Quantitative Analysis and Modeling of

Earth and Environmental Data:

Space-Time and Spacetime Data
Considerations - eBook PDF
Visit to download the full and correct content document:
https://ebooksecure.com/download/quantitative-analysis-and-modeling-of-earth-and-e
nvironmental-data-space-time-and-spacetime-data-considerations-ebook-pdf/
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.
10 1. Chronotopologic data analysis
Example 2.1 When studying air temperature
data all over the globe during 150 years in order
to describe global climate change (e.g., Jones
et al., 1986), one could calculate temperature
averages at a given moment (say, day), but these
numbers would not take into account the natural
chronotopologic variability of these data (at any
given moment, the range of temperatures
around the world is huge due to different local
climates, seasons, topography, whereas no tem-
perature data exist over large portions of the
oceans). In addition to the standard temperature
datasets, recent studies of global warming relied
on many other data sources, including data on
the timing of the first appearance of tree buds
in spring, greenhouse gas concentrations in the
atmosphere, and measurements of isotopes of
oxygen and hydrogen from ice cores (Egger
and Carpi, 2011). The study of such diverse
and multisourced datasets is essentially
interdisciplinary, i.e., it requires the integrated
efforts of scientists with a variety of backgrounds
and expertise, see the report of the Intergovern-
mental Panel on Climate Change (IPCC, 2007).
The profound inverse relationship that exists
between the notion of variability and that of
dependency is described by the following
relationship:
• Complementarity relationship: the presence of
higher (lower) attribute variability in a
chronotopologic domain would mean weaker
(stronger) attribute dependency in the domain.
For reasons that will become evident later,
complementarity assumes a key role in CTDA
practice. Another essential property of chrono-
topology is as follows:
• Dataset continuity: in principle, data can be
considered at any point of the chrono-
topologic domain.
It should be noticed that data continuity in the
above sense refers to the attribute chronotopol-
ogy rather than to the attribute value per se. In
the latter case, attribute value continuity means
that the attribute (e.g., chemical concentrations)
can assume any value within a specified range
to be distinguished from discrete attribute values
that can be counted (e.g., number of days an indi-
vidual has been exposed to a pollutant) and can
take a finite number of possible values. Although
chronotopologically continuous datasets play a
pivotal role in investigations in the fields of earth
monitoring, ocean research, environmental
risk assessment and planning, public health
management and decision-making, they are
usually not readily available and are often
difficult and/or expensive to acquire (e.g., in
mountainous and deep marine regions).
Example 2.2 The following real-world situa-
tions illustrate the significance of continuous
datasets and the need for CTDA techniques that
can effectively compensate for their lack.
• Environmental managers require chrono-
topologically continuous data over a region
to make effective and informed decisions
concerning the prevention and control of
severe pollution, and scientists need
accurate data that are well-distributed in a
chronotopologic domain to make justified
interpretations.
• The marine environment in Australia is a
typical case where seabed mapping, habitat
classification, and prediction of marine
biodiversity, essential for marine biodiversity
conservation, need reliable continuous data
of the marine environment. Yet, in most of
the Australian marine regions such data are
not available, and only sparsely and
unevenly scattered point samples have been
collected. Hence, chronotopologic
interpolation techniques (Section 4) are
essential for computing biophysical variables
at the unsampled locations.
• Botanists collect information that is discrete at
a particular set of sampling locations and
time instances, but has to be converted in
some way into useful dynamic map
representations. To achieve this goal one
 2 Chronotopologic variability, dependency and uncertainty
needs to manipulate attribute data in a way
that accurately estimates attribute values in
chronotopologic domains that have not
been sampled. In essence, one needs from
the information available to produce
continuous maps at unknown domains.
2.2 Quantitative chronotopology
At the center of CTDA is the quantitative rep-
resentation of a natural attribute that varies
jointly across space and time.
A natural attribute is represented as a mathe-
matical function X(p) of the point p varying in a
chronotopologic domain, where p consists of
spatial coordinates s ¼ (s1, …, sn), the number n
depending on the dimensionality of the region,l
and a temporal coordinate (time instance) t.m
A digression may be appropriate at this point.
To facilitate the writing, below we specify the
notation to be used throughout the book, once
and for all.
• Regular English letters are used to represent
mainly natural attributes and chronotopologic
arguments (space, time and spacetime); bold
English letters denote vectors (e.g., assigned
to points in a chronotopologic domain and
locations in a topologic region); Greek letters
represent attribute realizations; bold Greek
letters denote vector realizations; and
subscripts are used to identify knowledge
bases, specific datasets, geographical
coordinates and time instances.
As the following example illustrates, chrono-
topologically distributed data are often collected
from point sources p, which are defined in terms
of their spatial locations and time instances
throughout the domain of interest.
l
Usually, n ¼ 2 or 3.
m
In the same CTDA setting, “location” is a topologic entity, whereas “point” is a chronotopologic entity.
11
FIG. 2.1 An example of chronotopologic point data
coordinates.
Example 2.3 Two points p and p0 (dots) are
shown in Fig. 2.1. The spatial coordinates of
the point p are s ¼ (s1, s2) and the time instance
is t, which are always combined with chronoto-
pologic coordinates; similar is the case of point
p0 . Typical cases of chronotopologic point data
can be found easily when natural attributes are
studied such as, e.g., soil moisture, contaminant
concentration, disease incidence, ocean salinity,
sea surface chlorophyll concentration, tempera-
ture, and humidity. The distribution of natural
attribute values varies over an infinite number
of points, thus forming what is often called a
chronotopologic pollutant distribution (enviro-
mentalism), a disease spread pattern (epidemiol-
ogy), or a moving surface (geography). At these
points, the location on the earth’s surface may
change in elevation, orientation, or proximity to
a feature, and the associated time instance may
vary within a specified period (day, season, year,
etc.) or frequency. The term “associated” is used
here to emphasize the important issue that in the
chronotopologic context space and time are
linked and not independently varying.
As should be expected, a central objective of
CTDA is the development of mathematical
models that account for the chronotopologic
12 1. Chronotopologic data analysis

variability (or dependency) of the natural
attribute X(p) or X(s, t) in a rigorous and inter-
pretable manner. As it turns out, the following
mathematical operator plays a pivotal role in
CTDA.
Quantitative models of attribute variability
(dependency) represent dynamical (time-
dependent) processes across space described by
a mathematical operator S[X(pi), X(pj)] for any
points pi and pj.
Based on the form of the S-operator,n these
variability (dependency) models have different
properties that guide their implementation in
practice. Accordingly, some of them are data-
depended whereas some others are data-free.
We postpone a detailed consideration of the
attribute variability (dependency) models until
Chapters 4–6.o
2.2.1 Revisiting Tobler’s law
Remarkably, CTDA analysis may generate
some interesting albeit sometimes unexpected
results. For illustration, one such result is related
to the well-known 1st law of geographical sci-
ence, which consists of an intuitively
rather profound statement by Waldo R. Tobler
Tobler (1970):
@ Everything is related to everything else, but
near things are more related than distant
things.
As CTDA shows, this intuitively appealing
empirical statement does not apply to all
natural attributes, including those attributes
exhibiting periodic or wave-like variability,
where attribute values at longer distances and
separations apart may be substantively more
related than values at close distances and
separations.
Example 2.4 Two typical cases of this situation
are shown in Figs. 2.2 and 2.3.
• Fig. 2.2 (also, Example 3.21, Section 3 of
Chapter 5) shows the variogram γ X of sea
surface heights that exhibits a wave pattern
(the variogram is a function that measures
attribute variability across space and time,
Chapter 5). As is determined by the

FIG. 2.2 The variogram of sea surface heights exhibiting a wave variation pattern.

n
For technical details, see Eqs. (2.5a–2.5b), Section 2 of Chapter 5.
o
The covariance and variogram are the mainstream chonotopologic variability models expressed by the S-operator.
Additional models include the trivariance, contingogram and sysketogram.
p
As will be discussed in Chapter 5, the lower the variogram value the higher the chronotopologic correlation
(dependency) between the attribute values at the corresponding pair of points.
 2 Chronotopologic variability, dependency and uncertainty 13

FIG. 2.3 Data points pi (i ¼ 1, …, 6) and interpolation point p7. The attribute value is indicated by the color level of the dot
according to the scale legend on the right. The triplet within a parenthesis denotes coordinates pi ¼ (si1, si2; ti), and the italic
number is the contribution of datum at pi to interpolation at p7. The contribution of p6 is much larger than those of p1, p2, p4 and
p5 that are farther away from p7; but the contribution of p3 is larger than those of p4 and p6, which are both closer to p7 (this is
called the “screen effect”).

variogram γ X values,p the sea surface heights
at a space-time lag represented by
C (distance 5 m and separation 10 s) with
γ X ¼ 0.9614 m2 are more closely related than
heights at a lag B (distance 10 m and
separation 5 s) with γ X ¼ 0.9945 m2, and, in
turn, heights at a space-time lag B are more
closely related than heights at a lag
A (distance 5 m and separation 5 s) with
γ X ¼ 1.6160 m2.
• Chronotopologic interpolation would assign
bigger weights to faraway points than to
near ones.q This is, in fact, the case of Fig.
2.3 that represents the so-called screen effect,
where pi (i ¼ 1, …, 6) denote data points
and p7 the interpolated point (the triplet of
numbers within the parentheses indicate
space-time coordinates (si1, si2; ti)). The italic
numbers represent the contribution of each
datum at pi to the interpolated value p7.
A positive value represents a positive
contribution, whereas a negative value
represents a negative contribution, and the
interpolated value was obtained using the
space-time ordinary Kriging interpolation
technique (to be discussed in Chapter 8). As
indicated by these numbers, samples that
are further away from the attribute
interpolation point may have a greater
influence on the interpolated value than
points much closer to the interpolation
point.
Rather not surprisingly, then, these examples
seem to have illustrated numerically the follow-
ing assertion:
• The first half of the empirical Tobler’s law
(i.e., “Everything is related to everything
else”) is valid more often in practice than
the second half (i.e., “near things are more
related than distant things”).

q
These violations of Tobler’s law actually may be the starting point of valuable discoveries. After all, the progress of
any scientific field is based on the derivation of unexpected and previously unknown results.
14 1. Chronotopologic data analysis
As it turns out, this assertion points out
another significant issue that is worth revisiting
in later chapters.
2.3 Concerning real-world uncertainty
and its probabilistic description
In addition to chronotopologic attribute vari-
ability, another key characteristic of CTDA is the
frequent lack of in-situ certainty. Every investi-
gator and practitioner of almost every scientific
field can testify to this fact. Otherwise said,
humans do not seem capable of the kind of last-
ing immunity from doubt that the ideal of cer-
tainty requires.
Example 2.5 When studying changes in global
mean surface air temperature from 1861 to 1984,
Jones et al. (1986) used processing techniques to
correct for uncertainties and inconsistencies in
the historical data not related to climate. As they
notice, “Sea surface temperatures were mea-
sured using water collected in uninsulated, can-
vas buckets, while more recent data come either
from insulated bucket or cooling water intake
measurements, with the latter considered to be
0.3-0.7 °C warmer than uninsulated bucket
measurements.” Accounting for these uncer-
tainties is a complicated matter, since most sea
surface temperature data do not include a
description of what kind of bucket was used.
Moreover, the air temperature measurements
over the ocean were obtained aboard ships of
unknown type and size. But ship type and size
determined the height at which measurements
were obtained, which introduced additional
measurement uncertainties (temperatures can
change rapidly with height above the ocean).
What is required, then, is a comprehensible
and widely applicable notion of uncertainty.
As it turns out, however, this is not necessarily
the case in the real world.
2.3.1 Multifaceted notion
In fact, what makes matters of uncertainty so
complicated is that the following argument can
be made with actual certainty: There is not a
unique definition of uncertainty, which is a mul-
tifaceted notion (ontic, epistemic, or technical),
which may emerge in a number of distinct ways
and may refer to different things.
Given, then, that there is considerable uncer-
tainty about the definition of uncertainty itself,
the associated probability notion may refer to
more than one consideration:
❶ Probability as a mathematical quantity: a
number belonging to the range between 0
and 1 (or 0 and 100%) expressing the
frequency of occurrence of possible
attribute values.
❷ Probability as a state of mind: a state of
doubt or lack of certainty about the true
attribute value, or a state of incomplete
understanding of the phenomenon, or a
state of surprise regarding the outcomes of
experiments or future events.
❸ Probability as an irreducible (inherent)
property of the natural phenomenon:
representing, e.g., Heisenberg’s uncertainty
principle of phenomena in microscales.
Example 2.6 Uncertainty in science is closely
related to the fact that most data have a range
of possible values as opposed to a precise single
value). It might be enlightening to view matters
from a layman’s perspective. The layman’s clas-
sification of the laws of uncertainty rather distin-
guishes between four kinds, as described in
Table 2.1.
TABLE 2.1 Layman’s laws of uncertainty.
Law Description
Inevitability Almost every investigation step contains a
factor of uncertainty, it’s an inevitable part
of doing science in the real world
Invisibility The absence of evidence is not evidence of
absence
Obviousness In many data collections, the figure most
obviously correct could be the mistake
Interpretation No matter what the result, there is always
someone who could misinterpret it
 2 Chronotopologic variability, dependency and uncertainty
Yet, despite the considerable and multi-
faceted difficulties caused by the above aspects
of uncertainty, it is widely acknowledged that
to understand anything about any phenomenon
of nature, one must understand its uncertainty.
Accordingly, the next suggestion is a reasonable
one:
• Uncertainty-related assumptions about a
phenomenon should be made on the basis
of rational reasons, i.e., they should be
justified assumptions.
Rational reasoning shows that data uncer-
tainty does not necessarily imply that the data
are wrong and useless. Quite the contrary,
rigorous uncertainty quantification becomes
very important in this respect, since the magni-
tude of uncertainty offers an assessment of how
confident the investigators would be on the
accuracy of their data (i.e., higher uncertainty
implies lower confidence to the data).
Example 2.7 While the uncertainty level of
historical air temperature data increases going
further back in time, it has been reduced consid-
erably, for all practical purposes, in more recent
years starting around 1900 (Mann et al., 1998). In
another case, although one cannot specify the
exact amount of precipitation during next
month, nevertheless, one can assume, as practi-
cally certain, that there will be no major weather
pattern changes. Hence, the known weather
conditions of previous years provide good
grounds for reasonably assuming a certain prob-
ability rainfall distribution.
2.3.2 The limited notion of practical
certainty
Surely, the level of rigor associated with
uncertainty determination may also depend on
the study conditions and goals. In this respect,
the following notion could be useful in
applications:
15
A notion of practical certainty would be
appropriate (a) when the supporting argumenta-
tion is rational or justified enough to be relied
upon in practical deliberation and in deciding
what to do, or (b) when the stakes are low,
whereas a more rigorous or stricter uncertainty
notion may be appropriate when the stakes are
higher.
In CTDA, one can distinguish between
temporal vs. intemporal practical uncertainty
(i.e., certainty at a specific time vs. lasting cer-
tainty), as well as, between context-dependent
vs. context-independent practical certainty.
Example 2.8 While in certain applications it is
often sufficient to focus on the conditions
required for a phenomenon to be certain at a
time or within a certain context, in other applica-
tions, it is recognized that a phenomenon might
be certain at one time (when a clear and distinct
perception of the phenomenon has been estab-
lished) or within a particular context but not at
another time (when one no longer has or is not
attending to this clear and distinct perception)
or within a different context.
2.3.3 Sources of uncertainty
Real-world case studies encounter consider-
able uncertainties from different sources, which
is why CTDA studies put great emphasis on
the reliability and quality of the data used.
A list of key uncertainty sources is given in
Fig. 2.4, and a brief review of them follows
(see below items ❶ through ❹).
❶ One of the most common sources is in-situ
data uncertainty, which is linked to
experimental, observation and interpretation
errors, heterogeneity and scale variation, as
well as to limited or sparse data, see
Example 2.1 of Section 2.
❷ One of the most important sources is model
(or conceptual) uncertainty, i.e., whether or
16 1. Chronotopologic data analysis
FIG. 2.4 Uncertainty sources considered in CTDA.
not a mathematical model used by CTDA is
an adequate representation of the real
phenomenon.
Model uncertainty may be linked to the ontic
condition, which refers to an inherent feature of
the natural phenomenon, and the epistemicr
condition, which refers to incomplete under-
standing of the phenomenon.
Example 2.9 This kind of uncertainty may be
due to imperfections and idealizations (due to
ontic or epistemic conditions) made in the con-
struction of a model representing a physical
phenomenon, and to the choices of the relevant
probability functions to describe statistically the
model parameters.
Altera pars, perhaps not surprisingly, there are
several CTDA studies in which uncertainty is
linked to the opposite situation, as follows:
• Uncertainty may be linked to an unmodeled
source that often can cause errors and biases
that are chronotopologically correlated.
Example 2.10 Geomagnetic field studies con-
sider both measurement errors, i.e., instrument
accuracy (see data uncertainties from the Swarm
mission; Tøffner-Clausen et al., 2016) and errors
from unmodeled sources (even the most modern
r
The branch of philosophy that concerns itself with knowledge is called epistemology, or methodology, or philosophy
of science. It is the focus of some important developments in CTDA (see, also, subsequent chapters).
satellite field models do not fully capture the
ionospheric processes, particularly in the auro-
ral region; Finlay et al., 2017).
❸ Computational uncertainty could be the result
of imperfection of a numerical technique and
the finiteness of machine precision.
Example 2.11 Computational uncertainty may
characterize the sensitivity of a numerical prob-
lem solution taking into account the errors and
inaccuracies of inputs, algorithms, truncations,
and roundings.
❹ Parameter uncertainty quantifies the lack of
knowledge about the actual value of a
parameter, and is linked to data
erraticness, physical properties, process
optimization, design variables, rate
coefficients, and unknown boundary and
initial conditions (BIC).
Example 2.12 Parameters that are inputs to
computer models are uncertain when their exact
values can neither be computed in physical
experiments nor be inferred by statistical tech-
niques. When more than one parameter is
involved, their uncertainties are correlated and
their proper merging is needed to fully assess
uncertainty in an attribute derived from more
than one parameter.
 2 Chronotopologic variability, dependency and uncertainty
2.3.4 Consequences of uncertainty
Uncertainty can have serious consequences in
many areas of scientific inquiry. In particular,
each of the above uncertainty sources affects
basic issues of attribute modeling, estimation,
and mapping,s as follows:
• The reliability and accuracy of the
chronotopologic attribute maps that one
derives depend to a considerable extent on
the uncertainty conception that one uses.
• Another issue is the propagation of initial
uncertainties by physical laws (chaos
phenomena).
• Uncertainty has a profound influence on risk
analysis (environmental impacts, health
effects, social implications), decision-
making (optimality criteria), and planning
(multiple objectives, economic factors,
public policy).
In view of the above concerns, in CTDA stud-
ies investigators take into consideration not only
what the data suggest, but also the data quality
and the consequences of uncertainty as regards
various courses of action under the specific
circumstances.
2.3.5 Quantification of uncertainty
Another central objective of CTDA is the
development of quantitative models that
describe the uncertainty sources of the natural
attribute X(s, t) or X(p) in a rigorous and inter-
pretive manner. This is, in fact, the role of a vari-
ety of probability models representing the
corresponding multifaceted notion of probabil-
ity (introduced in Section 2.3.1 above).
Quantitative representations of attribute
uncertainty include the probability models
(probability density and distribution functions)
describing the occurrence conditions of possible
attribute values, and their statistical functions
(mean, variance, covariance, variogram).
s
We will revisit each one of these three CTDA components, at various parts of the book.
17
Example 2.13 When attribute uncertainty is
fully characterized by a normal (Gaussian) prob-
ability model, it can be partially summarized in
terms of the model’s standard deviation (also
called error bar) measuring model’s width.
Assessing the uncertainties of a posteriori proba-
bility model requires adequate observation error
assessment and prior parameter information.
Remarkably, the underlying probability notion
at the core of these models does not admit a unique
interpretation. As a matter of fact, two essential
interpretations may be considered, as follows:
Probability may be a fundamentally epistemic
concept, i.e., the probability of an attribute value
is relative to some body of knowledge. If a body
of knowledge is construed as a set of statements
about the phenomenon, then probability becomes
a logical concept.
Example 2.14 As a further classification of the
twofold probability characterization above,
Table 2.2 presents five commonly encountered
cases of probability interpretation that depend
on the context considered.
As it turns out, different probability models
have distinct properties that guide their imple-
mentation in practice. Some of them are more
informative but more difficult to develop,
whereas some others are easier to compute but
offer a limited representation of the actual attri-
bute uncertainty.
Example 2.15 To mention a well-known case,
there is a high demand for more informative
probabilistic models of weather forecasting
and extreme event warning (hurricanes, earth-
quakes, tsunamis, etc.); yet, admittedly, these
models become increasingly complex too.
At the moment, it seems appropriate to post-
pone until a later chapter the presentation of a
18 1. Chronotopologic data analysis
TABLE 2.2 Probability interpretations.
Probability
interpretation Contextual description
Classical The ratio of cases in which an attribute
occurs over all possible cases
Frequentist The value of classical ratio when
number of cases is very large (tends to
infinity)
Geometric The ratio of the domain in which an
attribute occurs over the entire domain
of interest
Subjective The result of personal experience,
preferences and intuition
Scientific The result of rational assessment based
on logical reasoning and scientific
knowledge
more detailed discussion of the quantitative
uncertainty models employed in scientific
applications.
3 Theory and evidence
A chief objective of CTDA is to adequately use
the existing body of knowledge in a specified
area of science in an efficient and systematic
way in order to better understand a phenome-
non and make predictions. Let us start with a
brief yet concise description of what constitutes
a “theory” and what constitutes “evidence”.
Theory is the product of constructive imagina-
tion and logical thinking (mathematical models, sci-
entific theories, physical laws, and hypotheses),t
whereas evidence is the product of inquisitive
observation and practical experience (evidential
support, perceptual experience, empirical data,
measurements, recordings, surveys) of various
degrees of reliability and uncertainty.
t
Classical examples of theories are Coulomb’s law of electrical attraction and repulsion, Maxwell’s electromagnetic
equations, and the ideal gas equations describing relations between temperatures and pressures of enclosed gases.
The above definitions emphasize both the clear
distinction between theory and evidence about the
natural attribute of interest and their strong com-
plementarity. Another interesting distinction
should be made between theory, which is about
predicting what has not been observed yet, and
inference, which is based on evidence alone and
cannot exceed the domain of the data it uses.
As regards scientific theories, they typically
develop gradually over time into increasingly com-
plex models representing concentrated human
knowledge that often spans several centuries.
Example 3.1 In physics, the original Newton’s
theory was advanced considerably during the
following three centuries, through the works
of Bohr, Einstein, Heisenberg, Higgs, and many
others to form today’s Standard Model.
This gradual increase in complexity may be
due to various reasons, i.e., as the understanding
of the phenomenon improves, new variables
need to be included in the theory, thus increas-
ing its complexity, or, a phenomenon may man-
ifest only at a higher level of complexity and not
at lower levels, or, a theory that is valid in one
scale maybe not so in a different scale. On the
other hand, a deeper physical understanding
can lead to theory simplification, i.e., by omit-
ting variables that play a minor role in the phe-
nomenon, or certain correlations are found to
have negligible effects.
3.1 The value of theory
The value of theory has been emphasized
throughout the centuries by many great
thinkers. According to Leonardo da Vinci:
@ He who loves practice without theory is like
the sailor who boards a ship without a rudder
and compass and never knows where he
may cast.
 3 Theory and evidence
In the words of Auguste Comte:
@ If it is true that every theory must be based
upon observed facts, it is equally true that
facts cannot be observed without the
guidance of some theory. Without such
guidance, our facts would be desultory and
fruitless; we could not retain them: for the
most part, we could not even perceive them.
And, writing about Charles Darwin, Lionel
Ruby (1966) notices:
@ Darwin and a fellow scientist were searching
for fossils in the north of England. They were
not aware of the glacial theory at the time.
Years later Darwin revisited the area, and
he was now astonished to discover how
clearly marked were the glacial ridges on
the rocks. He had not noticed them on his
earlier visit because he was not looking for
them.... Darwin was able to appreciate the
glacial markings only after he became
aware of the glacial theory.
Sufficiently motivated by these words of wis-
dom, next we examine some other valuable uses
of theory.
3.2 Against convenient slogans
Science-based CTDA ought to proceed with
caution and avoid the oversimplifying slogans
of purely data-driven analytics. Chief among
them is the convenient slogan encouraging
investigators to:
? Let the data speak.
Yet, surrendering oneself to the data is often a
naive policy with many hidden dangers (see,
also, discussion in Section 3 of Chapter 3). One
of these dangers is that those surrendering
themselves to this convenient slogan often fail
to appreciate the fact that an attribute dataset
is not just a set to be manipulated by statistical
techniques, their numerical correlations com-
puted, and purely number-driven (content-free)
19
conclusions drawn. Yet, reality is more compli-
cated and lies beyond the convenient slogan
above.
A dataset has a natural “soul” that needs to be
explored strictly within the specified phenome-
nal context so that theory-driven (content-
dependent) conclusions can be drawn.
The knowledgeable investigators understand
that the same numbers may have very different
meanings, depending on the physical character-
istics of the attribute they represent, and their
possible ranges are restricted by these character-
istics. Any number representing an attribute
value has units that assign to it a physical mean-
ing, and any data-processing technique should
be conditioned by this meaning.
Example 3.2 A negative temperature value
makes both numerical and physical sense. How-
ever, a negative permeability value, although
makes numerical sense it surely does not have
physical meaning. To declare that the ozone con-
centration at a certain location is, say, 5, is like to
provide a meaningless number, unless physical
units, ppm, are assigned to it. Furthermore, this
number should be consistent with other num-
bers representing concentrations that are linked
to it by a physical law.
Another hidden danger associated with the
convenient slogan above is the consideration
of apparent relationships among the data that
do not actually exist.
• Currently, the only way to avoid the traps of
such spurious correlations among the data is
the protection provided by a sound theory.
If, instead, one focuses on correlation and
ignores considerations of theory or mechanisms,
one may be engaging in flawed reasoning.
Therefore, if investigators really need a slogan,
then a much more appropriate one should be to:
? Let the theory speak about the data.
20 1. Chronotopologic data analysis
Indeed, it is a theory that can take data anal-
ysis beyond mere numerics and assign the nec-
essary meaning to the data and avoid common
numerical traps like spurious data correlations.
Also, theoretical developments may be more
captivating than empirical evidence in cases
where randomization and replication are diffi-
cult and several confounders exist.
3.3 Evidence unavailability
It is not surprising that the necessity of theory
in many real-world investigations is also estab-
lished on the basis of practical evidence-related
reasons. Indeed, people need to understand the
world, and when sufficient empirical evidence
is not available, they naturally tend to rely on
theory.
Theoretical developments may be necessary
because of unavailable evidence, i.e., in many
real-world cases, evidence is not available at the
current state.
Example 3.3 Evidence concerning the future
occurrence of a phenomenon is usually unavail-
able, and empirically inaccessible attribute
domains may exist.
As a matter of fact, logically the notion of
“future evidence” may be a contradiction in
terms, since the “evidence” expresses a current
state (i.e., one definitely talks about “current evi-
dence”), whereas the future refers to an unrea-
lizable yet state (which may or may not be
realized in the future, but this is not determined
with certainty).
Example 3.4 When investigators study the
weather conditions at a future time, they need to
rely heavily on theoretical developments in the
form of complicated weather models and hypoth-
eses, surely supported by past evidence, whereas
direct evidence of future events is profoundly
absent. Similarly, when one attempts to forecast
future developments in financial affairs or politics,
future evidence is, indeed, a contradiction in
terms, and one has no other option than to theorize
or hypothesize about possible future scenarios.
3.4 Synergy
The crux of the matter is that to understand
the world, one needs both theory and evidence.
One should take advantage of the available evi-
dence in all its forms (traditional and emerging)
but without ignoring the treasure of core knowl-
edge accumulated in scientific theories. And, the
deeper meaning of the distinct yet complemen-
tary features of theory and evidence discussed
above can be summarized as follows:
Beyond the distinct yet crucial roles theory and
evidence individually play in scientific inves-
tigation, perhaps even more important is the real-
ization that they act in synergy, i.e., there is an
intrinsic and essential relationship between
the two.
Some key features of theory and evidence,
together with some comments concerning their
interplay, are presented in Table 3.1.
Example 3.5 Interesting cases of the synergy
between theory and evidence in two different
scientific disciplines are as follows:
• In modern health sciences, there is an urgent
need to combine evidence bases with
theoretical and computational models in
order to adequately investigate the huge
amounts of data from healthcare and
medical research.
• In earth sciences, many applications merge
the information contained into geophysical
observations with that coming from
dynamical models, a process often known
as assimilation.
Although theory develops before any evi-
dence becomes available in the form of data,
yet it can be properly modified (updated or
 3 Theory and evidence
TABLE 3.1 Theory, evidence, and their interplay.
Theory feature
Theory is a scientific and systematic description of
knowledge that usually relies on a set of assumptions about
the phenomenon of interest
Theory is customarily represented as a collection of
sentences, propositions, beliefs and formulas, and their
logical consequences
Theory uses evidence (real data) to calibrate its parameters
Theory and theorizing are cornerstones of a scientific
discipline. Research findings and conclusions are usually
presented in the form of theories and hypotheses
A sound theory helps avoid both unanchored abstract
thought and a narrow empiricism that concerns itself only
with observation or measurement and unidirected data
gathering
A theory can be challenged and defended in different ways
and at different times as new empirical evidence and
investigative techniques are introduced
a
For example, to measure soil porosity one associates it with an appropriate concept; and proponents of a caloric account of heat do not understand heat
experiment results in the same way as those who view heat in terms of mean kinetic energy or radiation.
b
For example, if dark clouds are observed in the sky, there is a high likelihood of rain, but if dark clouds are observed when the temperature is far below freezing,
there may be no rain.
calibrated) when new and reliable evidence
emerges in the process of scientific inquiry.
Actually, this seems to be an appropriate proce-
dure for combining theory with evidence. In
practice, theory and evidence should rest com-
fortably with each other and they should also
reinforce each other (detecting data correlations
facilitates theory testing and refinement; and
combined with the increasing computing
power, the data may suggest different ways of
developing a theory).
• When significant differences are found
between theory and evidence, they can lead
to improvements in both the way a theory
21
Empirical evidence feature
Evidence is obtained by observing natural and
experimentally generated events, objects, and effects
Evidence sources are theory-laden observations and research
findings. If evidence is separated from theory, it is
disconnected from its origin and it may become devoid of
scientific meaning
Evidence-based practice applies the most credible and
reliable evidence available in order to understand and
describe a phenomenon
Evidence is a conceptual term, and, thus, it is linked to
knowledge fields that are conceptual. As such, the evidence
does not look the same to observers with different conceptual
resourcesa
Evidence depends on the relevant context. By changing the
context, the notion of evidence may change and even become
problematicb
Very few investigators would be convinced by a study
drawing an apparently interesting conclusion without a
plausible, at least, theory adequately fitting the relevant data
is constructed and the way the evidence is
obtained, including the type of evidence
that counts in scientific reasoning.
If a theory suggests that the current evidence is
inadequate, it implies that more research on that
specific subject is necessary, whereas when hard
data contradict a theory or the theory cannot
explain it, the theory must be revised accordingly.
Naturally, theory and practice have not been
developed to the same degree in all disciplines.
Example 3.6 Weather forecasting is a disci-
pline that is both theory- and practice-strong,
whereas public health is considered a theory-
weak and practice-strong discipline.
22 1. Chronotopologic data analysis
By way of a summary:
Evidence is necessary for science but not suf-
ficient. Evidence needs to be integrated with the-
ory in order to have any explanatory and
predictive value.
Ut ita dicam, a more appropriate goal should
be seeking “content-rich data” or “content-deep
data” rather than merely “big data” (Section 3 of
Chapter 3). A detailed discussion of several
important theory- vs. evidence-related topics
will be presented in Chapters 3 and 5.
4 Chronotopologic estimation and
mapping
While the study of attribute variation or
dependence (introduced in Section 2) aims at a
deep understanding of the chronotopologic
structure and laws of change of the natural attri-
bute of interest, another key CTDA objective is
the following.
Chronotopologic estimation seeks the deriva-
tion of estimates X ^ ðpÞ of an attribute X(p) at
unsampled points within the chronotopologic
domain of interest, using all available information
about the attribute.
Within the broad context of chronotopologic
estimation, in a space-time or a spacetime setting,
two very important special cases are considered:
• Chronotopologic interpolation that seeks to
fill in a domain with attribute values at
unsampled points.
• Chronotopologic extrapolation that seeks to
obtain attribute values beyond this domain.
Interpolation is clearly a much easier problem
than extrapolation.
4.1 Threefold conditions
It is worth stressing that in epistemic
(knowledge-theoretic) terms, chronotopologic
attribute estimation is not a categorical device,
but only a conditional one. Accordingly, there
are three main conditions of attribute estimation
that are worth keeping in mind:
❶ Attribute estimation is uncertain and
approximate.
❷ It depends on one’s knowledge (models,
theories, data) about the phenomenon of
interest.
❸ It is warranted only if the assumptions
made are reasonable and realistic.
Condition ❶ is an implication of the condi-
tionality of estimation (i.e., the data are too weak
to support categorical estimation of perfect reli-
ability), in which case the estimated attribute
values are uncertain and only approximations
of the real albeit unknown ones. Condition ❷
is about how much data will support how much
conditional estimation. Regarding condition ❸,
one should distinguish between the assump-
tions one rationally and justifiably makes about
the phenomenon and the modeling assumptions
one uses in estimation. Otherwise said, if the
characterization of the chronotopologic variabil-
ity and the in-situ uncertainty of the natural
attribute X(p) is valid and meaningful, then
the estimated values X ^ ðpÞ will be accurate and
informative.
As it turns out, the restrictions imposed on
estimation by the conditions ❶, ❷, and ❸
above may not necessarily keep estimation
models and techniques from being informative
and useful in practice. An instructive example
is appropriate here.
Example 4.1 We offer our readers two illustra-
tions of the above claim:
 4 Chronotopologic estimation and mapping
• That the law of gravity is not perfectly
reliable, because of wind currents and
other intervening factors, is not an excuse
for one to ignore its predictions and decide
to jump from a ravine. Any common-
sensical human being understands that
despite its imperfection, the law remains
reasonably valid for all practical purposes.
• Similarly, to decline to make critical climate
change estimates on the grounds that the
data is imperfect (as many investigators
have argued) would probably not be a good
decision, especially given the gravity of the
matter. As it turns out, investigators can
indeed draw sufficiently informative
inferences with the evidence we have.
4.2 Confidence and accuracy
On a relevant note, since chronotopologic esti-
mation is subject to error, it is necessary to assess
how much uncertainty can be tolerated and how
much confidence can be placed in the generated
attribute estimates. Arthur Rudolph, the scientist
who developed the Saturn V moon rocket,
expressed this critical issue very succinctly:
@ You want a valve that doesn’t leak, and you
try everything possible to develop one. But
the real world provides you with a leaky
valve. Then, you have to determine how
much leaking you can tolerate.
On practical grounds, then, the need for toler-
ance and confidence assessment leads to the
development of several estimation accuracy indica-
tors, which are absolutely necessary given the data
heterogeneity and uncertainty (Section 2). After
all, a general motto of scientific investigations is:
Being aware of how much one does not know
is often more important than being cognizant of
how much one knows.
23
Remarkably, the kind of “quality control”
introduced by the accuracy indicators is not part
of deterministic interpolation (see, e.g., the inverse
distance techniques to be discussed in Chapter 7).
Moreover, some techniques do not take into
account intrinsic properties of the interpolated
phenomena (e.g., rapid changes in the data with
distance vs. more gradual changes, or law-
obeying variations), as they only take account of
the location/instance of the data points. The
CTDA techniques to be discussed in the book will
try to address the above and similar issues.
4.3 Interpretation
We conclude with the following comment
concerning the interpretation of the CTDA out-
comes, like the chronotopologic attribute
maps.
In general, when investigators interpret the
attribute maps, they attempt to explain the uncov-
ered patterns and trends, bringing all of their
background knowledge, experience and skills to
bear on the question and relating their data to
existing scientific ideas.
Example 4.2 Weather forecasting is an
interpretation made by a meteorologist of data
collected by satellites using climate models.
Given the personal nature of the knowledge
the investigators rely upon, interpretation is
bound to be subjective to some extent, in the
sense that by bringing a different background,
thinking mode and experience to bear on the
same dataset, an investigator can come to very
different conclusions. A widely publicized case
in point is the global climate debate concerning
the presence or absence of a temperature trend
in time, which is the focus of the following
example.
24 1. Chronotopologic data analysis
Example 4.3 The study of global warming
turned out to be an arena of strong disagree-
ments concerning data analysis and interpreta-
tion. Jones et al. (1986) interpreted the changes
in global mean surface air temperature from
1861 to 1984 to imply a long-term warming trend
(e.g., they emphasized that 1980, 1981, and 1983
were the three warmest years in the entire
dataset). On the other hand, Lindzen (1990)
opposed this interpretation using a number of
arguments, like the inadequacy of the existing
data collection network to correct for data uncer-
tainty, substantial gaps in temperature coverage
(especially over the oceans), errors in data anal-
ysis, and inappropriate interpretation of the
global mean temperatures, thus, reaching the
opposite conclusion that there is no trend in
the data. This is a classical case where different
investigators made very different interpreta-
tions and drew very different conclusions from
the same datasets, as a result of their different
backgrounds, experiences, and thinking modes.
Ut aequum, disagreement is not uncommon in
scientific investigations. When it occurs, it rou-
tinely leads to more data collection and analysis,
but perhaps more importantly it should also
lead to the consideration and appreciation of
the distinct thinking modes used by the different
investigators.
5 A review of CTDA techniques
As already noted, there exist several CTDA
techniques (both S-TDA and STDA) available
in the relevant literature. Naturally, each tech-
nique has its own set of specific features and
assumptions.
5.1 Classification
The CTDA techniques to be discussed in this
book, in varying degrees of detail, belong to the
following distinct groups:
❶ Pure chronogeographic statistics techniques
(point center, diffusion, inverse metrics, etc.;
Chapters 4 and 7). They are concerned only
with the chronotopological characteristics
(coordinates, distances, instances) of the
phenomenon.
❷ Attribute chronogeographic statistics
techniques (Moran’s index, Geary’s index,
etc.; Chapters 4, 7, and 12). They are
concerned both with the chronotopological
characteristics of the phenomenon and its
physical values (mass, weight, temperature,
concentration, etc.). They include multiple
linear regression (MLR), land-use regression
(LUR), and linear mixed-effect models.
❸ Standard geostatistics techniques (ordinary
and simple Kriging, statistical regression,
etc.; Chapters 5 and 8). They are popular
techniques that focus on hard (uncertainty-
free) data, and they usually rely on rather
restrictive modeling assumptions (linear
estimators, Gaussian data distribution,
low-order statistics).
❹ Modern geostatistics techniques (Bayesian
maximum entropy, stochastic logic,
factoras, etc.; Chapters 6 and 9). They are
very versatile, incorporate core knowledge,
assimilate both hard and soft (uncertain)
data, and avoid the restrictive modeling
assumptions of the standard techniques
(nonlinear estimators and non-Gaussian
data distribution are automatically
considered).
❺ Machine learning techniques (artificial
neural networks, random forest, gradient
boost machine, etc.; Chapter 12). They are
often characterized as “black-box,”
because they do not involve confirmed
mathematical expressions representing the
substance of the natural attribute of
interest. Yet, due to their high
nonlinearity, self-adaptation, self-learning
and other features, they can fit very well
complex real world processes.
 5 A review of CTDA techniques
Each group of CTDA techniques has its pros
and cons. Although many factors affect a tech-
nique (including sample size, scale, sampling
design, data properties, and law of change),
there are not always consistent findings about
how exactly these factors affect the technique’s
performance (accuracy, informativeness, sub-
stantiveness). Therefore, it may not be always
an easy task to select an appropriate chronoto-
pologic modeling and interpolation technique
for a given input dataset. It is noteworthy that
different data interpolation techniques can gen-
erate different maps that are interpreted as
representing very different phenomena.
In view of the classification introduced, it is
often methodologically useful to make a key
distinction:
CTDA distinguishes between defining and
accompanying properties of a class of techniques,
where the former refer to properties that hold for
any member of the class and the latter to proper-
ties that hold only for a certain subset of the class.
The test whether a property is a defining one
or not consists of investigating whether its
absence would exclude a technique from inclu-
sion in the specific CTDA class. Chronotopology
is a defining property of the CTDA class,
whereas uncertainty is an accompanying prop-
erty, since it holds for some members of the
CTDA class (e.g., geostatistics, Chapter 8) but
not for others (e.g., deterministic chronogeo-
metric, Chapter 7). Depending on their defining
and accompanying properties, one may distin-
guish between techniques that are global vs.
local, exact vs. inexact, and formal vs. substanti-
ve.u CTDA techniques that have been developed
for and applied to various disciplines (physical,
biological, and social) are often data-specific or
u
The distinction between a formal and a substantive technique is that the former considers data as numbers assigning
no physical meaning to them, whereas the latter views data as physical quantities assigning different interpretations to
them, depending on the phenomenon.
25
attribute-specific. In fact, there exist several
chronotopologic estimation techniques available
in the relevant literature (mainly interpolation
rather than extrapolation techniques, Section 5
above).
Interestingly, of the several changes in the
meaning of the term CTDA over the years, a cer-
tain proportion is due to a tendency of accompa-
nying features to become gradually defining ones
(in some cases in addition to the original defining
properties, and some others displacing the origi-
nal defining properties). It is also possible that the
confusion between defining and accompanying
properties could produce misconceptions that
stand in the way of a realistic view of the phe-
nomenon. Misconceptions about the physical
world often originate in semantic shortsighted-
ness, but are usually liable to correction through
observation.
5.2 Computer technology
Naturally, one could not underestimate the
important contribution of technology in
computer-assisted CTDA. During the last few
decades, CTDA underwent a profound transfor-
mation from being manual (and tedious) to pre-
dominantly computer-based (this was made
possible mainly due to advances in computer
hardware during the 1990s).
• Computers play a major role in the
development and implementation of CTDA
techniques.
Among the advantages of computer software
is that it makes CTDA much more efficient and
effective, allowing improved numerical accu-
racy, higher speed, time-savings, and handling
complicated computations (see, also,
Section 8). Technology advances of various
26 1. Chronotopologic data analysis
forms have made it easier for investigators to
collect and organize evidence in ways that allow
systematic analysis and encourage worldwide
collaboration among groups of CTDA experts
and institutes.
We conclude our very brief review with a
word of caution. Although the continuous
development of more powerful computers in
the past has facilitated improvements in real-
world CTDA that would not have been possible
otherwise, in the view of many scholars, this
does not seem to necessarily be the case any-
more. Instead, it has been argued that processors
do not continue getting faster at the previous
pace; instead, it is hoped that the continuation
of the steady increase in power will come from
massively parallel machines.
6 Chronotopologic visualization
technology
After CTDA has been used to derive quantita-
tive data dependency measures and generate
numerical attribute estimates from isolated data,
the next task is usually to portray these results
on a continuous chronotopologic domain. In
such a context, the following objective is of par-
ticular interest:
Chronotopologic mapping is the act of repre-
senting the numerical CTDA results, involving
data and attribute estimates, in a continuous
and visually informative way.
Given its objective above, mapping is closely
linked to the following form of information
communication:
Visualization in the form of maps, graphs,
charts, diagrams, and animation is communica-
tion of the available CTDA information to the
interested individuals or groups of individuals
accurately and efficiently.
As such, visualization is enabling the act of
mapping in a very effective and efficient man-
ner. This brings to mind what Albert Einstein
once said about scientists’ ultimate goal:
@ The natural scientists try to develop, in the
fashion that suits him best, a simplified and
intelligible picture of the world; he then
tries to some extent to substitute this
cosmos of his for the world of experience,
and thus to overcome it. This is also what
the painter, the poet, or the speculative
philosopher try to do.
In the present case, the “intelligible picture
of the world” is represented by the properly
visualized attribute map. Surely, achieving
intelligibility in mapping depends on choosing
the kind of visualization tool that works best
given the study objectives. In this respect, mod-
ern visualization is a very powerful technology
that provides a large collection of tools that
can generate chronotopologic maps of varying
levels of sophistication (contour maps, domain
coloring maps, animated maps, etc.). The idea
is not so much to follow a formula, but rather
to be aware of the many different aspects of
visualizing a chronotopologic map or other
CTDA results.
Generally, experience has shown that visual-
ization is an incredibly effective medium for
informing or conveying a viewpoint or a mes-
sage across. Indeed, it is widely acknowledged
that the informativeness and usefulness of
CTDA results can be greatly improved by
means of powerful visualization techniques.
More specifically, based on the attribute data
at the sampled points and the estimated attri-
bute values at the unsampled points a detailed
and, hopefully, highly enlightening map can be
visualized. The emphasis is on the visualization
of attribute properties and relationships that
takes into account the chronotopologic localiza-
tion/instantanization of these properties and
relationships.
 7 The range of CTDA applications
6.1 The emergence of visual thinking
Humans are inherently visual. The human
brain has the ability to literally process and store
visual information faster than text, symbols,
numerics, and statistics. Visualization takes
complex information that most people don’t
have the time, skills, or attention span to compre-
hend, and transforms it into easily understood,
pleasant and visually focused presentations. In
this setting, then, another important human
cognition issue emerges that transcends mere
technological considerations:
The visualization of CTDA results requires the
adaptation of a new way of thinking, viz. visual
thinking, and, accordingly, to develop a deep
understanding of what it means to think visually.
The modern scientists are expected to be able
to think visually and use visualization technol-
ogy to its full potential so that they can generate
plots that allow them and others (the public,
stakeholders, decision-makers) to grasp informa-
tion in an efficient and physically meaningful
manner. Given that CTDA involves a variety of
knowledge bases, core and specificatory, shared
by many different individuals with different
backgrounds (scientific and cultural), this has
made visualization a very important common lan-
guage. One then comes to appreciate the claim
that if data is information (of various kinds) about
the world, then visualization is the creative com-
munication of this information. Therefore:
• A good visualization technique does not only
need to be technically adequate and
aesthetically pleasant, but contextually rich
and functional too.
The first two requirements focus on the eyes
and follow certain presentation rules of visual
grammar, whereas the last two requirements
focus on the mind and respond to context
(rather than set it). According to this perspec-
tive, visualization conveys facts and ideas
27
clearly and also sets them in the appropriate
context that can potentially influence people’s
views and decisions. Therefore, a good visual-
ization technique should add to data’s meaning
and/or clarity.
More about this interesting topic will be pre-
sented in the following chapters. In the mean-
time, let us note that increasingly sophisticated
and high-quality visualization technology
products become available nowadays that are
often both affordable and easy to use. The inter-
ested readers are referred to the relevant visual-
ization literature (e.g., Tufte, 2001; Knaflic,
2015; Healy, 2018; Kirk, 2016; and references
therein).
7 The range of CTDA applications
CTDA (in its S-TDA and STDA formulations)
provides useful quantitative tools of chrono-
topologic dependency characterization and
mapping. Besides the characterization and
visual perception of attribute distributions, it is
very important to translate the emerging chron-
otopologic patterns into useful objective and
measurable entities, which is why CTDA can
help the investigators answer key questions
linked to real-world problems of numerous
disciplines, such as those listed in Table 7.1.
The list in Table 7.1 is by no means exhaustive –
the readers may propose many other CTDA
issues that are clearly relevant to their fields of
interest. On the other hand, it may come as no sur-
prise to our readers that many fields of inquiry
have contributed to CTDA’s rise in its modern
forms, including cartography and surveying,
botanical studies of global plant distributions
and local plant locations, forestry, ethnological
surveys of population movement, human expo-
sure investigations, landscape ecological studies
of vegetation blocks, environmental studies of
population dynamics, biogeography research,
epidemiology, and disease spread assessment
combined with locational healthcare delivery
data, and space-time econometrics.
28 1. Chronotopologic data analysis
TABLE 7.1 List of scientific disciplines and related questions.
Discipline
Environmental science: Investigators need to know the
chronotopologic distribution of pollutant concentrations
during peak days
Earth sciences: Investigators need to derive estimates of a
wide range of physical attributes in large domains and
during long time periods
Mining: Data analysts need to estimate from a usually
limited set of samples of different periods the extension of a
mineral deposit in a site
Agronomy: Scientists often must analyze a region for
agricultural zoning purposes
Epidemiology: Health professionals collect data about
disease occurrence
Social studies: People would like to know whether there is
any chronotopologic concentration in regional theft
distribution
Lastly, as noted earlier, advances in TGIS
have contributed considerably to continuing
CTDA developments and wide applicability.
In return, TGIS popularity requires CTDA
methods that allow the transformation of quan-
titative results into the kind of systematic proce-
dures and qualitative representations that are
tractable to nonexpert users.
Questions
• Does pollution exceed certain thresholds
• In what parts of a region, and when and how frequently
does threshold exceedance happen
• Which are the most anticipated public health effects
• Which are the most efficient prevention and control
measures
• What is the fluctuation of water table elevation in a region
during the summer months
• How weather variables—such as temperature, air pressure,
relative humidity, precipitation, cloud cover, wind speed,
and direction—change from a region to the next, or from
current to future
• Can those samples be used to estimate the mineral
distribution in that site or to forecast its deposit in the
near future
• What is the accuracy of the estimates to be used in decision-
making and management
• How to choose the independent variables—soil, vegetation,
and geomorphology
• What is the dependency structure of such variables within
the domain
• How to determine what the contribution of each variable is
in order to define where the growth of each type of crop is
more adequate
• Does the distribution of disease cases develop a specific
chronotopologic pattern and which are its major features
• Is there any association with any pollution sources
• Do geographical disease patterns vary with time
• Should a state of emergency be declared, how long should
it last, and which regional hospitals should participate
• Are there any “hot spots” in this distribution
• Are thefts that occur at certain regions or during specific
seasons correlated to the socio-economic characteristics of
these regions
8 Public domain software libraries
Admittedly, most CTDA techniques have
been developed either for particular disciplines
or even for specific attributes based on the data
properties modeled. Several useful public
domain and commercial software are listed in
the Stochastic Analysis Software Library docu-
mentation (SANLIB, 1995).
 8 Public domain software libraries
More recent and in certain respects more
advanced CTDA software libraries in the public
domain include the following (the vast majority
of the numerical examples and the real-world
case studies in this book have been conducted
using these libraries):
① Spatiotemporal Epistemic Knowledge
Synthesis Graphical User Interface (SEKS-
GUI): This is a library interface that was
developed on a Matlab platform, and it
provides an application for interactive
CTDA analysis. The original space-time
data, processed data, and final results are
all available for users visualization. SEKS-
GUI can be downloaded from:
http://homepage.ntu.edu.tw/hlyu/
software/SEKSGUI/SEKSHome.html.
Example 8.1 The interface of SEKS-GUI is
shown in Fig. 8.1. In the “BME Spatiotemporal
Analysis” part of the software library, the proce-
dure is as follows: import hard data, specify soft
data type and format, import soft data, specify
predict locations or output grid, check for
co-located data, data detrending with Gaussian
FIG. 8.1 Interface of the SEKS-GUI library.
29
kernel (optional) and exploratory analysis, data
transformation (optional), spatial or spatiotem-
poral covariance computation and modeling,
BME estimation and visualization.
② Bayesian Maximum Entropy Graphical
User Interface (BMEGUI): This is a Python-
based GUI implementation for the BMELIB
library, and has a variety of components
for different types of analysis. BMEGUI can
be downloaded from:
https://mserre.sph.unc.edu/BMEGUI_
web/BMEGUI3.0.1_web/BMEGUI3.0.1_
WEB_2014.htm.
③ Space-Time Analysis Rendering with BME
(STAR-BME): This is, also, a Python-based
plugin interacted with Q Geographic
Information System software (QGIS,
https://www.qgis.org/en/site/). It has a
superior ability on rendering visualizations
within the GIS environment, which makes
it easier for mapping purposes. STAR-BME
can be downloaded from:
https://stemlab.bse.ntu.edu.tw/
wordpress/starbme/
30 1. Chronotopologic data analysis

FIG. 8.2 Interface of the STAR-BME library.

Example 8.2 The interface of the STAR-BME
software library is shown in Fig. 8.2. Briefly, the
process introduced by STAR-BME is as follows:
set the coordinate reference system, import hard
and soft data, data detrending with Gaussian ker-
nel or spatiotemporal mean (optional), spatiotem-
poral covariance modeling and estimation,
prediction points, BME prediction and
visualizations.
Ergo, a reasonable suggestion would be that
the readers visit the above sites and decide
which software library best fits their needs.
Moreover, the users of the software tools
should make sure that the underlying assump-
tions and limitations are well understood and
2. One can think of many natural quantities that
not hesitate to question their applicability in
the specific case study, if necessary. It is hoped
that the readers will be sufficiently encouraged
to apply one or more of these software tools into
their CTDA data to make sense of theory
and data.
Concluding, the discussion in this chapter is a
skeletal description of the various CTDA ele-
ments and features, with the understanding that
3. Consider the chronotopologic coordinates
much of the ensuing discussion in Chapters 2–13
will serve to flesh out the skeleton.
9 Practice exercises
1. How a physical law is presented by an
equation matters in determining what
conceptualizations are activated. Thus,
changes in symbolic structure or arrangement
may cause changes in conceptualization.
A difference in the arrangement of the
symbols in an equation could imply a
different organization of the body of
knowledge, which could, in turn, lead to
different interpretations. Newton’s second law
can be represented by two mathematically
equivalent equations, F ¼ mγ and γ ¼ mF , where
F is the total force on an object, m is its mass
and γ its acceleration. Explain why although
their mathematical formulations are the same,
the conceptualization of the two equations
could be quite different.
exhibit chronotopologic variation, such as
rainfall, vegetation, population density,
human mortality, economic wealth,
unemployment and crime rates. Suggest
attributes in natural sciences that
(a) do not vary in space,
(b) do not vary in time,
(c) vary neither in space nor in time.
associated with:
(a) the occurrence of volcano eruption,
(b) soil moisture samples,
(c) household density in the state of California,
(d) migration of swallows from Ireland to S.
Africa, and
(e) SARS disease incidences in East Asia.
 9 Practice exercises
Which ones of the above can be studied by
the CTDA techniques?
4. In one sentence, what is CTDA all about and
what kinds of problems can it solve, in your
view?
5. Describe natural attributes in your field of
interest characterized by chronotopologic
dependence.
6. Consider the following phenomena:
(a) the occurrence of your father’s birth,
(b) pollutant concentration samples,
(c) poverty levels in North Carolina counties,
(d) the battle of Marathon, and
(e) Plague mortality throughout India in the
early 1900s.
Which ones of the above can be studied
by CTDA techniques?
7. With which ones of the following
descriptions do you agree, if any?
There are scientific fields that:
(a) capitalize on chronotopologic depen-
dence to estimate attribute values at
unsampled points,
(b) are concerned with the study of
phenomena that vary in a composite
spacetime domain,
(c) focus on numerical techniques that deal
with chronotopology in conditions of
in-situ uncertainty,
(d) offer a way of describing the continuity
of natural phenomena and include
adaptations of regression techniques
to take advantage of this continuity.
8. In what way do S-TDA and STDA
distinguish themselves from classical data
analysis, and in what way are they
distinct from each other?
9. Fill the empty spaces (__) in the following
statement:
“Compared to the __ statistics that merely
examine the statistical distribution of a set
of sampled data, __ incorporates both the
statistical distribution of the samples and
the chronotopologic correlation among them.
Because of this difference, many real-world
31
problems are more effectively addressed
using __ techniques.”
10. Compare the chronotopologic variability
and dependency of the temperature
distribution in the countryside vs. the
industrial city center.
11. Design an investigation of the
chronotopologic variation of atmospheric
humidity in a geographical region during
a time period of your choice. List sources
that may generate uncertainty.
12. Provide some special situations or data that
cannot be analyzed by CTDA techniques.
13. List uncertainty sources associated with
different types of sampling (random,
regular, stratified, etc.).
14. Compare the variability of daily
temperature in May at a city of your
choice vs. the variability of daily
temperature during the entire
calendar year.
15. Compare the variability of daily humidity
in May at a specific city of your country
vs. the variability of daily humidity in
May at several cities of your country.
16. Provide examples of attribute distributions
that violate Tobler’s law of geographical
science.
17. Write a brief review of CTDA techniques.
18. Which is the branch of philosophy that
concerns itself with knowledge, and why
is it important for CTDA research
purposes?
19. Do you consider implementing CTDA
techniques, and in what specific fields?
20. Should CTDA be concerned with the
representation and analysis of common-
sense knowledge or with solving
problems designed to test research-
motivated axiomatic theories, or with both?
21. Like all types of human inquiry, CTDA too
can be influenced by the underlying
conceptual or methodological perspective.
For illustration, Table 9.1 illustrates some
32 1. Chronotopologic data analysis

TABLE 9.1 Differences between the positivist and the interpretivist approach to human inquiry.
Issue Positivist perspective Interpretivist perspective

Access to real- Direct Indirect

world
External reality Single Not single
Reality studied Objective knowledge Perceived knowledge
through
Focus of human On generalization and abstraction based on On the specific and concrete seeking to
inquiry hypothesis and theory understand the specific context
Investigation On description and explanation On understanding and interpretation
focus
Investigator’s • Detached, external observer • Experiences what is studying
role • Make clear distinction between reason and • Allows feeling and reason to govern actions
feeling. • Partially creates the object of study, the
• Aims to discover external reality rather than meaning of phenomena
create the object of study • Uses preunderstanding (intentional structure
• Use rational, consistent, and logical approach of feelings and thoughts)
• Maintain a clear distinction between facts and • Assumes a less clear distinction between facts
value judgments and value judgments
• Distinguishes between science and personal • Accepts influence from both science and
experience personal experience
Techniques used Predominantly quantitative (mathematical, Primarily nonquantitative methods
statistical) methods

key differences between the positivist and
interpretivist approaches. What kind of
approach do you use in your research?
Justify.
22. Do you agree or not with the view that truth-
seeking is the critical thinking skill that is
defined as the practice of seeking
knowledge, and following reason and
evidence, even if it disagrees with a deep-
stated belief system. Justify.
 C H A P T E R
2
Chronotopology theory
1 Introduction
Let us start by introducing an essential study
perspective pertaining to the main focus of the
present chapter regarding the role of chronos
(time) and topos (space) and that of their relation-
ship in CTDA:
The goal of a theory of chronotopology is to
link chronos and topos in a mathematically rigor-
ous and physically meaningful manner.
Whether the natural attribute of interest con-
cerns human populations or the physical envi-
ronment, chronotopology is of fundamental
importance. Historically, two distinct concep-
tions of chronotopology have been developed,
around which all the other CTDA notions
revolve:
• The classical conception, in which the
chronotopologic domain serves as an arena
for the natural phenomena to take place,
and which is independent from the
phenomena.
• The modern conception, according to which
the chronotopologic domain does not claim
Quantitative Analysis and Modeling of Earth and Environmental Data
https://doi.org/10.1016/B978-0-12-816341-2.00012-5
existence on its own, but it is intimately
linked to the natural phenomena and
cannot exist independently of them.
The implementation of a particular chrono-
topologic framework may require theoretical
and methodological commitments in different
fields of human inquiry, including physics,
mathematics, ontology, epistemology, cog-
nition and sociology. Chronotopology, then,
plays a key role in achieving certain key
objectives.
An operational chronotopology can achieve
certain real-world key objectives: determining
what kinds of natural attribute distributions exist,
assessing how they vary in-situ, describing the rel-
evant body of knowledge, and classifying the
attributes into different scientific disciplines.
In this respect, some initial characterizations
are in order regarding the main constituents
of an operational chronotopology. These
characterizations concern the correspondence
between elements of the material (real or phys-
ical) world and these of the chronotopologic
domain, as follows:
33 Copyright # 2022 Elsevier Inc. All rights reserved.
34 2. Chronotopology theory

❶ A material pointa (or particle) in the
material (or real) world consisting of
continuous matter and an immaterial
(abstract or geometric) pointb representing
only position and time instance in the
chronotopologic domain.
❷ A material body (or object) in the material
world and its abstract configuration in the
chronotopologic domain within which the
configuration resides and moves.
❸ A mobile point in the chronotopologic
domain representing the activity of a
material particle in the material world and
an immobile point in the chronotopologic
domain that merely serves as an attribute
argument (location and instance of the
occurrence of an attribute value).
Example 1.1 Fig. 1.1A shows a point P and a
body in the material world (consisting of a set
of material points) that are mapped into a point
p and a configuration, respectively, in the chron-
otopologic domain, i.e., the configuration of the
body is obtained by fixing the material points of
the body to abstract points on the chronotopolo-
gic domain. The study of this representation (i.e.,
a material point or body residing and moving in
the chronotopologic domain) is the focus of

FIG. 1.1 (A) Elements of the material (or real) world and their mapping in the chronotopologic domain; and (B) attribute
value and its mapping in the chronotopologic domain.

a
I.e., an entity having size (width, no length and no depth).
b
I.e., a sizeless entity in geometry having no size (no width, no length and no depth).
Another random document with
no related content on Scribd:
 The Beggar.
By Jules Bastien-Lepage.
The more he become master of his brush, the more the rustic
work haunted him. He was still a thorough countryman. Although he
had now at intervals the refinements of elegance and little bursts of
worldliness; although he had exchanged the modest atelier in the
Impasse du Maine for a house in the Quartier Monceau, the world
soon wearied him, and he was glad to go back to his village.
This six weeks’ absence, of which he speaks in his letter to his
friend Baude, was spent in an excursion to Venice, and in
Switzerland. He came back only half delighted, and brought back
only a few unimportant sketches.
 Italy and the splendours of Venetian art had left him cold. In this
world of history and mythology he was not at home. He sickened for
his meadows and his Meusian forests.
During his rapid visits to Paris in 1881 and 1882, the painting of
various portraits, notably that of Madame Juliette Drouet, and the
compulsory tax of visits and soirées occupied him almost entirely.
We saw but little of him. But these successes, and the adulation
lavished upon him in Parisian drawing-rooms, did not change him.
He was still the loyal, joyous comrade, faithful to old ties; very
good, very simple; happy as a child when he found himself in a circle
of intimate friends.
We were both members and even founders of an Alsace-Lorraine
dinner, the Dîner de l’Est, which was always given in summer in the
country. One of the last meetings at which he was present, took
place at the end of May, 1881.
A boat had been engaged, which was to take the diners to the
bridge at Suresnes, and to bring them back at night. When we
arrived at the landing-stage, a blind man was standing by the
footbridge, attended by a young girl, who held out her sebilla to the
passers-by.
“Come, gentlemen! all of you, put your hands in your pockets!”
gaily commanded Bastien, and he passed over first, preaching by
example. And the eighty, or a hundred guests of the Dîner de l’Est,
passed one after another over the footbridge, each one leaving in
the child’s sebilla a coin, large or small.
When we were on the deck, Bastien turned round to look at the
blind man and his girl, who were amazed at this unexpected windfall,
and were slowly counting their money.
“What a lovely group?” he said to me. “How I should like to paint
that child!”
While waiting for dinner we walked in the Bois de Boulogne. The
acacias and hawthorns were in flower. The lawns, newly shorn, gave
out a perfume of mown grass. Jules, joyfully drawing in this air
impregnated with country odours, laughed like a happy child.
 At that moment all was going well with him. His Mendiant had had
a great success at the Salon; his last visit to England had been very
prosperous; his head was full of fine projects for pictures. “It is good
to be alive!” he exclaimed, as he played with a flower he had plucked
from the bushes…. On the way back he gave himself up to all sorts
of roguish fun. Mounted on the prow of the boat he sang, with his full
voice, the Chant du Départ.
The vibrating tones resounded powerfully between the two
sleeping river banks; the sky was splendid, twinkling with
innumerable stars. From time to time Bastien lighted a rocket and
sent it up overhead, shouting a loud hurrah!
The fusée mounted slowly into the night, showering down many-
coloured sparks, then fell suddenly and sank in the dark water. Alas!
it was the image of the short and brilliant years that remained for him
to live.
 IV.

On the death of Gambetta, January 1, 1883, Bastien was

commissioned to make a design for the funeral car in which the great
orator was to be conveyed to Père Lachaise; he spent a week in the
little room at Ville d’Avray, painting the picture representing the
statesman on his deathbed. The cold was extreme at this time, and,
his work scarcely finished, he went away, feeling ill, to Damvillers,
where he hoped to finish the great picture he had began of L’Amour
au Village.
His native air, the simple life, and his mother’s loving care
restored him, and he began to work again with his usual eagerness.
Muffled in a warm jacket and a travelling cloak that covered him
down to the feet, he made his models pose for him in the piercing
days of February, in the little garden where he had already painted
the portrait of his grandfather. In March the work was well advanced,
and he invited me to go and see it at Damvillers before it was sent to
the Salon. I left Verdun on a freezing afternoon, accompanied by the
old friend who had walked with us through the Argonne, and we
were set down at Damvillers at night-fall. Our hosts were awaiting us
on the doorstep; the grandfather, always the same, with his Greek
cap and white beard, and his Socratic face; the painter and the little
mother, with smiles and outstretched hands.
Around them Basse the spaniel, and Golo and Barbeau were
bounding and barking joyfully to give us a welcome.
The next morning, early, we went up to the studio to see L’Amour
au Village, which was to go to Paris that day.
The subject of this picture is well known; it is one of the most real
and the most original that the artist has painted: the daylight is
waning; at the gate of a village garden, a lad of twenty, who has
been binding sheaves, and still wears his leggings of leather, is
talking, leaning against a fence, with a young girl, who turns her back
to the spectator; what he is saying to her may be guessed from his
awkward manner of twisting his stiff fingers, and also from the
attentive but embarrassed air of the young girl. One feels that they
are not saying much, but that love exhales from every word, so
difficult to speak. Around them summer spreads the robust verdure
of the country. The fruit trees stand lightly silhouetted against a
background of kitchen herbs, gently sloping up to the houses of the
village, whose brown roofs and pointed spire come against the soft
and misty twilight sky. All this, bathed in a subdued light, is
marvellously painted. The young girl, her short plaits falling over her
shoulders, her neck bent, the form of her back, so young, so
delicate, is an exquisite figure; the face of the young harvester, so
energetic, so ingenuously in love, is charming in expression; the
treatment of the hands, the bust, the dress, is masterly. There is in
this picture a true and manly poetry, which is strengthening and
refreshing, like the odour of ripe corn.
Bastien was glad to have completed this difficult work, and his
satisfaction enabled him to bear with cheerfulness the pains in his
loins, and the digestive troubles which were becoming more and
more frequent.
It was long since I had seen him so gay and unreserved. This
happy holiday-week spent at Damvillers was the pendant to the walk
through the Argonne. The sullen sky, continually blotted out by
chilling showers, allowed us few walks in the open air; but every
morning we went up to the studio. Jules dismissed the little sweep,
who was sitting for a picture that he had on hand, and, taking a sheet
of copper, he made us pose for an etching. I have this plate before
me now; it did not bite well. It represents the whole family, including
the grandfather, making a circle round our friend F., who, standing up
and very grave, is reciting one of La Fontaine’s fables. While I look at
it, I seem to hear again the merry laughter which filled the studio,
alternating with the rattling of the hail against the windows.
In the evening, after supper, we placed ourselves at the round
table, and played at Diable or Nain rouge. Jules, throwing away his
best cards, always managed to let the grandfather win; and when the
octogenarian, quite proud of his success, took up the stakes, he
would pat him on the shoulder, and cry out, with a merry twinkle of
the eye, “Ha! what a lucky man! he will ruin us all!” and the laughter
began again.
We did not go to bed till well on into the night, after having roused
the little domestic, Felix, who had dozed off in the kitchen while
copying a portrait of Victor Hugo.

Father Jacques, the Woodman.

By Jules Bastien-Lepage.
In the intervals of sunshine, Bastien-Lepage took us to visit “his
fields.” He had a peasant’s love for the land, and he employed his
gains in adding to the paternal domains. He had just bought an
orchard situated in the old moat of the town, which had belonged to
an unfrocked priest. He intended to build a châlet there, where his
friends, painters or poets, might come and live in their holidays and
dream at their ease. He explained to us with the delight of a child, his
plans for the future. When, with his portraits, he should have gained
an independent fortune, he would execute at his ease and in
freedom, the grand rustic pictures that he dreamed of, and among
others, that burial of a young village girl, for which he had already
made many notes and sketched the principal details. We only took
one long walk, and it was in those woods of Réville which form the
background of his landscape, Ripe Corn. The weather had remained
cold, and there were still patches of snow on the backs of the grey
hills, though the sun shone sometimes. Except a few downy buds on
the willows, the woods were without verdure; but the ploughed fields
had a beautiful brown colour; the larks sang; the tops of the beeches
began to have that reddish hue, which indicates the rising of the sap,
the swelling buds. “Look,” said Bastien to me, when we were in the
forest, “my Wood-cutter in the last Salon was reproached with want
of air…. Well, here we are in a wood, and the trees are still without
leaves, yet look how little the figure stands out from the undergrowth
of trees and bushes. There is a great deal of routine and prejudice in
that criticism of the perspective of my pictures done in the open air. It
is the criticism of people who have never looked at a landscape,
except crouching down or sitting. When you sit down to paint, you
naturally see things quite differently from the way you see them
standing. Sitting, you see more sky and you have more objects—
trees, houses, or living beings standing out sharply in silhouette
against the sky, which gives the illusion of a greater distance and a
wider atmosphere. But it is not in this way that we generally see a
landscape. We look at it standing, and then the objects, animate or
inanimate, that are nearest to us, instead of being seen in profile
against the sky, are silhouetted upon the trees, or upon the fields,
grey or green. They stand out with less clearness, and sometimes
mix with the background, which then, instead of going away, seems
to come forward. We need to renew the education of our eye, by
looking with sincerity upon things as they are in nature, instead of
holding as absolute truths the theories and conventions of the school
and the studio.”
All the afternoon passed thus happily away in friendly talking and
slow smoking along the wooded paths. The blackbirds were
whistling; from time to time we discovered a flower in the open
spaces, which showed that spring was surely coming; a wood
anemone, with its milk-white petals, or a branch of mezereon, with its
pink flowers opening before the leaves, and its Japanese
appearance.
Jules stopped and gathered a stem of black helebore. “Ah, how
beautiful!” he said. “How one would like to make a careful study of
these leaves—so decorative, so finely cut—of dark green, almost
brown, out of which comes this pale green stem, with its clusters of
greenish flowers edged with pale rose-colour. What lovely forms, and
what a variety of tender shades! This is what they ought to give as a
copy to the children in the schools of design, instead of the eternal
and wearisome Diana de Gabies!”

Sketch for Father Jacques

We did not return till evening, when there was a magnificent
sunset, which crimsoned the smoky roofs of Réville, and made the
light clouds scattered over the sky look like a strew of rose-leaves.
The next day was the last of my visit. We took leave after long
embraces, making fine plans for returning to Damvillers for the
September holiday, while the grandfather, shaking his hoary head,
murmured sadly, “Who knows if you will find me here?” And
Barbeau, and Golo, and Basse bounded and barked round the
omnibus that took us away with tremendous noise.
I did not see Jules again till a month later, at the opening of the
Salon, in front of L’Amour au Village, which had a full success. He
was ill, and complained of pains in the loins more acute than
formerly; then he suddenly disappeared mysteriously. The door of
the atelier in Rue Legendre was closed, and visitors were told that
the painter was gone into the country. We did not know till later that
he had hidden himself, to undergo a sharp and painful treatment,
and that, scarcely convalescent, he had gone to breathe the sea air
in Brittany, at Concarneau. He spent his days there, in a boat,
painting the sea, and forgetting his pains by the help of work.
When he came to see us again in October, he appeared to be
recovered; but digestion was still a difficulty, and his habitual gaiety
was, as it were, clouded over. His character was changed. There
were no more of those trenchant affirmations of which his comrades
sometimes complained; he was indulgent, and even affectionate,
much more than was usual with him. He did not stay long in Paris,
but hastened back to Damvillers, to get seriously to work again. He
arrived in time to be present during his grandfather’s last moments.
The old man departed loaded with years; but, though surely
expected, his death was a painful blow to the survivors. “The house,”
he wrote, “is empty more than one could believe. Only a few days
ago, at any moment, a door would open and the grandfather
appeared, without motive, without object, without speaking or being
spoken to; but the sight of his kindly face was enough. One kissed
him, and he went away, as before, without object, sitting down, going
into the garden, coming back, and always with the same kind face. I
remember now that he has been growing paler for some days…. No,
you can have no idea how empty the house is. I cannot get
accustomed to it. We often talk of him with my mother—with what
pleasure! It is not that we weep for him with tears; we reason about
it, and we appear resigned and courageous; but behind all that there
is a sad feeling of want, of absolute loss. It is the touch one wants….
I have been ill with it, and am so still. I have not been able to work;
to-day, for the first time, I went out to shoot larks; the weather was
fine, the sun was shining, and the country beautiful. This did me
good.”
Indeed, the health of the artist, far from improving, was becoming
daily more uncertain. “It is the digestive tube,” said he, “that is out of
order.” Nevertheless, he worked with his usual courage, overlooking
his Concarneau studies, planning a new picture, and only stopping to
go out shooting or to saunter through the woods.
“Our evening walks are the best part of the day”—(letter to Ch.
Baude, Nov. 27, 1883)—“that is, from the setting of the sun till it is
dark. Every night the spectacle is new. The programme changes with
the weather. Sometimes the subject of the piece is dramatic; the next
day it is soft and charming; and, with the constant rain, our inundated
meadows reflect the brilliant scenery. Can you imagine all our
pleasure, in your dingy Paris? The next morning is too slow in
coming; one wants so much to put down last night’s impression; so
that I am making a heap of sketches, and find much pleasure in it.
Then—here is a surprise!—I have a new picture on the way….
Guess!… The subject is a wounded deer taken by the dogs. The
scene is, naturally, the wood, and the wood at this time of year: only
a few leaves of brilliant yellow against the marvellous rosy-grey of
the branches of the trees; then the violet tone of the dead leaves
flattened on the soil, and a few green briars round a pool under a
willow. The place was not chosen by me. The deer chose it himself
to die there; for I killed him the other day, and he went there to be
taken, a hundred yards from where he was shot—just opposite the
spot where Minet killed a hare. It was then that this picture struck
me. Afterwards I sketched in and reconstructed the scene; and, as I
wanted a model, I killed a second deer….”
Here is a characteristic symptom: he who formerly only wrote the
shortest of notes, scribbled in haste at the corner of a table, now
sent long, expansive letters to his friends, showing signs of
redoubled love of life, of art, of the beauties of nature:—
“My dear friends” (Jan. 3, 1884), “if you could see your poor
Bastien, with this heap of letters to write, you would certainly say:
‘How he is changed!’… If my wishes had the extraordinary virtue of
fulfilling themselves, I should like that you, whom I love, should profit
by it, and that 1884 should bring health and happiness and success
to all. My mother’s wishes are the same as mine, and she rejoices
that we are to see you soon. Ah, my dear friend, what pleasure you
would have in living upon the woods, as I feed upon them now
almost every day, along with Golo and Barbeau! What marvellously
delicate tones! and the fading out of daylight, and when the evening
comes on! The woods are exquisitely fine, with their tall, dry, ivory-
coloured grasses; they are so tall in some of the open spaces that
they caress your face as you pass, and the cool touch upon your
face and hands, hot with walking, is a delicious sensation. I rarely
leave the woods before night, for I must send up a few salutes to the
wild ducks with my gun before going in. One hears them coming
from a great distance, but it is difficult to judge if they are far away or
near, from the peculiarity of their cry; so they have often passed, and
are already a good way off, before one finds out that one has missed
them.
“This is to let you know that I am not a stay-at-home, as you might
think. I find it important to walk a good deal, for in this way I regain a
little health. My stomach was beginning to get wrong, but it is
better!…”
A few days after this I met a mutual friend of ours. “Well,” he said
to me, “our poor Bastien is very ill…. They think it is hopeless.”
 V.

Indeed he was very ill. The treatment he had undergone in the

summer of 1883 had not been successful. The pains in the loins and
bowels had returned with greater violence at the end of January.
By the advice of his friend Dr. Watelet he again went to Paris in
March to consult Dr. Potain. Without any illusions as to the fatal
nature of the disease, the doctors thought that a change of air and of
climate might, morally and physically, produce good results. They
advised that he should go to Algiers for two months.
Bastien himself, seized with that longing desire for movement
which often torments invalids who are seriously ill, had experienced
a wish to go to the south. It was decided that he should start as soon
as possible for Algiers, accompanied by his servant Felix, and by his
mother.
On the morning of the day fixed for starting I went to the Rue
Legendre to say good-bye to him. He had gone to complete some
arrangements with his picture-agent. I found only Mme. Bastien, who
was occupied in filling the trunks which were scattered about the
studio. The brave little mother, who had never left her home at
Damvillers for more than a few days together, was preparing for this
long journey to an unknown country quite simply, with an apparent
tranquillity, as if she were going as far as Saint Cloud.
The hope that the change might be good for Jules was enough to
give her courage to face this upsetting of all her old ways of living.
Sometimes only, when she was carefully arranging the linen in the
trunk, the tears would rise to her eyes and a quiver of pain pass over
her lips.
Upon the chairs and against the walls were placed the recent
studies brought from Damvillers, and one felt one’s heart tighten at
the sight of these last works, where nature had been observed and
rendered with incomparable skill, penetration, and charm. They were
The Frog-fisher, The Little Sweep, The Washerwoman, The Pond at
Damvillers, The Edge of the Wood, The Church at Concarneau, and
that study of A Midnight Sky so original, with the clouds scattered
over an azure that was almost black.
At this moment Bastien-Lepage came in, and on seeing him walk
with difficulty into the studio, I was distressed at the change that had
come over him. His thin face had become quite bloodless; the skin of
his neck was peeling off; his hair seemed to have no life in it. His
questioning blue eyes expressed an anguish and weariness that was
heartrending. “Well,” said he, after having embraced me, “are you
looking at my studies? When people see them at George Petit’s,
they will say that the little Bastien could paint the landscape too,
when he gave himself the trouble!…” When I said to him that his long
absence that morning had made his mother anxious, he added quite
low, and taking me into one corner of the studio: “When one is going
to take a journey so far, one must prepare for it…. I wanted to put my
affairs in order. Poor little mother!” he went on; “she has been very
brave! Down at home she used to spend whole nights in rubbing me
for my rheumatism, and I let her think that it did me good…. Now,
perhaps the Algiers sun will cure me.” Hope alternated with
discouragement. During breakfast he recovered a little. I was to go to
Spain at the end of March; he urged me to change my plans, and to
join him in Algiers. We ended with a half-promise. We tried hard to
appear gay; we clinked our glasses as we drank to the hope of soon
meeting again, but each one felt his throat tighten, and turned away
to hide from the other his moist eyes. I left the house in the Rue
Legendre with my heart full of the saddest forebodings.
Jules left the same night for Marseilles. They had a good
crossing, and his first letter, dated March 17th, was reassuring:—
“My dear friends, there is no getting out of it; you must come, for a
thousand reasons. Here it is just like May in Paris. Everything is in
flower; and such flowers!—heaps of them, everywhere. The verdure
is delicate and grey, and, like patches, always well placed; the
outlines picturesque and new, the trees very dark green. And in the
midst of all this, upon the roads, the Arabs, of astonishing calmness
and splendid carriage, under their earth-coloured and ash-coloured
draperies—ragamuffins as proud as kings, and better dressed than
Talma. They all wear a shirt and burnous; not one is like another. It
seems as if each one, at every moment, gave expression to his
thought by his manner of draping his garment. It is once more the
triumph of blank truth over arrangement and conventionalism. The
sorrowful man, whether he wishes it or not, in spite of himself is not
draped like the gay. Beauty, I am convinced, is exact truth: neither to
the right nor to the left, but in the middle.
“All this without telling you we have hired a house at Mustapha
Superior. It is half Arab, half French, quite white, with an interior
court opening into a garden twice as big as that at Damvillers. The
garden is full of orange-trees, and lemon, almond, fig, and a quantity
of other trees, the names of which I do not know and probably never
shall. All this, not trim like a park, but left a little à la diable, like our
garden at home. Then we have the right of walking in a magnificent
garden which joins ours. We have at least eight rooms; in counting
them I thought of you. In all directions round this house there are
delightful walks within reach for invalid limbs; in short, it is a
Mahomet’s Paradise, … ‘moins les femmes.’ I have said nothing
about Kasbah, the old Arab town—my legs have only let me see it
from a distance as yet; but, my good friend, imagine that against a
morning sky you have, sometimes in the palest rose, sometimes in
silvery grey, sometimes in faint blue, and so on—everywhere against
the pearly sky—more or less elongated rectangles, placed
irregularly, but always horizontally, in the manner of a line of low hills,
and you will have the delicate colouring of the old town. One would
not suppose it was a town with habitations, so delicate is the tone of
it, but for some little holes of rare windows placed here and there.
One could not have a sensation more unexpected, and never a
sweeter and finer joy. So you must come! My mother is counting
upon it, and what, then, am I? What new things you could say about
all this! The sea was very fine at the beginning and end of our
crossing. Midway some of the passengers suffered: my mother and
Felix among them, but they got some sleep. We were twenty hours
in crossing, and we were not tired on arriving. Come, set off; start!…
A good embrace from my mother and from me.”
His first letter, as may be seen, was full of ardour. The climate of
Algeria did him good at first, and his sufferings seemed to be
relieved.
“I am preparing myself bravely for the ordeal by fire” (April letter to
Ch. Baude); “may my rheumatism take flight and depart with the
coming attack of the sun! When it is hot here, it is still quite bearable.
Apart from these calculations about the heat and these health
experiences, I am happy, even excited, by all that I have seen; and
yet I have only seen what any bagman might see who is busy about
the selling of his goods; but it has been enough to give me great
delight. What remains of the old Arab town is marvellous; one holds
one’s breath when, at a sudden turn, the vision reappears. For those
unhappy eyes that only see the colours on the palette, it is white; but
picture to yourself a long hill, rather high, with a depression in the
middle, and sloping as if to the sea, and this hill all covered with
elongated or elevated cubes of which one cannot distinguish the
thickness; all this remaining unnoticed by the eye that is ravished by
the delicate tone, rosy, greenish, pale blue, making altogether white
tinted with salmon.
“If one did not know it beforehand, one would never dream that
amongst these cubes of plaster thousands of men are walking,
talking, sleeping—men of noble manner, proud and calm, and with
something very like indifference or contempt for us. And they are
right. They are beautiful, we are ugly. What matter is it to me that
they are knaves! They are beautiful!…
“Yesterday I went to take a bath. I had to go three or four hundred
steps through streets full of merchants. In a passage a Jew was
selling silks, pearls and corals; in front of his shop, not two yards
wide, were three Arabs—an old man, another of middle age, the
third about seventeen. There they were, seated, attentive, calm,
wishing to buy, consulting together, making scarcely a gesture with
their hands, always kept at full length, but sitting quietly, never
hurrying, reflecting enormously, and keeping all the while under their
burnouses the softest, gentlest attitudes. The youngest was superb
—so handsome that mama was struck with it. ‘They are like beautiful
statues,’ said she. I could not understand the scene and the relations
that united these three Arabs. It was clear they were come to buy;
they had come down from the higher part of the town. They were
poor, for the youngest was in rags, and the burnouses of the others,
though not in rags, were very much worn; but they took such pains in
counting the little pieces of false coral that it was clear the Jew was
selling dear to these big children a thing of no value. The one of
middle age was counting on the table, with his flat hand by groups of
five, the little pieces of coral which he chose as he counted them;
thus adding each time five pieces to the heap that he drew towards
him.
“What strikes one is this simple colouring, these magnificent folds,
and then this serious childishness.
“I was not able to wait till the end of the scene. It was cold and
draughty in this passage, which brought me back to the fact of my
poor crazy legs. I long for the time when I shall be a man again; what
lovely things I shall see, and perhaps I shall do!”
April 23rd (to the same): “Now I take myself by the ear and drag
myself to the letter-paper, and all the needful things. Nothing is
wanting, neither the thousand things I have to say, nor above all the
tender affection that I keep in store for you.
“Emile says that you are coming, and soon: don’t be alarmed, you
will not melt in the hot sun. There are cool places in the garden,
where one can stretch oneself, with a magnificent landscape at one’s
feet. We have only had the heat since yesterday; you will see how
good you will find it, your muscles will relax, and you will go back
quite young. We will make some excursions together if I am up to it.
Any way there are plenty all round us to tempt you to make some.
“You have heard from Emile that I went to Blidah. I bore the little
journey very well at first, but I was tired afterwards. I am going to
begin to rest, and go slowly, in order that I may go farther. I have
scarcely done anything till now, for I don’t feel myself up to remaining
long in the same position, as a painter must, who thinks only of his
work.”
The health that he hoped for, and so anxiously waited for, did not
come. On the contrary, as the heat increased, Jules felt more unwell
and more fatigued. The last letter that he wrote to me reached me at
Granada, in that hotel, the “Siete Suelos,” where Fortuny and Henri
Regnault had lived. There was all through it a sentiment of touching
melancholy and discouragement.
“My good friends, this is delightful. It is too good to get your
photographs at the same time as your kind and affectionate letter. I
am glad you are going to Spain. Lucky fellows! Go along! while I,
who should so like to see a bull fight!… You had not time to come,
and indeed it was selfish to ask you. You could not have stayed more
than a few days. But that is to be done some day when I am no
longer a cripple, and when we can have two months before us. We
are comfortably settled here. At this moment I am writing to you
under the tent set up in the terraced court of our villa, with a
wonderful view before me. Placed a little to the left of a semicircle,
formed by the hills of Mustapha, 170 yards above the sea which
flows at their base, we have at every hour of the day, a different
landscape; for the sides of the hills are full of ravines, and the sun,
according to the time of day, throws their slopes into light, or makes
a network of shade, in a way quite peculiar to this corner of Africa.
Little villas gleaming in the sunshine or grey in the shade give effect
to the groups of verdure, the whole looking from the distance like a
rich embroidery, with bosses of green harmoniously arranged. All
this runs down toward the Gulf of Algiers, and trending away from
here forms Cape Matifou. Above are the crests of the Little Atlas, far
away, and lost in heaven’s blue; near by, sloping gardens spread out
their golden or silvery verdure, according as one looks upon olive or
eucalyptus. Add to this the perfume of the orange and lemon trees,
the pleasure of telling you that I embrace you all three, Tristan
included, that I am a little better, and you will have the state of my
heart.
“Enjoy yourselves,—and you, my dear forester, with your Toledo
eyes, what are you going to give to the world after all this delight of
sunshine and kindly fellowship and the loving union of the charming
trio that you make? It seems to me I have the heart and voice to
make a fourth—what say you? Ah! that shall be after the
rheumatism! Kindest regards from mama and from me. A last
embrace to all three of you.”
The improvement he had experienced on arriving in Algiers
ceased about the end of April. His strength and appetite gradually
failed; and at the end of May it was decided to take the invalid back
to France. He settled again in the Rue Legendre with the poor little
mother, who never left him afterwards. When I saw him again I was
shocked at the progress the disease had made. His thinness was
such that my unhappy friend was nowhere in the garments that were
made for his journey. His legs refused their service; he could no
longer work; and yet he kept a little hope. He had just begun a new
treatment, and talked of going into Brittany “as soon as he was
strong enough.” He drove every day in the Bois when the weather
was fine, and spent the rest of his day on cushions in the corner of
the studio, occupied in contemplating, with a heartrending look, his
studies hanging on the walls. This inaction was most distressing to
him.
“Ah!” cried he, “if I was told: They are going to cut off your two
legs, but after that you will be able to paint again, I would willingly
make the sacrifice….”
He could only sleep now with the help of injected morphine, and
he waited with impatience for the hour when a new supply should
give him some relief, and a factitious drowsiness should make him
forget his suffering.
In proportion as digestion became more difficult his appetite
became more capricious. He wanted to have dishes made which
reminded him of the cooking of his village; then, when they were
brought to him, he turned away disgusted, without tasting them.
“No,” said he, pushing aside the plate, “that’s not it; to have it good it
must be made down there, prepared by the Damvillers people, with
home-grown vegetables.” And while he was speaking one saw by his
moist eyes a sudden and painful calling up of the impressions of
former days; he saw all at once the old home, the gardens and
orchards of Damvillers at the fall of evening, the peaceful village
interiors at the time when the fires were lighted for the evening meal.
As the season advanced his strength decreased. In September
his brother was obliged to take him on his back to carry him to the
carriage, and he drove about slowly for an hour in the avenues of the
Bois. He could not read, and was easily wearied by conversation.
His nerves were become very irritable, and the slightest odours were
disagreeable to his sense of smell. His courage seemed to forsake
him; at the same time he was always wanting to know what others
thought of his illness. His blue eyes with their penetrating look
anxiously searched the eyes of his friends, and of his mother, who
never left his side. The heroic little woman did her best to
dissimulate, and was always smiling and affecting a cheerfulness
and a confidence which were painful to see; then, when she could
escape for a moment, she hastened into the neighbouring room and
melted into tears.
For months this cruel agony was thus prolonged. Bastien was
only a shadow of himself. On the 9th of December, during great part
of the night, he talked of Damvillers with his mother and his brother.
Then at about four in the morning he said to them, with a kiss,
“Come, it is time for children to sleep.” All three slept. Two hours later
Mme. B. was awakened by Jules, who asked for something to drink;
she rose, and brought him a cup of tea, and was alarmed on finding
that the invalid groped for the cup to guide it to his lips; he could no
longer see; but he still spoke and even joked about the difficulty he
had in moving his limbs.
Shortly afterwards he dozed, and sliding gently from sleep into
death, he expired at six in the evening, December 10, 1884.
I saw him next day lying on his mortuary bed, in the midst of a
thick covering of flowers. His poor emaciated face, with its sightless
and deeply sunk orbits, made him look like one of those Spanish
figures of Christ, fiercely cut in wood by Montanez.
On the 12th of December a long train of friends and admirers
accompanied his remains to the Eastern Railway Station, whence it
was conveyed to the Meuse. The next day, Sunday, the whole
population of Damvillers waited at the entrance of the town for the
funeral carriage, which brought back Bastien-Lepage to his native
place.
The sad procession advanced slowly on that road from Verdun
where the painter had loved to walk at twilight, talking with his
friends. A pale mist blotted out those hills and woods whose familiar
outlines he had so often reproduced. The cortège stopped before the
little church where he had intended painting his Burial of a Young
Girl. The morning was showery; the wreaths and festoons of flowers,
placed the night before on his coffin, were revived and refreshed by
the moisture; when they were heaped up upon the grave they
seemed to come to life again, and to send out with their renewed
perfume a last adieu from Paris to the painter of the peasants of the
Meuse.