Prioritizing the Components of the Disaster Resilient System Using Dematel and Anp for Urban Areas

Although many research studies have been performed on vulnerability assessment and hazard mitigation


Introduction
Recent incidents indicate that communities and people are increasingly becoming more vulnerable to disasters, and the risks are globally rising (Mayunga, 2007;Ainuddin and Routray, 2012).However, risk reduction and vulnerability are often ignored until the occurrence of disasters (Cutter et al, 2008b).Under these circumstances in which risks and uncertainties are growing, the concept of resilience has been defined as the managing of disturbance, unanticipated events, and change (Mitchell and Harris, 2012).
Resilience to natural hazards and disasters has been defined as the ability of communities exposed to hazards to resist, absorb, accommodate, and recover from the effects of a hazard in a timely and efficient manner, including through the preservation and restoration of its essential structures and functions (UNISDR 2009).
While the concept of resilience has been used in science since 1625 (Alexander, 2013), in the field of ecology; it was first introduced by Holling (1973) and has been applied in disaster risk reduction since 2000 (Alexander, 2013).
In fact, after the presentation of the Hyogo Framework for Action (1995), an evident paradigm shift has occurred from hazard assessment to vulnerability analysis and community resilience creation in recent hazard literature (Ainuddin and Routray, 2012).Although there is a variety of frameworks and models in disaster resilience, there is pg. 2 Sarmadia/ Environmental Science and Sustainable Development no consensus among researchers on the components and standard metrics of resilience.Moreover, the variables that should be utilized to measure resilience have not been determined (Ainuddin and Routray, 2012;Cutter et al., 2008b;Twigg, 2009).Vulnerability and resilience are indeed dynamic processes; however, they are often considered as static phenomena because of their measurement purposes (Cutter et al., 2008b).Recent literature (Norris et al., 2008;Mitchell, 2012) considers resilience as a process rather than an outcome and specifies resilience as a dynamic system.
In addition, various definitions have been presented for disaster resilience.Mayunga (2007) has summarized some selected definitions of the resilience concept published in disaster and hazard literature.A number of recent definitions have also been provided all of which indicate there is no consensus among researchers and practitioners on a common definition of the concept.It is evident that the definitions are diverse, reflecting the complex nature of the concept, and most of them emphasize the capacity to cope with disasters, disturbances, or emergencies.In other words, resiliency is often the flexible capacity of a system to absorb stresses and recover rapidly from emergencies.Accordingly, a resilient community has been defined as one that which can cope appropriately with disasters, has the highest level of flexible capacity, and will not be greatly influenced by disasters.A disaster-resilient community can also be defined as 'the safest possible community that we have the knowledge to design and build in a natural context' (Twigg, 2009).
Two kinds of strategies have been suggested; anticipation and resilience strategies.Anticipation strategies usually operate to resolve known problems, while resilience strategies better resolve unidentified problems (Wildavsky, 1988;Normandin et al., 2011).Resilience concentrates on community resources and methods of strengthening community capacities, instead of focusing on their vulnerability and deficiency (Twigg, 2009).On the other hand, vulnerability, as one of the most common elements of risk that is applied directly or indirectly to risk assessment methods, is often considered as contradictory to resilience (Twigg, 2009).However, resilience goes beyond the capacity component of vulnerability and the combination of these two concepts can lead to a more comprehensive understanding of disasters (Obrist et al., 2010).According to Godschalk (2003), a resilient city is a combination of physical systems and human communities.Resilient systems that can be applied to physical and social systems tend to be redundant, diverse, efficient, autonomous, strong, interdependent, adaptable, and collaborative (Godschalk, 2003).
A resilient community contains a variety of complex issues.Thus, many elements are involved in resilience due to which its description in a clear and organized manner is almost impossible.Therefore, the Disaster Resilient System (DRS) is defined as a resilient system that proposes and prioritizes appropriate dimensions and components of resilience against contingency earthquakes in urban areas based on experts' comments using Delphi, Decision Making Trial and Evaluation Laboratory (DEMATEL), and Analytic Network Process (ANP).It specifies the components that are vital for disaster (especially earthquake) resilience in urban areas with high levels of disaster as a result of earthquakes and should be considered by authorities and managers as their priorities.Some research has been conducted regarding disaster resilience in urban areas.
The economic and institutional resilience of four different districts of Tehran was evaluated by Rezaei (2014).He assessed institutional resilience (institutional context, institutional relations, institutional efficiency) and economic resilience (loss rate, loss compensation capacity, ability to recover) using Analytic Hierarchy Process (AHP) and Preference Ranking Organization Method for Enrichment Evaluation (PROMETHEE).He found that the loss rate indicator (0.383) and compensation capacity indicator (0.281) had the highest importance.

Finding the dimensions and components of DRS using the Delphi method
The first stage was the identification of appropriate criteria for the Disaster Resilient System in urban areas.The Delphi method was suggested for the demonstration of the dimensions and components of DRS.The Delphi method originated from the studies of the RAND Corporation in the 1950s and is known as a method that is applied to a wide variety of fields and a tool for expert problem-solving.The Delphi study is designed to answer questions with high uncertainty and based on speculation.Hence, only individuals with a high level of relevant knowledge are eligible to respond to these questions (Okoli and Pawlowski, 2004).
In the first step of the Delphi method, a group of relevant experts was identified and selected.For this research, a group of 26 experts in the fields of disaster management, urban planning, civil and environmental engineering, and geography were chosen.After the identification of experts, a relevant questionnaire should be designed.Therefore, in the second step, we selected 16 studies in the field of disaster resilience (Table 1) and classified their selected factors into 6 dimensions as main elements of DRS.Then, based on the obtained dimensions and components, the questionnaire was designed and sent to experts.The experts were asked to respond to the questionnaire and to offer feedback on the position of dimensions and components and consolidation or expansion of components, or to propose other new elements which they thought were important for DRS.
In the next step, the responses of experts were received and analyzed.Then, a new questionnaire was designed for the second round.The respondents were asked to revise their first response and answer other questions based on their responses to the first survey.They could change their responses or maintain their previous opinion.This process was repeated until more than 77% of respondents attained a component identical to that acceptable for DRS.Finally, 41 components in 6 main dimensions were selected (Table 2).

Determination of the intensity of the direct and indirect relationships between DRS components using DEMATEL
DEMATEL is one of commonly used techniques in multiple-criteria decision-making (MCDM) which is suitable for the analysis of the direction and strength of the direct or indirect relationships between defined components (Chen, Hsu, and Tzeng, 2011).The different steps of this method are described below.
Step 1: In this step, the direct relation average matrix should be calculated.The degree of the direct impact that each perspective i exerts on each perspective j is determined by respondents (Chen, Hsu, and Tzeng, 2011) using a scale ranging from 0 to 4 (0: no impact, 1: low impact, 2: medium impact, 3: high impact, 4: very high impact) (Chen, Lien, and Tzeng, 2010).The mean of the same perspective in the various direct matrices of the respondents is used to obtain an average matrix C (eq. 1) (Chen, Hsu, and Tzeng, 2011). C=[ Step 2: The initial direct impact matrix is calculated in this step.Therefore, the average matrix C should be normalized to obtain the initial direct impact matrix D using equations ( 2) and (3).

D=sC
(2) s=min Step 3: The total impact matrix is calculated in this step.By obtaining the normalized direct-impact matrix D, the total impact matrix T can be calculated using equation (4) as follows: Where I is the identity matrix.
Step 4: In this step, the sum of each row and column of matrix T are obtained using equations ( 5) and ( 6).]  * 1 (5) = the sum of the ith row of matrix T   = the sum of the jth column of matrix T Therefore,   illustrates the sum of direct and indirect impacts of component iIon the other components.On the other hand,   shows the sum of direct and indirect impacts that component j has received from the other components (Chen, Hsu, and Tzeng, 2011).Moreover, di+rj indicates the intensity of the relationship between components, and di-rj indicates the intensity of impacts among components (Azizi et al., 2014).

Applying DEMATEL data to ANP to find weighted supermatrix using Supper Decision software
Finding the intensity and interrelation between the components and dimensions in a complicated system such as DRS is very difficult.DEMATEL is not only used to find interrelation between factors, but is also used to specify the more accurate weights of the pairwise comparison (Chen, Hsu, and Tzeng, 2011).Moreover, due to its nonlinear structure, ANP can manage interdependencies between clusters (dimensions) and nodes (components of each dimension).ANP is one of the MCDM methods, is an extension of AHP, and was introduced by Saaty in 1996 and 2001.A combination of DEMATEL and ANP is increasingly used in different subjects such as equity investment (Lee et al., 2011), environment watershed plans (Chen, Lien, and Tzeng, 2010), hot spring hotel performance (Chen, Hsu, and Tzeng, 2011), brand marketing (Wangand Tzeng, 2012), and entrepreneurship policy (Tsai and Kuo, 2011).
The first step of ANP is defining a conceptual model which determines interconnections among dimensions and components.The second step is a pairwise comparison between components using Super Decisions software.The Super Decisions software is a mathematical theory that applies ANP to find dependence and feedback of dimensions and components.In this step, pairwise comparison is performed through number coding (from 1 to 9) in Super Decisions software (Fig. 1) (Azizi et al., 2014).The next step is the calculation of the primary supermatrix which is obtained from weights of the pairwise comparison of the previous step.Equation 7indicates the general form of the supermatrix.
Eq. 7: This equation has 6 dimensions and 41 components.Therefore, it is a 41*41 supermatrix in which Wij denotes the connection between the ith component and the jth component.In the final step, the dimension weights are calculated pg. 5 Sarmadia/ Environmental Science and Sustainable Development by weighting the primary supermatrix, which is obtained by multiplying the dimension weights matrix by the primary supermatrix (Azizi et al., 2014).

Discussion and Results
As mentioned previously, the Delphi method was used to find the relevant dimensions and components of disaster resilience in urban areas.Therefore, the opinions of 26 experts, mostly with experience in disaster management and primary knowledge of resilience, were applied.We tried to include various perspectives in our Delphi process to evaluate the results.
In the first design of the DRS, six dimensions and relevant components were introduced based on the 16 mentioned studies (Table 1).They were used to design a structured questionnaire.The experts were asked to comment on DRS and their components (dimensions and components that are thought to be appropriate for DRS).The structured questionnaire was sent to experts by email, fax, and rarely by mail based on their convenience.In addition to the questionnaire, the main base and origin of each dimension and component were provided to experts for their knowledge.Then, they were asked to modify the classification and the position of components if they had new ideas or comments.These comments were gathered and analyzed by the researchers.Identical responses were removed and other perspectives were consolidated to provide a new questionnaire.In the second round, experts provided their comments on the modified questionnaire.Finally, after 4 rounds, an agreement was reached by experts and more than 77% of participants accepted the dimensions and components of Table 2 as the elements of DRS.Although some of these components such as land use can belong to different dimensions, they were placed in only 1 dimension based on experts' comments.In the second step, the experts were asked to specify the extent of the impact of each component on other components through the numbers 0 to 4 (0: no impact; 1: very low impact; 2: low impact; 3: high impact; 4: very high impact) regarding the earthquake condition and resources of urban areas.Therefore, based on the obtained data, DEMATEL techniques, and programming by MATLAB software, the intensity and direction of the relationship between components were calculated.The outcomes were applied to introduce a conceptual model in ANP. Figure 2 illustrates the overall defined conceptual model of DRS in Super Decisions software.It shows how dimensions are connected and therefore impact each other.As mentioned previously, connections and intensity among components (connected to or from each component) were computed by importing DEMATEL outcomes data and defining user formulas to this software.In the next step, pairwise comparisons were performed between dimensions and components to specify the importance of dimensions and components to each other for urban areas.Finally, the results of the model were obtained as a weighted supermatrix, normalized components, and overall synthesized priorities for alternative dimensions in the following charts and tables.The intensity between dimensions is shown in Figure 3. Data in columns indicate the impact of each dimension on others.It can be seen that the economic dimension of urban areas has the most impact on structural and infrastructural dimensions (0.21837, 0.213053), and then, on the institutional and planning dimension (0.130875).This means that a strong economic dimension results in powerful and resistant infrastructural and structural dimensions in urban areas.
It also illustrates the impact of the economic dimension on other dimensions which is relatively high in comparison with other dimensions.The environmental dimension also has a high impact on structural and infrastructural dimensions (0.17263, 0.184595).Obtained data indicate that the infrastructural dimension is more influential on structural resilience (0.303017) and economic resilience (0.153694) in urban areas.The high impact of planning and socio-cultural dimension on economic dimension can also be seen in urban areas.However, in the next rank, planning is more influential on structural and infrastructural dimensions.It should be noted that the socio-cultural dimension has a reasonable impact on planning.Finally, the structural dimension also has a high impact on the infrastructural dimension, and then, with a significant difference, on the planning dimension.
pg. 10     The next analysis of the DRS was the analysis of normalized components by dimension which has been illustrated in table 4.This table represents the components that should be noticed for resilience in urban areas and illustrates the prioritized components of urban areas.It has been shown that the CEP is the priority of urban areas and the most effective component of the system.This means that experts believe there are some deficiencies in the CEP for urban areas.This also shows the role of emergency plans in resilience in urban areas.With minor differences, sustainability stands in second place.The next 8 important components for urban areas are, respectively, public facilities, municipal finance, lifelines, industrial and commercial units, residential units, critical infrastructure, community competence, and social network.It is evident that the majority of priorities are related to structural and infrastructural resilience in urban areas.Among the final 10 priorities, the outstanding role of social components can be observed in urban areas resilience.

Conclusion:
Since the use of the resilience concept in disaster risk reduction (Alexander, 2013), various models and frameworks have been proposed by scholars and scientists.Moreover, resilience has been analyzed in different fields, from environment to engineering and disasters.Although there is a variety of frameworks and models in disaster resilience, there is no consensus among researchers on components and standard metrics of resilience and the variables that should be used to measure resilience have not been determined (Ainuddin and Routray, 2012, Cutter et al., 2008b, Twigg, 2009).In this article, the physical and nonphysical dimensions of resilience were examined for urban areas Disaster Resilient System and the relationship between selected components and dimensions was specified using DEMATEL and ANP.
pg. 14 Sarmadia/ Environmental Science and Sustainable Development Whenever an emergency occurs, communities in urban areas are confronted with disturbances and interruptions; therefore, destruction and entropy are created based on community vulnerability.The community system of urban areas is resilient and has the capacity to cope with disasters if DRS is performed and implemented accurately.The selected elements, which consisted of 6 dimensions and 41 components, were determined using the Delphi method.ANP and DEMATEL were used to find the interactions between components in addition to their weights.Although ANP determines interactions between components, it cannot specify the weights of the components (Azizi et al., 2014).Therefore, the application of DEMATEL was suggested in combination with ANP to increase accuracy.As a result, the direction and the intensity of the relationship between selected dimensions and components were calculated using DEMATEL.Finally, the weight and importance of the components were obtained through ANP.It should be mentioned that due to the time-consuming process of responding to the questionnaire, it was very difficult to convince participants to take their time to complete it.
One of the key findings of this research is dimensional interrelationships.According to the findings, structural and infrastructural dimensions have the most impact on each other.Furthermore, the institutional and planning dimension and socio-cultural dimension have the most impact on the economic dimension.Furthermore, the economic dimension also has a noticeable impact on structural and infrastructural dimensions.
According to experts' comments and the analysis conducted using ANP and DEMATEL, the first prioritized component for contingency earthquake of urban areas is a CEP.This is mostly due to the lack of existence of a CEP for urban areas which should contain response plans, continuity of operation plans, recovery plans, preparedness and contingency plans, urban plans, disaster risk reduction plans, and infrastructure protection plans.These are pieces of a puzzle that should be defined with respect to each other.It is suggested that the TDMMO, as a responsible organization in disaster management in urban areas, provide a CEP for urban areas as one of the vital criteria in responding to this devastating disaster in urban areas.For the second priority of urban areas resilience, the sustainability indexes of urban areas should be recognized and defined.Moreover, in the first 10 prioritized components, the important role of public facilities resilience, lifelines resilience, industrial and commercial resilience, residential resilience, and critical infrastructural resilience was noticeable.This finding indicates that, besides other responsible organizations and agencies, TDMMO should implement projects to improve the resilience of the mentioned components.
Finally, experts believe that community competence and social networks are in a reasonable condition in urban areas; however, they have not been well-implemented in the CEP for urban areas.Therefore, it is suggested that authorities and managers consider these components as nonphysical components of resilience in their programs and decisions.

Fig 1 :
Fig 1: Pairwise comparison in Super Decisions software

Fig 2 :
Fig 2: The conceptual model of DRS in Super Decisions software

Table 1 :
Models and frameworks of resilience risk and vulnerability to hazards, level and diversity of economic resources, and equity in resource distribution); social capital (received social support, perceived social support, social embeddedness (informal ties), organizational linkage and cooperation, citizen participation, leadership and roles (formal ties), sense of community, and attachment to place); community competence (community action, critical reflection and problem-solving skills, flexibility and creativity, collective efficacy empowerment, and political partnership); information and communication (narratives, responsible media, skills and infrastructure, and trusted sources of information) Buckle (2006)ibutes (health, income, age, gender, skills, network, and lifestyle choices); infrastructure status (coverage, accessibility, and reliability); community attributes (networks, amenities, and facilities); economic status and trends (growth or decline, employment levels, and innovation); demographic status and trends (age structure, immigration, and gender balance); environmental status (sustainability, diversity, and pollution); geographic attributes (remoteness, topography, and weather)Buckle (2006)shared community values, knowledge of hazards, established social infrastructure, positive socioeconomic trends, partnerships, resources, and skills.CSIRO(2007) metabolic flows (production, supply, and consumption chains); governance networks (institutional structures and organizations); social dynamics (demographics, human capital, and inequity); built environment (ecosystem services in urban landscapes) Mayunga (2007) Social capital (trust, norms, and networks); economic capital (income, savings, and investment); human capital (education, health, skills, and knowledge/information); physical capital (housing, public facilities, and business/industry); natural capital (resource stocks, land and water, and ecosystem) Twigg (2009) Governance (policy, planning priorities and political commitment, legal and regulatory systems, integration, institutional mechanisms, capacities and structures, allocation of responsibilities, partnerships, and accountability and community participation); risk assessment (hazards/risk data and assessment, vulnerability and impact data and assessment, and scientific and technical capacities and innovation); knowledge and education (public awareness, knowledge and skills, information management and sharing, education and training, cultures, attitudes, motivation, and learning and research); risk management and vulnerability reduction (environmental and natural resource management, health and wellbeing, sustainable livelihoods, social protection, financial instruments, physical protection, structural and technical measures, and planning regimes); disaster preparedness and response (organizational capacities and coordination, early warning systems, preparedness and contingency planning, emergency resources and infrastructure, emergency response and recovery, participation, voluntarism, and accountability).values-cohesion,and faith-based organization); economic (employment, value of property, wealth generation, and municipal finance); institutional (participation, hazard mitigation plans, emergency services, zoning and building standards, emergency response plans, interoperable communications, and COOP); infrastructure (lifelines and critical infrastructure, transportation network, residential housing stock and age, and commercial and manufacturing establishments); community competence (local understanding and risk, counseling services, absence of psychopathologies, health and wellness, and quality of life).leadershipand politics, structural and societal changes, physical location, age, attitudinal factors, health, income, gender, long-term commitment, social networks, government policies, short-term recovery, capable agencies, reaccumulation of capital, and long-term rehabilitation Wilson (2013)Proposed the use of government channels to promote community-based disaster prevention, coordinate the timing of policy implementation, define the rights and responsibilities of various stakeholders, and control funding to ensure proper implementation of community-based

Table 2 :
Proposed dimensions and components of DRS

Table 3 :
Urban areas weighted supermatrix Table3illustrates the weighted supermatrix which is the result of pairwise comparison of components.This table also represents the connection or disconnection of components with each other and the impacts of each component on the other components.The impact of each component on the other components is shown in each column.It can be seen that potential hazards and exposure, natural capital, and geographic attributes of urban areas have the highest impacts on the other components.The comprehensive emergency plan (CEP) is also one of the vital components of urban areas, which is under the influence of most components.Charts 1 to 6 illustrate overall synthesized priorities of urban areas for the dimensions which have been represented in 3 forms; ideal, normal, and raw which are based on three different methods by idealzing or normalizing the values.The first normal priority in the economic dimension is municipal finance.Sustainable livelihood and economic growth are, respectively, in the second and third rank (Chart 1).Sustainability, natural capital, and land use are the priorities of the environmental dimension (Chart 2).As can be observed in Chart 3, the infrastructural priorities are critical infrastructure, lifelines, and public facilities which are close to each other.The CEP is the priority of the planning and institutional dimension.Governance networks and institutional mechanisms, capacities, and structures are the Chart 4).Community competence and social network are also close in socio-cultural dimension and demographics are the subsequent priority (Chart 5).Chart 6 illustrates that industrial and commercial units and residential units are in the highest normal priority.

Table 4 :
Prioritized components of resilience in urban areas