A Spatial Decision Support System for Modeling Urban Resilience to Natural Hazards

Abstract

A major component of urban management is studying and evaluating urban resilience in order to minimize the effects of natural hazards. This is because of the increasing number of natural hazards occurring worldwide. A spatial decision support system is presented for modeling urban resilience and selecting resilient zones in response to natural hazards. This system is implemented based on 22 criteria, grouped into three categories: demographics, infrastructure, and environmental. The criteria are then standardized using minimum and maximum methods, and their importance is determined by the analytical hierarchy process (AHP). The resilience maps in various scenarios are prepared using the ordered weighted average (OWA) method. Flow accumulation (distance from fault), vulnerable population density (vulnerable population density), and distance from road network (material type) were regarded as the most important criteria for flood resilience (earthquake resilience) from environmental, demographic, and infrastructure criteria, respectively. There are different areas that are considered to have very low resilience depending on the risk attitude. According a pessimistic scenario, 1% of Tehran’s area has very low resilience, while according to an optimistic scenario, 38% has very low resilience. This system can be used by urban planners and policymakers for the purpose of improving resilience to natural hazards in low-resilience areas.

1. Introduction

Over the past few decades, natural hazards have become increasingly common and their destructive effects (both economic and human) have increased dramatically [1]. Global statistics indicate that 40% of social and economic damage is caused by natural hazards each year [2]. There are several types of natural hazard that occur throughout the world, including earthquakes, landslides, floods, droughts, fires, tornadoes and severe storms [3]. In the period from 2006 to 2021, natural hazards affected approximately 271 million people per year. They caused approximately 70,000 deaths and $135 billion in damage around the world [4]. As a result of inadequate infrastructure, social injustices, physical growth, population density, and inappropriate spatial planning, urban areas are often prone to natural hazard’s risks [5]. Currently over half of the world’s population lives in cities. More than 70% of the world’s population is expected to live in urban areas by 2050 [6]. In light of this, it becomes increasingly necessary to pay particular attention to the reduction of risk and to urban adaptation.

Natural hazards such as earthquakes and floods are the most common. Earthquakes are considered one of the most devastating and unforeseen natural hazards in terms of their effects and risks regarding human life [7]. In comparison to other types of environmental disasters, earthquakes rank second in terms of their effects. In the last 90 years, Iran, which is considered a country prone to earthquakes, has experienced 18 earthquakes with a magnitude greater than 7 Richter, resulting in severe economic and social damage and high deaths. There were an estimated 40,000 to 50,000 deaths from the 1990 northern Iran earthquake, one of the most destructive earthquakes in Iran [8]. Similarly, floods are also a serious natural hazard which are mainly caused by changes in land use, vegetation removal, soil erosion, drainage system limitations, and the occupation of floodplains or flood-prone areas [9]. It is anticipated that the number of people who will be affected by river floods worldwide will increase to 54 million by 2030 as a result of socio-economic development and climate change [10]. Therefore, it is important to take proactive steps to mitigate the risk of floods to protect communities from potential devastation.

Increasing resilience is one of the most appropriate measures for preparing cities for these hazards. Resilience is defined by Knaapen et al. [11] and Dimelli [12] as the ability of a system or group at risk to adapt and recover from the effects of risk through the maintenance and restoration of fundamental functions and structures. Since the majority of resilience studies primarily focus on physical conditions, it has been recognized that a variety of aspects (economic, social, educational and cultural) can contribute to physical damage [13]. Recent developments have led to the emergence of an alternative approach that takes a social perspective. This approach takes into account the vulnerability of individuals and communities, as well as their economic, social, and cultural capabilities [14]. According to Bruno et al. [15], resilience refers to the ability of social units to reduce the risks associated with natural hazards. They identified three aspects of resilience: the ability to reduce the consequences of failure, the ability to reduce the probability of failure, and the ability to reduce recovery time. As cities continue to grow and become denser, as well as facing threats from natural hazards, it is essential to pay attention to and improve resilience within cities. This should be considered an essential component of urban management as a means of reducing disaster risk. In the event of natural disasters or a lack of resources, urban environments with a high level of resilience will not suffer from disruption. In the context of natural hazards, spatial analysis can provide valuable information to managers, municipalities, and urban planners regarding potential risks. As a result of this information, urban environments can reduce the potential negative social and economic impacts of future natural disasters.

The preparation of resilience maps can be improved by integrating multi-criteria decision making (MCDA) with geographic information systems (GIS), as demonstrated in previous studies [16,17,18,19,20]. While GIS is a tool for analyzing, editing, and managing spatial data, multi-criteria decision-making methods involve combining and transforming criteria maps and decision-makers’ priorities to obtain valuable decision-making information [21,22]. To provide useful information for spatial decision-making, the GIS-MCDA method combines map criteria with values obtained by decision-makers’ judgment [23,24,25]. The combination of GIS and MCDA has been used in some studies to prepare resilience maps. Tayyab et al. [26] employed a hierarchical model combined with GIS to assess urban flood resilience in Peshawar, Pakistan. Kaaviya & Devadas [27] used MCDA and GIS to determine water resilience in India in their study. In order to develop a resilience map, they considered a variety of topographic, physiological, environmental, ecological, and infrastructural factors. He et al. [28] evaluated fire resilience in Yuxi city, China, based on the analytic network process (ANP) and the preference ranking organization method for enrichment evaluations (PROMETHEEE II). Doorga et al. [16] described a methodology for flood resilience assessment in Port Louis, Mauritius based on the analytic hierarchy process (AHP). Their strategies included building an underground water storage tank in the Champ de Mars, creating flood refugee camps in Vallee Pitot and Cite La Cure, revising building and urban planning regulations, replacing the Caudan underpass with a flyover, and building a diversion tunnel to connect the Saint Louis River to the Grand River to strengthen flood resilience.

One of the MCDA methods is ordered weighted average (OWA), which is a multivariable hybrid method. In the OWA method, the decision makers’ priorities and subjective evaluations are considered [29]. By changing the parameters and criteria, different maps and scenarios can be predicted [30]. As part of a decision-making problem, risk-averse individuals emphasize the inappropriate features of an option and risk-taking individuals emphasize the appropriate features and use them as a criterion for their selection [31]. As a result of this method, it is possible to calculate the level of risk taking and risk aversion, which can then be used to determine the final option. For the past few decades, OWA-based GIS has been utilized for land use analysis such as health care [32], residential-quantity assessment [33], natural-based tourism [34], parking [35], and urban traffic management [36]. GIS-OWA integration has been used for a variety of socio-economic applications, including environmental monitoring [37], natural hazards [10,38], renewable energy [22,39], water resource management [40], and landfill location [41].

Based on prior studies, it is evident that only one natural hazard was considered in the preparation of each resilience map. Many studies around the world have simultaneously investigated several natural hazards in recent years. For example, landslides, erosion, earthquakes, and floods in Greece [42], earthquakes and landslides in India [43], climate hazards in Chile [44], and floods, avalanches, and rock falls in Iran [45]. Depending on the characteristics of each risk and their interactions, one risk may trigger the occurrence of another risk. Tehran, the capital of Iran, has numerous faults around and within it, worn-out areas, a high population density, non-earthquake resistant buildings, and unbalanced physical expansion, all of which contribute to the selection of Tehran as a study area. On the other hand, the steep slopes in the east and north of Tehran, the reduced permeability of the soil, the extent of the city from 700 to 2200 m and the absence of an efficient sewage system are the primary reasons for which Tehran has been chosen as a study area for reducing and managing flood risks. Thus, the main goal of this study is to evaluate the spatial resilience of Tehran city against flood and earthquake risks separately and in combination with each other. For this purpose, a spatial multi-criteria decision-making system with the ability to evaluate resilience based on different mental attitudes is used. Policy makers and planners can take advantage of combined resilience maps in order to develop effective geo-environmental programs and manage resources in a sustainable approach. In addition, according to the capabilities of the spatial multi-criteria decision-making system used, the results of this research can be practical and useful for different stakeholders in different decision-making situations.

2. Materials and Methods

2.1. Study Area and Data Collection

The capital and largest city of Iran, Tehran, is located near the Alborz mountain range in the north of the country (Figure 1). Its area is 730 square kilometers, ranking 27th on the list of the world’s largest cities [46]. In this city, the population fluctuates during the day and at night. Thus, it has a population of over 13 million during the day and over 9 million at night. In other words, it consists of 5.17% of the total population of the country [47]. Compared to the previous decade, Tehran’s population has grown by 1.4%, which indicates an increasing trend. Due to the fact that most of the rivers and canals in Tehran are directed towards the urban area floods may occur in the city of Tehran if heavy rain occurs. Moreover, this city is located on a number of faults, including the North Ray Fault, the North Tehran Fault, and the Mosha Fault [48].

The data used in previous studies differ due to various factors such as availability, geographic location of the study area, and the opinion of relevant experts. This study utilized two spatial databases (remote sensing products and organization-derived data) for the development of flood and earthquake resilience maps. The effective criteria for flood and earthquake resilience are categorized into three main groups: environmental, demographic, and infrastructure. Based on the capabilities of each layer at each stage, ArcGIS 10.3 software was used to produce and analyze information layers and prepare maps. The details of the data used are presented in

Table 1. General characteristics of the data used in this study.

Criterion	Sub-Criterion	Hazard Type	Format	Source
Environmental	Elevation	Flood/Earthquake	Raster	https://earthexplorer.usgs.gov/ (accessed on 25 December 2022)
	Slope	Flood/Earthquake	Raster	Extracted from the Digital Elevation Model (DEM)
	Aspect	Flood	Raster	Extracted from the DEM
	Stream network	Flood	Vector	https://frw.ir/ (accessed on 18 November 2022)
	Flow accumulation	Flood	Raster	Extracted from the DEM
	Rainfall	Flood	Raster	https://wapor.apps.fao.org/ (accessed on 25 November 2022)
	Vegetation density	Flood	Raster	https://earthexplorer.usgs.gov/ (accessed on 25 November 2022)
	Impervious surfaces	Flood	Raster	https://earthexplorer.usgs.gov/ (accessed on 25 November 2022)
	Fault	Earthquake	Vector	https://gsi.ir/ (accessed on 18 November 2022)
Demographic	Total population density	Flood/Earthquake	Vector	https://www.tehran.ir/ (accessed on 18 December 2022)
	Vulnerable population density	Flood/Earthquake	Vector	https://www.tehran.ir/ (accessed on 18 December 2022)
	Working population density	Flood/Earthquake	Vector	https://www.tehran.ir/ (accessed on 18 December 2022)
	Literate population density	Flood/Earthquake	Vector	https://www.tehran.ir/ (accessed on 18 December 2022)
Infrastructure	Power transmission lines	Flood/Earthquake	Vector	https://fa.ncc.gov.ir/ (accessed on 18 November 2022)
	Fuel station	Flood/Earthquake	Vector	https://www.tehran.ir/ (accessed on 18 November 2022)
	Fire station	Flood/Earthquake	Vector	https://www.tehran.ir/ (accessed on 18 November 2022)
	Pharmacy	Flood/Earthquake	Vector	https://www.tehran.ir/ (accessed on 18 November 2022)
	Medical centers	Flood/Earthquake	Vector	https://www.tehran.ir/ (accessed on 05 November 2022)
	Road network	Flood/Earthquake	Vector	https://fa.ncc.gov.ir/ (accessed on 05 November 2022)
	Material type	Earthquake	Vector	https://www.tehran.ir/ (accessed on 05 November 2022)
	Skeleton type	Earthquake	Vector	https://www.tehran.ir/ (accessed on 05 November 2022)
	Public transport station	Flood/Earthquake	Vector	https://www.tehran.ir/ (accessed on 05 November 2022)

.

There were 82 academic and organizational specialists included in the statistical population for this study. For flood resilience, 42 experts have provided their opinions (including experts in flood engineering and natural disasters, urban management, natural resources engineering, and GIS analysts) and for earthquake resilience, 40 experts have provided their opinions (including experts in earthquake engineering and natural disasters, structural engineering, urban management, urban planners, and GIS analysts). Out of 82 experts, 12 experts have bachelor’s education, 30 experts have master’s education, and 40 experts have doctorate education and above. Additionally, their age range is between 28 and 60 years.

2.2. Methodology

As shown in Figure 2, the research method in this study consists of 5 steps. The first step of the analysis is to identify the most relevant factors for each risk based on the literature review and experts’ opinions, and then collect and pre-process the data. In the second step, maps of environmental, demographic, and infrastructure criteria will be prepared using spatial analysis. The third step involves standardizing the criteria and calculating the weights and importance of the criteria by applying the AHP method. In the fourth step, the output of step (3) was entered into the OWA model and resilience maps were prepared for floods and earthquakes in each scenario. As a fifth step, flood and earthquake resilience maps were combined.

2.2.1. Standardization of Criteria

As a result of the different nature of factors as well as the inconsistency of units and dimensions, they cannot be used directly in the modeling process. [49]. The “slope” criterion, for example, is expressed as a percentage and the “distance from road network” criterion is expressed as km. In order to be evaluated in comparison with each other, the layers must be converted into comparable units. Equation (1) is used in the case of criteria whose high values decrease resilience (such as elevation), whereas Equation (2) is used in the case of criteria whose low values decrease resilience (such as distance from fault).

$$Y_i=\frac{y_i-y_{\min }}{y_{\max }-y_{\min }}$$

(1)

$$Y_i=\frac{y_{\max }-y_i}{y_{\max }-y_{\min }}$$

(2)

where, Y_i is the standardized value of criterion i, y_i is the initial value of criterion i, y_max is the maximum value of criterion i, and y_min is the minimum value of criterion i [50].

2.2.2. AHP

The AHP method, which was invented by Saaty [51], is one of the most popular methods of MCDA. Using AHP makes it possible to combine qualitative and quantitative criteria when faced with several options and decision-making factors [52]. By comparing alternatives and decision criteria in pairs or pairwise, this method enables us to combine qualitative and quantitative criteria simultaneously [53]. Due to this, decision makers can focus solely on comparing two criteria or options without external interference. Furthermore, the pairwise comparisons provide valuable information for the investigated problems and rationalize the decision-making process since the respondent only compares two factors [54]. The AHP method generally includes the following steps [55]:

• Define the problem and express its goals in a clear manner (hierarchical tree design).
• Separate the defined problem into decision elements (criteria and options).
• Form a decision-making matrix using pairwise comparisons.
• Use the eigenvalue method to estimate the relative weight of decision-making criteria.
• Calculate the compatibility rate of matrices to ensure the compatibility of decision-makers’ judgments.
• Collect the decision weights for the overall ranking of the criteria.

For pairwise comparisons, Saaty and Vargas [56] define a suitable measurement scale. Verbal judgments are expressed as a degree of importance: equal importance = 1, somewhat more importance = 3, much more importance = 5, very much more importance = 7, absolutely more importance = 9, and intermediate values = 2, 4, 6, and 8. The inverse relationship is expressed by the reciprocal of these numbers. The compatibility index (CI) is calculated in Equation (3).

$$\mathrm{CI}=\frac{{\lambda }_{\max }-n}{n-1}.$$

(3)

where, n is the number of pairwise comparison matrix criteria and ${\lambda }_{\max }$ is the greatest eigenvalue of pairwise comparison matrix. The consistency index of a randomly generated reciprocal matrix is known as the random index (RI). The average RI value for matrices is between 1 and 15. When the consistency ratio (CR) is less than or equal to 0.1, the system is considered to have acceptable consistency; when the CR is greater than 0.1, decision makers should review their judgments. After this, CR is calculated using Equation (4).

$$\mathrm{CR}=\frac{\mathrm{CI}}{\mathrm{RI}}.$$

(4)

2.2.3. OWA

In the OWA method, different operators and sets of ordered weights can be used to apply to a wide range of multi-criteria evaluation strategies [57]. A major objective of the OWA method is to provide decision makers with a quantitative assessment of a range of very optimistic to very pessimistic scenarios. Thus, a variety of risk attitudes among decision makers will be assessed as a factor influencing the final result [58]. Ranking criteria and dealing with the uncertainty resulting from their interaction is the purpose of OWA [59]. As a result of the OWA method, a continuous ranking of scenarios occurs between an intersect operator (risk-averse) and a union operator (risk-taking) [60]. This method is similar to the WLC method but includes two sets of weights. The first set of weights controls the relative contribution of a specific criterion, while the second set of weights controls the aggregation rank of the weighted criteria [61]. The advantage of the OWA method is that the decision-maker is able to produce a wide range of maps and different solutions and scenarios by rearranging and changing criteria [62]. In contrast to Boolean logic overlap, the intersect operator (AND) indicates a low risk of decision making and the union operator (OR) indicates a high risk of decision making [63]. Using this method, a wide range of risk scenarios can be prepared between community and shared operators; this is described using Equation (5).

$$\mathrm{OWA}={\displaystyle \sum }_{j=1}^n\left\{\frac{w_j{v}_j}{{{\displaystyle \sum }}_{j=1}^n{w}_j{v}_j}\right\}{Z}_{ij}$$

(5)

where, v_j is the order weight, w_j is the criterion weight, and Z_ij is the attribute value of the j criterion of the i pixel that has been standardized and hierarchically assigned [64].

3. Results and Discussion

Based on the matrix of pairwise comparisons, Figure 3a,b illustrates the weight given to each of the effective flood and earthquake resilience criteria, respectively. Among the environmental, demographic, and infrastructure criteria for flood resilience, flow accumulation (0.21), vulnerable population density (0.35), and distance from road network (0.23) have the highest weight and aspect (0.03), literate population density (0.15), and distance from public transportation stations (0.04) have the lowest weight. Furthermore, among the environmental, demographic, and infrastructural criteria for earthquake resilience, the distance from fault (0.47), the density of vulnerable population (0.32), and the material type (0.21) have the highest weight and elevation (0.18) and distance from public transportation stations (0.01) have the lowest weight. The weight of environmental, demographic and infrastructure criteria for preparing flood (earthquake) resilience maps is 0.33 (0.35), 0.45 (0.45), and 0.22 (0.20), respectively. The consistency rate for the pairwise comparison matrix was less than 0.1, indicating consistency and reliability among the experts. To achieve the CR value, the evaluation process of experts’ opinions was repeated three times.

Related spatial analyses have been used to prepare criterion maps. For example, the “Euclidean distance” spatial analysis was used to produce the “distance from fault” map. After preparing the criterion map, the minimum and maximum standardization method was employed to scale the criterion values (Figure 4, Figure 5 and Figure 6).

The minimum standardization method was used for distance from stream network, vegetation density, working population density, literate population density, distance from fuel stations, distance from power transmission lines, and distance from faults. Meanwhile, the maximum standardization method was used for distance from fire station, distance from medical centers, distance from pharmacy, distance from road network, distance from public transport station, aspect, total population density, vulnerable population density, impervious surfaces, precipitation, materials, and skeleton type. All criteria have standardized values between 0 and 1. Therefore, for each criterion, 0 indicates low resilience and 1 indicates high resilience.

The standardized maps, along with the weights extracted using the AHP method and the order weights, were defined as inputs to the OWA model. A flood and earthquake resilience map with environmental, demographic and infrastructure dimensions (Figure 7 and Figure 8), was prepared as a result of this study. Based on standardized values (between 0 and 1), resilience maps were divided into five classes: 0–0.2 (high resilience), 0.2–0.4 (high resilience), 0.4–0.6 (moderate resilience), 0.6–0.8 (low resilience), and 0.8–1 (very low resilience). In multiple dimensions of resilience maps, it was observed that areas with very low resilience increased as the degree of optimism increased, while areas with very high resilience decreased.

Maps of flood resilience in the environmental dimension indicate that the majority of areas with low resilience are found in the north and northeast. The most critical reason is the high elevation and slope of the northern regions as well as the concentration of rivers in the northeast of Tehran. Conversely, areas with high resilience are strategically located in the center and south of Tehran. In the maps of earthquake resilience in the environmental dimension, the northern and southern areas of Tehran were placed in the class of very low resilience. This is due to the fact that these areas are the closest to fault lines. As a result of the demographic dimension, Tehran’s central, southern, and eastern areas display very low resilience to both floods and earthquakes. The most significant reason is the migration of people from other provinces to these areas due to the low price of rent and land. These areas also have a high density of commercial and economic activities. In the infrastructural dimension of resilience maps, it is evident that the suburbs of the city, particularly in the northwest and southeast regions, have low resilience to floods and earthquakes. These areas are far from medical centers and fire stations. They are also characterized by an excessive level of urban decay.

By generating different scenarios, the OWA method allows planners and managers to make decisions based on conditions (financial, time, and environment). Planners and managers typically have three types of attitudes: risk-taking attitude, neutral attitude, and risk-averse attitude. Planners and managers with a risk-taking attitude are more inclined towards an optimistic scenario. In this view, areas with low resilience encompass a significant portion of the study area, and there is no limit to financial resources available. In contrast, when planners and managers are faced with financial and time constraints, they tend to consider a pessimistic scenario where decision making is low risk.

For each dimension (environmental, demographic, and infrastructure) of flood and earthquake resilience, Figure 9 provides the percentage of area of different classes in pessimistic, neutral, and optimistic scenarios. Increasing optimism leads to an increase in the percentage of classes with low resilience and a decrease in the percentage of classes with high resilience. In other words, there is an inverse relationship between the area of the low resilience class and an increase in the optimistic scenario. Generally, under pessimistic decision making, the largest area was for the class with high and very high resilience. In the conditions of neutral decision-making, the largest area was for the class with medium resilience. A class with low and very low resilience occupied the largest area under optimistic decision making. It is estimated that for the flood resilience map in the environmental dimension of the pessimistic scenario, the area of the very low, low, moderate, high, and very high classes is 0.6, 3.7, 7.8, 49.9, and 38.1 percent, respectively. In the neutral scenario, it was 8.1, 20.3, 30.2, 20.5, and 20.9 percent, and in the optimistic scenario, it was 36.4, 35.6, 16, 9.2, and 2.8 percent, respectively.

A pessimistic scenario is more suited to risk-averse planners and managers, since, in this scenario, only a few areas of the study area are selected as being of very low and low resilience. However, these areas have inadequate conditions for all criteria, and managers at these locations are more likely to achieve optimal results. A risk-neutral planner and manager will utilize a balanced decision-making approach in retrofitting and reconstruction projects, finding credit allocation options that are compatible with all effective criteria. There is a 50% probability that the efficiency of these places will reach its expected level in the hazard occurrence. As the standards are average, there should be an increase in quality and safety in these places. In risk-taking planning and management, optimistic scenarios are more common. Under this scenario, even if an area fails to meet optimal conditions for criteria, it is placed very low and very low in the resilient class. As a consequence, a substantial part of the study area with minimal quality is placed in the low and very low resilience class. Therefore, this scenario should only be considered as a last resort.

Individual resilience maps (i.e., floods and earthquakes), created using the OWA method, were used to produce a composite multi-risk resilience map by combining two different risk maps (in environmental, demographic, and infrastructure dimensions). In the end, five resilience classes were identified (Figure 10), ranging from very low resilience to very high resilience. Based on a combination of risk maps, it is apparent that northern regions have a very low level of resilience to flood and earthquake risks when compared to other regions. A considerable area of the central region also falls into this class.

The percentage of the area of different classes of resilience resulting from the combination of risks is shown in

Table 2. The percentage of the area that belongs to different resilience classes.

Scenarios	Very Low	Low	Moderate	High	Very High
Pessimistic	1	8	18	31	42
Neutral	12	17	24	24	23
Optimistic	38	26	17	12	7

. Based on the results of the multi-hazard map prepared for the city of Tehran, it is evident that under pessimistic, neutral, and optimistic scenarios, 0.73, 0.47, and 0.29 percent of the total area is classified as the high and very high resilience classes, respectively. These classes are considered to be the safest against floods and earthquakes. In addition, the area of classes with low and very low resilience in pessimistic, neutral, and optimistic scenarios is 0.9, 0.29, and 0.45 of the total area of the study area, respectively. As a result of pessimistic, neutral, and optimistic scenarios, the largest area will be for classes with very high resilience, moderate resilience, and very low resilience, respectively.

The areas of very low resilience and population density for 22 districts of Tehran are shown in

Table 3. Percentage of very low resilience area in different scenarios and population density for 22 districts of Tehran.

Districts	Population Density	Pessimistic	Neutral	Optimistic
1	10,270	2.716	48.298	99.235
2	14,999	0.975	21.397	55.791
3	11,986	0.010	10.531	47.347
4	15,710	0.090	22.507	68.240
5	15,337	1.531	19.753	52.506
6	12,643	0.006	0.018	3.310
7	20,374	0.000	0.000	0.000
8	35,806	0.000	0.000	0.137
9	9757	0.000	0.000	0.004
10	41,921	0.000	0.000	0.032
11	25,987	0.000	0.000	0.000
12	14,954	0.000	0.000	0.001
13	18,712	0.000	0.007	5.333
14	28,912	0.000	0.000	0.112
15	23,372	0.018	0.828	51.840
16	15,466	0.016	0.161	14.883
17	39,568	0.000	0.017	0.148
18	11,901	0.001	0.001	0.001
19	13,366	0.000	0.105	7.477
20	18,990	0.009	0.784	46.601
21	3887	0.000	0.688	16.370
22	3610	0.189	20.263	68.891

. Across all 22 districts of Tehran, the percentage of the class with very low resilience increases as optimism increases. In Tehran, districts 1, 4, 2, 22, 5, and 3 have the highest percentages of area classified as very low resilience out of 22 districts. Districts 7, 11, 12, 18, 9, and 10 have the smallest areas in the class of very low resilience. In districts 4, 5 and 2, which have a high area of very low resilience, flood and earthquake risk planning is more important. This is due to the high population density of these districts as well as the high area of very low resilience.

4. Conclusions

Currently, human settlements, particularly metropolises, face a number of serious and basic issues, including natural hazards. The identification and preparation of a map of the resilience areas of cities may be beneficial. Urban planners and managers must consider a variety of risk management cycles at various stages in the planning and management process. This will minimize the time and amount of recovery after a hazard has occurred. To identify resilience areas to floods and earthquakes, a combination of OWA and AHP was used. The decision support system developed in this study was implemented in the form of three decision scenarios (pessimistic, neutral, and optimistic). As a result, the system generates different scenarios based on levels of optimism and pessimism. In this study, it was shown that the weight of effective criteria determines the level of resilience in different regions. In determining resilience levels, demographic and infrastructure criteria have the highest and lowest weights and importance, respectively. The area of classes with low resilience decreases in the pessimistic scenario, while it increases in the optimistic scenario. Thus, planners and managers may select the appropriate scenario based on their attitude (risk-averse or risk-taking). As a practical guide to determining how a city is going to respond to natural hazards, it is beneficial to examine urban resilience and prepare location maps for its assessment. A number of the results can assist planners in developing spatial planning strategies. Additionally, authorities may use these results to devise disaster risk reduction strategies and policies. For example, they can use the approach introduced in this study to identify vulnerable areas that should be prioritized during resilience-building activities. This study has a limitation as it relies on experts’ opinions when determining the priority of the criteria. The agreement between different experts can pose an uncertainty regarding the priority of the criteria. Furthermore, one of this study’s limitations is the inability to obtain real-world data from areas affected by floods and earthquakes, which may facilitate the evaluation of study results and the improvement of resilience modeling. Using large group decision-making methods in future studies is suggested as a means of reducing the uncertainty associated with weight determination due to a limited number of expert opinions. It is also possible to use the system proposed in this study to assess the resilience of other natural hazards in a flexible manner based on stakeholder attitudes in future studies.