Temporal and Spatial Di ﬀ erences in the Resilience of Smart Cities and Their Inﬂuencing Factors: Evidence from Non-Provincial Cities in China

: Based on the sample data of 81 non-provincial smart cities in China in 2017, the comprehensive evaluation index of the resilience of sample cities is calculated by using the entropy method, and the spatial di ﬀ erences of di ﬀ erent factors on the resiliency are analyzed by using the geographical weighted regression (GWR) model. The conclusions are as follows: Firstly, the comprehensive evaluation index of the resilience of smart cities presents a spatial distribution characteristic of decreasing from the east to the west. At the same time, the resilience comprehensive index, the public infrastructure resilience capacity index, the economic development resilience index, the social security resilience index, and the ecological environment resilience index of smart cities have obvious agglomeration e ﬀ ects on their geographical spaces. Secondly, the public infrastructure resilience capacity index and the ecological environment resilience index are both low with a discrete distribution, while the economic development resilience index and the social security resilience index are both high with a concentrated distribution. Thirdly, di ﬀ erent factors have signiﬁcantly positive e ﬀ ects on the resilience of smart cities. In particular, the public infrastructure capacity resilience index decreases from the north to the south with the spatial distribution pattern of concentration, the economic development resilience index and the ecological environment resilience index of smart cities decrease from the east to the west with a concentrated spatial distribution pattern, and the social security resilience index of smart cities decreases from the southwest to the northeast with a concentrated spatial distribution pattern. Therefore, it is necessary to enhance awareness of smart cities, strengthen the driving force of science and technology innovation, strengthen public infrastructure and service construction, and continuously improve the rapid resilience of smart cities.


Introduction
With the acceleration of the urbanization process in China, the problems of urbanization, such as water pollution, water shortage, urban haze, and urban flooding are becoming more and more prominent at the same time as benefits are provided by urban development. As traditional urban management has been difficult to use to meet the real needs of the environment and social development, this concept is facing a major challenge. Smart cities are the key means for modern urban governance [1][2][3]. However, cities that are smart do not necessarily effectively respond to rapid urbanization and their own growth.
Business Machines Corporation (IBM) came up with the idea of the "smart planet", which also included the concept of the "smart city". The cores of the concept are "perception", "interconnection", and "intelligence", which have attracted worldwide attention once the concept was put forward [5]. In the view of IBM, the "smart city" is consciously and actively controlling the trend of urbanization and combining advanced information and communication technology to integrate the core systems of the cities, such as people, commerce, transportation, communication, water, and energy, so as to make the whole city as a grand "system of the system" and operate in a smarter way, thus creating a better life for people in the city, and promoting the harmonious and sustainable growth of the city [6][7][8]. Caragliu et al. think that smart cities are meant to invest in human capital, social capital, and traditional and modern communications infrastructure through participatory governance, to promote sustainable growth in the economy, to improve the quality of life of the population, and to push forward the smart management of natural resources [9].
At present, there is no unified specific definition of smart city. For example, Anavitarte and Tratz-Ryan argue that the smart city is the government's use of "information and communications technology" (ICT) in different public service areas with urban function to provide sustainable and efficient services to citizens [10]; Batty et al. propose that the smart city mainly coordinates and integrates ICT and traditional infrastructure in cities through new electronic technology in cities [11]. However, the basic connotation of the smart city has been generally recognized, in which the use of a new generation of information technology manages the operation of the city in an integrated and systematic way, so that the functions of the city can be coordinated with each other and are beneficial to provide high quality development space for the enterprises in the city. In essence, the smart city is a complex system, which has the characteristics of adaptability. Besides, the smart city can interact with the environment and other subjects and constantly "learn" or "accumulate experience" in the process of continuous interaction and change to its own structure and behavior according to the experiences learned [12]. In the process of urban development, the most direct embodiment of this adaptive behavior is to repair itself in the process of resisting external interference, and finally to achieve sustainable development.

Urban Resilience
The concept of resilience stems from the field of engineering, which has been used by scholars in "risk and emergency management" in recent years [13]. In response to natural and human-made disasters facing human society, urban resilience has been used to measure the city's ability to withstand external "disturbances", with more emphasis on early preparedness and recovery. Meerow et al. believe that urban resilience is an ability of the urban system and all its social ecological and social technology networks on the temporal and spatial scales to maintain or quickly restore the required functions under interference, adapt to changes, and quickly change the systems that limit current or future adaptability [14]. The definition emphasizes that urban resilience focuses on managing and adapting to current conflicts and pressures.
Many scholars have carried out relevant research on urban resilience. Urban resilience mainly includes five aspects: the motivations and purposes [14], the resilience boundary of urban system [15,16], the time range of resilience [17][18][19], the content of resilience [20,21], and the subjects of the implementation of resilience [22,23]. These studies are all carried out around the connotations of urban resilience, while the problems of those five aspects will affect urban resilience. More importantly, how to judge the level of urban resilience is still pending. In addition, compared with ordinary cities, smart cities have their own unique features, for example, they are more dependent on information technology, and their response is faster in the process of external conflicts and crises. However, there is little research on the resilience of smart cities at present. Therefore, it is of great theoretical and practical significance to study the influencing factors of the resilience of smart cities and to build a systematic scientific evaluation model [24].

Data
Marana et al. argue that urban resilience should include critical infrastructure dependency, climate change, and social dynamics [25]. To achieve rapid recovery, Hernantes et al. constructed the resilience maturity model (RMM) with five divisions of starting, moderate, advanced, robust, and vertebrate [26]. Unlike ordinary cities, smart cities are driven by three factors, namely community, technology, and policy, which directly affect productivity, sustainability, accessibility, wellbeing, livability, and governance [27]. Different drivers may lead to risks with smart city development. Based on the case of five smart cities, Bunders and Varro argue that big data can also bring development risks to smart cities [28]. Further, based on the reality of China, Zhu et al. construct a disaster resilience evaluation index system including infrastructure, economy, society, the public, and the environment, and hold that the differences of above five aspects lead to the difference of disaster resilience capacity of smart cities in China [29].
Based on the research results of existing scholars, this paper divides the resilience evaluation index system of the smart city into urban public infrastructure capacity, urban economic development capacity, social security capacity, and ecological environment capacity, which are described as follows: (1) The public infrastructural resilience capacity of smart cities. This is the underlying resilience capacity of the smart city, which can evacuate and assist quickly during catastrophes such as earthquakes, storms, and so on. It includes urban public transport resilience capacity, urban flood discharge capacity, and urban medical assistance capacity. In this paper, urban public transport resilience capacity is expressed by the number of buses and the urban road area, which reflects the capacity of smart cities to mitigate the disaster; the urban flood discharge capacity is expressed by the length of the urban drainage pipeline, which reflects the capacity of smart cities to deal with flood disasters; the urban medical assistance capacity is expressed by hospital beds and practitioners in the city, reflecting the medical resilience capacity of the city in disaster emergency.
(2) The economic development resilience capacity of smart cities. The economic development resilience capacity means that government and residents have more economic development capacity to resist catastrophes and can inherit resilience relying on the technological innovation capacity quickly. It includes three parts: economic resilience capacity, scientific and technological innovation resilience capacity, and employment resilience capacity. Among them, the economic resilience capacity is expressed by the registered population number at the end of the year, per capita GDP, and the proportion of tertiary industries by GDP, which reflects the realistic basis of economic development of smart cities. Scientific and technological innovation resilience capacity is expressed by the intensity of research and development (R&D), the number of general college students, and patent applications. Furthermore, employment resilience capacity is indicated by the registered unemployment rate in cities and towns.
(3) The social security resilience capacity of smart cities. This is the public economic security capacity of the smart city, reflecting basic economic protection during catastrophes. It includes three parts: pension resilience capacity, medical insurance resilience capacity, and unemployment insurance resilience capacity, which are expressed by the number of participants in basic old-age insurance, the number of participants in basic medical insurance, and the number of insured persons in unemployment insurance, respectively.
(4) The ecological environment resilience capacity of smart cities. This is the external protection capacity of the smart city, meaning that the higher the quality of the ecological environment, the lower the possibility of a catastrophe. The ecological environment resilience capacity includes urban greening potential and urban ecological cleaning capacity. In this paper, urban greening potential is expressed by the greening coverage rate of the urban built-up area, and urban ecological cleaning capacity is expressed by comprehensive utilization rate of general industrial solid waste, the centralized treatment rate of sewage treatment plants, and the harmless disposal rate of domestic waste.
From 2012 to 2015, the Ministry of Industry and Information Technology of China has released three lists of smart cities, and a total of 287 cities or counties (districts) were selected as pilot projects for the construction of smart cities; no list of smart cities has been released since 2016. Therefore, this paper took smart cities released in 2015 and before as the investigation sample. Considering the availability of data and the reality of non-provincial smart cities, 81 smart cities were selected. At the same time, considering the accumulation of the construction effect of smart cities, only the year of 2017 was selected as the investigation period. Without special explanation, all the data in this paper came from the China Urban Statistical Yearbook, the China Regional Economic Statistical Yearbook, and the China Urban and Rural Construction Statistical Yearbook (2018), all retrieved from the Easy Professional Superior (EPS) electronical database. The selection and description statistics of variables are shown in Table 1.

The Entropy Evaluation Method
Compared with other methods, the entropy method can effectively avoid the subjectivity of index weighting, and objectively conduct evaluations [30][31][32]. Therefore, referring to the literature above, the entropy method was used to calculate the comprehensive evaluation index of resilience of smart cities.
Where x i represents the standardized value of x i , the negative standardized x i only needs to change the molecule of x i into (max{x 1 , . . . , x n } − min{x 1 , . . . , x n }), p i represents weight of index, e j is the entropy of index j, and n is the sample size. Hence, we can get weights of different samples w j as follows: In Formula (2), d j is the utility value of the index j. Then we can get the composite index of the resilience capacity of smart cities, SMRT i :

The Geographical Weight Regression Model
The traditional econometric model can only reflect the statistical relationship among variables as a whole, but cannot reflect the regional effect among variables, while the spatial econometric model can make up for the shortcomings of the traditional econometric model and fully reflect the differences in spatial geographical impact among variables, and especially can show all variables' coefficients of the smart city in the sample [33][34][35]. Therefore, the geographical weighted regression (GWR) model was used to analyze the spatial differences of the influence of various factors on resilience capacity of smart cities. In Formula (4), y i represents the resilience indexes of smart cities, and x ij represents the driving factors; (m i ,n i ) represents the spatial position of sample i, and µ i is the random error; β represents the regression coefficient, which is calculated as follows: In Formula (5), W ij represents spatial distance weight between different sample points, which is determined by the following Gaussian function: w ij = exp[−(c ij /b) 2 ], where C ij represents Euclidean distance between sample point i and sample point j; b is the bandwidth, which determines the weight of distance between any two smart cities and the optimal bandwidth determined by the cross confirmation method (CV):

Results and Discussion
In this section, we first analyze the spatial distribution of the comprehensive evaluation index of resilience capacity of smart cities, and then reveal the spatial different characteristics of factors that affect the resilience capacity of smart cities. Figure 1a shows the spatial distribution of the comprehensive resilience index of smart cities. From the figure, we can see that in 2017, the comprehensive resilience index of sample smart cities decreased from east to west, and the resilience capacity of smart cities in the eastern coastal areas was significantly higher than that in the central and western regions and Northeast China.

The Spatial Distribution of Resilience Index
First, the number of smart cities with the highest resilience index was small and relatively concentrated. Only seven smart cities were located in the first range of the highest comprehensive resilience index, and the order from large to small was Suzhou, Shenzhen, Qingdao, Wuxi, Ningbo, Dalian, and Nantong, most of which are developed cities along the eastern coast, among which, Suzhou, Shenzhen, Qingdao, and Wuxi have comprehensive resilience indexes of more than 60.
Second, the number of smart cities with the second highest resilience index was small and relatively concentrated. There were 18 smart cities located in the interval of the second highest value of the comprehensive resilience index, which were discretely distributed in the eastern and central regions, while most were in the east.
Third, the number of smart cities with median resilience index was large and relatively concentrated. There were 27 smart cities in the middle and third range of the comprehensive evaluation index, which were scattered in the central area and Inner Mongolia, while mainly located in the central area.
Fourth, the number of smart cities with low resilience index was large and relatively concentrated. There were 29 smart cities in the fourth range of the comprehensive evaluation index, which were discretely distributed in the western, northeast, and central regions, and mainly in the western regions.
Furthermore, we analyze the spatial distribution of public infrastructural resilience index of smart cities.
First, the public infrastructure resilience index of smart cities was small and discrete. In Figure 1b, smart cities located in the highest value range of infrastructure resilience index were few and scattered in the eastern area. Smart cities with the second highest value were few and discrete in the eastern, central, and northeast regions. There were many smart cities with medium values, which were scattered in the southeast and northeast, and were relatively concentrated in the central region. There were also many smart cities in the lowest value range, which were relatively concentrated in the central and northeast regions and scattered in the western region.   Second, values of economic development resilience index of smart cities were high, and the distribution of those cities was relatively concentrated. In Figure 1c, smart cities, which were located in the highest value of the economic development resilience index, were small and discrete in the east, and a few in northeast and central regions; there were relatively more smart cities in the sub high-value range, and they were concentrated in the eastern and central regions; smart cities with median values were more abundant and relatively concentrated in the central and western regions; and smart cities with the lowest values were less abundant and were distributed in the western, central, and northeast regions.
Third, values of the social security resilience index were low, and the distribution of those cities was relative concentrated. As seen in Figure 1d, the number of smart cities located in the highest-value section of social security resilience index was small and discretely distributed in the eastern region; the number of smart cities in the sub-high-value section was also small and concentrated in the east and scattered in the northeast area; smart cities with medium value were distributed in the central and eastern regions; smart cities in the lowest-value section were large and distributed in western, central, and northeast China.
Fourth, values of the ecological environmental resilience index of smart cities were high, and the distribution was relatively concentrated. As seen in Figure 1e, smart cities with the highest ecological environmental resilience index were the most numerous and relatively concentrated in the eastern and central regions and scattered in the western region. Smart cities in the sub-high value range were more abundant, and most of which were concentrated in the central region and scattered in the eastern region. Smart cities in the medium value region were few and scattered in the western and central regions, as well as the northeast region. Smart cities in the lowest value range were the least abundant and were confined to the west, central, and northeast regions. Table 2 shows the spatial Moran I index of the resilience capability of smart cities. From Table 2, we can see that the Moran I index of the resilience capability of smart cities, the public infrastructure resilience index, the economic development resilience index, the social security resilience index, and the environmental ecological resilience index were significantly positive at a confidence level of 1%, indicating that all of those indices had obvious agglomeration effects in geographical space. Furthermore, based on Formula (4), spatial differences in the impact of influencing factors on the resilience capacity of smart cities were analyzed. The robustness results of the model are shown in Table 3. Except for the ecological environment resilience index, the R 2 values for the public infrastructure resilience index, economic development resilience index, and social security resilience index were all greater than 0.6, indicating that the estimated results could reflect the statistical relationship among variables well.  Figure 2a shows the spatial distribution of the influence of public infrastructure resilience index on the resilience capability of smart cities. From Figure 2a, we can see that the coefficient of public infrastructure resilience index was significantly positive, and the mean value was 2.22, which indicates that improvement of public infrastructure is beneficial to promote the resilience capacity of smart cities. The better the public transport infrastructure, the easier it is to promote the rapid restoration of order in smart cities. In the event of natural disasters such as earthquakes and floods, we rely on the construction of big data central systems of smart cites to smartly improve the passage of urban traffic and the accuracy and timeliness of medical assistance, and to enhance the flood discharge capacity of the city. In space, the coefficient of public infrastructure decreased from west to east and had a concentrated spatial distribution pattern, which indicated that the positive influence of public infrastructure of smart cities in western China on urban resilience was higher than that in other areas. In western China, the population of smart cities was generally smaller than that in other areas, the urbanization rate was low, and the public transport infrastructure was relatively weak, which led to the marginal effect of public infrastructure on urban resilience capability being higher than that in other areas.

The Spatial Differences of Factors Affecting the Resilience of Smart Cities
In Figure 2b, the coefficient of economic development resilience index of smart cities was generally significantly positive, and the mean value was 2.55, indicating that economic development resilience ability has a significant positive impact on the resilience of smart cities. The better the economic development, especially the stronger the scientific and technological innovation foundation and the innovation driven development capacity, the easier it is to realize the rapid economic recovery of disaster cities. In space, the coefficient of economic development resilience decreased from east to west and has a relatively concentrated spatial distribution pattern, which indicated that positive influence of economic resilience of smart cities in the eastern region was higher than that in other areas. In the eastern region, especially in the smart cities along the eastern coast, such as Suzhou, Shenzhen, Ningbo, and Wuxi, the intensity of R&D and per capita patent applications were significantly higher than those in other regions. The foundation of urban economic development driven by R&D was solid. At the same time, per capita GDP in these areas was also significantly higher than that in other areas, and the quality of economic resilience was therefore high. Furthermore, the promoting effect of economic development on urban resilience in this kind of area was higher than that in other areas.
In Figure 2c, the coefficient of social security resilience index of smart cities was generally significantly positive, and the mean value was 4.12, indicating that social security had a significant positive effect on the resilience capacity of smart cities. After disasters, smart cities with strong social security resilience capacity can guarantee life safety of residents in time through the convenient and rapid response of medical assistance system. At the same time, through the big data system of complete pension, medical care, and employment security, it can timely distribute living, medical, and unemployment security benefits to residents to ensure them basic needs, and quickly restore the order of smart city. In space, the coefficient of social security resilience capability decreased from north to south and had a relatively concentrated spatial distribution, which indicated that the positive influence of social security of smart cities in northern China on urban resilience was higher than that in other areas. Due to the restriction of economic development and fiscal income, the level of social security of smart cities in the northern and northeastern areas, such as Jinchang, Zhangye, Liaoyuan, Siping, and Tonghua, was lower than that in the south, especially in the coastal economically developed areas, which led to the marginal effect of social security on urban resilience being higher than that in other areas.
In Figure 2d, the coefficient of ecological environment resilience index of smart cities was significantly positive, and the mean value was 2.16, indicating that the ecological environment had a significant positive effect on resilience capacity of smart cities. Smart cities with good ecological environments, especially with more greening space and higher air quality, provide a better ecological basis, and by using the internet of things, can speed up the restoration of smart cities in catastrophes. In space, the coefficient of ecological environment decreased from east to west and had a relatively concentrated spatial distribution, which indicated the positive influence of the ecological environment on urban resilience in smart cities in eastern China was higher than that in other areas. The main reason was that ecological environment level of the eastern region was obviously lower than that of other areas, due to previous overdevelopment, especially the problems of urban greening and air quality, which led to the marginal effect of ecological environment of smart cities on the urban resilience in the eastern region being higher than that in other areas.

Conclusions
Based on the sample data of 81 non-provincial smart cities in China in 2017, the comprehensive resilience indexes of smart cities were calculated by the entropy method, and spatial differences of influencing factors on urban resilience were analyzed by the GWR model. The conclusions are as follows: First, the comprehensive resilience indexes of smart cities show the spatial distribution characteristics of decreasing from east to west, which are consistent with Zhu et al. [29]. The resilience capacity of smart cities in the eastern coastal area is significantly higher than that in the central, western, and the northeast regions. At the same time, in geographical space, smart city resilience capacity index, public infrastructure resilience capacity index, economic development resilience capacity index, social security resilience capacity index, and environmental ecological resilience capacity index have obvious agglomeration effects.
Second, spatial distributions of influencing factors of different smart cities are obviously different. The index of public infrastructure resilience capacity in smart cities is small and discrete, the index of economic development resilience capacity of smart city is high and relatively concentrated, the index of social security resilience capacity of smart city is small and relatively concentrated, while the index of ecological environment resilience capacity of smart city is high and relatively concentrated.
Third, different influencing factors have significant positive influences on resilience capacity of smart cities, and the spatial distribution of influence coefficients is quite different [36]. Coefficients of public infrastructure resilience capacity index are significantly positive and show a decreasing spatial distribution pattern from west to east, indicating that the positive influence of public infrastructure resilience capacity of smart cities in western China on urban resilience is higher than that in other areas [37]. Coefficients of economic development resilience capacity index of smart cities are significantly positive and show a decreasing spatial distribution pattern from east to west, which indicates that the positive influence of economic development resilience capacity of smart cities on urban resilience in the eastern region is higher than that in other areas. Coefficients of social security resilience capacity index of smart cities are significantly positive and show a decreasing spatial distribution pattern from north to south, which indicates that the positive influence of social security resilience capacity of smart cities on urban resilience in northern China is higher than that in other areas. Coefficients of the ecological environment resilience capacity index of the smart city are significantly positive and show spatial distribution patterns of decreasing from east to west in turn and concentration, indicating that the positive influence of ecological environment resilience capacity on urban restoration capacity of the smart city in the eastern region is higher than that in other areas [38].

Implications
Recently, earthquakes, flood disasters, and air pollution have occurred frequently in China, and construction of smart cities gives us a meaningful and practical method to predict and solve these catastrophes using smart tools. Based on the above conclusions, in order to continuously improve the resilience capacity of smart cities, this paper puts forward three suggestions: First, the government should continue to strengthen the construction of big data central systems in smart cities. By increasing the data integration of smart cities in smart transportation, smart fire, and social security, smart brains of cities covering a variety of disaster recovery information can be constructed, and big data and cloud computing can be used to improve the resilience capacity of smart cities [39].
Second, the government should continue to strengthen the role of scientific and technological innovation in the resilience and development of smart cities. The investment in scientific and technological R&D needs to be continuously improved, and the training of scientific and technological talents should be strengthened [40]. Especially, the northeast and mid-western regions should make up for the shortcomings of scientific and technological innovation and attach importance to the supporting role of scientific and technological innovation in economic resilience of smart cities. At the same time, the eastern region needs to increase investment in major basic science and technology to consolidate the foundation driven by scientific and technological innovation.
Third, the government should step up efforts to withstand disasters in social security and infrastructure construction [41]. The construction of public infrastructure and public medical service facilities in the central and western regions, and also Northeast China, should be strengthened, and the capacity of smart cities to fight risks should be improved so rapid recovery can be achieved. In addition, it is necessary to improve the coverage of social security such as medical care and unemployment and constantly enhance the social security resilience capacity of smart cities in the central and western regions, and also Northeast China.

Limitations and Further Research
There are the following limitations: first, not all the smart cities in China were used in this paper due to lack of data availability. Second, we did not use panel data, since the construction year was 2015. Third, we did not analyze the influence with microscopic variables but focused on macroscopic variables.
In the future, we can continue to study from the following two aspects: The first is to further collect time series data for spatiotemporal analysis. The second is to try to add more variables from different angles to analyze the relevant problem of smart cities in multi-dimensional ways.