for open access to
calculation 243 topographic data 318
 Contents vii
3 Broad impacts from openly available 12. Reproducible topographic analysis
topographic data 320 Stuart W.D. Grieve, Fiona J. Clubb, and Simon M. Mudd
4 OpenTopography overview and
impact 321 1 Topographic analysis and (reproducible)
5 OpenTopography partnerships 328 geomorphology 339
6 Lessons learned and challenges for 2 Scientific reproducibility 340
supporting open access to topographic 3 Reproducibility in the context of topographic
data 328 analysis for geomorphology 344
7 Outlook 331 4 Barriers to reproducible topographic analysis 354
8 Conclusions 333 5 Making topographic analysis reproducible 357
Acknowledgments 333 6 Conclusions 362
References 333 References 362

Index 369
This page intentionally left blank
 Contributors
J. Ramón Arrowsmith School of Earth and
Space Exploration, Arizona State University,
Tempe, AZ, United States
Ilaria Barone Dipartimento di Geoscienze,
Università di Padova, Padova, Italy
Jacopo Boaga Dipartimento di Geoscienze,
Università di Padova, Padova, Italy
Antony G. Brown Tromso University Museum,
UiT The Artic University of Norway, Tromsø,
Norway; Geography and Environmental
Science, University of Southampton, South-
ampton, United Kingdom
Laura Busato Department of Agricultural Sci-
ences, University of Naples Federico II, Naples,
Italy
Maria Clementina Caputo IRSA CNR, Bari,
Italy
Giorgio Cassiani Dipartimento di Geoscienze,
Università di Padova, Padova, Italy
Fiona J. Clubb Durham University, Durham,
United Kingdom
David Cockcroft Archaeological Research Ser-
vices Ltd, Bakewell, DE, United Kingdom
Christopher J. Crosby UNAVCO, Boulder, CO,
United States
Sara Cucchiaro Department of Land, Environ-
ment, Agriculture and Forestry, University of
Padova, Legnaro, Padova, Italy
Lorenzo De Carlo IRSA CNR, Bari, Italy
Rita Deiana Dipartimento di Beni Culturali
(dBC), Università di Padova, Padova, Italy
Gian Piero Deidda Dipartimento di Ingegneria
Civile, Ambientale e Architettura, Università di
Cagliari, Cagliari, Italy
Marc-Henri Derron ISTE—Institute of Earth
Sciences, Risk-Group, GEOPOLIS-3793, Uni-
versity of Lausanne, Lausanne, Switzerland
ix
Anette Eltner Institute of Photogrammetry and
Remote Sensing, Technische Universit€ at
Dresden, Dresden, Germany
Daniel J. Fallu Tromso University Museum,
UiT The Artic University of Norway, Tromsø,
Norway
Baptiste Feldmann Univ Rennes, CNRS,
Nantes-Rennes Topo-bathymetric Lidar plat-
form, OSUR, UMS 3343, Rennes, France
Stuart W.D. Grieve Queen Mary University of
London, London, United Kingdom
Steven Hancock University of Edinburgh, School
of GeoSciences, Edinburgh, United Kingdom
Michel Jaboyedoff ISTE—Institute of Earth
Sciences, Risk-Group, GEOPOLIS-3793, Uni-
versity of Lausanne, Lausanne, Switzerland
Dimitri Lague Univ Rennes, CNRS, Geosciences
Rennes, UMR 6118, Rennes, France
Mihai Ciprian Mărgărint Department of Geog-
raphy, Faculty of Geography and Geology,
Alexandru Ioan Cuza University of Iași, Iași,
Romania
David T. Milodowski University of Edinburgh,
School of GeoSciences; University of Edin-
burgh, National Centre for Earth Observation,
Edinburgh, United Kingdom
Simon M. Mudd University of Edinburgh,
School of GeoSciences, Edinburgh, United
Kingdom
Viswanath Nandigam San Diego Supercom-
puter Center—UC San Diego, La Jolla, CA,
United States
Mihai Niculiță Department of Geography,
Faculty of Geography and Geology, Alexandru
Ioan Cuza University of Iași, Iași, Romania
Maria Teresa Perri Dipartimento di Geoscienze,
Università di Padova, Padova, Italy
x Contributors
Matteo Rossi Engineering Geology (LTH),
Lund University, Lund, Sweden
Sonia Silvestri University of Bologna, Depart-
ment of Biological, Geological and Environ-
mental Sciences, Bologna, Italy; Duke
University, Nicholas School of the Environ-
ment, Durham, NC, United States
Giulia Sofia Department of Civil & Environ-
mental Engineering, University of Connecticut,
Storrs, CT, United States
Claudio Strobbia Realtimeseismic SA, Pau,
France
Paolo Tarolli Department of Land, Environ-
ment, Agriculture and Forestry, University of
Padova, Legnaro, Padova, Italy
Giulio Vignoli Dipartimento di Ingegneria
Civile, Ambientale e Architettura, Università
di Cagliari, Cagliari, Italy
Clive Waddington Archaeological Research
Services Ltd, Bakewell, DE, United Kingdom
Pengzhi Zhao Earth & Life Institute, Universite
Catholique de Louvain, Louvain-la-Neuve,
Belgium
 Foreword
This latest volume on Remote Sensing of
Geomorphology edited by Paolo Tarolli and Si-
mon Mudd gives a refreshing new look at a
group of topics and processes that have
greatly intrigued quite a number of scientists
most interested in using new technologies to
investigate the landforms of our home
planet. Once global overviews of regional
landforms became a common endeavor more
than 30 years ago (Short and Blair, 1986), the
use of myriads of satellite sensing systems
and new technologies and methods to assess
various environmental parameters became
more common. As the imagery platforms
and technologies continue to improve, in
fact, the methodologies developed here can
also be used to assess a variety of extraterres-
trial bodies as well. No doubt we will con-
tinue to use the Earth-bound term “geo”
morphology to refer to the landforms on
many other such bodies in space as well, even
though that would be a bit of an etymological
misnomer. Still, the newer methodologies
discussed in this book do point to the many
interesting ways of looking at near-surface
and surficial landforms, and continue to
break new ground.
For people who are relatively new to these
technologies, the rather arcane terminology,
even obscure jargon, and profuse uncertain
acronyms can be somewhat disheartening
to those not prepared to work with the
practitioners of these disciplines. Neverthe-
less, probably in any profoundly new area
of technology such as this assessment of
remote sensing of geomorphology, a certain
willingness to tolerate a measure of personal
xi
confusion while immersing oneself in the
details can help facilitate understanding
later. Most scientists seem to be able to wade
through the technical weeds, as it were, pro-
vided that the results one achieves at the end
lead to new viewpoints and useful results.
This volume of papers by a number of
specialists stewarded by Tarolli and Mudd
can be viewed as such because they offer
some introduction into new methodologies,
understandings, and terminologies. Sorting
out the blizzards of acronym names is just
one of the ancillary benefits.
Currently, most geomorphologists are at
least reasonably familiar with digital eleva-
tion models (DEMs) of natural and anthropo-
genic topography as well as various types of
scanning to obtain variable images of objects
or ground surfaces. Less well understood by
many scientists, for example, are the consid-
erable variations also representing topogra-
phy through the use of multiple different
means such as digital terrain model (DTM),
digital surface model (DSM), and triangular
irregular network (TIN), and the fact that
these usages differ from one country to an-
other. Of course, once an image or a picture
of any feature is obtained in a digital form
capable of being measured and assessed in
different ways, then the science gained can
be quite impressive. Once resolutions of
topographic representations become detailed
enough, certain tell-tale landform structures
emerge from noisy datasets and significant
understandings can be obtained of process
mechanics and chronologies of superposition
events. In combination with the diverse new
xii
surficial age-dating techniques now avail-
able, the many varieties of remote sensing
of geomorphology offer more advanced
assessments of the geomorphology of any
place.
Structure-from-motion (SfM) photogram-
metry, hyper-scale, three-dimensional land-
form models, high-resolution topography
(HRT), laser scanning or light detection and
ranging (LiDAR) point clouds of topographic
data obtained from the space, aerial, or terres-
trial devices, unmanned aerial vehicles (UAV)
or drones, geomorphic change detection
(GCD), and DEMs of differences (DoD) are
all new remote-sensing methodologies ap-
plied to various fluvial, agricultural, landslide,
archeological, anthropogenic, subsurface, and
other geomorphic processes discussed in
Foreword
this volume. These diverse new methodolo-
gies are not exhaustive coverage of new dis-
ciplines and methodologies, or of satellite
platforms, but do present useful discussions
that will enable readers to better understand
many of the new remote-sensing technologies.
John F. Shroder, Jr.
Editor-in-Chief
Developments in Earth Surface Processes
October 25, 2019
Reference
Short, N.M., Blair Jr., R.W., 1986. Geomorphology From
Space: A Global Overview of Regional Landforms.
NASA Scientific and Technical Information Branch,
National Aeronautics and Space Administration,
Washington, DC.
 Introduction to remote sensing
of geomorphology
The Earth’s surface has fascinated scien-
tists for centuries. For well over a hundred
years, scientists have speculated about the re-
lationship between surface topography and
the processes that lead to landforms. The first
topographic maps at a national scale were
published by France in the late 18th century,
but these did not become widespread until
the late 19th century. At that time, a number
of scientists began speculating on quantita-
tive relationships between uplift, erodibility,
hydrology, and sediment transport. Survey-
ing by national agencies produced contour
maps that could be used to extract data such
as slope profiles or drainage areas; Ordnance
Survey of the UK began producing contour
maps of scale 6 in to the mile (approximately
1:10,000) in the late 1920s and early 1930s,
and the USGS (US Geological Survey) began
national mapping at 1:24,000 scale. Testing
of hypotheses developed early in the 20th cen-
tury began in earnest as workers in the 1930s,
1940s, and 1950s began using intensively
measured landscapes: the Perth Amboy bad-
lands made famous by the seminal work of
Schumm (1956) were mapped at a scale of
1 in to 10 ft by Strahler and Coates in 1948.
The collection of such data was labor
intensive, however. This is no longer the case.
Advances in the field of physics have yielded
instruments that can collect vast quantities of
data remotely: the volume of data at our fin-
gertips at present is beyond the wildest
dreams of late 19th century scientists. Global
topographic data have gone from 1 km
xiii
resolution in 1996 to a global 90 m dataset
in 2004, a global 30 m dataset in 2009,
down to a global 5 m dataset in 2018. In par-
allel, LiDAR (light detection and ranging)
technology allows very high-density topo-
graphic and point cloud data to be collected
using both terrestrial and airborne instru-
ments. Point densities greater than 20 points
per square meter are now routinely col-
lected with submeter (and sometimes sub-
centimeter) accuracy using airborne LiDAR
instruments.
The impact of satellites and terrestrial
and airborne LiDAR instruments cannot be
understated, but the cost of satellites means
that new missions are primarily funded by na-
tional agencies, and collection of LiDAR data
relies on relatively expensive instruments. In
contrast, nearly every mobile phone contains
a camera, and drone-mounted cameras are
far less expensive than LiDAR instruments.
The advances in the structure-from-motion
(SfM) photogrammetric technique, where
multiple images are used to create point
clouds, have also advanced to the point of
being routine. Today, anyone with a camera
can produce dense point clouds of three-
dimensional objects.
Remote-sensing instruments are useful for
a range of applications relevant to geomor-
phology that go far beyond measuring topog-
raphy. Radar interferometry can be used to
detect very subtle (centimeter-scale) ground
motions; radar and multispectral data can
be used to detect vegetation. A range of
xiv Introduction to remote sensing of geomorphology
geophysical methods can be used to detect
what lies beneath the ground surface. Col-
lectively, these remote-sensing techniques
allow geomorphologists to quantify a rich
array of landscape properties that can help
them understand both intrinsic and extrinsic
factors that shape the landscape, and can help
them understand landscape history.
In this book you will find chapters
reviewing and exploring state-of-the-art tech-
niques in remote sensing relevant to geomor-
phology. We hope that the chapters will
serve as both a reference for experienced
practitioners, and a guide to geomorpholo-
gists looking to use remote-sensing tech-
niques to benefit their studies.
We first have several chapters on specific
techniques: Chapter 1 describes advances in
the SfM photogrammetric techniques that
allow generation of 3D terrain models using
overlapping images acquired from different
perspectives with standard compact cameras
(including smartphone cameras) and geo-
referencing information. Chapter 2 describes
the opportunities offered by the green laser;
such lasers (characterized by a wavelength
of λ ¼ 532 nm) can penetrate shallow water
and therefore provide bathymetry of rivers,
lakes, and estuaries.
We also have chapters giving an overview
of the remote-sensing instruments and data-
sets used to quantify specific landscape prop-
erties. Chapter 4 explores the instruments
used and techniques available for generating
topographic data from space. This chapter
reviews the accuracy and availability of topo-
graphic datasets and discusses the implica-
tions for geomorphic research. Chapter 5
explores the remote-sensing techniques that
can quantify features of living organisms that
are likely to influence, and be influenced by,
geomorphic processes; the focus is on vege-
tation. Chapter 3 presents a review of geo-
physical methods for the characterization of
shallow subsurface.
The book contains chapters on specific
applications of remote-sensing data as well.
Chapter 8 explores the potential of terres-
trial laser scanners (TLS) to solve problems
in fluvial geomorphology, synthesizing
examples of data acquisition, processing
methods, and applications. This instrument
offers an unprecedented combination of sub-
centimeter resolution that allows workers to
capture the geometry of individual pebbles
and quantify precisely the spatial variability
of channel evolution. Chapter 7 explores
the capability of laser scanning in the quan-
tification of volumes, understanding mech-
anisms, and timing of landslide and rockfall
events. The basics of LiDAR performance
are reviewed and an overview of the advan-
tages and limitations of this 3D data acqui-
sition technique are presented. Chapter 10
combines UAV optical imagery and LiDAR
data to evaluate the rate of process for four
reservoir bottom gully systems between two
temporal frames. Chapter 9 explores how
high-resolution topography can help under-
stand how humans are increasingly modify-
ing the Earth’s surface. The chapter focuses
on agricultural landscapes and shows how
new remote-sensing technologies (e.g., air-
borne LiDAR), available to the public, can
provide a better understanding of the inter-
action between anthropogenic elements,
potential erosion, and associated sediment
delivery. Chapter 6 shows how drones and
the structure-from-motion technique can
be used to quantify the history of land use
and land modification in an archeologi-
cal context. Geoarcheological studies have
benefitted from new technological develop-
ments in remote-sensing technologies that
have become an integral and important
part of archeological research. In particular,
structure-from-motion (SfM) photogram-
metry is one of the most successful emerging
techniques in high-resolution topography
(HRT) and provides exceptionally fast,
 Introduction to remote sensing of geomorphology
low-cost, and easy 3D survey for geoscience
applications.
Finally, the book includes two chapters on
open data. Chapter 11 describes the efforts of
Opentopography.org to archive and distrib-
ute high-resolution topographic data. The
authors highlight the fact that open access
to these data and a cyberinfrastructure plat-
form that enables users to discover, manage,
share, and process them increases the impact
of investments on data collection and cata-
lyzes scientific discovery. Furthermore, open
and online access to data enables broad
interdisciplinary use of high-resolution topo-
graphy across academia and in communities
such as education, public agencies, and com-
mercial sector. Chapter 12 discusses tech-
niques to ensure processing of remotely
sensed data in geomorphic and other applica-
tions is reproducible in the chapter. They pre-
sent clear definitions of the terms “replicable”
xv
and “reproducible” for geomorphic research
and communicate the importance of perform-
ing reproducible analysis of remotely sensed
topographic data.
As editors, we are genuinely excited about
the opportunities afforded by the range of
remote-sensing data that is now easily acces-
sible to geomorphologists. We hope that by
offering examples of various datasets avail-
able for geomorphic investigations we can
spur even more uptake of remotely sensed
data than is currently the case.
Paolo Tarolli
Department of Land, Environment, Agriculture
and Forestry, University of Padova,
Legnaro (PD), Italy
Simon M. Mudd
University of Edinburgh, School of GeoSciences,
Edinburgh, United Kingdom
This page intentionally left blank
 C H A P T E R

1
Structure from motion
photogrammetric technique
Anette Eltnera, Giulia Sofiab
a
Institute of Photogrammetry and Remote Sensing, Technische Universit€at Dresden, Dresden,
Germany bDepartment of Civil and Environmental Engineering, University of Connecticut, Storrs,
CT, United States

O U T L I N E

1 Introduction 1 3 Reconstructing processes across

1.1 Brief historical summary and state space 12
of the art 2
4 Reconstructing processes
1.2 Reasons for success in
in time 14
geomorphological surveys 4
4.1 Past and real-time reconstruction 14
2 Method 5 4.2 Time-lapse imagery for 4D change
2.1 Choosing suitable settings to comply detection 16
with the application at hand 6
5 Final remarks 18
2.2 Accuracy considerations in
geomorphological applications 8 References 18
2.3 Direct geo-referencing (DG) for
Further reading 24
flexible UAV applications 9

1 Introduction

Structure from motion (SfM) photogrammetry provides hyper-scale three-dimensional

(3D) landform models using overlapping images acquired from different perspectives with
standard compact cameras (including smartphone cameras) and geo-referencing informa-
tion. As applied to the remote sensing of geomorphology, it is not so much a single technique,

Developments in Earth Surface Processes, Volume 23 1 © 2020 Elsevier B.V. All rights reserved.
https://doi.org/10.1016/B978-0-444-64177-9.00001-1
2 1. Structure from motion photogrammetric technique

but rather a workflow employing multiple algorithms developed from computer vision,
traditional photogrammetry, and more conventional survey techniques (Carrivick et al.,
2016). Recent literature has provided reviews on the importance of SfM in geosciences
(Carrivick et al., 2016; Eltner et al., 2016; Smith et al., 2016) or specific scientific contexts (Man-
cini et al., 2013; Dietrich, 2016; Entwistle et al., 2018). This contribution builds on the existing
literature, to provide a showcase of the technology, relevant to the remote sensing of
geomorphology.

1.1 Brief historical summary and state of the art

The roots of SfM lie in two key fields: photogrammetry and computer vision. When tech-
niques from these fields are combined with both automation and precision, the result is a
comprehensive tool (Pierrot-Deseilligny and Clery, 2011) for geomorphological applications.
Photogrammetry is a relatively old technique (Slama et al., 1980). In this field, the reconstruc-
tion efforts of pioneers in the 1840s initially attempted using a pair of ground cameras separated
by a fixed baseline and followed by applications using cameras for estimating the shape of the
terrain from ground and aerial photographs (Maybank, 1993). With the introduction of aero-
planes and space photography, the development of photogrammetry flourished, with 2D pho-
tographs used to rectify images into appropriate coordinates, or mosaicking multiple frames to
estimate structures or ground elevation. In a parallel effort, the computer vision community
provided the first early algorithms for 3D scene reconstructions by stereo images (Marr and
Poggio, 1976) or to pioneer work on motion-based reconstruction (Ullman, 1979).
The prime formalisms derived in these two communities provided the most important
foundational theory for the SfM community. However, advances in SfM have been spurred
mostly due to the wide range of modern applications. A search in the academic publications
database Web of Sciences (WoS) for Structure from Motion (made in August 2018) delivered
>3000 records since the early 1980s (Fig. 1), covering as many as 125 fields of study.
Computer science and artificial intelligence is the category with the most counts of that
phrase. Engineering is ranked second, remote sensing is fourth, and geosciences is currently
ranked sixth. This wide range of applications of SfM results in research with different goals,
hence emphasizing multiple ways of addressing SfM problems in space and time. The com-
puter vision field features much older publications than other fields, with the first papers
published in the 1980s (Bolles et al., 1987) introducing a technique for building a 3D description
of a static scene from a dense sequence of images, and the latest (Zhu et al., 2018) discussing new
methods for bundle adjustment (the optimization method needed to simultaneously retrieve
the image pose parameters from overlapping images considering corresponding image points).
Notably, the geosciences have only started producing publications incorporating SfM photo-
grammetry in the past decade, but with improvements in the technique moving at an incredible
speed: note that a similar search in 2015 by Carrivick et al. (2016) ranked Geosciences in the
ninth position. In this field, the first work was published (according to WoS) by Heimsath
and Farid (2002). Here, results from three unconstrained photographs characterized hillslope
topography, and yield to an estimated surface with errors of the order of 1 m. In comparison,
one of the last papers published in the field at the time of the search (Smith and Warburton,
2018) illustrates that topographic data from SfM photogrammetry (with errors on the scale
FIG. 1 “Structure from motion” search in academic databases: first 25 results and number of records per discipline (as of August 2018).
4 1. Structure from motion photogrammetric technique

of <1 mm) inherits enough information to analyze the relationship between geomorphological
process and form, at the microscale (few millimeters).
These few examples show an evolution of SfM photogrammetry in time and topics. In the
computer vision field, the emphasis remains on methods for obtaining information from im-
ages, whereas the evolution of SfM photogrammetry is different in geosciences. Early SfM
photogrammetry studies in geosciences emphasized the accuracy of reconstruction, whereas
modern geosciences applications focus more on the information that can be retrieved from
such analyses.

1.2 Reasons for success in geomorphological surveys

For geomorphological studies, the availability of a high-resolution topographic dataset is
fundamental, particularly so for those systems characterized by a complex morphology. We
find four main reasons for the success of SfM photogrammetry in geomorphology: (i) spatial
accuracy and temporal frequency, (ii) cost; (iii) speed and ease of use. A further reason for
SfM’s success, although it is still in its exploratory phase, is and (iv) the possibility of involv-
ing citizens in science. These points are intrinsically interrelated and build on each other to
determine the success of the technique in geosciences.
In geosciences, SfM photogrammetry is a workflow that is virtually independent of spatial
scale (Carrivick et al., 2016), it allows potentially unlimited temporal frequency (Carrivick
et al., 2016) and can provide point-cloud data comparable in density and accuracy to those
generated by terrestrial and airborne laser scanning at a fraction of the cost (Westoby
et al., 2012). It offers therefore exciting opportunities to characterize surface topography in
unprecedented detail, allowing workers to detect elevation, position, and volumetric or areal
changes that are symptomatic of earth surface processes across spatial (see Section 3) and
temporal (see Section 4) scales.
When speaking about the costs of a SfM photogrammetry application, they can vary
depending on sensors, survey design, and ground control points (GCPs)—when present.
SfM photogrammetry sensors are based on consumer-grade cameras, or even smartphones
(Micheletti et al., 2014; Prosdocimi et al., 2016; Sofia et al., 2017), which can be handheld or
mounted on UAV systems. The sensors, mounting systems or cameras can vary substantially
in price and complexity, but the trade-offs between these and the quality of the resulting data
are not well constrained (Cook, 2017). In general, however, the availability of these sensors, and
the opportunity of applying SfM photogrammetry to satellite images (Sofia et al., 2016), histor-
ical photographs or opportunistic sensors (see Sections 4.1 and 4.2), has drastically reduced the
costs of surveys with respect to airborne or terrestrial laser scanners or GNSS. Geo-referencing
forms a fundamental part of topographic surveys and, for SfM photogrammetry work, dense
deployments of carefully measured GCPs are usually used, which can represent a substantial
proportion of the overall survey effort (James et al., 2017a). However, new applications are also
evaluating the opportunity of directly referenced surveys (see Section 2.3).
The availability of free or low-cost fully automated photogrammetric software, of cameras
of any level (from reflex to smartphones), and the recent increase of drones also in the private
and public sectors (News organizations, journalists, and private citizens have employed
UAVs to provide glimpses of natural disaster, for example), allows just about anyone to
generate 3D models for various purposes (Remondino et al., 2017). Processing of the data,
 2 Method 5
in fact, does not necessarily need proprietary software, e.g., (AgiSoft, 2010), but numerous
open-source photogrammetric (OSP) software, e.g., OpenMVG (https://github.com/
openMVG/openMVG), OpenDroneMap (https://github.com/OpenDroneMap), MicMac
(http://logiciels.ign.fr/?Micmac), VisualSFM (Wu, 2011, 2013), SF3M (Castillo et al., 2015),
and 3D data processing tools, e.g., CloudCompare (Girardeau-Montaut, 2015) or MeshLab
(Cignoni et al., 2008) among others, are emerging.
Geographic research is nowadays a data-rich environment, where the most recent advance is
not just the resolution of the data, but the variety and the rapidity with which we can capture
georeferenced data (Miller and Goodchild, 2015). Citizen science can improve research, but it
suffers from necessitating specialized training and simplified methodologies that reduce re-
search output (Raoult et al., 2016). The ease of the use of SfM photogrammetry with a range
of sensors can enable the opportunity of participatory and opportunistic crowdsourced sens-
ing, facilitating the involvement of crowd communication. It is important to underline, how-
ever, that this comes to a hidden cost: the majorities of image-based users are often unaware
of strengths and weaknesses of the used methodology and software, employing it much like
a black-box where they can drop photographs in one end and retrieve a (hopefully) completed
3D model on the other end. It is fundamental, therefore, to provide geospatial tools integrated
with appropriately designed instructional materials (Sofia et al., 2017).

2 Method

The workflow of SfM photogrammetry can be put in a nutshell as follows (e.g., James and
Robson, 2012; Smith and Vericat, 2015; Eltner et al., 2016; Schonberger and Frahm, 2016): In
the first step features are detected in each image and matched between overlapping frames
(e.g., using the SIFT algorithm from Lowe, 2004). These homologous image points are used in
a second step to reconstruct the image network geometry in an iterative bundle adjustment
(e.g., Snavely et al., 2006). During this phase, intrinsic camera parameters, describing the in-
terior camera geometry (focal length and principle point plus additional distortion parame-
ters), and extrinsic parameters, describing the position (three shifts) and orientation (three
rotations) at which images have been captured, are estimated. Furthermore, 3D object coor-
dinates in an arbitrary coordinate system are calculated from the 2D image coordinates of the
homologous image points, creating a sparse point cloud. With the knowledge about the image
network geometry, it is possible to retrieve a dense point cloud, which comprises the calcu-
lation of a corresponding 3D point for almost each image pixel. For a summary of dense
matching algorithms, we refer to Remondino et al. (2014). The resulting 3D point cloud
can be geo-referenced during the adjustment, and the additional information can be consid-
ered to optimize intrinsic and extrinsic camera parameters, or afterwards with a similarity
transformation, thus having no further potential for improvement of the adjustment.
Although algorithmic advances and software tools make the application of SfM photo-
grammetry simple in its usage for topographic reconstruction, basic knowledge about pho-
togrammetric principles are still required for a robust accuracy assessment (e.g., Carbonneau
and Dietrich, 2017) to avoid potential bias in the 3D model leading to misinterpretation of
geomorphological forms and processes. The increased awareness in this regard is highlighted
by increased interest in proper parameter settings and their effect on the final model as illus-
trated in the next section.
6 1. Structure from motion photogrammetric technique

2.1 Choosing suitable settings to comply with the application at hand

Various influences occur on the quality of the final 3D reconstructed surface model using
SfM photogrammetry (James et al., 2019). Careful considerations are necessary during both
data acquisition and processing. Different impacts on model quality, trade-offs, and guide-
lines to achieve most suitable surface models in geo-scientific applications are discussed in
detail by Eltner et al. (2016) and Smith et al. (2016). Thus, this section builds on those reviews,
and summarizes in detail key elements and the related recent literature, providing sugges-
tions to improve SfM photogrammetry models.

2.1.1 Image quality

Image quality is considered to be of great importance because SfM photogrammetry relies
on the successful detection and matching of image features, which is one of the main tasks of
photogrammetry (Gruen, 2012). Because image quality significantly influences these first
steps, making sharp and well-exposed images are the basis for accurate further data
processing (O’Connor et al., 2017). Thus, in order to obtain reliable 3D models, it is important
to start by choosing the right camera and the most suitable configuration for optimized image
capture (Mosbrucker et al., 2017). It is important to note that each parameter setting can im-
prove image quality, and the optimal choice is a trade-off between camera settings that con-
sider the application at hand (Mosbrucker et al., 2017; O’Connor et al., 2017). The main points
for an optimal image quality (highlighted by Mosbrucker et al., 2017 and O’Connor et al.,
2017) are summarized here:
– Images should be captured in RAW format rather than JPEG, due to significantly higher
bit-depth, e.g., 12–16-bit vs 8-bit image information, respectively.
– Cameras with larger sensors should be favored because they enable a higher signal to noise
ratio, as pixels are generally larger and thus more light can be captured.
– The dynamic range of the camera is the camera’s ability to resolve the brightest (saturation
level) and darkest (minimum level of detection) signals, which depends on the resolution
of the analogue to digital signal converter. This range should be set as high as possible, to
allow to capture the entire range of luminance of an observed scene.
– Regarding lenses, a good trade-off between overlap and distortion effects has to be chosen.
For instance, wider angled lenses allow for higher image overlap, but mostly also depict
higher radial distortions.
– For close-range applications, depth-of-field has to be considered, and therefore aperture
should be chosen correspondingly.
– Furthermore, exposure settings are important, which can be evaluated using the exposure
triangle with ISO, aperture and shutter speed at each corner. ISO should be chosen as low
as possible because less noise and a higher dynamic range are the consequences. Shutter
speed should also be as low as possible to avoid blur due to motion but still receive enough
light at the sensor. These settings change with different lenses, object distances, and
moving objects.
– Finally, it should be noted that images with high quality are also achievable with compact
cameras if fixed lenses and large sensors are considered, which is important considering
pay-load aspects in UAV applications.
 2 Method 7
A detailed description of the data and its processing enables a comprehensive assessment
of 3D model retrieval. Thus, for better evaluation and comparability of image quality, data
including metadata about settings during image acquisition should be made available in
an open access repository (O’Connor et al., 2017). This could complement the documentation
spreadsheet introduced by Eltner et al. (2016) that aims to record data-acquisition settings
during the field campaigns and parameter setting during subsequent data processing.

2.1.2 Ground sampling distance

The distance between the camera and the area of interest influences the accuracy and res-
olution of the reconstructed surface model, revealing an inverse relationship between dis-
tance and model accuracy (Smith and Vericat, 2015; Eltner et al., 2016). However, instead
of referring to this distance value alone, Mosbrucker et al. (2017) suggest also considering
ground sampling distance (GSD), which describes the ratio between the distance in image
space to the distance in object space. Different cameras with different focal length and differ-
ent sensors (and thus pixel pitch) lead to different GSDs, even when objects are captured from
the same distance.

2.1.3 Image network geometry

The orientation and position from which images are taken is a key aspect of a reliable 3D
reconstruction. Images should have a high overlap from different perspectives. The distances
between images (which is called the base) should be big enough to avoid glancing ray inter-
sections due to very small parallax angles. At the same time, images should not be taken too
far apart to avoid changes in the image content appearance so great that no homologous
points are detected. Each point for which a 3D model is to be retrieved should be seen in
at least three images. The more images the better due to increasing redundancy in image mea-
surements. Furthermore, the image network geometry should comprise convergent images, if
possible, to avoid systematic errors such as domes (James and Robson, 2014). Other advice
regarding an ideal geometry to avoid unfavorable error propagation include capturing the
area of interest from different distances (Micheletti et al., 2014) and cross-flight stripes in
the case of UAV imagery (Gerke and Przybilla, 2016).

2.1.4 Camera parameter choice during bundle adjustment

Deciding which parameters are to be considered during bundle adjustment, and with what
weights, is essential for a robust model reconstruction from overlapping images. James et al.
(2017a) demonstrated that estimating too many camera model parameters during bundle ad-
justment can lead to over-parameterization and thus errors in the final model. For instance, in
many applications, two radial distortion parameters are sufficient although more values
could be implemented. Remondino et al. (2012) previously discussed the relevance of choos-
ing the correct number of parameters. They observed dome effects for SfM software tools that
estimated the interior camera geometry for each image and suggested using only one interior
camera model if one camera has been utilized to capture the images. Similar conclusions were
also supported by Rosnell and Honkavaara (2012). A potential approach to check for over-
parameterization is to consult correlation values between estimated camera parameters: they
should be low. Furthermore, the significance of each estimated parameter can be consulted to
check for over-fitting (James et al., 2017b).
8 1. Structure from motion photogrammetric technique

2.1.5 Referencing: GCP weights and distribution

The precision and distribution of referencing and control data, i.e., ground control points
(GCPs) and checkpoints (CPs, that are GCPs not implemented during the bundle adjustment),
respectively, are important to guarantee and control the quality of the final scaled model. The
weights of the precision of image measurements of GCPs and tie points have to be chosen
accordingly, to avoid model errors due to over-fitting at the GCPs. Furthermore, reprojection
errors at the CPs should not be much higher than at the GCPs (James et al., 2017a).
Generally, GCPs should be surrounding the area of interest. Also, they need to be well dis-
tributed. A minimum of four GCPs is necessary for increased accuracies, with errors in height
increasing with increasing distance to GCPs (Tonkin and Midgley, 2016). Recent advances in
direct geo-referencing, where models are referenced directly considering the orientation and
position from which cameras were triggered, indicate that GCPs might become less important
in future applications in geomorphology (see Section 2.3).

2.1.6 Exterior influences

Surface properties, e.g., texture, and illumination conditions influence feature detection
and matching significantly. Overcast conditions are preferred to strong shadows. Regarding
surface properties, on the one hand, surface texture has to be sufficient, e.g., snow is less suit-
able due to potentially missing contrast, but on the other hand, it should not be too
complex, e.g., vegetation, whose appearance changes distinctively within shortest distances
and minimal changes of perspectives. Recent studies, however, have shown that it is possible
to reconstruct single blades of grass if the number of images is high enough (Kr€ ohnert
et al., 2018).

2.2 Accuracy considerations in geomorphological applications

Due to the many parameters that influence the accuracy of the final 3D model derived from
the SfM-photogrammetry approach, error reliability is not as high as, for instance, point
clouds derived from terrestrial laser scanning (TLS). Therefore, the need for robust error
modeling is important when using SfM photogrammetry, especially considering the variety
of applications in geomorphology at varying spatiotemporal scales. When performing error
modeling, distinctions should be made between error reproducibility, i.e., error behavior un-
der different conditions, and error repeatability, i.e., error behavior under the same condi-
tions (Goetz et al., 2018). Furthermore, it is important to distinguish between constraining
3D accuracies due to internal and external causes (James et al., 2017b). Internal precision is
influenced by image network geometry and tie-point measurements, whereas external pre-
cision relies on actual geo-referencing. Recent studies have focused on modeling of error be-
havior of SfM-photogrammetry data to improve data quality in geomorphic studies (James
et al., 2017a,b; Wasklewicz et al., 2017).
SfM photogrammetry is not as rigorous in regards to the precision weights when
compared to traditional photogrammetry, and therefore improvements to the accuracy of
the final SfM-DEM (digital elevation model) are still possible if photogrammetric principles
beneath SfM photogrammetry are considered (James et al., 2017a). James et al. (2017a) provide
a workflow to consider and minimize errors when using SfM photogrammetry, which they
 2 Method 9
illustrate with significantly improved error to distance ratios for two case studies. Thereby,
the weight consideration of reference accuracy of GCPs in object space, as well as image mea-
surement accuracy of tie points and GCPs in image space are important to avoid falsely fitting
during bundle adjustment (James et al., 2017a).
Errors are spatially highly correlated when using SfM photogrammetry (James et al.,
2017b), which is in contrast to other high-resolution topography methods such as TLS, where
spatially independent, error behavior is assumed (e.g., Abellán et al., 2009; Kromer et al.,
2015). Thus, rather than using one level of detection (LoD) applied to the entire DEM of dif-
ference for multitemporal change detection, consideration of spatial correlation is important
(James et al., 2017b). James et al. (2017b) use Monte Carlo simulation to calculate precision
maps, which they combined with an adopted M3C2 algorithm, which already considers a var-
iable LoD depending on the complexity of the terrain (Lague et al., 2013) to estimate a spa-
tially correlated error of the SfM-photogrammetry point cloud. However, it should be noted
that precision maps are not able to detect systematic errors such as domes (e.g., Eltner and
Schneider, 2015) and thus independent reference data, e.g., CPs, are needed for a robust ac-
curacy estimation (James et al., 2017b).

2.3 Direct geo-referencing (DG) for flexible UAV applications

Recent advances in the field of devices, combined with advances in the retrieval of accurate
sensor orientation and position of the camera during image acquisition, have the potential to
enable SfM-photogrammetry applications based on UAV imagery that does not require
GCPs. This could potentially revolutionize collection of SfM-photogrammetry data in remote
or dangerous areas, or areas under very frequent observation. Here we discuss the potential
for direct geo-referencing (DG) for aerial platforms.
Benassi et al. (2017) divided geo-referencing into indirect sensor orientation (InSO), direct
sensor orientation (DSO), and integrated sensor orientation (ISO). InSO, or indirect geo-
referencing (IG), uses tie points, GCPs and bundle adjustment to reference the data, and po-
tentiality also considers camera self-calibration. DSO uses solely camera orientation/position
estimates, which complicates reliable camera self-calibration, resulting in potential systematic
errors due to unresolved image block deformations. ISO considers camera position and ori-
entation as well as tie-points to perform bundle adjustment. Furthermore, a few GCPs might
be considered when using ISO, which can be important if self-calibration is also performed. In
this study, we refer to DG as a method that incorporates both DSO and ISO. In general, DG
refers to the direct implementation of estimated orientation and position information of the
central projection center of the camera during image capturing to reference image-based re-
construction products (Pfeifer et al., 2012).
Utilizing UAV data with DG has great advantages because access to changing or danger-
ous environments for the purpose of including GCPs will not be needed. Inclusion of GCPs is
still one of the main limitations for flexible UAV applications (e.g., Carbonneau and Dietrich,
2017; Forlani et al., 2018). Furthermore, IG implies high demands regarding the GCP distri-
bution (e.g., James et al., 2017a; Tonkin and Midgley, 2016), because low-cost position and
orientation estimation devices, as well as low-cost cameras, demand robust GCP networks
for reliable adjustment during 3D reconstruction (Gerke and Przybilla, 2016).
10 1. Structure from motion photogrammetric technique

FIG. 2 Error estimates (standard deviation or RMSE) of case studies using direct geo-referencing or integrated sen-
sor orientation (considering 1 to a maximum of 4 GCPs) are related to round sampling distance (GSD) considering
case studies by Rehak et al. (2013); St€ ocker et al. (2017); Eling et al. (2015); Mian et al. (2015); Rehak and Skaloud (2016);
Forlani et al. (2018); Benassi et al. (2017); Gabrlik et al. (2018); and Gerke and Przybilla (2016). If GCP and no GCP were
evaluated within one study, solely, the case for GCP included is illustrated.

Most current low-cost UAVs are equipped with GNSS devices that do not enable real-time
kinematic processing (RTK) or postprocessing kinematic (PPK) to correct the GNSS signal
leading to accuracies of the final 3D model in meter-ranges (Turner et al., 2014) or
dm-ranges (Gerke and Przybilla, 2016; Hugenholtz et al., 2016; St€ ocker et al., 2017). However,
if RTK- or PPK-GNSS is possible, this will result in a high potential for DG of UAV data (Bláha
et al., 2012) and cm-ranges can be achieved (Fig. 2). Furthermore, using RTK- or PPK-GNSS
can help to decrease image block deformations significantly (Gerke and Przybilla, 2016), and
thus systematic errors such as domes in the 3D model (James et al., 2017b) might be mitigated.
Generally, an integrated GNSS and IMU (inertial measurement unit capturing angular
changes and accelerations) approach is used in support of DG to allow for precise flight-
trajectory reconstruction. This provides knowledge about the position, attitude, and velocity
of the UAV during image capturing (Chiang et al., 2012; Pfeifer et al., 2012; Eling et al., 2015).
Due to the weight constraints of UAVs, microelectromechanical systems (MEMS) are utilized
as low-cost IMUs (Pfeifer et al., 2012).

2.3.1 Achievable accuracies

Errors of the SfM-photogrammetry result reach about 0.1% of flying height for low-cost
GNSS devices with no kinematic processing (Carbonneau and Dietrich, 2017). However,
the picture is different for RTK- or PPK-GNSS applications. To evaluate the accuracies of
the final 3D model, studies using DG with RTK- or PPK-GNSS have been compared. It has
to be noted that different studies utilize different parameters, e.g., some use lever arm and
 2 Method 11
boresight corrections and others do not consider their influence. Error ranges are high, i.e.,
between 2 and 20 cm, and errors increase with increasing GSD (Fig. 2).
Furthermore, these studies reveal that error improves significantly if camera calibration is
performed with self-calibration instead of a pre-/postcalibration (e.g., from 55.3 to 4.1 cm in
Gabrlik et al., 2018). Also, GCP consideration is important because the studies show that if at
least one GCP can be implemented, the error decreases strongly (e.g., from 10 to 3 cm in
Forlani et al., 2018). Beside absolute error, relative error (ratio between measured error
and object distance, Eltner et al., 2016) is diverse for all studies, as well. Average ratio amounts
are 1:1300 with a standard deviation of 700.

2.3.2 Guidelines for DG applications

If direct DG is performed, several aspects have to be accounted for:
– Pre- or postflight camera calibration should be considered if no GCPs are possible because
residuals of the self-calibration of the interior camera geometry propagates into the final 3D
model in object space, which can be compensated for by exterior orientation when GCPs
are used in the bundle block adjustment (Gerke and Przybilla, 2016; Forlani et al., 2018).
Also, systematic errors are absorbed by exterior camera orientations and positions and thus
estimated positions and orientations do not coincide anymore with actual physical image
network geometry during image capturing, which, however, does not matter if the focus is
on the final accuracy of object points (Cramer et al., 2000).
– If possible, at least one GCP should be included, which is relevant for estimating interior
camera geometry during self-calibration, allowing for almost identical results to IG
applications (Benassi et al., 2017; Forlani et al., 2018; Gabrlik et al., 2018). Using many more
GCPs compared to just one indicates no further improvement of the final 3D model (Gerke
and Przybilla, 2016).
– Specific flight patterns should be chosen, especially if camera self-calibration is aimed for,
to avoid unfavorable parameter correlation, i.e., use cross-flights, especially if the terrain
comprises no large height shifts (Gerke and Przybilla, 2016), fly at different heights, and/or
capture convergent images (James et al., 2017a).
– The importance of weights given to the parameters of exterior orientation has to be
considered. Orientation estimates of the sensor are still not sufficient for high weights on
angles and hence achieve best results if high weights are assigned to the position but low
weights on attitude during bundle adjustment (St€ ocker et al., 2017). Furthermore, the
choice of a set of parameters has an impact on height accuracy, and using the parameter
weights can lead to different accuracies with different software, potentially due to different
consideration of weights of observations (Benassi et al., 2017).
– Offsets between projection center of the camera and the position of the GNSS receiver [i.e.,
lever arm, and orientation of the IMU, i.e., boresight (angular misalignments)] need to
be estimated (Chiang et al., 2012). Furthermore, synchronization issues between the
camera shutter release and GNSS/IMU signal logging have to be considered (Gabrlik
et al., 2018).
Ultimately, the best practice for flexible and accurate direct referencing of UAV data is the
combination of traditional aerial triangulation and implementation of directly measured,
12 1. Structure from motion photogrammetric technique

sensor-position information. Thereby for highest accuracies, position estimates are used as
approximation values during adjustment, in combination with tie points and very few GCPs,
if the terrain allows for it (Chiang et al., 2012). Estimates of camera position mitigate block
deformation issues (Gerke and Przybilla, 2016) and support tie-point detection in areas of
unfavorable texture (St€ ocker et al., 2017), whereas GCPs can be further used as checkpoints
to enable reliable error estimates.

3 Reconstructing processes across space

Accurate, precise, and rapid acquisition of topographic data is fundamental to many sub-
disciplines of physical geography (Smith et al., 2016). Conceptually, the patterns of earth sur-
face processes detected in any topographic dataset are a function of scale. The scale of a study
can relate to the overall area encompassed by an investigation (extent) or the size of the in-
dividual units of observation (process resolution, which we define as grain). In SfM-
photogrammetry applications in geomorphology, several fundamentally different extents
and grains concerning processes are known, but the boundaries or thresholds among them
may be fuzzy.
At the smaller grain, for example, the rigorous modeling and quantification of soil-water
erosion processes require detailed information about the topography of the land surface with
appropriate resolution and accuracy. Thanks to SfM photogrammetry, this microscale grain
can be assessed at multiple extents (hillslope, plot, and micro-plot scale; see Fig. 3), allowing
quantification of detailed physical changes of soils and their influence on surface morphology
even at submillimeter resolution (Kaiser et al., 2018). Among the challenges of SfM photogram-
metry in this type of analyses, we can mention the establishment of a common and sufficient
reference system for the different DEMs considered, determination of errors in the generation
of DEMs, selection of appropriate criteria to obtain reliable changes, error propagation, and
validation of the procedure by comparing the results with actual sediments collected during
the experiment (Gessesse et al., 2010; H€ansel et al., 2016; Glendell et al., 2017; Prosdocimi et al.,
2017; Balaguer-Puig et al., 2017a,b; Eltner et al., 2018; Tarolli et al., 2019). A further problem is
that many geoscience processes associated with soil surface microtopography occur on natu-
rally vegetated surfaces, but few guidelines exist for the acquisition and treatment of SfM pho-
togrammetry data on vegetated surfaces (Nouwakpo et al., 2015).
Increasing the grain, remote sensing in fluvial geomorphology using SfM photogrammetry
has increased significantly in last 5 years (Entwistle et al., 2018), with many recent advances
in, for example, river restoration (Marteau et al., 2017; Woodget and Austrums, 2017). From
the smallest to the largest scale, SfM photogrammetry has been proven useful in laboratory
flumes (Morgan et al., 2017), for grain size measurements (Micheletti et al., 2014; Bertin and
Friedrich, 2016; Pearson et al., 2017), for erosion assessment (Prosdocimi et al., 2016;
Hemmelder et al., 2018; Jugie et al., 2018) or river ice quantification at embankment level
(Alfredsen et al., 2018), and to study riverbed evolution (Lane et al., 2003; Javernick et al.,
2014; Dietrich, 2016; Cook, 2017). Recently, a further “grain” investigated in science has been
underwater bathymetry reconstruction, with the pioneer works by Woodget et al. (2015, 2017)
and Dietrich (2017). SfM photogrammetry in subaerial studies can provide consistent results
if systematic errors due to refraction impact are accounted for (Mulsow et al., 2018). Partama
et al. (2018) found that using coregistered image sequences or video frames to mitigate the
 3 Reconstructing processes across space 13

FIG. 3 Surface reconstruction across scales from UAV imagery (1 cm resolution) illustrated for a hillslope. (A) At
hillslope scale, topographic features such as slope are measurable. (B) At the next scale traces for tillage (across-slope)
and local potential accumulation spots become obvious. (C) At the last scale single aggregates, and, e.g., their rele-
vance for roughness, can be evaluated.

effects of waves and water reflections could increase the size of reconstructed areas under
difficult observation conditions. This method has the potential to significantly boost SfM-
photogrammetry applications for bathymetric measurements. Overall, monitoring changes
on stream channels with SfM photogrammetry gives a more complete spatial perspective
than the traditional method of cross sections when quantifying small-scale geomorphic
change.
Another random document with
no related content on Scribd:
 In Pace Conquiescare
Toward midnight, Paul Duval emerged from 355 McDougal Street,
quietly closed the battered door behind him, and descended the
steps. He slouched along the street, with the brim of his felt hat,
which dangled over his ears, flapping up and down to the rhythm of
his stride. Probably, he thought, some one would take him for a
murderer or a burglar—although, to be sure, such people use
automobiles in this twentieth century. Paul was especially conscious
that the policeman leaning against a post of the elevated railway, did
peer at him searchingly, whistled something, and twirled his stick
meditatively. But perhaps all this was fancy, aided by the dim light of
the arcs.
It was, however, likely that Paul carried with him a remnant of the
atmosphere of the death-chamber he had just left—the green-walled
room in the rear of 355 where Hanaré Tierens had died—and that
the remembrance of this most recent experience created in his mind
a marked sensitiveness to ghostly things such as policemen and
Greenwich Village arc-lights. That calm, livid face, with its peculiarly
French nose, had passed through some experience of which Paul, at
least, knew nothing. He still felt the pressure of Hanaré’s hand,
which he had held until the last moment. It had relaxed and become
dead. What a world of truth and wonder was there in that moment,
that relaxation!
Few men, Paul thought, had ever passed through emotions such
as his own had been. It was bad enough to see one’s old friend and
adviser die; to feel a hand relax, the way Hanaré’s had; to realize
that it belonged no longer to a friend or an adviser. This, Paul
reflected, was bad enough. But there had also been a girl—Hanaré’s
daughter; a girl whom Paul had passionately loved for the last five
years; a girl whose drawn, white face stood out now in his memory,
like a ghost, to aggravate the torture in his heart. These two had sat
facing each other during the last hours, when the doctor had gone,
and the rest of the house was asleep. They had not exchanged a
word. The tragedy had been heightened by the silence. Paul had
expressed his love too often for her to be able to forget, even at this
time, the intensity of his passion. And once, when their eyes met, he
knew that in her young heart one more sorrow had thus been added
to her present burden—a sympathy for him, and a feeling of almost
shame that she could not respond to his love.
Then there had been a frightful kind of mental telepathy which
carried even his most involuntary thoughts over to her. How could he
help thinking that since she was now alone, without her father, she
might accept him as a lover and a protector? How could he avoid
extending his sympathy for her distress into a conviction that, since
she needed comfort, some overt expression of his love was justified?
Indeed, once when she had laid her head in despair upon the dying
man’s breast, Paul had stretched out his hand and stroked her hair.
She had, then, taken his hand in hers, pressed it, and released it.
The situation only seemed to strengthen the barrier between them,
and to make them even more intensely conscious of it.
These thoughts flowed slowly through Paul’s mind, now that he
was out on the street, walking toward his apartment. He cursed
himself for his selfishness and for bringing into a death-chamber
such passions and emotions, thereby to heighten a young girl’s
distress. What if they were the passions of a lifetime! What if they
had caused him inexpressible suffering! He was none the less a
selfish brute, immersed in his own selfishness.
Upon passing a quick-lunch room, he decided to enliven his tired
mind by indulging in some coffee and doughnuts. He opened the
door, walked past the shiny, white-topped tables, and approached
the counter. Here he was at once surprised by the beauty of the girl’s
face, which confronted him, and which stood out against the
background of coffee containers, cups, saucers, shredded wheat
boxes, and the like, as though an inhabitant of his dreams had been
transposed to this earthly environment. Paul, who was sleepy and
dazed, stared at her until she was forced to drop her lashes and hide
from him the blue depths of her eyes.
 “I am Paul Duval,” he said, in his absent-minded way, “and I
should like some coffee.”
The girl turned and drew it from the container. Paul watched her—
the slim back and the delicate, white skin which showed through her
fine blouse.
“A strange world,” he sighed as she turned toward him with the
coffee. “Isn’t it?”
He had not meant it offensively. And as she gazed into his vague,
grey eyes, saw the sallow cheeks and the whimsical expression on
the mouth, she divined that he was not talking about the world but
about himself.
“Do you think so?” she smiled.
“Don’t you?”
“I think you are a strange person!” she laughed, turned on her heel
gayly, and pretended to busy herself with something below the
counter. Paul noticed that her voice was quite cultured.
“Do you work here—always?” he asked.
“One week,” she passed nonchalantly from one little task to
another. “This is my first night’s work—my last, too—Fred was sick.”
“Who’s Fred?” Paul found it difficult to keep up with her.
“Night-man. Did you want anything else?” She leaned across the
counter, exposing her slim arms, and a pair of delicate hands. She
looked up at him, laughing.
Stupidly he remembered that he had come there to drink coffee.
He fumbled for his cup.
“No,” he said, “—eh—that is—I’d like some doughnuts.”
She procured the doughnuts, and Paul reluctantly shambled off to
a nearby table, where he sat down, facing the counter. It was indeed
strange that this girl should exert so much attraction over him. He
had seen beautiful women before during his twenty-six years of
varied existence. However, he remembered with a smile that since
women had meant anything to him at all, he had been in love: first
with a stolidly serious young lady, who was now married to a man
much older than herself, and then with Marie Tierens. This latter
affair had been going on for the past five or six years. It had become
his ideal. It had given Paul the conviction that if a man is going to
marry a decent woman,—well—the least he can do is to be decent
himself. At the heart of civilization, he thought, lay the unitary
standard. And thus he had crossed in safety numerous pitfalls which
present themselves to the average hack-writer—the small dealer in
ideas.
But to-night, ah, well—even one’s deepest ideals are shattered at
times. The excruciating emotions of the past few hours had left him
like a rudderless ship, adrift in a sea of bewildering passions. Hanaré
was gone now. Without Hanaré life could never be the same. And
Hanaré’s daughter had changed. She had become an independent
woman. There was defiance in her eyes, instead of that ancient
girlishness which had always kept hope alive in Paul’s heart. Indeed,
the world had changed. For better or for worse, Paul, too, had
changed.
In those intense moments of a man’s lifetime, wherein the past,
together with the ideals which epitomize the past, are relinquished
and a new method of life undertaken—in those moments a man is
not fully conscious of all that he is doing. He moves in response to
the predominant feeling in his heart. And there opens up before him
new and unexplored vistas of life, at the other end of which he hopes
to find some sort of Eldorado. To-night Paul was craving for beauty.
Beauty to alleviate the coarseness of the death-chamber. Beauty to
help him forget the face and the eyes of a girl who could no longer
truly be called “his” girl.
He was awakened from his short reverie by a voice close beside
him. “Good-night,” it said cheeringly. Paul looked up to behold the
girl of the counter, in a blue serge dress, with a dark blue hat slanted
to one side of her head. She waved to him as she passed. Paul
gathered his queer legs together, and arose.
“Eh—are you going?”
 The girl turned. “Yes,” she said, “I’m only on duty until one o’clock.”
“Really? Is it that late?” He felt for his watch, but could not find it.
“Good-night,” said the girl again.
“I say,” said Paul, as though he were embarrassed, “perhaps—well
—wouldn’t it be rather nice if I were to take you home? We could—
eh—go somewhere—dance—first.”
“Dance! At one o’clock?” she laughed. “I don’t think we could.”
“Of course.”
She made no move to go, nor did Paul, who was standing close to
her. At length she took hold of his arm.
“Well, are you coming?”
They walked out of the restaurant and down the deserted street
together.

The night spent itself. Some sort of a dawn crept across the city
and touched the edges of the windows in the rear of 355 McDougal
Street. As the grey light penetrated the room on the third floor, a girl,
who had been lying across the body of a dead man, arose, looked
stupidly about her, rubbed her eyes, and went over to the window
where she gazed across the damp Greenwich Village roofs. She
thought, perhaps, that she was going mad, with this silence which
penetrated her whole nature, like the cold dawn that had just
penetrated the night. But, strangely enough, it was not altogether her
own loneliness, nor yet the painful sense of loss at the death of her
father, nor even the ghostliness of his figure on the bed, that was
thus driving her toward insanity. Rather, it was the remembrance of
Paul’s face, the knowledge of his suffering for her, and the feeling
that, although she could never love him—really love him, as she had
pictured love in her girlhood dreams—still, the death of her father
had removed the last tangible excuse which she had to offer him.
She felt that it was not right to add to his sorrow for Hanaré’s death,
a still larger grief caused by her own selfishness.
 She smiled tearfully as she gazed out of the window. Why, she
was making it appear like a case of duty!—and, of course, no one
ought to marry for duty. Actually, it was not altogether a case of duty.
Actually, she was alone—and afraid of her own loneliness. Indeed,
the image of Paul came to her like a light shining through the
darkness. He was forced upon her, by the strength of circumstances.
Hanaré was gone now. Without Hanaré life could never be the same.
Paul had become essential to her very existence. Love him or not,
he was essential to her existence.
As the sun rose and the day wore on, and she went about her
necessary tasks, it seemed to her almost as though she loved Paul.
She had never had a feeling quite so compelling as this. Before her
father’s death, she had never wished to marry. She had
contemplated some sort of a career, with her painting and her
sculpture, which she inherited from Hanaré. Besides, her father had
needed her. He had been a solitary man, with few friends, a dreamy
personality, and so absent-minded that he required her constant
attention. Thus life had seemed to her best, close to her father’s
side, managing the little household, and doing her art at her leisure.
The thought of children to take care of revolted her. And as yet no
passion had entered her life, sufficiently powerful to make this
secluded existence seem trivial or repulsive. Nor was there anything
about Paul Duval to attract her strongly. He was the nicest and
kindest man in the world, and he loved Hanaré; but for a husband—
well—what was the use of a husband, anyway?
She felt differently now. She wanted Paul. Yet all day he stayed
away.
Toward evening there came to her again the sensation that she
was going mad. It was simply inhuman of Paul to leave her alone like
this. There had been, of course, the neighbors, who offered their
sympathetic assistance, and who tried to comfort this strange, silent
girl, whom none of them understood. But because of her yearning for
Paul, the neighbors only aggravated her nervous sorrow. And
although she had consented to sleep with an elderly woman in
another part of the house, until her father should be buried,
nevertheless, late that night, she felt herself irresistibly drawn to
Hanaré’s stiff corpse; and she crept into the ghostly room, in her
night-gown, to appease that unnatural craving. This was about ten
o’clock. She sat for some minutes on the edge of the bed, but could
find no consolation. Suddenly she jumped up with the wild resolve to
go to Paul’s apartment and find out what had happened to him.
She reflected, as she slipped on her clothes, that this was a most
unwomanly course of action. She was impelled toward it by the
almost inhuman nature of her circumstances. She hoped Paul would
understand. She hoped nothing had happened to him. Perhaps she
could even be of some comfort to him, in this recent sorrow which so
obviously depressed him.
In fact, as she made her away along the winding streets of
Greenwich Village, Marie began to feel almost exultant. A new joy
entered her heart, because she was relieving herself of intolerable
burdens, and because, too, she was bringing to Paul a surprise-
present for which he had been waiting many years. She began
timidly to picture to herself Paul’s expression, first upon seeing her,
and later—perhaps even days later—when he should realize what
this new resolve of hers meant to both of them. She found herself
immensely relieved at the thought of transferring her small
belongings from her present dreary apartment to his own. Her
collection of books, her pictures, yes, and even her paints and her
sculptor’s tools—all these she would show to Paul as belonging to
both of them together. In his eyes and in his mouth would come that
look of appreciation for things which were such precious
possessions. It would be inexpressible relief! A happy life! They were
both dreamers—
She arrived, a trifle breathless, at his apartment, which was four
stories up in a brick building that boasted of no elevator. She
knocked several times on the thin, wooden door, but no one
answered. So she tried the door knob, found that it was unlatched,
opened the door timidly, and gazed in. There was a vestibule leading
into the sitting-room, and since the latter was lighted, she proceeded
on tip-toe toward it. Upon entering she perceived a long, narrow
room, hazy with tobacco smoke and heavy with the odor of stale
whiskey. The bric-a-brac and furniture were in a state of disorder.
There were a couple of empty bottles on the table—glasses and
books. She perceived a thin, sallow figure, sprawled out in the morris
chair, staring at her in a glazed way, like a dead man.
“Paul!” she cried.
Paul moved slowly, blinked his eyes, shuddered. “Eh?”
“Paul!”
“I should not have wished you to see me like this,” he said, as
though her coming were the most natural thing in the world.
“Are you—all right?”
“Yes, a little dazed.”
“Why have you been drinking?”
He sighed and bowed his head, in a tired way, until his chin
touched his disordered bow-tie. But he did not answer.
Impulsively Marie ran over to his side and knelt there, with her
arms upon the chair.
“Are—are you drunk, Paul?” She had rarely seen drunken men.
He raised his head then and looked into her eyes, which were so
close to his. “No,” he said. “I have been drinking, but I am not drunk.
I am merely dazed, by death, and by life—but mostly by life. Life is
so strange. Have you never thought that?”
“Yes.”
“No—no—no! Not the way I have thought it. You only know the
half of life—Hanaré’s half. You have inherited, now, Hanaré’s
domain. Innocent, childish Hanaré! You are the mistress of his
innocence and his naïveté. But it will never—never—never be the
same again.”
“Paul! What do you mean?”

“O Welt du bist so nichtig!

Du bist so klein, O Rom!”
 They lapsed into silence then, for Marie saw that he was in one of
those unintelligible moods, which had often come upon him, but
which she had never seen so pronounced.
“Why did you stay away all day?” she asked.
“Have I been away all day? I had forgotten.”
“I have been lonely and miserable, Paul.”
“I am sorry. I have forgotten.”
“How could you forget?”
“How? There are plenty of ways to forget.” He arose and strode up
and down the room restlessly.
“How could I forget?” She looked at him as though he had
wounded her.
“I don’t know.”
They were silent while Paul continued to walk up and down. At
length he proceeded.
“Certainly my being there wouldn’t have helped much, would it? It
isn’t as though you had ever allowed me to love you or comfort you!
God knows, I’ve been ready to do so—any time. I thought you hated
me. Do you?”
“Yes,” she replied, “at this moment I hate you intensely.”
“Why did you come here, then—if you hate me?”
“O Paul, Paul! Because I needed you!”
He stared at her. This woman! “I didn’t suppose anybody needed
me now, except the devil.”
He saw then that she was crying and that he had hurt her
tremendously. He saw distinctly that he had been unjust. But his
mind could not piece together the broken fragments of the situation.
He, too, had been unjustly treated: it was not fair for a woman to
allow a man to love her for six years, and to hold herself away from
him merely for the sake of her own career—her own whimsical
happiness. He felt that in the hour of need Marie had not been with
him. He felt this even more keenly than his own cruelty toward her
now.
“My God!” he exclaimed, in the midst of his meditation. “What
twenty-four hours will do!”
There was undisguised bitterness in his words; a bitterness which
Marie, conscious of the unprecedence of her behavior, construed as
an expression of his scorn for what she called her “unwomanliness”.
Her excited mind only served to intensify the horrid picture which she
had drawn of herself. To think that she had come this way to Paul, of
all people! Even the awful atmosphere of her father’s death-chamber
could not excuse her for doing so. She wished that she could hide
herself away. She was ashamed of her body—her very existence.
But Paul was not thinking of these things. He was merely
astounded at the change that the night had wrought in himself.
“I wish to hell Hanaré hadn’t died,” he exclaimed suddenly, and
without any reason for it. “Life is nothing but a constant attempt to
adjust ourselves to the tragedy of existence. Since we cannot tell to-
day what will happen to-morrow, we never quite succeeded in our
adjustments: and so, there’s always a tragedy. We go on and on—
like that!”
He felt master of himself now. But Marie supposed that he was
lecturing her. There was an element of brutality in it.
“If we were automatons,” Paul proceeded, as though the sound of
his own voice helped to drive away the real tragedy behind—“if we
were all automatons, who acted out one day the same as any other,
incapable of making fools out of ourselves,—why then, life might be
worth living. But some fool of a God—a fool God—gave us this
power to make mistakes. Marie, for the past six years both our lives
have been mistakes. And now just see what you have done—and
what I have done.”
Marie stood facing him, and clenched her fists.
“Paul Duval, you are undoubtedly the most unfeeling man in the
world—the most pitiless—the most un—unreasonable. I know I’m a
little fool! Do you suppose I have no sensibilities? Do you sup—?
Oh, heaven!” She fell back again into the armchair, weeping.
The situation between them had changed tremendously in one
night, because his ideals had become incompatible with her ideals,
his life had lost that simplicity and innocence which they had once
shared together. Paul found that his love for her, just yesterday so
vivid and passionate, had changed, and had converted itself into a
red and golden derelict of the past, which he still loved, though in a
different way. Like Dante, his love for a face and a living body had
transformed itself into an intellectual remembrance—an ideal—a
hope which, while it might later be fulfilled in some immortal
existence, had lost, once and forever, its earthly potency. Just as the
death of Beatrice had forced Dante to relinquish the earthly passion,
so the death of Hanaré, which brought the confusing emotions of last
night, had led Paul to reconceive Marie and transform her into a
vanished reality, an ideal, rather than a living being. He had tasted,
now, that side of life which does not permit of the more refined loves.
Indeed it was a strange position to be in: and the tragedy of it lay
in the fact that he could never make it clear to Marie why he had
done as he had, and why the relationship between them was now
changed. Tremendous, this change!—almost infinite in character.
Especially, she would never understand how it had come about so
quickly. He sighed. “With questions like this,” he said, “of life and
death—time has little, if anything, to do.”
He began to reflect that the course he had taken was an evil one.
And although the forces which had led him along this course were
still potent, nevertheless the sudden apparition of Marie into the
midst of them recalled his old life with her, if nothing else. And this
feeling, that he had better go back, repent, and, if possible, forget the
slight digression of the night before, grew upon him, just as a
glimmer of light, which increases in intensity, turns at last into a ball
of fire. He even came to the conclusion that it was his duty to marry
her. He felt that he had no right to add to her sorrow for Hanaré’s
death, a still larger grief caused by his own selfishness.
 He smiled then. He was making it out to be a wretched case of
duty—and of course no one ought to marry for duty. Actually it was
not altogether a case of duty. Actually he was alone in a new sea of
conflicting passions, lost ideals, and hopelessness,—and he was
afraid of his own loneliness. Indeed, the picture of Marie sitting there
came to him as a light shining through the darkness. He no longer
loved her. True. But he felt that in her he could find some salvation
from the horrible destiny which immediately confronted him, and a
relief from his present wretchedness.
“Marie,” he said suddenly. “You know that I have loved you!”
“Yes. You have said so. I believe it.”
“Last night there came into my life something which you could not
understand—which I cannot explain now—which some day I hope to
forget.”
She looked up at him, anxiously, as though fearing unknown
things.
“Ah, don’t look at me that way. Let the past take care of the past.
You shall know some day. I will tell you.”
“Why don’t you tell me now?”
“Because you would not understand—you would not appreciate—
nor could I tell it as it is.”
“You only arouse my worst fears by talking this way,” she said. “I
came to you as a friend, for consolation. I came in order to forget
that horrible room. I wanted your companionship—perhaps for
always. But you have only succeeded in making me more disturbed.
I do not understand you.”
He went over to her chair, and sat down beside her, and put his
arm around her.
“Come,” he said. “You and I must escape the tragedy of our
existence. Together we will fly away from it. You will forget that room,
and I—I shall forget myself.”
 She drew away from him a little—from his impetuosity. “I don’t love
you, that way,” she said.
“Great heaven! Nor do I love you that way, any more. You are too
idealistic, Marie. Marriage, for you and me, is no longer an ideal, but
a necessity. We will escape, that way. We will rest in peace and
Hanaré’s death will be forgotten.”
She made no reply, but sat there as though meditating. Suddenly,
from far out in the city, came the boom of a clock—a lonely thing
beating the hour of midnight. It awoke Paul to realities. And, although
he had so far been master of the situation, he now lost control of
himself, and cried: “Twelve o’clock! You must go now, you must go!”
And as she stared at him, mystified, he cried again, “You must go,
you must go!”
He took her arm, and she arose. They stood facing each other.
“Promise me,” he said, “that you will marry me—to-morrow.”
She dropped her eyes. Impulsively he took her in his arms and
kissed her—not passionately, but as he would a little child. And then
he led her toward the vestibule.
Even then there was a knock at the door. He did not answer. He
looked at Marie, and she at him.
“Who is that?” she asked.
He turned bitterly away. “Nobody! Fate!”
The door opened, since he had given no answer. There was a
moment of suspense while the visitor was hidden in the vestibule.
Then the girl of the counter, looking extraordinarily pretty, came
toward them. She started imperceptibly upon seeing Marie, but
regained her composure.
“Good evening,” she said.
“Good evening,” Paul replied, but did not move.
The girl proceeded to take off her hat in front of the mirror, and to
make herself at home, adjusting her brown hair prettily and helping
herself to a cigarette from the box on the table. “You look tired,” she
said “And the room’s a mess. Is there no one to clean up for you?”
Paul did not answer.
Marie could do nothing but stare. She stared at the girl and then at
Paul. Suddenly she ran toward the door and was gone.
“Who was that?” asked the girl.
“That is Hanaré Tierens’ daughter. Why did you come?”
“You told me to come at twelve.”
Paul sat down and put his hands over his face. “You have ruined
everything,” he said.
“So you told me last night.”
“Did I? I had forgotten.”
“You said an old friend had died—and that you were in love with
his daughter.”
“Really? I was quite frank, wasn’t I?” Her reference to Marie
exasperated him.
“Don’t be cross.” She came up close to him and put her hand upon
his shoulder.
“I am not going to kiss you,” said Paul, anticipating her.
“I haven’t asked you to, have I?”
“No; but I’m going to marry Ma—Hanaré’s daughter.”
“There! I knew you had something like that in your mind! You look
so—so determined,” she laughed, in spite of her obvious vexation.
“I am determined.”
“What are you going to do with me?”
He started. “You? You aren’t going to make a fuss about it, are
you?”
“That’s impudent!” She turned away from him and sat down upon
the sofa restlessly.
 “I’m sorry.”
“Oh, you needn’t be! Last night you let me know well enough what
you thought of me. But I don’t mind, because I know that is what I
am. I can’t remember ever being anything else; though, of course, if
one is to be a man’s mistress, one has the right of choosing the man.
I prefer education to ignorance, and a decent amount of politeness to
mere brutality.”
“I am not merely brutal!” He stood up and faced her as he said it.
“It’s you who are brutal—or at least you will be before you have done
with me. Women are all cruel, because they understand men so well.
Our souls are torn first by one and then by the other. I should like to
make you see, however, that I have a duty to perform.”
“Duty?” She arose from the sofa, and came to stand beside him
again.
“A duty to myself and—to some one else.”
“There’s a difference between duty and love, isn’t there?” This time
she appeared to have no scruples, for she put her arms around him,
frankly, and stood looking up at him. He held himself rigidly away
from her.
“I am not going to kiss you,” he said again. That appeared to be
his last defense in any case.
She made no answer to that, except to deny it with her eyes.
“You’re a witch!” he cried, drawing away from her suddenly. “A
positive instrument of the devil. Leave me alone!”
“Why?”
“Why? Because I am going to marry Marie. Because, for Hanaré’s
sake, I’m going to forget you. I was a fool last night.”
“Last night you said you’d rather be a fool than a wise man. I
suppose you have forgotten that also.”
He cringed. “No,” he said, “I have not forgotten that.” He began
wandering about the room as though hunting for something, picking
up piles of papers, looking underneath the tables and chairs.
 “What are you hunting for?” she asked.
“My hat.”
“Do you want to go out?”
“I’m going out. You can do as you please. I’m going over to see
Marie.”
“Marie? At this time of night?”
He found his hat behind the sofa. He put it on, pulling the soft brim
far down over his eyes.
“Good-night,” he said casually, as he walked out of the room. He
was not fully conscious that there was anybody else there. He had
one determination. In view of that everything else was forgotten.
Vain resolve! The windows at 355 McDougal Street were brilliantly
lighted, policemen were in front of the door, people stared, and there
was general confusion. Some one had been killed. Paul learned,
finally, that Marie had committed suicide.
He felt, rather, that he had committed murder.
Thus, when he return to his own apartment and found the girl still
there, he was glad to forget Hanaré’s death that way instead.
RUSSELL WHEELER DAVENPORT.
 Portfolio
Melody

My lady sitteth on a shrine

And dreameth beauteously.
She dreameth much, her deep eyes shine
Like stars on a quiet sea.
And to watch her hands so soft and white
Is a never-ending, sweet delight.
Lady of Day, Lady of Night,
Queen of the World is she.
FRANK D. ASHBURN.

Inspiration
The smoking-room gave a terrific lurch. As if the motion had been a
signal, Carlos Bentley abruptly broke off his sentence, at the same
time removing his hand from the arm of his companion’s chair.
Although the big steamer recovered almost immediately from the
unexpected blow, Carlos continued to remain silent, his gaze
wandering uncertainly around the comfortable room. But he did not
notice particularly the brown sleekness of the leather chairs nor the
subtle masculinity of the lighting. He was wondering whether he had
not again let his tongue run away with his good taste in allowing it to
run on over the history of his past two weeks to this gentleman to
whom he had introduced himself. That was one of Carlos’ bête-noirs
—a cheerful frankness and lack of reserve that made him
communicate things he wished later he had kept to himself. But after
all the fellow had looked lonely and— A polite question which
interrupted his train of thought finished by driving the self-reproaches
from his mind. He answered the question at some length.
“Oh, yes! We spent six months in Paris. I got to know the place
quite well—well enough to get tired of it. I’m looking forward to New
York as a change. If it hadn’t been for my wife, I’d have come back
before, but she insisted on our staying—for my own good, she said.
You see, I went over to study art—portraits mainly. Spent hours
every day looking at pictures and trying to copy them.”
“Do you plan to take up art as a profession?” asked his
companion, knocking the ashes from his pipe. He was an elderly
man who had an air of demanding confidences with a view of solving
any difficulties connected with them from the depths of a thoughtful
urbanity.
Carlos hesitated a moment.
“Yes,” he said finally, “I expect to. That’s my ultimate aim. But, of
course, after all this studying I’ll want a bit of a rest—say a month or
so. Then I’ll be ready to get down to work.”
The other nodded a thoughtful assent. Then—
“You’ll pardon the remark, but—you have an income, I take it.”
Bentley nodded.
“Very fortunate, very fortunate indeed. So many poor devils have
to start with literally nothing but their talent. You’re unusually
blessed. Well, I must be getting to bed. We dock early to-morrow, I
believe. I’ve enjoyed talking to you immensely, and you’ll pardon my
leaving so abruptly, won’t you? Good-night.”
Carlos stood gazing after him a moment; then, turning away, went
off in the direction of his own stateroom. He had an uneasy feeling
that the man had not quite approved of him, although he was unable
to explain what he himself had said that could have given ground for
such an opinion.
When he got to his stateroom, he found a message that his wife
had left on his bureau before going to bed. It had come by wireless
that evening and was from his father. On opening it, he read:
 “Meet you at pier. Glad you are settling down to work at
last.
Dad.”

Carlos laughed softly. Just like his father to mention work, even in
a wireless. It occurred to him that everyone, ever since he was a
boy, had been wanting him to work. They had all told him what great
things they expected of his talent if he would only use it. His mother
had cherished a letter from a boyhood schoolmaster, which dwelt in
glowing terms on his artistic ability, while at the same time it decried
his indolence. His wife had refused many suitors as importunate and
more wealthy than he because she was in love with him, and
believed that her love could make him fight for the success which
was expected of him. Well, his father was right—it was time to start
work. They had had enough disappointments in him, and now he
must do something to make them proud of him. It wouldn’t be hard.
In an exceedingly virtuous mood Carlos bent over and kissed his
sleeping wife. What a wonderful girl Eloise was, and what a trump to
have believed in him enough to have married him. He would work as
he never had before as soon as they got settled in New York. With
which resolutions he got into bed to dream of painting portraits for
the kings and queens of Europe.

Four months later in a studio-apartment in the low Fifties a wet

paint-brush was hurled viciously at a small statue of the Laocoon. It
struck the largest figure full in the face with a comforting smack, and
clattered to the floor. Carlos Bentley had been trying to do a portrait.
Eloise, who in lieu of a regular model had been sitting for him,
started at the sound, then relaxed her pose. She was an appealing
figure with a touch of dynamic force in the aggressive tilt of her chin
that made Carlos, jokingly and yet half-seriously, call her his will-
power; at this moment she seemed to be bracing herself as if to
meet something.
“Why, Carlos dear, what is the matter?” she asked, approaching
her husband doubtfully.
 Carlos stood before a half-finished picture removing his painting
jacket, which he hurled into a corner before turning to his wife.
“I’m going to stop,” he said impatiently. “I don’t seem to feel in a
mood for it to-day somehow. Besides we’ve been working for quite a
while and we need a rest.” His eyes met hers half-defiantly, as if he
were expecting some remonstrance. Then he added, “Come on
down to a show, dear. We can do some more to-night on this.”
His wife turned away.
“I don’t care to go down, Carlos,” she answered slowly, “and I had
hoped you’d want to work this afternoon. We’ve only been up here a
little over an hour. Won’t you stay a little longer? You were just
beginning to get the right feeling in the picture. I know you were.”
Carlos laughed and kissed her.
“There’s plenty of time for the picture and it’s too wonderful an
afternoon to stay indoors. I’m going out for a walk. Sorry you won’t
come.” He slammed the door as he went out.
Eloise sat down dejectedly on a straight chair. Her lips trembled
until she could hardly keep from crying. For seven weeks this same
thing had happened continuously until she was sick to death of trying
to fight against it. Every day Carlos had alternated playing around
the city with attempts to work which always ended like to-day. In all
that time he had only finished one picture—but it had been good,
and had shown the talent that was being wasted. If only she knew
some way to touch the spark to that talent. Eloise found herself
wondering whether perhaps she had not undertaken a task too
difficult even for her love. It seemed as if Carlos utterly lacked the
requisite energy to produce what he was capable of. With a sigh she
turned to putting the studio in order.
Meanwhile Carlos, after wandering out onto the street, had set off
in the direction of the park. The refreshing air of a sunny autumn
afternoon soon cleared his brain, but there was still an uneasy
feeling in the back of his mind. He felt that he ought to be working,
yet was unable to, and he knew vaguely that he was not happy even
in the freedom of the moment. In this contradictory frame of mind he