1 Introduction
Against the backdrop of accelerating global urbanization, population and various economic activities are inevitably highly concentrated in cities, and urban disasters are frequent leading to increasingly prominent urban safety issues (
Lu et al., 2022). Global climate change has triggered a variety of extreme weather phenomena which have become a major challenge that threatens the safety of residents’ lives and property and affects the sustainable development of cities (
Weilnhammer et al., 2021). Therefore, improving urban disaster prevention and mitigation capacity is crucial to sustainable urban development, and the study of social vulnerability to urban disasters provides a new perspective for the cause of urban disaster prevention and mitigation.
In the context of the new urbanization strategy, it has become an important issue in urban geography to explore how to achieve sustainable urban development and build a, “safe and livable”, urban living environment from the perspective of vulnerability. With the widely accepted view that, “vulnerability originates from human beings themselves” (
Zhan et al., 2018), more scholars have begun to pay attention to the impact of human social systems on vulnerability, and in this context, social vulnerability has become a key direction in the study of vulnerability to urban disasters. Derived from anthropology and social ecology, traditional social vulnerability research aims to understand the mechanisms of resilience and vulnerability of disaster-prone groups, emphasizing that social vulnerability is to some extent a product of social inequality, focusing on analyzing which social factors shape or influence the disaster sensitivity of various groups of people, and affecting their response capacity. With the deepening of the research, the single idea of characterizing the attributes of vulnerable populations can no longer meet the needs of regional social vulnerability assessment. Social vulnerability is not only related to the physical and socio-economic attribute characteristics of the population, but also closely related to the background environment of social life where people are located, such as social networks, the level of urban public service facilities, building structure type, population density, government organization factors, etc. Therefore, it has been a consensus among scholars to carry out urban social vulnerability research in an integrated dimension by taking urban space as a carrier and considering the social, architectural and institutional environmental characteristics of the region where people are located.
Social vulnerability assessment is still in the exploratory stage and lacks a unified assessment model (Table 1). As the simplest algorithm to quantify social vulnerability, the social vulnerability index method proposed by Cutter has been widely used (
Cutter, 2003;
Cutter, 2024). This method can effectively reveal the spatial and temporal patterns of social vulnerability in a region, identify the most vulnerable units or populations, is applicable to different geographic environments and spatial and temporal scales, and is characterized by its simplicity of thinking and methodology (
Gu et al., 2018;
Majumder et al., 2023;
Han et al., 2024;
Qi et al., 2024). However, the one-dimensionalized assessment values allow for over-generalization of multidimensional information on social vulnerability, and key factors of regionally differentiated vulnerability are not adequately captured. Urban activities and spatial development have significant non-equilibrium, leading to heterogeneous characteristics of social vulnerability traits in different regions. Under the influence of factors such as the pattern of patchy territorial landscapes, urban functional zoning, and socio-spatial difference, as a multidimensional, multilevel system attribute characteristic of urban disaster social vulnerability. With the complexity of the composition of vulnerability elements and the diversity of the combination of different geographical spatial vulnerability elements, there is a differentiated social vulnerability characteristics within the urban space, showing obvious spatial heterogeneity characteristics (
Xiao et al., 2022;
Zhang et al., 2022;
Qin et al., 2023). As a simple example, the vulnerability of ecologically fragile areas within cities due to blind urban expansion and encroachment on ecologically fragile areas is significantly different from the vulnerability induced by high-density population agglomeration in urban core areas in terms of vulnerability mechanisms and characteristics. The traditional one-dimensional quantitative assessment approach is likely to group the two together as a class of highly fragile areas, whose vulnerability is not adequately reflected. In addition, scale is an important topic in geographic research, and the social vulnerability of urban disasters is characterized by an obvious spatial scale dependence. The social vulnerability of urban disasters has different manifestations or characteristics at different spatial scales. Conducting vulnerability studies at the urban and regional scales is conducive to grasping the vulnerability dynamics of the country and the region, which in turn is conducive to providing references to the layout of urban or regional resources and facilities for disaster prevention and relief at the macro-strategic level (
Chen et al., 2012;
Liu and Liang, 2014;
Wen et al., 2016). However, macro-scale vulnerability studies cannot reflect the changes of social vulnerability within the small scale (or continuous space), and it is difficult to provide effective information for accurate disaster prevention, mitigation, and post-disaster recovery and reconstruction. In view of the background characteristics of spatial variability of social vulnerability in different areas of a city or a region, and taking into account the current reality of the need for urban management, it is urgent to carry out urban disaster vulnerability studies on the street or community level (
Richmond et al., 2015;
Tu et al., 2023;
Wu et al., 2024), thus helping to deepen people’s understanding of urban disaster risk analysis and key elements and processes (
Yin et al., 2020;
Huang et al., 2023;
Yin et al., 2024).
Traditional vulnerability assessment methods, such as the composite index method and the social vulnerability index, are easy to operate and have a wide range of applicability through simplification and weighted summation to produce a single vulnerability index. However, this method tends to over-generalize multidimensional information, ignores spatial heterogeneity and regional differences, and is difficult to reflect the specific changes in small-scale scales, and it has limitations in capturing the complexity of social vulnerability traits and in supporting fine-grained management decision-making. Compared with traditional vulnerability assessment, spatial heterogeneity assessment not only emphasizes the differences within and between regions, taking into account factors such as geographic location, environmental conditions and infrastructure, but also builds a more complex multidimensional indicator system from multiple dimensions to reflect the complexity and diversity of social vulnerability, while providing new perspectives and paths for the sustainable development of social systems. Spatial heterogeneity assessment focuses more on spatial changes at the micro level and, by refining spatial units, develops more targeted risk management and disaster mitigation measures for specific regions and improves the effectiveness and precision of policies. Vulnerability assessment based on spatial heterogeneity emphasizes the cross-fertilization of multiple fields, such as geography, ecology, and sociology, and recognizes and takes into account changes in vulnerability over time and space, enabling decision makers to adopt more flexible and adaptive strategies to deal with various potential threats.
This paper tries to start from the idea of social vulnerability research, considering the multidimensional attribute characteristics of urban social vulnerability and the limitations of the existing quantitative assessment means, based on the perspective of spatial heterogeneity, integrating the vulnerability research theory into urban geography research. It attempts to construct a kind of assessment idea and method that can show the spatial differentiation pattern of urban social vulnerability by combining the quantitative analysis with the qualitative analysis. The empirical study is carried out in Beijing as an example, in order to identify the spatial characteristics of urban social vulnerability in Beijing, explore the hotspots of urban social vulnerability and their main influencing factors, and put forward the differentiated strategies for urban social vulnerability regulation, to provide scientific references and decision-making support for urban disaster risk management and sustainable development.
2 Methodology
2.1 Construction of a framework for assessing the heterogeneity of urban disaster social vulnerability
At present, scholars adopt a variety of indirect methods to conduct urban disaster social vulnerability research. There are many factors affecting social vulnerability, including rapid population growth, low levels of education, health status, beliefs and practices, unbalanced gender ratios, low socio-economic levels, the built environment, social infrastructure, and lack of access to resources and services (
Zhou et al., 2014). Currently, commonly accepted factors include age, gender, employment, and socio-economic status (
Hung et al., 2024;
Stolte et al., 2024). Higher economic conditions can significantly improve the ability of the population to cope with and withstand disasters, with higher-income groups tending to have more capital and being able to rebuild and recover more quickly after a disaster, while lower-income groups are more vulnerable to disasters because of their limited resources, thus increasing the vulnerability of society as a whole (
Blaikie et al., 2004). Most vulnerability studies take the natural elements as a starting point and give less consideration to the elements of the built environment. Therefore, based on the understanding of the connotation of social vulnerability, this paper establishes a social vulnerability assessment index system from two aspects, namely, socio-economic factors and factors of the built environment, and elaborates on the influence mechanism of regional social vulnerability (Table 2).
2.2 Interpretative Structural Modeling (ISM) - dimensionality reduction of vulnerability evaluation indicators
Interpretative Structural Modeling (ISM) was first applied in the field of systems engineering research. The model believes that there is no complete parallelism between the influencing factors that trigger social events, there are intrinsic links between many factors, and there is also a causal relationship between different factors (
Teng et al., 2022). ISM can be used to effectively transform the complex relationship between the influencing factors into a direct and clear structural relationship, taking into account the complex coupling characteristics of vulnerability, on the basis of emphasizing the independence of the evaluation indicators, it is also necessary to carry out a complementary analysis of their correlation relationship. This study adopts the explanatory structural modeling method to extract the core indicators of vulnerability from the qualitative dimension based on the identification of the hierarchical relationship of the vulnerability indicator system and the internal conduction process, and based on the correlation relationship among the indicators, as a supplementary analysis of the quantitative method. The specific steps are as follows.
1) Determine the relevant statistical indicator variables according to the social vulnerability indicator system, adopt the method of expert consultation to qualitatively assess the role of the relationship between the indicator variables, and construct the system topology diagram of the indicator system (Fig. 1).
2) Based on the neighboring relationship between the indicator elements, the relationship matrix of the evaluation indicators is further constructed. Utilizing the Boolean algorithm to find out the reachability matrix between the elements and separating it into independent subsystems or decomposing it into hierarchical subsystems, and finally obtaining the recursive structural model can reveal the hierarchical relationship between the elements of the evaluation indicators (Fig. 2). Among them, the top-level elements can be regarded as impact indicators that directly act on the level of social vulnerability, as candidates for core indicators of social vulnerability under the qualitative analysis dimension. Combined with the candidate indicators selected in the aforementioned quantitative way, the core indicators of social vulnerability for spatial clustering are determined.
2.3 Spatial clustering
The spatial clustering method for multidimensional attribute elements was constructed using the grouping analysis tool developed by ArcGIS. When the spatial-temporal constraints of the clustering units are specified, the system provides a more convenient solution by finding the optimal classification mode through the connectivity graph (minimum spanning tree). When spatial constraints are specified, the system uses the SKATER (Spatial “K” luster Analysis by Tree Edge Removal) method to cluster the analyzed units.
For the problem of determining the optimal number of classification groups in traditional clustering, the grouping analysis tool constructs the Calinski-Harabasz pseudo F-statistic (
Wang et al., 2006) that measures the validity of assessing units when there are as many as 15 clustering groups (Fig. 3). This is a ratio that reflects the similarity within groups and the dissimilarity between groups, where the maximum value corresponds to the number of classes as the optimal clustering scheme. On the basis of determining the optimal classification results, by outputting the parallel box plots of the distribution of each clustering unit in the multi-attribute fields, the multi-attribute aggregation features of the clustering units can be identified and named, and also can be analyzed in comparison with the results of the traditional clustering analysis, to determine the final clustering scheme, which is used for the identification of the characteristics of the spatial differentiation pattern of urban social vulnerability.
3 Beijing urban disaster social vulnerability assessment
3.1 Construction of Beijing street-scale urban disaster social vulnerability assessment index system
On the basis of the urban disaster social vulnerability index system, this study selects the following vulnerability evaluation indicators based on the principles of science, comprehensiveness, relevance, feasibility and simplicity, and taking into account the availability, updatability, and exhaustiveness of the relevant data for the analysis of social vulnerability of Beijing’s urban street scale (Table 3).
Data source: Beijing Statistical Yearbook 2020, Beijing Regional Statistical Yearbook, Statistical Bulletin on National Economic and Social Development, Beijing 6th Population Census Townships and Streets Volume, Beijing Government Data Network, Gaode Map POI data.
From the level of spatial clustering method, the attribute characteristics of the clustering unit should be as representative as possible, and should not be too much. The spatial clustering research on the social vulnerability of disasters at the street scale in Beijing needs to select and reduce the dimensionality of the indicator system to be constructed. Combined with the explanatory structural modeling method constructed in this study for the construction of the hierarchical relationship of the indicator system, based on the correlation relationship between the indicators (Table 4), the core indicators of vulnerability are extracted from the qualitative dimension as a complementary analysis to the aforementioned quantitative method. As can be seen in Fig. 3, there is a correlation between the sex ratio and some indicators, and there is a strong significance between the population density, family size, construction land, and other indicators, so this study finally identified eight indicators for spatial clustering study, including the proportion of mobile population, dependency coefficient, proportion of higher education, the density of the road network for avoiding hazards, the density of hazardous point sources, the lighting index, the number of 10000 people’s hospitals, and the proportion of ecological land.
3.2 Spatial clustering analysis of urban disaster social vulnerability in Beijing
Based on the selected social vulnerability indicators, spatialization of the data was carried out, and a preliminary analysis of the spatial differentiation pattern of social vulnerability to disasters in Beijing was conducted at the level of each indicator (Fig. 4).
First, from the perspective of mobile population, the spatial differentiation pattern of mobile population at the street scale level in Beijing is relatively obvious. With the main mobile population concentrated between the third and fourth ring roads of Beijing, the mobile population in the inner-city areas and the peri-urban areas is at a medium level, and the proportion of the mobile population in the remote suburban counties is low. The population dependency coefficient, on the other hand, shows an opposite pattern, with higher levels in the six inner-city districts and suburban districts, and relatively lower levels in the far suburban districts. As for the population with higher education, the proportion of inner-city areas is higher, and the proportion of remote suburban counties is mostly less than −1.5 standard deviation, except for the central town streets in some districts and counties, indicating that the population with higher education has obvious spatial differences at the street level in Beijing. The escape routes used for urban disaster risk avoidance, on the other hand, present a higher density of road network in the far suburban counties, while the density in the near suburban counties is not high, and the northern region is better than the southern region, so the road infrastructure construction in the near suburban areas and the southern region of Beijing needs to be further strengthened. In terms of the sources of risk within the city, the key areas of concern still lie in the peri-urban areas, while the risk level in the remote suburban counties is low, and the inner-city areas are at an average level, which is also consistent with the pattern of disasters and accidents in Beijing in recent years. For example, most of the fires occurred in the peri-urban areas, which have a larger proportion of mobile population, and there are more fire hazards in these areas. In addition, from the viewpoint of urban construction development, the lighting index is a good indicator, so in the process of factor screening, this indicator is retained, while the proportion of construction land is excluded. As can be seen from the figure, the intensity of Beijing's lighting index gradually moves outward from the inner city, forming an obvious circle structure, with the intensity of lighting gradually decreasing, and significantly lower than −2.5 standard deviations in the northern mountainous area and the western mountainous area. In addition, medical facilities are more abundant in the inner-city area, while the situation of some townships in the remote suburban counties is not optimistic, which will exacerbate the social vulnerability to disasters in these areas. In terms of ecological protection land use, there is a staggered distribution of high-value and low-value areas, and the proportion of ecological land use is low in most areas. Ecological land use plays an important role in reducing the risk of urban disasters and improving the quality of the urban environment, so this indicator is also a key indicator that affects the level of social vulnerability to disasters in Beijing. Combined with the previous analysis, at the street scale level, there are significant differences in the key factors of social vulnerability faced by different streets, so it is necessary to categorize the process, rather than simply combining the indicators, and determine the level of social vulnerability of the street by the value of the vulnerability index.
The spatial clustering analysis of the above indicators divides the disaster social vulnerability of Beijing street-scale units into five categories. To better identify the main features of social vulnerability of each category, a wind rose diagram (Fig. 5) is made for the eight indicators of each category, which facilitates the comparative analysis, and because of the obvious differences in the data of each indicator in terms of the scale of the indicators, the indicators are first processed without the scale of the indicators. From the figure, it can be seen that the spatial unit of type 1 is at a relatively low level in indicators such as the density of roads to avoid danger and the number of hospitals for 10,000 people, and the level of the light index is not high, and these areas belong to the remote suburbs and counties of Beijing, and the infrastructure support is relatively weak, so the spatial unit of this type is named as infrastructurally vulnerable (IFV). The spatial unit of type 2 has relatively high values of the light index and the proportion of the mobile population, and this area belongs to the near suburbs of Beijing, so the type unit is named as migrant population spatial agglomeration vulnerability (MPSAV). Type 3 spatial unit has a high level in the proportion of ecological land use and the density of risk point sources, most of these areas are important ecological protection zones in Beijing, and the distribution of geologic hazards and other disaster risk sources in the area is relatively large, so the type unit is named as environmentally and ecologically vulnerable area (EEVA). Type 4 spatial unit is similar to the spatial unit of type 2, with a high level of lighting index and proportion of mobile population, but the proportion of population with higher education in this spatial unit is significantly higher than that of type 2, which indicates that the overall education and culture level of the population in this area is higher than that of the second spatial type. Additionally the area is located in the northern part of the core urban area of Beijing, which has a high intensity of development and construction and is densely populated, so the type of spatial unit is named as spatial development agglomeration vulnerability (SDAV). Type 5 spatial unit, from the viewpoint of spatial location, is mainly located in the core area of the old city, in addition, from the viewpoint of the indicator level, the population dependency coefficient of the region is higher, and the level of mobile population is relatively low, and the proportion of the population with higher education is not high, which indicates that the population vitality in the region is relatively weak, and in addition, in fact, there is a serious population aging and other demographic problems, so they are named as demographic structure decline vulnerability (DSDV).
Although this study summarizes the different types of zones, in terms of type designation, which summarizes the main vulnerability characteristics of these areas, it should be noted that there are other potential vulnerability factors in different types of areas, as well. IFV, the proportion of ecological land use is not high, and the population dependency coefficient is also a vulnerability factor that needs to be paid attention to in this type of area. MPSAV, with a low proportion of highly educated population and a low density of sheltered roads, also becomes a potential factor affecting this type of zone. EEVA, with a low proportion of highly educated population, indicates that the region, in addition to strengthening ecological environmental protection and management, how to implement and popularize disaster prevention and mitigation knowledge is also an issue that needs to be focused on. DSDV, in addition to the structural problems of the population, the region's high-intensity construction of man-made environments has led to a low level of land for ecological protection, so efforts need to be made to improve the quality of the human environment. SDAV, although the main problem of the region is high-intensity spatial development, the number of sources of disaster risk in the region is high, but the density of roads to avoid danger is relatively low, so within this type of zone, in addition to controlling the intensity of spatial development and construction, there is a greater need to improve the safety of the built environment in the region, so as to reduce the vulnerability type.
Through the spatial clustering study, the urban disaster social vulnerability traits of different areas in Beijing can be identified from the perspective of spatial heterogeneity. From the perspective of quantitative distribution (Table 5), MPSAV has the largest number of 99 streets, followed by IFV with 75 streets, and DSDV is also a more prominent problem, with 65 street units in Beijing as well. In addition, in terms of the spatial extent of the area (Fig. 6), IFV has the largest spatial extent, accounting for about half of the total area of the area, followed by MPSAV, EEVA has the middle range of distribution, DSDV and SDAV have a relatively small geographic distribution range. Therefore, the main core problems faced by Beijing are the large-scale mobile population and the uneven urban development and construction, and the problems of intra-regional population development and the fragile ecological environment should not be ignored.
4 Conclusions and discussion
4.1 Conclusions
On the basis of sorting out the urban disaster social vulnerability indicator system, based on the principles of scientificity, comprehensiveness, relevance, feasibility and conciseness, and considering the availability, updatability and exhaustiveness of the relevant data and information for the analysis of the social vulnerability of the urban street scale in Beijing, 12 vulnerability evaluation indicators are selected at the levels of the socio-economic environment and the built facility environment. Before conducting a spatial clustering research on the street scale level of the Beijing, it is necessary to select and downsize the indicator system to be constructed. Combined with the explanatory structural modeling method to construct the hierarchical relationship of the indicator system, based on the correlation relationship between the indicators, this study finally determines eight indicators for spatial clustering, including the proportion of migrant population, dependency coefficient, proportion of higher education, the density of the road network, the density of the hazardous point sources, the lighting index, the number of hospitals per 10000 people, and the proportion of the ecological land use.
Based on the selected social vulnerability indicators, spatialization of the data was carried out to obtain the distribution pattern map of street disaster social vulnerability indicators at each indicator level, and to obtain the characteristics of the divergence pattern of the 8 indicators. A spatial clustering analysis was carried out for the 8 indicators to classify the disaster social vulnerability of the street scale unit in Beijing into 5 categories, and wind rose diagrams were made for each category of the 8 indicators, so that the main characteristics of social vulnerability of each category could be better identified. Type 1 is infrastructurally vulnerable, Type 2 is migrant population spatial agglomeration vulnerability; Type 3 is environmentally and ecologically vulnerable area; Type 4 is spatial development agglomeration vulnerability; and Type 5 is demographic structure decline vulnerability. At the same time, the existence and potential vulnerability factors in different types of districts were identified, and the urban disaster social vulnerability traits of different areas in Beijing were identified from the perspective of spatial heterogeneity. In terms of quantitative distribution, there are 99 streets with the highest number of mobile population agglomeration vulnerability units, followed by 75 streets with infrastructure weakness vulnerability units, and 65 streets with demographic decline vulnerability units. In addition, in terms of spatial scope, infrastructure weakness vulnerability has the largest spatial scope, accounting for about half of the total area of the district, followed by mobile population agglomeration vulnerability, and the ecological environment vulnerability is in the middle of the distribution scope, while the demographic decline vulnerability and the spatial development agglomeration have a relatively small geographic distribution scope. Therefore, the main core problems faced by Beijing are the large-scale mobile population and the uneven urban development and construction issues, and the problems of intra-regional population development and the fragile ecological environment should not be ignored.
4.2 Discussion
Starting from the idea of social vulnerability research and based on the perspective of spatial heterogeneity, this study constructs an assessment idea and method that can show the spatial differentiation pattern of urban social vulnerability. Through the research method combining quantitative analysis and qualitative analysis, an empirical study was conducted with Beijing as an example, revealing the spatial heterogeneity pattern of urban disaster social vulnerability characteristics in Beijing. The study makes up for the shortcomings of traditional vulnerability assessment methods and provides an important reference for urban planning, disaster risk management, and social vulnerability governance. In this paper, administrative units are used as the basic unit for spatial clustering analysis, which is conducive to the effective articulation and implementation of policy measures. Although the shape and size of administrative units may vary depending on factors such as historical background and geographic conditions, which may lead to some bias in the clustering process, given that this study focuses on the assessment of urban social vulnerability, it will have no effect on urban plains and little impact on the results of mountainous areas.
Future research can be carried out in the following areas to further advance the field of urban social vulnerability. First, the applicability and universality of the assessment method can be further promoted and verified, and it can be applied to more cities to provide wider reference and reference for social vulnerability research in different cities. Second, the accuracy and resolution of the data can be improved by increasing the spatial and temporal resolution of the data using, for example, advanced geographic information systems (GIS) and remote sensing technology, to reflect more accurately the characteristics and changes in urban social vulnerability. This will improve the credibility and accuracy of the assessment results. Thirdly, a multidimensional and multiscale urban vulnerability assessment framework can be explored, taking into account multiple factors, such as natural, social and economic factors, in order to gain a more comprehensive understanding of the risks and vulnerability issues facing cities. Such a research approach can provide deeper insights and more comprehensive decision support for urban planning and disaster management. Finally, there is a need to develop appropriate policy and planning measures to improve the coping capacity and resilience of urban social vulnerability and to promote sustainable urban development and social safety. The research results should be fully communicated with government departments and relevant stakeholders to provide a scientific basis for policy formulation and decision-making. Meanwhile, the research results can also be applied in the fields of urban planning, disaster management, and public safety to provide practical support for urban disaster prevention, mitigation, and emergency response. Through sustained research efforts, we can further deepen our understanding of the nature and characteristics of urban social vulnerability and provide a more scientific basis for urban planning and disaster management. This will provide important reference and guidance for us to better understand and respond to the challenges of urban social vulnerability.
PDF
(3494KB)
