China is in a rapid urbanization stage: in the next 20 years, more than 60% of the population will live in the nation’s cities (Fang, 2009). With the concentration of the population and the economy, cities must manage the higher frequencies and risks of various hazards and are becoming more vulnerable. In addition, new-type urbanization requires the reduction of urban vulnerability and the improvement of urbanization quality. Therefore, research on urban vulnerability is an important scientific tool to explore the harmonious coexistence of the human-environmental systems, and urban vulnerability has become a key topic that needs to be strengthened in the forefront of research.

The concept of “vulnerability” originated from studies on natural disasters (Janssena et al., 2006). In the 1970s, White (1974) and Burton (1978) introduced this concept in their studies of natural disasters. Later, in the 1980s, this concept was expanded to research on starvation and food security, poverty, and development. For example, Sen (1981) adopted this concept in research on starvation and food security, and Chambers (1989) and Swift (1989) introduced it into the research field of poverty and development. In the 1990s, researchers concentrating on natural disasters began to pay attention to the effect of human activities on global environmental change, in particular, their effect on vulnerability to climate change. Liverman (1990) used the concept of “vulnerability” in research on global environmental change. During this period, a rapidly increasing number of studies in the field of human dimensions of global environmental change focused on vulnerability (Janssena et al., 2006; Janssen, 2007). In addition, research on vulnerability to climate change plays a pivotal role in the development of vulnerability research. Scholars have persistently tracked the studies of the Intergovernmental Panel on Climate Change (IPCC) on vulnerability to climate change and have analyzed the concept of vulnerability proposed by the IPCC (Watson et al., 1996; Watson et al., 1998; McCarthy et al., 2001). Moreover, research on poverty was also advanced during this period, which not only considered poverty caused by low income according to the traditional definition but also extended the research scope to the “poverty” of developmental capacity (Roberts and Yang, 2003). Referencing a power-losing theory emerging in food security research, developmental economic scholars conducted studies on vulnerability associated with sustainable livelihoods and poverty (Adger, 2006). In the 21st century, the vulnerability of coupled systems, particularly the coupled human-environment system and the coupled socio-ecological system, began to attract attention, leading to numerous studies on vulnerability on a specific spatial scale of coupled systems. In general, evolution of the “vulnerability” concept resulted from the processes of content extension and multidisciplinary comprehensive integration (Kelly and Adger, 2000). In particular, “vulnerability” in research on the human-environment system has become a collective concept (Newell et al., 2005), which includes the components of risk, fragility, marginalization, natural disaster, sensitivity, adaption and response, adaptive potential, resilience, etc. (Birkmann, 2007; Liverman, 2000). Humane studies have extended this concept to “vulnerability” in the pure sense of economic social systems (Adger et al., 2004; Adrianto and Matsuda, 2002). Related studies have emphasized the identification of vulnerable populations or groups and vulnerable regions and have focused on intraregional or interregional comparative analyses of vulnerability under the same disturbance background. These studies have generally been more focused on the economic, political, management system, and cultural factors that easily affect either human societies or groups, along with the sociocultural mechanisms and countermeasures that can be used to restore the resilience of both natural environmental systems and economic social systems (Li and Zhang, 2011).

Until now, domestic and international scholars have conducted both multidimensional and multilevel studies on urban vulnerability in different fields. In addition to a traditional risk-hazard research viewpoint, urban vulnerability research has achieved much in various fields, including sustainability science, socioecological resilience research, adaptability science, and complex human-environmental systems. The vulnerability studies in different fields are associated and have overlapped with each other, forming the core of vulnerability research. Urban vulnerability research has covered a broad range of topics, including social vulnerability to environmental hazards (Siagian et al., 2014; Cutter et al., 2003; Cutter, 1996), urban vulnerability to terrorist attacks (Walter et al., 2007), urban vulnerability to climate change (Grasso et al., 2014; Renaud and Perez, 2010; Hay and Mimura, 2006), urban internal vulnerability to heat (Joyce et al., 2014), urban vulnerability to natural disasters (Zhou et al., 2014; Usha, 2012), etc. Urban vulnerability research has involved cities in both developed and developing countries, such as the United States (Cutter et al., 2003), Australia (Abbas and Fahim, 2015), Chile (Muller et al., 2011), Kenya (Oluoko, 2011), South Korea (Koh, 2011), China (Yi et al., 2014), Bangladesh (Seraj et al., 1997), Vietnam (Nhuan et al., 2014), and Samoa (Grasso et al., 2014). The research methods used for urban vulnerability include data envelopment analysis (DEA), the geographic information system (GIS) ArcGIS, scenario analysis, the image-merging method, the vulnerability-function- based method, etc. As that research has become more profound, urban vulnerability research has evolved from the assessment of vulnerability associated with natural environmental deterioration to the investigation of the impact of human activities on vulnerability. In addition, the related research has also evolved from the simple assessment and passive acknowledgment of vulnerability-associated damage to the social economy to the active exploration of countermeasures for either coping with or avoiding vulnerability. Moreover, the focus of urban vulnerability research has shifted from traditional concepts, such as urban hazard vulnerability and eco-environmental vulnerability, to new concepts that are more closely associated with coupled systems, such as human-environment and socioecological systems (Fang et al., 2007).

Urban vulnerability research emerged in China more than two decades later than in other countries. However, with the acceleration of China’s rapid urbanization, chronic urban problems have become prominent, leading to an increasing number of urban vulnerability studies. According to the published data, China’s vulnerability research primarily focused on eco-environmental vulnerability and hazard vulnerability, which has made great progress in assessing eco-environmental vulnerability (Cai et al., 2009), classifying vulnerable eco-environmental systems (Liu and Li., 2007), and assessing vulnerability to natural disaster (Liu et al., 2010). Since the 21st century, scholars have attempted to investigate the socioeconomic vulnerability of a specific region under a specific disturbance, focusing on economic social vulnerability and sustainable development in special types of cities, including resource-based cities, coastal cities, and tourist cities (Su et al., 2008). However, the related studies primarily evaluated only the vulnerability of a single urban subsystem. In recent years, researchers began to pay attention to the vulnerability of urban complex systems, such as the vulnerability of urban economic social systems (Zhao and Zhang, 2006), the vulnerability of human-environment systems (Tian et al., 2013), urban vulnerability and sustainable development (Holling, 2007), and the vulnerability of urban and regional development (Zhang, 2006). However, until now, only a relatively small number of studies on the above aspects have been conducted.

Because of the complexity of cities and vulnerability concepts, studies in different fields have focused on different aspects of urban vulnerability. Consequently, the research content of urban vulnerability is scattered and relatively independent, leading to a lack of comparability among the research data and resulting in tremendous difficulties in summarizing the conclusions through comparison of independent research data (Wang, 2013). Therefore, from the perspective of sustainable development, this paper proposes a comprehensive concept of urban vulnerability; according to this concept, urban vulnerability refers to the coping capacity to resist multiple aspects of disturbances from various internal and external natural and human factors—e.g., resources, ecological environmental, economic, and social development—during the urban development process. A city is considered vulnerable if its anti-disturbance capacity is below a threshold level. Urban vulnerability is generally a collective reflection of four types of vulnerability, namely, resource vulnerability, eco-environmental vulnerability, economic vulnerability, and social vulnerability (Figure 1). Based on the proposed concept of urban vulnerability, we attempted to construct a research framework and a comprehensive assessment system of urban vulnerability. A nation-scale comprehensive assessment was performed to clarify the overall status and spatial distribution pattern of urban vulnerability in China and to explore vulnerability characteristics in cities of different sizes. This study tries to provide the government with scientific evidence for the formulation and perfection of related policies and regulations for reducing and coping with

urban vulnerability and for creating sustainable development.

In this study, using 288 Chinese cities at the prefectural level and above as study subjects, an index system for a comprehensive assessment of urban vulnerability was established from four dimensions (urban resources, environment, economy, and society) that involved 36 corresponding parameters. Standard values of urban vulnerability parameters were determined using the following approaches: (1) directly adopting the standard values proposed in the international or national survey reports; (2) referencing the current actual values in the domestic and international advanced areas; (3) basing those values on the current calculation results; and (4) referencing the currently available data in authoritative literature. The weights of the urban vulnerability index (VI) and relevant parameters were calculated using the expert scoring method and analytic hierarchy process (AHP) with the support of entropy technology (Fang, 2000). Based on the obtained results, an index system and standard parameter values for comprehensive assessment of urban vulnerability were established (Table 1).

3.2.1 The measurement model for subfactor indexes

Urban subfactor VIe are listed in the criteria hierarchy, which can be calculated based on the arithmetic mean values of corresponding parameters using the below equation：

where X_i is the standardized value of a parameter, m is the number of parameters included in a subfactor index, and UVI_e is the subfactor index of urban vulnerability. There are 10 subfactor indexes, including urban resource utilization VI (RUVI), urban resource assurance VI (RTVI), urban ecological VI (ECVI), urban environmental VI (ENVI), urban economic structure index (ESI), economic efficiency index (EEI), urban economic innovation index (EII), urban human development index (HDI), urban infrastructure VI (IFVI), and urban social environmental VI (SEVI).

3.2.2 The measurement model for subsystem indexes

Urban subsystem VIs are listed in the target hierarchy, which can be calculated using a weighting method following the below equation:

where (UVI_e)_i is the subfactor VI; n is the number of subfactors included in a subsystem VI; Wi is the weight of a subfactor in the criteria hierarchy; and UVI_s is the urban subsystem VI, including resource VI (URVI), eco-environmental VI (UEnVI), economic VI (UEcVI), and social VI (USVI). Because the urban economic structure index (ESI), economic efficiency index (EEI), urban economic innovation index (EII), and urban human development index (HDI) are negative parameters, they were calculated using the values of “1-ESI,” “1-EEI,” “1-EII,” and “1-HDI.”

3.2.3 The measurement model for comprehensive assessment

Urban comprehensive VI was calculated using the weighted summation of urban subsystem VIs following the below equation：

where (UVI_s)_i is the subsystem VI, Q_i is the weight of each subsystem in the target hierarchy, p is the number of subsystem indexes included in the comprehensive VI, and UVI is the urban comprehensive VI.

Based on the calculated values, urban comprehensive VI (UVI) was separated into five grades according to the natural breaking points, that is, lower vulnerability, low vulnerability, moderate vulnerability, high vulnerability, and higher vulnerability (Table 2).

The average UVI of 288 Chinese cities at the prefectural level and above is 0.5481, and their overall vulnerability is moderate. Shenzhen has the lowest UVI of 0.1221, followed by Beijing’s urban UVI of 0.3622. Baiyin City, a city facing resource exhaustion, has the highest UVI, 0.6762. There are remarkable differences in UVI among individual cities, as evidenced by the data indicating that the highest value is more than five times the lowest value. According to the UVI, urban vulnerability was divided into five grades based on the natural breaking points. The average UVIs of the five grades of cities are 0.4239, 0.5005, 0.5514, 0.5940, and 0.6462, and the average UVI of the high-vulnerability cities is 2.9 times the average UVI of the low-vulnerability cities, indicating distinct differences among cities at different vulnerability grades.

The hierarchical analysis of the number of cities demonstrated that the numbers of cities with low vulnerability, moderate vulnerability, and higher vulnerability are relatively large, accounting for 19.79%, 42.36%, and 26.04% of the total number of cities, respectively, and the sum of the cities at these three vulnerability grades accounted for 88.19% of the total number of cities. The numbers of the lower-vulnerability cities and the higher-vulnerability cities are comparable at 20 and 14, accounting for 7% and 5% of the total number of cities, respectively. In China, urban vulnerability shows a “more in the middle and fewer at the ends” distribution pattern. Various problems arose during rapid urbanization, and the cities have shown vulnerability to different extents.

The cities with lower or low vulnerability are mostly located in the eastern coastal zone, the moderate-vulnerability cities are mostly located in Central China, and the cities with high or higher vulnerability are mostly located in Western China. In terms of overall urban vulnerability, a statistical analysis of cities with different grades of vulnerability in Eastern, Central, and Western China(The region to which a city belongs is determined based on the region to which the corresponding provinces (or autonomous regions or municipalities directly under the Central Government) belongs. Eastern China includes Liaoning, Hebei, Beijing, Tianjin, Shandong, Jiangsu, Shanghai, Zhejiang, Fujian, Guangdong, and Hainan (11). Central China includes Heilongjiang, Jilin, Shanxi, Henan, Anhui, Hubei, Hunan, and Jiangxi (8). Western China includes Inner Mongolia, Shaanxi, Gansu, Ningxia, Qinghai, Xinjiang, Sichuan, Yunnan, Guizhou, Tibet, Chongqing, and Guangxi (12).) revealed significant interregional differences. The average UVI of the cities in Eastern, Central, and Western China is 0.5161, 0.5564 and 0.5911, respectively. In terms of the vulnerability grades, cities with GradeⅠ and GradeⅡ vulnerability were concentrated in the east and accounted for more than half of China’s cities with the same grade (85.00% and 68.42%, respectively).A larger number of cities with GradeⅢ and GradeⅣ vulnerability are distributed in Central China and accounted for 52.46% and 37.33% of those cities, respectively. The GradeⅤ cities are concentrated in the west (Figure 2). The analysis of the proportion of cities with different grades of vulnerability in each region shows that the proportion of Grade Ⅲ cities (40.86%) is the largest in Eastern China, followed by GradeⅡ and GradeⅠ, and GradeⅤ cities are scarce in the east. Central China is dominated by GradeⅢ and GradeⅣ cities, which account for 58.71% and 25.69%, respectively. Western China is dominated by GradeⅣ and GradeⅤ cities, which account for 56.25% and 18.75%, respectively (Figure 3).

Interregional differences are also demonstrated by Figure 4, in which urban vulnerability shows a westbound increasing trend. Moreover, urban resource vulnerability, eco-environmental vulnerability, economic vulnerability, and social vulnerability all show spatial differentiation consistent with that of urban overall vulnerability, showing a low-to-high order of Eastern China<Central China<Western China (Figure 5). In particular, there is a large difference in the vulnerabilities of resource utilization, environmental protection, and infrastructure (Table 3).

The spatial “gradient-distribution”-based heterogeneous feature of China’s urban vulnerability results from the impacts of multiple factors, both historical and environmental. From the perspective of natural conditions, compared to Eastern China, Western China has much poorer water and soil resources and climate conditions: the surface vegetation is mostly distributed among steppe and desert, and geologic and topographic features are complicated. In Western China, natural disasters frequently occur, environmental conditions are poor, and the eco-environment is vulnerable. These disadvantageous conditions are not conducive to human habitation and agricultural production, resulting in a poverty situation that has long existed in the region. Since the implementation of the reformation and opening-up policy, China’s unbalanced development strategy has drastically promoted the rapid growth of Eastern China, making this region a new economic growth point but further broadening the gap in economic development between the eastern and central-western regions. Moreover, differences in economic development have led to increased interregional differences in social development, such as education, medical care, and infrastructure. Since 2000, with the implementation of the Development Campaign of the Western Regions, economic growth in Western China has significantly accelerated, along with essential improvement in urban and rural living standards, along with education and infrastructure status. However, against the background of nationwide rapid economic growth, particularly that of Eastern China, the difference between the eastern and western regions did not decrease and instead increased in some aspects. During the period from 1979 to 2001, Eastern China contributed 57.8% and Central China contributed 27.9% toward GDP, whereas Western China contributed only 14.3%. Taken together, differences in natural conditions led to differences in economic development, and differences in the total economy and economic development led to differences in demographic and social development. Through this process, the combined effects of natural, economic, and social factors resulted in the spatial heterogeneous characteristics of urban vulnerability.

In Figure 4, the blank areas indicate the absence of cities at a prefectural level and above, such as autonomous prefectures. As shown in this figure, cities with low urban vulnerability are concentrated, showing a “clustered” distribution pattern. The areas with low urban vulnerability include the Harbin-Changchun area, the central and southern Liaoning area, the Beijing-Tianjin-Hebei area, the Shandong Peninsula, the Yangtze River Delta, the Pearl River Delta, the middle Yangtze River area, the western Taiwan Straits, and the Chengdu-Chongqing area. This distribution pattern is consistent with the spatial distribution of the major city clusters in China (Figure 6). In the areas with a clustered distribution, cities with a low UVI and low vulnerability include Changchun and Harbin (in the Harbin-Changchun area); Shenyang and Dalian in the central and southern Liaoning area; Beijing and Tianjin (in the Beijing-Tianjin- Hebei area); Qingdao, Yantai, and Jinan (in the Shandong Peninsula); Shanghai, Nanjing, Wuxi, Ningbo, Yangzhou, Suzhou, Changzhou, Taizhou, Zhoushan, and Zhenjiang (in the Yangtze River Delta); Shenzhen, Zhuhai, Xiamen, Guangzhou, Zhongshan, and Dongguan (in the Pearl River Delta); Wuhan and Changsha (in the middle Yangtze River area); Fuzhou and Xiamen (on the western Taiwan Straits); and Chengdu and Chongqing (in the Chengdu-Chongqing area). Clearly, the cities with lower or low vulnerability are mostly distributed within the city-cluster areas of Eastern China, and only a few are capital or major cities of the provinces in centralwestern China. Cities with high or higher vulnerability are primarily concentrated in the old industrial base of northeast China, the arid area of northwest China, and the southwest China (Figure 7). Developing and expanding city clusters are not only a major method of reducing urban vulnerability but are also an important approach to promoting healthy urbanization and improving the developmental level of urbanization.

Using the metropolitan population as an indicator of city size, the relationship between city size and urban vulnerability was explored. From the perspective of individual cities (Figure 8), there is no definitive corresponding relationship observed between urban vulnerability and city size; that is, UVI does not significantly decrease with the increasing metropolitan population. However, if cities are divided into different levels based on metropolitan population, the average UVIs of small cities (below 200,000), medium-sized cities (200,000- 500,000), large cities (500,000-1000,000), very large cities (1-2 million), and super-large cities (more than 2 million) are 0.5717, 0.5697, 0.5625, 0.5461, and 0.4903, respectively, demonstrating a decreasing trend with increasing city size (Figure 9). However, the UVI differences among the small cities, medium-sized cities, and large cities are insignificant, whereas the difference between very large cities and small cities is relatively large. In other words, from the perspective of individual cities, larger cities do not necessarily have lower vulnerability, and both medium-sized and small cities are suitable for economic development and human life. Nevertheless, overall, the bigger the city size, the lower the urban overall vulnerability, which is conducive to urban sustainable development.

China’s cities at a prefectural level and above were divided into two classes according to their urban functions: resource-based cities and comprehensive cities(As reported in “A Survey on Economic Structural Transformation of Resource-Based Cities in China,” published in 2002 by the Academy of Macroeconomic Research, State Development Planning Commission, China has 118 resource-based cities, including 47 prefectural-level cities and 71 county-level cities. Based on these data, the cities at a prefectural level and above in China can be roughly divided into two categories: resource-based cities and comprehensive cities. China has 47 resource-based cities and 241 comprehensive cities.). The average UVI is 0.5703 among the resource-based cities and 0.5440 among the comprehensive cities. Based on the grading criteria of the comprehensive assessment system of urban vulnerability, the resource-based cities overall have a relatively high vulnerability, whereas the comprehensive cities overall have a moderate vulnerability. In terms of resource types, the resource-based cities are further divided into coal-based cities, nonferrous-metallurgical-industry-based cities, ferrous-metallurgical-industry-based cities, petroleum-based cities, logging-industry- based cities, and other types of cities. The average UVIs of cities with different types of resources are in the descending order of coal-based cities>metallurgical-industry-based cities>logging-industry-based cities>other types of cities>petroleum-based cities (Figure 10). Moreover, compared with the comprehensive cities, the resource-based cities have significantly higher UVIs. In addition, among the resource-based cities, cities with different resource types have different vulnerability levels, and cities with a single type of resource are more vulnerable than cities with multiple types of resources.

To explore the relationship between urban economic growth and vulnerability, the metropolitan GDP growth rate was used as an indicator of cities’ economic growth. The GDP growth rate in 2011 was the slowest in Karamay City (3.5%), followed by Langfang (4.0%), whereas Qingyang had the fastest growth (30.3%), followed by Wuzhou (27.1%). Correspondingly, the urban VIs of these four cities is 0.4653, 0.5300, 0.5583, and 0.5550. Regardless of the large differences in the GDP growth rate among these four cities, their UVIs are not significantly different, indicating that economic growth does not necessarily correspond to the level of urban vulnerability. In addition, as shown in Figure 11, urban UVI irregularly fluctuates and does not decrease with the increase of the metropolitan GDP growth rate, suggesting that there is not necessarily a connection between the two. The correlation coefficient is only 0.15, indicating a low correlation between the two. Therefore, cities’ economic growth rate cannot reflect the level of urban vulnerability.