Journal Pre-proof

A location-allocation optimization model for post-earthquake emergency

shelters using network-based multi-criteria decision-making

Moslem Dehnavi Eelagh, Rahim Ali Abbaspour

PII: S2772-6622(24)00034-1
DOI: https://doi.org/10.1016/j.dajour.2024.100430
Reference: DAJOUR 100430

To appear in: Decision Analytics Journal

Received date : 12 December 2023

Revised date : 4 February 2024
Accepted date : 14 February 2024

Please cite this article as: M.D. Eelagh and R.A. Abbaspour, A location-allocation optimization
model for post-earthquake emergency shelters using network-based multi-criteria decision-making,
Decision Analytics Journal (2024), doi: https://doi.org/10.1016/j.dajour.2024.100430.

This is a PDF file of an article that has undergone enhancements after acceptance, such as the
addition of a cover page and metadata, and formatting for readability, but it is not yet the definitive
version of record. This version will undergo additional copyediting, typesetting and review before it
is published in its final form, but we are providing this version to give early visibility of the article.
Please note that, during the production process, errors may be discovered which could affect the
content, and all legal disclaimers that apply to the journal pertain.

© 2024 The Author(s). Published by Elsevier Inc. This is an open access article under the CC
BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
 Journal Pre-proof

A Location-Allocation Optimization Model for Post-Earthquake

Emergency Shelters Using Network-Based Multi-Criteria Decision-

of
Making
Moslem Dehnavi Eelagh1, Rahim Ali Abbaspour2*

pro
1
Ph.D. Student in Geospatial Information Systems, School of Surveying and Geospatial
Engineering, College of Engineering, University of Tehran, Tehran, Iran
([email protected])
2
Associate Professor, School of Surveying and Geospatial Engineering, College of Engineering,
University of Tehran, Tehran, Iran ([email protected])

re-
lP
rna
Jou

*
Corresponding author
 Journal Pre-proof

Abstract

It is imperative for residents to have access to safe facilities after an earthquake. This ensures the

of
safety and well-being of individuals while minimizing the risks posed by aftershocks. This
research has used multi-criteria decision-making (MCDM) and network-based analysis to select
and allocate emergency shelters (ESs). Several ESs are initially chosen as potential candidates. A
weighting process is then used to evaluate various criteria, including proximity to the fault, fire
stations, hospitals, main roads, the area of the ESs, and the population’s vulnerability. The centers

pro
are evaluated and ranked using the CRiteria Importance Through Intercriteria Correlation
(CRITIC) and Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS)
methods. The Maximized Weighted Capacitated Coverage with a Radius (MWCCR) problem is
used to address location and allocation issues for varying ESs using option rankings. The findings
showed that increasing the number of centers does not always lead to a higher level of service
delivery, assuming a consistent service delivery radius. The distribution of the centers is more
crucial. Additionally, line density analysis is used to evaluate traffic conditions in the study area,
assisting in finding areas with heavy traffic flow. When the radius of access for ESs is assumed to
re-
be small, the main roads bear less additional traffic, and with the increase of the radius, the amount
of traffic on the main roads gradually increases. This is valuable information for emergency
services following an earthquake.

Keywords: Multi-criteria decision-making, network-based model, emergency shelters, maximize

weighted capacitated coverage with a Radius, TOPSIS; CRITIC
lP
rna
Jou
 Journal Pre-proof

1-Introduction

Earthquakes, as natural disasters, can cause extensive damage. Due to their potential to cause

of
widespread damage to critical urban infrastructure, they are a significant challenge for large cities.
Therefore, organizing crisis planning, including implementing policies that govern urban
decisions, is very important to respond effectively to earthquakes [1, 2]. After an earthquake, one
of the priorities is to quickly and safely evacuate citizens to protect them from potential hazards

pro
such as earthquakes, fires, and the spread of infectious diseases [3]. Emergency shelters (ESs) are
safe places with necessary services for evacuating citizens during earthquakes [4]. Therefore, it is
crucial to construct secure and appropriately located emergency evacuation facilities to mitigate
risks following an earthquake. These facilities should have sufficient resources to support affected
citizens [5]. Researchers face challenges in selecting the appropriate location for emergency
evacuation due to multiple criteria. The city officials should allocate a budget to adapt the facilities
and create a wide range of services in these centers. Therefore, it is crucial to determine the optimal
location and quantity of ESs to minimize costs. Selecting these centers' locations involves a
location-allocation problem with two components. A positioning component determines the
re-
placement of ESs and outlines how these centers serve the population in the allocation section [6,
7]. Therefore, to determine appropriate locations for ESs, a solution must be selected to address
the location allocation problem.

Various approaches have been used in studies on location allocation for ES planning. Some
approaches have focused on allocating ESs to citizens in a single-objective manner, considering
only one criterion, such as time or distance. Many studies commonly use single-objective
lP
techniques such as p-median, p-center, and maximum coverage. Multi-objective and hierarchical
models have been used in some studies. Additionally, geographic information systems (GIS)
methods, such as the Voronoi diagram, can be used [8, 9].

The study starts with a background review using a single-objective approach, followed by
explanations related to the research. The study [10] addresses the issue of assigning ESs to demand
rna

points as a set covering problems. Also, a time limit or interval has been considered for each
demand center, and linear programming (LP) has been used to model the problem. The study [11]
used LP to maximize citizen coverage in evacuation planning. A GIS was utilized for this purpose.
The study [12] proposes a modified particle swarm optimization (MPSO) algorithm for allocating
ESs. The simulated annealing (SA) algorithm is used to overcome local optima and enhance search
capability. The study [13] has developed a model to determine the optimal location for ESs and to
assign evacuees to them using the shortest routes. The goal is to minimize the total evacuation
time. The model that was created is a nonlinear mixed-integer programming model. The second-
order cone programming technique has been utilized to solve the model. The proposed model
Jou

considers the significance of both the quantity and location of ESs and the creation of indicators
for efficiency and fairness. In a study by [14], a method based on mixed integer linear
programming is proposed for selecting the location of ESs. The mathematical model presented
aims to maximize the minimum weight of ESs. In a study by researchers [15], the location and
allocation of ESs in the first district of Tehran were examined using GIS. They utilized the
Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS), a straightforward
clustering method, along with two meta-heuristic algorithms: particle swarm optimization (PSO)
and ant colony optimization (ACO). The TOPSIS method and simple clustering have been used to
 Journal Pre-proof

select candidate centers based on multiple criteria. These criteria include slope, area, population
density surrounding the centers, distance from faults, distance between centers, center utilization,
distance from roads, and parcel distance. The point allocation approach has been utilized to assign

of
weights to the criteria. The allocation problem was then solved using the two aforementioned meta-
heuristic algorithms. The study [8] proposed multi-level location models to determine the optimal
emergency location in Jianchuan, China, based on the severity of the emergency. The objective is
to minimize transfer and construction expenses while maximizing the coverage rate. The study

pro
utilized GIS and weighted Voronoi diagram (WVD) models to analyze the capacity of ESs.

Multi-objective models have also been widely used in the problem of location-allocation of ESs,
some of them are explained below. In a study by [16], a model for aid distribution was developed
using the multi-objective optimal planning method. This model is designed to assist in planning
aid systems for natural disaster scenarios. This model had three objectives: minimizing total cost
and travel time while maximizing the minimum satisfaction. In their study, the authors [17] discuss
the simultaneous consideration of selecting the location of the ES and regional planning of the
service areas. A bi-objective model is proposed to minimize the total traveled distance and cost
re-
while considering capacity and proximity constraints. A non-dominated sorting genetic algorithm
(NSGA) has been developed for solving bi-objective problems. The Pareto-optimal strategy and
feasibility-based rule have been used to balance objectives. The study [18] implemented a multi-
objective model to allocate residents to shelters during earthquakes in Beijing, China, using it as a
case study. The objectives of this model were to reduce the evacuation time from residential areas
to a specific ES and to minimize the total area. Two limitations were considered regarding the
capacity and service radius of ES. The PSO algorithm was initially modified by implementing the
lP
von Neumann structure and subsequently used to solve the problem. In a study [20], a
mathematical model was implemented with two objectives: minimizing the total evacuation time
and reducing the total area of ESs. A PSO algorithm was employed to solve the problem.

Some studies have focused solely on location and aimed to identify distribution centers for
emergency evacuation, relief, and rescue during crises. Multi-criteria decision-making tools have
rna

been utilized for this purpose. In a study by [21], the analytical hierarchy process (AHP) method
based on TOPSIS and the goal programming (GP) model were used to identify the best location
for distribution centers. The TOPSIS method was utilized to rank places based on risk score, total
area, population, and distance from the center. The GP model proposes two goals: maximizing
points from TOPSIS and minimizing the number of distribution centers required to fulfill all
demands. In a study by [22], criteria were classified based on the crisis management standard using
GIS. The aid centers in the study area were assessed using standard criteria. The matrix was used
to measure the research criteria using the entropy method (EM). The Preference Ranking
Organization METhod for Enrichment Evaluations (PROMETHEE) was used to determine the
Jou

optimal options.

A more complete literature review for the years 2018 to 2023 is shown in Table 1. Studies are
classified based on problem definition, model, objective, ES evaluations, application, and scale.
As mentioned, the problem can simultaneously be considered as location or location-allocation.
Also, the single-objective or multi-objective model of the problem is meant and various objectives,
which can be monetary or non-monetary, can be considered in these issues. In some studies, the
candidate ESs are evaluated and weighted regarding structural and spatial characteristics, shown
 Journal Pre-proof

in the corresponding column. Also, the issue of locating and allocating ESs can be important in
various crises such as floods, earthquakes, infectious diseases, etc. The meaning of scale in the last
column is the extent of the study, and studies that have been carried out on urban scales are shown

of
with micro, and other studies considered in higher extent are shown with macro in Table 1.

Table 1- Literature review based on problem, medel, objectives, and applications.

Objectives
Paper Year Problem definition Model ES evaluations Application Scale

pro
Monetary Non-monetary
[22] 2018 Location allocation Single - Min total time - Flood Micro
[23] 2018 Location allocation Single - Min total distance TOPSIS Earthquake Micro
Min total distance
Min total time Weighted
[24] 2018 Location allocation Multi -
Min number of shelters method
Earthquake Micro
Min total shelter area
[25] Min total shelter area
2019 Location allocation Multi - - Earthquake Micro
Min total distance
[26] Min total weighted distance
Max cover demand
2021 Location allocation Single - - Flood Macro
Min distance or time

[27]

[28]
2021

2021
Location allocation

Location
Multi

-
re-
Min total cost

-
Max attendance
Min total time
Min number of shelters

-
-
Multiple
Criteria
Evaluation
Flood

Flood
Traffic
Macro

Micro

[29] 2021 Location allocation Multi Min total cost Min total time - Flood Macro
[30] GIS Spatial
Evacuation
2021 - - - Analysis and - Micro
simulation
Statistics
lP
[31] Min total time
2021 Location allocation Multi - Max the suitability fuzzy-VIKOR Earthquake Micro
Min number of sites
[32] 2022 Location Multi - - AHP Earthquake Simulated
[33] Min total time
2022 Location allocation Multi - - COVID-19 Micro
Max demand satisfaction
[34] Min investment
2022 Location allocation Multi - - - Micro
Min total time
[35] 2022 Location Multi Min total cost Min total time - Earthquake Micro
rna

[36] Suitability
Min total shelter area
2022 Location allocation Multi - evaluation Earthquake Micro
Min total distance
method
[37] 2022 Location allocation Single - Max min weight - Earthquake Micro
[38] Location allocation
2022 Multi Min total cost Min total time fuzzy method Earthquake Micro
routing
[39] 2023 Location allocation Single - Max cover demand - - Micro
[40] Min total distance Flood
2023 Location allocation Single - Macro
Max cover demand Landslide
[41] Min number of new shelters
2023 Location Multi Min total cost - Earthquake Micro
Min the square
Jou

[42] k-means
2023 Location - - - - Macro
clustering
[43] Large Group
Decision-
2023 Location - - - Earthquake Micro
Making
(LGDM)

The structure of this study is similar to the study [36], which has used a two-stage approach with
the difference that in this study, the approach of CRiteria Importance Through Intercriteria
Correlation (CRITIC) and TOPSIS is used for weighting and ranking ESs. Also, as is shown in
Table 1, Previous studies used more straightforward weighting methods. The criteria used in this
 Journal Pre-proof

study included the distance from the fault, fire stations, hospitals, main roads, the area of ESs, and
population vulnerability. The population vulnerability layer is a new criterion not explored in
previous research. After considering the criteria, the ESs have been ranked using the TOPSIS

of
method. The problem of maximizing weighted capacitated coverage with a radius (MWCCR) is
used to optimize the locations and allocations of ESs. This problem has been solved to determine
the number of required centers, the percentage of population covered, the occupancy percentage
of ESs, and the distance traveled. Moreover, the number of individuals covered is categorized by

pro
age and gender. The traffic conditions were analyzed using line density analysis to evaluate various
allocation methods. This type of analysis, essential for crisis management, has not been used in
previous studies.

This article is organized as follows. Preliminaries are described in Section 2. Materials and
methods including calculation of population vulnerability, CRITIC and TOPSIS methods are
described in Section 3. Results and discussion that include case study description, ranking of
candidate hubs description, location-assignment results, and traffic growth forecast are presented
in Section 4. This article ends with the conclusion of the model implementation in section 5.

2- Preliminaries
re-
In this section, all related symbols used in this article are presented. In addition, the problem is
defined, and its mathematical modeling is explained.
2-1- Definitions
lP
Numerous studies and research have been conducted on the issue of locating and allocating ESs.
This issue aims to select multiple temporary ES locations. Each center will be selected based on
its capacity to serve a population that can easily reach it. Therefore, we assume the set
S = 1, 2,..., N  as the ESs and the set D = 1, 2,..., M  as the demand centers or population centers.
In this case, we are searching for the set A = {(𝑠, 𝑑)|𝑠𝜖𝑆 ∗ , 𝑑𝜖𝐷∗ }. Set A is proposed as a solution
rna

to the location-allocation problem, where each element represents an allocation from ES s to

demand center d. Moreover, S* is a set of selected centers based on specific criteria before
allocation This set has N* members, which N*≤N. Moreover, D* represents the set of demand
centers covered by ESs. The number of demands is M*, which M*≤M. The choice of set A as a
solution can vary depending on the problem's goals and the criteria for selecting ESs. In this study,
we examine the set W = w1 , w2 , w3 ,..., wN  , which determines the weight of each corresponding
ES. The weights can be determined based on different criteria. Specific centers will likely be
selected when solving the location allocation problem. For instance, in Figure 1, when selecting
Jou

one of the three ESs based on the highest service coverage within a specific radius, the weighted
mode chooses the ES with the highest weight. In contrast, in the unweighted state, a center is
randomly selected.
 Journal Pre-proof

of
pro
re-
Figure 1 – Schematic illustration of the effect of weight on Ess allocation

2-2- Maximize Weighted Capacitated Coverage with a Radius (MWCCR)

The section explains MWCCR modeling. This model, like any other, is based on the following
assumptions.
lP
• Demand centers are population blocks where the population is concentrated at a central point.

• An earthquake occurred at night, and the entire population is in residential areas.

• Each demand center can be assigned to only one emergency ES, and the entire population is
allocated to one ES in each assignment.
rna

• Emergency ESs include mosques, schools, and public parking lots, which are assumed to have
adequate earthquake resistance.

This model aims to maximize the number of people served while considering each service center's
capacity and coverage radius constraints. In the modeling process, we encounter three sets of ESs
(I), demand centers (J), and weight of centers (W). We are also seeking allocations (A) that cover
the highest demand. The study includes nomenclature, indices, parameters, decision variables, and
proposed modeling as follows:
Jou

Nomenclature set
I Set of all ES locations
J Set of all demand locations
W Set of all weight of centers
Indices

iI -
jJ -
 Journal Pre-proof

Parameters

wi Weight of ES point iI

of
pj The population of demand point jJ
Cj The capacity of the ES point i  I

dij Transfer distance between the ES point and the demand point jJ
R Cover radius for each ES point

pro
Decision variable

xij 1, if demand point j is allocated to ES point i and zero, otherwise

yi 1, if ES point i is selected in the location-allocation process and zero, otherwise

max  wi p j xij (1)

iI jJ

x  1 j  J (2)
iI

p x
jJ
ij

j ij  Ci yi i  I

dij xij  R i  I , j  J
re- (3)

(4)

xij , yi  0,1 i  I , j  J (5)

lP
The goal of the problem is to maximize the number of people served, as stated in Equation (1).
Equation (2) restricts the problem of assigning each demand to only one ES. Equation (3) is
associated with the capacity and limits the allocation of each ES to no more than its capacity.
Equation (4) ensures that each center is assigned only the requests placed within a specific
neighborhood radius. The Equation (5) is related to the binary decision variables. In this study,
which aims to allocate emergency evacuation centers to residents after an earthquake, the capacity
rna

of each center has been determined using Equation (6):

Ai
Capi = (6)
S
Where (the capacity of the ith service center) is based on the number of people, Ai the total area
of the infrastructure of the service center, and S is the area required for each person.
3- Material and methods
Jou

In this section, we will introduce the proposed research methodology. According to Figure 2, raw
data has been collected, including information on faults, fire stations, hospitals, main roads,
population blocks, and candidate ESs. Various layers of information were generated using analysis
and GIS tools that relied on distance and neighborhood factors. One layer of information in this
study was population vulnerability, which was determined using criteria such as gender, age, and
marital status. The layers created in the previous stage were utilized to assign weights to the ESs.
The criteria weighting and ranking of ESs were done using CRITIC and TOPSIS methods,
 Journal Pre-proof

respectively. The MWCCR model was implemented after considering the criteria. Allocation plans
were determined as a result. An analysis was conducted on the population served and the additional
traffic generated during the crisis, using the results obtained in this stage. The following section

of
outlines various components of the proposed method, including the calculation of population
vulnerability, CRITIC, and TOPSIS.

pro
re-
lP

Figure 2 – Flowchart of the proposed method

rna

3-1- Population vulnerability calculation

The vulnerability of city residents to accidents varies depending on factors such as age, gender,
and marital status [44]. One of the criteria considered in this study for ranking ESs was the
vulnerability of individuals. Vulnerable areas with more vulnerable populations are more likely to
be selected and ranked higher during the location process. The population's vulnerability at the
city block level was determined using Equation 7.
Jou


t1,2,3
( ptwvtw + ptm vtm )
Vi = (7)
Pi
The population is divided into three groups: people under six years old (t=1), between 6 and 65
years old (t=2), and over 65 years old (t=3). Is the population vulnerability of the ith block, Pi is
the total population of the ith block, ptw is the population of women in the tth group, and vtw is the
vulnerability weight of women in the tth group. Moreover, ptm , vtm are the population of men and
 Journal Pre-proof

their vulnerability in the tth group, respectively. Moreover, the group of people between 6 and 65
years old are divided into four groups: single (s=1), married (s=2), widowed (s=3), and divorced

of
(s=4), and the vulnerability weight for This category of people obtained from Equations 8 and 9:


s1,2,3,4
p2ms v2ms
v2m = (8)
p2m


pro
p2ws v2ws
s1,2,3,4
v2w = (9)
p2w
Where p2ms and v2ms are the population of men between 6 and 65 years and their vulnerability are in
the sth set, respectively. The p2ws and v2ws are the population of women between 6 and 65 years and
their vulnerability in the sth set, respectively.
3-2- CRITIC
re-
The CRITIC technique, as suggested by Diakulaki et al. in 1995 [45], is an objective method for
determining the weight of indicators. This technique utilizes statistical measures of dispersion and
correlation to assign weights to indicators. Indicators with a higher standard deviation and less
dependent on various criteria are considered more important and given more weight. The technique
consists of seven steps, which will be explained in the following paragraphs. The decision matrix
should be normalized using the fuzzy method as the first step. Based on this method, the values
lP
range from 0 to 1, and Equation 10 is used.
aij − min(aij )
rij = j  1, 2,3,..., n (10)
max(aij ) − min(aij )

Where rij is the normalized value of the ith row and jth column of the decision matrix, aij is the
rna

value of the jth index of the ith option, and n is the number of indices. Moreover, the second step
calculates the standard deviation value for each index. The standard deviation is calculated for the
values of each column in the decision matrix at this stage. For this purpose, Equation 11 is used.

 (r ij −  j )2
(11)
j = i =1

m
Jou

Where is the standard deviation value of the jth index,  j is the average of the jth index, and m is
the number of options. In the third step, the non-dimensionalized decision matrix calculates the
correlation coefficient between each pair of indicators. Spearman's method is commonly used to
calculate correlation when the values are ranked. The result of this step is a correlation matrix with
dimensions. Spearman's correlation coefficient is represented by Equation 12.
 Journal Pre-proof

n
6 d 2jh

of
j =1 (12)
rjh = 1 −
m(m 2 − 1)
Where is the rank difference between the jth and hth indexes. In the fourth step, the non-correlation
matrix is calculated. To calculate the non-correlation matrix, subtract 1 from the values of the
correlation matrix. In the fifth step, the sum of the columns of the non-correlation matrix is

pro
calculated. Furthermore, in the sixth step, the standard deviation value of each index is multiplied
by the sum of the corresponding column, as shown in Equation 13.

n
C j =  j   I jh (13)
h =1
Where is the value of the range of the jth row and the hth column of the non-correlation matrix.
Moreover, the higher the value, the higher the importance and weight of the relevant index. Finally,
re-
in the seventh step, the values are divided by their sum to reach the weight values.
3-3- TOPSIS

The TOPSIS technique, introduced by Huang and Yun in 1981 [46], is an efficient method in
multi-criteria decision-making(MCDM). The effectiveness of this technique relies on the distance
between the ideal positive and negative responses. The closer an option is to the ideal positive
lP
solution, and the further it is from the ideal negative answer, the more suitable it becomes. The
ideal solution is hypothetical and has the highest value among all options. The ideal negative
solution is the one that has the worst values in all possible scenarios. The essential information for
utilizing this technique includes the decision matrix and the criteria weights. The weights of the
criteria were determined using the CRITIC method. The decision matrix should be normalized
using the Euclidean method to implement this technique, as described in Equation 14.
rna

aij
rij =
J

a 2 (14)
ij
j =1

Where rij is the normalized value, aij is the value of the ith option in the jth criterion, and the
number of criteria is J. In the next step, the weighted normalized value is calculated by multiplying
the weight of each criterion by its corresponding normalized value, as described in Equation 15.
Jou

tij = rij  w j (15)

Where w j is the weight corresponding to the jth criterion. Afterward, it is essential to specify the
ideal positive and negative solutions should be specified. The ideal positive solution is to maximize
positive criteria and minimize negative criteria. The negative ideal solution is defined as the
highest value for negative criteria and the lowest value for positive criteria. After determining the
positive and negative ideal solutions, the distance between the options should be calculated using
Equations 16 and 17. These distances are denoted as d + and d − , respectively.
 Journal Pre-proof

 (t

of
d+ = ij − s +j ) 2 (16)
j =1

J
d− =  (t
j =1
ij − s −j ) 2 (17)

pro
Where s +j and s −j are respectively positive and negative values of the ideal solution in the jth
criterion. The next step is to calculate the relative proximity value for all options, which is done
using Equation 18.

d−
Ci = (18)
d+ + d−

Where is the degree of relative closeness to the ideal solution for the ith option. This numerical
re-
index ranges from 0 to 1, with a higher value indicating a closer proximity to the ideal solution.
Therefore, the options should be sorted in descending order based on their Ci value to rank them
according to this index.

4- Results and discussion

4-1- Study area
lP
The proposed method in this study has been implemented in a part of Tehran’s District 12.
Demographic block data, candidate ES, and the road network were used for this analysis. The
calculation of the center of the polygon for each population block assumed that the population is
concentrated at the center of each block. The demand centers were identified as these points. Figure
3 illustrates the study area and the data utilized in this study.
rna
Jou

Figure 3 – Study area

 Journal Pre-proof

According to Figure 3, 332 population centers with a population of 78,499 people have been
considered, along with 65 candidate centers for ESs. The ESs were obtained from the Tehran
Municipality website, which lists multiple ESs for each region. Various regional schools, mosques,

of
and parking lots have been identified as potential locations for ESs. Various criteria have been
considered when selecting ESs from potential centers. The criteria for determining location
suitability include proximity to fault lines, fire stations, hospitals, main road networks, area of Ess,
and population vulnerability, shown in Figure 4.

pro
)a( re- (b)
lP

)c( )d(
rna

)e( )f(
Jou

Figure 4 - The layers used include: (a) distance from the fault, (b) distance from fire stations, (c) distance
from the hospital, (d) distance from the main roads, (e) area, and (f) Population vulnerability

4-2- Ranking of candidate centers

The TOPSIS method was used to rank the candidate centers. Before implementing this method,
assessing the weight and importance of various criteria is essential. This study used a CRITIC to
assign weights to the layers. According to the CRITIC, a weight value was obtained for each layer,
as indicated in Table 2.
 Journal Pre-proof

Table 2- The weight of different layers for ranking candidate centers for ESs

Layer name Layer weight

of
Distance from the fault 0.18
Distance from fire stations 0.19
Distance from the hospital 0.19
Distance from the main roads 0.17
Area 0.12

pro
Population vulnerability 0.15

After determining the weight values of the layers, the TOPSIS method was used to rank the
potential centers. First, a decision matrix was created, and the values associated with each criterion
were normalized. Positive and negative ideals were identified based on their influence on decision-
making. The criteria for positive criteria include the distance from the fault, the area, and the
vulnerable population. The most suitable ES is determined by selecting the highest value for each
factor. Moreover, considering the minimum values for distance from the hospital, distance from
the main road, and distance from the fire station, these are considered positive ideals. The negative
re-
ideal is approached similarly, with the lowest value considered for the first three layers and the
highest value for the second three layers. The distance between each option and the positive and
negative ideals was calculated as d + and d − , respectively. Table 3 shows the distances and index
C values used for the final ranking.

Table 3- The weight of different layers for ranking candidate centers for ESs
lP
rank d+ d− C
1 0.000682 0.011469 0.943882
2 0.003817 0.007128 0.651266
3 0.003661 0.005286 0.590804
4 0.004069 0.004471 0.52351
.. .. .. ..
. . . .
63 0.010815 0.002185 0.168068
rna

64 0.012161 0.001905 0.135431

65 0.010301 0.001256 0.108704

As indicated in the table, the highest value of C was associated with the first rank, indicating that
it is the most similar to the positive ideal and the least similar to the negative ideal compared to
the other options. The value of C decreased over time and reached its lowest point at rank 65.
This ranking has been used to locate and allocate ESs.
Jou

4-3- Location - Allocation of ESs

In the previous stage, the CRITIC and TOPSIS methods were utilized to rank the candidate ESs,
resulting in a C value for each ES. In this section, which focuses on the location and allocation of
ESs, C values are used as weights for each potential center. The higher the weight of a center, the
more likely it is to be selected during the location process. For instance, the center with a rank of
one had a C value of 0.943882, whereas the center with a rank of two had a C value of 0.651266.
Therefore, the probability that rank one is chosen is 1.5 (0.943882/0.651266) times that of
choosing rank 2.
 Journal Pre-proof

The modeling for the location-allocation problem considers two parameters: the neighborhood
radius and the number of candidate ESs. To analyze the impact of these parameters on allocation
results, three different neighborhood radius values (500, 1000, and 1500 meters) were considered

of
for each candidate ES. It was assumed that only the population within each neighborhood could
be relocated to the ES. Additionally, the model was evaluated using a range of 10 to 65 candidate
ES as input. Figure 5 shows the changes in allocation percentage, total distance traveled by
citizens, and average occupancy percentage of the selected ESs based on the radius of the

pro
neighborhood and the number of ESs.

re-
)a(
lP
rna

)b(
Jou
 Journal Pre-proof

of
pro
)c(
Figure 5- Chart of changes of different allocation statistics based on the number of different centers and for
the coverage radius: (a) 500 meters, (b) 1000 meters, and (c) 1500 meters
re-
As depicted in the graphs in Figure 5, the percentage of population coverage has increased in all
scenarios with the increase in the number of ESs, which is expected. However, the changes were
not linear, and the rate of increase was higher at the beginning of the chart than at the end. At the
end of the corresponding diagram, there have been minor changes. Although the population is not
fully covered, the expansion of medical centers at the end of the graph has not impacted the
increase in coverage. The reason for this may be related to the size of the neighborhood.
lP
Specifically, as the radius of the neighborhood increases, the graph tends to converge more
quickly. In a situation where convergence has occurred, the appropriate distribution of centers is
much more important than the quantity of centers. This issue is consistent with the results of the
study [42] so that the authors also concluded that the increase in the number of ESs is insufficient
regardless of their distribution. Also, many studies have specifically addressed the importance of
ESs distribution like [26, 30, 31] . This justification can also be considered for the distance-traveled
rna

graph. Regarding the chart depicting the average occupancy percentage of the centers, it can be
observed that in all cases, there has been a decrease in value towards the end of the chart. This
trend may be attributed to the size of the neighborhood. Even though there are still vacancies in
the centers, the allocation has been prevented due to the neighborhood distance constraint, and the
center still needs to be filled. With the expansion of the neighborhood radius, the occupancy rate
has also risen, surpassing 90% when the radius reaches 1500 meters. Of course, it should be
considered that in critical situations such as an earthquake, the emergency centers should be
located close enough to ensure that citizens can access emergency services quickly. Assuming the
Jou

selection of 40 ESs, maps related to the allocation and non-allocation for the studied area have
been produced and evaluated. Figure 6 shows the allocation maps for the neighborhood distances
of 500, 1000, and 1500 meters and the number of 40 ESs.
 Journal Pre-proof

of
pro
re- )a(
lP
rna

)b(
Jou
 Journal Pre-proof

of
pro
)c(
re-
Figure 6- Allocation maps based on 40 ESs and per neighborhood radius: (a) 500 meters, (b) 1000 meters, and
(c) 1500 meters.

The allocation statistics for different age groups are presented in Table 4. This includes children
under six, men and women between 6 and 65, and men and women over 65. The statistics are
based on neighborhoods with a 500, 1000, and 1500 meter radius. Based on this, the age groups
lP
under six and over 65, who are highly vulnerable, have been prioritized for service when
considering a larger neighborhood radius. In addition, women received higher services (in
percentage terms) than men.

Table 4 - Allocation statistics for different age and gender groups

men women
Boy under girl under men over women
rna

between 6 between 6
coverage radius six years six years 65 years over 65
and 65 and 65
old old old years old
years old years old
person 993 952 9950 10246 1075 1088
500 m
Percent 32.24 32.29 30.54 31.81 32.13 32.82
person 1602 1552 15991 16086 1816 1812
1000 m
Percent 51.99 52.66 49.08 49.94 54.26 54.63
person 2066 1992 20598 20457 2359 2304
1500 m
Percent 67.06 67.54 63.22 63.51 70.52 69.48
Jou

4-4- Prediction of increased traffic

After solving the allocation problem for coverage distances of 500, 1000, and 1500 meters, the
allocation lines were used as network routes to predict the increase in traffic level after the
earthquake caused by the transfer of residents to ESs. Line density analysis was utilized to
generate the traffic map. This analysis focuses on the distribution of allocation lines on the road
network, particularly in areas with high residential properties. The results of the analysis provide
 Journal Pre-proof

a general idea of areas with traffic. Figure 7 displays maps illustrating the traffic ranges for
coverage distances of 500, 1000, and 1500 meters.

of
pro
re- (a)
lP
rna

(b)
Jou
 Journal Pre-proof

of
pro
(c)
re-
Figure 7- Prediction maps of areas with increased traffic for coverage distances: (a) 500 meters, (b) 1000
meters, and (c) 1500 meters

The traffic situation becomes more intense as the coverage distance for ESs increases, as shown
in Figure 7. It is expected that there will be minimal traffic near the ESs within a 500-meter
distance, while heavy traffic is not expected on the city's main roads. Additionally, traffic will
disperse quickly because these centers are situated close to their designated areas. The traffic is
lP
heavier at distances of 1000 and 1500 meters. The transportation network includes sections of
main roads serving other essential service departments, such as emergency and fire departments.
This can cause disruptions, especially since the operations of these services should not be disturbed
during that time.

5- Conclusion
rna

ESs are essential for promoting peace and ensuring the safety of individuals during earthquakes
and potential aftershocks. These centers' locating and coverage areas should be optimized to
prioritize citizen safety and maximize accessibility. This study aimed to identify and allocate ESs
in a part of Tehran's 12th district. Potential ESs were initially selected based on specific criteria,
including their proximity to fault lines, fire stations, hospitals, main roads, ES areas, and
population vulnerability. The criteria were weighted, and the centers were ranked using the
CRITIC and TOPSIS methods.
Jou

According to the results obtained using the CRITIC method, the criteria of distance from the
hospital and fire department had the highest weight (0.19). In contrast, the area criteria had the
lowest weight (0.12). The TOPSIS method has been used to rank the ESs by assigning weights to
the criteria. The ranked centers were utilized to address the issue of location and allocation. In the
location problem, priority was given to centers with higher ratings. The location and allocation
issues have been resolved for different numbers of ESs. Based on the results, the service delivery
radius is assumed to remain constant. Simply increasing the number of centers does not guarantee
better service delivery. The distribution is crucial. When the coverage radius was set to 500 meters,
 Journal Pre-proof

increasing the number of ESs decreased the occupancy percentage of the centers. This indicates
that the centers were not fully utilized because they were more than 500 meters away from the
residents. If the coverage radius is small, the distribution of centers becomes more significant. This

of
requires considering multiple centers in different locations. In cases where the radius was set at
1000 and 1500 meters, there was an increase in the percentage of service delivery and occupancy
of ESs. The service coverage percentage within a 1500-meter radius was only about 60%, well
below the desired value of 80%. It is possible to expand the distance to increase this case's

pro
coverage. It is important to note that as the distance from centers increases, quick and easy access
becomes more complex, which is crucial during times of crisis.

Line density analysis has been used to evaluate traffic conditions and predict areas with high traffic
volume caused by the transfer of residents to ESs in the studied area. According to this analysis,
reducing the service radius leads to decreased traffic, suggesting that services can be conducted
more efficiently. Emergency response after an earthquake is crucial. Another reason expanding
the coverage radius is not a viable solution for improving service delivery is the possibility of
traffic and disruptions that may occur when relocating emergency services during crises.
re-
Therefore, this research introduces the conflict between the number of people served and the
increase in traffic, which should be appropriately evaluated in the decision-making process
because although the increase in the radius of the neighborhood increases the service to citizens,
it can disrupt other services such as fire department, emergency, etc.

In this research, it is assumed that the population is concentrated in the center of population blocks,
and perhaps higher-resolution information can lead to better planning. However, due to the
lP
unavailability of building data separately, it has not been possible to implement. Also, due to the
unavailability of information on other urban uses, this study only considered residential use,
therefore it is assumed that the earthquake occurred at night when most of the residents were in
residential use. Also, if there is time information about the population in different blocks, it is
possible to simulate the population flow and solve the location-allocation problem based on that.
In this research, the types of structures related to ESs and their resistance to earthquakes are not
rna

considered as a weight component due to the unavailability of this information. Also, despite the
availability of structural information, it is possible to simulate the destruction of buildings and the
blocking of roads and the simultaneous evaluation of the effect of traffic and blocking of the road
network after an earthquake.

In future studies, researchers can address the problem by exploring different assumptions. For
instance, let us consider a situation where an earthquake strikes during the day, causing residents
to be scattered in various locations. Other techniques, like VIseKriterijumska Optimizacija I
Kompromisno Resenje(VIKOR), can be used to weigh and rank ESs and compare their results
Jou

with the method used in this study. Additionally, considering time factors when determining
location and allocation can result in a more accurate simulation of crisis conditions. This enables
better decision-making in the location and allocation of ESs. It is also essential to consider other
aspects of crisis management, such as allocating medical centers, firefighting, and rescue centers,
as these tasks and services are interconnected. This would help achieve the study's objectives.

References
 Journal Pre-proof

[1] K. Das, R. Lashkari, and N. Biswas, “Disaster Assessment and Mitigation Planning: A Humanitarian Logistics Based
Approach,” Industrial Engineering and Management Systems, vol. 12, no. 4, pp. 336-350, 2013.
[2] S. Mi, Y.-J. Piao, R. Zhang, and T.-D. Cho, “The Campus of University Earthquake Disaster Prevention Planning-The
Research of Spatial Pattern Based on GIS,” Journal of Environmental Science International, vol. 23, no. 7, pp. 1213-

of
1221, 2014.
[3] L. Zhao, H. Li, Y. Sun, R. Huang, Q. Hu, J. Wang, and F. Gao, “Planning ESs for urban disaster resilience: An integrated
location-allocation modeling approach,” Sustainability, vol. 9, no. 11, pp. 2098, 2017.
[4] V. Bayram, and H. Yaman, “ES location and evacuation route assignment under uncertainty: A benders decomposition
approach,” Transportation science, vol. 52, no. 2, pp. 416-436, 2018.
[5] A. Frimpong, “Sheltering and housing recovery after disasters: Dissecting the problems of policy implementation and

pro
possible solutions,” International Journal of Humanities and Social Science, vol. 1, no. 20, pp. 1-12, 2011.
[6] N. Aras, M. Orbay, and I. K. Altinel, “Efficient heuristics for the rectilinear distance capacitated multi-facility Weber
problem,” Journal of the Operational Research Society, vol. 59, no. 1, pp. 64-79, 2008.
[7] J.-P. Arnaout, “Ant colony optimization algorithm for the Euclidean location-allocation problem with unknown number
of facilities,” Journal of Intelligent Manufacturing, vol. 24, no. 1, pp. 45-54, 2013.
[8] Y. Shi, G. Zhai, L. Xu, Q. Zhu, and J. Deng, “Planning ESs for Urban Disasters: A Multi-Level Location–Allocation
Modeling Approach,” Sustainability, vol. 11, no. 16, pp. 4285, 2019.
[9] C.-A. Tai, Y.-L. Lee, and C.-Y. Lin, "Earthquake disaster prevention area planning considering residents' demand." pp.
381-385.
[10] C. Toregas, R. Swain, C. ReVelle, and L. Bergman, “The location of emergency service facilities,” Operations research,
vol. 19, no. 6, pp. 1363-1373, 1971.
[11]

[12]

[13]
re-
M. Ye, J. Wang, J. Huang, S. Xu, and Z. Chen, “Methodology and its application for community-scale evacuation
planning against earthquake disaster,” Natural hazards, vol. 61, no. 3, pp. 881-892, 2012.
F. Hu, W. Xu, and X. Li, “A modified particle swarm optimization algorithm for optimal allocation of earthquake ESs,”
International Journal of Geographical Information Science, vol. 26, no. 9, pp. 1643-1666, 2012.
V. Bayram, B. Ç. Tansel, and H. Yaman, “Compromising system and user interests in ES location and evacuation
planning,” Transportation research part B: methodological, vol. 72, pp. 146-163, 2015.
[14] F. Kılcı, B. Y. Kara, and B. Bozkaya, “Locating temporary ES areas after an earthquake: A case for Turkey,” European
journal of operational research, vol. 243, no. 1, pp. 323-332, 2015.
[15] B. Saeidian, M. S. Mesgari, B. Pradhan, and M. Ghodousi, “Optimized location-allocation of earthquake relief centers
lP
using PSO and ACO, complemented by GIS, clustering, and TOPSIS,” ISPRS International Journal of Geo-Information,
vol. 7, no. 8, pp. 292, 2018.
[16] G.-H. Tzeng, H.-J. Cheng, and T. D. Huang, “Multi-objective optimal planning for designing relief delivery systems,”
Transportation Research Part E: Logistics and Transportation Review, vol. 43, no. 6, pp. 673-686, 2007.
[17] F. Hu, S. Yang, and W. Xu, “A non-dominated sorting genetic algorithm for the location and districting planning of
earthquake ESs,” International Journal of Geographical Information Science, vol. 28, no. 7, pp. 1482-1501, 2014.
[18] X. Zhao, W. Xu, Y. Ma, and F. Hu, “Scenario-based multi-objective optimum allocation model for earthquake ESs using
a modified particle swarm optimization algorithm: a case study in Chaoyang District, Beijing, China,” PloS one, vol. 10,
no. 12, pp. e0144455, 2015.
rna

[19] X. Zhao, W. Xu, Y. Ma, L. Qin, J. Zhang, and Y. Wang, “Relationships between evacuation population size, earthquake
emergency ES capacity, and evacuation time,” International Journal of Disaster Risk Science, vol. 8, no. 4, pp. 457-470,
2017.
[20] Polat, E.G., Distribution Centre Location Selection for Disaster Logistics with Integrated Goal Programming-AHP based
TOPSIS Method at the City Level. Afet ve Risk Dergisi, 2022. 5(1): p. 282-296.
[21] D. Darvishi, H. Ahmadi Choukolaei, and S. Shafaee, “Assessing the location of relief centers using a combination of
multi-criteria decision-making methods and GIS (case study: district 18 of Tehran),” International journal of research
in industrial engineering, vol. 11, no. 1, pp. 16-29, 2022.
[22] F. Haghpanah, and H. Foroughi, “Optimal shelter location-allocation during evacuation with uncertainties: a scenario-
based approach,” arXiv preprint arXiv:1802.05775, 2018.
[23] B. Saeidian, M. S. Mesgari, B. Pradhan, and M. Ghodousi, “Optimized location-allocation of earthquake relief centers
Jou

using PSO and ACO, complemented by GIS, clustering, and TOPSIS,” ISPRS international journal of geo-information,
vol. 7, no. 8, pp. 292, 2018.
[24] W. Xu, X. Zhao, Y. Ma, Y. Li, L. Qin, Y. Wang, and J. Du, “A multi-objective optimization based method for
evaluating earthquake shelter location–allocation,” Geomatics, Natural Hazards and Risk, vol. 9, no. 1, pp. 662-677,
2018.
[25] Y. Ma, W. Xu, L. Qin, X. Zhao, and J. Du, “Emergency shelters location-allocation problem concerning uncertainty
and limited resources: a multi-objective optimization with a case study in the Central area of Beijing, China,”
Geomatics, Natural Hazards and Risk, 2019.
[26] M. Rahman, N. Chen, M. M. Islam, A. Dewan, H. R. Pourghasemi, R. M. A. Washakh, N. Nepal, S. Tian, H. Faiz, and
M. Alam, “Location-allocation modeling for emergency evacuation planning with GIS and remote sensing: A case
study of Northeast Bangladesh,” Geoscience frontiers, vol. 12, no. 3, pp. 101095, 2021.
 Journal Pre-proof

[27] P. Praneetpholkrang, and S. Kanjanawattana, “A multi-objective optimization model for shelter location-allocation in
response to humanitarian relief logistics,” The Asian Journal of Shipping and Logistics, vol. 37, no. 2, pp. 149-156,
2021.
M. J. Alam, M. A. Habib, and E. Pothier, “Shelter locations in evacuation: A Multiple Criteria Evaluation combined

of
[28]
with flood risk and traffic microsimulation modeling,” International Journal of Disaster Risk Reduction, vol. 53, pp.
102016, 2021.
[29] P. Praneetpholkrang, and S. Kanjanawattana, “A Novel Approach for Determining Shelter Location-Allocation in
Humanitarian Relief Logistics,” International Journal of Knowledge and Systems Science (IJKSS), vol. 12, no. 2, pp.
52-68, 2021.
[30] W. Chen, Y. Shi, W. Wang, W. Li, and C. Wu, “The spatial optimization of emergency shelters based on an urban-

pro
scale evacuation simulation,” Applied Sciences, vol. 11, no. 24, pp. 11909, 2021.
[31] S. Geng, H. Hou, and Z. Zhou, “A hybrid approach of VIKOR and bi-objective decision model for emergency shelter
location–allocation to respond to earthquakes,” Mathematics, vol. 9, no. 16, pp. 1897, 2021.
[32] Y. Wang, and Z. Xu, “A multi-objective location decision making model for emergency shelters giving priority to
subjective evaluation of residents,” International Journal of Computers Communications & Control, vol. 17, no. 4,
2022.
[33] X. Ren, and J. Tan, “Location allocation collaborative optimization of emergency temporary distribution center under
uncertainties,” Mathematical Problems in Engineering, vol. 2022, 2022.
[34] L. He, and Z. Xie, “Optimization of urban shelter locations using bi-level multi-objective location-allocation model,”
International Journal of Environmental Research and Public Health, vol. 19, no. 7, pp. 4401, 2022.
[35] G. Zhong, G. Zhai, and W. Chen, “Optimization of Shelter Location Based on a Combined Static/Dynamic Two-Stage

[36]

[37]
re-
Optimization Methodology: A Case Study in the Central Urban Area of Xinyi City, China,” ISPRS International
Journal of Geo-Information, vol. 11, no. 4, pp. 262, 2022.
Y. Ma, B. Liu, K. Zhang, and Y. Yang, “Incorporating multi-criteria suitability evaluation into multi-objective
location–allocation optimization comparison for earthquake emergency shelters,” Geomatics, Natural Hazards and
Risk, vol. 13, no. 1, pp. 2333-2355, 2022.
L. Eriskin, and M. Karatas, “Applying robust optimization to the shelter location–allocation problem: a case study for
Istanbul,” Annals of Operations Research, pp. 1-47, 2022.
[38] S. Aghaie, and B. Karimi, “Location-allocation-routing for emergency shelters based on geographical information
system (ArcGIS) by NSGA-II (case study: Earthquake occurrence in Tehran (District-1)),” Socio-Economic Planning
lP
Sciences, vol. 84, pp. 101420, 2022.
[39] W. Zhu, H. Xing, and W. Kang, “Spatial Layout Planning of Urban Emergency Shelter Based on Sustainable Disaster
Reduction,” International journal of environmental research and public health, vol. 20, no. 3, pp. 2127, 2023.
[40] S. Bera, K. Gnyawali, K. Dahal, R. Melo, M. Li-Juan, B. Guru, and G. Ramana, “Assessment of shelter location-
allocation for multi-hazard emergency evacuation,” International Journal of Disaster Risk Reduction, vol. 84, pp.
103435, 2023.
[41] G. Zhong, Y. Lu, W. Chen, and G. Zhai, “Multi-objective optimization approach of shelter location with maximum
equity: an empirical study in Xin Jiekou district of Nanjing, China,” Geomatics, Natural Hazards and Risk, vol. 14, no.
1, pp. 2165973, 2023.
rna

[42] Z. Zhang, Y. Hu, W. Lu, W. Cao, and X. Gao, “Spatial accessibility analysis and location optimization of emergency
shelters in Deyang,” Geomatics, Natural Hazards and Risk, vol. 14, no. 1, pp. 2213809, 2023.
[43] A. R. Bakhshi Lomer, M. Rezaeian, H. Rezaei, A. Lorestani, N. Mijani, M. Mahdad, A. Raeisi, and J. J. Arsanjani,
“Optimizing emergency shelter selection in earthquakes using a risk-driven large group decision-making support
system,” Sustainability, vol. 15, no. 5, pp. 4019, 2023.
[44] F. Dehghan Farouji, and Z. Bahramian, “Vulnerability Zoning of Tehran in Natural Disasters with respect of
Population Groups,” Disaster Prevention and Management Knowledge (quarterly), vol. 10, no. 4, pp. 362-351, 2021.

[45] D. Diakoulaki, G. Mavrotas, and L. Papayannakis, “Determining objective weights in multiple criteria problems: The
critic method,” Computers & Operations Research, vol. 22, no. 7, pp. 763-770, 1995.
[46] C.-L. Hwang, K. Yoon, C.-L. Hwang, and K. Yoon, “Methods for multiple attribute decision making,” Multiple attribute
Jou

decision making: methods and applications a state-of-the-art survey, pp. 58-191, 1981.
 Journal Pre-proof

Highlights
 Present an integrated multi-criteria decision-making and network-based model to solve the

of
location-allocation problem.
 Consider population vulnerability based on age, gender and marital status in order to weigh
emergency shelters.
 Solve the location-allocation problem by considering weights for emergency shelters
 Study the relationship between the coverage radius of emergency shelters and the number of

pro
citizens served.
 Investigate the possibility of blockage and traffic in case of increasing coverage radius of the
emergency shelters.

re-
lP
rna
Jou
 Journal Pre-proof

Declaratio if ioterettt

of
☒ The authors declare that they have no known competng fnancial interests or personal relatonships
that could have appeared to infuence the work reported in this paper.

☐ The authors declare the following fnancial interests/personal relatonships which may be considered
as potental competng interests:

pro
re-
lP
rna
Jou