International Journal of Disaster Risk Science (2023) 14:768–781 [Link].

com
[Link] [Link]/13753

ARTICLE

Urban Earthquake Vulnerability Assessment and Mapping

at the Microscale Based on the Catastrophe Progression Method
Deniz Gerçek1 · İsmail Talih Güven2

Accepted: 4 October 2023 / Published online: 6 November 2023

© The Author(s) 2023

Abstract
Vulnerability assessment and mapping play a crucial role in disaster risk reduction and planning for adaptation to a future
earthquake. Turkey is one of the most at-risk countries for earthquake disasters worldwide. Therefore, it is imperative to
develop effective earthquake vulnerability assessment and mapping at practically relevant scales. In this study, a holistic
earthquake vulnerability index that addresses the multidimensional nature of earthquake vulnerability was constructed. With
the aim of representing the vulnerability as a continuum across space, buildings were set as the smallest unit of analysis. The
study area is in İzmit City of Turkey, with the exposed human and structural elements falling inside the most hazardous zone
of seismicity. The index was represented by the building vulnerability, socioeconomic vulnerability, and vulnerability of the
built environment. To minimize the subjectivity and uncertainty that the vulnerability indices based on expert knowledge are
suffering from, an extension of the catastrophe progression method for the objective weighing of indicators was proposed.
Earthquake vulnerability index and components were mapped, a local spatial autocorrelation metric was employed where
the hotspot maps demarcated the earthquake vulnerability, and the study quantitatively revealed an estimate of people at
risk. With its objectivity and straightforward implementation, the method can aid decision support for disaster risk reduction
and emergency management.

Keywords Catastrophe progression method · Earthquake vulnerability index · Hotspots · Microscale · Turkey

1 Introduction earthquake disasters worldwide for urban mortality and eco-

nomic loss (Brecht et al. 2013). One of the most damaging
Given the specific nature of its rapid onset, unpredictability, earthquakes in the country, the İzmit earthquake of 1999
and destruction potential, earthquakes pose a serious threat (Mw 7.8, also called the Marmara earthquake) caused the
to cities, placing their assets and inhabitants at risk. Earth- death of over 18,000 people and 37,000 people were injured,
quakes do not affect every country, region, city, or even part and about 95,000 houses were severely damaged (Bağcı
of a city equally (Wisner et al. 2004). Although developed et al. 2000). Even though it has not been long past 1999, an
countries are at risk for their high-valued assets, developing earthquake hazard is not a remote possibility for the city. A
countries are more at-risk regarding fatalities, causalities, major, Mw ≤ 7.6 event is expected on the Marmara segment
and short- and long-term impacts (Pelling 2003; Peduzzi of the fault (west) in the next half century with a probability
et al. 2009). of approximately 50% (Şengör et al. 2005). İzmit City is
Turkey is one of the developing countries highly sus- expected to be affected by an earthquake in this segment.
ceptible to earthquakes. It is the most at-risk country for Under the threat of a probable earthquake, the seismic
risk of the city remains a critical issue to be addressed.
Although the city has developed since 1999, there is still
* Deniz Gerçek
denizgercek@[Link] evidence of needs for improvement regarding the building
stock, socioeconomic status of the inhabitants, and the built
1
Department of City and Regional Planning, Faculty environment.
of Architecture, İzmir Institute of Technology, 35430 Izmir, This study aimed at vulnerability assessment and map-
Turkey
ping to demarcate vulnerable places encompassing diverse
2
Department of Geophysics Engineering, Kocaeli University, components that produce vulnerability at the part of İzmit
41001 Kocaeli, Turkey

13
Vol:.(1234567890)
International Journal of Disaster Risk Science 769

City with very high seismicity using a data-driven approach, (Barbat et al. 2009). Social vulnerability is described as
advancing visual communication of the results, and making socioeconomic status and characteristics that make people/
a crude estimate of the inhabitants that are at the highest societies susceptible to hazards (Cutter et al. 2003). Even
risk. though physical and social conditions may have diverse
courses of change over space and time, they are considered
1.1 Disaster Risk and Vulnerability Concept to be inextricably linked together to produce high-risk con-
ditions, where the former is indicative of the latter (Rashed
Disaster risk is a product of co-existence of hazard, expo- and Weeks 2003). Socioeconomic conditions often force
sure, and vulnerability. While hazard and exposure are com- particular communities to live in hazard-prone areas and
ponents of disaster risk that are barely modifiable, vulnera- homes of poor structural condition (McEntire 2012) and
bility is. In this regard, there has been a shift from hazards to are considered the causes of the physical vulnerability in
a vulnerability/resilience paradigm (Haque and Etkin 2006; many cases (Barbat et al. 2009). Built environment capaci-
McEntire 2012) and an increased recognition that disaster ties, determined by the grain, density, and distribution of
risks on human society cannot be reduced without a focus facilities such as parks, and so on, affects location-specific
on vulnerability and its quantitative assessment. vulnerabilities (Wamsler et al. 2013). These inequalities
Vulnerability is a complex and multifaceted concept across the space produce location-specific vulnerabilities
(Bohle 2001), and there are many conceptualizations of the communicated formerly as “hazard of places” (Cutter 1996).
term, as gathered in the review by Diaz-Sarachaga and Jato- A common interest of vulnerability assessment is repre-
Espino (2019). Among many others, a widely recognized senting the diversity of vulnerability across space. Congru-
definition suggests that vulnerability reflects the suscepti- ently, mapping of earthquake vulnerability has become a
bility or the intrinsic predisposition to be affected or the trend in disaster risk reduction studies. The complex and
conditions that favor or facilitate damage (Cardona 2004). multifaceted nature of the vulnerability concept is also
These predispositions and conditions for seismic vulner- reflected in a broad spectrum of spatial scales, for exam-
ability typically exhibit physical, environmental, and socio- ple, national, regional, local, and household levels (Diaz-
economic aspects (Carreño et al. 2007; Barbat et al. 2009). Sarachaga and Jato-Espino 2019; Jaimes et al. 2022). Studies
We describe seismic vulnerability in this study as physical, that assess intracity variation of vulnerabilities commonly
socioeconomic, and environmental susceptibilities or condi- use census units, for example, ward or neighborhood, or the
tions of limited capacity that facilitate harm in an earthquake census blocks, as mapping units. Studies that take smaller
event. units as a basis, for example, buildings, are scarce and usu-
ally focus on the structural vulnerability (Zhai et al. 2019;
1.2 Vulnerability Assessment Method and Indices Pavić et al. 2020), or they are based on standalone commu-
nity surveys, rather than open census data (Birkmann 2006;
Vulnerability assessment has become a method to commu- Ebert et al. 2009). However, in urban environments of com-
nicate the disaster risk to the decision makers for risk reduc- plexity, interactions across space generate spatial variations
tion and adaptation (Birkmann 2006; Lindlay et al. 2011; of characteristics regardless of the administrative units or
Fekete 2019). A map of vulnerability supplemented with an census tracks that are for management purposes. Assigning
estimate of the rate of people at risk is key to loss modeling an aggregate measure of characteristics of elements to units
and emergency and disaster risk management (Aubrecht or groups, without any relation between them being demon-
et al. 2012). Vulnerability is a multidimensional construct strated, may also lead to methodological error called “eco-
that cannot be measured easily with a single variable (Cutter logical fallacy” (Jones and Andrey 2007), which may cause
and Finch 2008). A Multi-criteria decision analysis (MCDA) drawing inaccurate conclusions about the vulnerability and
with weight allocation is the most widely used technique the households/individuals that exhibit it. At the microscale,
to address urban vulnerability (Diaz-Sarachaga and Jato- it becomes essential to represent spatial variation based on
Espino 2019) where a composite of multiple quantitative the smallest unit, that is, buildings.
variables with an aggregative formula results in a single In constructing a vulnerability index, the selection of
index score, called vulnerability index. indicators and the weighting are considered major challenges
Several physical, environmental, and socioeconomic fac- (Zhang et al. 2017; Ziarh et al. 2021). Although numerous
tors are considered responsible for producing the vulnerabil- studies have used equal weights (Zhang et al. 2017; Diaz-
ity of places (Carreño et al. 2007; Cutter and Finch 2008; Sarachaga and Jato-Espino 2019), a common disposition
Carreño et al. 2017). Although physical vulnerability that is that indicators’ contribution to overall vulnerability is
represents the fragility of the physical structures is crucial, not equal, and weighing the indicators based on their rela-
comprehensive or holistic earthquake vulnerability covering tive importance is more likely to represent the real situa-
overall predispositions is a far more encompassing concept tion. There are broadly two approaches to weighing—the

13
770 Gerçek and Güven. Urban Earthquake Vulnerability Assessment and Mapping at the Microscale

subjective and objective methods—that both have certain CP method advancing entropy weights for objective ranking
drawbacks. The subjective methods—the so-called knowl- of the indicators.
edge-driven or expert-based methods—rely on the experts’ The current study is motivated to objectively address the
rankings or a consensus. The vast amount of studies that multidimensional nature of earthquake vulnerability. We
employ knowledge-driven methods such as AHP, TOPSIS, acknowledge that earthquake vulnerability exhibits physi-
and so on, however, suffers from the subjectivity that intro- cal, socioeconomic, and environmental aspects and adopted
duces uncertainty and controversy into an index (Chen et al. a holistic approach that accommodates the below items as
2010; Zhang et al. 2017; Rodcha et al. 2019). The objective components of earthquake vulnerability index:
methods—the so-called data-driven methods—attempt to
overcome the subjectivity of the knowledge-based meth- 1. Building vulnerability, as the structural fragility of the
ods by determining weights based on the data. However, exposed residential building stock (construction system,
weights derived from data may not exactly reflect the actual period, plan irregularity, and so on);
weights of indicators and may deviate the results from the 2. Socioeconomic vulnerability that produces susceptibili-
real situation (Zhang et al. 2021). To overcome the challenge ties for individuals and households (age, gender, house-
of weighing, this study adopted the catastrophe progression hold structure, income, and so on); and
(CP) method that uses relative importance of indicators, 3. Vulnerability of the built environment that may reduce
rather than direct use of hard to determine weights (Zhang the coping capacity (building density, distance to assem-
et al. 2017). bly areas, and so on).
Catastrophe progression, as a data-driven approach stem-
ming from the study of dynamic systems, combines catas- To conduct the analysis at a level of detail that can rep-
trophe theory with fuzzy mathematics (Cheng et al 1996) to resent the vulnerability as a continuum across space and
develop a fuzzy membership function of system state (Gao avoid ecological fallacy, vulnerability mapping at the micro-
et al. 2020). In CP, responses to changes in the internal scale, taking buildings as the smallest unit of analysis, was
values of each factor are intrinsically evaluated consider- adopted. Spatial clustering that advances communication by
ing their ranking of importance that reduces subjectivity adding analytical abilities (Aksha et al. 2019) was employed.
(Ahmed et al. 2015; Xue et al. 2022). Song et al. (2020) A classification was conducted on the vulnerability scores
and Mostafa (2022) stated that the CP method has been to reveal an estimate of the number of people that are highly
demonstrated to have a unique advantage in dealing with vulnerable.
uncertainty and is increasingly employed in holistic indi- In the present study, indicators of earthquake vulnerabil-
ces, particularly the vulnerability indices in the last decade ity were chosen based on the relevance and availability of
(for example, Zhang et al. 2017; Ziarh et al. 2021; Zheng the data at the building and household scales. We particu-
and Huang 2023). The CP method is also characterized by larly sought to employ data available or can be calculated
its perceptiveness to gradual changes in a system that may using GIS tools at no cost.
cause sudden shifts, resembling a catastrophe. Conventional
data-driven methods that use calculus to reach a vulner-
ability index score, however, may fail to address multiple 2 Materials and Methods
characteristics of an individual or neighborhood that interact
and amplify each other to a fragile condition. We consider A holistic earthquake vulnerability assessment and mapping
vulnerability as an example to this. Intrinsically, the CP at a microscale taking buildings as minimum mapping unit
method favors the relative importance of control variables was proposed. The catastrophe progression (CP) method was
where, ranking indicators regarding their importance is an adopted for its perceptiveness to gradual changes that may
issue to be addressed (Shen et al. 2020; Du et al. 2022). evolve into highly vulnerable conditions and merit in mini-
The CP executes hierarchical recursive calculations apt for mizing subjectivity. In this section, the study area and the
a system tree. Therefore, rather than ranking indicators all vulnerability index components and indicators are described.
at once, grouping indicators with shared context into cat- The CP method and its steps are also explained.
egories and then ranking them is a compromise approach
(Shen et al. 2020). The ranking is either knowledge-based 2.1 Study Area
or data-driven. By employing data-driven methods, for
example, the mean square difference method (Jin and Zhang İzmit City of Kocaeli Province is located at the İzmit Bay
2020), weights based on components share on total variance of the Marmara Sea, in the northwest of Turkey. The city is
in principal component analysis (PCA) (Wood et al. 2010; in the hinterland of megacity Istanbul and is the industrial
Aksha et al. 2019), and entropy weights (Wu et al. 2022), heart of Turkey. The city center is close to the sea level,
subjectivity can be avoided. We adopted an extension of the with altitude rapidly increasing towards the north of the

13
International Journal of Disaster Risk Science 771

city where limited settlement activity has occurred, and 6.56 ­km2 was 12,651 persons/km2, indicating quite a dense
densification has taken place in the immediate surround- settlement pattern.
ings of the city center. The city is located within the North
Anatolian Fault Zone (NAFZ) extending all the way in 2.2 Earthquake Vulnerability Components
northern Turkey. and Indicators at the Microscale
Due to the high alluvial sediment thickness (250–500 m)
in the region, soil amplification is extreme (Özalaybey Acknowledging that earthquake vulnerability exhibits physi-
et al. 2008). The Vs30 velocity below 360 m/s in the study cal, socioeconomic, and environmental aspects, we iden-
area indicates the two most hazardous soil groups, D and tified indicators for the components of earthquake vulner-
E, according to the NEHRP (National Earthquake Hazard ability based on a literature review as follows and also in
Reduction Program) soil classification. The soil charac- Table 1.
teristics as sub-components of the seismic hazard suggest
that a part of the city center is a 1st level seismic hazard 2.2.1 Building Vulnerability
zone (Özalaybey et al. 2008) as shown in Fig. 1b (in red).
The severity significantly reduces at the surrounding parts The indicators of the building vulnerability in this study
of the zone. Therefore, the study area was restricted to the primarily stem from the building classification method
urban area intersecting with this hazard zone. As the hazard of rapid visual screening (RVS). A building classification
is steady across the area, all residential building stock and groups buildings with similar materials and seismic force-
inhabitants within the study area are presumed to be exposed resisting systems together, facilitating the fast identification
to earthquake. of a building’s susceptibility to an earthquake (FEMA 2015).
The study was conducted for the residential building The RVS and other more sophisticated methods use many
stock in the study area that accounts for 75% of all build- parameters and sometimes necessitate in-situ tests, which are
ings, that is, 6540 buildings accommodating about 83 thou- not practically applicable at the urban scale. Hence, quick
sand people by 2021. The gross density in the study area of methods of assigning vulnerability scores to buildings are

Fig. 1  Location of the study area on the earthquake hazard map of Turkey (a), and the study area in the 1st level seismic hazard zone (red) (b).
Source a AFAD (2023) and b adapted from Özalaybey et al. (2008)

13
772 Gerçek and Güven. Urban Earthquake Vulnerability Assessment and Mapping at the Microscale

Table 1  Earthquake vulnerability index: components, indicators, and data properties

Components Indicators (Xi) Measurement scale Data range Sign
(per building)

Building vulnerability (VPhy) Construction system (x1) Categorical 1, 2, 3, 4, 5 +

Construction period (x2) Categorical 1, 2 +
Building position (x3) Categorical 1, 2, 3, 4 +
Plan irregularity (x4) Continuous 0.478–1 +
Height/floor area (x5) Continuous 0.02–0.63 +
Socioeconomic vulnerability (VSoc) Rate of age 65+ (x6) Continuous 0–100% +
Rate of age ≤ 5 (x7) Continuous 0–66.66% +
Rate of females (x8) Continuous 0–100% +
Rate of single 65+ households (x9) Continuous 0–100% +
Rate of female-headed households (x10) Continuous 0–100% +
Rate of age ≤ 5 and ≥ 3 households (x11) Continuous 0–100% +
Market value of real estate (TRY/m2) (x12) Continuous 11,797–200,019 –
Rate of socially aided households (x13) Continuous 0–100% +
Built environment vulnerability (VEnv) Density of buildings (per building in 50 m ring buffer) Continuous 0.48–71% +
(x14)
Density of people (per building in 50 m ring buffer) (x15) Continuous 0.55–689 +
Floor/parcel ratio (x16) Continuous 2.55–100% +
Mean NDVI (per building in 50 m ring buffer) (x17) Continuous 0.089–0.461 –
Distance to assembly areas (m) (x18) Continuous 100–1600 +

often more convenient (Barbat et al. 2009; Fischer et al. “Relative position” is the type of adjacency of the build-
2022). Since there is no complete list of RVS indicators for ing where an isolated building responds better to earthquake
the study area, available data necessary for building vul- compared to attached buildings, and the position of a build-
nerability assessment were used in building classifications ing, for example, in the middle, at the end, or at the corner
(Silva et al. 2022; Zhang et al. 2023). of a block, matters (FEMA 2015; RYTEIE 2019). An exten-
Some of the structural attributes were obtained from the sion of the technique “proximity” to automatically determine
local municipality, which mainly archives these for building the relative positions of the buildings was employed. “Plan
permits and the data are stored on a building-by-building irregularity” is a geometric measure of the extent to which a
basis. “Construction system” of existing buildings readily building plan is regular and symmetrical to reduce torsional
available as FEMA building types S3, C2, C1, C3, and URM response (Davidson 1997). “Minimum bounding rectangle”
were ranked according to the basic scores of FEMA 154 for a convex hull of the building vertices was calculated
(FEMA 2015) from 1 to 5 respectively, where 5 represents to quantify the irregularity in plan. The “height/floor area
the highest seismic vulnerability of an existing building in ratio” was calculated using the building height divided by
the study area. Most common buildings in the study area the floor area, where the higher ratio is an indication of the
were of C3 type (concrete frame with unreinforced masonry fragility of a building.
infill walls) with moderate to high vulnerability, and URM
(unreinforced masonry bearing wall buildings) with high 2.2.2 Socioeconomic Vulnerability
vulnerability. The “construction period” was available only
as before and after 2007, which is a milestone of the release Most of the socioeconomic indicators in this study originate
of the most comprehensive and legally binding regulation on from census data at the household level of the Citizenship
seismic building codes in Turkey (DBYBHY 2007). Affairs of the Republic of Turkey. Specific age groups or
Other indicators are about buildings’ physical proper- gender, namely elderly, children, and female are common
ties that may cause defects in their seismic response. They groups deemed socially vulnerable in most studies (Wood
are relative position (adjacency), plan irregularity (FEMA et al. 2010; Armaş et al. 2017; Zhang et al. 2017; Sauti et al.
2015), and height/floor area ratio that makes tiny buildings 2021). Based on the census data, the occupancy rate of
highly vulnerable (Shakya et al. 2018). We calculated these elderly, children, and females were calculated per building.
indicators using metrics via GIS tools (ArcGIS 10.8) for Households with single elderly (Fischer et al. 2022), female-
repeatability. headed (Wood et al. 2010; McEntire 2012), and with three or

13
International Journal of Disaster Risk Science 773

more children (Armaş et al. 2017), which were considered influencing factors (Cheng et al. 1996). The status of the
disadvantaged, were calculated for their rate in the buildings’ state variable is formulated regarding the relative impor-
total households. Rate of the disadvantaged groups ranged tance of the control variables intrinsically by catastrophe
between 0 and 100% except for the children that may not models’ fuzzy membership functions (normalization) rather
occupy a whole building. People with low income are more than user-defined weights (Ahmed et al. 2015; Zhang et al.
susceptible to hazards (Cutter et al. 2003; Wood et al. 2010). 2017). Catastrophe progression was employed after stand-
As the income data at the household level were unavail- ardization of the raw values of indicators, followed by nor-
able, two proxy indicators (Ebert et al. 2009)—the “current malization for the estimation of fuzzy membership functions
market value of the real estate” and “households receiving for the state variable, which constitutes the core part of the
social aid”—were obtained from the municipality records. CP method.

2.2.3 Vulnerability of the Built Environment 2.3.1 Standardization

Besides physical and socioeconomic components, the urban For resolving incompatibility and imbalanced weighing, all
fabric also affects location-specific vulnerabilities (Pelling indicators were standardized using the range standardization
2003). High concentration and density create susceptible method and set to a dimensionless range between 0 and 1. As
conditions (Sauti et al. 2021; Jaimes et al. 2022). In this the CP method is designed for positively related control vari-
component of the human-environment interface, three den- ables, all indicators positively related to vulnerability were
sity indicators—the density of buildings, the density of peo- transformed using Eq. 1, and indicators negatively related
ple, and the density of the parcel (floor/parcel ratio)—were to vulnerability were transformed using Eq. 2.
adopted.
Greenness that was selected as an indicator is not a direct xi − xmin
xi� = (1)
measure of vulnerability, but a proxy for the quality of the xmax − xmin
built environment and the abundance of green and tree cover
that may provide services (Ebert et al. 2009). Greenness xmax − xi
xi� = (2)
was obtained from Normalized Difference Vegetation Index xmax − xmin
(NDVI) using Landsat 8 images of July 2021. Assembly
areas are of crucial importance for increasing coping capaci- where xi is the original value of the indicator X , xmax and xmin
ties after an earthquake (Allan and Bryant 2011). Hence are the maximum and minimum values of the indicator, xi′ is
each building’s distance to assembly areas was taken as an standard values of the indicator X , and xi� ∈ [0, 1].
indicator for the component. All of the indicators of the built
environment component were calculated using proximity 2.3.2 Normalization Using Catastrophe Models
analysis. Buffers excluding the building itself (ring buffer)
was produced to calculate zonal statistics for each residential In the CP method, the system is expressed with potential
building. function via the state variables and the external control vari-
Earthquake vulnerability index is represented by the three ables. The normalization formulas are obtained by decom-
components and the selected set of 18 indicators (x1, x2, ..., posing the bifurcation set equation (Zhang et al. 2017). The
x18) for each case (building), as seen in Table 1. The first descriptions and normalization formulas of the catastrophe
three physical indicators are ordinal categorical variables models describe all possible discontinuities of a single state-
with rank order that are commonly used as input in CP (e.g., dependent variable controlled by up to five control varia-
Jenifer and Jha 2017; Ahmed et al. 2015). The signs indi- bles, namely fold, cusp, swallowtail, butterfly, and wigwam
cate whether the indicator is positively (+) or negatively (−) (Cheng et al. 1996; Wood et al. 2010) (Table 2).
related to vulnerability. As shown in Table 2, x represents response variables,
and a , b , c , d , and e represent control variables. There
2.3 The Catastrophe Progression Method are two principles in implementing normalized formulas:
complementary and non-complementary. If there is a cor-
The CP method is based on the catastrophe theory devel- relation between the control variables (a, b, c, … ) , vari-
oped by Thom (1975) and deals with discontinuous changes ables are assumed to fill up the deficiency of each other,
caused by continuous changes that typically characterize and they are complementary. Therefore, ( their mean)value
catastrophic phenomena. In the catastrophe theory, system can be used for the system, that is, x = xa + xb + xc ∕3. If
function variables are divided into state variables (depend- there is no apparent correlation between the control vari-
ent), which are the internal variables of the system, and ables (a, b, c, … ), control variables cannot be replaced with
control variables (independent), which are the external each other, and they are non-complementary. Therefore,

13
774 Gerçek and Güven. Urban Earthquake Vulnerability Assessment and Mapping at the Microscale

Table 2  Descriptions of catastrophe models

Catastrophe models State Control Potential function Normalization formula
vari- variables
able

Fold 1 1 f (x) = x3 + ax xa = x1∕2

Cusp 1 2 4 2
f (x) = x + ax + bx xa = x1∕2, xb = x1∕3
Swallowtail 1 3 f (x) = x5 + ax3 + bx2 + cx Bifurcation xa = x1∕2, xb = x1∕3,xc = x1∕4
Butterfly 1 4 6 4 3
f (x) = x + ax + bx + cx + dx 2 xa = x1∕2, xb = x1∕3, xc = x1∕4, xd = x1∕5
Wigwam 1 5 7 5 4 3
f (x) = x + ax + bx + cx + dx + ex2
xa = x1∕2, xb = x1∕3, xc = x1∕4,
xd = x1∕5 , xe = x1∕6

Source Cheng et al. (1996) and Woodcock (2006)

the smallest value{of control } variables is used for the sys- where n is the number of samples, m is the number of indica-
tem, i.e., x = min xa , xb , xc (Jenifer and Jha 2017; Zhang tors, and 0 ≤ 𝜔j ≤ 1, Σ𝜔j = 1. Entropy weight calculation is
et al. 2017). In catastrophe models, the system’s behavior is usually followed by a final step of the comprehensive index:
determined by the priority of control variables. Therefore, an m
essential step before normalization by catastrophe models is
∑
Sj = yij 𝜔i (6)
the sorting of control variables in priority order (Shen et al. j=1
2020; Du et al. 2022). In an attempt to avoid the subjectiv-
ity of user-defined ranking, the entropy weights approach However, in the present study, Sj was not used, instead,
was adopted. weights (𝜔i ) were employed to rank indicators per category.

2.4 Entropy Weighted Ranking

3 Results and Discussions
Entropy weight employs statistical variation to determine
weights of factors that control a dependent variable (Jaafari The system (state variable), earthquake vulnerability herein,
et al. 2014). The smaller the information entropy of an index, was divided into subsystems, building vulnerability (VPhy),
the greater the variation degree of the index value. This socioeconomic vulnerability (VSoc), and built environment
means that when a particular indicator has more differentia- vulnerability (VEnv), each comprising their own indica-
tion degree, the greater its role in the holistic evaluation and tors. As an initial step of the catastrophe progression (CP)
hence the greater its weight (Gao et al. 2020). method, a tree-like structure of the system was established
The steps to calculate the entropy weight of each control that portrays the multidimensionality and the hierarchy of
variable are shown in Eqs. 3–6. The first step is the stand- the subsystems and the control variables of the earthquake
ardization of the original values. Pij is the standardized value vulnerability index (EVI).
of ith index, jth sample:
xij 3.1 Definition of the Index System
Pij = ∑n (3)
j=1 xij Some studies rank indicators all at once (for example, Yari-
yan et al. 2020). However, ranking indicators that are very
Ei is the information entropy value of ith index:
different from each other, like comparing an indicator of
∑n
P .lnPij social vulnerability with an indicator of building vulnerabil-
j=1 ij
Ei = (4) ity, is not convenient. We followed an approach of defining
lnn
an index system with categories and subcategories and rank-
As the natural logarithm function ln(x), x ∈ [R ∣ x > 0]. ing indicators under their parent category (Shen et al. 2020).
lnPij is undefined where Pij = 0. Therefore, lnPij is set to 0 In the calculation of CP membership function values, the
where Pij = 0 for convenience (Zhu et al. 2020). 𝜔i is the complementary and non-complementary principle applies to
entropy weight of the ith index: the correlated and uncorrelated indicators, respectively. By
convention, for uncorrelated indicators, CP membership is
1−E
𝜔i = ∑m � i � (5) assigned as the minimum membership value. However, the
i=1 Ei empirical results in the present study caused CP membership
function values for subcategories assigned zero for many

13
International Journal of Disaster Risk Science 775

Fig. 2  Earthquake vulnerability index (EVI) system

cases (buildings). This is attributed to cases with indicator Target is divided into subsystems of VPhy, VSoc, and
values equal to zero, especially in the category of social vul- VEnv (components), sub-categories at the 3rd and 4th
nerability—for example, there may be no elderly persons in levels. The indicators x 1, x 2, …, x 18 are at the 5th and
a building—and according to the non-complementary rule, the basic level. The indicators under subcategories were
zero is assigned to the upper categories in the index sys- ordered regarding their entropy weights ( 𝜔i ), and the sub-
tem, finally ending up with zero vulnerability for numerous systems as well regarding aggregate entropy weights ( 𝜔 )
cases. We considered this situation obscuring the real merit of their control variables. The EVI system with categories,
of the data and hence followed a course of grouping corre- entropy weights, and ranking of indicators is shown in
lated indicators together that are also semantically related. Fig. 2.
We explored the correlation of the indicators besides their
semantics and employed correlation information to sub-
group indicators as well. Thus, we organized subgroups as 3.2 The Catastrophe Progression for Earthquake
child-related (x7 and x11), gender-related (x8 and x10), and Vulnerability Index, Mapping, and Evaluations
elderly-related (x6 and x9) that are correlated. The other sub-
category of VSoc, the income indicators, were both semanti- Four normalization formulas of the CP method were used
cally and statistically related (x12 and x13). Some indicators in this study based on the number of control variables,
under VEnv both show a semantic relationship and correla- named Fold catastrophe, Cusp catastrophe, Swallowtail
tion. Hence we subgrouped those indicators as (x14, x15, and catastrophe, and Wigwam catastrophe, respectively, as
x16). On the other hand, although having correlation with presented in Fig. 2.
other indicators, x17 and x18, were considered semantically Following the workflow, CP membership function val-
representatives of diverse issues—for example, distance to ues for each case (building) in the study area at multiple
assembly areas (x18) have no causal relationship with density levels were obtained. Membership values of the buildings
indicators, such as the density of buildings or people. Hence across the study area were mapped at selected levels of
those indicators were set independent based on semantics. the 2nd and the 1st (the target EVI). Spatial clustering
Similarly, although they may correlate, indicators of VPhy using Getis Ord Gi* on the resulting index score maps
represent different characteristics of buildings, making their was employed to demarcate clusters of high vulnerabil-
subgrouping inapt. Therefore, all five indicators of VPhy (x1, ity. Autocorrelation of each component was tested using
x2, x3, x4, and x5) were set independent. global Moran’s I to understand the strength of the spatial
In the present study, a five-level index system was pro- context. Pearson’s correlation was employed to assess the
posed; the first-level index being the target, that is EVI. relationship among the vulnerability components. Finally,

13
776 Gerçek and Güven. Urban Earthquake Vulnerability Assessment and Mapping at the Microscale

Fig. 3  Catastrophe progression membership values for building vulnerability (VPhy) (a), and Getis-Ord Gi* hotspot analysis for VPhy (b)

an estimate of people at risk is provided based on the EVI the most vulnerable cases. However, it is difficult to deter-
classification. mine class breaks based on synthetic CP scores as they do
not refer to actual vulnerability. Studies that employ CP use
3.2.1 Earthquake Vulnerability Mapping various classification methods including standard deviation
(Zhang et al. 2017), natural breaks (Ziarh et al. 2021; Zheng
Figures 3, 4, and 5 depict the resulting maps for each com- and Huang 2023), quantile (Shen et al. 2020), and bespoke
ponent at the 2nd level. Figure 6 depicts the comprehensive classification methods (Su et al. 2011). The quantile method
earthquake vulnerability index (EVI) maps. is adequate for evenly distributed data and natural breaks is
The CP scores of index maps were classified to better adequate for unevenly distributed data that show clustering.
communicate the results visually and to aid in quantifying The EVI scores for the study resemble a normal distribution

Fig. 4  Catastrophe progression membership values for socioeconomic vulnerability (VSoc) (a), and Getis-Ord Gi* hotspot analysis for VSoc (b)

Fig. 5  Catastrophe progression membership values for built environment vulnerability (VEnv) (a), and Getis-Ord Gi* hotspot analysis for VEnv
(b)

13
International Journal of Disaster Risk Science 777

Fig. 6  Catastrophe progression membership values for earthquake vulnerability index (EVI) (a), and Getis-Ord Gi* hotspot analysis for EVI (b)

that is attributed to several recursive normalization proce- mean of the dataset, which can enable results to be compared
dures of CP. Therefore, standard deviation classification was with a future situation of the same study area or with another
employed since it is best suited for data that conforms to a study area given the same dataset.
normal distribution (Longley et al. 2015). Standard devia- Based on standard deviation classification, five catego-
tion classification reveals how values are dispersed from the ries were defined: High Vulnerability (> 1.5 Std. Dev),
mean and is more reliable in terms of determining highest- High-Medium Vulnerability (0.5–1.5 Std. Dev), Medium
vulnerability cases, whereas other methods may undesirably Vulnerability (− 0.5 to 0.5 Std. Dev), Medium-Low Vul-
involve near-mean values into dispersed classes. Standard nerability (− 1.5 to − 0.5 Std. Dev.), and Low Vulner-
deviation classes were obtained by subtracting or adding ability (< − 1.5 Std. Dev). For the hotspot maps, color red
certain standard deviation (for example, 0.5, 1, 1.5) from the represents clusters of high vulnerability, and color blue

13
778 Gerçek and Güven. Urban Earthquake Vulnerability Assessment and Mapping at the Microscale

Table 3  Spatial autocorrelation Autocorrelation

and bivariate correlation of
vulnerability components Global Moran’s I Correlation VPhy VSoc VEnv EVI

VPhy 0.234** Pearson correlation 1 0.213** 0.065** 0.829**

VSoc 0.404** Pearson correlation 1 0.058** 0.682**
VEnv 0.542** Pearson correlation 1 0.237**
EVI 0.386** Pearson correlation 1

VPhy Building vulnerability, VSoc Socioeconomic vulnerability, VEnv Built environment vulnerability, EVI
Earthquake vulnerability index
**Significant at 99% confidence interval

Table 4  An estimate of the Vulnerability class Buildings Residents (official) People (nighttime) People (daytime)
quantity of the buildings and
people at risk in the study area High 194 1959 1469 600
High-medium 2027 20,538 15,403 5949
Medium 2476 35,143 26,357 9450
Medium-low 1271 17,918 13,438 4587
Low 572 7438 5578 1695
Total 6540 82,996

represents clusters of low vulnerability. Gray colors indi- is not correlated with the other two vulnerability dimensions
cate no statistically significant clustering. (physical and social).
The results show that earthquake vulnerability is more
concentrated in the western half of the study area in general. 3.2.3 Population at Risk
But the three components of vulnerability—VPhy, VSoc,
and VEnv—differ in localizations for the particular vulner- Given the necessity of quantification for enhanced risk
abilities they present. understanding, a crude estimation of the population at risk
was implemented. As the vulnerability is a spatiotemporal
construct due to density, human activities, and mobility
(Aubrecht et al. 2012), daytime and nighttime estimates were
3.2.2 Earthquake Vulnerability Evaluations
provided (Table 4).
The number of residents in official records was used
Besides the mapping of earthquake vulnerability, spatial
as basic data. However, people do not fully occupy their
autocorrelation was explored for a better understanding
homes at all times. Nighttime occupancy is usually
of the extent to which the vulnerability scores correlate in
accepted as approximately 75% (Ara 2014; Wei et al.
relation to space, that is, expressing clustering, dispersion,
2017) of total population. Daytime occupancy can be
or randomness. The global Moran’s I test results show that
represented as “home population” that is described by
all the components have statistically significant clustering
people that usually stay at home, that is, elderly, children,
across space. Ordered from greatest to least, spatial auto-
and unemployed (Ara 2014). For this study, nighttime
correlation degree is VEnv > VSoc > VPhy (Table 3). Built
population was calculated as 75% of residents (official),
environment vulnerability (VEnv) has the greatest spatial
and daytime population was calculated for each class as
autocorrelation, as its indicators are density-related metrics
number of elderly (age 65+), number of children (age
representing spatial context. The fact that VSoc's spatial
≤ 5), and unemployed approximated as 13.5% of the
autocorrelation is higher than that of VPhy is related to simi-
active population (age 15–64) according to the statisti-
lar communities showing clustering, wheareas the buildings
cal records of the city for the relevant year (TUIK 2021).
of common characteristics showing less thereof.
Crude estimate of people at high risk (that is, vulner-
Pearson’s correlation of the components was also
ability classes high and high-medium) constitute 20% and
explored. The results show that building vulnerability and
7.8% of the total population (82,996 people) for nighttime
social vulnerability are moderately and significantly cor-
and daytime, respectively.
related (Table 3), indicating coherence between social and
physical vulnerabilities. But built environment vulnerability

13
International Journal of Disaster Risk Science 779

4 Conclusion localities of the vulnerability represented by the convolu-

tions of the dimensions of vulnerability across space, the
In this study, a holistic earthquake vulnerability assess- method is considered robust despite the missing data and
ment and mapping that addresses the multidimensional the use of proxies.
nature of earthquake vulnerability at the microscale was
Acknowledgments This study was supported by the Disas-
conducted. The main contributions of this study are: (1) ter and Emergency Management Presidency under Project No.
minimizing subjectivity in the weight assignment by AFAD-UDAP-Ç-19-06.
employing an extension of the catastrophe progression
method through entropy based ranking of the indicators Open Access This article is licensed under a Creative Commons Attri-
bution 4.0 International License, which permits use, sharing, adapta-
where the index structure was improved using correlation tion, distribution and reproduction in any medium or format, as long
and semantics; (2) vulnerability mapping at the microscale as you give appropriate credit to the original author(s) and the source,
taking buildings as the smallest unit that is scarce in the provide a link to the Creative Commons licence, and indicate if changes
literature; (3) improvement of earthquake vulnerability were made. The images or other third party material in this article are
included in the article's Creative Commons licence, unless indicated
maps using spatial clustering technique; and (4) estima- otherwise in a credit line to the material. If material is not included in
tion of people at risk for effective allocation of resources the article's Creative Commons licence and your intended use is not
for emergency planning and disaster risk reduction. permitted by statutory regulation or exceeds the permitted use, you will
Catastrophe progression method avoiding subjectivity need to obtain permission directly from the copyright holder. To view a
copy of this licence, visit [Link]
demonstrated a unique advantage in dealing with uncer-
tainty and provided a straightforward workflow to generate
vulnerability scores. Index scores are synthetic relative
representations of vulnerability calculated using the catas-
References
trophe normalization formulas. Therefore, drawing con-
clusions such as a building may collapse or is extremely AFAD (Afet ve Acil Durum Yönetimi Başkanlığı / Disaster and Emer-
deprived by means of the vulnerability scores on building- gency Management Presidency). 2023. Turkey earthquake hazard
by-building basis is inherently flawed. map (Türkiye Deprem Tehlike Haritası). [Link]
The empirical results show that the western part of the TDTH/​main.​xhtml. Accessed 10 Jul 2023 (in Turkish).
Ahmed, K., S. Shahid, S. Harun, T. Ismail, N. Nawaz, and S. Shamsu-
study area is highly vulnerable. Evaluating the components din. 2015. Assessment of groundwater potential zones in an arid
separately, the results show that social vulnerabilities clus- region based on catastrophe theory. Earth Science Informatics
ter more across space compared to physical vulnerabili- 8(3): 539–549.
ties of the buildings. Buildings might be renewed/recon- Aksha, S.K., L. Juran, L.M. Resler, and Y. Zhang. 2019. An analysis of
social vulnerability to natural hazards in Nepal using a modified
structed individually in time but the social context does social vulnerability index. International Journal of Disaster Risk
not change as much in the study area. Physical and social Science 10(1): 103–116.
vulnerabilities show bivariate correlation, which verifies Allan, P., and M. Bryant. 2011. Resilience as a framework for urban-
coherence of physical and social vulnerabilities, formerly ism and recovery. Journal of Landscape Architecture 6: 34–45.
Ara, S. 2014. Impact of temporal population distribution on earthquake
signified with the idea of “hazard of places.” The EVI loss estimation: A case study on Sylhet, Bangladesh. International
scores classified into vulnerability classes reveal that about Journal of Disaster Risk Science 5(3): 296–312.
20% and 7.8% of all residents are at risk at nighttime and Armaş, I., D. Toma-Danila, R. Ionescu, and A. Gavriş. 2017. Vul-
daytime, respectively. This crude estimate of the people nerability to earthquake hazard: Bucharest case study, Romania.
International Journal of Disaster Risk Science 8(2): 182–195.
at risk provides a relative measure and a basis for disaster Aubrecht, C., S. Freire, C. Neuhold, A. Curtis, and K. Steinnocher.
risk and emergency management initiatives. 2012. Introducing a temporal component in spatial vulnerability
Although the microscale analyses at the building level analysis. Disaster Advances 5(2): 48–53.
provide a great amount of spatial detail, the availability Bağcı, G., A. Yatman, S. Özdemir, and N. Altın. 2000. The earthquakes
causing damage in Turkey (Türkiye’de Hasar Yapan Depremler).
of data at the building and household levels is an issue. Earthquake Research Bulletin (Deprem Araştırma Bülteni) 69:
The present study needs to include some of the indicators 113–126.
deemed important in vulnerability assessment that were Barbat, A.H., M.L. Carreño, L.G. Pujades, N. Lantada, O.D. Cardona,
unavailable at the household level—for example, educa- and M.C. Marulanda. 2009. Seismic vulnerability and risk evalu-
ation methods for urban areas: A review with application to a pilot
tion and employment. Some of the indicators were substi- area. Structure and Infrastructure Engineering 6(1–2): 17–38.
tuted with proxies—for example, the current market value Birkmann, J. 2006. Measuring vulnerability to promote disaster-
of real estate was used in the assessment instead of income resilient societies: Conceptual frameworks and definitions. In
that was unavailable. With all the critical indicators avail- Measuring vulnerability to natural hazards: Towards disaster
resilient societies, ed. J. Birkmann, 9–54. Tokyo: United Nations
able, the method may more likely represent the actual situ- University Press.
ation. However, as the study intended to demarcate the Bohle, H. 2001. Vulnerability and criticality: Perspectives from social
geography. IHDP Update 2(1): 3–5.

13
780 Gerçek and Güven. Urban Earthquake Vulnerability Assessment and Mapping at the Microscale

Brecht, H., U. Deichmann, and H.G. Wang. 2013. A global urban Haque, C.E., and D. Etkin. 2006. People and community as constitu-
risk index. Policy research working papers 6506. Washington, ent parts of hazards: The significance of societal dimensions in
DC: The World Bank. hazards analysis. Natural Hazards 41(2): 271–282.
Cardona, O.D. 2004. The need for rethinking the concepts of vulner- Jaafari, A., A. Najafi, H.R. Pourghasemi, J. Rezaeian, and A. Sattarian.
ability and risk from a holistic perspective: A necessary review 2014. GIS-based frequency ratio and index of entropy models for
and criticism for effective risk management. In Mapping vulner- landslide susceptibility assessment in the Caspian forest, northern
ability: Disasters, development and people, ed. G. Bankoff, G. Iran. International Journal of Environmental Science and Tech-
Frerks, and D. Hilhost, 37–51. London: Earthscan. nology 11(4): 909–926.
Carreño, M., O.D. Cardona, and A.H. Barbat. 2007. Urban seismic Jaimes, D.L., C.R. Escudero, K.L. Flores, and A. Zamora-Camacho.
risk evaluation: A holistic approach. Natural Hazards 40(1): 2022. Multicriteria seismic hazard and social vulnerability assess-
137–172. ment in the Puerto Vallarta metropolitan area, Mexico: Toward
Carreño, M.L., O.D. Cardona, A.H. Barbat, D.C. Suarez, M.P. Perez, a comprehensive seismic risk analysis. Natural Hazards 116(2):
and L. Narvaez. 2017. Holistic disaster risk evaluation for the 2671–2692.
urban risk management plan of Manizales, Colombia. Interna- Jenifer, M.A., and M.K. Jha. 2017. Comparison of analytic hierar-
tional Journal of Disaster Risk Science 8(2): 258–269. chy process, catastrophe and entropy techniques for evaluating
Chen, Y., J. Yu, and S. Khan. 2010. Spatial sensitivity analysis of groundwater prospect of hard-rock aquifer systems. Journal of
multi-criteria weights in GIS-based land suitability evaluation. Hydrology 548: 605–624.
Environmental Modelling & Software 25(12): 1582–1591. Jin, X., and W. Zhang. 2020. The optimization of objective weight-
Cheng, C.-H., Y. Liu, and Y. Lin. 1996. Evaluating a weapon system ing method based on relative importance. In Proceedings of the
using catastrophe series based on fuzzy scales. Proceedings of 2020 5th International Conference on Mechanical, Control and
the 1996 Asian Fuzzy Systems Symposium, 11–14 December Computer Engineering (ICMCCE), 25–27 December 2020, Har-
1996, Kenting, 212–217. bin, China, 1234–1237. [Link]
Cutter, S.L. 1996. Vulnerability to environmental hazards. Progress 2020.​00271.
in Human Geography 20(4): 529–539. Jones, B., and J. Andrey. 2007. Vulnerability index construction: Meth-
Cutter, S.L., and C. Finch. 2008. Temporal and spatial changes odological choices and their influence on identifying vulnerable
in social vulnerability to natural hazards. Proceedings of the neighbourhoods. International Journal of Emergency Manage-
National Academy of Sciences 105(7): 2301–2306. ment 4(2): 269–295.
Cutter, S.L., B.J. Boruff, and W.L. Shirley. 2003. Social vulnerabil- Lindlay, S., J. O’Neill, N. Lawson, R. Christian, and M. O’Neil. 2011.
ity to environmental hazards. Social Science Quarterly 84(2): Climate change, justice and vulnerability. York: Joseph Rowntree
242–261. Foundation.
Davidson, R. 1997. EERI annual student paper award a multidiscipli- Longley, P.A., M.F. Goodchild, D.J. Maguire, and D.W. Rhind. 2015.
nary urban earthquake disaster risk index. Earthquake Spectra Geographic information science and systems. New York: Wiley.
13(2): 211–223. McEntire, D. 2012. Understanding and reducing vulnerability: From
DBYBHY (Deprem Bölgelerinde Yapılacak Binalar Hakkında Yönet- the approach of liabilities and capabilities. Disaster Prevention
melik / Regulation on buildings to be constructed in earthquake and Management: An International Journal 21(2): 206–225.
zones). 2007. [Link] Mostafa, M.M. 2022. Five decades of catastrophe theory research: Geo-
20070​306-3.​htm. Accessed 10 Jul 2023. graphical atlas, knowledge structure and historical roots. Chaos,
Diaz-Sarachaga, J.M., and D. Jato-Espino. 2019. Analysis of vulner- Solitons & Fractals 159: Article 112078.
ability assessment frameworks and methodologies in urban areas. Özalaybey, S., E. Zor, M.C. Tapırdamaz, A.R. Tarancıoğlu, B. Erkan,
Natural Hazards 100(1): 437–457. A. Karaaslan, E.M. Alpaslan, E.S. Ergintav, and E. Tan. 2008. Soil
Du, X., X. Liu, and Y. Zihang. 2022. Risk assessment of agricultural classification and seismic hazard assessment project for Kocaeli
drought based on catastrophe progression method: A China case Province (Kocaeli İli için Zemin Sınıflaması ve Sismik Tehlike
study. SSRN. [Link] Değerlendirme Projesi). TÜBİTAK.
id=​40340​84. Pavić, G., M. Hadzima-Nyarko, and B. Bulajić. 2020. A contribution
Ebert, A., N. Kerle, and A. Stein. 2009. Urban social vulnerability to a UHS-based seismic risk assessment in Croatia – A case study
assessment with physical proxies and spatial metrics derived from for the city of Osijek. Sustainability 12(5): Article 1796.
air and spaceborne imagery and GIS data. Natural Hazards 48(2): Peduzzi, P., H. Dao, C. Herold, and F. Mouton. 2009. Assessing global
275–294. exposure and vulnerability towards natural hazards: The disas-
Fekete, A. 2019. Social vulnerability (re-)assessment in context to ter risk index. Natural Hazards and Earth System Sciences 9(4):
natural hazards: Review of the usefulness of the spatial indicator 1149–1159.
approach and investigations of validation demands. International Pelling, M. 2003. The vulnerability of cities: Natural disasters and
Journal of Disaster Risk Science 10(2): 220–232. social resilience. London: Earthscan.
FEMA (Federal Emergency Management Agency). 2015. Rapid visual Rashed, T., and J. Weeks. 2003. Exploring the spatial association
screening of buildings for potential seismic hazards: A handbook, between measures from satellite imagery and patterns of urban
3rd edn. [Link] vulnerability to earthquake hazards. International Archives of the
earth​quakes_​rapid-​visual-​scree​ning-​of-​build​ings-​for-​poten​tial-​ Photogrammetry, Remote Sensing and Spatial Information Sci-
seism ​ i c- ​ h azar ​ d s-a- ​ h andb ​ o ok- ​ t hird- ​ e diti ​ o n- ​ fema-p- ​ 1 54. ​ p df. ences XXXIV 7(W9): 144–152.
Accessed 10 Jul 2023. Rodcha, R., N.K. Tripathi, and R.P. Shrestha. 2019. Comparison of
Fischer, E., A.E. Biondo, A. Greco, F. Martinico, A. Pluchino, and A. cash crop suitability assessment using parametric, AHP, and
Rapisarda. 2022. Objective and perceived risk in seismic vulner- FAHP methods. Land 8(5): Article 79.
ability assessment at an urban scale. Sustainability 14(15): Article RYTEIE (Riskli Yapıların Tespit Edilmesine İlişkin Esaslar / Principles
9380. for identification of risky buildings). 2019. [Link]
Gao, S., H. Sun, M. Ma, Y. Lu, Y. Yao, and W. Liu. 2020. Vulnerability gov.​tr/​db/​altya​pi/​iceri​kler/r-​skl--​yapil​arin-​tesp-t-​ed-​lmes-​ne-​il-​
assessment of marine economic system based on comprehensive sk-n-​esasl​ar-​20190​21813​4628.​pdf. Accessed 6 Oct 2023.
index and catastrophe progression model. Ecosystem Health and Sauti, N.S., M.E. Daud, M. Kaamin, and S. Sahat. 2021. GIS spa-
Sustainability 6(1): Article 1834459. tial modelling for seismic risk assessment based on exposure,

13
International Journal of Disaster Risk Science 781

resilience, and capacity indicators to seismic hazard: A case study Woodcock, A. 2006. Stereographic reconstuctions of the catastrophe
of Pahang, Malaysia. Geomatics, Natural Hazards and Risk 12(1): manifolds of the cuspoids: (4) the wigwam. In A geometrical study
1948–1972. of the elementary catastrophes, ed. A.E.R. Woodcock, and T. Pos-
Şengör, A.M.C., O. Tüysüz, C. İmren, M. Sakınç, H. Eyidoğan, N. ton, 193–211. Berlin: Springer.
Görür, X. Le Pichon, and C. Rangin. 2005. The North Anatolian Wu, J., X. Chen, and J. Lu. 2022. Assessment of long and short-term
Fault: A new look. Annual Reviews 33: 37–112. flood risk using the multi-criteria analysis model with the AHP-
Shakya, M., H. Varum, R. Vicente, and A. Costa. 2018. Seismic vul- entropy method in Poyang Lake basin. International Journal of
nerability assessment methodology for slender masonry struc- Disaster Risk Reduction 75: Article 102968.
tures. International Journal of Architectural Heritage 12(7–8): Xue, J., J. Yan, and C. Chen. 2022. Combining catastrophe technique
1297–1326. and regression analysis to deduce leading landscape patterns for
Shen, D., T. Qian, Y. Xia, Y. Zhang, and J. Wang. 2020. Micro-scale regional flood vulnerability: A case study of Nanjing, China.
flood hazard assessment based on catastrophe theory and an inte- Frontiers in Ecology and Evolution 10: Article 1002231.
grated 2-D hydraulic model: A case study of Gongshuangcha Yariyan, P., H. Zabihi, I.D. Wolf, M. Karami, and S. Amiriyan. 2020.
detention basin in Dongting Lake area, China. ISPRS Interna- Earthquake risk assessment using an integrated fuzzy analytic
tional Journal of Geo-Information 9(4): Article 206. hierarchy process with artificial neural networks based on GIS: A
Silva, V., S. Brzev, C. Scawthorn, C. Yepes, J. Dabbeek, and H. Crow- case study of Sanandaj in Iran. International Journal of Disaster
ley. 2022. A building classification system for multi-hazard risk Risk Reduction 50: Article 101705.
assessment. International Journal of Disaster Risk Science 13(2): Zhai, Y., S. Chen., and Q. Ouyang. 2019. GIS-based seismic hazard
161–177. prediction system for urban earthquake disaster prevention plan-
Song, F., X. Yang, and F. Wu. 2020. Catastrophe progression method ning. Sustainability 11(9): Article 2620.
based on M-K test and correlation analysis for assessing water Zhang, P., X. Li, and C. Liu. 2023. Impact of spatial scale and build-
resources carrying capacity in Hubei Province. Journal of Water ing exposure distribution on earthquake insurance rates: A case
and Climate Change 11(2): 556–567. study in Tangshan, China. International Journal of Disaster Risk
Su, S., D. Li, X. Yu, Z. Zhang, Q. Zhang, R. Xiao, J. Zhi, and J. Wu. Science 14(1): 64–78.
2011. Assessing land ecological security in Shanghai (China) Zhang, H., Y. Sun, W. Zhang, Z. Song, Z. Ding, and X. Zhang. 2021.
based on catastrophe theory. Stochastic Environmental Research Comprehensive evaluation of the eco-environmental vulnerability
and Risk Assessment 25: 737–746. in the Yellow River Delta wetland. Ecological Indicators 125:
Thom, R. 1975. Structural stability and morphogenesis. Reading, MA: Article 107514.
WA Benjamin. Zhang, W., X. Xu, and X. Chen. 2017. Social vulnerability assess-
TUIK (Türkiye İstatistik Kurumu / Turkish Statistical Institute). 2021. ment of earthquake disaster based on the catastrophe progression
Labor force statistics, 2021. [Link] ata.​t uik.​g ov.​t r/​B ulten/​ method: A Sichuan Province case study. International Journal of
Index?p=I​ sgucu-I​ stati​ stikl​ eri-2​ 021-4​ 5645. Accessed 10 Jul 2023. Disaster Risk Reduction 24: 361–372.
Wamsler, C., E. Brink, and C. Rivera. 2013. Planning for climate Zheng, J., and G. Huang. 2023. Towards flood risk reduction: Com-
change in urban areas: From theory to practice. Journal of monalities and differences between urban flood resilience and risk
Cleaner Production 50: 68–81. based on a case study in the Pearl River Delta. International Jour-
Wei, B., G. Nie, G. Su, L. Sun, X. Bai, and W. Qi. 2017. Risk assess- nal of Disaster Risk Reduction 86: Article 103568.
ment of people trapped in earthquake based on km grid: A case Zhu, Y., D. Tian, and F. Yan. 2020. Effectiveness of entropy weight
study of the 2014 Ludian Earthquake, China. Geomatics, Natural method in decision-making. Mathematical Problems in Engineer-
Hazards and Risk 8(2): 1289–1305. ing 2020: Article 3564835.
Wisner, B., P.M. Blaike, T. Cannon, and I. Davis. 2004. At risk: Natural Ziarh, G.F., M.D. Asaduzzaman, A. Dewan, M.S. Nashwan, and S.
hazards, people’s vulnerability and disasters. London: Routledge. Shahid. 2021. Integration of catastrophe and entropy theories for
Wood, N.J., C.G. Burton, and S.L. Cutter. 2010. Community varia- flood risk mapping in peninsular Malaysia. Journal of Flood Risk
tions in social vulnerability to Cascadia-related tsunamis in the Management 14(1): Article e12686.
U.S. Pacific Northwest. Natural Hazards 52(2): Article 369389.

13