Vous êtes sur la page 1sur 15

This article was downloaded by: [Chinese University of Hong Kong]

On: 09 February 2015, At: 05:19


Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,
37-41 Mortimer Street, London W1T 3JH, UK

Annals of GIS
Publication details, including instructions for authors and subscription information:
http://www.tandfonline.com/loi/tagi20

A GIS-based multi-criteria decision-making approach


for seismic vulnerability assessment using quantifier-
guided OWA operator: a case study of Tehran, Iran
a b c
Milad Moradi , Mahmoud Reza Delavar & Behzad Moshiri
a
Department of Surevying and Geomatic Engineering, College of Engineering, University of
Tehran, Tehran, Iran
b
Center of Excellence in Geomatic Engineering in Disaster Management, Department of
Surveying and Geomatic Engineering, College of Engineering, University of Tehran, Tehran,
Iran
c
Control and Intelligent Processing Center of Excellence, School of Electrical and Computer
Click for updates Engineering, College of Engineering, University of Tehran, Tehran, Iran
Published online: 21 Oct 2014.

To cite this article: Milad Moradi, Mahmoud Reza Delavar & Behzad Moshiri (2014): A GIS-based multi-criteria decision-
making approach for seismic vulnerability assessment using quantifier-guided OWA operator: a case study of Tehran, Iran,
Annals of GIS, DOI: 10.1080/19475683.2014.966858

To link to this article: http://dx.doi.org/10.1080/19475683.2014.966858

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained
in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no
representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the
Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and
are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and
should be independently verified with primary sources of information. Taylor and Francis shall not be liable for
any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever
or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of
the Content.

This article may be used for research, teaching, and private study purposes. Any substantial or systematic
reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any
form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://
www.tandfonline.com/page/terms-and-conditions
Annals of GIS, 2014
http://dx.doi.org/10.1080/19475683.2014.966858

A GIS-based multi-criteria decision-making approach for seismic vulnerability assessment using


quantifier-guided OWA operator: a case study of Tehran, Iran
Milad Moradia, Mahmoud Reza Delavarb* and Behzad Moshiric
a
Department of Surevying and Geomatic Engineering, College of Engineering, University of Tehran, Tehran, Iran; bCenter of Excellence
in Geomatic Engineering in Disaster Management, Department of Surveying and Geomatic Engineering, College of Engineering,
University of Tehran, Tehran, Iran; cControl and Intelligent Processing Center of Excellence, School of Electrical and Computer
Engineering, College of Engineering, University of Tehran, Tehran, Iran
(Received 14 May 2014; accepted 31 August 2014)
Downloaded by [Chinese University of Hong Kong] at 05:19 09 February 2015

Earthquakes are one of the most destructive disasters that cause huge damage to structures and human beings. It is not
possible to predict the exact location, duration and intensity of an earthquake; however, the probable damages may be
predicted in advance. Tehran has suffered from a number of non-standard aged buildings, a high population density and the
existence of a number of faults which have made it significantly vulnerable to the risk of earthquakes. In this research,
Tehran’s earthquake physical vulnerability map has been produced. The main contribution of this research is to take the
effect of the optimism degree into account in the problem of seismic vulnerability assessment. The ordered weighted
averaging operator is applied to provide a wide range of answers from pessimistic to optimistic solutions based on multi-
criteria analysis. The results are then compared with the granular computing and intuitionistic fuzzy (Int_Fuzzy) models that
indicate 91% and 89% similarity, respectively.
Keywords: multi-criteria decision making; GIS; OWA operator; earthquake; vulnerability assessment; Tehran

Introduction probable damages inflicted by earthquakes can be esti-


Urban areas have always been threatened by natural dis- mated (Rashed and Weeks 2003; Duzgun et al. 2011). In
asters such as hurricanes, tsunamis and earthquakes. addition, an efficient post-disaster response can be planned
Earthquake is the most damaging natural disaster in Iran based on the estimated vulnerability maps of mega cities.
(Aghamohhamdi et al. 2013). The Iranian Plateau, which Tehran which is home to approximately one quarter of the
is part of the Eurasian plate, is adjacent to the Alborz total population of the country has three major known and
Mountains on the north and is bounded by the Arabian many unknown faults within its boundaries, which make it
platform to the south (Berberian and Yeats 2001). The extremely vulnerable to earthquakes (Samadi Alinia and
stress caused by the convergence between the Eurasian Delavar 2011). Earthquakes of 6.0 magnitude or larger are
and Arabian plates increases the probability of fault acti- expected in Tehran about every 150 years. The last large
vation in this area (Berberian and Yeats 2001). Located in one shook the city approximately 170 years ago, and,
a seismically active region in Asia, Iran has a number of therefore, experts argue that a large earthquake should be
earthquakes every year, some of which are large enough to expected in the near future (Silavi et al. 2006).
destroy a large number of houses and urban facilities and Being an ill-structured problem, urban seismic vulner-
cause huge damages to infrastructure. Ineffective human ability assessment does not have a unique solution
response to earthquakes seems to be one of the main (Rashed and Weeks 2003). Intensive researches have
factors aggravating these damages. Constructing buildings been undertaken recently to propose a reliable solution
with sub-standard materials, construction of towns and for this problem, all of which fall into two main categories
cities near the major faults, overcrowded urban areas and that include using earthquake modelling software to esti-
the lack of an effective disaster relief and management mate loss and using the multi-criteria decision-making
plan are the major factors that contribute to further losses (MCDM) methods (Duzgun et al. 2011). Two examples
(Chini, Pierdicca, and Emery 2009; Duzgun et al. 2011). It of software applications used for modelling earthquakes
is impossible to avoid earthquakes or even predict their are HAZUS and SELENA. HAZUS, developed by the
exact location, time, duration and intensity. However, the Federal Emergency Management Agency (FEMA), is the

*Corresponding author. Email: mdelavar@ut.ac.ir

© 2014 Taylor & Francis


2 M. Moradi et al.

most famous seismic loss estimation software. It enables determine the street seismic vulnerability that is widely
experts to estimate the extent of human, physical and applicable in post-earthquake route planning and mobility
economic loss through many predefined earthquake sce- management. Previous researches seem to have magnifi-
narios (Tantala et al. 2008; Schmidtlein et al. 2011; Remo cent improvements in earthquake vulnerability assess-
and Pinter 2012). SELENA (SEismic Loss EstimatioN ment. However, they did not take the optimism degree
from a logic tree Approach) is an open-source software into account, which is one of the most important factors in
developed by the International Center for Geohazards, multi-criteria decision-making. Considering different
which facilitates the seismic vulnerability assessment for levels of optimism, different results may be obtained. In
metropolitan areas (Molina, Lang, and Lindholm 2010; this research, Tehran’s seismic vulnerability map is pro-
Yang, Zhan, and Li 2011). Although applying loss estima- duced based on seven different optimism degrees asso-
tion software has a number of benefits including using ciated with certain linguistic quantifiers.
standard data formats and having lower levels of compu-
tational errors, the MCDM methods have the advantage of
empowering experts to assess the seismic vulnerability Materials and methods
In order to find the best choices, most of the decision-
Downloaded by [Chinese University of Hong Kong] at 05:19 09 February 2015

through any earthquake scenario. In addition, by using


MCDM algorithms, experts can tailor the conditions making problems in the GIS, such as site selection and
used for the modelling to the city’s specifications (Kolat allocation, require the decision-maker to consider the
et al. 2006; Mohanty et al. 2007; Kolat, Ulusay, and Suzen scores of alternatives (attribute values) in multiple dimen-
2012). The proposed model in this research uses the sions, which is a spatial multi-criteria decision-making
MCDM methods in order to have a better control over problem (Jankowski 1995; Greene et al. 2011). MCDM
parameters and results. seems to be the most important part of the decision sup-
A number of researches have been undertaken to pro- port systems in GIS (Jiang and Eastman 2000). Spatial
pose a seismic vulnerability model for the Tehran MCDM is a process that extracts useful information
Metropolitan Area (TMA). Aghataher, Delavar, and (needed in decision-making) from the attribute values of
Kamalian (2005) introduced some important spatial fac- geospatial data. Although GIS and MCDM are two sepa-
tors affecting the seismic vulnerability and proposed a rate areas of research, they can complement each other.
fuzzy-AHP1 model to specify layer weights through a GIS facilitates the visualization and management of geos-
pairwise comparison. They assessed the physical vulner- patial data, while MCDM provides a set of algorithms in
ability, which is the vulnerability of urban structures and order to help users make an informed decision (Eldrandaly
facilities to earthquakes. In addition, they determined the 2013). By combining these two areas of research, a power-
most vulnerable areas of Tehran. Due to the prevalence of ful tool for GIS-based MCDM is generated (Boroushaki
various kinds of uncertainty in input data, Silavi et al. and Malczewski 2010). In a number of problems such as
(2006) proposed a vulnerability model based on intuitio- the vulnerability assessment, there may not be real data to
nistic fuzzy logic, which takes the indeterminacy of mem- extract knowledge from. Therefore, experts’ knowledge is
bership functions into account, to overcome the the only source of information for these problems.
shortcomings of the fuzzy-AHP model. Moreover, they Boolean overlay and weighted linear combination
evaluated the earthquake human vulnerability, which is (WLC) have frequently been used in spatial multi-criteria
the mortality rate caused by an earthquake (Silavi et al. decision-making. In fact, they represent two main cate-
2006). Although the recent earthquake in Tehran caused gories of traditional solutions for a spatial MCDM pro-
no damage significant enough to calculate the accuracy of blem. In the Boolean overlay method, a decision-maker
the proposed model, the sensitivity of the model to input (expert) defines a threshold which is used to transform all
errors could be estimated using a variety of methods. of the values into a zero or one set. In other words, the
Jahanpeyma et al. (2007) introduced an error propagation Boolean overlay method uses binary maps in its decision-
model based on the Monte Carlo simulation which illus- making process, while WLC uses the standardized value
trates the effect of uncertain input data on the final seismic of the original data as well as weight vector in order to
vulnerability of urban areas. In order to resolve the contra- analyse the problem and rank the alternatives (Greene
dictions inside the classification rules, expressed by an et al. 2011). In addition, in the WLC method, experts are
expert, Samadi Alinia and Delavar (2011) minimized the supposed to determine the weights of the layers (criteria),
entropy of rules using the granular computing (GrC) which express the relative importance of the map layers in
approach. They asked the expert to classify a set of sample comparison to each other. However, in the Boolean over-
statistical units. Then, they proposed a model to extract the lay method, binary map layers could be aggregated using
classification rules with minimum entropy. They used “AND” or “OR” operators, while a weighted average
these rules to classify all the statistical units in Tehran could be used to aggregate the layers and produce an
urban area. Hashemi and Alesheikh (2011) used a overall map in WLC method (Al-Adamat, Diabat, and
Geospatial Information System (GIS)-based model to Shatnawi 2010). Both of these two traditional methods
Annals of GIS 3

•Determining important criteria


1

•Specifying cost and benefit layers


2

•Normalizing raw data


3

•Selecting the decision sterategy (Linguistic Quantifier)


4

•Obtaining order weights


5

•Aggregating different criteria to obtain the vulnerability degree


6
Downloaded by [Chinese University of Hong Kong] at 05:19 09 February 2015

•Classification of vulnerability degrees and producing vulnerability map


7

Figure 1. The proposed steps of vulnerability assessment based on the OWA operator.

are covered in ordered weighted averaging (OWA) operator. where I is a number in the interval of [0,1] and In is input
Therefore, OWA method is widely used in GIS-based vector consisting of n numbers between 0 and 1.
MCDM (Malczewski 2006, 1996). In OWA operator
when the weight vector (W) is like [0,0,…,0,1], the operator
changes into an “AND” operator and when W is in the form
of [1,0,0,…,0], an “OR” operator is created and for all other OWA operator
kinds of W, the WLC method is applied. In this research, In a decision-making process, one important issue is to
OWA operator is applied to produce a seismic vulnerability determine how many criteria are going to be satisfied.
map through an MCDM process. The procedure of the Traditionally, there are two extreme operators, AND and
proposed model is illustrated in seven steps in Figure 1. OR. Yager (1988) indicated that the satisfaction of one
First, important criteria are specified based on experts’ (OR) or all (AND) of the criteria as two very extreme
opinion. Then, experts determine the criteria that have a cases of the decision-making strategies is not sufficient for
direct relationship with seismic vulnerability (benefit cri- a proper judgement. In some problems, satisfaction of half
teria) and those that have an inverse relationship (cost of the criteria may be sufficient. Based on this idea, Yager
criteria). In this research, the attribute values are then proposed a new family of aggregation operators that can
extracted from the 1996 census data since more recent implicitly adjust the level of ‘ANDness’ and ‘ORness’ in
census data were not available. Next, order weights are combining the alternatives scores. He entitles this new
calculated based on the selected linguistic quantifier. By operator as OWA operator. The OWA operator is an
aggregating the ordered attribute values using the calculated aggregation operator defined in Equation (2) (Yager 1988):
weight vector, a relative seismic vulnerability degree is
assigned to each statistical unit. Finally, the vulnerability X
n
F ða1 ; a2 ; . . . ; an Þ ¼ wj bj
degrees are classified and visualized in ArcGIS software. j¼1
In many decision-making problems, the score of
alternatives should be aggregated among different cri- ¼ w1 b1 þ w2 b2 þ . . . þ wn bn ; wi 2 ½0; 1
teria. Then, alternatives should be ranked based on their Xn
and wi ¼ 1
aggregated score. Finding an appropriate aggregation
i¼1
operator has been a challenging issue, and a number of
(2)
aggregation operators have been proposed based on dif-
ferent aspects of the decision-making problem. An aggre-
gation operator is a mapping from n-dimensional space where n is the number of criteria, ai is the score of
to a one-dimensional space, defined in Equation (1) as the alternative in ith criterion, bj is the jth largest element
(Yager 1988): in the input vector a, wj is weight of the jth input and F
is the aggregated score of the vector a. In a decision-
making process with m alternatives, the alternative with
F : I n ! I; I 2 ½0; 1 (1) the largest value of F is the most preferred choice. OWA
4 M. Moradi et al.

operator satisfies monotonicity, stability, associativity, operator is capable of satisfying a predefined number or
idempotency, identity and neutrality (Fodor, Marichal, percentage of the criteria. This predefined number is indi-
and Roubens 1995). In this model, F is the calculated cated by a linguistic quantifier, which is a natural language
vulnerability degree, and statistical units are then classified express used by the decision-maker to specify the strategy
based on the aggregated value of F. of decision-making. Generally, there are two kinds of
Weight vector is the original parameter of the OWA quantifiers including absolute and relative quantifiers.
model and also implements the decision strategy chosen Statements such as ‘At least 6’, ‘Not more than 3’,
by the decision-maker (Yager, Kacprzyk, and Beliakov ‘Almost 4’ and ‘About 8’ are all examples of absolute
2011). OWA includes a variety of operators including quantifiers, while relative quantifiers do not specify the
maximum, minimum, weighted average and geometric number of criteria directly and instead use expressions
average. The optimism degree determines the act of the such as ‘At least one’, ‘Few’, ‘Many’, ‘Some’ and ‘All’
operator. An optimistic decision-maker tends to choose an to determine the decision strategy (Yager 1996). For
alternative that has the minimum distance to the idealistic example ‘Some’ means ‘some of the criteria must be
point. However, a pessimistic decision-maker tends to satisfied by an acceptable alternative’, ‘All’ turns the
OWA into an AND operator and ‘At least one’ indicates
Downloaded by [Chinese University of Hong Kong] at 05:19 09 February 2015

choose the one with maximum distance from the unde-


sired state (Zarghami and Szidarovszky 2008). In an OR operator. A linguistic quantifier, Q(r), is a number
Equation (2), the weight vector is multiplied by the vector between 0 and 1 which represents how compatible is
of descending-order alternative scores. Therefore, greater satisfying of r% of criteria with the expert’s point of
values at the beginning of the weight vector cause the view. For example, if a decision-maker chooses ‘Few’ as
greater optimism degree (risk acceptance). Yager (1988) a quantifier, Q(0.2) = 95% means an alternative that satis-
defined the optimism degree as shown in Equation (3): fies 0.2 of the criteria is 95% preferable with the decision
strategy of ‘Few’. There are a number of ways to define a
1 X n
linguistic quantifier. Yager (1996) proposed a family of
OðW Þ ¼ ðn  jÞwj (3)
n  1 j¼1 quantifiers known as regular increasing monotone (RIM),
using the exponential function. RIM quantifiers are
defined in Equation (4):
where W is the weight vector and n is the number of elements
in W (number of criteria). The optimism degree (ORness) is
always between 0 and 1. O(W) = 1 results in OR (union) QðrÞ ¼ rα ; α > 0 (4)
operator while O(W) = 0 leads to AND (intersection) opera-
tor, and O(W) = 0.5 turns the OWA into simple weighted
where Q is a quantifier of the unit interval [0,1], r
average (Yager 1988). The optimism degree entirely
represents the degree of compatibility and α is the opti-
depends on values in weight vector. Different OWA opera-
mism degree, the only parameter of the quantifier. By
tors have different methods of extracting the order weights,
changing α, all kinds of quantifiers are built (see
and, therefore, have different optimism degrees.
Figure 2). When α tends to zero, OWA acts exactly like
a MAX operator while at the other extreme, OWA changes

Linguistic quantifiers
In order to define an OWA operator, a unique weight 1.2
vector should be defined. Although there are a variety of
methods to obtain order weights, they can be summarized 1
into three major categories including linguistic quantifiers,
maximum entropy and learning methods (Xu 2005). Yager 0.8 α = 100
(1988) proposed linguistic quantifiers to specify the num- α=5
Q(r)

ber of criteria that are going to be satisfied. Fullér and 0.6 α=2
α=1
Majlender (2003) introduced the minimal variation
0.4 α = 0.5
method instead of maximum entropy that tends to mini-
α = 0.2
mize the standard deviation of weights. Taking the advan-
0.2 α = 0.01
tage of being similar to natural language, we use linguistic
quantifiers to calculate the weight vector. In this proce-
dure, the decision-maker could efficiently involve his/her 0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
preferences into the model by using expressions such as
r
‘All’, ‘Half’ and ‘At least one’.
Although the Max and Min operators could only Figure 2. Linguistic quantifiers and their corresponding opti-
satisfy ‘All’ and ‘At least one’ of the criteria, the OWA mism degree.
Annals of GIS 5

Table 1. Weight vector obtained based on each linguistic quantifier.

Decision strategy Linguistic quantifier α Weight vector (n = 6) Operator

Optimistic At least one 0.01 [1 0 0 0 0 0] Max (OR)


Few 0.2 [0.69 0.10 0.06 0.05 0.04 0.03] %
Some 0.5 [0.40 0.16 0.12 0.10 0.09 0.08] %
Neutral Half 1 [0.16 0.16 0.16 0.16 0.16 0.16] WLC
More 2 [0.02 0.08 0.13 0.19 0.24 0.30] %
Pessimistic Most 5 [0.00 0.00 0.02 0.09 0.26 0.58] %
All 100 [0 0 0 0 0 1] Min (AND)

into a MIN operator and for α = 1, the linguistic quantifier weight vectors with the same ORness measure. In order
‘half’ is built. to overcome this problem, they proposed another mea-
Downloaded by [Chinese University of Hong Kong] at 05:19 09 February 2015

Table 1 represents a number of frequently used fuzzy sure, called the dispersion (or entropy) of weight vec-
quantifiers and their corresponding α. The major function- tor. Dispersion determines the tendency of OWA
ality of linguistic quantifiers is to specify the optimism weights to take all the information (representing by
degree (ORness) in a model by determining the order the vector of attribute values) into account. The higher
weights. Yager (1988) proposed Equation (5) to obtain the entropy is, the more the aggregated value depends
the OWA weight vector using the linguistic quantifiers: on all inputs. Dispersion is introduced in Equation (8)
    (Yager 1988):
j j1
wj ¼ QRIM  QRIM ; j ¼ 1; 2; . . . ; n (5)
n n X
n  
H ðW Þ ¼  wj ln wj (8)
j¼1
where j is the criterion number, n is the total number
of criteria and wj is the weight of jth criterion. By employ-
ing Equations (4) and (5), order weights are calculated The dispersion measure is always in the unit interval
in Equation (6) as (Yager 1988): [0,1]. The zero value of dispersion represents a uniform
 α   weight vector while one means its intensive variation. In
j j1 α order to use all the information content in the input
wj ¼  ; j ¼ 1; 2; . . . ; n (6)
n n vector, dispersion of the weight vector should be max-
imized. To solve this problem, O’Hagan (1988) proposed
where α is the ORness or optimism degree. Table 1 also an optimization method as shown in Equations (9)
represents the weight vector for a decision-making pro- and (10):
blem with six criteria. Weights are derived using Equation
(6). Yager (1999) proposed Equation (7) for calculating X
n  
the optimism degree: max : H ðW Þ ¼  wj ln wj (9)
j¼1

ð1 ð1
1
OðW Þ ¼ QRIM ðrÞdr ¼ rα dr ¼ (7) subject to:
1þα
0 0

1 X n Xn

where O(W) is the optimism degree of the vector W OðW Þ ¼ ðn  jÞwj ; wj 2 ½0; 1 and wj ¼ 1
n  1 j¼1 j¼1
derived using the RIM quantifier and Equation (6) when
n (the number of criteria) tends to infinity. In this article, (10)
we calculated the weights for seven different linguistic
quantifiers to investigate the influence of the optimism O’Hagan (1988) nominated an operator whose
degree on earthquake vulnerability assessment. weights are associated with a certain optimism degree,
the maximal entropy-OWA (ME-OWA) operator. In order
to get better results and use all information content in the
Measure of dispersion input data (vector of attribute values), the dispersion of
Yager and Filev (1999) demonstrated that an ORness weight vector is maximized in this article. Table 2 indi-
measure is inadequate to represent a unique weight cates the dispersion of weight vectors derived from
vector. They proved that there may be several distinct Equation (6).
6 M. Moradi et al.

Table 2. Weight vectors and their corresponding dispersion and ORness.

Linguistic quantifiers Weight vector Dispersion ORness

At least one [1 0 0 0 0 0] 0 1
Few [0.69 0.10 0.06 0.05 0.04 0.03] 1.03 0.83
Some [0.40 0.16 0.12 0.10 0.09 0.08] 1.32 0.68
Half [0.16 0.16 0.16 0.16 0.16 0.16] 1.64 0.5
More [0.02 0.08 0.13 0.19 0.24 0.30] 1.57 0.33
Most [0.00 0.00 0.02 0.09 0.26 0.58] 0.93 0.17
All [0 0 0 0 0 1] 0 0

Study area The most outstanding principal structure in the neigh-


In recent years, a sharp increase in the population and a bourhood of Tehran is the NTF, which begins from
rapid growth of Iran’s economy has led to a significant Lashgarak (in the east) and ends in Kan (in the west)
Downloaded by [Chinese University of Hong Kong] at 05:19 09 February 2015

growth of Tehran. As a result, in the last two decades, (Figure 3). It can be considered as a small section of a
Tehran has attracted special attention as the hub for Iran’s collection of blocks that is located in the NE–SW thrust of
educational, industrial and financial activities (Tayyebi, Alborz. Having a length of approximately over 35 km, the
Pijanowski, and Tayyebi 2011). More than 10 million peo- fault at its eastern end is connected to the Mosha Fault
ple live in Tehran, most of whom reside in the southern part (Tchalenko 1975).
of the city, where most of the buildings are aged and non-
standard. Moreover, Tehran has not had a large earthquake
since 1830. As a result, a future earthquake may be very Contributing criteria
destructive (Hashemi and Alesheikh 2011). Seismologists A large number of contributing criteria are introduced in
investigating earthquakes in Tehran believe that a probable seismic loss estimation models. Table 3 illustrates 14
earthquake would be in the magnitude range of 6.2–7.3 factors (four main groups of criteria) applied in previous
(Modified Mercalli Intensity) (Hamzehloo, Vaccari, and seismic loss estimation models. In the current research,
Panza 2007). Tehran is surrounded by three major faults due to limited accessibility to data, only six criteria are
(Figure 3) including Mosha Fault, North Tehran Fault included into the model.
(NTF) and North Rey Fault (Ritz et al. 2003; Solaymani The age of a building is one of the most important
Azad, Ritz, and Abbassi 2011). criteria that make it vulnerable against earthquake (Silavi

Figure 3. Major faults around TMA.


Annals of GIS 7

Table 3. Contributing factors in seismic vulnerability assessment.

ID Type of factor Factor Reference

1 Structure Height of the buildings Samadi Alinia and Delavar (2011)


2 Number of floors Aghataher, Delavar, and Kamalian (2005)
3 Construction materials Molina, Lang, and Lindholm (2010)
4 Structure quality Panahi, Rezaie, and Meshkani (2014)
5 Seismic Resonance Coefficient of buildings Anbazhagan et al. (2010)
6 Adjacent Distance to flammable facilities Panahi, Rezaie, and Meshkani (2014)
7 Distance to active faults Samadi Alinia and Delavar (2011)
8 Percentage of old population Walker et al. (2014)
9 Population Percentage of dependent population Rashed and Weeks (2003)
10 Percentage of uneducated population Aghamohammadi et al. (2013)
11 Other Soil liquefaction Hashemi and Alesheikh (2011)
12 Peak ground acceleration Hashemi and Alesheikh (2011)
Downloaded by [Chinese University of Hong Kong] at 05:19 09 February 2015

13 Slope Walker et al. (2014)


14 Soil type Molina, Lang, and Lindholm (2010)

et al. 2006). Considering that the first building code in Iran their datum. The study area is located in the UTM
was implemented in 1966, buildings constructed before Zone 39 N. Attribute values of 10 sample statistical
this period have been considered as non-standard build- units which are randomly selected are showed in Table 5.
ings in this research. The first standard code for strength-
ening buildings against earthquakes came into effect in
1988. Therefore, the proportion of the buildings built Normalization of attribute values
between 1966 and 1988 has been added to the model as
A number of MCDM algorithms like OWA need the entire
an important factor.
attribute values to be in a comparable scale. Without this
In addition, the quality of materials is noticeably
condition, aggregation of different criteria is impossible
important. The height of a building is the third effective
because they may be in different units and have different
factor (Jahanpeyma et al. 2007). Faults are the main
ranges. Normalization is a procedure that transforms all
source of earthquakes, and destruction is related to the
the scores in different criteria into the interval of [0–1].
distance from active faults (Hashemi and Alesheikh
One of the most common techniques of normalization is
2011). This article assumes that the NTF has been acti-
maximum score. Based on this method, all of the scores in
vated and has caused an earthquake. Slope can also be
a certain criterion should be divided by the maximum of
mentioned as another effective criterion (Aghataher,
that criterion. Therefore, normalized values are calculated
Delavar, and Kamalian 2005). In this article, due to
using Equation (11) (Malczewski 1999):
limited access to updated data, the 1996 census data are
used. Table 4 illustrates the data used in this article. All ai
the data, obtained from Iranian National Statistical a0i ¼ (11)
amax
Center, is in the format of Environmental Systems
Research Institute shape file and have WGS-1984 as where ai is the ith great value between all the scores in the
criterion, amax is the maximum value of the criterion and ai′
is the normalized value (Malczewski 1999). Although max-
Table 4. Important criteria used in this model. imum score is extremely useful for the benefit criteria,
Criterion Definition Type which are desired to be larger, in the case of the cost criteria,
which should be minimized, this method is not sufficient
Slope Average slope of each statistical unit Benefit (Malczewski 1999). In order to normalize the cost criteria,
Bld_Less4 Percentage of buildings having less Cost Equation (12) have been used (Malczewski 1999):
than four floors
Bld_More4 Percentage of buildings having four Benefit ai
floors and more a0i ¼ 1  (12)
Bld_Bef66 Percentage of buildings built before Benefit amax
1966
Bld_Bet66_88 Percentage of buildings built Benefit where ai′ is the normalized value of the cost criterion, ai is
between 1966 and 1988 ith great value among all alternatives and amax is the
Dist_Fault Geometric distance of statistical Cost
units to North Tehran Fault maximum score in a certain criterion. Table 6 shows the
normalized attribute values of the 10 sample units. In the
8 M. Moradi et al.

Table 5. Attribute values for 10 randomly selected statistical units in 1996 census.

ID Slope (%) Bld_Less4 (%) Bld_More4 (%) Bld_Bef66 (%) Bld_Bet66_88 (%) Dist_Fault (km)

776 14 17 8 52 13 6.751
894 1 4 0 84 6 3.890
186 26 2 0 91 19 1.435
666 4 88 0 97 0 5.476
522 6.62 9 1 63 5 2.054
233 17.66 46 10 84 24 1.947
1474 1 36 10 75 0 6.210
2518 0 86 36 60 0 16.735
2190 0 37 5 82 0 10.225
3146 0 96 20 54 2 22.885
Downloaded by [Chinese University of Hong Kong] at 05:19 09 February 2015

Table 6. Normalized attribute values for the 10 randomly selected statistical units.

ID Slope Bld_Less4 Bld_More4 Bld_Bef66 Bld_Bet66-88 Dist_Fault

776 0.53 0.17 0.22 0.53 0.54 0.71


894 0.03 0.04 0 0.86 0.25 0.84
186 1 0.02 0 0.93 0.79 0.94
666 0.15 0.91 0 1 0 0.77
522 0.25 0.09 0.02 0.64 0.20 0.92
233 0.67 0.47 0.27 0.86 1 0.92
1474 0.03 0.37 0.27 0.77 0 0.73
2518 0 0.89 1 0.61 0 0.27
2190 0 0.38 0.13 0.84 0 0.56
3146 0 1 0.55 0.55 0.083 0

seismic vulnerability assessment problem, in order to find quantifier needs at least one of the criteria to be satisfied,
the vulnerable areas, slope, percentage of weak buildings which is equal to ‘OR’ operator. The weight vector
with less than four floors, percentage of buildings that w1 = [1, 0, 0, 0, 0, 0] is obtained based on the quantifier
have four floors and more, percentage of buildings built ‘At least one’ (see Table 2). Multiplying this weight vector
before 1966 and percentage of buildings built between by descending-order vector of the input criteria results in
1966 and 1988 are the benefit criteria. However, the dis- the maximum attribute value. Therefore, the combined
tance of a building to the NTF is considered as a cost value (vulnerability degree) and maximum attribute value
criterion. will be the same. Under this strategy, 2200 statistical units
out of 3175 units have been classified in ‘Very Vulnerable’
class and only 151 statistical units are considered to be
Results and discussion ‘Very Low Vulnerable’ in the case of NTF activation. In
Figure 4 illustrates seven different solutions for the seis- addition, the impact of distance to NTF is very noticeable
mic vulnerability assessment problem based on the differ- here and causes approximately all of the northern statis-
ent decision strategies employed. Optimistic strategies tical units to be classified as ‘Very Vulnerable’ (Figure 4).
result in a higher vulnerability degree in comparison to Using the quantifiers ‘Few’ and ‘Some’, an optimistic
pessimistic ones, because optimistic quantifiers tend to decision about the seismic vulnerability problem has been
assign higher weights to higher attribute values and vice made. Based on these quantifiers, higher weights are
versa. Consequently, the number of units classified in assigned to the higher attribute values and vice versa.
‘Very Vulnerable’ and ‘Vulnerable’ classes is increasing Therefore, it leads to a combined value near the maximum
in optimistic decision strategies. However, in the pessi- value. The weight vectors w2 = [0.69 0.10 0.06 0.05 0.04
mistic decision strategies, most of the units are classified 0.03] and w3 = [0.40 0.16 0.12 0.10 0.09 0.08] are
in ‘Very Low Vulnerable’ and ‘Low Vulnerable’ classes. obtained using ‘Few’ and ‘Some’ quantifiers, respectively
Table 7 illustrates the vulnerability degree of the 10 sta- (Table 2). Although 44% of the total units still correspond
tistical units obtained by multiplying the weight vectors by to the ‘Very Vulnerable’ class, the number of units in
descending-order vector of the attribute values. ‘Very Low Vulnerable’ class increased by 304 to 455
Using the quantifier ‘At least one’ means that an using the quantifier ‘Few’. ‘Some’ is a more pessimistic
extremely optimistic decision strategy is expected. This quantifier than ‘Few’ so that the results of ‘Few’ are more
Annals of GIS 9
Downloaded by [Chinese University of Hong Kong] at 05:19 09 February 2015

Figure 4. Seven different Tehran’s seismic vulnerability maps based on the different quantifiers using OWA operator.
10 M. Moradi et al.

Table 7. Aggregated seismic vulnerability degrees for the 10 selected statistical units based on the seven different decision strategies.

ID At least one Few Some Half Many Most All

776 89.7 79.2 57.3 37.6 22.0 9.7 0.50


894 91.8 81.5 57.8 37.7 22.4 10.8 0.52
186 96.9 80.6 55.8 36.5 21.8 11.2 1.55
666 85.0 79.6 60.4 40.29 23.6 10.5 0.56
522 81.3 68.6 49.2 32.7 19.6 9.3 1.44
233 92.6 79.9 63.4 46.9 31.5 17.8 6.65
1474 87 70.4 46.4 29.1 16.6 7.3 0.38
2518 95.0 82.2 58.3 37.8 21.9 9.6 0.51
2190 94.0 80.7 57.5 37.1 21.3 9.38 0.49
3146 89.4 71.7 50.9 32.5 18.9 8.3 0.44

similar to the ‘Max’ operator, while the vulnerability versa. Under the strategy of ‘More’, over 85% of units are
Downloaded by [Chinese University of Hong Kong] at 05:19 09 February 2015

degree obtained by ‘Some’ is more comparable to WLC. classified in ‘Vulnerable’, ‘Moderate’ and ‘Low
Moreover, the impact of distance to NTF decreased in Vulnerable’ classes, while based on ‘Most’ quantifier,
comparison to ‘At least one’, and some of the northern only 53% of the total statistical units are classified in
statistical units are not considered to be‘ Very Vulnerable’ these classes. Therefore, the ‘Most’ quantifier has a ten-
anymore (see Figure 4). dency to provide more extreme results than ‘More’ (see
The quantifier ‘Half’ is a neutral quantifier and corre- Figure 4).
sponds to a strategy that 50% of the criteria should be When the expert wants all of the criteria to be satisfied
satisfied by an acceptable alternative. Based on this quan- by an acceptable result, ‘All’ should be chosen as a lin-
tifier, equal weights are assigned to the high and low guistic quantifier. Therefore, the quantifier ‘All’ corre-
attribute values. Therefore, w4 = [0.16 0.16 0.16 0.16 sponds to the ‘AND’ operator. The weight vector
0.16 0.16] should be multiplied by the descending vector w7 = [0 0 0 0 0 1] is calculated using Equation (6)
of the input attribute values. In this strategy, the combined which means that the weight of all attribute values is
value does not tend to be extremely high or extremely low. zero except the minimum one. The number of units in
Consequently, most of the statistical units are classified in ‘Very Low Vulnerable’ class increased by 824 to 2181 so
‘Moderate’ class (see Figure 5). that over 68% of total units are in this class (see Figure 5).
‘More’ and ‘Most’ are quantifiers that solve the pro- In addition, the impact of distance to NTF increased and
blem with 33% and 17% of optimism degree, respectively caused almost all of the southern statistical units, which
(see Table 2). They are pessimistic quantifiers that provide have a relatively large distance to NTF, to be assumed as
a range of answers which fall between ‘AND’ and ‘WLC’. ‘Very Low Vulnerable’ units (see Figure 4).
Using Equation (6), w5 = [0.02 0.08 0.13 0.19 0.24 0.30] Figure 5 illustrates the results achieved from this
for ‘More’ and w6 = [0.00 0.00 0.02 0.09 0.26 0.58] for research in addition to those presented by Samadi Alinia
‘Most’ quantifiers are obtained. Both quantifiers tend to and Delavar (2011) using GrC. Most of the urban statis-
assign higher weights to lower attribute values and vice tical units in this method are classified in ‘Moderate’ class

2500
Number of Statistical Units

2000
1500
1000
500
0
At Least Granular
Few Some Half Many Most All
One Computing
Very Low Vulnerable 151 455 90 327 559 1357 2181 280
Low Vulnerable 178 118 442 758 968 767 174 489
Moderate 219 459 889 1012 997 747 557 1023
Vulnerable 427 726 1263 897 367 180 160 834
Very Vulnerable 2200 1417 491 181 284 124 103 547

Figure 5. Number of statistical units in each class based on the different quantifiers using OWA.
Annals of GIS 11

which means an approximately neutral approach was similar vulnerability pattern is used in both methods.
applied. The number of statistical units classified in Moreover, the behaviour of neutral quantifiers such as
‘Vulnerable’ and ‘Low Vulnerable’ classes is 834 and ‘Half’, ‘Some’ and ‘Many’ is better confirmed by the
489, respectively, and nearly 18% of statistical units are GrC results in comparison to very optimistic and very
‘Very Vulnerable’ according to this model. In addition, pessimistic quantifiers. The comparison between OWA
GrC is generally less sensitive to distance to NTF in and Int_Fuzzy indicates that the similarity between the
comparison to the OWA operator. two methods reaches 0.89% for ‘Half’ and has its mini-
One of the most important issues in MCDM is valida- mum in 0.59% for ‘At least one’. It shows that a neutral
tion of the model. Validation indicates the extent to which strategy is applied in the Int_Fuzzy model.
the model is simulating the real-world conditions. In order This article proposed an urban earthquake physical
to validate an MCDM model, the results should be com- vulnerability assessment model based on the OWA opera-
pared with the real-world evidence. However, the last tor that provides a wide range of answers. In addition,
large earthquake occurred in TMA struck the city in linguistic quantifiers are used to implement different deci-
1830 (Hashemi and Alesheikh 2011). Therefore, valida- sion strategies based on different optimism degrees.
Downloaded by [Chinese University of Hong Kong] at 05:19 09 February 2015

tion is applicable only through the comparison of results Initially, the most important criteria affecting the seismic
with the other MCDM models. In this article, Spearman vulnerability degree are identified. Then, a set of seismic
correlation coefficient (r) is applied to measure the asso- vulnerability degrees are obtained using seven linguistic
ciation between vulnerability degrees obtained for each quantifiers by aggregating the attribute values of the six
statistical unit based on GrC, Int_Fuzzy and OWA. important criteria.
Spearman correlation coefficient is calculated using The most outstanding advantages of the proposed
Equation (13) (Zar 1972): model are as follows:
"
P #
6  Ni¼1 D2i (1) Including the optimism degree as a key factor into
r ¼1 (13)
N  ð N 2  1Þ the decision process. An optimistic decision-
maker tends to choose an alternative which has
where Di is the difference between vulnerability degrees in the shortest distance to the ideal solution.
the first method and the second one for the ith statistical However, the alternative with largest distance to
unit and N is the total number of statistical units in TMA. negative ideal solution is the most preferable one
The value of Spearman correlation coefficient r ranges for a pessimistic decision-maker. This article used
from −1.00 to +1.00. The positive values indicate a direct the OWA model in order to provide five distinct
correlation between the results of the two MCDM meth- decision strategies for Tehran’s seismic vulnerabil-
ods, while negative values mean that there is an inverse ity assessment between these two extreme cases.
relationship between the results. The zero value of the (2) Integrating the OWA operator with linguistic
Spearman correlation coefficient shows that there is no quantifiers. There are a variety of methods to
correlation between the results of two MCDM methods. obtain order weights: (1) minimum variation and
Table 8 indicates the Spearman correlation coefficient maximum dispersion, which have a tendency to
values (r) between linguistic quantifiers, GrC and maximize the trade-off between the criteria, in
Int_Fuzzy. It can be concluded from Table 8 that the order to involve all of the criteria equally (Wang
OWA operator using quantifier ‘Some’ has the maximum and Parkan 2005); (2) using linguistic quantifiers
likeness to GrC method (r = 0.91). In addition, the mini- to describe the expert’s opinion on ‘how many
mum value of Spearman correlation coefficient between criteria should be satisfied by the most preferable
GrC and linguistic quantifiers is 0.67 which ensures that a alternative’. The most prominent superiority of

Table 8. The Spearman correlation coefficient values for linguistic quantifiers, GrC and Int_Fuzzy.

Method At Least One Few Some Half Many Most All GrC Int_Fuzzy

At Least One 1.00 0.86 0.82 0.78 0.62 0.57 0.49 0.70 0.59
Few 0.86 1.00 0.88 0.81 0.65 0.61 0.52 0.78 0.65
Some 0.82 0.88 1.00 0.83 0.73 0.67 0.62 0.91 0.77
Half 0.78 0.81 0.83 1.00 0.85 0.83 0.80 0.86 0.89
Many 0.62 0.65 0.73 0.85 1.00 0.90 0.83 0.79 0.84
Most 0.57 0.61 0.67 0.83 0.90 1.00 0.85 0.73 0.79
All 0.49 0.52 0.62 0.80 0.83 0.85 1.00 0.67 0.71
GrC 0.70 0.78 0.91 0.86 0.79 0.73 0.67 1.00 0.86
Int_Fuzzy 0.59 0.65 0.77 0.89 0.84 0.79 0.71 0.86 1.00
12 M. Moradi et al.

this method is the similarity of linguistic quanti- The results demonstrate that a large number of north
fiers to conversation language. Therefore, the Tehran statistical units and a few southern ones are very
experts could easily specify their preferences with- vulnerable based on approximately all decision strategies.
out using numbers and equations. Moreover, using these results in urban planning, especially
(3) Normalization of the attribute values. A number of in urban development and after earthquake search and
MCDM algorithms use normalized values of the rescue planning, could facilitate sustainable development
criteria, while others use the original value. from an urban disaster management perspective, which is
Normalization makes the attribute values with dif- substantially necessary in smart mega city growth and
ferent units comparable with each other. management.
In addition, the comparison of the results with those of
On the other hand, the major limitations of the pro- GrC and Int_Fuzzy model using Spearman correlation
posed model are: coefficient indicates that the vulnerability degrees obtained
from neutral quantifiers are extremely certified with the
(1) Lack of up-to-date data. One of the most impor- two models. However, ‘All’ and ‘At least one’ quantifiers
Downloaded by [Chinese University of Hong Kong] at 05:19 09 February 2015

tant issues in the case of spatial MCDM is the have only 67% and 70% similarities to the GrC and 0.71%
availability of up-to-date data. Data used in this and 0.59% similarities to the Int_Fuzzy model,
article are related to a census taken in 1996. respectively.
Consequently, the model represents the vulnerabil- Although the proposed model provides promising
ity degree of approximately 18 years ago. results which address a number of concerns to some
(2) Neglecting the interrelation between faults. There extent, new issues in disaster management necessitate
are at least three main known faults around further research. Future works may include a number of
Tehran. However, in the current research, it is issues as follows:
assumed that the earthquake was caused by the
activation of NTF only. In other words, the activa- (1) Group decision-making. Different experts not only
tion of one fault has a complicated impact on the have different ideas about the same topic, but they
other faults, which has been neglected here. also may be wrong in some cases. Using more
than one expert in decision-making may provide
more accurate results and better which ensures the
Conclusion reliability of the decision strategy. Some exten-
This article indicates the applicability of the quantifier- sions of the OWA operator such as intuitionistic
guided OWA operator in earthquake physical vulnerability fuzzy OWA (Su et al. 2012) and density-induced
assessment in TMA. Linguistic quantifiers are used in OWA (Ma and Guo 2011) can be employed to
order to enable disaster management experts to apply enable group decision-making.
any decision strategy by selecting the appropriate quanti- (2) Adding a number of layers such as streets, gas
fier. In addition, this model is able to include both quali- stations and aqueducts. Data layers used in this
tative and quantitative information about decision strategy study were limited due to data availability.
into the model by using linguistic quantifiers, which are Future researches may improve this model by
powerful means of human–computer interaction. In other including more data layers indicating the loca-
words, the model proposed in this research specified the tion of flammable structures and the transporta-
earthquake vulnerability regarding the optimism degree. tion network.
Six important criteria are included in this model, which (3) Investigating earthquake human vulnerability.
cover five benefit criteria: (1) average slope of each sta- Human seismic vulnerability is as critical as phy-
tistical unit; (2) percentage of buildings having less than sical seismic vulnerability. If an earthquake hap-
four floors; (3) percentage of buildings having four floors pens at night, more losses would occur in
and more; (4) percentage of buildings built before 1966; residential areas, while fewer fatalities would
(5) percentage of buildings built between 1966 and 1988; occur in commercial and industrial ones.
and one cost criterion: the distance of statistical units to Therefore, human seismic vulnerability plays a
NTF. Then, the weight vectors of these six attribute values key role in earthquake search and rescue opera-
were obtained based on the selected quantifiers. tions and precise information about human vulner-
Integrating the attribute values using these weight vectors ability improves the efficiency of finding injured
led to a seismic vulnerability degree of each statistical people. This article only proposed earthquake phy-
unit. Then, the vulnerability degrees obtained by this sical vulnerability regarding the optimism degree.
method were classified in five classes using ‘Equal Including optimism degree in human vulnerability
Intervals’ in ArcGIS software. could be the subject of future studies.
Annals of GIS 13

Finally, by applying natural language terms, this model Information Science 27 (12): 2455–2482. doi:10.1080/
improves the efficiency of the previous models in the 13658816.2013.815356.
case of including experts’ opinion in mathematical Fodor, J., J.-L. Marichal, and M. Roubens. 1995.
“Characterization of the Ordered Weighted Averaging
computations. Besides, this model could be applied in Operators.” IEEE Transactions on Fuzzy Systems 3 (2):
other areas of disaster management, such as flood 236–240. doi:10.1109/91.388176.
and tsunami vulnerability assessment, with partial Fullér, R., and P. Majlender. 2003. “On Obtaining Minimal
modification. Variability OWA Operator Weights.” Fuzzy Sets and
Systems 136 (2): 203–215. doi:10.1016/S0165-0114(02)
00267-1.
Greene, R., R. Devillers, J. E. Luther, and B. G. Eddy. 2011. “GIS‐
Funding Based Multiple‐Criteria Decision Analysis.” Geography
This work was supported by Deputy for Research, College Compass 5 (6): 412–432. doi:10.1111/j.1749-8198.2011.
of Engineering, University of Tehran [grant number 42/6/ 00431.x.
Hamzehloo, H., F. Vaccari, and G. F. Panza. 2007. “Towards a
8103002]. Reliable Seismic Microzonation in Tehran, Iran.” Engineering
Geology 93 (1–2): 1–16. doi:10.1016/j.enggeo.2007.05.001.
Downloaded by [Chinese University of Hong Kong] at 05:19 09 February 2015

Hashemi, M., and A. A. Alesheikh. 2011. “A GIS-Based Earthquake


Note Damage Assessment and Settlement Methodology.” Soil
1. Analytical hierarchy process. Dynamics and Earthquake Engineering 31 (11): 1607–1617.
doi:10.1016/j.soildyn.2011.07.003.
Jahanpeyma, M., M. R. Delavar, M. R. Malek, and N. Kamalian.
2007. “Analytical Evaluation of Propagation of Uncertainty
References in Assessment of Seismic Vulnerability of Tehran Using
Aghamohammadi, H., M. S. Mesgari, A. Mansourian, and D. Geospatial Information System.” Proceedings of ISSDQ
Molaei. 2013. “Seismic Human Loss Estimation for an 2007, edited by A. Stein, Enschede, June 13–15, 5p.
Earthquake Disaster Using Neural Network.” International Jankowski, P. 1995. “Integrating Geographical Information
Journal of Environmental Science and Technology 10 (5): Systems and Multiple Criteria Decision-Making Methods.”
931–939. doi:10.1007/s13762-013-0281-5. International Journal of Geographical Information Systems
Aghataher, R., M. R. Delavar, and N. Kamalian. 2005. 9 (3): 251–273. doi:10.1080/02693799508902036.
“Weighing of Contributing Factors in Vulnerability of Jiang, H., and J. Ronald Eastman. 2000. “Application of Fuzzy
Cities against Earthquakes.” Paper presented at Proceedings Measures in Multi-Criteria Evaluation in GIS.” International
of Map Asia Conference, Jakarta, August 22–25, 6p. Journal of Geographical Information Science 14 (2): 173–
Al-Adamat, R., A. Diabat, and G. Shatnawi. 2010. “Combining 184. doi:10.1080/136588100240903.
GIS with Multicriteria Decision Making for Siting Water Kolat, Ç., V. Doyuran, C. Ayday, and M. Lütfi Süzen. 2006.
Harvesting Ponds in Northern Jordan.” Journal of Arid “Preparation of a Geotechnical Microzonation Model Using
Environments 74 (11): 1471–1477. doi:10.1016/j.jaridenv. Geographical Information Systems Based on Multicriteria
2010.07.001. Decision Analysis.” Engineering Geology 87 (3–4): 241–
Anbazhagan, P., K. K. S. Thingbaijam, S. K. Nath, J. N. 255. doi:10.1016/j.enggeo.2006.07.005.
Narendara Kumar, and T. G. Sitharam. 2010. “Multi- Kolat, C., R. Ulusay, and M. Lutfi Suzen. 2012. “Development
Criteria Seismic Hazard Evaluation for Bangalore City, of Geotechnical Microzonation Model for Yenisehir
India.” Journal of Asian Earth Sciences 38 (5): 186–198. (Bursa, Turkey) Located at a Seismically Active Region.”
doi:10.1016/j.jseaes.2010.01.001. Engineering Geology 127: 36–53. doi:10.1016/j.enggeo.
Berberian, M., and R. S. Yeats. 2001. “Contribution of 2011.12.014.
Archaeological Data to Studies of Earthquake History in the Ma, F.-M., and Y.-J. Jun Guo. 2011. “Density‐Induced Ordered
Iranian Plateau.” Journal of Structural Geology 23 (2–3): Weighted Averaging Operators.” International Journal of
563–584. doi:10.1016/S0191-8141(00)00115-2. Intelligent Systems 26 (9): 866–886. doi:10.1002/int.20500.
Boroushaki, S., and J. Malczewski. 2010. “Using the Fuzzy Malczewski, J. 1996. “A GIS-Based Approach to Multiple
Majority Approach for GIS-Based Multicriteria Group Criteria Group Decision-Making.” International Journal
Decision-Making.” Computers & Geosciences 36 (3): 302– of Geographical Information Systems 10 (8): 955–971.
312. doi:10.1016/j.cageo.2009.05.011. doi:10.1080/02693799608902119.
Chini, M., N. Pierdicca, and W. J. Emery. 2009. “Exploiting Malczewski, J. 1999. GIS and Multicriteria Decision Analysis.
SAR and VHR Optical Images to Quantify Damage Hoboken, NJ: John Wiley & Sons.
Caused by the 2003 Bam Earthquake.” IEEE Transactions Malczewski, J. 2006. “Ordered Weighted Averaging with Fuzzy
on Geoscience and Remote Sensing 47 (1): 145–152. Quantifiers: GIS-Based Multicriteria Evaluation for Land-
doi:10.1109/TGRS.2008.2002695. Use Suitability Analysis.” International Journal of Applied
Duzgun, H. S. B., M. S. Yucemen, H. S. Kalaycioglu, K. Celik, Earth Observation and Geoinformation 8 (4): 270–277.
S. Kemec, K. Ertugay, and A. Deniz. 2011. “An Integrated doi:10.1016/j.jag.2006.01.003.
Earthquake Vulnerability Assessment Framework for Urban Mohanty, W. K., M. Yanger Walling, S. K. Nath, and I. Pal.
Areas.” Natural Hazards 59 (2): 917–947. doi:10.1007/ 2007. “First Order Seismic Microzonation of Delhi, India
s11069-011-9808-6. Using Geographic Information System (GIS).” Natural
Eldrandaly, K. A. 2013. “Exploring Multi-Criteria Decision Hazards 40 (2): 245–260. doi:10.1007/s11069-006-0011-0.
Strategies in GIS with Linguistic Quantifiers: An Extension Molina, S., D. H. Lang, and C. D. Lindholm. 2010. “Selena–An
of the Analytical Network Process Using Ordered Weighted Open-Source Tool for Seismic Risk and Loss Assessment
Averaging Operators.” International Journal of Geographical Using a Logic Tree Computation Procedure.” Computers &
14 M. Moradi et al.

Geosciences 36 (3): 257–269. doi:10.1016/j.cageo.2009. Engineering 28 (10–11): 812–835. doi:10.1016/j.soildyn.


07.006. 2007.10.012.
O’Hagan, M. 1988. “Aggregating Template or Rule Antecedents Tayyebi, A., B. C. Pijanowski, and A. H. Tayyebi. 2011. “An
in Real-Time Expert Systems with Fuzzy Set Logic.” In Urban Growth Boundary Model Using Neural Networks,
Twenty-Second Asilomar Conference on Signals, Systems GIS and Radial Parameterization: An Application
and Computers, vol. 2, 681–689, October 31–November 2. to Tehran, Iran.” Landscape and Urban Planning 100 (1–
IEEE. doi:10.1109/ACSSC.1988.754637. 2): 35–44. doi:10.1016/j.landurbplan.2010.10.007.
Panahi, M., F. Rezaie, and S. A. Meshkani. 2014. “Seismic Tchalenko, J. S. 1975. “Seismotectonic Framework of the
Vulnerability Assessment of School Buildings in Tehran City North Tehran Fault.” Tectonophysics 29 (1–4): 411–420.
Based on AHP and GIS.” Natural Hazards and Earth System doi:10.1016/0040-1951(75)90169-9.
Science 14 (4): 969–979. doi:10.5194/nhess-14-969-2014. Walker, B. B., C. Taylor-Noonan, A. Tabbernor, H. Bal, D.
Rashed, T., and J. Weeks. 2003. “Assessing Vulnerability to Bradley, N. Schuurman, and J. J. Clague. 2014. “A Multi-
Earthquake Hazards through Spatial Multicriteria Analysis Criteria Evaluation Model of Earthquake Vulnerability in
of Urban Areas.” International Journal of Geographical Victoria, British Columbia.” Natural Hazards 65 (1): 97–
Information Science 17 (6): 547–576. doi:10.1080/ 113.
1365881031000114071. Wang, Y.-M., and C. Parkan. 2005. “A Minimax Disparity
Remo, J. W. F., and N. Pinter. 2012. “Hazus-MH Earthquake Approach for Obtaining OWA Operator Weights.”
Downloaded by [Chinese University of Hong Kong] at 05:19 09 February 2015

Modeling in the Central USA.” Natural Hazards 63 (2): Information Sciences 175 (1–2): 20–29. doi:10.1016/j.
1055–1081. doi:10.1007/s11069-012-0206-5. ins.2004.09.003.
Ritz, J. F., S. Balescu, S. Soleymani, M. Abbassi, H. Nazari, K. Xu, Z. 2005. “An Overview of Methods for Determining OWA
Feghhi, E. Shabanian, H. Tabassi, Y. Farbod, and M. Weights.” International Journal of Intelligent Systems 20
Lamothe. 2003. “Determining the Long-Term Slip Rate (8): 843–865. doi:10.1002/int.20097.
along the Mosha Fault, Central Alborz, Iran. Implications Yager, R. R. 1988. “On Ordered Weighted Averaging
in Terms of Seismic Activity.” Paper read at Proceeding of Aggregation Operators in Multicriteria Decision making.”
the 4th International Conference on Seismology and IEEE Transactions on Systems, Man, and Cybernetics 18
Earthquake Engineering. (1): 183–190. doi:10.1109/21.87068.
Samadi Alinia, H., and M. R. Delavar. 2011. “Tehran’s Seismic Yager, R. R. 1996. “Quantifier Guided Aggregation Using OWA
Vulnerability Classification Using Granular Computing Operators.” International Journal of Intelligent Systems 11
Approach.” Applied Geomatics 3 (4): 229–240. (1): 49–73. doi:10.1002/(SICI)1098-111X(199601)11:1<49::
doi:10.1007/s12518-011-0068-7. AID-INT3>3.3.CO;2-L.
Schmidtlein, M. C., J. M. Shafer, M. Berry, and S. L. Cutter. Yager, R. R. 1999. “Nonmonotonic OWA Operators.” Soft
2011. “Modeled Earthquake Losses and Social Vulnerability Computing-A Fusion of Foundations, Methodologies and
in Charleston, South Carolina.” Applied Geography 31 (1): Applications 3 (3): 187–196.
269–281. doi:10.1016/j.apgeog.2010.06.001. Yager, R. R., and D. P. Filev. 1999. “Induced Ordered Weighted
Silavi, T., M. R. Delavar, M. R. Malek, N. Kamalian, and K. Averaging Operators.” IEEE Transactions on Systems, Man
Karimizand. 2006. “An Integrated Strategy for GIS-Based and Cybernetics, Part B (Cybernetics) 29 (2): 141–150.
Fuzzy Improved Earthquake Vulnerability Assessment.” doi:10.1109/3477.752789.
Proceedings of Conference, ISPRS, The Second Yager, R. R., J. Kacprzyk, and G. Beliakov. 2011. Recent
International Symposium on Geo-information for Disaster Developments in the Ordered Weighted Averaging Operators:
Management (Gi4DM), Goa, September 25–26, 6p. Theory and Practice. Vol. 265. Berlin: Springer.
Solaymani, A. S., J.-F. Ritz, and M. R. Abbassi. 2011. “Left- Yang, Y., F. Benjamin Zhan, and L. Lin 2011. “Estimating Seismic
Lateral Active Deformation along the Mosha–North Losses of Schools Using SELENA: The Case of Wenchuan
Tehran Fault System (Iran): Morphotectonics and Earthquake.” Paper presented at 19th International
Paleoseismological Investigations.” Tectonophysics 497 Conference on Geoinformatics, Shanghai, June 24–26, 1–6.
(1–4): 1–14. doi:10.1016/j.tecto.2010.09.013. IEEE. doi:10.1109/GeoInformatics.2011.5980908.
Su, Z.-X., G.-P. Xia, M.-Y. Chen, and L. Wang. 2012. “Induced Zar, J. H. 1972. “Significance Testing of the Spearman
Generalized Intuitionistic Fuzzy OWA Operator for Multi- Rank Correlation Coefficient.” Journal of the American
Attribute Group Decision Making.” Expert Systems with Statistical Association 67 (339): 578–580. doi:10.1080/
Applications 39 (2): 1902–1910. doi:10.1016/j.eswa.2011. 01621459.1972.10481251.
08.057. Zarghami, M., and F. Szidarovszky. 2008. “Fuzzy Quantifiers in
Tantala, M. W., G. J. P. Nordenson, G. Deodatis, and K. Jacob. Sensitivity Analysis of OWA Operator.” Computers &
2008. “Earthquake Loss Estimation for the New York City Industrial Engineering 54 (4): 1006–1018. doi:10.1016/j.
Metropolitan Region.” Soil Dynamics and Earthquake cie.2007.11.012.

Vous aimerez peut-être aussi