Sustainable Cities and Society 36 (2018) 272–285

Contents lists available at ScienceDirect

Sustainable Cities and Society

journal homepage: www.elsevier.com/locate/scs

The Resilience to Emergencies and Disasters Index: Applying big data to MARK
benchmark and validate neighborhood resilience capacity
⁎
Constantine E. Kontokosta , Awais Malik
Dept. of Civil & Urban Engineering and Center for Urban Science & Progress, New York University, 1 MetroTech Center, 19th Floor, Brooklyn, NY 11201, United States

A R T I C L E I N F O A B S T R A C T

Keywords: Resilience planning and emergency management require policymakers and agency leaders to make difficult
Urban resilience decisions regarding which at-risk populations should be given priority in the allocation of limited resources. Our
Emergency management work focuses on benchmarking neighborhood resilience by developing a unified, multi-factor index of local and
Disaster recovery regional resilience capacity: the Resilience to Emergencies and Disasters Index (REDI). The strength of the REDI
Community resilience
methodology is the integration of measures of physical, natural, and social systems – operationalized through the
Big data
Anomaly detection
collection and analysis of large-scale, heterogeneous, and high resolution urban data – to classify and rank the
relative resilience capacity embedded in localized urban systems. Feature selection methodologies are discussed
to justify the selection of included indicator variables. Hurricane Sandy is used to validate the REDI scores by
measuring the recovery periods for neighborhoods directly impacted by the storm. Using over 12,000,000 re-
cords for New York City’s 311 service request system, we develop a proxy for neighborhood activity, both pre-
and post-event. Hurricane Sandy had a significant and immediate impact on neighborhoods classified as least
resilient based on the calculated REDI scores, while the most resilient neighborhoods were shown to better
withstand disruption to normal activity patterns and more quickly recover to pre-event functional capacity.

1. Introduction communities compare, and how to quantify the impact of resilience-

Hurricane Sandy's devastation led to 147 deaths, over 650,000
homes destroyed, and left 8.5 million people along the U.S. East Coast
without power (Sullivan, 2012). In New York City alone, Sandy caused
43 deaths and over $19 billion in damage (Bloomberg, 2013). This
historic storm particularly exposed New York City's vulnerability to
coastal flooding, and left 6800 evacuees assigned to shelters, including
1800 patients from chronic care facilities (Gibbs & Holloway, 2013).
The aftermath of this extreme event resulted in a comprehensive set of
proposals aimed at achieving resilience through ‘protection’ and ‘ac-
commodation’ in the event of a similar disaster (McArdle, 2014).
Hurricane Sandy brought resilience improvement measures to the
forefront of urban policy and planning, making the process of identi-
fying vulnerable communities and quantifying their resilience capacity
critical to effective emergency management and long-term resilience
investments.
Resilience planning and emergency management require policy-
makers and agency leaders to make difficult decisions regarding which
at-risk populations should be given priority in the allocation of limited
resources. By extension, policymakers need to understand how com-
munity resilience capacity changes over time, how different
related investments and mitigation strategies. Our work presents a
methodology to use diverse, large-scale urban data to identify, quantify,
and benchmark neighborhood resilience through a unified, multi-factor
index of local and regional resilience capacity: the Resilience to
Emergencies and Disasters Index (REDI).
The REDI score is a benchmark of relative neighborhood resilience
capacity within and between municipalities, and can be used to prior-
itize investment and funding needs across multiple dimensions of
physical, social, economic and environmental conditions. The index is
also intended to measure progress over time in increasing local resi-
lience capacity, and to provide a performance measure to estimate the
return on investment of resilience capacity-building measures. The
strength of the proposed REDI score is that it combines measures of
physical, natural, and social infrastructure systems to classify and rank
the relative resilience capacity embedded in localized urban systems at
high spatial resolution. This approach recognizes the importance of
critical local community attributes that impact the ability to respond to
and recover from emergencies and disasters.
The neighborhood REDI score is normalized on a scale of 1–100 to
measure the deviation of a given neighborhood from a reference region
mean. The reference region can be shifted to coincide with the political

⁎
Corresponding author.
E-mail addresses: ckontokosta@nyu.edu (C.E. Kontokosta), awais.malik@nyu.edu (A. Malik).

http://dx.doi.org/10.1016/j.scs.2017.10.025
Received 23 June 2017; Received in revised form 13 October 2017; Accepted 21 October 2017
2210-6707/ © 2017 Elsevier Ltd. All rights reserved.
C.E. Kontokosta, A. Malik
and governance boundary of a particular agency's jurisdiction. A score
of 100 represents the highest relative resilience capacity. ArcGIS and
Python (Pandas) serve as the spatial data integration and visualization
platforms, while temporal data are visualized using Tableau, and
ARIMA time-series analysis is performed using Minitab. In this ex-
ploratory study of New York City, we collect and analyze a range of
data sources provided by city, state and federal agencies, including the
Department of City Planning, Department of Transportation,
Metropolitan Transit Authority, Office of Emergency Management,
Federal Emergency Management Agency, Department of Information
Technology & Telecommunications, Department of Finance, and the
United States Census Bureau, among others. The REDI score metho-
dology is designed to be scalable to all municipalities across the U.S.
and potentially global cities, provided the requisite underlying data are
available at the appropriate spatio-temporal resolution. The metho-
dology also allows for customization of the REDI algorithm to account
for local priorities through indicator weighting.
This paper begins with a literature review of previous work relevant
to community resilience metrics and measurement, and highlights the
challenges involved in creating a unified resilience capacity index. The
subsequent section describes the REDI methodology: the collection and
integration of publicly-available data, the selection and extraction of
relevant indicator variables, the formulation of REDI scores for each
neighborhood, and the data cleaning process required to remove out-
liers and areas with incomplete information. Following the metho-
dology description, a discussion of the major findings of the analysis
and its implications for urban resilience policy, planning, and decision
support is presented. The paper continues with a validation model of
the REDI methodology using “big data” in the form of 12,000,000
“311” service request records before, during, and after Hurricane Sandy
in New York City. Using service request data as a proxy for neighbor-
hood activity patterns, the recovery period from this event is measured
and compared for the most resilient and least resilient communities
identified by their REDI scores. The paper concludes with a discussion
of limitations and recommendations for future research.
2. Literature review
The notion of “resilience” has been the subject of several contested
definitions (Aldunce, Beilin, Handmer, & Howden, 2014), resulting in
divergent views on what it should encompass (Cutter, 2016a;
Linkov & Florin, 2016; Linkov & Palma-Oliveira, 2017), and how it
should be measured (Cutter, Ash, & Emrich, 2014; Winderl, 2014).
Holling (1973) describes resilience as a “measure of the persistence of
systems and ability to absorb change and disturbance and still maintain
the same relationships between populations and state variables.” The
National Academies of Sciences defines the resilience of a system as “its
ability to plan and prepare for, absorb, respond to, and recover from
disasters and adapt to new conditions” (Ganin et al., 2016). The Na-
tional Institute of Standards and Technology (NIST) characterizes
community resilience as “the ability of a community to prepare for
anticipated hazards, adapt to changing conditions, and withstand and
recover rapidly from disruptions” (López-Cuevas, Ramírez-Márquez,
Sanchez-Ante, & Barker, 2017). To operationalize these definitions,
Bruneau et al. (2003) propose four key dimensions of community re-
silience: robustness, redundancy, resourcefulness and rapidity. The
“resilience capacity” of a community, therefore, is the inherent set of
features that enable that community to effectively respond to and re-
cover from extreme events (Foster, 2012).
To add to the epistemological complexity around the term resi-
lience, vulnerability – the propensity of systems to incur adverse shocks –
is often considered the opposite side of the same coin (Bates,
Angeon, & Ainouche, 2014; Gallopin, 2006; Weichselgartner, 2001), as
regions of greater vulnerability tend to be the least resilient
(Bergstrand, Mayer, Brumback, & Zhang, 2015). In urban environ-
ments, both the susceptibility of a system to potential harm
273
Sustainable Cities and Society 36 (2018) 272–285
(vulnerability) and its ability to quickly and effectively ‘bounce back’
from any damage (resilience capacity) must be accounted for in efforts
to create the “sustainable networks of physical systems and human
communities” (Godschalk, 2003) that define resilient cities (Meerow,
Newell, & Stults, 2016). The focus of this study is on measuring resi-
lience capacity, independent of neighborhood vulnerability, although it
is recognized that many indicators of resilience capacity may also in-
fluence an area's risk exposure.
Urban communities can be viewed as the complex, dynamic inter-
actions of physical, social, economic, and environmental systems
(Cavallaro, Asprone, Latora, Manfredi, & Nicosia, 2014; McPhearson
et al., 2016; Norris, Stevens, Pfefferbaum, Wyche, & Pfefferbaum,
2008). Community boundaries are considered porous, with most re-
sidents moving frequently across and between neighborhoods for work
and leisure activities (Berkes & Ross, 2013). Consequently, a neighbor-
hood's resilience capacity cannot be considered completely independent
of the resilience capacity of its surrounding neighborhoods. Since most
spatial geographies are defined by political or U.S. Census boundaries,
it is important to pair any quantitative neighborhood resilience mea-
sures with local knowledge of the region through partnerships with the
municipal government and the public that resides in those neighbor-
hoods.
Several indices, frameworks, and conceptual models have been de-
veloped to quantify resilience (see Appendix A). Most of these models
either lack data at the spatial granularity needed to adequately re-
present urban neighborhoods – resulting in several studies that instead
use counties or other large administrative divisions (Bergstrand et al.,
2015; Cimellaro, Solari, & Bruneau, 2014; Cutter, Boruff, & Shirley,
2003; Lam, Reams, Li, Li, & Mata, 2015; Miles & Chang, 2011; Sherrieb,
Norris, & Galea, 2010) – or the methods do not define the community's
spatial boundaries at all, thereby implying a spatially-scalable property
of the proposed methodologies (Cutter et al., 2008). Other approaches
are specifically applied to certain communities, such as coastal regions,
making them less generalizable to other geographic areas (Fox-Lent,
Bates, & Linkov, 2015; Islam, Swapan, & Haque, 2013; Orencio & Fujii,
2013; Razafindrabe, Parvin, Surjan, & Shaw, 2009). Feature selection
methods used to justify the indicators included in previous models
range from literature reviews and applied domain knowledge to sta-
tistical correlation analysis. One significant limitation of previous work
is the omission of any empirical validation method to test the efficacy of
the proposed methods (Bakkensen, Fox-Lent, Read, & Linkov, 2017; Cai,
Lam, Zou, Qiang, & Li, 2016; Tate, 2012).
Existing approaches to measuring neighborhood resilience can be
categorized into one or more of four domains: (1) social infrastructure
and community connectivity, (2) physical infrastructure, (3) economic
strength, and (4) environmental conditions (Jordan & Javernick-Will,
2013). The social infrastructure and community connectivity domain
consists of demographic indicators and social services that either signify
social vulnerability or highlight community cohesiveness. Suggested
social resilience enhancement strategies include increased civic en-
gagement to enable effective and efficient emergency management
operations (Aldrich & Meyer, 2015; Burnside-Lawry & Carvalho, 2015;
Godschalk, 2003; Magis, 2010), and greater policy engagement leading
to effective land use zoning changes and infrastructure investments in
vulnerable regions (Dale, Ling, & Newman, 2010; Ernstson et al., 2010;
Wagner, Chhetri, & Sturm, 2014). A number of social vulnerability
measurement techniques have been proposed (Cutter et al., 2003;
Madrigano, Ito, Johnson, Kinney, & Matte, 2015; Sherrieb et al., 2010;
Van Zandt et al., 2012). Tate (2012) provides an assessment of the
major social vulnerability index configurations using global sensitivity
analysis, and highlights the significant challenge of attempting to va-
lidate these indices using external data. López-Cuevas et al. (2017)
proposed a method to benchmark the “mood steady state” of a com-
munity using online social networks (OSNs), and measured community
resilience by gauging the changes in these steady states when a disaster
occurred.
C.E. Kontokosta, A. Malik
The physical infrastructure domain consists of critical facilities and
structural systems that are either most vulnerable to disasters or pro-
vide crucial support services to mitigate their effects. Hosseini, Barker,
and Ramirez-Marquez (2016) provide an extensive review of the lit-
erature on resilience measurement methodologies for physical infra-
structure systems. Most methodologies in this domain are applied to
specific physical infrastructure systems, such as buildings
(Bonstrom & Corotis, 2012; Matthews, Sattler, & Friedland, 2014),
highways and bridges (Bocchini, Frangopol, Ummenhofer, & Zinke,
2013; Brownjohn & Aktan, 2013), emergency facilities (e.g. hospitals,
fire stations) (Cimellaro, Reinhorn, & Bruneau, 2010; Hemond & Robert,
2012; Simpson, 2008), transportation hubs (Shafieezadeh & Burden,
2014), and utility (water, power, gas) distribution systems (Chang,
2003; Cimellaro et al., 2014; Francis & Bekera, 2014; Gonzales & Ajami,
2017; Koraz & Gabbar, 2017; Ouyang, Dueñas-Osorio, & Min, 2012;
Zimmerman, Zhu, de Leon, & Guo, 2017). A number of flexible frame-
works exist that can be applied to multiple infrastructure systems
(Ayyub, 2014a, 2014b; Bocchini et al., 2013; Croope & McNeil, 2011;
Francis & Bekera, 2014; Robert, Morabito, Cloutier, & Hémond, 2015;
Sharkey et al., 2015; Timashev, 2011; Zimmerman et al., 2017).
However, the specificity of these approaches to physical infrastructure
systems makes their application to other neighborhood resilience do-
mains limited, thus constraining a comprehensive measure based on
these methods.
Briguglio, Cordina, Farrugia, and Vella (2009) define economic re-
silience as “the policy-induced ability of an economy to recover from or
adjust to the negative impacts of adverse exogenous shocks and to
benefit from positive shocks.” The economic strength of a community
has historically been measured through indicators such as unemploy-
ment, poverty, income inequality, and economic diversity. Several
economic resilience metrics have been proposed (Briguglio et al., 2009;
Foster, 2007; Guillaumont, 2009; Martin, 2011; Rose, 2004;
Simmie & Martin, 2010; Zhang & Wan, 2009). However, most of these
frameworks share the same constraints as the methodologies belonging
to the physical infrastructure domain in that they are sector- or in-
dicator-specific, which limits their adaptability to a comprehensive,
unified measure of resilience.
Several resilience frameworks have been developed to quantify the
response to natural hazards, including hurricanes and coastal storms
(Adger, Hughes, Folke, Carpenter, & Rockström, 2005; Gong, 2014;
Grayson & Pang, 2014; Highfield, Peacock, & Van Zandt, 2014; Lam,
Qiang, Arenas, Brito, & Liu, 2015; Lam, Reams, et al., 2015; Wagner
et al., 2014), floods (Highfield & Brody, 2012), earthquakes (Bruneau
et al., 2003; Cimellaro et al., 2014; Kojima, Fujita, & Takewaki, 2014;
Miles & Chang, 2011; Shafieezadeh & Burden, 2014), droughts (Maity,
Sharma, Nagesh Kumar, & Chanda, 2012; Ranjan & Athalye, 2009), and
heat waves (Madrigano et al., 2015). Certain local characteristics (such
as building density, open space, proximity to flood zones, tree cover,
etc.) can be used to determine both the environmental vulnerability of a
neighborhood and its resilience capacity (Jabareen, 2013).
While each domain has its own set of methodologies for defining
resilience metrics, a number of composite frameworks have been es-
tablished (Angeon & Bates, 2015; Barker, Ramirez-Marquez, & Rocco,
2013; Bozza, Asprone, & Manfredi, 2015; Cai et al., 2016; Cutter,
2016b; Cutter, Burton, & Emrich, 2010; Islam et al., 2013; Linkov et al.,
2013; Renschler et al., 2010). A significant challenge associated with
forming such comprehensive indices is feature selection to reduce the
dimensionality of the indicator set (Angeon & Bates, 2015). Another
problematic decision regarding the indicator variables is assigning their
weights. McPhearson et al. (2016) advocate the social-ecological-tech-
nical systems (SETS) approach that allows “equal emphasis on the so-
cial, ecological and technological dynamics of urban systems.” Cutter
et al. (2010) also argue for equally-weighted indices for two reasons:
(1) this method provides transparency and is intuitive to a range of end-
users and (2) they were unable to determine a justification for favoring
one indicator over another. Methods for determining weights are either
274
Sustainable Cities and Society 36 (2018) 272–285
subjective or data-reliant, and the weighting rationale does not always
match the priorities of policy-makers (Cutter et al., 2010). This would
suggest, then, the importance of a model that allows for user-assigned
weights of individual indicators and groups of indicators across infra-
structure domains.
Previous attempts to quantify neighborhood resilience across mul-
tiple dimensions have been constrained by the absence of any empirical
or systematic effort to validate the findings (Bakkensen et al., 2017).
Though a framework may present a theoretically or anecdotally logical
abstraction, its utility can only be tested in practice (Bakkensen et al.,
2017; Manyena & Gordon, 2015). Attempts to validate results include
comparison with socioeconomic community indicators that were not
used in the model (Sherrieb et al., 2010), internal validation of models
using global sensitivity analysis (Tate, 2012), correlation of disaster
indices with derived loss functions (Bakkensen et al., 2017; Peacock
et al., 2010) or other statistical techniques such as discriminant analysis
(Cai et al., 2016). However, true model validation of resilience indices
requires measurement of post-disaster recovery after the occurrence of
an actual disaster event using independent proxies for normal and
atypical activity (Ganin et al., 2016; Granot, 1995; Henry & Ramirez-
Marquez, 2012).
3. Data & methodology
The resilience capacity of a locality can be viewed as the ability of
that neighborhood to respond to and recover from an event – in-
dependent of its nature or magnitude – through the minimization of
potential impacts and a more rapid return to “normal” activity.
Therefore, resilience capacity is a function of a community's capital
(infrastructure, people, and resources) to limit the disruption to equi-
librium (pre-event activity levels). This approach to measuring resi-
lience capacity is visualized in Fig. 1. The y-axis shows the community's
activity level or functionality, and the x-axis is time, divided into pre-
event, impact (during event), recovery, and post-event segments. For a
given event of magnitude (Aimpact = Aequilibirum − Aevent), community
resilience is measured by the shape of the recovery curve and the
minimization of time to post-event equilibrium, given by
(trecovery = t2 − t1) (Linkov et al., 2014). While it is assumed that a re-
turn to pre-event levels is a signal of recovery, it is possible that equi-
librium activity may permanently shift in response to the event, and a
new post-event equilibrium is established.
The REDI methodology consists of four components: (1) creating a
geospatial resilience data repository, (2) selecting relevant indicator
variables, (3) cleaning the REDI dataset, and (4) formulating the REDI
scores.. Each component is described below.
3.1. Creation of a geospatial resilience data repository
Our first task is to create a geospatial repository by collecting and
integrating relevant datasets. We identify and select data sources that
are widely available and can be accessed for the broad application of
our approach. Table 1 provides the list of open data sources collected
Fig. 1. Conceptual depiction of resilience capacity with event occurring at time t0.
C.E. Kontokosta, A. Malik
Table 1
Data sources for geospatial repository.
Data sources [Web link]
US Census Bureau [https://www.socialexplorer.com/]
NYC Department of City Planning (DCP) BYTES of the BIG APPLE™[https://www1.nyc.gov/site/
planning/data-maps/open-data.page]
Metropolitan Transit Authority (MTA) – NYC Subways
Stations & Routes: [https://spatialityblog.com/2010/07/08/mta-gis-data-update/#datalinks]
Entrances: [https://data.cityofnewyork.us/Transportation/Subway-Entrances/drex-xx56]
Federal Emergency Management Agency (FEMA) Modeling Task Force (MOTF) – Hurricane
Sandy Impact Analysis [https://www.arcgis.com/home/item.html?id=
307dd522499d4a44a33d7296a5da5ea0]
NYC Emergency Management Hurricane Evacuation Centers [https://data.cityofnewyork.us/
Public-Safety/Hurricane-Evacuation-Centers/ayer-cga7]
NYC 311 Service Requests (2010–2015) [https://nycopendata.socrata.com/Social-Services/311-
Service-Requests-from-2010-to-Present/erm2-nwe9]
NYC Directory of Parks Properties [https://data.cityofnewyork.us/City-Government/Parks-
Properties/rjaj-zgq7]
NYC Street Tree Census
Manhattan: [https://data.cityofnewyork.us/Environment/Street-Tree-Census-Manhattan-/e6n3-
m3vc]
Brooklyn: [https://data.cityofnewyork.us/Environment/Street-Tree-Census-Brooklyn-/ztcw-
bzc8]
The Bronx: [https://data.cityofnewyork.us/Environment/Street-Tree-Census-Bronx-/bkyy-g74a]
Queens: [https://data.cityofnewyork.us/Environment/Street-Tree-Census-Queens-/4wcq-bg6s]
Staten Island: [https://data.cityofnewyork.us/Environment/Street-Tree-Census-Staten-Island-/
pvq6-utmq]
specifically for the New York City proof-of-concept and the corre-
sponding datasets included from each source. ArcGIS and Python
(Pandas) are used to integrate these datasets and form the basis for the
geospatial repository. The datasets are separated by borough to enable
easy collection and storage in ArcGIS Map Packages. The primary data
geoprocessing steps taken to merge these datasets based on spatial re-
lations – using both overlay and proximity methods – include common
functions such as buffer, clip, intersection, union, dissolve and spatial
join.
3.2. Selection of indicator variables
In much of the literature studying urban spatial dynamics, census
tracts are typically used as a geographic proxy for neighborhoods, as
operationalized in a recent analysis of the social vulnerability of New
York City and in the rich urban planning literature on neighborhood
change (Kontokosta, 2014; Kontokosta, 2015; Madrigano et al., 2015;
Wallace & Wallace, 2008). Census tracts are chosen as the REDI score
spatial unit of analysis for two reasons: (1) most socio-economic data
are already available at this geographic scale removing the need for
areal interpolation and associated margins of error, and (2) census tract
boundaries provide a reasonable geographic delineation of New York
City neighborhoods and have been used in previous studies as proxies
for neighborhoods in cities (Wallace & Wallace, 2008).
After integrating the datasets and selecting the geographic unit of
analysis, twenty-four (24) indicator variables are selected to compute
275
Sustainable Cities and Society 36 (2018) 272–285
Datasets selected from source
• 2010 Decennial Census
• American Community Survey 2009–2013 (5-year estimates)
• Primary Land Use Tax Lot Output (PLUTO) Zoning Characteristics
• Administrative
Area)
and Political Districts (2010 Census Tracts without Water
• Selected Facilities and Program Sites (Schools, Libraries, Police Stations,
Fire Stations, Health Care Services, Residential Developmental Disabilities
Services, Adult Social Services & Child Social Services)
• Hurricane Evacuation Zones
• Subway Stations & Entrances
• Subway Routes
• GIS Shapefile of Hurricane Sandy Storm Surge in NYC
• GIS Shapefile of locations of Hurricane Evacuation Centers
• Daily 311 complaint totals extracted for each census tract
• NYC Parks Shapefiles
• GIS Shapefiles of locations of street trees for each borough
the REDI scores for each neighborhood in New York City based on
previous literature, the availability of the data across multiple geo-
graphies, and the results of the correlation matrix discussed below.
Table 2 presents the selected variables, as well as a justification for why
that particular feature was included in the index. These indicator
variables are assigned to one of four dimensions: (1) social infra-
structure and community connectivity, (2) physical infrastructure, (3)
economic strength, or (4) environmental conditions. Each indicator
variable is also assigned a +1 or −1 weight depending on whether a
higher value for that indicator variable enhances (+1) or decreases
(−1) resilience capacity for a given neighborhood. Although equal
weights are the default indicator weighting in the model, the method
can accommodate variable weights based on the priorities of a parti-
cular agency or jurisdiction.
While the initial selection of indicator variables is based on a review
of the literature and available data at the appropriate spatial scale, a
robust resilience capacity measurement framework requires a feature
selection methodology to identify the most parsimonious set of in-
dicators that significantly contribute to capturing the resilience capa-
city of urban neighborhoods. While most of the existing resilience
studies rely on theoretical expectations or domain-specific knowledge
to select indicators, a few analyses have attempted to use statistical
correlation methods, machine learning algorithms, and other techni-
ques for feature selection (see Appendix A). Here, Pearson's correlation
coefficients are computed for all 24 indicator variables, and plotted on a
heat map (see Fig. 2). None of indicator variable pairs had a Pearson's
C.E. Kontokosta, A. Malik
Table 2
List of 24 indicator variables used to compute REDI scores grouped by dimension.
Indicator variable [code] (data source)
Social infrastructure & community connectivity
Population Density [PopDen] (US Census Bureau)
Percent Population under 18 & over 65 years [Un18Ov65] (US Census
Bureau)
Percent of Non-Family Households with Single Occupancy [NonFHLone]
(US Census Bureau)
Percent of Non-Family Households with Under 18 Occupants [HHUnder18]
(US Census Bureau)
Percent of Vacant Housing Units [VacUnit] (US Census Bureau)
Percent Population Over 25 with Bachelor's degree [BachOvr25] (US Census
Bureau)
Percent Population Over 3 Not Enrolled in School [Ov3NoSchl] (US Census
Bureau)
Percent Population with no Health Insurance Coverage [NoHealth] (US
Census Bureau)
Density of Adult Social Services Centers [AdSS] (NYC DCP)
Density of Child Social Services Centers [Chss] (NYC DCP)
Density of Residential Developmental Disabilities Services Centers [RDDS]
(NYC DCP)
Density of Libraries [Libr] (NYC DCP)
Physical infrastructure
Distance to Nearest Fire Station from Tract Center [fire] (NYC DCP)
Distance to Nearest Police Station from Tract Center [police] (NYC DCP)
Distance to Nearest Health Services from Tract Center [health] (NYC DCP)
Number of Subway Stations in 1-mile radius from Tract Center [SubSt]
(MTA)
Number of OEM Evacuation Centers in 1-mile radius from Tract Center
[OEMEvac] (NYC Emergency Management)
Economic strength
Unemployed Population Over 16 in Labor Force [unemploy] (US Census
Bureau)
Gini Index for Income Inequality [Gini] (US Census Bureau)
Lack of Economic Diversity – Percent of Employed Population in Single
Largest Economic Sector [EconDiver] (US Census Bureau)
Environmental conditions
Percent of Tract covered by Hurricane Sandy flood [PctSandy] (FEMA
MOTF)
Tree Density [trees] (NYC Open Data – DoITT)
Building Density [BldgDen] (NYC DCP – PLUTO)
Percent of Census Tract's Land Use categorized as “Open Space” – PLUTO
Land Use Category 9 [PctOpen] (NYC DCP – PLUTO)
correlation coefficient of 0.7 or greater, which is the suggested
threshold for collinearity concerns used by Cutter et al. (2010). The
most positively correlated variables were Percent Population under
18 & over 65 years and Percent of Non-Family Households with Single
Occupancy, with a Pearson's coefficient of 0.613. Percent Population
Over 25 with Bachelor's Degree has the highest negative correlation
with Percent of Non-Family Households with Occupants Under the Age
of 18 (−0.549) and Percent Population under 18 & over 65 years
(−0.529). Based on these results, the full feature set is included in the
model.
3.3. Data cleaning
Before the REDI scores can be calculated and visualized, certain
data cleaning measures are conducted. First, census tracts with missing,
Sustainable Cities and Society 36 (2018) 272–285
Justification [reference] (weight)
Denser regions are harder to evacuate (Cutter et al., 2003) (−1)
Extremes of the age spectrum have mobility constraints that affect evacuation measures (Cutter
et al., 2003; Tate, 2012) (−1)
Such individuals have decreased support and ability to cope with additional financial costs or time
demands (−1)
These are minors/dependent populations without ready access to adult/parental support (−1)
Negatively impacts community cohesion (Van Zandt et al., 2012). Although Cutter et al. (2010)
argue that it is a positive measure of shelter capacity (−1)
Higher educational attainment is linked to higher socioeconomic status, better understanding of
warning information, greater access to recovery information, and increased problem-solving
capacity (Cutter et al., 2003) (+1)
These populations are difficult to identify and their invisibility leads to post-disaster neglect (Cutter
et al., 2003) (−1)
No health insurance negatively affects recovery (Cutter et al., 2010) (−1)
Populations dependent on social services require additional support before and after a disaster
(Cutter et al., 2003) (−1)
Same rationale as Adult Social Services (−1)
Patients residing in these centers require special attention during evacuation measures (Tate, 2012)
(−1)
Libraries serve as community resource centers (+1)
Smaller distance means easier access to emergency services (Cutter et al., 2008) (−1)
Same rationale as Fire Stations (−1)
Same rationale as Fire Stations (−1)
Measure of emergency evacuation redundancy (+1)
Measure of sheltering capacity (Cutter et al., 2010) (+1)
Limited access to resources for recovery (Tate, 2012) (−1)
Greater income disparity negatively effects community economic strength (Sherrieb et al., 2010)
(−1)
A single-sector dependent economic base is more vulnerable (Cutter et al., 2010) (−1)
Sandy affected areas are still recovering and therefore, have lower resilience capacity (−1)
High street tree density has environmental benefits. Although, they can also be a hazard during
wind storms (Burnside-Lawry & Carvalho, 2015) (+1)
Increasing urban compactness enhances the urban heat island effect and decreases options for
water retention and infiltration. Also, greater building density can lead to greater losses during a
hazard event (Cutter et al., 2003). Although, greater density is more energy and resource efficient
(Jabareen, 2013) (−1)
Greater open space is less vulnerable to disasters and enables easier establishment of shelters and
evacuation points (+1)
incomplete or incorrect information (e.g. tracts with zero population,
tracts with missing Gini coefficients from US Census, etc.) are removed.
Second, tracts with populations less than 500 people are removed, as
census tracts are designed to generally have a population between 1200
and 8000 people (US Census Bureau) and these low population tracts
are not associated with residential neighborhoods. Finally, tracts that
are mostly parks (e.g. Central Park, Prospect Park, etc.), tracts assigned
the Neighborhood Tabulation Area (NTA) label of “parks-cemetery-
etc.”, and tracts with greater than 10% land use zoning attributed to
“Open Space and Outdoor Recreation” (PLUTO Land Use Zone 9) are
removed, as these largely open areas serve as outliers in the environ-
mental conditions category (see Results and Discussion). The cleaned
dataset includes 2089 (96 percent) of the 2166 New York City census
tracts.
276
C.E. Kontokosta, A. Malik
3.4. Formulation of REDI scores
The REDI scores for a particular community at a given time are
calculated as follows:
n Lower Manhattan, a core commercial business district. This reflects
1
REDIj = ⎛ ⎞ ∑ (wi*zij)
⎝n⎠ i=1
where REDIj is the REDI score for locality j, n is the total number of
indicator variables, wi represents the weight for indicator i, and zij is the
normalized value for indicator i in locality j given by:
x ij − x ik
zij =
σik
where xij is the value of indicator i for locality j, x ik is the mean value of
indicator i for reference region k (such that j ∈ k), and σik is standard
deviation of indicator i for reference region k, operationalized here as
the borough/county.
The calculated REDI scores are normalized to a range of 1–100,
where 1 represents the census tract with the lowest resilience capacity
and 100 the highest, with respect to the reference region average. The
reference region is used to account for potentially significant local
disparities, particularly relating to demographic and economic factors.
The composite REDI score is calculated in one of two ways: (1) the sum
of each indicator variable with equal weights normalized to the 1–100
range (Equal Weights), and (2) the sum of indicator variables in each of
the four categories (Categorical Weights) calculated separately, nor-
malized to the 1–100 range, and then summed and renormalized to the
1–100 range. The categorical weights method provides a means to give
equal significance to each of the four categories, and removes any bias
given to a category with a greater number of indicator variables. It is
possible to add a weighting factor to each category to account for user-
selected priorities across the four dimensions of resilience capacity.
4. Results and discussion
After the REDI scores are computed for each census tract, ArcGIS is
used to visualize the results. In each visualization, the scores are sorted
into five bins of equal distribution. Fig. 3 shows the REDI scores for
277
Sustainable Cities and Society 36 (2018) 272–285
Fig. 2. Pearson correlation coefficients for all 24 indicator variables.
New York City using the Equal Weights (a) and Categorical Weights (b)
methods with the most and least resilient tracts highlighted (c).. Both
maps highlight the general trend that neighborhood resilience capacity
gradually decreases as a function of the neighborhood's distance from
observed spatial patterns of infrastructure density and socioeconomic
conditions across the City.
4.1. Quartile distributions
The first exploratory analysis involves constructing radar plots of
the quartile distributions of each REDI dimension, and the indicator
variables within each domain. Fig. 4 shows these quartile distributions
for all New York City tracts used in our analysis (top row), the top 10
performing tracts according to the Categorical Weights REDI score
(middle row), and the bottom 10 performing REDI tracts (bottom row).
Quartiles increase from 0% at the center of the radar plots to 100% at
the circumference. The top row is the first check to ensure that all di-
mensions and the indicator variables in each dimension have suitable
ranges in their respective distributions, which should completely fill the
polygons when the entire dataset is represented on the radar plots. The
middle row shows indicators with +1 weights generally falling inside
the fourth quartile (75% and above), and indicators with −1 weights
similarly falling inside the first quartile (25% and below). The opposite
is observed, as expected, in the bottom row for the bottom 10 REDI
tracts. The solid black line indicates the dimensions associated with the
theoretical maximum or minimum values when looking at the top REDI
score tracts and the bottom REDI score tracts, respectively.
4.2. Categorical REDI scores
Fig. 5 shows the REDI scores for each of the four categories of
neighborhood resilience. In terms of social infrastructure and commu-
nity connectivity, Harlem, South Bronx, Corona, Williamsburg and
Borough Park stand out as neighborhoods with concentrations of census
tracts having low REDI scores. The physical infrastructure REDI scores
further highlight the trend found in the overall REDI scores that
neighborhood resilience capacity decreases as the neighborhood's
C.E. Kontokosta, A. Malik Sustainable Cities and Society 36 (2018) 272–285

Fig. 3. REDI scores for New York City: (a) equal weights, (b) categorical weights and (c) top 5 & bottom 5 categorical weights REDI tracts & Sandy surge areas highlighted.

distance from Lower Manhattan increases. This is a function of access to
public transportation, and the greater spatial dispersion of city service
infrastructure. In terms of economic strength, Harlem, Morningside
Heights, most of the Bronx, Far Rockaways, and neighborhoods along
the Brooklyn-Queens border (Bushwick, Ridgewood, Cypress Hills, etc.)
have the least resilience capacity. These neighborhoods exhibited high
poverty levels and unemployment rates during the study time frame.
The results for environmental conditions are rather deceptive. While
it may seem upon initial examination that most of the city has low
resilience capacity based on environmental conditions, this is not ne-
cessarily the case. Note that very few neighborhoods have REDI scores
in the 61–100 range, which means that these REDI scores are skewed
towards the lower end of the normalized distribution. Indicator vari-
ables with large variation in values – such as percent of tract in flood-
prone areas or percent of tract that is “Open Space” – may have con-
tributed to this distribution. It would be possible to include additional
indicator variables in this category to limit the effects of extreme values
in any single indicator, or to employ further data cleaning measures to
identify and remove a larger range of potential outliers. However, the
results for this category highlight the relatively poor natural environ-
ment on the City with respect to resilience capacity.
4.3. Development of REDIviz
In order for the REDI to be adopted by government agencies as a
decision-making tool, and to ensure accessibility by a range of stake-
holders that may use these scores, including the general public, an in-
teractive, web-based visualization platform offers transparency in the
dissemination of data-driven tools and can support user engagement
(Kontokosta, 2013; Kontokosta & Tull, 2016). A web-based prototype of
this tool, named REDIviz, has been developed (see Fig. 6).

Fig. 4. Radar charts of quantile distributions of REDI Dimensions and Indicator Variables within each Dimension for (a)–(e) All census tracts, (f)–(j) Top 10 REDI tracts, and (j)–(o)
Bottom 10 REDI tracts.

278
C.E. Kontokosta, A. Malik
Fig. 5. New York City categorical REDI scores – (a) social infrastructure, (b) physical infrastructure, (c) environmental conditions, and (d) economic strength.
5. REDI validation using Hurricane Sandy & 311 service request
data
It is expected that neighborhoods identified as having the least re-
silience capacity (e.g lowest REDI scores) will experience both a greater
initial disruption and a longer recovery time when compared to
neighborhoods with higher REDI scores impacted by the same event.
279
Sustainable Cities and Society 36 (2018) 272–285
For New York City, Hurricane Sandy provides a useful test case of an
extreme event for this validation. The challenge is to quantify “normal”
neighborhood activity, and thus be able to measure deviations from it,
on a sufficiently granular time scale to assess the impact of a particular
event. We consider a number of potential proxies from high spatial and
temporal resolution data sources, including building energy and water
consumption (Martani, Lee, Robinson, Britter, & Ratti, 2012), Wi-Fi
Fig. 6. Screenshot of REDIviz Visualization tool (Available at: http://serv.
cusp.nyu.edu/projects/REDIviz).
C.E. Kontokosta, A. Malik
Fig. 7. (a) Daily, (c) weekly and (e) monthly average 311 service requests for bottom five tracts (left column) and (b) daily, (d) weekly and (f) monthly average for top five (right column)
census tracts based on categorical weights REDI scores.
connection and usage activity (Kontokosta & Johnson, 2017;
Kontokosta, 2016; Martani et al., 2012), social media activity (Kats,
Qian, Kontokosta, & Sobolevsky, 2017; Stewart & Wilson, 2016), public
transit ridership (Barry, Newhouser, Rahbee, & Sayeda, 2002; Reddy,
Lu, Kumar, Bashmakov, & Rudenko, 2009), and 311 service requests
(Wang, Qian, Kats, Kontokosta, & Sobolevsky, 2017; O’Brien, 2016).
Each source has limitations. Building energy and water consumption
data are not available at the frequency and granularity needed given
sample size requirements. Wi-Fi usage data holds promise, although
such data are not widely available. Social media data, such as Twitter,
provides an interesting glimpse into neighborhood dynamics based on
the location and frequency of activity; however, geocoded records ac-
count for only approximately three percent of all Twitter activity. In
addition, social media data are skewed demographically and socio-
economically, thus potentially limiting the representativeness of the
data with respect to the entire population. Data from transit ridership
can also provide a viable proxy for neighborhood activity. With respect
to measuring the impact of an event; however, transit data may present
an endogeneity problem in that ridership may be affected by system-
level closures or re-routing caused by the event itself.
New York City's 311 service request system provides a mechanism
for citizens to report problems or other unsafe conditions by phone or
online (Kontokosta, Hong, & Korsberg, 2017). These requests
280
Sustainable Cities and Society 36 (2018) 272–285
(complaints) provide one measure of neighborhood activity: it can be
expected that given a community's demographic profile and physical
infrastructure (building, land use, topography, etc.) characteristics, a
regular pattern of complaints can be observed. To explore this, and the
potential to utilize these complaints as a proxy for neighborhood ac-
tivity, 311 service request records are collected from 2010–2015 from
New York City's Open Data portal, which accounts for more than
12,000,0000 observations. All request types are included, and absolute
and normalized request activity aggregated to the census tract level are
analyzed. From Fig. 1, we consider 311 as an inverse measure of
community activity (A), with the expectation that an anomalous event
would yield greater service request volume during both the impact and
recovery stages.
Fig. 7 shows the average number of daily, weekly and monthly 311
service requests from January 2010 to December 2015 for the bottom
five (left column) and top five (right column) census tracts, based on
their Categorical Weights REDI score. The top five tracts represent the
most resilient neighborhoods according to the REDI score and the
bottom five tracts represent the least resilient. Hurricane Sandy is in-
dicated by the vertical reference line labeled ‘29-Oct-12’. For the
communities with the lowest REDI scores, the average number of re-
quests is clearly affected by Hurricane Sandy, as significant peaks are
observed immediately after the event (impact period) on all three time
C.E. Kontokosta, A. Malik Sustainable Cities and Society 36 (2018) 272–285

Fig. 8. Average (a) monthly, (b) weekly and (c) daily number of 311 service requests for top five (green) and bottom five (red) sandy-affected REDI tracts. (For interpretation of the
references to color in this figure legend, the reader is referred to the web version of the article.)

1-year forecast generally hold as there is no significant change in

Table 3 complaints post-Sandy in the next year. We do observe a gradual rise in
Daily, weekly and monthly statistics of pre and post Sandy 311 complaints. the volume of complaints starting approximately a year after Sandy,
primarily due to population increases in the subject areas. For the
Pre-Sandy 311 calls Post-Sandy 311 calls bottom five tracts, the immediate spike in the number of complaints
Average of Top 5 Sandy- Daily Avg.: 3.7 Daily Avg.: 4.4 post-Sandy is significantly outside of the 95% forecast intervals of the
affected REDI tracts Daily Std. Dev.: 1.6 Daily Std. Dev.: 2.0 ARIMA(1,0,1) model, and the recovery time (trecovery) to return to the
Weekly Avg.: 25.6 Weekly Avg.: 30.8 forecast interval is approximately three months, with no significant
Weekly Std. Dev.: Weekly Std. Dev.: 7.3 long-term increasing trend in call volumes.
4.4
Monthly Avg.: Monthly Avg.: 134.3
Finally, to account for the discrepancy in the volume of complaints
110.9 between the most resilient and least resilient tracts, the weekly and
Monthly Std. Dev: Monthly Std. Dev: monthly requests were normalized by indexing the number of com-
12.2 23.2 plaints to the time period immediately preceding Hurricane Sandy (the
Average of Bottom 5 Daily Avg.: 1.3 Daily Avg.: 1.6 week before 29th Oct, 2012 in case of weekly requests, and the month
Sandy-affected REDI Daily Std. Dev.: 1.0 Daily Std. Dev.: 1.4 of October in the case of monthly requests). The number of requests for
tracts Weekly Avg.: 9.4 Weekly Avg.: 11.4 these two times are set to a reference value of 100, and the number of
Weekly Std. Dev.: Weekly Std. Dev.: 6.7
2.8
complaints for other time periods are normalized with respect to this
Monthly Avg.: 40.9 Monthly Avg.: 49.7 value to indicate a percentage change. Fig. 10 shows the monthly and
Monthly Std. Dev: Monthly Std. Dev: weekly normalized number of complaints for both groups, and the
7.1 22.4 difference in behavior attributable to Hurricane Sandy. The bottom five
REDI tracts returned to “normal” activity by January 2013 (trecovery = 3
months), while the top five REDI tracts remained unaffected. This
scales. Conversely, the average number of requests for the top five REDI
analysis provides an approach to empirical validation of the REDI
tracts (i.e. the most resilient neighborhoods) shows no significant shock
method in effectively measuring the resilience capacity of New York
due to Hurricane Sandy.
City neighborhoods, and the practical application of the REDI scores in
To further validate the REDI methodology, only those census tracts
evaluating a community's ability to rebound from a given event.
that were directly impacted by Sandy and located within the storm
surge area are analyzed. Fig. 8 shows the average number of monthly,
weekly and daily requests for the top five and bottom five Sandy-af- 6. Conclusions and future research
fected census tracts with respect to their Categorical Weights REDI
scores. Table 3 lists the mean and standard deviation for the pre-event The Resilience to Emergencies and Disasters Index (REDI) estab-
and post-event time periods for both groups. For the least resilient lishes a new framework to comprehensively quantify neighborhood-
tracts, a spike in the number of requests immediately following Hurri- level urban resilience capacity. As a proof-of-concept, the REDI method
cane Sandy (impact period) is observed that represents a greater than is applied to New York City census tracts as a proxy for neighborhood
three standard deviation perturbation above the pre-event mean. The boundaries. Tracts that are found to have the greatest and least resi-
recovery time (trecovery) needed for the least resilient tracts to return to lience capacity with respect to social infrastructure, physical infra-
stable post-event activity is approximately three months. The average structure, economic strength, and environmental conditions have been
number of requests for the most resilient tracts shows no significant highlighted. The least resilience communities are considered to be less
shock due to Hurricane Sandy across all three time scales. equipped to respond to and recover from future events given the se-
This variation in response to Sandy in the least and most resilient lected feature set of indicator variables.
tracts is further highlighted in Fig. 9, which shows the weekly average The REDI methodology involves creating a comprehensive geospa-
of 311 complaints for both groups (same as Fig. 8b) and the 1-year tial resilience data repository, identifying and extracting relevant in-
ARIMA(1,0,1) forecasts post-Sandy (29-Oct-12). The selection of p, d dicator variables at the appropriate spatial scale, cleaning the merged
and q parameters in the ARIMA model was made using standard time- dataset, calculating individual REDI scores, visualizing the results, and
series analysis techniques such as plotting autocorrelation (which validating the model. Hurricane Sandy served as the test case to vali-
showed a gradual decrease after 1 significant lag resulting in q = 1), date the REDI approach, and 311 service requests are used as the data
partial autocorrelation (resulted in similar observation after first sig- proxy for measuring neighborhood activity at the requisite temporal
nificant lag meaning p = 1), and a positive test for over-differencing granularity. Preliminary analysis shows that the REDI scores effectively
after differencing the data once (hence d = 0). No methods to account characterize resilience capacity for New York City neighborhoods, as
for various seasonal trends have been included. While the 95% con- evidenced by the validity tests across pre-event, impact, recovery, and
fidence intervals of the two ARIMA(1,0,1) models generally account for post-event periods.
pre-Sandy activity in both the top five and bottom five REDI tracts, We acknowledge the potential limitations of this analysis and op-
these forecast intervals diverge post-Sandy. For the top five tracts, the portunities for further work. First, the REDI model could be re-
formulated to include additional attributes that account for the spatio-

281
C.E. Kontokosta, A. Malik
Fig. 9. Weekly average 311 service requests and 1-year ARIMA(1,0,1) forecasts starting from Sandy for (a) top five and (b) bottom five Sandy-affected REDI tracts.
Fig. 10. Normalized avg. of highest-scoring (green) and lowest-scoring (red) REDI tracts w.r.t (a) monthly and (b) weekly 311 service requests. (For interpretation of the references to
color in this figure legend, the reader is referred to the web version of the article.)
temporal dynamics of the city. This would necessitate time-varying
indicators derived from high-frequency temporal data available at a
monthly, weekly, or daily cadence. Second, the weights of each in-
dicator variable could be iteratively improved to reflect the relative
impact of that variable on the neighborhood's resilience capacity.
Although our correlation analysis indicates independence of the in-
cluded indicators, additional sensitivity analysis of each variable's sig-
nificance to the overall REDI score can refine the applied weights (Tate,
2012). Finally, an expanded version of the neighborhood REDI could
account for the resilience capacity of surrounding neighborhoods. In
practice, a neighborhood with low resilience capacity, but surrounded
by high-resilience areas, should perform better (and recover more ef-
fectively) than a similar neighborhood adjacent to other neighborhoods
with limited resilience capacity. Additional validation events could be
used to test and quantify this effect.
An important contribution of this work is our validation metho-
dology. Although exploratory, it holds promise by defining neighbor-
hood activity levels and patterns using big data and non-traditional,
high-resolution datasets (Sevtsuk & Ratti, 2010). Using 311 service re-
quests as a proxy involves a degree of uncertainty that can be affected
by sample bias based on a particular neighborhood's propensity to
complain, which can be influenced by demographic and cultural fac-
tors, as well as technical issues of connectivity and access to technology.
However, this measure provides a reasonable proxy for neighborhood
activity and effectively functions as a preliminary validation of the
REDI approach.
Understanding neighborhood resilience capacity is not simply a
282
Sustainable Cities and Society 36 (2018) 272–285
theoretical endeavor; it is equally a practical one, and must be fostered
through community-based planning (Berkes & Ross, 2013). Therefore, it
is recommended that future investment decisions regarding enhancing
the resilience capacity of neighborhoods utilize the proposed REDI
methodology as a preliminary screening tool in order to identify which
communities require the most immediate attention and where return on
investment may be highest. The REDI scores provide both a spatial and
temporal comparative benchmark of resilience capacity that can be
used to evaluate resource allocation decisions across neighborhoods, as
well as to track a neighborhood's resilience capacity over time. The
integration of the four components of resilience capacity also allow
policymakers and community organizations to prioritize across phy-
sical, social, economic, and environmental dimensions of resilience.
The REDI score, then, can be used to provide an important bench-
marking capability for understanding and measuring local resilience
capacity.
Acknowledgements
The authors thank Schuyler Poukish and William Vidal for their
research assistance, and Bud Griffis, Magued Iskander, and Ilan Juran
for their comments and feedback. This work was supported, in part, by
a grant from the New York State Office of Storm Recovery through the
New York State Resiliency Institute for Storms and Emergencies (RISE).
Any opinions, findings, and conclusions expressed in this paper are
those of the authors and do not necessarily reflect the views of any
supporting institution.
C.E. Kontokosta, A. Malik Sustainable Cities and Society 36 (2018) 272–285

Appendix A. Comparing regional resilience measurement methods

Resilience Resilience Geographic Spatial and temporal Feature selection Validation methodc
measurement domainsa scale of scales methodb
method resilience
measure

Angeon and Bates Soc, Country Spatial: Countries; CA Comparison with similar previous
(2015) Econ & Env Temporal: Years work
Bates et al. (2014) Soc, Country Spatial: Countries; CA N/A
Econ & Env Temporal: Years
Bergstrand et al. Soc Community Spatial: Counties; ML (Principal Correlation Analysis between SoVI
(2015) Temporal: N/A Components Analysis) and CRI; Moran's I measure of spatial
auto-correlation
Briguglio et al. Econ Country Spatial: Countries; CA Correlation Analysis with GDP per
(2009) Temporal: N/A capita
Bruneau et al. (2003) Soc, Community N/A N/A Measurement of Post-Earthquake
Phy & Econ Recovery Time
Cai et al. (2016) Soc, Phy, Community Spatial: Census Block ML (K-means Regression analysis between
Econ & Env Groups; Temporal: clustering) & Stepwise resilience scores and indicators
Decade (2000-2010) discriminant analysis extracted from discriminant analysis
Cimellaro et al. Phy City Spatial: Cities; CA (Infrastructure Restoration curves for power, water
(2014) Temporal: Days Interdependency Index and gas delivery
using modified CCF)
Cutter et al. (2003) Soc, Community Spatial: Counties; CA Correlation Analysis between
Phy & Econ Temporal: N/A SoVI & Presidential Emergency
declarations
Cutter et al. (2010) Soc, Community Spatial: Counties; CA N/A
Phy & Econ Temporal: N/A (Pearson's) & Cronbach's
alpha
Flanagan, Gregory, Soc Community Spatial: Census Tracts; N/A Comparison of SVI with Katrina
Hallisey, Temporal: N/A drowning probabilities & mail
Heitgerd, and deliveries
Lewis (2011)
Foster (2012) Soc & Econ Community Spatial: Metropolitan N/A N/A
Areas (MSA); Temporal:
N/A
Fox-Lent et al. Soc, Phy, Community Spatial: Rockaway N/A N/A
(2015) Econ & Env Peninsula (NY);
Temporal: Months
Granot (1995) Soc & Econ Community N/A N/A N/A
Guillaumont (2009) Econ Country Spatial: Countries; N/A N/A
Temporal: Years
Highfield et al. Soc, Phy, Community Spatial: Households; CA N/A
(2014) Econ & Env Temporal: N/A
Lam, Qiang, et al. Soc, Country Spatial: Countries; ML (K-means Inferential validation through
(2015) Econ & Env Temporal: Years clustering & Discriminant classification functions
analysis)
Lam, Reams, et al. Soc, Community Spatial: Counties; ML (K-means Inferential validation through
(2015) Econ & Env Temporal: Decade clustering & Discriminant classification functions
(1998-2008) analysis)
Madrigano et al. Soc, Community Spatial: Census Tracts; CA Logistic regression with heat wave
(2015) Phy & Env Temporal: Years (2000- days predicting index quintiles
2011)
Miles and Chang Soc, Phy, Community Spatial: Public Use N/A N/A
(2011) Econ & Env Microdata Areas
(PUMA); Temporal:
Weeks
Orencio and Fujii Soc, Phy. Community Spatial: Coastal Analytical hierarchy Weighted Linear Combination of
(2013) Econ & Env communities in process (AHP) & Delphi AHP & ICBRR
Philippines; Temporal: technique
N/A
Peacock et al. (2010) Soc, Phy, Community Spatial: Counties; CA Correlation Analysis between CDRI,
Econ & Env Temporal: N/A CRIs and disaster indicators

283
C.E. Kontokosta, A. Malik
Razafindrabe et al. Soc, Phy, City Spatial: Cities;
(2009) Econ & Env Temporal: N/A
Renschler et al. Soc, Phy, Community N/A
(2010) Econ & Env
Sherrieb et al. (2010) Soc & Econ Community Spatial: Counties;
Temporal: N/A
Simpson (2008) Soc, Phy, City Spatial: Cities;
Econ & Env Temporal: N/A
Van Zandt et al. Soc Community Spatial: Census Block
(2012) Groups; Temporal: N/A
References
Adger, W. N., Hughes, T. P., Folke, C., Carpenter, S. R., & Rockström, J. (2005). Social-
ecological resilience to coastal disasters. Science, 309(5737), 1036–1039.
Aldrich, D. P., & Meyer, M. A. (2015). Social capital and community resilience. American
Behavioral Scientist, 59(2), 254–269.
Aldunce, P., Beilin, R., Handmer, J., & Howden, M. (2014). Framing disaster resilience:
The implications of the diverse conceptualisations of “bouncing back”. Disaster
Prevention and Management, 23(3), 252–270.
Angeon, V., & Bates, S. (2015). Reviewing composite vulnerability and resilience indexes:
A sustainable approach and application. World Development, 72, 140–162.
Ayyub, B. M. (2014a). Quantification and valuations of resilience for emergency man-
agement. Sustainable Development of Critical Infrastructure, 8, 91.
Ayyub, B. M. (2014b). Systems resilience for multihazard environments: Definition,
metrics, and valuation for decision making. Risk Analysis, 34(2), 340–355.
Bakkensen, L. A., Fox-Lent, C., Read, L. K., & Linkov, I. (2017). Validating resilience and
vulnerability indices in the context of natural disasters. Risk Analysis, 37(5),
982–1004.
Barker, K., Ramirez-Marquez, J. E., & Rocco, C. M. (2013). Resilience-based network
component importance measures. Reliability Engineering & System Safety, 117, 89–97.
Barry, J., Newhouser, R., Rahbee, A., & Sayeda, S. (2002). Origin and destination esti-
mation in New York City with automated fare system data. Transportation Research
Record: Journal of the Transportation Research Board, 1817, 183–187.
Bates, S., Angeon, V., & Ainouche, A. (2014). The pentagon of vulnerability and resi-
lience: A methodological proposal in development economics by using graph theory.
Economic Modelling, 42, 445–453.
Bergstrand, K., Mayer, B., Brumback, B., & Zhang, Y. (2015). Assessing the relationship
between social vulnerability and community resilience to hazards. Social Indicators
Research, 122(2), 391–409.
Berkes, F., & Ross, H. (2013). Community resilience: Toward an integrated approach.
Society & Natural Resources, 26(1), 5–20.
Bloomberg, M. (2013). A stronger, more resilient New York.
Bocchini, P., Frangopol, D. M., Ummenhofer, T., & Zinke, T. (2013). Resilience and sus-
tainability of civil infrastructure: Toward a unified approach. Journal of Infrastructure
Systems, 20(2), 04014004.
Bonstrom, H., & Corotis, R. (2012). Structural reliability and sustainable resilience.
Structures congress, ASCE, 2268–2278.
Bozza, A., Asprone, D., & Manfredi, G. (2015). Developing an integrated framework to
quantify resilience of urban systems against disasters. Natural Hazards, 78(3),
1729–1748.
Briguglio, L., Cordina, G., Farrugia, N., & Vella, S. (2009). Economic vulnerability and
resilience: Concepts and measurements. Oxford Development Studies, 37(3), 229–247.
Brownjohn, J., & Aktan, E. (2013). Improving resilience of infrastructure: The case of
bridges. Structures congress 2013: Bridging your passion with your profession, ASCE,
1812–1821.
Bruneau, M., Chang, S. E., Eguchi, R. T., Lee, G. C., O’Rourke, T. D., Reinhorn, A. M.,
Shinozuka, M., Tierney, K., Wallace, W. A., & von Winterfeldt, D. (2003). A frame-
work to quantitatively assess and enhance the seismic resilience of communities.
Earthquake Spectra, 19(4), 733–752.
Burnside-Lawry, J., & Carvalho, L. (2015). Building local level engagement in disaster risk
reduction: A Portuguese case study. Disaster Prevention and Management, 24(1),
80–99.
Cai, H., Lam, N. S.-N., Zou, L., Qiang, Y., & Li, K. (2016). Assessing community resilience
to coastal hazards in the lower Mississippi river basin. Water, 8(2), 46.
Cavallaro, M., Asprone, D., Latora, V., Manfredi, G., & Nicosia, V. (2014). Assessment of
urban ecosystem resilience through hybrid social-physical complex networks.
Computer-Aided Civil and Infrastructure Engineering, 29(8), 608–625.
Chang, S. E. (2003). Evaluating disaster mitigations: Methodology for urban infra-
structure systems. Natural Hazards Review, 4(4), 186–196.
Cimellaro, G. P., Reinhorn, A. M., & Bruneau, M. (2010). Framework for analytical
quantification of disaster resilience. Engineering Structures, 32(11), 3639–3649.
Cimellaro, G. P., Solari, D., & Bruneau, M. (2014). Physical infrastructure inter-
dependency and regional resilience index after the 2011 Tohoku earthquake in
Japan. Earthquake Engineering & Structural Dynamics, 43(12), 1763–1784.
Croope, S., & McNeil, S. (2011). Improving resilience of critical infrastructure systems
postdisaster: Recovery and mitigation. Transportation Research Record: Journal of the
Transportation Research Board, 2234, 3–13.
284
Sustainable Cities and Society 36 (2018) 272–285
N/A N/A
N/A N/A
CA Performance Comparison with SoVI
Inductive approach to N/A
measuring relative
importance
CA Correlation Analysis with estimated
response, impact & recovery
Cutter, S. L. (2016a). The changing context of hazard extremes: Events, impacts, and
consequences. Journal of Extreme Events, 3(02), 1671005.
Cutter, S. L. (2016b). The landscape of disaster resilience indicators in the USA. Natural
Hazards, 80(2), 741–758.
Cutter, S. L., Ash, K. D., & Emrich, C. T. (2014). The geographies of community disaster
resilience. Global Environmental Change, 29, 65–77.
Cutter, S. L., Barnes, L., Berry, M., Burton, C., Evans, E., Tate, E., & Webb, J. (2008). A
place-based model for understanding community resilience to natural disasters.
Global environmental Change, 18(4), 598–606.
Cutter, S. L., Boruff, B. J., & Shirley, W. L. (2003). Social vulnerability to environmental
hazards. Social Science Quarterly, 84(2), 242–261.
Cutter, S. L., Burton, C. G., & Emrich, C. T. (2010). Disaster resilience indicators for
benchmarking baseline conditions. Journal of Homeland Security and Emergency
Management(1).
Dale, A., Ling, C., & Newman, L. (2010). Community vitality: The role of community-level
resilience adaptation and innovation in sustainable development. Sustainability, 2(1),
215–231.
Ernstson, H., van der Leeuw, S. E., Redman, C. L., Meffert, D. J., Davis, G., Alfsen, C., &
Elmqvist, T. (2010). Urban transitions: On urban resilience and human-dominated
ecosystems. Ambio, 39(8), 531–545.
Flanagan, B. E., Gregory, E. W., Hallisey, E. J., Heitgerd, J. L., & Lewis, B. (2011). A social
vulnerability index for disaster management. Journal of Homeland Security and
Emergency Management, 8(1).
Foster, K. A. (2007). A case study approach to understanding regional resilience.
Foster, K. A. (2012). In search of regional resilience. Urban and Regional Policy and Its
Effects: Building Resilient Regions, 4, 24–59.
Fox-Lent, C., Bates, M. E., & Linkov, I. (2015). A matrix approach to community resilience
assessment: An illustrative case at rockaway peninsula. Environment Systems and
Decisions, 35(2), 209–218.
Francis, R., & Bekera, B. (2014). A metric and frameworks for resilience analysis of en-
gineered and infrastructure systems. Reliability Engineering & System Safety, 121,
90–103.
Gallopin, G. C. (2006). Linkages between vulnerability, resilience, and adaptive capacity.
Global Environmental Change, 16(3), 293–303.
Ganin, A. A., Massaro, E., Gutfraind, A., Steen, N., Keisler, J. M., Kott, A., Mangoubi, R., &
Linkov, I. (2016). Operational resilience: Concepts, design and analysis. Scientific
Reports, 6, 19540.
Gibbs, L., & Holloway, C. (2013). Hurricane Sandy after action: Report and recommendations
to Mayor Michael R. Bloomberg.
Godschalk, D. R. (2003). Urban hazard mitigation: Creating resilient cities. Natural
Hazards Review, 4(3), 136–143.
Gong, J. (2014). A remote sensing-based approach for assessing and visualizing post-
sandy damage and resiliency rebuilding needs. Construction research congress 2014:
Construction in a global network, ASCE, 1259–1268.
Gonzales, P., & Ajami, N. K. (2017). An integrative regional resilience framework for the
changing urban water paradigm. Sustainable Cities and Society, 30, 128–138.
Granot, H. (1995). Proposed scaling of the communal consequences of disaster. Disaster
Prevention and Management: An International Journal, 4(3), 5–13.
Grayson, J. M., & Pang, W. (2014). The influence of community-wide hurricane wind
hazard mitigation retrofits on community resilience. Structures congress 2014, ASCE,
1392–1402.
Guillaumont, P. (2009). An economic vulnerability index: Its design and use for inter-
national development policy. Oxford Development Studies, 37(3), 193–228.
Hemond, Y., & Robert, B. (2012). Preparedness: The state of the art and future prospects.
Disaster Prevention and Management: An International Journal, 21(4), 404–417.
Henry, D., & Ramirez-Marquez, J. E. (2012). Generic metrics and quantitative approaches
for system resilience as a function of time. Reliability Engineering & System Safety, 99,
114–122.
Highfield, W. E., & Brody, S. D. (2012). Evaluating the effectiveness of local mitigation
activities in reducing flood losses. Natural Hazards Review, 14(4), 229–236.
Highfield, W. E., Peacock, W. G., & Van Zandt, S. (2014). Mitigation planning: Why ha-
zard exposure, structural vulnerability, and social vulnerability matter. Journal of
Planning Education and Research. http://dx.doi.org/10.1177/0739456X14531828.
Holling, C. S. (1973). Resilience and stability of ecological systems. Annual Review of
Ecology and Systematics, 1–23.
Hosseini, S., Barker, K., & Ramirez-Marquez, J. E. (2016). A review of definitions and
measures of system resilience. Reliability Engineering & System Safety, 145, 47–61.
Islam, M. S., Swapan, M. S. H., & Haque, S. M. (2013). Disaster risk index: How far should
C.E. Kontokosta, A. Malik
it take account of local attributes? International Journal of Disaster Risk Reduction, 3,
76–87.
Jabareen, Y. (2013). Planning the resilient city: Concepts and strategies for coping with
climate change and environmental risk. Cities, 31, 220–229.
Jordan, E., & Javernick-Will, A. (2013). Indicators of community recovery: Content
analysis and Delphi approach. Natural Hazards Review, 14(1), 21–28.
Kats, P., Qian, C., Kontokosta, C., & Sobolevsky, S. (2017). Twitter Activity Timeline as a
Signature of Urban Neighborhood. arXiv preprint arXiv:1707.06122.
Kojima, K., Fujita, K., & Takewaki, I. (2014). Building earthquake resilience in sustainable
cities in terms of input energy. Sustainable Cities and Society, 12, 46–62.
Kontokosta, C. E. (2013). Energy disclosure, market behavior, and the building data
ecosystem. Annals of the New York Academy of Sciences, 1295(1), 34–43.
Kontokosta, C. E. (2014). Mixed-income housing and neighborhood integration: Evidence
from inclusionary zoning programs. Journal of Urban Affairs, 36(4), 716–741.
Kontokosta, C. E. (2015). Do inclusionary zoning policies equitably disperse affordable
housing? A comparative spatial analysis. Journal of Housing and the Built Environment,
30(4), 569–590.
Kontokosta, C. E. (2016). The quantified community and neighborhood labs: A frame-
work for computational urban science and civic technology innovation. Journal of
Urban Technology, 1–18.
Kontokosta, C. E., & Johnson, N. (2017). Urban phenology: Toward a real-time census of
the city using Wi-Fi data. Computers. Environment and Urban Systems, 64, 144–153.
Kontokosta, C. E., Hong, B., & Korsberg, K. (2017). Equity in 311 reporting:
Understanding socio-spatial differentials in the propensity to complain. Bloomberg
data for good exchange 2017.
Kontokosta, C. E., & Tull, C. (2016). Energyviz: Web-based eco-visualization of urban
energy use from building benchmarking data. 16th international conference on com-
puting in civil and building engineering.
Koraz, Y., & Gabbar, H. A. (2017). Risk analysis and self-healing approach for resilient
interconnect micro energy grids. Sustainable Cities and Society.
López-Cuevas, A., Ramírez-Márquez, J., Sanchez-Ante, G., & Barker, K. (2017). A com-
munity perspective on resilience analytics: A visual analysis of community mood. Risk
Analysis.
Lam, N. S., Qiang, Y., Arenas, H., Brito, P., & Liu, K.-b. (2015). Mapping and assessing
coastal resilience in the Caribbean region. Cartography and Geographic Information
Science, 42(4), 315–322.
Lam, N. S., Reams, M., Li, K., Li, C., & Mata, L. P. (2015). Measuring community resilience
to coastal hazards along the northern Gulf of Mexico. Natural Hazards Review, 17(1),
04015013.
Linkov, I., Bridges, T., Creutzig, F., Decker, J., Fox-Lent, C., Kröger, W., Lambert, J. H.,
Levermann, A., Montreuil, B., & Nathwani, J. (2014). Changing the resilience para-
digm. Nature Climate Change, 4(6), 407–409.
Linkov, I., Eisenberg, D. A., Bates, M. E., Chang, D., Convertino, M., Allen, J. H., Flynn, S.
E., & Seager, T. P. (2013). Measurable resilience for actionable policy. Environmental
Science & Technology, 47(18), 10108–10110.
Linkov, I., & Florin, M. (2016). International Risk Governance Council (IRGC) resource guide
on resilience. Lausanne, Switzerland: EPFL International Risk Governance Center.
Linkov, I., & Palma-Oliveira, J. M. (2017). Resilience and risk: Methods and application in
environment, cyber and social domains. Springer.
Madrigano, J., Ito, K., Johnson, S., Kinney, P. L., & Matte, T. (2015). A case-only study of
vulnerability to heat wave-related mortality in New York city (2000–2011).
Environmental Health Perspectives, 123(7), 672–678. http://dx.doi.org/10.1289/ehp.
1408178.
Magis, K. (2010). Community resilience: An indicator of social sustainability. Society and
Natural Resources, 23(5), 401–416.
Maity, R., Sharma, A., Nagesh Kumar, D., & Chanda, K. (2012). Characterizing drought
using the reliability-resilience-vulnerability concept. Journal of Hydrologic
Engineering, 18(7), 859–869.
Manyena, S. B., & Gordon, S. (2015). Bridging the concepts of resilience, fragility and
stabilisation. Disaster Prevention and Management, 24(1), 38–52.
Martani, C., Lee, D., Robinson, P., Britter, R., & Ratti, C. (2012). Enernet: Studying the
dynamic relationship between building occupancy and energy consumption. Energy
and Buildings, 47, 584–591.
Martin, R. (2011). Regional economic resilience, hysteresis and recessionary shocks.
Journal of Economic Geography, lbr019.
Matthews, E. C., Sattler, M., & Friedland, C. J. (2014). A critical analysis of hazard re-
silience measures within sustainability assessment frameworks. Environmental Impact
Assessment Review, 49, 59–69.
McArdle, A. L. (2014). Storm surges, disaster planning, and vulnerable populations at the
urban periphery: Imagining a resilient New York after superstorm sandy. Idaho Law
Review, 50, 19–47.
McPhearson, T., Pickett, S. T., Grimm, N. B., Niemelä, J., Alberti, M., Elmqvist, T., Weber,
C., Haase, D., Breuste, J., & Qureshi, S. (2016). Advancing urban ecology toward a
science of cities. BioScience, 66(3), 198–212. http://dx.doi.org/10.1093/biosci/
biw002.
Meerow, S., Newell, J. P., & Stults, M. (2016). Defining urban resilience: A review.
Landscape and Urban Planning, 147, 38–49.
Miles, S. B., & Chang, S. E. (2011). Resilus: A community based disaster resilience model.
Cartography and Geographic Information Science, 38(1), 36–51.
285
Sustainable Cities and Society 36 (2018) 272–285
Norris, F. H., Stevens, S. P., Pfefferbaum, B., Wyche, K. F., & Pfefferbaum, R. L. (2008).
Community resilience as a metaphor, theory, set of capacities, and strategy for dis-
aster readiness. American Journal of Community Psychology, 41(1–2), 127–150.
O’Brien, D. T. (2016). Using small data to interpret big data: 311 reports as individual
contributions to informal social control in urban neighborhoods. Social Science
Research.
Orencio, P. M., & Fujii, M. (2013). A localized disaster-resilience index to assess coastal
communities based on an analytic hierarchy process (AHP). International Journal of
Disaster Risk Reduction, 3, 62–75.
Ouyang, M., Dueñas-Osorio, L., & Min, X. (2012). A three-stage resilience analysis fra-
mework for urban infrastructure systems. Structural Safety, 36, 23–31.
Peacock, W. G., Brody, S. D., Seitz, W. A., Merrell, W. J., Vedlitz, A., Zahran, S., Harriss, R.
C., & Stickney, R. (2010). Advancing resilience of coastal localities: Developing, im-
plementing, and sustaining the use of coastal resilience indicators. A final report. Hazard
Reduction and Recovery Center.
Ranjan, R., & Athalye, S. (2009). Drought resilience in agriculture: The role of techno-
logical options, land use dynamics, and risk perception. Natural Resource Modeling,
22(3), 437–462.
Razafindrabe, B. H., Parvin, G. A., Surjan, A., & Shaw, R. (2009). Climate disaster resi-
lience: Focus on coastal urban cities in Asia. Asian Journal of Environment and Disaster
Management, 1(1).
Reddy, A., Lu, A., Kumar, S., Bashmakov, V., & Rudenko, S. (2009). Entry-only automated
fare-collection system data used to infer ridership, rider destinations, unlinked trips,
and passenger miles. Transportation Research Record: Journal of the Transportation
Research Board, 2110, 128–136.
Renschler, C. S., Frazier, A., Arendt, L., Cimellaro, G.-P., Reinhorn, A. M., & Bruneau, M.
(2010). A framework for defining and measuring resilience at the community scale: The
PEOPLES resilience framework. MCEER Buffalo.
Robert, B., Morabito, L., Cloutier, I., & Hémond, Y. (2015). Interdependent critical in-
frastructures resilience: Methodology and case study. Disaster Prevention and
Management, 24(1), 70–79.
Rose, A. (2004). Defining and measuring economic resilience to disasters. Disaster
Prevention and Management: An International Journal, 13(4), 307–314.
Sevtsuk, A., & Ratti, C. (2010). Does urban mobility have a daily routine? Learning from
the aggregate data of mobile networks. Journal of Urban Technology, 17(1), 41–60.
Shafieezadeh, A., & Burden, L. I. (2014). Scenario-based resilience assessment framework
for critical infrastructure systems: Case study for seismic resilience of seaports.
Reliability Engineering & System Safety, 132, 207–219.
Sharkey, T. C., Nurre, S. G., Nguyen, H., Chow, J. H., Mitchell, J. E., & Wallace, W. A.
(2015). Identification and classification of restoration interdependencies in the wake
of hurricane sandy. Journal of Infrastructure Systems, 04015007.
Sherrieb, K., Norris, F. H., & Galea, S. (2010). Measuring capacities for community re-
silience. Social Indicators Research, 99(2), 227–247.
Simmie, J., & Martin, R. (2010). The economic resilience of regions: Towards an evolu-
tionary approach. Cambridge Journal of Regions, Economy and Society, 3(1), 27–43.
Simpson, D. M. (2008). Disaster preparedness measures: A test case development and
application. Disaster Prevention and Management: An International Journal, 17(5),
645–661.
Stewart, M. C., & Wilson, B. G. (2016). The dynamic role of social media during hurricane
#sandy: An introduction of the STREMII model to weather the storm of the crisis
lifecycle. Computers in Human Behavior, 54, 639–646.
Sullivan, D., & Uccellini L. Hurricane/post-tropical cyclone sandy, October 22–29, 2012.
NOAA Service Assessment, NOAA 1.
Tate, E. (2012). Social vulnerability indices: A comparative assessment using uncertainty
and sensitivity analysis. Natural Hazards, 63(2), 325–347.
Timashev, S. (2011). Resilience and preparedness of critical infrastructures. Proceedings
from International Conference on Vulnerability and Risk Analysis and Management.
ICVRAM 2011 and the International Symposium on Uncertainty Modeling and Analysis,
ISUMA.
Van Zandt, S., Peacock, W. G., Henry, D. W., Grover, H., Highfield, W. E., & Brody, S. D.
(2012). Mapping social vulnerability to enhance housing and neighborhood resi-
lience. Housing Policy Debate, 22(1), 29–55.
Wagner, M., Chhetri, N., & Sturm, M. (2014). Adaptive capacity in light of hurricane
sandy: The need for policy engagement. Applied Geography, 50, 15–23.
Wallace, D., & Wallace, R. (2008). Urban systems during disasters: Factors for resilience.
Ecology and Society, 13(1), 18.
Weichselgartner, J. (2001). Disaster mitigation: The concept of vulnerability revisited.
Disaster Prevention and Management: An International Journal, 10(2), 85–95.
Wang, L., Qian, C., Kats, P., Kontokosta, C., & Sobolevsky, S. (2017). Structure of 311
service requests as a signature of urban location. PloS one, 12(10), e0186314.
Winderl, T. (2014). Disaster resilience measurements: Stocktaking of ongoing efforts in de-
veloping systems for measuring resilience. Tech. rep. United Nations Development
Programme.
Zhang, Y., & Wan, G. (2009). How precisely can we estimate vulnerability to poverty?
Oxford Development Studies, 37(3), 277–287.
Zimmerman, R., Zhu, Q., de Leon, F., & Guo, Z. (2017). Conceptual modeling framework
to integrate resilient and interdependent infrastructure in extreme weather. Journal of
Infrastructure Systems, 23(4), 04017034.