Urban Resilience Assessment Using Hybrid MCDM Model Based on DEMATEL-ANP Method (DANP)

Resilience is an essential strategy for building the capacity of communities and cities. There have been many changes in attitudes toward hazards. In effect, the global community's perspective has shifted from focusing on reducing vulnerability to enhancing resilience in times of crisis. This study used a GIS-based DANP model to study the resilience of districts in Tehran to hazards. First, the criteria affecting resilience are selected according to four dimensions (environmental, socioeconomic, physical and institutional) using the Delphi method. After two rounds of selection, 11 criteria included disasters and natural disasters, water resources, environmental pollution, topography, urban infrastructure, land use, green space, employment rate, population and education, health status, and insurance coverage were selected. Then, a DEMATEL model followed by ANP was applied to determine the internal relationship between the criteria. Then, GIS overlay was performed to provide visual output. The DEMATEL results show that disasters and disasters in terms of environment, urban infrastructure in terms of physical aspects and employment rate in terms of socio-economic aspects are the most important criteria affecting the ability to urban resilience. In addition, 54.7% of the total urban area is classified into very low to moderate resilience categories, which need special attention. The results of the sensitivity analysis show that the sub-criteria of the vulnerability of critical infrastructures in the event of earthquake and the vulnerability of structures to an earthquake have the highest rate of change, which will definitely have a greater impact on the results of resilience. Moreover, the sub-criteria of the number of educational centers and life expectancy have the lowest change rates. This research provides new perspectives to help urban planners understand the causal relationship between dimensions and criteria, to better understand resilience.


Introduction
Over the last few decades, the world population has grown exponentially (Choi & Labhsetwar, 2020). The urban population accounted for 55% of the total population of the world in 2018, which is predicted to reach 65% by 2050 (Desa 2019;Foroozesh et al., 2022). Sustainable cities provide a safe environment for economic activity, opportunity, and creativity. However, several hazards have posed a threat to this security by creating crises and causing vulnerability in the structure and function of cities (EC, 2010). The vulnerability of a city leads to irreparable damage to physical, environmental, social, economic, and institutional dimensions, which can threaten the foundation of a city (Fatemi et al., 2020;Formetta & Feyen, 2019). Generally, disaster risk management has resulted in actions to avoid new hazards and minimize and manage current hazards, disaster risk management combined with strengthening urban resilience is now recognized as a proper practice for preventing and responding to the consequences of a disaster (Benson, 2016;UNDRR, 2019). Therefore, disaster risk management along with sustainable development perspectives seek to increase resilience among communities and individuals in response to hazards (Benson, 2016;Uitto & Shaw, 2016). In this view, it is essential to perform strategies for urban resilience which can decrease disaster vulnerability by changing challenges to opportunities (Moraci et al., 2018).
Resilience is one of the significant principles of sustainable development (Liang & Li, 2020). Resilience, originally meaning as jumping back and bounce-back (original meaning from the Latin ''resilire '', ''resalire'', ''resilio,'') and the ability of recovery and restoration (Annarelli et al., 2020). Resilience mostly focuses on the bounce-back of the system to the equilibrium overtime after a disruption (Döring et al., 2015). There are various descriptions of resilience in previous studies. Holling (1973) first used the word 'resilience' and defined it as a measure of the persistence of systems and their ability to absorb change and disturbance and maintain the same relationships between populations or state variables. Wagner and Breil (2013) state that resilience is the general capacity and ability of a system to resist pressure, survive, modify, and bounce back from a disaster. In addition, Martin-Moreau and Menascé (2018) stated that the resilience of a system is its ability to reconstruct itself and recover its balance after being disordered. In this way, resilience describes not only the capacity to resist but also the ability to recover after a shock and return to the previous state. The common point of all these definitions is the reversibility of capacity, adaptation, disaster recovery, absorption, resilience, and survival, which could be applied to deal with disruption and adaptation of society. These disorders include crises, stresses, and shocks.
The concept of resilience has been constantly developing and could refer to resilience for adapting in the face of climate change (Rajkovich & Okour, 2019), hazard risk management and resilience (Etinay et al., 2018;Galderisi & Treccozzi, 2017), resilience in energy and environmental security (Keskinen et al., 2019), urban resilience (Mohamadzadeh et al., 2020), and social-ecological system resilience (Garmestani et al., 2019;. The review of the research literature indicates that resilience is a comprehensive concept with several dimensions. Moving toward a comprehensive definition of this concept for considering all the influencing factors can eventually lead to resilience in systems. Therefore, it is highly important to determine the factors which influence urban resilience, which can help successful planning and improve adaptation and management of changing conditions. Knowing that how much a city is resilient toward hazards before occurring such events is of great importance. There are two main approaches for urban resilience measurement in the literature. Decision makers are interested in assessing the resilience of a designed landscape before a disaster occur. But this is a difficult process because resilience can be measured only after a disaster (Cariolet et al., 2019). Despite widespread interest in resilience, there is relatively little research implementing and/or testing disaster resilience measures (Cutter, 2016). Some researchers have demonstrated two different concepts of a city's resilience, against which different measurement methods are concerned. Adaptive and inherent resilience. Inherent resilience is the resilience that exists before or before the event in the community. It is often used as a benchmark to measure results and change over time. Adaptive resilience is the ability of individuals, stakeholders or groups to learn and respond to changes caused by a hazardous event (Cutter, 2016). Two main approaches, bottom-up and topdown, have been distinguished to measure landscape resilience (Cutter, 2016). The data used for bottom-up approach is mostly non-spatial and therefore cannot lead to a map (Cariolet et al., 2019). Two main suggested indicators for this approach include the Urban Resilience Index developed by International Environment and Disaster Management Laboratory in Kyoto (Kabir et al., 2018) and the City Resilience Index (Arup International Development, 2015). In inherent resilience mapping, several methods use the concept of vulnerability to view low vulnerability as high resilience (Highfield, 2014). Several other methods follow the capacity approach, where the number of variables based on the resilience property is used to construct resilience indices (Cariolet et al., 2019). The resilience capacity index (Foster, 2007) is the most commonly used method in this category. Most of the proposed approaches are not sufficient to generate a resilience map. Baseline Resilience Indicators for Communities (BRIC) , Nexus City Index (NCI) (Schlör et al., 2018), Resilience to Emergencies and Disasters Index (REDI) (Kontokosta & Malik, 2018) are the methods used to generate a resilience map. Out of all the methods proposed in the research literature, only a few are applicable on a city-wide scale. The majority of studies on urban resilience do not consider the entire urban ecosystem and use assessments that are limited to specific systems (Tong, 2021). Additionally, there is a lack of interview-based methods in the literatures (Tong, 2021). The preparation of hybrid methods and subsystem resilience measures is one of the research gaps pointed out by the researchers. Furthermore, it has become increasingly common to use spatial data to assess the resilience of cities recently. As we know, Geographic Information Systems (GIS) enable thematic map and graph visualization that can improve data recording, analysis, aggregation, and policymaking at various spatial and temporal resolutions (Kong et al., 2022).
Urban resilience can be affected by a number of factors, which is important for developing a measurement model. In different studies, different criteria have been presented to assess resilience. Some studies have examined a limited number of criteria in various aspects, while others have examined more criteria. In some studies only three dimensions including society, economy and ecology were regarded for urban resilience measurements (for example, Fu et al., 2021;Sajjad et al., 2021). Sun et al. (2021) proposed 11 different factors in physical, ecological, location and society dimensions. Ostadtaghizadeh et al. (2015) propose five dimensions to measure urban resilience: physical, natural, economic, institutional and social. Tong (2021) states that five important aspects are presented in related studies including: built environment and infrastructure; environmental materials and resources; society and well-being; economy; and finally governance and institutions. However, none of these studies provide an exhaustive list of factors affecting urban resilience due to the lack of a guiding theoretical framework, leading to a significant knowledge gap (Huang et al., 2021).
Tehran is the largest city and capital of Iran with more than 9.2 million of inhabiting population. Tehran suffers from widespread pressures on natural resources and ecological life support systems due to rapid population growth, high concentration of industries, governmental organizations, services, and utilities, the existence of seismic faults, rusty and old places, Groundwater level reduction, the volume of traffic and air pollution (Moghadas et al., 2019). This issue has led to environmental, economic, social, and cultural problems which pose serious challenges for planners and decision-makers in strategic urban planning. Some of these challenges in Tehran megacities include decreased capacity to absorb contaminants, the imbalance between population and urban infrastructure, high resource consumption, intensive land use, and inequalities of urban districts in urban per capita. As per above-mentioned gaps in the researches concerning urban resilience the present study aims to present a hybrid model to assess the criteria affecting urban resilience in the Tehran metropolis. There is a lack of comprehensive understanding of the factors affecting urban resilience and especially the mechanism of their influence in the literature. To fill this gap, the DEMATEL (Decision making trial and evaluation laboratory) integrated with ANP (Analytic Network Process) was used in this study. DEMATEL is a powerful tool to help team decision-making by studying cause-and-effect relationships between several factors. It analyzes the relationship between any two elements in the system using matrices and graph theory (Wang et al., 2018). Considering the fact that several complicated and interrelated factors influence assessing urban resilience, this study applies a hybrid MCDM (Multi-criteria decision making) approach by integrating the DEMATEL, and ANP. Therefore, this study represents one of the first attempts to develop an index-based measurement using a hybrid DEMATELbased ANP (DANP) and GIS method for resilience assessment for Tehran, the capital city of Iran. In the DANP method, the dependent relationships between the criteria and the relative importance of each criterion are determined. More specifically, the present study mainly aims to: • Provides a comprehensive list of indicators to measure urban resilience. • Build a network structure model for each criterion and dimension. • Combine MCDM approaches with GIS • Explore dependencies and drivers, and elaborate on the interactions among features that influence Tehran's resilience. • Gain expertise in measuring urban resilience through interviews • Measure resilience in a sub-system scale (districts) • Two models of MCDM (DEMATEL-based ANP) manage great complex data and provide adequate results for determining resilient areas of Tehran and manage the high volume of data with various natures in a short time

Theoretical Framework
Resilience, as a dynamic process relying on the inherent and adaptive capacities of a society, helps decision-makers deal with unforeseen shocks or stresses by developing strategies and action plans (Moghadas et al., 2019). Resilience is a multi-dimensional approach. Thus, assessing resilience structure should ideally consider all various dimensions and criteria of an urban system to improve the integrity and content validity of the assessment system (Sharifi & Yamagata, 2016). In this study, four dimensions of resilience are considered based on theoretical and experimental principles: social-economic, institutional, physical, and environmental dimensions. The first component is the social and economic dimensions. Social resilience is the capacity of communities and social groups to withstand external social shock toward improving social capacity to resist environmental, social, economic, or political challenges (Maclean et al., 2014;Maguire & Hagan 2007;Khalili et al., 2015). In other words, it expresses the capacity of social groups to recover and be reversible in response to disasters (Keck & Sakdapolrak, 2013;Kwok et al., 2016). In economics, resilience is the response and adaptation of people and societies to disasters, which can enable them to decrease the cost of probable damages to make financial recovery possible (Hallegatte, 2014;Rose & Liao, 2005). The second component is the physical dimension which includes assessing the community response and post-disaster recovery capacity. The physical system must be able to continue to play its role and function under the pressure of danger. A city without a resilient physical structure would damage greatly by unforeseen stress and shocks (Cutter et al., 2010). The third component is the institutional dimension which includes features related to risk reduction and planning. Here, resilience is affected by the capacity of communities and the use of local people to reduce risks for creating links within an organized community and improving and protecting social systems in a community (Godschalk, 2003). The environmental dimension of resilience concerns paying attention to issues such as environmental pollution caused by human activities and reducing the potential to absorb pollutants in urban systems (Cutter et al., 2014;Moghadas et al., 2019). It is worth noting that each of the socio-economic, institutional, physical, and environmental dimensions f. Table 1 and Fig. 1 present the dimensions, criteria, and sub-criteria which affect resilience based on the literature review.

Study Area
Tehran as the capital of Iran is one of the largest metropolises in the world. Official estimates indicate that 9.2 million people live in Tehran. Tehran metropolis is located between the latitudes of 35°36' to 31°44' N and the longitudes of 51°1 7' to 51°33' E. Urban is divided into 22 districts, 123 regions, and 374 neighborhoods (Fig. 2). Some statistics related to these districts are shown in Table 2.

Identifying the Criteria
Resilience assessment is a multi-criteria decision-making approach that evaluates resilience in different dimensions. To assess the degree of urban resilience, it is essential to identify the factors and crises which make urban vulnerable to natural and anthropogenic disasters. Therefore, it is necessary to pay careful attention for determining an appropriate and clear set of criteria for assessing urban resilience. In this study, a questionnaire with 53 criteria in four dimensions was designed according to a comprehensive literature review, experts' opinions, local conditions, and data accessibility. The Delphi technique was used to screen the selection criteria.
The Delphi method is a process for reaching an agreement on a matter between a panel of specialists (Hsu & Sandford 2010). This method is widely used in research to collect ratings anonymously from geographically dispersed experts, with acceptable results over time (Jordan & Javernick-Will, 2013). Previous researches have demonstrated the benefits of this approach and highlighted four main advantages for it namely, (1) ' the anonymity of experts; (2) the possibility of change in the decision made by experts at each round; (3) providing feedback so that the results of previous courses can be exchanged; (4) statistical analysis of responses; It eventually leads to a full review of each expert's responses (Cerè et al., 2019). The panel in this research included 20 experts including professional experts in government organizations and academic experts from various geographical locations. Table 3 shows the demographic information of the survey participants. Past research has shown that two or three rounds are sufficient to achieve a reasonable result in the Delphi method. The number of rounds depends on the level of consensus on all items, whenever 70% of the experts reached an agreement (Xiaorong et al., 2020). In this study, the Delphi technique was implemented in two rounds. The initial screening was conducted to determine a list of criteria that could be used for assessing resilience. In this step, the criteria with arithmetic means of less than 3 were eliminated (Table 4). In the second screening, criteria and sub-criteria were approved and respondents were presented with a 5-level Likert scale ranging from 'Strongly disagree' to 'Strongly agree' and were asked to rate the importance of the criteria for urban resilience.
Accordingly, four dimensions, 11 criteria, and 32 subcriteria were employed for assessing urban resilience ( Table 5). The four influential dimensions are environment, socio-economic, physical, and institutional aspects. The eleven criteria include disasters and natural hazards, water resources, environmental pollutants, topography, urban infrastructure, land use, green space, employment rate, population and education, health status, and insurance type.

Spatial Analysis using GIS
Each criterion should be provided, collected, and scrutinized as a map layer in a GIS-based environment. For this purpose, the following GIS analytical techniques were used: • Data and statistics for economics, health care, demographics, economic partnership, education, and consumption patterns were obtained from the Management and Planning Organization of Iran (MPO). • Topographic indicators, roads, infrastructure, land use, and green space were gathered from the National Cartographic Center of Iran (NCC).
• Disasters and natural hazards criteria were gathered from the National Disaster Management Organization (NDMO). • Data for the environmental and water resource criteria were collected from the Department of Environment and Water Resources Management Organization (WRMO) of Tehran Metropolis, respectively.
Finally, all vector layers were converted to raster format with a 30 m spatial resolution to apply GIS-MCDA functions. All layers were standardized to different measuring scales. This standardization was based on the cost and benefits context of the criteria, which could minimize or maximize the criteria based on their values and their significance for assessing sustainability (Mohamadzadeh et al., 2020). Several data normalization techniques exist in the literature such as; linear scaling, median-MAD normalization, truncation, z-score), logarithmic scaling, sigmoid double, Tan-h normalization and many more (Kutty et al., 2022). The Z-score method was used to standardize the criteria on a scale of 1-9. Next, the criteria weights were measured on different importance of criteria. The DANP approach was used to calculate the weights of the In recent decades, due to the occurrence of various disasters and environmental hazards of natural and human origin and the creation of adverse human and ecological effects, the approach of improving city resilience against natural disasters has been significant Xu et al. (2020), Alizadeh and Sharifi (2020) Water resources Maintaining and increasing the quality and quantity of water resources is important in urban planning to promote sustainable water management and make cities safer and more resilient Roach et al. (2018), Li et al. (2016) Environmental pollutants The consequences of environmental pollution affect the physical and mental health of citizens. Furthermore, by threatening the health of individuals as the social capital of society and reducing social health, it causes severe damage to the social and economic structure of that society and reduces urban resilience Green space The consequences of urban development and the complexities of today's environmental problems that have afflicted many urban communities, the existence of green space and its expansion are more necessary than ever. In fact, green spaces with their ecological, economic, and social functions, increase the quality of the urban environment and hence improve the resilience of the city Ni'mah & Lenonb (2017)

Population and education
Increasing the population of urban areas beyond its capacity provides physical, economic, and environmental problems for planners and citizens. Furthermore, one of the most significant actions to minimize vulnerability and promote urban resilience is to raise public awareness and education about the hazard and resulting harm that might occur when it happens, as well as how to deal with it Newport and Jawahar (2003), Chou and Wu (2014) Health status Improving the health situation and providing appropriate facilities and infrastructure according to the population is one of the basic aspects in improving the quality of healthy life and increasing the resilience of society against the crisis Bhandari and Alonge (2020), Thomas et al. (2013) Insurance The insurance industry is one of the factors increasing society's resilience to natural disasters. In critical conditions, insurance provides peace of mind, compensation for economic costs, and facilitates the return of the pressurized system to normal conditions as defined in resilience, resilience provides the ability to return from difficult situations and repair the system de Vet and Eriksen (2020), Surminski et al. (2016) criteria. DANP is a combined model of DEMATEL and ANP techniques (Azarnivand & Chitsaz, 2015;Chen et al., 2010). In this study, the DEMATEL model was applied to evaluate the internal relationship between the criteria, and ANP was employed to assign the weights. Then, the Weighted Overlay technique was utilized to overlay the layers. Finally, the natural break classification was used to classify the resilience layer into three classes: low, Fig. 1 The sub-criteria effective in urban resilience assessment moderate, and high. Figure 3 shows the schematic flowchart of the urban resilience assessment in the Tehran metropolis.

DEMATEL-based Analytic Network Process
The DANP is a decision support tool that can sufficiently provide relationships and interdependence between various subjects to facilitate problem-solving. This model can verify the interdependence of variables and attributes and build a relationship that reflects the characteristics of an essential system and evolutionary trend (Chiu et al., 2013). DANP is a hybrid MCDA that combines the DEMATEL and the ANP (Liu et al., 2014). DEMATEL solves the problem of influencing factors within a multi-factor interleaving system (Uygun et al., 2015). By using graph theory and matrix tools, the DEMATEL method can calculate the cause and effect of each factor and convert the relationships among the factors into a structural model to visually represent the interdependence among them. The DANP model adopts a composite influence matrix, instead of normal pairwise comparison matrices within the ANP, to discover the weight of each factor (Wang et al., 2018). The main steps of the DANP method are as follows:

DEMATEL Method
Step 1: The five-point scale is used for pairwise comparisons by considering the level of impact of particular dimensions. The measurement criteria of 0, 1, 2, 3, and 4 are applied for indicating no, very low, low, high, and very high influence, respectively (Tseng, 2009).
Step 2: The direct-influence matrix is created concerning the degrees of relative effects stemming from the pair comparisons in Eq. (1). An n9n direct-influence matrix A is obtained based on the directly observed relations, where aij indicates the degree of impact of the i factor on the j factor (Taghizadeh Herat et al., 2012).
A ¼ a 11 a 1j a 1n a i1 a ij a in a n1 a nj a nn 2 4 3 5 ð1Þ Step 3: The normalized matrix obtains from Eqs. (2) and (3): s=min 1 max P n j¼1 a ij ,    Awareness of city managers of threatening risks 3 3.75  Step 4: Deriving the full relationship matrix T from Eq. (4): Step 5: Summing each column and row to obtain D and R Eqs. (5) and (6): Here, d i shows the sum of each row in T and the rows indicate the degrees of direct and indirect effects over the other criteria, and r j is considered as the sum of each column in T, where columns refer to the effect degree from other criteria (Lee et al., 2011).
Step 6: Evaluating a threshold value for disregarding the minor effects is essential for separating the relationship structure of the factors and obtaining the Network Relation Map (NRM) (Yang & Tzeng, 2011).

ANP Method
Step 1 In this step, a conceptual model is designed and the relationships between/among clusters and nodes are evaluated.  Step 2 In this stage, two elements are compared by decision-makers. Pairwise comparisons are established based on the marks ranging from 1 to 9 (Vasiljević et al., 2012). A reciprocal value of each number is applied for indicating the inverse comparison. The values of pairwise comparisons are specified in the comparison matrix and the local priority vector is stemmed from the eigenvector. Like the AHP (Analytic hierarchy process), the consistency of the pairwise matrix should be less than 0.1 (Sener et al., 2011).

Implement a Multi Criteria Evaluation
Step 3 The weights created from the previous steps are introduced into the supermatrix including the entire network components, which shows their inter-relationships. In this step, the supermatrix is called the initial supermatrix. Eq. (7) indicates the general form of the supermatrix.
Step 4 The cluster weights should be calculated for weighing the initial supermatrix. The initial supermatrix can be measured by multiplying the cluster weights matrix by an initial supermatrix after obtaining the cluster weight matrix (Nekhay et al., 2009). The newly obtained matrix is known as the weighted supermatrix.
Step 5 Finally, the weighted supermatrix n times are multiplied by itself until reaching the limit supermatrix (Nekhay et al., 2009).

Sensitivity Analysis
Sensitivity analysis involves measuring the impact of unforeseen changes and policies on the final outcome of decision-making (Foroozesh et al., 2022). The sources of uncertainty in the results of MCDM methods are related to different types of errors, many of which are weighting criteria, data, knowledge of the system, and influential expert decisions, affect the results. (Feizizadeh and Blaschke, 2014;Myagmartseren et al., 2017). One of the most commonly used methods for sensitivity analysis in relevant studies is the OAT (One-at-a-time) method. In this method, the weight of a single criterion is modified according to the specified values while the other criteria remain unchanged. To perform the sensitivity analysis, all criteria were changed by ± 6, ± 12, ± 18 (Foroozesh et al., 2022).

Result
Efficient factors for assessing urban resilience were determined based on the literature review and area conditions. A part of the criteria layer is presented in Fig. 4. The DEMATEL-based ANP method was used since that multiple criteria had different effects on the process of assessing urban resilience. Considering the fact that the ANP model investigates the relative importance of criteria in the network, it was essential to apply the DEMATEL approach to resolve the issues of interactions or interdependence among the criteria. Thus, the current study has three results that involved: building the NRM, determining the weights of criteria, and generating an Urban Resilience Map (URM).

Building the Network Relation Map (NRM)
The first result was the dependent relationships between the criteria. Tables 6 and 7 show the amount of interaction between the criteria and dimensions. Figure 5 shows an Influential network relation map (INRM) for a visual representation of the four dimensions and their criteria.
The value of (d i -r j ) listed the criteria in cause-and-effect classes. As can be seen in Table 6 and Fig. 5, the environmental dimension (D1) has the highest (d i -r j ) value with 1.18, which has a direct impact on other dimensions. In addition, the influential impact degree of (D1) is 5.95 (Table 6), which is ranked as the second-highest degree among all causal dimensions. The 'Socio-economic (C)' dimension has a significant impact on other cause group dimensions with the second highest (d i -r j ) value of 0.61. Additionally, (C) has the first highest r i value (6.01) among the causal dimensions in terms of prominent impact degree. The Institutional (D) aspect has the lowest value of (d i -r j ) with (-1.22), which is the most vulnerable to influence. Generally, A affects C, B, and D dimensions (A ? {C B D}), C affects B and D dimensions (C ? {B D}), and B affects D dimensions (B ? {D}). Understanding these cause-and-effect relationships helps planners and urban managers make decisions for solving resilience issues. For instance, urban planners should first take action to develop the D (Environmental) dimension, and then C (Socioeconomic), B (Physical), and D (Institutional) dimensions, respectively. Furthermore, the causal diagram shows that among the four dimensions, the socio-economic (C) dimension has the highest prominence (d i ? r j ) value with 11.42 (Fig. 5). Prominence ranking is listed from the highest to the lowest value as follows: C (Socio-economic), D (Institutional), A (Environmental), and B (Physical). Figure 5 and Table 7 demonstrate the most significant causal criteria. As can be seen, A1 affects A3, A2, and A4 criteria (A1 ? {A3 A2 A4}) in the environmental dimension. According to Table 7, the (d i -r j ) values of the A1 (Disasters and natural hazards) and A3 (Environmental pollutants) criteria are positive. Hence, this criterion is classified as the cause group. Disaster and natural hazard criteria (A1) have the maximum positive value (d i -r j ) (1.11), and indicate that it has a significant effect on all of the criteria. Further, the A2 (Water resources) and A4 (Topography) criteria were located in the effect class due to the negative amount of (d i -r j ) values. Topography (A4) has the minimum negative value (d i -r j ) (-1.06) and receives a significant effect from cause class criteria. The results indicate that urban managers in Tehran can improve urban resilience in the environmental dimension by identifying types of natural hazards such as floods, earthquakes, and landslides and preparing vulnerability maps to identify safe situations for citizens in hazard situations.
Further, B1 affects B2 and B3 criteria (B1 ? {B2 B3}) in the physical dimension. This finding indicates that the (d i -r j ) values of the B1 (Urban infrastructure) and B2 (Land use) are positive between these criteria and can influence all criteria. Urban infrastructure (B1) has the highest (d i -r j ) value with 1.19, which impacts the other sub-criteria of the physical dimension. In addition, the results show that B3 (Green space) is negative, has the minimum negative value (d i -r j ) (-1.57), and is affected by other criteria. Therefore, urban managers can improve urban resilience in the physical dimension by observing international standards of design and planning for providing the infrastructure and facilities. Additionally, they can protect vital and infrastructural public facilities through reconstruction.
C1 affects C2, C3, and C4 criteria (C1 ? {C2 C3 C4}). The employment rate (C1) is another significant criterion since the (d i -r j ) value is the maximum value (1.13) in the socio-economic dimension. Furthermore, the results indicate that C3 (Health status) and C4 (Insurance type) are negative and are influenced by other criteria. C4 has the minimum negative value (d i -r j ) (-1.26). Therefore, urban managers can improve urban resilience by stabilizing the economic activities in the district, providing employment facilities, and taking measures to attract investment to diversify economic activities and increase people's financial capacity. Similar influential relationships could also be defined for the other criteria, as illustrated in details in Fig. 5. Fontela and Gabus (1976) stated that due to internal relationships between factors, more attention should be paid to the cause class criteria due to their impact on the effect class criteria. This method is a useful tool for urban decision-makers to identify priorities for increasing urban resilience. Urban decision-makers can define regular prevention programs based on causal factors to increase urban resilience. Table 8 presents a preventive program for the most important causal factors in urban resilience according to the Tehran situation.

Determining the Weights of Criteria
The second result is related to determining the weights of the criteria. Based on the performance matrices of the -DEMATEL method, ANP was used to measure the criteria weight in a network structure. Finally, the limits of the super-matrix W a were applied to obtain the criteria weights presented in Table 9 and Fig. 6.
As shown in Table 9, the environmental dimension with a weight of 0.27 has a more important effect on urban resilience in comparison to other dimensions. It is noteworthy that the weights of the percentage of critical infrastructure vulnerabilities in the event of a natural disaster (earthquake) (0.077), the degree of the vulnerability b Fig. 4 Part of the criteria layer are used in urban resilience assessment of the structure to hazards (0.061), the worn-out texture rate (0.058), the employment rate (0.054), the poverty line (0.05), and the population density (0.047) are the most important sub-criteria in comparison to other sub-criteria. Therefore, these six sub-criteria are the key factors for assessing urban resilience.

Population and education
Promoting economic activities in the district for economic recovery after the crisis Increase business and investment opportunities to increase urban economic resilience Implementing regulations that limit building density to increase resilience Implementing regulations that limit population density to increase resilience Educating citizens to prepare for a crisis through photos, posters, seminars, etc Implement programs for citizens to identify hazards, increase awareness of hazards, safety training, etc

Discussion
Urban resilience to natural hazards is a complex issue. Measuring inherent resilience plays an important role in designing a more resilient city. It deals with the preemergence of an event and topic of interest to planners. It can be effectively measured by finding appropriate criteria  and combining them. The method proposed in this paper uses 32 criteria spread across four dimensions to measure a city's resilience to natural hazards. The main contribution of this study was to find a comprehensive list of urban resilience indicators and to provide a hybrid approach in which expert knowledge is involved in figuring out urban resilience. As mentioned in the previous literature, the role of expertise in mapping methods has been overlooked (Tong, 2021), which has been covered in this article. There is a lack of a method of measuring urban resilience at the city scale. The framework proposed in this study is well suited for measuring the city and its subsystems and can be a city-wide specific approach. Combining different methods with their associated advantages can provide more powerful tools than single models. Through the model proposed in this paper, a new combined method is introduced to measure the urban resilience of megacities, in which the advantages of the MCDM approach, the Geographical information system and expertise have been reasonably combined. Figure 7 shows the spatial pattern of Tehran's metropolis resilience as well as the districts which need additional attention. The spatial patterns of resilience rates show that districts with very high to high resilience (Districts 4, 1, 2, 5, and 22) are located in the northwest to northeast parts of the study area (Table 10). The household economy situation in these districts is moderate to high. These districts include mostly old residents and immigrants with high economic and social status. Furthermore, access to welfare facilities, urban infrastructure, social services, and large green spaces such as Chitgar, Koohsar, and Kan forest parks for local communities is better than other districts. The districts with moderate resilience (Districts 3,6,21,18,19,16,15,20,and 13) are mostly located in the southern, southwestern, and western areas of the study area. According to Fig. 8, 17% of the total area of Tehran is located in 'very low' and 'low' resilience classes. Districts with very low to low resilience (Districts 7,8,9,10,11,12,14,and 17) are in the central and eastern areas of the study area. These districts have a very low to a low degree of resilience due to high population density, wornout texture, the vulnerability of infrastructure to hazards (earthquake), and air pollution because of compact urban structure. Furthermore, a wide range of activities, government bureaus, and offices of large industries are located in these areas due to the location of large shopping complexes and commercial land uses, which has led to low land-use mixing in these areas. On the other hand, due to important faults such as the Mosha or North Tehran fault and the possibility of a magnitude 7 earthquake in Tehran, the designs made according to the new regulations in these areas are not responsive. Compact patterns and reducing the number of open spaces is one of the effective factors in reducing the resilience of these areas. Also, the study of medical uses indicates an imbalance in the distribution of these services in these areas and, consequently, reduced access and lack of rapid response in the event of a crisis. The results of the DANP method showed that in the field of disasters and natural hazards, the sub-criteria of the percentage of the vulnerability of critical infrastructure at the time of the earthquake (0.077) and the degree of vulnerability of structures to earthquake (0.061) are the most important which was consistent with the results of the study of Namjooyan et al. (2020). They conducted a study titled Improving the resilience of Tehran metropolis against natural catastrophes with a focus on earthquakes. The findings of the study showed that the importance of the physical dimension of infrastructure is correlated with the priority of dimensions to increase earthquake resilience in Tehran's 12th district, with a correlation of 0.462. Eshgi et al. (2018) evaluated physical resilience against possible earthquakes in region one of Tehran. The ANP method is used in this study. The difference between the above study and our study is the combined application of ANP and DEMATEL methods in order to investigate the causal relationships between the criteria and determine the effects of the criteria, as well as consider all dimensions of resilience. This study aimed to provide opportunities for urban planners and decision-makers to reflect their decisions based on the DANP method. The advantages of DEMA-TEL based ANP technique for decision-makers could be stated as follows: • It is a proper method of MCDM that is used to identify the pattern of causal relationships between the variables. • This method has clarity and transparency for showing the interrelationships between an extended range of criteria as compared to the network analysis approaches, which can help experts express their views on the direction and intensity of the effects among factors with more knowledge. • It expresses the direction and intensity of the effects between the criteria in a quantitative way by presenting diagrams for visual interpretation. • It determines the importance and weight of the factors by considering all available factors and dimensions. • Structuring complex factors in the form of cause-andeffect groups is one of the most important functions, which could be considered as one of the most important reasons for its widespread use in problem-solving processes. By dividing a wide range of complex factors into cause-and-effect groups, it puts the decision-maker in a better position to understand relationships, which can result in a greater understanding of the position of factors and their role in the interaction process. • The ANP technique used in this study is a suitable tool for network ranking and does not require a hierarchical structure. Thus, it shows the more complex relationships between different levels of a decision in a network. In fact, it considers the interactions and feedback between the criteria and provides a suitable framework for analyzing the problem.

Conclusion
Today, there are many changes in attitudes toward hazards, and the views of the world community have shifted from focusing on reducing vulnerability to increasing resilience in response to hazards. Accordingly, urban decision-makers focus on hazard reduction programs by developing and strengthening the characteristics of resilient cities and communities which are in line with sustainable development goals. The present study aimed to assess urban resilience to use management and prevention programs before the occurrence of natural and human hazards by using the combined Delphi and ANP-DEMATEL method for identifying and evaluating important urban resilience criteria, determining the internal relationships between the criteria, and assessing their impacts on each other. This study aimed to determine at least a set of appropriate and clear criteria based on experts' opinions by the Delphi method. This method decreases the decision time for urban planners. Considering the Delphi findings, four dimensions, 11 criteria, and 32 sub-criteria were employed for assessing urban resilience. As Kharat et al. (2016) and Shi et al. (2020) maintain, the Delphi method is an effective method for measuring criteria and finding optimal solutions which can effectively convert vague, subjective, and linguistic data from experts' opinions into quantitative and logical results. In the present study, the DEMATEL combination with ANP approaches was used to determine the interdependent relationships among the criteria and their importance. Azizi et al. (2014) stated that ANP cannot assign the strengths and internal relationships between the criteria and does not pay attention to this issue, which could cause the model results to deviate from the real situation. To overcome this shortcoming, DEMATEL was applied along with ANP. ) stated that the DEMATEL based ANP method can correct the deficiency of the ANP method and reflect the interdependent feedback relationships between the factors, which could ensure that the results are scientific and reasonable. Based on the DEMATEL method, A1 (Disasters and natural hazards) in the environmental dimension, B1 (Urban infrastructure) in the physical dimension, and C1 (The employment rate) in the socio-economic dimension had the highest degree of effect (d i -r j ). The results of this study revealed that DEMATEL based ANP is a good approach for better understanding the issues and can help urban planners make accurate decisions. The findings help urban planners consider the criteria of the cause group for defining priority prevention programs to increase urban resilience. The results of this study show that the most important factors affecting urban resilience are those related to the vulnerability of infrastructure and the wear and tear of urban structures. Therefore, it is suggested that city managers pay special attention to improving the standards, securing urban infrastructure and reducing the city's worn out in order to improve the city's resilience to environmental hazards, risks and damages. In this study, different criteria in several dimensions were used to show the degree of resilience against unpredictable stresses and shocks. However, the unavailability and inaccessibility of data for some criteria were the limitations of the study. Consequently, some of the criteria should be omitted since they might influence the obtained results. For instance, the institutional dimension, which plays a significant role in the preparedness and planning stage for hazard resilience was measured by only two criteria to this limitation. Integrated conjoined collaboration between governmental sections and between government ministries, resource sharing between different urban subsystems, transparency, accountability and participation in the public budget clearly governing regulations and rules and a number of other criteria can be taken into account in this respect. Finally, it is suggested that further research be conducted on the resilience of the influencing factors in each of the dimensions introduced by the DANP method.
Funding This research did not receive any specific grant from funding agencies in public, commercial, or not-for-profit sectors.

Declarations
Conflicts of interest The authors have no conflict of interest to declare.