Identification of the Key Influencing Factors of Urban Rail Transit Station Resilience against Disasters Caused by Rainstorms

Abstract

Improving the ability of the urban rail transit system to cope with rainstorm disasters is of great significance to ensure the safe travel of residents. In this study, a model of the hierarchical relationship of the influencing factors is constructed from the resilience perspective, in order to research the action mechanisms of the influencing factors of urban rail transit stations susceptible to rainstorm disaster. Firstly, the concept of resilience and the three attributes (resistance, recovery, and adaptability) are interpreted. Based on the relevant literature, 20 influencing factors are discerned from the 3 attributes of the resilience of urban rail transit stations. Subsequently, an interpretative structural model (ISM) is utilised to analyse the hierarchical relationship among the influencing factors. The key influencing factors of station resilience are screened out using social network analysis (SNA). Combined with ISM and SNA for analysis, the result shows that the key influencing factors are: “Flood prevention monitoring capability”; “Water blocking capacity”; “Flood prevention capital investment”; “Personnel cooperation ability”; “Emergency plan for flood prevention”; “Flood prevention training and drill”; “Publicity and education of flood prevention knowledge”; and “Regional economic development level”. Therefore, according to the critical influencing factors and the action path of the resilience influencing factors, station managers can carry out corresponding flood control work, providing a reference for enhancing the resilience of urban rail transit stations.

1. Introduction

In China, there have been significant developments due to the constantly prosperous economy and unprecedented urbanisation, making the traffic congestion problem of large cities gradually severe [1]. Nonetheless, urban rail transit, with its advantages of size, speed, comfort, affordable pricing, and environmental benefits [2,3], can sufficiently alleviate the traffic congestion of megacities [4]. Therefore, in the past decade, China has increasingly built more urban rail transit lines. According to the China Association of Metros data in 2020, the route length of urban rail transit in mainland China reached 7978.19 km [5], which is 2.21 times larger than that of Germany, the second-longest route length in the world, measuring 3604.16 km [6].

However, urban rail transit, especially its underground stations, are more vulnerable to disasters caused by heavy rain. Recently, due to extreme weather conditions, such as the frequent occurrence of rainstorms, China has been most seriously affected by rainstorm disasters [7]. Rainstorms are the most common and disaster-prone natural disasters that have the most significant impact, with the broadest range and the highest frequency, on the urban transportation system and have caused substantial economic losses and severe security threats to urban development and public life [4,8]. For instance, on 28 June 2018, the Guangfu station of the Chengdu Metro was hit by heavy rain, resulting in water intrusion into the station, and the station was forced to stop operation. On 21 May 2020, due to heavy rainfall in Guangzhou, the flood water invaded many stations along Guangzhou Metro Line 13, resulting in the shutdown of Line 13, which has affected passengers’ travel. On 20 July 2021, a rare extreme rainstorm occurred, whereby flood water poured onto Zhengzhou Metro Line 5, resulting in a train being besieged by the flood. Following an intense rescue mission, 14 passengers did not survive the disaster.

Therefore, identifying the key influencing factors of the resilience of urban rail transit stations against disasters caused by rainstorms plays a significant role in ensuring the regular operation of the station, which can help station managers to take corresponding protective measures. Previous studies have researched the influencing factors of the resilience of urban rail transit stations against rainstorm disasters from different perspectives, such as comprehensive evaluation, scenario simulation analysis, and flood control measures analysis. For example, Lyu et al. [9] conducted a flood risk assessment on the Guangzhou metro system considering 12 influencing factors, including rainfall, elevation, slope, and metro line density; Yu et al. [10] evaluated the waterlogging risk of Chengdu Metro Line 4 from 6 influencing factors, such as the drainage pipe network, surrounding terrain, and regional situation; and Lyu et al. [11] observed 16 influencing factors, including rainfall, slope, land use type, and river density, to evaluate the waterlogging risk of the Shenzhen metro system. Furthermore, Wang et al. [4] considered 20 influencing factors, such as land subsidence, rainstorm intensity, subway line density, entrance height, and passenger flow, to evaluate the flood risk level of the Beijing subway system. In order to assess the flood risk of the Shanghai metro system, 19 influencing factors, such as the green area ratio, type of exit, drainage capacity, and GDP were considered in the work of Lyu et al. [12]. In the scenario simulation analytical type, Quan et al. [13] focused on the impact of a rainstorm in different return periods on the Shanghai Metro. Lyu et al. [14] analysed the impact of different rainfall intensities on the Shanghai central metro system. Forero-Ortiz et al. [15] simulated the flood water depth on the impact of transport service continuity of Barcelona Metro Line 3. In terms of the analytical type of flood control measures, Aoki et al. [16] analysed the impact of different flood control equipment on the subway station’s flood control capacity and considered the entrance to the underground station as the most affected part of the station by rainstorms.

The above research results mainly analyse the flood risk analysis of urban rail transit stations, primarily focusing on external influencing factors, such as rainfall intensity, entrance attributes, land subsidence, and surrounding terrain, to research the impact of flood disasters on urban rail transit stations. However, there is a lack of research that considers or identifies the relationship of the critical influencing factors of urban rail transit systems. Urban rail transit is a complex system affected by external factors and internal subsystems, making the factors affecting the resilience of urban rail transit stations against rainstorms numerous and complicated [17]. Therefore, it is essential to study its action path and key influencing factors to better analyse the impact of rainstorms on urban rail transit underground stations.

Furthermore, most of the existing studies focus on the stations vulnerable or unable to withstand the adverse effects of rainstorms [18]. Recently, resilience theory is a significant theory in the field of disaster prevention and mitigation, which can be used to explore the ability of a system to resist, recover, and adapt to natural disasters [19,20,21,22], and has become a hotspot in the field of disaster prevention and mitigation in urban systems [23]. Therefore, enhancing the resilience of urban rail transit stations can improve the disaster-bearing capacity and self-recovery of urban rail transit systems when exposed to rainstorms, to improve the reliability and safety of urban rail transit system operations.

However, there are few studies on the resilience of urban rail transit systems against rainstorms. Thus, from the perspective of resilience, this study aims to identify and analyse the critical influencing factors of the resilience of urban rail transit systems against rainstorms by combing the interpretative structure model method (ISM) and social network analysis method (SNA), to provide a reference for improving the ability of the station to deal with disasters caused by rainstorms.

The subsequent parts of this paper are organised as follows: Section 2 introduces the concept of resilience and determines the influencing factors of the station’s resilience; Section 3 details the methods of ISM and SNA; in Section 4, ISM and SNA are applied for empirical analysis; and, finally, the conclusion is obtained in Section 5.

2. Selection of the Influencing Factors

2.1. Resilience

“Resilience” originated from “resilio” in Latin, which means “returning to the original state”. Later, English borrowed the word and it evolved into “resile” [24]. In 1973, the ecologist Holling [25] applied, foremost, resilience theory in system ecology to express the idea that the ecosystem can operate normally or restore itself to a state of equilibrium after being disturbed by the external world. Since then, the concept of resilience has gradually extended from the ecological field to other fields, including engineering, society, and medicine [26,27,28]. However, researchers in different fields have different definitions of resilience and emphasise different aspects of its multiple meanings, and the details are shown in

Table 1. Definition of resilience in various fields.

Institution/Author	Year	Subject Area	Definition of Resilience	Emphasis
Pimm [29]	1984	Ecological	Considered that resilience is the speed at which the system returns to equilibrium following a perturbation.	Return to equilibrium.
Klein et al. [30]	1998	Engineering	The self-organising capacity to maintain the actual and latent functions in the face of constant changes in the outside environment.	Maintain the actual and potential functions.
Paton et al. [31]	2001	Society	The system’s ability to cope with challenges and changes in the face of external disruptions to maintain its regular operation.	Cope with challenges and changes, and maintain its normal operation.
Godschalk et al. [32]	2003	Engineering	The sustainable network of physical systems and human communities.	The sustainable network.
Pickett et al. [33]	2003	Ecological	The ability of a system to return to its stable equilibrium point after disturbance.	Return to their stable equilibrium point.
Campanella et al. [34]	2006	Society	The capacity of a city to rebound from destruction.	Rebound.
Resilience Alliance [35]	2007	Ecological	The ability of the urban system to digest and absorb external interference, maintain the initial structure, and maintain crucial functions.	Digest, absorb, and maintain.
Cutter et al. [36]	2008	Society	The properties of a social system in the event of a disaster: the ability to cope with the disaster, the ability to recover from the disaster, and the ability to adapt after the disaster.	Resistance, recovery, and adaption.
Intergovernmental Panel on Climate Change (IPCC) [37]	2012	Disaster	The ability of a system and its parts to anticipate, absorb, accommodate, or recover from the effects of a perilous event in a timely and efficient manner, including through ensuring the preservation, restoration, or enhancement of its essential basic structures and performs.	Absorb, accommodate, recover, preserve, and restore.
Brown et al. [38]	2012	Society	The capacity of an individual, community, or institution to dynamically and effectively respond to shifting climate circumstances while continuing to function at an acceptable level.	Respond to shifting climate circumstances, continuing to function at an acceptable level.
Meerow et al. [23]	2016	Urban	The ability of a system to maintain or rapidly return to desired functions in the face of a disturbance, adapt to change, and quickly transform systems.	Maintain, return, adapt, and transform.
Hoterová et al. [20]	2021	Transport	The ability of a system to resist, absorb, and adapt to adverse events.	Resist, absorb, and adapt.	Resist, absorb, and adapt.
Chen et al. [19]	2021	Urban	The ability of a system to resist, absorb, adapt, and recover from danger in time to hedge its impact when confronted with external shocks.	Resist, absorb, adapt, and recover.

.

In recent years, based on the concept of resilience, the application of resilience in infrastructure has gradually become a research hotspot, such as urban resilience, transport resilience, and urban rail transit resilience. Huang et al. [39] studied the critical influencing factors on urban resilience and its affecting mechanisms in China. Chen et al. [21] constructed a new resilience model to study urban resilience in Taiwan, by collecting empirical data from Taiwan during the Morakot. Based on the “scale–density–morphology” resilience framework, Feng et al. [40] evaluated the urban resilience of Shenyang city. Testa et al. [41] assessed the resilience of transportation networks against extreme events by applying topological graph properties, by using the transportation network of the New York City metropolitan area as a case study. Zhang and Wang [42] utilised the WIPW as the network performance metric for relieving the risk of surface road transportation networks from the perspective of resilience. Sohouenou and Neves [43] offered an assessment skeleton for road network resilience to research the impacts of link–repair sequences on the resilience of road networks. Considering resilience properties, Leobons et al. [44] proposed a framework of indicators for measuring the resilience of urban transportation systems. For urban rail transit resilience, Li et al. [2] evaluated the network resilience of the Beijing subway by accounting for the measurement model of passenger flow and network topological structure. Taking four subway station projects of Chengdu Metro Line 11 as the research object, Liu et al. [22] adopted the projection pursuit method to measure the resilience of subway station projects against waterlogging disasters. These scholars considered resilience as the ability of a system to maintain its stability in the face of external disturbances and to quickly recover from damage.

However, there is no consensus on the concept of resilience and most scholars and institutions agree that resilience encompasses or is associated with the meaning of resistance, recovery, and adaptability [19,20,21,22]. Therefore, the resilience of urban rail transit systems against rainstorms is also considered, in the present study, from the perspective of the three aspects. It is defined as follows: when suffering from unpredictable and abrupt catastrophes, the urban rail transit system can bear and absorb the harmful effects that are the result of the disaster, to preserve the primary function of the system to the maximum extent. Through the self-healing, studying, and organisation of the system, the system can be restored to a normal state; moreover, in the face of similar disasters in the future, the system will suffer minor damage and recover faster, that is to say, the system will show more vital adaptability.

Therefore, this paper summarises the influencing factors of urban rail transit station resilience against rainstorms in the context of the three attributes of resilience.

2.2. Three Attributes of the Station’s Resilience

2.2.1. Resistance

Resistance usually refers to the external invasion resistance and immunity of the human body. The application of resistance in disaster science is considered as the system’s ability to withstand and absorb the disturbance of natural catastrophes [20,22,45]. On account of this, the resistance of urban rail transit stations refers to the ability of the station to absorb the adverse effects of rainstorms to guarantee the station’s essential functions are not damaged to the greatest extent.

The entrance is the passageway connecting the underground station and the ground. In most cases, the flood water enters the station from the entrance. Therefore, different types of entrances and the height of the steps have different roles in the resistance of the invasion of the water. Moreover, the higher the green coverage around the station, the stronger the soil’s water storage capacity and the stronger the effect of reducing the rainwater collection speed. By enhancing the drainage capacity of the municipal pipe network, the ponding around the station can be drained away rapidly, reducing the probability of flood disasters around the entrance and reducing the flood prevention pressure of the station. Good flood prevention monitoring measures can effectively monitor the waterlogging around the entrance of the station: once the danger of flooding is detected, the staff can set the water blocking board, sandbags, and other flood prevention materials at the entrance to carry out the water blocking work, ensuring the dryness and cleanliness of the station. Furthermore, when the flood danger becomes severe, the number of casualties can be reduced by implementing effective evacuation methods.

2.2.2. Recovery

The original meaning of “recovery” refers to the process by which a spring returns to its normal state after being stressed. According to the Chinese dictionary, the meaning of “recovery” is the restoration of a system to its previous or stable state. Therefore, the recovery of urban rail transit stations refers to the prompt restoration of an unbalanced state to an anticipated stable state when the station is damaged. The “stable state” can be the primary state or a new equilibrium state [21,22,45]. It emphasises that the station can operate smoothly and normally after recovery.

When encountering ponding, the station will enter a state of recovery. At this time, the station’s drainage equipment plays a significant role. The stronger the drainage capacity, the faster the speed of cleaning up the ponding. The number of staff, the proficiency level of staff, and the coordination ability also accelerate the station’s recovery rate. The power supply and communication systems’ guaranteed capability can assist the staff’s flood prevention work and ensure maintenance during the recovery period. The investment in flood prevention capital can not only replace the damaged station equipment in time, but can also provide flood control subsidies and rewards to the staff to promote the rapid implementation of recovery work.

2.2.3. Adaptability

Adaptability can be understood as “adaptation”, which is interpreted in the Chinese dictionary as the pleasant state in which the individual can adjust to various environments, ways, times, people, and things. Moreover, Andresen and Gronau [46] considered adaptability as the capability to modify one’s structure to adjust to changes. Furthermore, Thieken et al. [45] considered adaptability as a system’s capacity to study one-time events and adjust the status. It can go beyond the previous state of facing disasters.

On account of the definition mentioned above, the adaptability of urban rail transit stations is understood as the ability of the station to adjust its structure through self-healing, learning, and organisation when resisting and absorbing the impact of rainstorms to better cope with different levels of severity of the rainstorms in the future.

Establishing an appropriate flood prevention emergency plan can result in the rapid enforcement of flood prevention arrangements in the station according to the existing provisions when facing flood disasters, or adjustments, according to similar plans, to improve the station’s adaptability to the rainstorm. Through the training of flood control, the staff can find the loopholes in the flood prevention work and make timely corrections and adjustments. Through the publicity and education of flood prevention methods, the staff can improve the attitude of individuals toward flood prevention and distinguish the priorities of flood prevention projects. Additionally, the rail passengers can actively cooperate with the flood prevention work. The maintenance of the station can keep the stable state of flood prevention equipment when encountering flood danger, so that the flood prevention work of the station can be carried out smoothly. The development of the regional economy provides financial support for improving the flood prevention facilities of the urban system, which not only enhance the flood prevention capacity of the area around the station, but also strengthen the implementation of the flood prevention measures of the station. These factors play a crucial role in the station’s first response to the disasters following a rainstorm and play a significant part in future flood prevention work. Therefore, under the influence of these factors, the station has a richer flood prevention experience and faster response when facing similar rainstorms, which make the station more adaptable.

Based on the understanding, as explained above, of the factors that influence the resilience of urban rail transit stations, this paper selected 20 factors that influence a station’s resilience against rainstorms, which is presented in

Table 2. Influencing factors of the station’s resilience.

Aspects	Number	Influencing Factor	Description	Source
Resistance	F1	Type of entrance	Based on the location and the external structure of the entrance, the types of the entrance are divided into the hidden entrance, enclosed entrance, semi-enclosed entrance, and open entrance.	[12]
	F2	Height of entrance steps	Their specific values can represent the height of the entrance steps. A larger value symbolises the more robust ability to resist rainstorms.	[12,47]
	F3	Flood prevention monitoring capability	Flood prevention monitoring capability means the possibility that the relevant departments forecast the occurrence of a rainstorm by using equipment, such as hydrometeorological receiving equipment and disaster monitoring equipment.	[4]
	F4	Water blocking capacity	Water blocking capacity is determined by constructing the anti-retaining plate, waterproof groove, and civil enclosure at the entrance.	[4,48]
	F5	Drainage capacity of municipal pipe network	Drainage facilities determine the drainage capacity of the municipal pipe network outside the station; the higher standard of drainage facilities means a stronger drainage ability.	[10,49]
	F6	Emergency material reserve	Emergency materials include a dedicated power supply, emergency light, radio walkie-talkie, portable escalator, flood protection shovel, and, water baffle.	[22,50]
	F7	Evacuation efficiency	Evacuation efficiency means completing personnel evacuation within the unit time while facing a rainstorm.	[47,49]
	F8	Surrounding green coverage	Surrounding green coverage is the situation of greening with the unit area.	[12]
Recovery	F9	Station drainage capacity	Station drainage capacity is determined by drainage facilities, such as roof drainage and the drainage ditch at the entrance.	[12,22]
	F10	Number of flood prevention staff	The number of flood prevention staff can directly reflect the recovery ability of the station through the efficiency of rescue.	[50]
	F11	Staff proficiency	The staff proficiency is mainly determined by the education and years of work experience of related staff, which can also directly reflect the proficiency of the station’s ability to recover through the efficiency of rescue.	[22,51]
	F12	Flood prevention capital investment	The flood prevention capital investment can directly reflect the recovery ability of the station through the efficiency of rescue from the perspective of revenue.	[17,48]
	F13	Power supply system guarantee capability	Power supply system guarantee capability is mainly determined by the power generator and backup power, which can directly reflect the recovery ability of the station through the efficiency of rescue from the perspective of equipment.	[52,53]
	F14	Communication system guarantee capability	Communication system guarantee capability can be assessed by the speed of receiving instructions, which can directly reflect the recovery ability of the station through the efficiency of rescue from the perspective of equipment.	[51]
	F15	Personnel cooperation ability	Personnel cooperation ability can directly reflect the recovery ability of the station through the efficiency of rescue from the perspective of individuals.	[17]
Adaptability	F16	Emergency plan for flood prevention	The emergency plan for flood prevention contains an emergency evacuation plan, personnel deployment plan, material distribution plan, and rescue implementation rules. It can help improve the adaptability of the station during the occurence of rainstorms.	[17,22]
	F17	Flood prevention training and drill	Flood prevention training and drill can help relevant departments address rainstorms in an orderly manner.	[22]
	F18	Publicity and education of flood prevention knowledge	Publicity and education of flood prevention knowledge can promote the comprehensive quality of staff to improve the adaptability from the perspective of individuals.	[48,54]
	F19	Regional economic development level	The regional economic development level indirectly determines the investment into and equipment supplied to the metro system. The higher regional economic development level means a greater possibility of adapting to rainstorms.	[4,12,55]
	F20	Maintenance of flood prevention equipment	Maintenance of flood prevention equipment can reduce the possibility of further damage caused by equipment while facing rainstorms to promote adaptability.	[48,54]

.

3. Methodology

3.1. Interpretative Structural Modelling (ISM)

3.1.1. Conceptual Interpretation of ISM

Interpretative structural modelling (ISM) is a method used to transform complex and fuzzy models into well-defined models, and was first proposed and applied by Warfield [55] in 1973. This method has been widely applied to analyse the interactive relationship between factors with a hierarchical structure [56,57]. For example, using the integrated ISM-MICMAC model, Kinker et al. [57] built the structural model of service quality barriers to understand the interrelation between service quality barriers. Ali et al. [58] utilised the ISM approach to identify the contextual relationships between the barriers that are imposed by learning six sigma implementations in the supply chain. Patel et al. [59] used mixed ISM and fuzzy MICMAC to describe the contextual relationships and interrelationships between circular economy enablers. ISM can describe the essence of the system better than manifestation modes, such as text, tables, and formulas [60]. The interrelationships between the factors affecting informal learning were studied in the work of Suresh and Balajee [61], by employing the ISM method. By referring to these studies, this paper will adopt the method to hierarchise the factors that influence the resilience of urban rail transit stations against rainstorms, by employing the following four steps.

3.1.2. Construction of the ISM Model

Determine the Pair-Wise Relationship between the Influencing Factors

Identifying the pair-wise relationship is the first step for applying ISM, which is usually assessed based on expert opinions by conducting semi-structured interviews. All the experts are invited to judge whether factor i affects factor j with four symbols (V, A, X, and O).

(a)

V indicates that factor i affects factor j;

(b)

A indicates that factor j affects factor i;

(c)

X indicates that there is the mutual impact of factors i and j;

(d)

O indicates that there is no impact between factors i and j.

Based on the collected answers from experts, the structural self-interaction matrix (SSIM) is established to reflect the pair-wise relationship between the factors.

Calculate the Reachability Matrix

The developed SSIM is then transferred into a binary matrix called the initial reachability matrix (IRM) for further calculation by converting the four symbols (V, A, X, and O) of the SSIM to 0 and 1. The conversion rules are given in

Table 3. The rules for conversion.

SSIM	(i,j)Entry	(j,i)Entry
V	1	0
A	0	1
X	1	1
O	0	0

.

The initial reachability matrix (IRM) only displays the direct relationships between the influencing factors. A final reachability matrix M can present both direct and indirect relationships, usually attained by employing the transitivity rule [62,63]. The transitivity rule indicates that if factor i affects factor j, and factor j affects factor k, then i virtually affects k. The matrix M can be obtained by adding the transitivity to the IRM through the Boolean operation, which involves the self-multiplication of the matrix until it reaches a stable state. Software, such as MATLAB, can help this calculation process for generating matrix M.

Level Partitions Analysis of the Reachability Matrix

In order to establish the hierarchical structure model of influencing factors, it is necessary to determine the level of each factor in the hierarchical structure. To do this effectively, the antecedent set T(F_i), reachability set R(F_i), and intersection set R(F_i) ∩ T(F_i) of each influence factor must first be identified. The reachability set R(F_i) is the set of all factors affected by factor F_i, which is obtained by the column factor with the value of “1” in the corresponding row of factor F_i; the antecedent set T(F_i) represents the set of all factors affecting factor F_i, which is obtained by the row factor with the value of “1” in the corresponding column of factor F_i; and the intersection R(F_i) ∩ T(F_i) represents the set of factors that affect each other, which is obtained by the intersection of the antecedent set T(F_i) and reachability set R(F_i) [64,65].

According to the ISM principle, an intersection set should be placed on the top factors with the same reachability set. In other words, factors at the top level are always affected by other factors. Once the factors on the topmost level are identified, they will be discarded until all the RFs have been partitioned in the ISM structure.

Establishment of the Interpretation Structure Model

Based on the result of the level partitions, a directed hierarchical structure diagram can be constructed to present the chain of influence of each factor. The first layer of influencing factors is at the top of the structure diagram.

3.2. Social Network Analysis (SNA)

3.2.1. Conceptual Interpretation of SNA

Social network analysis (SNA) is a relational data research method based on network structure analysis [66]. SNA can describe the relationship between nodes in a network composed of nodes and edges, explaining the relationship between nodes and corresponding quantitative indicators [67]. That is to say, based on the node relationship data, applying network quantitative indicators and relationship network topology, SNA can study the structural characteristics of the network and the status of nodes in the network [68]. Therefore, SNA provides an analytical method for understanding the behavioural changes of the interactions between actors, and a unique insight into the understanding and development of the research of network structure analysis theory [69].

Currently, SNA is widely applied in sociology, economics, ecology, and other fields. For instance, Prell et al. [70] distinguished the central stakeholders in natural resource management by employing the SNA method: Marks et al. [71] utilised SNA to identify the critical staff in child care centres for obesity preventions; Hauck et al. [72] used SNA to recognise the critical stakeholders for agricultural biodiversity governance; and Mbaru and Barnes [73] utilised SNA to identify the key players for four types of natural resource conservation objectives.

There is a relatively complex correlation between the factors that influence the resilience of urban rail transit stations against rainstorms, and each interactive influencing factor constitutes a factor network [74]. Therefore, considering the influencing factors of a station’s resilience as nodes in the network, and regarding the relationship between the influencing factors of a station’s resilience as the connecting line of interaction between the nodes, SNA metrics, like degree centrality and closeness centrality, is adopted to identify the key influencing factors of a station’s resilience against rainstorms, in the present paper. In this network structure, the relationship between the influencing factors includes being affected by other factors and influencing other factors, which are all considered as a connection between two influencing factors.

3.2.2. SNA Metrics

Degree Centrality

In a network, the connection strength between a node and other nodes is different, indicating that its level of importance is also different. At this point, the importance of a node in the entire network can be measured using degree centrality. The larger the degree centrality value, the more nodes associated with this node, and the more critical this node is in the network [68,69]. Currently, for comparative analysis, normalised degree centrality is regarded as a more standard calculation form [75], and its calculation formula is:

$$C_{\mathrm{D}}\left(N_i\right)=\frac{{\displaystyle \sum _{j=1}^{n}e(N_j,N_i)}}{n-1}$$

(1)

where C_D(N_i) denotes the normalised degree centrality of node i; e(N_j, N_i) denotes the number of connections between node i and node j, it is equal to 1 if node i is connected to node j; and n is the number of nodes in the network, where n − 1 denotes the maximum probable number of connections for the node in the network.

Closeness Centrality

In a network, there can be a case in which one node has a high degree of centrality, showing a solid connection with the surrounding nodes but a faint connection with other nodes. Therefore, degree centrality cannot fully describe the role played by a node in the entire network.

Considering the indirect relationship between nodes, closeness centrality is used to describe the accessibility and convenience of a node in the entire network [68]. It is obtained by calculating the sum of the shortest distance between this node and other nodes. The distance is expressed here as the number of edges between two nodes. If the distance between a node and other nodes is relatively short, the node has a high closeness centrality. In other words, the smaller the closeness centrality of a node, the closer the connection with other nodes in the network, and the closer the node is to the centre of the network [69]. In order to facilitate the comparison, relatively compact centrality is also used to calculate the value, and its calculation formula is as follows:

$$C_{\mathrm{C}}\left(N_i\right)=\frac{n-1}{{\displaystyle \sum _{j=1}^{n}d\left(N_i,N_j\right)}}$$

(2)

where C_C(N_i) denotes the reciprocal of the close centrality of node i. The larger the value, the smaller the close centrality of the point, and the more the point is in the centre of the network. Moreover, d(N_j, N_i) denotes the minimum number of edges connecting node i and node j; n is the number of nodes in the network, where n − 1 denotes the maximum probable number of connections for the node in the network.

4. Empirical Analysis

4.1. The Analysis of the ISM Results

4.1.1. Determine the Pair-Wise Relationship between the Influencing Factors

According to Section 3.1, semi-structured interviews with experts can be conducted to determine the interrelationships between these twenty influencing factors. As opined by Shen et al. [76], ISM emphasises the quality of respondents rather than their quantities. It does not require a large sample of respondents for the data survey. In order to ensure the experts are sufficient, this study selected 20 experts from the fields of academia, business, and government with strong experience and knowledge of urban rail transit stations. Furthermore, the research group conducted a pre-survey with these experts to understand whether or not they were familiar with these twenty factors. Finally, five experts were selected to participate in the semi-structured interviews. A few studies also adopted five experts for the purpose of analysis and showed reliable results [62,63]. The profiles of these five experts are presented in

Table 4. Background information of the experts.

Expert	Occupation	Education	Work Experience (Years)	Employer
A	Station manager	Masters deree	9	Chongqing Rail Transit (Group) Co., Ltd., Chongqing, China
B	Station operator	Bachelors degree	8	Chongqing Rail Transit (Group) Co., Ltd., Chongqing, China
C	Officer	Masters degree	10	Chongqing Urban Administration Bureau
D	Professor	Doctorate	12	Chongqing University, China
E	Professor	Doctorate	7	Southwest University, China

.

Based on the feedback of whether factor i affects factor j with the four symbols (V, A, X, and O), according to the principle in Section 3.1, the structural self-interaction matrix (SSIM) was generated. When the experts presented significant differences in the relationship between one pair of influencing factors, this paper applied the “minority gives way to the majority” principle to determine the interrelationships [62]. The structural self-interaction matrix (SSIM) of the factors influencing the resilience level of the station is shown in

Table 5. The structural self-interaction matrix (SSIM).

	F1	F2	F3	F4	F5	F6	F7	F8	F9	F10	F11	F12	F13	F14	F15	F16	F17	F18	F19	F20
F1	-	O	O	V	O	O	V	O	O	O	O	O	O	O	O	O	O	O	O	O
F2			O	V	O	O	V	O	O	O	O	O	O	O	O	O	O	O	O	O
F3				V	O	O	V	O	V	A	A	A	A	A	A	A	A	O	O	A
F4					A	A	O	O	O	A	A	A	A	A	A	A	A	O	O	A
F5						O	O	A	O	O	O	O	O	O	O	O	O	O	A	O
F6							O	O	V	O	O	A	O	O	O	A	A	O	O	O
F7								O	O	A	A	O	A	A	A	A	A	A	O	O
F8									O	O	O	O	O	O	O	O	O	O	A	O
F9										A	A	A	A	A	A	A	A	O	O	A
F10											A	A	O	O	X	A	A	O	O	O
F11												O	O	O	X	A	A	O	O	O
F12													V	V	O	X	V	X	A	V
F13														O	A	A	A	O	O	A
F14															A	A	A	O	O	A
F15																A	A	A	O	O
F16																	X	A	A	V
F17																		A	A	V
F18																			A	O
F19																				O
F20																				-

.

4.1.2. Calculate the Reachability Matrix

According to the SSIM, an initial reachability matrix (IRM) can be developed using the conversion rules presented in Table 3, which is shown in

Table 6. The initial reachability matrix (IRM).

	F1	F2	F3	F4	F5	F6	F7	F8	F9	F10	F11	F12	F13	F14	F15	F16	F17	F18	F19	F20
F1	0	0	0	1	0	0	1	0	0	0	0	0	0	0	0	0	0	0	0	0
F2	0	0	0	1	0	0	1	0	1	0	0	0	0	0	0	0	0	0	0	0
F3	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0
F4	0	0	0	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0
F5	0	0	0	1	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0	0
F6	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0
F7	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0
F8	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0
F9	0	0	1	1	0	0	1	0	1	0	0	0	0	0	1	0	0	0	0	0
F10	0	0	1	1	0	0	1	0	1	1	0	0	0	0	1	0	0	0	0	0
F11	0	0	1	1	0	1	0	0	1	1	0	0	1	1	0	1	1	1	0	1
F12	0	0	1	1	0	0	1	0	1	0	0	0	0	0	0	0	0	0	0	0
F13	0	0	1	1	0	0	1	0	1	0	0	0	0	0	0	0	0	0	0	0
F14	0	0	1	1	0	0	1	0	1	1	1	0	1	1	0	0	0	0	0	0
F15	0	0	1	1	0	1	1	0	1	1	1	1	1	1	1	0	1	0	0	1
F16	0	0	1	1	0	1	1	0	1	1	1	0	1	1	1	1	0	0	0	1
F17	0	0	0	0	0	0	1	0	0	0	0	1	0	0	1	1	1	0	0	0
F18	0	0	0	0	1	0	0	1	0	0	0	1	0	0	0	1	1	1	0	0
F19	0	0	1	1	0	0	0	0	1	0	0	0	1	1	0	0	0	0	0	0
F20	0	0	0	1	0	0	1	0	0	0	0	0	0	0	0	0	0	0	0	0

.

Based on the IRM, this study applies the MATLAB R2014a software to conduct the Boolean operation, to calculate the final reachability matrix M. The result is shown in

Table 7. The final reachability matrix M.

	F1	F2	F3	F4	F5	F6	F7	F8	F9	F10	F11	F12	F13	F14	F15	F16	F17	F18	F19	F20
F1	1	0	0	1	0	0	1	0	0	0	0	0	0	0	0	0	0	0	0	0
F2	0	1	0	1	0	0	1	0	0	0	0	0	0	0	0	0	0	0	0	0
F3	0	0	1	1	0	0	1	0	1	0	0	0	0	0	0	0	0	0	0	0
F4	0	0	0	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0
F5	0	0	0	1	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0
F6	0	0	0	1	0	1	0	0	1	0	0	0	0	0	0	0	0	0	0	0
F7	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0	0	0	0
F8	0	0	0	1	1	0	0	1	0	0	0	0	0	0	0	0	0	0	0	0
F9	0	0	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0	0
F10	0	0	1	1	0	0	1	0	1	1	1	0	1	1	1	0	0	0	0	0
F11	0	0	1	1	0	0	1	0	1	1	1	0	1	1	1	0	0	0	0	0
F12	0	0	1	1	0	1	1	0	1	1	1	1	1	1	1	1	1	1	0	1
F13	0	0	1	1	0	0	1	0	1	0	0	0	1	0	0	0	0	0	0	0
F14	0	0	1	1	0	0	1	0	1	0	0	0	0	1	0	0	0	0	0	0
F15	0	0	1	1	0	0	1	0	1	1	1	0	1	1	1	0	0	0	0	0
F16	0	0	1	1	0	1	1	0	1	1	1	1	1	1	1	1	1	1	0	1
F17	0	0	1	1	0	1	1	0	1	1	1	1	1	1	1	1	1	1	0	1
F18	0	0	1	1	0	1	1	0	1	1	1	1	1	1	1	1	1	1	0	1
F19	0	0	1	1	1	1	1	1	1	1	1	1	1	1	1	1	1	1	1	1
F20	0	0	1	1	0	0	1	0	1	0	0	0	1	1	0	0	0	0	0	1

.

4.1.3. Level Partitions of the Reachability Matrix

Based on the principle presented in Section 3.1, the antecedent set T(F_i), reachability set R(F_i), and intersection set R(F_i) ∩ T(F_i) of each influence factor can be determined. Take F₄ as an instance, where the reachability set R(F₄) = {F₄}; the antecedent set T(F₄) = {F₁, F₂, F₃, F₄, F₅, F₆, F₈, F₁₀, F₁₁, F₁₂, F₁₃, F₁₄, F₁₅, F₁₆, F₁₇, F₁₈, F₁₉, F₂₀}; and the intersection set R(F_i) ∩ T(F_i) = {F₄}. Likewise, the reachability sets, antecedent sets, and intersection sets for other influence factors can be obtained, as shown in

Table 8. Relationship set and the hierarchical division of the influencing factors.

Factor: Fi	Reachability Set: R(Fi)	Antecedent Set: T(Fi)	Intersection Set: R(Fi) ∩ T(Fi)	Level: Li
F1	1,4,7	1	1	2
F2	2,4,7	2	2	2
F3	3,4,7,9	3,10,11,12,13,14,15,16,17,18,19,20	3	2
F4	4	1,2,3,4,5,6,8,10,11,12,13,14,15,16,17,18,19,20	4	1
F5	4,5	5,8,19	5	2
F6	4,6,9	6,12,16,17,18,19	6	2
F7	7	1,2,3,7,10,11,12,13,14,15,16,17,18,19,20	7	1
F8	4,5,8	8,19	8	3
F9	9	3,6,9,10,11,12,13,14,15,16,17,18,19,20	9	1
F10	3,4,7,9,10,11,13,14,15	10,11,12,15,16,17,18,19	10,11,15	4
F11	3,4,7,9,10,11,13,14,15	10,11,12,15,16,17,18,19	10,11,15	4
F12	3,4,6,7,9,10,11,12,13,14,15,16,17,18,20	12,16,17,18,19	12,16,17,18	5
F13	3,4,7,9,13	10,11,12,13,15,16,17,18,19,20	13	3
F14	3,4,7,9,14	10,11,12,14,15,16,17,18,19,20	14	3
F15	3,4,7,9,10,11,13,14,15	10,11,12,15,16,17,18,19	10,11,15	4
F16	3,4,6,7,9,10,11,12,13,14,15,16,17,18,20	12,16,17,18,19	12,16,17,18	5
F17	3,4,6,7,9,10,11,12,13,14,15,16,17,18,20	12,16,17,18,19	12,16,17,18	5
F18	3,4,6,7,9,10,11,12,13,14,15,16,17,18,20	12,16,17,18,19	12,16,17,18	5
F19	3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20	19	19	6
F20	3,4,7,9, 13,14,20	12,16,17,18,19,20	20	4

.

The principle of “R(F_i) ∩ T(F_i) = R(F_i)” is applied to assign the factors of each level. For example, F₄, F₇, and F₉ in Table 8 meet this principle. Thus, these three factors should be placed on Level I. By removing these three factors, F₁, F₂, F₃, F₅, and F₆ are in line with the principle, and they are partitioned to Level 2. The same procedure is repeated to identify the level partitions with the rest of the factors. Table 8 presents the results of the level partitions of the twenty factors.

4.1.4. Establishment of the Interpretation Structure Model

Using the level partitions of each influencing factor in Table 8, the hierarchical structure of each influencing factor can be drawn, as shown in Figure 1. Some of the cross-level influences can be transmitted via the middle-level factors, so this paper makes a corresponding merger and simplifies some cross-level influence paths.

4.1.5. The Results of the ISM

On account of the hierarchy analysis of each influencing factor, presented in Figure 1, the “Evacuation efficiency” (F₇), “Water blocking capacity” (F₄), and “Station drainage capacity” (F₉) is on the first level, which has the most direct impact on the resilience of the station. The bottom influencing factor is the “Regional economic development level” (F₁₉), which can directly or indirectly affect the other factors, and has a far-reaching impact on the resilience of the station. It can be regarded as the source influencing factor of the resilience of the station. The factors ranging from the second to fifth level, in the middle layer, cannot directly affect the resilience of the station; however, they can directly affect the factors on the top level, and can also transfer the influence of the factors on the bottom level to the top level to play a part in connecting the bottom top levels.

4.2. The Analysis of the SNA Results

According to Table 5, the relationship between the factors influencing the station’s resilience is determined. Hence, using the relationship data in Table 5, the network structure of the factors influencing the station’s resilience can be drawn using Gephi software, as shown in Figure 2. In the figure, the red node represents the resistance factor, the green node represents the recovery factor, and the blue node represents the adaptability factor. Additionally, the node’s size indicates the degree centrality value, and the larger the point, the greater the degree centrality value.

Calculated using Formulas (1) and (2), Pajek software is applied to calculate each node’s degree centrality and closeness centrality in the factor network. For instance, take F₁, where the sum of e(N_j, N₁) for F₁ (node 1) is equal to 2, n is 20, and the sum of d(N_j, N₁) for F₁ (node 1) is equal to 39, thus:

C_D(N₁) = 2/(20 − 1) = 0.1053,
C_C(N₁) = (20 − 1)/39 = 0.4872.

The specific results for all the factors are shown in

Table 9. The normalised degree centrality of each influencing factor.

Attribute	Factors	Degree Centrality		Closeness Centrality
		CD(Ni)	Rank	CC(Ni)	Rank
Resistance	F1	0.1053	6	0.4872	6
	F2	0.1053	6	0.4872	6
	F3	0.6316	2	0.7037	2
	F4	0.7368	1	0.7917	1
	F5	0.1579	5	0.5135	5
	F6	0.2632	4	0.5588	4
	F7	0.5789	3	0.6551	3
	F8	0.1053	6	0.3958	8
Recovery	F9	0.5789	3	0.6129	6
	F10	0.5263	4	0.6333	3
	F11	0.4737	5	0.6129	6
	F12	0.7368	1	0.7308	1
	F13	0.4737	5	0.6333	3
	F14	0.4737	5	0.6333	3
	F15	0.6842	2	0.6786	2
Adaptability	F16	0.8947	1	0.8261	1
	F17	0.8421	2	0.8261	1
	F18	0.3684	4	0.5938	4
	F19	0.3158	5	0.5588	5
	F20	0.4211	3	0.6129	3

. Since there are three attributes of resilience influencing factors, this paper sorted the three attributes, respectively, to select the key influencing factors of each attribute.

Degree centrality indicates the importance of a node in the network, and relative centrality indicates how closely a node is related to other nodes. A node can be considered as a critical node in a network if it is essential and closely related to the other nodes. Therefore, by combining the degree centrality and closeness centrality, presented in Table 9, this paper selects the top two influencing factors from the three attributes, respectively, as the key influencing factors of the station’s resilience.

In the ranking of resistance attributes, the top two influencing factors ranked by degree centrality are “Water blocking capacity” (F₄) (C_D(N₄) = 0.7368) and “Flood prevention monitoring capability” (F₃) (C_D(N₃) = 0.6316). Subsequently, in the ranking of recovery attributes, the top two influencing factors ranked by degree centrality are “Flood prevention capital investment” (F₁₂) (C_D(N₁₂) = 0.7368) and “Personnel cooperation ability” (F₁₅) (C_D(N₁₅) = 0.6842). Finally, in the ranking of adaptability attributes, the top two influencing factors ranked by degree centrality are “Emergency plan for flood prevention” (F₁₆) (C_D(N₁₆) = 0.8947) and “Flood prevention training and drill” (F₁₇) (C_D(N₁₇) = 0.8421).

Likewise, in terms of closeness centrality, the selected influencing factors from the three attributes are consistent with those selected according to degree centrality, indicating that they are both critical and quite compactly related to the other nodes.

4.3. Discussion

According to the results of the ISM, the hierarchical structure of the influencing factors is sorted into six levels, as shown in Figure 1. Combining the ISM and SNA results, the comparison of the importance in the same level can be carried out, as shown in Figure 3.

4.3.1. The Critical Influencing Factors for Level I

As shown in Figure 3a, the water blocking capacity, evacuation efficiency, and station drainage capacity belong to the first level, which can directly influence the resilience of urban rail transit stations when facing rainstorms. The water blocking capacity stands in the most significant place due to its highest degree of centrality and closeness centrality, followed by evacuation efficiency and station drainage capacity, with a little distance. Water blocking capacity and station drainage capacity belong to the category of drainage facility, which is considered as the most crucial factor in assessing the resilience of cities. At the same time, rainstorms and flood disasters occur [77]. Considering the necessity and uncertainty of individuals, evacuation efficiency is proposed to be assessed and managed from population sensitivity [78]. Therefore, emphasising the construction and improvement of water blocking facilities can directly and efficiently promote the resilience of the urban rail transit systems against rainstorm disasters.

4.3.2. The Critical Influencing Factors for Level II

The results in Figure 3b indicate that there are five factors in level II: type of entrance, the height of the entrance steps, flood prevention monitoring capability, the drainage capacity of municipal pipe network, and emergency material reserve, which all belong to the dimension of resistance. Based on the SNA results, flood prevention monitoring capability is the first. The emergency material reserve and drainage capacity of municipal pipe networks, the type of entrance, and the height of entrance steps are the last. Flood prevention monitoring capability plays a role in the initial stage of a disaster as the result of a rainstorm d. The emergency material reserve is conducive to the process of rainstorms occurring, promoting resilience from addressing rainstorms. The other three factors can be summarised as the external engineering of urban rail transits, which directly impacts the resilience of urban rail transits during the occurrence of rainstorms [11,12]. Therefore, more attention should be paid to what departments in the urban rail transit system can address the disaster induced by rainstorms. In contrast, the engineering construction of urban rail transit stations has met the standards.

4.3.3. The Critical Influencing Factors for Level III

As shown in Figure 3c, surrounding green coverage, power supply system guarantee capability, and communication system guarantee capability are on level III. It can be easily found that power supply system guarantee capability and communication system guarantee capability are more important than surrounding green coverage. However, the power supply system and communication system guarantee capabilities play a protective role when a rainstorm affects the station, and the impact on the resilience of the station is not so direct. Showing that the same reason with level II, power supply system guarantee capability and communication system guarantee capability work in the process of dealing with damage caused by rainstorms. However, surrounding green coverage is conducive to the prevention of urban rail transit stations from being affected by rainstorms, with a relative lower importance to evaluate the risk of floods as a result of rainstorms [77,78,79]. Similarly, more care should be taken to reduce the damage to urban rail transit stations caused by rainstorms. Therefore, the influencing factors of Level III are relatively less important for the station’s resilience.

4.3.4. The Critical Influencing Factors for Level IV

As shown in Figure 3d, there are four factors on this level: the number of flood prevention staff, staff proficiency, personnel cooperation ability, and maintenance of flood prevention equipment. Except for the maintenance of flood prevention equipment, the other three factors are all in the dimension of recovery. The four factors’ degree centrality and closeness centrality values are relatively higher and closer, indicating the greater importance and a minor difference in this level. The number of flood prevention staff, staff proficiency, and personnel cooperation ability reflect that individuals significantly impact the resilience of urban rail transit stations against rainstorms. However, it is not the most direct factor [77]. Therefore, strengthening staff management and training seems to be conducive to promoting the resilience of urban rail transit stations [80].

4.3.5. The Critical Influencing Factors for Level V

Figure 3e demonstrates four factors on level V: flood prevention capital investment, emergency plan for flood prevention, flood prevention training and drill, publicity and the education of flood prevention knowledge. At this level, flood prevention capital investment, emergency plans for flood prevention, and flood prevention training and drill are good performers in the degree centrality and closeness centrality, playing a pivotal role in reducing flash floods [81]. Although the degree centrality value and closeness centrality value of publicity and the education of flood prevention knowledge seem to be relatively lower, well-trained individuals equipped with the correct knowledge about disaster prevention can help others respond appropriately during disasters and prevent life-threatening results [82]. Hence, investment, emergency plans, training, and education cannot be ignored.

4.3.6. The Critical Influencing Factors for Level VI

As shown in Figure 3f, only the regional economic development level belongs to this level. According to the ISM results, regional economic development, at the bottom of the hierarchical structure, is the most indirect but essential factor. Economic and social development is closely related to various human activities [83], which also plays a critical role in disaster prevention and mitigation [78]. Therefore, strengthening the development of the economy and society can significantly contribute to the promotion of the resilience of urban rail transit stations.

5. Conclusions

In general, with the background of disasters caused by rainstorms, the ISM model was applied to hierarchically classify the intricate influencing factors of the station’s resilience to present an explicit, concrete, and intuitive network structure. Meanwhile, the SNA model was utilised to select the key influencing factors of the station’s resilience from the network of influencing factors. On account of the three attributes of resilience, including resistance, recovery, and adaptability, as well as the critical influencing factors of each level in the ISM, this paper ultimately selects eight critical influencing factors of the station’s resilience against rainstorms, which include: “Flood prevention monitoring capability” (F₃); “Water blocking capacity” (F₄); “Flood prevention capital investment” (F₁₂); “Personnel cooperation ability” (F₁₅); “Emergency plan for flood prevention” (F₁₆); “Flood prevention training and drill” (F₁₇); “Publicity and education of flood prevention knowledge “ (F₁₈); and “ Regional economic development level “ (F₁₉).

This study provides a feasible solution that constructs the hierarchical structure of influencing factors for the station’s resilience by adopting the ISM method. This research provides significant references to help the station managers recognise the action path and mechanism of the factors influencing its resilience. In addition, this study presents a workable solution that screens out the key influencing factors of the station’s resilience by employing SNA, which allows the station’s managers to recognise which influencing factors require more attention. Therefore, this paper offers a reference for station managers to carry out targeted flood control work, which also delivers an idea to reinforce the resilience of urban rail transit stations to rainstorm disasters. Nonetheless, the limitations to this study are acknowledged because the generalisation of the data caused by methodology forced us to select five experts for analysis. In the future, the research team will collect more research data from quantifying the impact of factors on the resilience of urban rail transit stations when facing rainstorms or other disasters.