22601
22601
com
https://textbookfull.com/product/petascale-analytics-large-
scale-machine-learning-in-the-earth-sciences-1st-edition-
ashok-n-srivastava/
OR CLICK HERE
DOWLOAD EBOOK
https://textbookfull.com/product/large-scale-data-analytics-chung-yik-
cho/
textbookfull.com
https://textbookfull.com/product/big-data-analytics-for-large-scale-
multimedia-search-stefanos-vrochidis/
textbookfull.com
https://textbookfull.com/product/machine-learning-paradigms-advances-
in-learning-analytics-maria-virvou/
textbookfull.com
https://textbookfull.com/product/predictive-analytics-with-microsoft-
azure-machine-learning-barga/
textbookfull.com
Large Scale Data Handling in Biology Karol Kozak
https://textbookfull.com/product/large-scale-data-handling-in-biology-
karol-kozak/
textbookfull.com
https://textbookfull.com/product/recent-developments-in-machine-
learning-and-data-analytics-ic3-2018-jugal-kalita/
textbookfull.com
https://textbookfull.com/product/migrating-large-scale-services-to-
the-cloud-1st-edition-eric-passmore/
textbookfull.com
https://textbookfull.com/product/paleoclimatology-frontiers-in-earth-
sciences-1st-edition-gilles-ramstein/
textbookfull.com
https://textbookfull.com/product/feature-engineering-for-machine-
learning-and-data-analytics-first-edition-dong/
textbookfull.com
Large-Scale Machine Learning
in the Earth Sciences
Chapman & Hall/CRC
Data Mining and Knowledge Discovery Series
Series Editor: Vipin Kumar
University of Minnesota
Department of Computer Science and Engineering
Minneapolis, Minnesota, U.S.A.
PUBLISHED TITLES
Accelerating Discovery: Mining Unstructured Information for Hypothesis Generation
Scott Spangler
Machine Learning and Knowledge Discovery for Engineering Systems Health Management
Ashok N. Srivastava and Jiawei Han
Mining Software Specifications: Methodologies and Applications
David Lo, Siau-Cheng Khoo, Jiawei Han, and Chao Liu
Edited by
Ashok N. Srivastava
Ramakrishna Nemani
Karsten Steinhaeuser
MATLABⓇ is a trademark of The MathWorks, Inc. and is used with permission. The MathWorks does not warrant the
accuracy of the text or exercises in this book. This book’s use or discussion of MATLABⓇ software or related products does not
constitute endorsement or sponsorship by The MathWorks of a particular pedagogical approach or particular use of the
MATLABⓇ software.
CRC Press
Taylor & Francis Group
6000 Broken Sound Parkway NW, Suite 300
Boca Raton, FL 33487-2742
For permission to photocopy or use material electronically from this work, please access www.copyright.com
(http://www.copyright.com/) or contact the Copyright Clearance Center, Inc. (CCC), 222 Rosewood Drive, Danvers, MA 01923,
978-750-8400. CCC is a not-for-profit organization that provides licenses and registration for a variety of users. For organizations
that have been granted a photocopy license by the CCC, a separate system of payment has been arranged.
Trademark Notice: Product or corporate names may be trademarks or registered trademarks, and are used only for identifica-
tion and explanation without intent to infringe.
Foreword .............................................................................................................. ix
Editors ................................................................................................................. xi
Contributors ........................................................................................................xiii
Introduction ......................................................................................................... xv
vii
viii Contents
9 Semantic Interoperability of Long-Tail Geoscience Resources over the Web ......... 175
Mostafa M. Elag, Praveen Kumar, Luigi Marini, Scott D. Peckham, and Rui Liu
Index ......................................................................................................201
Foreword
The climate and Earth sciences have recently undergone a rapid transformation from a data-poor to a
data-rich environment. In particular, massive amounts of climate and ecosystem data are now available
from satellite and ground-based sensors, and physics-based climate model simulations. These information-
rich data sets offer huge potential for monitoring, understanding, and predicting the behavior of the Earth’s
ecosystem and for advancing the science of global change.
While large-scale machine learning and data mining have greatly impacted a range of commercial appli-
cations, their use in the field of Earth sciences is still in the early stages. This book, edited by Ashok
Srivastava, Ramakrishna Nemani, and Karsten Steinhaeuser, serves as an outstanding resource for any-
one interested in the opportunities and challenges for the machine learning community in analyzing these
data sets to answer questions of urgent societal interest.
This book is a compilation of recent research in the application of machine learning in the field of Earth
sciences. It discusses a number of applications that exemplify some of the most important questions faced
by the climate and ecosystem scientists today and the role the data mining community can play in answer-
ing them. Chapters are written by experts who are working at the intersection of the two fields. Topics
covered include modeling of weather and climate extremes, evaluation of climate models, and the use
of remote sensing data to quantify land-cover change dynamics. Collectively, they provide an excellent
cross-section of research being done in this emerging field of great societal importance.
I hope that this book will inspire more computer scientists to focus on environmental applications, and
Earth scientists to seek collaborations with researchers in machine learning and data mining to advance
the frontiers in Earth sciences.
ix
Editors
Ashok N. Srivastava, PhD, is the vice president of Big Data and Artificial Intelligence Systems and the
chief data scientist at Verizon. He leads a new research and development center in Palo Alto focusing on
building products and technologies powered by big data, large-scale machine learning, and analytics. He
is an adjunct professor at Stanford University in the Electrical Engineering Department and is the editor-
in-chief of the AIAA Journal of Aerospace Information Systems. Dr. Srivastava is a fellow of the IEEE, the
American Association for the Advancement of Science (AAAS), and the American Institute of Aeronautics
and Astronautics (AIAA).
He is the author of over 100 research articles, has edited four books, has five patents awarded, and has
over 30 patents under file. He has won numerous awards including the IEEE Computer Society Technical
Achievement Award for “pioneering contributions to intelligent information systems,” the NASA Excep-
tional Achievement Medal for contributions to state-of-the-art data mining and analysis, the NASA Honor
Award for Outstanding Leadership, the NASA Distinguished Performance Award, several NASA Group
Achievement Awards, the Distinguished Engineering Alumni Award from UC Boulder, the IBM Golden
Circle Award, and the Department of Education Merit Fellowship.
Dr. Ramakrishna Nemani is a senior Earth scientist with the NASA Advanced Supercomputing Division
at Ames Research Center, California. He leads NASA’s efforts in ecological forecasting to understand the
impacts of the impending climatic changes on Earth’s ecosystems and in collaborative computing, bringing
scientists together with big data and supercomputing to provide insights into how our planet is changing
and the forces underlying such changes.
He has published over 190 papers on a variety of topics including remote sensing, global ecology,
ecological forecasting, climatology, and scientific computing. He served on the science teams of several
missions including Landsat-8, NPP, EOS/MODIS, ALOS-2, and GCOM-C. He has received numer-
ous awards from NASA including the Exceptional Scientific Achievement Medal in 2008, Exceptional
Achievement Medal in 2011, Outstanding Leadership Medal in 2012, and eight group achievement
awards.
Karsten Steinhaeuser, PhD, is a research scientist affiliated with the Department of Computer Science
and Engineering at the University of Minnesota and a data scientist with Progeny Systems Corporation.
His research centers around data mining and machine learning, in particular construction and analysis
of complex networks, with applications in diverse domains including climate, ecology, social networks,
time series analysis, and computer vision. He is actively involved in shaping an emerging research area
called climate informatics, which lies at the intersection of computer science and climate sciences, and his
interests are more generally in interdisciplinary research and scientific problems relating to climate and
sustainability.
xi
xii Editors
Dr. Steinhaeuser has been awarded one patent and has authored several book chapters as well as numer-
ous peer-reviewed articles and papers on these topics. His work has been recognized with multiple awards
including two Oak Ridge National Laboratory Significant Event Awards for “Novel Analyses of the Simu-
lation Results from the CCSM 3.0 Climate Model” and “Science Support for a Climate Change War Game
and Follow-Up Support to the US Department of Defense.”
Contributors
xiii
xiv Contributors
Our society faces an unprecedented challenge in understanding, modeling, and forecasting Earth sys-
tem dynamics that involve complex interactions among physical, biological, and socioeconomic systems.
Addressing these challenges requires the ability to incorporate potentially massive data sets taken at mul-
tiple physical scales from sensors that are geospatially distributed, which can provide information at
different sampling intervals. In addition, computer simulations generate data that need to be reconciled
with observational data at different spatial and temporal scales.
In recent years, researchers in Earth science, computer science, statistics, and related fields have devel-
oped new techniques to address these challenges through interdisciplinary research. The purpose of this
book is to provide researchers and practitioners in these fields, a broad overview of some of the key chal-
lenges in the intersection of these disciplines, and the approaches that have been taken to address them.
This highly interdisciplinary area of research is in significant need for advancement on all fronts; our hope
is that this compilation will provide a resource to help further develop these areas. This book is a collection
of invited articles from leading researchers in this dynamic field.
The book begins with the chapter entitled, “Network Science Perspectives on Engineering Adaptation to
Climate Change and Weather Extremes” by Bhatia and Ganguly. The authors observe that extreme climate
and weather events can be correlated and can cause significant stress on human-built and natural systems,
which in turn can lead to negative impacts on human populations and the economies in which they live.
They discuss in detail the idea of teleconnections, in which spatiotemporal events spanning enormous
geographical distances can show high degrees of correlation. Thus, extreme climate and weather events in
one location can induce other potentially extreme events in a distant location. The authors emphasize the
use of methods developed by network science to understand and discover such teleconnections.
Chapter 2, entitled “Structured Estimation in High Dimensions: Applications in Climate,” continues the
theme of making predictions based on observational data. Goncalves and Banerjee focus on two specific
problems: first, the authors address the problem of making temperature and precipitation predictions over
nine landmasses given measurements taken over the oceans. The authors show that by imposing known
structure on the prediction variables, the prediction quality can exceed prior baselines. The second focus of
the chapter is on combining predictions made by global climate models using a multitask learning approach
based on structured learning yielding predictions, which significantly exceed competing methods.
Chapter 3, by McQuade and Monteleoni entitled, “Spatiotemporal Global Climate Model Tracking”
describes the use of ensemble machine learning models to combine predictions of global climate models
using information from spatial and temporal patterns. They explore the fixed-share algorithm and provide
an extension that allows the prediction weights of the global climate models to switch based on both spatial
and temporal patterns. This approach has thematic commonality with the preceding chapter, although it
addresses the problem using Markov random fields.
Chapter 4 features a discussion on “Statistical Downscaling in Climate with State-of-the-Art Scalable
Machine Learning” in which Vandal, Bhatia, and Ganguly explore the important problem of projecting
xv
xvi Introduction
the output of simulations made at low spatial resolution to higher spatial resolutions necessary for study-
ing the impacts of climatic changes on socioeconomic and ecological systems. They discuss techniques
such as sparse regression, Bayesian methods, kriging, neural networks, and other transfer function-based
techniques to address the problem. In addition, the authors discuss deep belief networks and provide a
comparison of this method against other methods on a precipitation prediction problem.
In Chapter 5, Johnson and Chawla provide an overview of methods to understand and predict the
proliferation of biological species due to changes in environmental conditions. In their chapter entitled,
“Large-Scale Machine Learning for Species Distributions,” the authors provide an overview of this pressing
problem and discuss correlative methods to predict species distributions based on environmental variables.
These methods are contrasted with causal approaches that rely on significant domain knowledge regarding
the species in question and the impact of environmental variations on those species. The authors provide
a theoretical framework for the correlative approaches and show how this problem can be reduced to a
familiar classification problem.
Chapter 6 addresses the problem of “Using Large-Scale Machine Learning to Improve Our Under-
standing of the Formation of Tornadoes.” In this chapter, McGovern, Potvin, and Brown show how
spatiotemporal relational random forests can be used to analyze massive four-dimensional data sets to
detect and potentially differentiate storms that will lead to tornadoes from those that do not. The challenge
that the researchers address is multifold due to the fact that the data sets are massive while the processing
must be done in near real-time to yield actionable results. In addition to trying to address the actionability
of the results, the authors also address the problem of providing potential explanations for tornadogenesis.
Ganguly et al., utilize a critical innovation in neural networks in Chapter 7 entitled, “Deep Learning for
Very High-Resolution Imagery Classification.” The authors address the problem of classifying images that
have very high resolution using a class of deep learning algorithms. The authors show the performance
of stacked autoencoders, deep belief networks, convolutional neural networks, and deep belief neural net-
works with a special set of features derived from the underlying data used as inputs. The study is performed
on a modern cloud-based computing platform and the NASA Earth Exchange (NEX) platform and reveals
that the feature extraction method coupled with the deep belief neural network has a significantly higher
classification performance compared to the other methods.
Chapter 8 by Kumar et al., addresses the critical problem of unmixing spectral signals in remote sensing
images of land cover. By the inherent nature of remote sensing data, any given image pixel, regardless of the
imaging resolution can be composed of many underlying sources—grass, trees, cement, water, etc. Land
cover classification techniques can be improved significantly through the application of algorithms that
use principled methods to decompose the spectral signals to reveal the underlying sources. This chapter
features a detailed analysis of numerous unmixing algorithms followed by a comprehensive set of results
across several data sets. Given the proliferation of medium- to high-resolution global satellite data sets,
these algorithms will play a key role in the monitoring of our planet for detecting changes in forest cover,
urbanization, and surface water.
Chapter 9 addresses the “Semantic Interoperability of Long-Tail Geoscience Resources over the Web.”
In this chapter, Elag, Kumar, Marini, and Peckham apply long-tail distributions, which are a well-known
concept in business, economics, and other disciplines in the context of numerous small, highly curated
data sets that an individual scientist may have on their local network. While each data set by itself has a
certain scientific value, the authors present the idea that the totality of the information contained in union
of all of these data would present an invaluable source of information across the Earth sciences. For such
data sets to be easily accessible, the authors discuss methods to improve the interoperability of the data sets
using semantic web mining. They provide concrete examples of potential approaches to embed context
and other relevant information in a geosemantics framework.
The editors are thankful to the authors of this book who have given a detailed discussion of press-
ing research issues covering a diverse array of topics ranging from combining global climate models to
new methods for creating semantic interoperability across data sets, large and small. We believe that these
Introduction xvii
contributions will form an important resource for interdisciplinary research in the Earth and computer sci-
ences. Developing a deeper understanding of the issues facing our planet through the pursuit of scientific
inquiry based on observable data is more pressing now than ever before.
MATLABⓇ is a trademark of The MathWorks, Inc. and is used with permission. The MathWorks does not
warrant the accuracy of the text or exercises in this book. This book’s use or discussion of MATLABⓇ
software or related products does not constitute endorsement or sponsorship by The MathWorks of
a particular pedagogical approach or particular use of the MATLABⓇ software.
1
Network Science
Perspectives on
Engineering Adaptation
to Climate Change and
Weather Extremes
1.1 Introduction.........................................................................1
1.2 Motivation ...........................................................................2
1.3 Network Science in Climate Risk Management .........................3
New Resilience Paradigm
1.1 Introduction
Weather and hydrological extremes, which may be exacerbated by climate variability and change [1,2],
severely stress natural, engineered, and human systems. What makes these climate hazards particularly
worrisome is their rapidly changing nature, along with our lack of understanding of the hazards attributes
that may matter the most for impact analyses. With climate change as threat multiplier, the weather
extremes have devastating impacts on essential functionality of infrastructure networks, specifically, life-
lines including transportation networks, power-grid, water distribution networks, and communication
networks [3]. The failure of one of the lifeline systems may cause massive social disruption, and localized
damage may trigger a cascade of failure within the network or across network multiplex [4,5]. Environ-
mental impacts of such infrastructures also span spatial scales from regional to national to global. For
example, fossil fuel combustion in power generation facilities can result in heat island effects, all of which
can have profuse impact on humans and biodiversity in the natural systems. The impacts of climate hazards,
and hence preparedness and management of natural hazards as well as climate adaptation [6] and to a
great extent mitigation, crucially depend upon our understanding of the interdependencies of the complex
systems. While component-based risk management frameworks have long been used to inform the dis-
aster mitigation and management, there is a growing realization that the conventional risk management
framework should be coupled with a system-based resilience approach that can confront the ever-growing
complexity and interdependencies of infrastructure systems. Challenges stem from the attributes of the
data, the systems under consideration, and the nature of the problems. One concern is the ability to deal
with “Big Data,” typically defined through attributes such as volume, variety, and velocity [7]. The interest in
extremes or large changes, often unprecedented, requires the extraction of signatures or precursors of rare
1
2 Large-Scale Machine Learning in the Earth Sciences
events and anomalies. Interrelationships among systems motivate the need for correlative and predictive
analyses [8], to understand processes such as spatiotemporal complexity, cascading failures [9], and flow
or spreading phenomena [10]. A combination of the best-available methods in data sciences (e.g., machine
learning), process models (e.g., agent-based systems), and physics models (e.g., global climate models) is
required to address these challenges. Machine learning in particular needs to be able to look for extremes,
anomalies, and change from massive and diverse (“Big”) data while being aware of complex dependence
structures. This has motivated recent advances in the machine learning literature in areas such as rare
event analytics and extreme value theory, network science, and graphical models and algorithms specif-
ically designed for massive spatiotemporal data. Complex networks, including their theory and efficient
computational implementations [11], are among the tools that are increasingly being used by theoretical
and applied machine learning researchers to solve the urgent societal problems discussed in this chapter.
1.2 Motivation
Unprecedented end correlated weather extremes, and drastic regional climate changes, are no longer sur-
prising; legacy infrastructures are crumbling, while multilayer lifeline networks magnify stresses through
percolation within and cascading failures across layers; natural environments from wetland to marine
ecosystems are degrading beyond the stage when they can protect coastlines or cities, or recover from bio-
diversity loss, while population and assets continue to be moved in vulnerable urban or coastal regions [12];
and indirect multiplier effects can be large given increased connectivity [13]. Crucial issues that need to
be understood in this context are generation and/or exacerbation of extreme stresses, resilience (robust-
ness and recovery) of the interconnected built and natural systems to these stresses, and consequences
of the stresses and systemic resilience elements across impacted natural, infrastructural, and societal sys-
tems. Key knowledge gaps include understanding the generation processes [14] and statistical attribute
of the extremes, potential changes in the nature of stressors and/or stressed and impacted systems, and
uncertainties (including intrinsic variability) associated with our understanding of these entities.
In the context of climate-related hazards, there has been increasing recognition of spatial and spa-
tiotemporal correlation [15] and tail dependence among extreme events. Therefore, understanding the
dependence structure among extreme events (such as large-scale extremes and anomalies and intense local
storms, global warming, and extreme precipitation events [16]) is important to understand impacts of
these stressors on infrastructure and natural systems (stressed systems), and human lives and economies
(impacted systems). Correlation and interdependency in climate systems over large spatial scales are char-
acterized by teleconnection patterns, which refer to a recurring and persistent large-scale pressure and/or
circulation pattern that spans over vast geographical area [17]. The organization of teleconnections plays a
crucial role in stabilization of climate patterns, and hence delineating the global changes in architecture of
these patterns could be useful to identify the signature of global climate change. Temporally, climate sys-
tems are affected by the slow responding subsystems such as oceans, and hence climate variability usually
exhibits long-term memory, which means the present climate states may have long-term influences on the
states in far future [18]. Thus, the dominance of long-term processes suggests that any future exceedance
in threshold of sea-level rise regionally might be exceeded earlier or later than from anthropogenic change
alone as a result of melting glaciers and snowcaps [19]. Also, the simultaneous occurrence of extreme
events, such as simultaneous storms and floods at different locations, has serious impact on risk assessment
and mitigation strategies.
A strong feedback that exists between the weather extreme-induced stresses, and built and natural sys-
tems severely impacts these systems [20], resulting in loss of their essential functionality. These impacts
are further exacerbated by aging of critical infrastructure systems and lifelines, habitat loss and mass
extinctions in ecosystems as a result of increasing urbanization, and high population pressure. Infras-
tructure systems, which are “backbone” of prosperity and quality of life in modern societies, are highly
networked in nature, and failure of some specific components can result in the disruption of entire system’s
functionality via cascading failure [21]. Hence, failure in one system can percolate across the multiple sys-
tems, resulting in the extreme state of disruptions by “triggering a disruptive avalanche of cascading and
Network Science Perspectives on Engineering Adaptation 3
Stressed systems:
natural environment
Stressors
Stressed systems:
built environment
Impacted systems:
society
FIGURE 1.1 Strong interdependencies exist between the stressors (e.g., hurricane, shown in the figure), stressed sys-
tems (e.g., ecosystems and lifeline networks for coastal megacities), and the impacted systems (e.g., urban communities
and financial institutions).
escalating failures” [4,5]. Similarly, loss of a small fraction of species from a natural ecosystem causes a
series of extinction of other dependent species, putting these systems at the risk of ecological community
collapse [22].
The climate- and weather-related extremes not only expose built environments to multiple hazards but
also bring about myriad consequences for humans and society. Community structure and connectivity
in the society are the key factors that determine how different shocks and disruptions, both natural and
man-made, impact the society. With the arrival of new tide of “soft technologies” [23], the societies are
getting less connected physically and strongly connected virtually. While closely operating societies may
be more adaptable and resilient to these shocks [24], the vulnerability of distant people is strongly corre-
lated through global environmental changes, global market linkages, and flows of information as shown in
Figure 1.1.
Past few decades have witnessed the steep rise in population in coastal regions, megacities, and adjoining
regions, which is expected to continue for coming decades. These populations are highly sensitive to many
hazards and risks: flood, climate change-induced sea-level rise, and disease outbreaks. Also, the severe pop-
ulation stress in these cities as a result of mass migration and urban-centric growth puts extreme stress on
already crumbling infrastructure systems of these megacities. The combined effect of over-virtualization
of social networks and mass migrations coupled with evolving nature of hazards, specifically in the back-
drop of climate change, has made these social networks dynamic in space and time, thus contributing to
the “nonstationarity” in social networks [25]. Table 1.1 summarizes the role of “nonstationarity” and “deep
uncertainty” across the three systems under consideration.
TABLE 1.1 Common Attributes of the Stressors, Stressed, and Impacted Systems
Attributes →
and systems ↓ Correlation Extremes Nonstationarity Deep uncertainty
Climate change Teleconnections [17]; More intense heat Trends in Climate chaos [29];
and weather multivariate and waves [26]; persisting extremes [28] predictive surprises [30];
extremes multiscale dependence; cold extremes [27]; climate variability [31]
(stressors) long-term memory [8] intensifying precipitation
extremes [16]
Lifeline and Interdependent Cascades [34]; first-order Time-evolving Topological diversity [38];
natural systems systems [32]; transition [35]; ecological networks [37] parameter uncertainty [39];
(stressed) mutualistic ecological community collapse [36] chaos in ecosystems [29]
networks [33]
Social systems Social network Extreme stress on Mass Uncertain data [25];
(impacted) analysis [40] communities [40] migrations [40] unanticipated policies [41]
systems. However, knowledge acquired from network-based representation and analyses should be supple-
mented and complemented with the information that aids our understanding about individual systems. For
example, in case of climate-related stressors, while network can give dynamical understanding of climate
oscillators and teleconnections, data sciences, statistics, and local-scale physics [14] still play a pivotal role
in generation of actionable insights. In case of infrastructure systems, both built and natural, the network-
based knowledge needs to be complemented and supplemented with data acquisition, data representation,
information management, real-time data ingestion, and offline data analyses. Finally, in impacted sys-
tems, while network-based solution framework can bridge the key knowledge gaps in our understanding
of societal resilience, this knowledge may not be sufficient enough to inform the policy, econometrics, and
incentive structures for decision making.
The three systems alluded to above and represented in Figure 1.2, specifically, stressors (climate change
and weather extremes), stressed (lifelines and natural environment), and impacted systems (societal and
human) share two important attributes. First, the systems are dependent, often in a complex manner, and
further, they may be either subject to or cause/exacerbate extremes dependent in space and time, termed as
“correlated extremes.” Second, recent evolution of these system lead to what has been called “nonstation-
arity with deep uncertainty”, or considerable changes in patterns (giving rise to situations where history
is no longer a sufficient guide to future) and unwieldy uncertainty (a mix of known unknowns, unknown
unknowns, and intrinsic systemic variability).
Despite the increasing complexity in threats and interdependencies in technical, social, and econom-
ical systems, conventional risk management frameworks have focused on strengthening these specific
components to withstand the identified threats to an acceptable level and hence preventing the overall
system failure. As outlined in the commentary in Nature Climate Change [6], enabling these capacities
requires specific methods to define and measure resilience and modeling and simulation techniques for
highly complex systems to develop resilience engineering. Figure 1.3 shows the resilience management
framework with risk as a central component. Risk of climate-related hazards results from the interaction
of climate-related hazards with the vulnerability and exposure of natural, built, and human systems [42].
It is noted that analogy can be drawn between the three elements of risks (Figure 1.3) with the three
layers shown in Figure 1.2. That is, threats are analogous to climate stressors, vulnerabilities to the stressed
systems, and exposed systems to the impacted systems, which are social systems in the present con-
text. However, given the strong bidirectional feedback among the three systems referred to in Figure 1.2,
this analogy may not always hold true since specific system can assume any of the three roles in risk
management framework.
s
Extreme
rt a in ty
” unce
s a n d “Big Data
ange network tremes
Regional ch Climate ather ex
ge and we
hydrology : climate
ch a n
Stressors
ctures
infrastru
Critical
&
urces tworks
Key reso cture ne
/g re e n infrastru o systems
e c te d b lue/grey t lifelin es and ec
n enden
Intercon interdep
systems:
Stressed
redness
& prepa
o c ia l n etworks l incentives
S
Financia ance
l govern
Regionaaction plans
nse
rd respo ve s
ia and ha z a e incenti
Socia l m e d
a n d g o vernanc
nsions
an dime
cy: hum
Society and poli
FIGURE 1.2 (See color insert.) A multilayered and unified framework can represent the coupled system where
(a) climate change and weather or hydrological extremes are the stressors; (b) interdependent lifeline and envi-
ronmental systems are the stressed systems; and (c) social networks of communities and regions are the impacted
systems. This representation enables a unified quantitative framework, but particular instantiations of this network
are necessary and sufficient to answer specific questions.
Risk Vulnerability
re
Antifragility
su
Essential functionality
po
Ex
Suboptimal recovery
ry
Re
Impact area
ove
sp
on
Rec
es
Time
FIGURE 1.3 In system resilience framework, risk can be interpreted as the total reduction in critical functionality,
while resilience is related to absorption of the stress, rate of reduction in essential functionality, and recovery of essential
functionality after adversity. The postrecovery state of the system—measured in terms of essential functionality—can
be better, equal to, or worse than prehazard state of the system.
6 Large-Scale Machine Learning in the Earth Sciences
functionalities that are crucial, understand and improve the interconnectivity of interdependent net-
works, enhance social networks, and bridge the gaps in policy shortfalls to mitigate and recover from the
unforeseeable impacts the evolving stressors, including climate change, can produce. It should be noted
that resilience is not an alternate to either risk management or engineering design principles for designing
specific components. Rather, resilience is a complementary attribute to improve the traditional risk man-
agement framework, which is useful and widely accepted tool in mitigating the foreseeable and anticipated
stress situations.
A roadmap to enable resilient design practices requires (1) understanding the correlations and evolving
nature of adverse events that can impact the functionality of infrastructure systems and/or society; (2)
approaches for computing robustness of the highly complex systems; (3) identifying the most important
(or influential) components within individual systems; and (4) development of new simulation techniques
to understand the resilience (response and recovery) of isolated as well as interdependent systems. The
next section discusses how network science-based technology Stack (Net tech Stack) can serve as a unified
framework to understand and hence quantify resilience for highly complex and interdependent systems.
However, the instantiations of each of these systems, represented by networks, are necessary to address
specific challenges. The representation and model can be combined with sensor and data engineering,
computational process models, data science methods, and uncertainty propagation in forward or feedback
processes to develop intervention strategies for monitoring and early warning, managing cascading failures
and graceful recovery, proactive designing, and developing vulnerability and risk models.
Stressed systems
Network multiplex
Catastrophic cascades
First-order transition
Resilience
Individual networks
Random networks
Scale-free networks
Robustness Small-world property
Evolving networks
Impacted systems
Stressors
Climate networks Network Degree
node weight
link strength
FIGURE 1.4 (See color insert.) Complex network-based technology stack in combination with sensor and data engi-
neering, computational process models, and data science methods can potentially serve as a unified framework to meet
these challenges. Tools to model single and interdependent networks come from network science and contribute to
insights for the lifelines and ecosystems. Since all the three systems—stressors, stressed, and impacted systems—share
common attributes of correlation and interdependence, fundamental network science breakthroughs and state of the
art leads to novel adaptations or (in some cases) customization and then onward to new science or engineering insights.
of networks emerging in various domains of science and technology are similar to each other, enabling the
applications of common set of mathematical and machine learning-based tools to explore these similarities
making network science applicable to the heterogeneous systems. For example, many real-life systems such
as transportation networks [44] and communication networks [45] exhibit a exponentially decaying tail in
degree distribution and hence characterized as “scale-free” networks [43].
Although real-world infrastructure systems are highly interdependent and interconnected in nature,
large number of studies have modeled and analysed the networks as single networks with underlying
assumption that these networks can be isolated from each other. However, in interdependent networks,
relatively localized failure in one system can percolate through the interconnected web or network of net-
works [5], triggering a “disruptive avalanche of cascades” [4]. Despite the fact that network science-based
methods have advanced our understanding of individual and interdependent systems, percolation-based
8 Large-Scale Machine Learning in the Earth Sciences
models are a very stylized representation of the network’s response to the damage and hence lack the
realism to capture many of the features that contribute to the resilience and robustness of real-world sys-
tems [4]. Moreover, given the strong bidirectional feedback that exists between the climate-induced stres-
sors, stressed, and impacted systems, enabling resilient design practices requires a coherent understanding
of the feedbacks within and across these systems, network science-based frameworks, in combination
with sensor and data engineering and computational process models, which can lead to new scientific
and engineering insights.
TABLE 1.2 Conceptual Representation of “Telescoping Systems of Systems” for Infrastructure Systems Operating at
Spatial Scales Ranging from System to Regional to National and Global Scales
National
Regional and
Air transportation: USA Power-grid and megaurban
transportation: India
Multiscale built-natural
Global infrastructures
interconnected and telescoping
engineered systems
System level
global circulation models [GCMs]), can provide more specific detailed simulations for systems operating
at diverse spatial scales.
1.6 Conclusion
The challenges in the coupled system-of-systems comprised of the stressors (climate change and extreme
weather), stressed (built-natural lifelines), and impacted (human and social) systems are twofold: corre-
lated and unprecedented extremes on highly interconnected systems as well as nonstationarity and deep
uncertainty in the systems. Nonstationarity in this context goes beyond statistical definitions to large
changes where history is no longer a sufficient guide to the future and engineering and planning principles
need to change accordingly, while deep uncertainty refers to situations where translation to likelihood-
based risks and hence risk-based design may not be feasible, thus motivating flexible design principles.
While complex network-based frameworks, including network science technology stack and “telescoping
systems of systems,” could potentially help enable the resilient design practices in changing climate scenar-
ios. Given the promise these climate networks have exhibited to understand the characteristics of highly
complex systems, it may not be even necessary and sufficient to represent the systems’ state. For example,
in context of social networks, though network science-based methodologies have delved crucial informa-
tion about architecture of these networks, it does not take into behavioral, social, and cultural attributes of
these systems into account. Similarly, in context of climate-related stressors, while understanding of cou-
pling in climate systems may yield new insights about stability of climate systems, processes that generate
extremes may be synoptic or mesoscale. This means that the network science should not serve as “the ham-
mer and the nail” solution to the challenges discussed for each of the three systems. Enabling the resilient
design and engineering practices, specifically in the context of climate change and weather-related haz-
ards, needs to integrate the system-level visualization with sensor and data engineering, computational
process models, data science methods, and uncertainty propagation in forward or feedback processes.
A 2014 survey at the World Economic Forum (WEF) identified “failure of climate change mitigation and
adaptation” and “greater incidence of extreme weather events (e.g., floods, storms, fires)” as two of the top
ten global risks of highest concern. The U.S. National Academy of Engineering (NAE) lists “restore and
improve urban infrastructures” and “engineer the tools for scientific discovery” as two of fourteen grand
10 Large-Scale Machine Learning in the Earth Sciences
Inform
ed decisi ed
ov nes on
li s
gui r
Imp
de
Vicious circle: uous circ
irt
Economic
rewards
le
Cycle of under-preparedness and disruptions:
V
Vulnerability to change worsened by uncertainty
engi
N o er in
Fragility to disruptions amplified by connectivity
ne
ve
l g
cy/socie
Poli t ering Pl a n n e d
myopia al engine i n t e r v e nt i o n s
s Climate-aware systems
l al
of built-natur
ve
disince ia
nti
c
b
Finan
Ou practices
es
t
tdated
n ua
te ing Action under uncertainty aids adaptation to change
pl e er
an
s E n g i n ati o n Managing interdependence improves resilience
st a g n
challenges in engineering. The broad vision for future research is to create a consensus for addressing the
challenges described in this chapter. A critical momentum is needed (Figure 1.6) to prevent from plunging
into a vicious cycle of economic disincentives, policy myopia, and social inertia, resulting in stagnation
of engineering and science, to a virtuous cycle of resilient infrastructures, climate preparedness, sustained
economic growth, social justice, and greater security.
References
1. Climate change: Cold spells in a warm world. Nature, 472(7342):139–139, April 2011. Available at
http://www.nature.com/nature/journal/v472/n7342/full/472139d.html.
2. Aiguo Dai. Increasing drought under global warming in observations and models. Nature Climate
Change, 3(1):52–58, January 2013.
3. Udit Bhatia, Devashish Kumar, Evan Kodra et al. Network science based quantification of resilience
demonstrated on the Indian railways network. PLOS ONE, 10(11):e0141890, 2015.
4. Alessandro Vespignani. Complex networks: The fragility of interdependency. Nature,
464(7291):984–985, April 2010.
5. Jianxi Gao, Sergey V. Buldyrev, H. et al. Percolation of a general network of networks. Physical Review
E, 88(6):062816, December 2013.
6. Igor Linkov, Todd Bridges, Felix Creutzig et al. Changing the resilience paradigm. Nature Climate
Change, 4(6):407–409, June 2014
7. Amit Sheth. Transforming Big Data into smart data: Deriving value via harnessing volume, vari-
ety, and velocity using semantic techniques and technologies. In 2014 IEEE 30th International
Conference on Data Engineering (ICDE), pp. 2–2, March 2014.
8. Karsten Steinhaeuser, Auroop R. Ganguly, and Nitesh V. Chawla. Multivariate and multiscale
dependence in the global climate system revealed through complex networks. Climate Dynamics,
39(3–4):889–895, June 2011.
9. Zhenning Kong and E.M. Yeh. Resilience to degree-dependent and cascading node failures in ran-
dom geometric networks. IEEE Transactions on Information Theory, 56(11):5533–5546, November
2010.
Exploring the Variety of Random
Documents with Different Content
Kello yhdentoista vaiheilla aamupäivällä tuli pääkaupunki näkyviin
muurien saartamana, joita vielä suojasi ympärysmitaltaan lähes
parinkymmenen kilometrin pituinen vallihauta. Kadut olivat leveitä ja
suorakulmaisesti suunniteltuja tasaisella maaperällä, ja keskellä
kohosi iso torni, jonka pohjoislaidan täytti kuninkaan palatsi. Tätä
laajaa rakennusryhmää ylemmäksi kohosi eräänlainen parveke,
lyhyen matkan päässä siitä neliömäisestä alasta, joka oli varattu
uhritoimituksille. Juhlapäivinä tältä parvekkeelta heitettiin kansan
käsiin vankeja, jotka oli sidottu pajukoreihin, ja oli helppoa kuvitella,
kuinka raivoisasti onnettomat uhrit revittiin kappaleiksi.
Tällä kertaa hän ei ollut yksin, sillä hän ei olisi ehtinyt katkaista
kaikkien päitä. Hänen lähelleen oli ryhmittynyt satakunta pyöveliä,
jotka olivat tottuneet yhdellä iskulla sivaltamaan kaulan poikki.
— Vai niin, niiden tekee mieli maistaa tätä lisää! Tom Turner
huudahti.
Siten saivat setä Prudence ja Phil Evans tietää, että he olivat nyt
jättäneet pohjoisen pallonpuoliskon ja joutuneet eteläiselle. Tähän
tapahtumaan ei kuitenkaan liittynyt sentapaisia temppuja ja
juhlamenoja kuin joillakin sota- ja kauppalaivoilla.
Selityksenä tähän ilmiöön ei voinut olla muu kuin sähkö, sillä oli
mahdotonta olettaa sen johtuvan kalojen kutusärkästä tai valtavasta
pieneliöiden kasaumasta.
Sää oli tyyni. Muutamin paikoin taivaalla oli paksuja mustia pilviä,
jotka olivat yläosastaan nystyrämäisiä ja päättyivät jäykkään,
täydellisesti vaakasuoraan viivaan. Näistä pilvistä pisti esille
pitkulaisia kyhmyjä, jotka näyttivät kärjillään vetävän puoleensa
vettä, joka poreili niiden alla, kohoten kuin nestemäinen pensaikko.
Itsestään oli myös selvää, että jos päivä alkoikin pidetä talvisesta
päivänseisauksesta, joka eteläisellä pallonpuoliskolla on kesäkuun
21. päivänä, niitä toisaalta yhä enemmän lyhensi Albatrossin yhä
jatkuva lento napamaita kohti.
Siellä oli todellakin vene. Sen purje roikkui velttona pitkin mastoa.
Kun ei tuullut, sitä ei enää voinut ohjata, eikä kukaan sen miehistä
näkynyt jaksavan pidellä airoa.
Ei mitään vastausta.
— Keitä te olette?
Robur tiesi sen. Hän oli myös selvillä siitä, että varovaisuus vaati
pakenemaan hirmumyrskyä sen vetovoiman piiristä ylempiin
ilmakerroksiin. Tähän asti hän oli siinä aina onnistunut. Mutta nyt ei
sopinut hetkeäkään viivytellä, kenties ei minuuttiakaan.
Oli selvää, että jos myrskyn rajuus vielä yltyisi, Albatross olisi sen
käsissä ohjauskyvyttömänä, ja silloin sen ahmaisisi kitaansa jokin
niistä pyörremyrskyistä, jotka kiskovat maasta puita juurineen,
repivät irti kattoja ja kaatavat kiviseiniä.
Our website is not just a platform for buying books, but a bridge
connecting readers to the timeless values of culture and wisdom. With
an elegant, user-friendly interface and an intelligent search system,
we are committed to providing a quick and convenient shopping
experience. Additionally, our special promotions and home delivery
services ensure that you save time and fully enjoy the joy of reading.
textbookfull.com