Quantitative Analysis and Modeling of

Earth and Environmental Data:

Space-Time and Spacetime Data
Considerations 1st Edition Jiaping Wu
Visit to download the full and correct content document:
https://ebookmass.com/product/quantitative-analysis-and-modeling-of-earth-and-envir
onmental-data-space-time-and-spacetime-data-considerations-1st-edition-jiaping-wu/
More products digital (pdf, epub, mobi) instant
download maybe you interests ...

Predictive Modeling in Biomedical Data Mining and

Analysis Sudipta Roy

https://ebookmass.com/product/predictive-modeling-in-biomedical-
data-mining-and-analysis-sudipta-roy/

Applied Data Analysis and Modeling for Energy Engineers

and Scientists

https://ebookmass.com/product/applied-data-analysis-and-modeling-
for-energy-engineers-and-scientists/

NoSQL and SQL data modeling: bringing together data,

semantics, and software Hills

https://ebookmass.com/product/nosql-and-sql-data-modeling-
bringing-together-data-semantics-and-software-hills/

Volume III: Data Storage, Data Processing and Data

Analysis 1st ed. 2021 Edition Volker Liermann (Editor)

https://ebookmass.com/product/volume-iii-data-storage-data-
processing-and-data-analysis-1st-ed-2021-edition-volker-liermann-
editor/
Big Data Analysis of Nanoscience Bibliometrics, Patent,
and Funding Data (2000-2019) 1st Edition Yuliang Zhao

https://ebookmass.com/product/big-data-analysis-of-nanoscience-
bibliometrics-patent-and-funding-data-2000-2019-1st-edition-
yuliang-zhao/

Data Modeling and Database Design 2nd Edition, (Ebook

PDF)

https://ebookmass.com/product/data-modeling-and-database-
design-2nd-edition-ebook-pdf/

Numerical Methods in Environmental Data Analysis Moses

Eterigho Emetere

https://ebookmass.com/product/numerical-methods-in-environmental-
data-analysis-moses-eterigho-emetere/

Data Treatment in Environmental Sciences 1st Edition

Edition Valérie David (Auth.)

https://ebookmass.com/product/data-treatment-in-environmental-
sciences-1st-edition-edition-valerie-david-auth/

Data Wrangling on AWS: Clean and organize complex data

for analysis Shukla

https://ebookmass.com/product/data-wrangling-on-aws-clean-and-
organize-complex-data-for-analysis-shukla/
QUANTITATIVE ANALYSIS AND MODELING
OF EARTH AND ENVIRONMENTAL DATA
This page intentionally left blank
 QUANTITATIVE
ANALYSIS AND
MODELING OF EARTH
AND ENVIRONMENTAL
DATA
Space-Time and Spacetime Data
Considerations

JIAPING WU
Zhejiang University, China

JUNYU HE
Zhejiang University, China

GEORGE CHRISTAKOS
Zhejiang University, China
San Diego State University, United States
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
Copyright © 2022 Elsevier Inc. 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: 978-0-12-816341-2

For information on all Elsevier publications

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

Publisher: Charlotte Cockle

Acquisitions Editor: Amy Shapiro
Editorial Project Manager: Lindsay Lawrence
Production Project Manager: Kumar Anbazhagan
Cover Designer: Victoria Pearson

Typeset by STRAIVE, India

They were good days, there have been good days.
Jiaping Wu, Junyu He, and George Christakos
This page intentionally left blank
 Contents

Preface ix 4 Chrono-geographic clustering and hotspot

(coldspot) analysis 136
5 Practice exercises 145
1. Chronotopologic data analysis
5. Classical geostatistics
1 From topos to chronotopos 1
2 Chronotopologic variability, dependency 1 Historical introduction 149
and uncertainty 9 2 Random field theory 155
3 Theory and evidence 18 3 Covariography and variography 165
4 Chronotopologic estimation and mapping 22 4 Chronotopologic block data analysis 200
5 A review of CTDA techniques 24 5 Practice exercises 206
6 Chronotopologic visualization technology 26
7 The range of CTDA applications 27
8 Public domain software libraries 28 6. Modern geostatistics
9 Practice exercises 30
1 Toward a theory-driven CTDA 213
2 Knowledge bases revisited 216
2. Chronotopology theory 3 Integrating lawful and dataful statistics 229
4 Rethinking chronotopologic dependence 243
1 Introduction 33
5 Applications 253
2 Basic chronotopologic notions 35
6 Practice exercises 261
3 Chronotopologic metric modeling 47
4 Metric effects on chronotopologic attribute
interpolation 52 7. Chronotopologic interpolation
5 Practice exercises 55
1 Introduction 267
3. CTDA methodology 2 Deterministic chronotopologic interpolation
techniques 273
1 Methodologic chain 57 3 Statistical chronotopologic interpolation
2 About knowledge 66 techniques 282
3 Big data: Why learn, if you can look it up? 81 4 Practice exercises 291
4 Attribute data scales 88
5 Emergence of chronotopology-dependent 8. Chronotopologic krigology
statistics 92
6 More on chronotopologic visualization 96 1 The emergence of geostatistical Kriging 293
7 Practice exercises 98 2 1st Kriging classification 299
3 Second Kriging classification: point,
4. Chrono-geographic statistics chronoblock and functional 320
4 Mapping accuracy indicators and
1 Introduction 101 cross-validation tests 323
2 CGS of data point information 102 5 Applied krigology: benefits and concerns 339
3 CGS of chrono-geographic attribute values 121 6 Practice exercises 341

vii
viii Contents

9. Chronotopologic BME estimation 12. DIA models

1 Epistemic underpinnings 345 1 Introduction 431
2 Mathematical developments 346 2 Machine learning 433
3 An overview of real world BME case studies 356 3 Linear regression techniques 434
4 Practice exercises 377 4 Artificial neural network 438
5 Practice exercises 444
10. Studying physical laws
13. Syntheses of CTDA techniques
1 The important role of physical PDE in CTDA 385 with DIA models
2 BME solution of a physical law 389
3 BME solution of an epidemic law 397 1 A broad synthesis perspective 449
4 Comparing core and specificatory 2 A synthesis of the STP and BME techniques 452
probabilities 402 3 A synthesis of the STP-BME technique with
5 Practice exercises 405 the LUR and ANN models 459
4 A synthesis of the BME technique with the
MLR and GWR models 464
11. CTDA by dimensionality reduction 5 Epilogue 471
1 The motivation 407 6 Practice exercises 472
2 The space-time projection (STP) method 408
3 Noteworthy STP features 427 References 477
4 Practice exercises 428 Index 485
 Preface
As its title dictates, the subject of this book is
the chronotopologic analysis and modeling of
natural phenomena, that is, phenomena that
vary as functions of both their spatial (topos)a
and the temporal (chronos)b coordinates, in real-
istic conditions of in situ uncertainty. The need to
study such phenomena, in particular, has led to
significant developments in chronotopologic
data analysis (CTDA), which is an important
part of the book.
The notions, techniques, and thinking modes
discussed in the book aim at improving the
understanding of the chronotopologic laws of
change underlying the available numerical data-
sets, while taking into consideration all relevant
core and site-specific knowledge bases, which
are subject to multi-sourced uncertainties and
associated measurement or observational errors
(conceptual, technical, computational). The core
knowledge bases include scientific theories,
physical laws, and models, whereas site-specific
knowledge bases may have various forms and
sources, including hard measurements, soft
observations, secondary information, and auxil-
iary variables (ground-level measurements, sat-
ellite observations, scientific instrument records,
protocols and surveys, empirical graphs and
charts). Understanding the spatial distribution
and temporal dynamics of knowledge bases
such as the above is a great challenge to the
elucidation of crucial questions in many physi-
cal, health, and social disciplines, including
geology, hydrology, geophysics, geography,
a
The term “topos” (see topography, topology etc.) was the ancient Greek term for space.
b
The term “chronos” (see chronology etc.) was the ancient Greek term for time.
ix
environmental science, agronomy, ecology,
public health, epidemiology, economics, public
policy, and risk management.
The book views the difference between train-
ing and education in terms of the following met-
aphor: The way to get people build a ship is not
to teach them carpentry, but to inspire them to
long for the infinite immensity of the sea.
Accordingly, the book’s methodology has four
stages: The first stage is diagnosis (knowing
through), that is, identifying the basic features
of the phenomenon of interest, the second stage
is cardiognosis (knowing the heart of the phe-
nomenon), that is, appreciating what is at stake
and choosing the appropriate mode of thinking,
the third stage is prognosis (foreknowledge), that
is, accurate prediction of certain important
aspects of the phenomenon, and the fourth stage
is epignosis (improved knowledge), that is, draw-
ing important conclusions, inferences, and
interpretations.
This methodology has some interesting
implications that become evident throughout
the book, including the following: new routes
open when one type of data crosses another;
potentially difficult problems are considered
from alternative standpoints, assessing how
well different methods can handle them, and
how an improved problem solution is obtained
by a synthesis of methods acting in synergy; the
presentation of worth noticing personal conclu-
sions based on experience and insight is
facilitated.
x Preface
To achieve its particular goals, the book’s
focus is the rigorous presentation of the support-
ing theory, followed by a comprehensive and
balanced fusion of theory-driven and data-
driven techniques. Accordingly, the assertions
presented in the book have theory-based content
(the expression of a proposition about the real
phenomenon being in a certain way) and are
made in a data-based context (including the
available evidential support and real world
conditions). In melioribus annis, this approach
would have seemed natural. In praesenti annis,
however, it is necessary to stress the importance
of this approach, since the excessive use of
technological “black-box” nowadays runs the
risk to eventually create impoverished human
“black-boxes.” Several numerical examples
and actual case studies are included so that
the readers gain a hands-on experience and
valuable insight concerning the implementation
of the presented notions, models, and tech-
niques in the real world. Practice exercises at
the end of each chapter can help the readers
learn more from the text and hone their critical
and practical skills.
Individual topics addressed in the book may
be probably found in separate publications, but,
as far as we know, there is not any particular
book that covers all these topics in a complete,
systematic, and integrative manner as the pre-
sent book does. Arguably, then, readers with sci-
entific background or engineering training in
the aforementioned disciplines would appreci-
ate a systematic presentation in one volume of
the most important quantitative concepts,
models and techniques studying the chronoto-
pologic behavior (combined spatial distribution
and temporal dynamics) of natural phenomena
under conditions of uncertainty and site-specific
measurement errors (i.e., the vast majority of
phenomena encountered in the various scien-
tific and engineering disciplines).
We would like to thank our families for their
patience, and Ms Lindsay C. Lawrence of Else-
vier for her continuing encouragement during
the book-writing project.
Jiaping Wu
Junyu He
George Christakos
 C H A P T E R

1
Chronotopologic data analysis

1 From topos to chronotopos A scientific paradigm is a framework of con-

The first section of the book starts with a
broad discussion of developments in the collec-
tion, analysis, and interpretation of data, from
the case where the data values are assumed to
vary as functions of the spatial (topos)a coordi-
nates to the case in which they are assumed to
vary as functions of both the spatial (topos)
and the temporal (chronos)b coordinates. The
latter case leads to the unified notion of chron-
otopology, where the data values may vary in a
separate (space-time) or a composite (spacetime)
manner.
1.1 Scientific paradigms
But, before focusing on chronotopology mat-
ters, and in order to put these matters in due per-
spective, the discussion should start with a
reference, albeit a brief one, to a fundamental
notion of human inquiry, as follows:
cepts, assumptions, theories, thinking modes,
results, and practices that define a scientific disci-
pline at any particular period of time.c
Historically, the evolution of the scientific
paradigm has undergone certain major develop-
mental phases:
① It started, a few thousand years ago, with the
purely empirical paradigm that focused on
the description of natural phenomena
using purely empirical means.
② This was followed, a few hundred years ago, by
the theory-laden experimentation paradigm
when a theoretical component (involving
theories, models, laws, generalizations) was
added to the empirical component (in which
systematic experimentation played a key
role).
③ During the last few decades, the
computation-dominated paradigm

a
The term “topos” (see topography, topology, etc.) was the ancient Greek term for space.
b
The term “chronos” (see chronology, etc.) was the ancient Greek term for time.
c
The word παρ
αδειγμα (paradeigma) has been used in famous texts, such as Plato’s Timaeus, as the model or the pattern
that God used to create the cosmos.

Quantitative Analysis and Modeling of Earth and Environmental Data 1 Copyright # 2022 Elsevier Inc. All rights reserved.
https://doi.org/10.1016/B978-0-12-816341-2.00005-8
2 1. Chronotopologic data analysis

emerged that is characterized by the This is a serious matter to anyone who takes a
addition of a significant computational moment to examine it with an attentive mind.
component (since theories of complex Therefore, we will revisit it in various parts of
phenomena became too complicated to the book.
solve analytically, numerical simulations
needed to be generated).
1.2 Interplay of science and mathematics
Furthermore, it has been argued by many
No readers of this book, most of whom are
investigators (scientists and nonscientists alike)
presumably quants (i.e., quantitative analysts),
that we currently enter the era of yet another
need to be reminded of the all-important rela-
paradigm:
tionship between two key components of
④ The big data-driven paradigm that seeks to human inquiry (see Fig. 1.1):
apply computational models to breathtaking
amounts of data obtained by instruments or
generated by simulators, then processed by Mathematics, which focuses on how to develop
the software, resulting in information stored abstract representations of the real world, and
in computers. science, which is concerned with the inverse
process, i.e., with how to use these abstract repre-
According to this paradigm, insight is gained sentations to obtain useful knowledge about the
through a self-reinforcing loop between experi- real world.d
mental data and statistical analysis (Succi and
Coveney, 2019). As should be expected,
serious objections to the purely data-driven Admittedly, this book uses a considerable
paradigm exist. Among them, Lewis H. Lapham amount of mathematics in the form of analysis,
has suggested that: modeling, and estimatione notions and tech-
@ Data mining engineers have no use for the niques. Compared to the complexity of the real
meaning and value of words. They come to
bury civilization, not to praise it.
In response to these and similar concerns, a
challenge to the purely data-driven perspective
has emerged, as follows:
• The big data perspective outlined above
makes extravagant claims, which, once
properly discarded, a synergistic synthesis
of large datasets with a sound scientific
theory could plausibly lead to a sensical FIG. 1.1 The opposite directions of mathematics and
paradigm-shift. science.

d
Yet, the two have some interesting sociological differences, including the fact that in science each generation of
investigators undoes the existing structure (replaces the theories of the previous generation with new ones etc.),
whereas in mathematics each generation adds a new piece to the existing structure.
e
The term “estimation” generally includes interpolation, extrapolation and prediction. The distinction between these
three kinds of chronotopologic estimation is critical: in interpolation and extrapolation attribute estimates are sought at
points, respectively, within and outside the chronotopologic sampling domain, and in prediction attribute estimates
are sought at points within the spatial sampling area but outside the temporal sampling period. The issue is discussed
in various parts of the book (e.g., Chapter 7).
 1 From topos to chronotopos
world, these notions and techniques are not as
complex as some investigators may seem to
believe. Those who do not believe so, it is rather
due to the fact that they may not realize how
complicated real life actually is.
The links of mathematics with ordinary life
are rather obvious. Many mathematical ideas
are ways of mathematizing ordinary ideas,
e.g., the idea of a derivative that is a straightfor-
ward mathematization of the ordinary idea of
instantaneous change. Beyond this rather simple
notion, many complex mathematical concepts
eventually come down to direct perceptual
experience.
Since science, a natura eius, is about finding
the analogy and identity in the most remote
parts, it should be noticed at the outset that this
book’s own interpretation of Fig. 1.1 is as
follows:
The interworkings of science and mathemat-
ics should be carefully considered in both direc-
tions: not only mathematics should be used in
scientific investigations, but science should be
used in mathematical investigations too.
Accordingly, investigators should closely
fuse mathematical symbology with physical
meaning, resulting in a powerful and productive
structure. That is, scientific applications require
us to integrate our understanding of the real
world with symbolic relations of mathematics,
thus adding meaning and structure to both. Sci-
ence puts meaning to mathematics, adding
additional levels of structure, interpretation,
and even tools. This book, then, is more inter-
ested in the scientifically meaningful employ-
ment of mathematical tools than in the purely
formal features of the tools. A direct consequence
of the above considerations is the warning that
looking at an equation and treating it purely as
a matter of mathematical grammar, while
neglecting to enrich its symbols with physical
meaning, leaves one vulnerable to errors.
3
Example 1.1 Assume one is given the equation
E ¼ Fq, relating the electric field E with force F and a
q
point charge q. If q is divided by 2, i.e., 2, what is the
corresponding E value? Focusing on the mathe-
matical grammar of the equation, the obvious
answer is E ¼ 2F q . This answer is physically incor-
rect, though, since it neglects the essential knowl-
edge that F is proportional to q, in which case the
correct answer is that there is no change in the
value of E. This error happens when one (who
probably lacks the necessary background knowl-
edge) looks at the equation and treats it purely as a
matter of mathematical grammar, neglecting to
enrich its symbols with physical meaning.
Since this book’s perspective is that an equa-
tion’s mathematical grammar should be inte-
grated with physical meaning, our focus will
not be only on technical deductions, but also
on the rich variety of inductive inferences asso-
ciated with our readers’ ordinary experience.
The same argument is obviously valid in terms
of data, i.e., data should not be treated purely
as numbers, neglecting to enrich it with the
due physical meaning. In this respect, the fol-
lowing general conclusion is reached:
• Readers are encouraged to avoid falling into
the trap of reasoning purely in terms of
mathematical grammar and numerics, thus
missing clues to a different interpretation
given by physical meaning.
The payoff of this approach, integrating
mathematical grammar and numerics with
physical meaning, is comprehensible, since it
can produce new knowledge, add to the existing
body of knowledge in a particular scientific
field, and, in many cases, scientific findings
can be transferred into useful technology.
1.3 Natural attributes
Quantitative analysis, as practiced in sci-
ences, relies on the tools of the discipline of
applied mathematics, which studies the varia-
tional characteristics of empirical observations
4 1. Chronotopologic data analysis
or measurementsf associated with natural attri-
butes. Before we proceed, we need a working
definition of the latter:
A natural attribute is a measurable or observ-
able entity characterizing selected aspects of a
natural phenomenon (physical, biological, social)
that vary over a region during a time period.
Simply put, everything that varies across
space and/or time in nature and can be mea-
sured (using simple devices or sophisticated
equipment of various kinds) or observed (via
perceptual experience) can be considered a nat-
ural attribute. In this setting, known (measured
or observed) attribute values are characterized
as attribute data.
Example 1.2 There is a wide variety of such
natural attributes, including pollutant concen-
tration, topographic elevation, ocean surface
temperature, rainfall intensity, human expo-
sure, disease incidence, population mortality,
land-use variables, commodity prices, and
regional poverty indicators.
In cases such as the above, attribute character-
istics of prime importance that need to be ade-
quately defined and understood are the
following:
• The variational characteristics describe
attribute dynamics,g regional changes, and
interrelationships due to the underlying
mechanisms of the phenomenon under
study (Section 2).
Indeed, data variation across space and time is
a dominant characteristic that has tremendous
impacts on the performance of certain quantita-
tive analysis and modeling aspects. For example,
the accuracy of attribute interpolation at
f
The distinction between measurement and observation is discussed later in the book.
g
The term dynamics here refers to the temporal change of the attribute.
unsampled points (discussed in Section 4) may
decrease considerably with increasing data vari-
ation. Accordingly, understanding the attribute’s
variational characteristics based on the available
data is a great challenge to the elucidation of cen-
tral questions in many physical, social, and health
disciplines, including environmentology, geol-
ogy, agronomy, ecology, geography, public
health, epidemiology, economics, social policy,
and risk management.
1.4 Kinds of scientific data analysis
As regards the practice of data analysis in sci-
ences, where there is a strong physical depen-
dence of data values on location and time,
three kinds of studies have historically emerged
(in ascending order of modeling sophistication):
❶ Studies concentrating solely on the variation
of data across space, known as spatial data
analysis (SDA), or solely on temporal data
variation, known as time data analysis (TDA).
❷ Studies focusing on some kind of separable
space-time data analysis (hereafter denoted
as S-TDA), in the sense that they consider
space and time in isolation.
❸ Studies favoring a composite spacetime
data analysis (denoted as STDA), in the
sense that space and time are considered
as an integrated whole, meaning that the
integration obeys the physics of the
phenomenon.
In the above settings, space and time individ-
ually become three- and one-dimensional
projections of the four-dimensional domain,
respectively. Conversely, the classical three-
dimensional geometry of SDA becomes a
four-dimensional geometry in space-time analy-
sis. Due to its wide popularity, let us comment a
little further on SDA. Although it may be seen,
 1 From topos to chronotopos
methodologically, as a special case of S-TDA,
there are considerable differences, as will become
clear in the following. Moreover, SDA was
historically developed many decades before
S-TDA. Hence, one can find a plethora of SDA
techniques—including inverse distance, splines,
trend surface, statistical regression, and geostatis-
tics Kriging (Chapters 7 and 8)—that have been
successfully implemented in many disciplines.
1.4.1 Geostatistics and time-series statistics
SDA techniques did not develop uno ictu tem-
poris. A case in point is the following well-
known SDA field:
Geostatistics consists of a collection of concepts
and techniques that study spatially distributed
attributes based on a space-dependent statistical
theory that was built on pre-existing results formu-
lated in different scientific disciplines with varying
objectives in mind.
Indeed, the roots of geostatistics go back to
the 1940s–60s as a group of spatial correlation,
regression, and mapping techniques used in
forestry and earth sciences with considerable
success, see the pioneering work of Mat
ern
(1947) and Matheron (1965). Remarkably, sev-
eral decades later the field was rediscovered
and renamed spatial statistics.h In any case, many
important advances have been made since the
1960s, and the range of geostatistics or spatial
statistics applications nowadays is vast (the
interested readers are referred to the relevant
literature).
TDA, on the other hand, generally focuses on
the study of sequences of attribute data ordered
in time (time is often considered the indepen-
dent variable). In particular, the systematic sta-
tistical study of this kind of time data was the
focus of a well-known TDA field:
h
A well-known spatial statistics book is that of Noel Cressie (Statistics for Spatial Data. J Wiley, NY, 1991).
5
Time series analysis (TSA) is a collection of
concepts and techniques for studying the statisti-
cal characteristics of time series data and making
attribute forecasts for future times.
Historically, the earliest theoretical TSA
developments can be traced back to the pio-
neering work of Yule (1927) and Walker
(1931). The next major development was the
work of Box and Jenkins (1976) presenting the
complete time series modeling process (specifi-
cation, estimation, diagnostics, and forecast-
ing). One can find several TSA models in the
relevant literature, including moving averages,
exponential smoothing of various degrees,
analysis of variance, autoregressive integrated
moving average, among many other models.
There have been numerous TSA applications in
almost any scientific discipline, including econo-
metrics, finance, geophysics, seismology, signal
processing, pattern recognition, and meteorol-
ogy. The interested readers are referred to the
very rich literature on the subject.
Example 1.3 Both SDA and TSA capitalize on
spatial and temporal correlations, respectively.
To most scientists, the fact that closely spaced
or temporally separated samples tend to be sim-
ilar is not surprising, since such samples are
likely to be influenced by similar physical
processes.
An obvious disadvantage of SDA and TSA is
that they both neglect any cross space-time data
associations, and the reason of this neglect may
be a physical justification, an apparently plausi-
ble approximation, or simply a matter of conve-
nience. Yet, reality often does not strictly satisfy
the SDA or TSA assumptions. Instead, strong
correlations may occur simultaneously across
space and time:
6 1. Chronotopologic data analysis
• In reality, the links between attribute data
values can be critically affected by space-
time cross-effects,i and the corresponding
data distributions can be highly skewed.
In a human sense, after all, to represent mat-
ters in a purely spatial manner is to freeze time
and to deny the dynamism implicit in living.
A similar criticism holds when representing
matters in a purely temporal manner that denies
spatial dependency. By way of a summary, then,
the following conclusion is drawn based on the
above considerations:
• Both SDA and TSA encounter profound
challenges to their methodologies that have
been based on notions and models that are,
respectively, timeless-space (SDA) and
space-independent (TSA).
It is reasonably anticipated that the adequate
responses to these challenges could lie in richer
approaches that account for both space and time
in a systematic way. Such approaches are intro-
duced in the remaining of this chapter and they
are, in fact, the subject matter of this book.
1.4.2 S-TDA and STDA
For the above reasons, the advent of S-TDA
has been seen as a significant development over
SDA and TSA because it offers a more accurate
quantitative representation of the real world,
including a more realistic study of the space-
time distribution of data from natural attributes.
Indeed, many useful S-TDA models have been
developed with emphasis on the quantitative
expression of attribute properties and relation-
ships that take into account the space-time local-
ization of these properties and relationships in a
direct way. The notions and methods discussed
in the book address these and similar issues.
i
This means that, due to space-time cross effects, data values that are further apart in space but closer in time may turn
out to be more similar than those that are closer in space but further apart in time, and vice versa.
Example 1.4 Space-time adaptive processing
extends adaptive antenna techniques to proces-
sors that simultaneously consider spatial
domain information (in the form of signals
received on multiple antenna array elements)
and temporal domain information (in the form
of multiple pulse repetition periods) of a coher-
ent processing interval.
Although S-TDA is very useful in many prac-
tical applications, its key limitation is rather pro-
found: S-TDA essentially suggests seeing the
world in slices. I.e., the S-TDA models, by defi-
nition, consider space and time as separate argu-
ments. In layman terms, then, the need for the
composite spacetime perspective proposed by
STDA is based on the following rather straight-
forward argument:
In many cases, it is hard to pin down the truth
when one is obliged to see the world in slices,
because snapshots may conceal as much as they
make plain.
Otherwise said, STDA also studies the distri-
bution of natural attributes, but, while S-TDA
views space and time as separate arguments,
STDA considers a composite spacetime argu-
ment. Classical statistics assumes nonspatiality
and that the samples are independent from
one another, whereas S-TDA and STDA account
for spatiality and temporality, although in dif-
ferent ways. STDA is obviously a more recent
development than S-TDA.
Undoubtedly, S-TDA and STDA offer
improved representations of reality, which is
why one can find a considerable number of
interesting real-world publications of S-TDA
and STDA in a variety of scientific journals
representing different and often unconnected
 1 From topos to chronotopos
disciplines. Yet, there is a very limited number
of books on S-TDA, and even fewer on STDA
theory and applications.j This is somewhat sur-
prising, since there is a great potential and need
for comprehensive and systematic presentations
of S-TDA and STDA in practice, due to the
obvious fact that the vast majority of real data
vary across both space and time. Moreover,
improved technology is becoming available that
facilitates the in-situ implementation of S-TDA
and STDA, e.g., low-cost Temporal Geographic
Information Systems, TGIS, with user-friendly
interfaces (Christakos et al., 2002).
1.5 The unifying chronotopologic data
analysis notion
It is now time to return to the notion of
chronotopology briefly introduced at the begin-
ning of this chapter. It should be brought to the
reader’s attention that, for methodological and
presentation reasons, in this book a unifying
notion will be favored, on occasion, depending
on the context (e.g., when the properties of
interest are shared by both S-TDA and STDA).
It is methodologically appropriate and occa-
sionally convenient to use the general term chron-
otopologick data analysis (CTDA) that includes
both cases of S-TDA and STDA.
Generally, CTDA aims at extracting implicit
knowledge such as chronotopologic relations
and patterns that may not be explicitly stored
in the relevant databases. This extraction relies
on the merging of scientific knowledge and log-
ical principles. CTDA distinguishes itself from
classical data analysis in that it associates with
j
Spatiotemporal data analysis books include Chistakos G (Spatiotemporal Random Fields, Elsevier, the Netherlands,
2017; Modern Spatiotemporal Geostatistics. Oxford UP, NY, 2000); and Christakos G and Hristopulos DT (Spatiotemporal
Environmental Health Modeling, Kluwer Acad, MA, 1998).
k
The term “chronotopologic” includes the cases of space-time and spacetime data variability.
7
each attribute of interest physically meaningful
chronotopologic features. In particular, while
classical statistics assumes that the samples are
independent from one another, CTDA considers
sample dependence governed by physics, it is
not tied to a distribution model that assumes
that all samples of a population are normally
distributed, and it is devoted to the interpreta-
tion of uncertainties caused by various in-situ
conditions (Section 2 below).
Due to the increasing availability of datasets
that are simultaneously location and time
dependent, their systematic quantitative analy-
sis and modeling provided by CTDA are of
great importance to a variety of scientific and
engineering disciplines, including earth, ocean
and atmospheric sciences, environmental and
ecological engineering, health and epidemio-
logical studies, risk assessment, social policy
and financial management.
1.5.1 CTDA conditions
As a matter of fact, it should be emphasized at
the outset that CTDA can be used in the study of
natural attributes that satisfy these conditions:
① They are measurable or observable.
② They vary within a well-defined chrono-
topologic domain.
③ They are characterized by chronotopologic
dependence.
④ They occur in conditions of in-situ
uncertainty.
In other words, the CTDA concepts and tech-
niques should be used in the study of natural
phenomena when meaningful values of the attri-
bute may occur at every point in the chronotopo-
logic domain of interest that can be potentially
measured or observed.
8 1. Chronotopologic data analysis
Example 1.5 While CTDA concepts and tech-
niques can be implemented to study a large
variety of natural attributes (Example 1.2 ear-
lier), they may not be necessarily useful when
points represent merely the presence of events
(e.g., earthquake occurrence), people, or some
physical object (e.g., volcanoes, buildings).
CTDA methods are based on a rigorous prob-
lem definition, which is then used to formalize a
CTDA protocol in order to answer the questions
of interest. In this respect, CTDA is a systematic
inquiry to produce new knowledge, refine or
validate the existing knowledge regarding the
natural attributes considered, and generate attri-
bute estimates.
• The interdisciplinary mantra of CTDA is to
not work in silos and to value that which is
cross-cutting.
In this setting, CTDA surely benefits from
developments in TGIS technology. This tech-
nology allows the chronotopologic visualiza-
tion of a variety of attributes such as
individual populations, quality of life indices,
pollutant distribution, disease spread, and
company sales over time in the region of inter-
est. To achieve that, it is enough to have a
chronotopologic database, and the TGIS is
capable of presenting an animated set of col-
ored maps that provide informative visualiza-
tions of the chronotopologic pattern of the
attributes of interest.
Example 1.6 CTDA applications currently
span many disciplines, with their methods vary-
ing in relation to the specific questions being
addressed, whether predicting air pollution,
examining suspiciously high frequencies of dis-
ease events, or handling the vast data volumes
being generated by the Global Positioning System
(GPS) and Satellite Remote Sensing (SRS).
1.5.2 Generic CTDA notions
CTDA is a broad framework for studying natu-
ral attribute distributions, linking attribute coordi-
nates, in-situ conditions and values. Considerable
information exists individually on each of these
items, which must be brought together to assure
consistency in concepts and methods. CTDA pro-
vides some initial principles and notions, as
follows.
• The chronotopologic unit is the basic
building block for the CTDA of space- and
time-specific natural attributes.
The choice of the chronotopologic unit should
be based on a relevant set of criteria (physical,
ecological, social, etc.) and it may affect the
CTDA results.
Example 1.7 A chronotopologic unit may be
associated with observable characteristics of ter-
restrial, marine and atmospheric domains. It
may correspond to the smallest or elementary
working unit that is representative of these
characteristics.
Information on natural attributes may be
available in several different scales (say, city,
county and state), and an adequate CTDA may
require conveying information from one scale
to another (see, also, Section 4 of Chapter 3). This
also includes the methods of transferring infor-
mation from one chronotopologic point to
another.
• Chronotopologic scaling is the process of
conveying information from one chrono-
topologic scale to another.
Scaling may take different forms: Down-
scaling (conveying information from larger to
smaller chronotopologic domains), upscaling
(conveying information from smaller to larger
domains), and transferring (conveying informa-
tion from one point to another).
Example 1.8 A general existing ecosystem
data analysis framework operates at the national
level, whereas a site-specific study may require a
much finer chronotopologic scale. Investigators
need to understand when and how to convey
information from the point (e.g., pollution mon-
itoring) to the chronoregion (e.g., city popula-
tion exposure) scale.
 2 Chronotopologic variability, dependency and uncertainty
Adequate information gathering is important
in understanding CTDA results as well as for
communication purposes.
• Chronotopologic aggregation is the process
of reducing certain measures characterizing
the chronotopologic attribute distribution
into simpler ones.
Aggregation may refer to chronotopologic units
or the corresponding natural attribute values. Ade-
quate aggregation relies on a satisfactory under-
standing of the underlying mechanisms, inter-
relationships, and their relative importance. These
measures may be the same across different scales,
or they may differ between scales, in which case
the appropriate links between them need to be
established. Also, different CTDA models and
techniques may have their own chronotopologic
units, scaling notions, and aggregation tools.
Example 1.9 Aggregation may consist in com-
bining several data variability measures into a
simpler set of measures. Aggregation may take
the form of indicators, indices, summary statis-
tics, and graphs. Disparate measures may be
aggregated using a reference state or a common
factor, such as unit of measurement.
1.5.3 Fourfold CTDA objectives
In light of the above deliberations, we chose
to conclude this first section of the book by mak-
ing some suggestions regarding the main objec-
tives of an adequate CTDA study.
Whatever the specifics of the real-world case
under investigation, the ultimate judgment of
progress in CTDA should be measurable results
obtained during a reasonable time based on bet-
ter science and more powerful technology.
The adequacy of the above fourfold objective
will become apparent throughout the book with
varying emphasis on its individual objective,
depending on the context. After all, it should
9
be kept in mind that this is an instructional book
that relies on sound theory, scientific reasoning,
and substantive knowledge, rather than on mere
data massaging techniques and popular quick
fixes in challenging times.
2 Chronotopologic variability,
dependency and uncertainty
As CTDA techniques become powerful tools
in many fields (earth sciences, oceanography,
geography, natural resource management,
biological conservation, and societal processes),
adequate quantitative characterizations of good
quality chronotopologic datasets of natural attri-
butes are increasingly needed.
2.1 Conceptual chronotopology
Accordingly, some further comments con-
cerning the notions of chronotopology and its
geometric dynamics are in order at this point.
We start with the following two key notions:
Chronotopologic variability refers to the
degree of the attribute’s joint change (fluctuation)
across space and time, whereas chronotopologic
dependency refers to the degree of the attribute’s
joint smoothness (connectivity) across space
and time.
That is, although both notions are concerned
with differences between attribute values at dif-
ferent points, there is an inverse relationship
between these two notions. In the real world,
chronotopologic variability or dependency takes
up particular forms that are functions of certain
factors, like the nature of the phenomenon and
the chosen representation of its chronotopologic
domain. The attribute data used in a study may
come from diverse, and sometimes unexpected,
sources set apart by their distinct variability
(dependency) features.
Another random document with
no related content on Scribd:
Archive Foundation, how to help produce our new eBooks, and how
to subscribe to our email newsletter to hear about new eBooks.