Analyzing Spatial–Temporal Patterns and Driving Mechanisms of Ecological Resilience Using the Driving Force–Pressure–State–Influence–Response and Environment–Economy–Society Model: A Case Study of 280 Cities in China

Abstract

Unveiling the spatial and temporal distribution of urban ecological resilience and analyzing the configuration paths for enhancing its levels are crucial for promoting sustainable development in China. Our study integrates the DPSIR and EES models, considering the causal relationships between systems affecting urban ecological resilience while also examining their internal structures. Based on this, we construct an evaluation system for urban ecological resilience indicators. Utilizing the entropy-TOPSIS method, we assess the ecological resilience index (ERI) across 280 Chinese cities from 2011 to 2021, and the kernel density estimation and Markov chain are used to study the evolution process while the magnitude and source of spatial–regional differences are examined by the Dagum Gini coefficient decomposition method. Additionally, we empirically investigate the driving mechanisms toward high ERI with the focused stepwise quantitative case analysis (fsQCA) method based on the technology–organization–environment (TOE) framework. The results find that the ERI in China shows a tendency of moderate growth in variability, with an obvious gradient distribution: higher levels in the eastern and southern and lower levels in the western and northern regions. Also, ERI exhibits evolutionary features of increasing polarization and inter-regional differentiation. Spatial disparities gradually increase with fluctuations, driven primarily by transvariation density and intra-regional differences, contributing to a dual non-equilibrium state of east–west and north–south directions. Achieving a high ERI is influenced by various antecedent variables interacting with each other, and there are three predominant driving paths among these variables, with the level of informatization playing a central role in each pathway.

1. Introduction

Economic development is inherently intertwined with the ecological environment [1,2,3]. Since China initiated its reform and opening-up policy in 1978, it has garnered global attention due to its remarkable achievements. However, this progress has been accompanied by detrimental effects on the ecosystem, leading to numerous unavoidable disturbances and ecological issues, affecting people’s livelihoods and obstructing China’s path to modernization [4,5]. In recent years, the escalating frequency of urban droughts and flood stress has exacerbated urban ecological risks, prompting widespread concern regarding the enhancement in urban ecosystems’ resistance, adaptability, and recovery capacity. Given the current circumstances, the primary focus of urban ecological protection and governance should pivot from mitigating ecological vulnerability to enhancing ecological resilience [6,7,8,9,10,11]. Therefore, it is imperative to establish a systematic and scientifically grounded ecological resilience evaluation system to objectively assess the present situation, delineate regional variances and evolution trajectories, and elucidate the driving mechanisms conducive to high ecological resilience.

Research on ecological resilience traces back to 1973 when Holling first introduced the concept of resilience to the field of ecology, evolving to emphasize ecosystems’ ability to defend, adapt, and recover from external disturbances, ultimately returning to their original state or forming a new stable state [6,12,13,14,15]. In recent years, scholars have increasingly focused on evaluating ecological resilience due to a series of ecological crises arising from urbanization and ecological fragility. The existing literature on constructing ecological resilience indicator systems primarily follows two main approaches. First, the evaluation system is based on the inherent properties of ecological resilience, focusing on three main characteristics: resistance, adaptability, and recovery [16]. In 2022, an evaluation system based on resistance and adaptability was constructed by Shi et al. to measure the ecological resilience of the Beijing–Tianjin–Hebei urban agglomeration in China [15]. Li et al. believed that the resilience of an ecological environment is influenced by pollution from socioeconomic activities and by efforts in ecological restoration and environmental management, which precisely demonstrates the ecosystem’s defensive resistance, adaptive recovery, and capacity for transformation [9]. Second, from a human response perspective, the evaluation system integrates urban ecological resilience with urban planning, incorporating indicators such as city size, density, and morphology, which has led to research on garden cities and sponge cities [17,18,19,20,21]. Furthermore, scholars investigate the interconnected development of economic growth, urbanization, and ecological resilience from a coupling perspective, and accordingly construct an evaluation system [22,23].

From the perspective of spatiotemporal characteristics of ecological resilience, scholars often focus on the provincial level across the country, dividing it into three regions, east, central, and west, to study regional differences in ecological resilience. Hu et al. explored the regional unevenness of ecological resilience at the provincial level based on a framework of “resistant, absorptive and restorative capacity”, and they found that ecological resilience was higher and more contiguous in the eastern region, while it was lower and more clustered in the western region [24]. Some scholars also use specific river basins or urban agglomerations as examples to study the levels and differences in ecological resilience within regions [15,25]. For example, Huang et al. investigated the spatial–temporal evolutionary characteristics of ecological resilience in the Yellow River Basin, and the results showed that there is a narrowing trend of the spatial differences [16]. Common research methods include descriptive statistics, an exploratory spatial data analysis, the Dagum Gini coefficient method, and the Moran’s index to reveal these spatiotemporal characteristics.

As scholars deepen their understanding of ecological resilience, they have increasingly delved into its underlying driving mechanisms. This exploration has predominantly centered on technological and environmental dimensions, with the environmental aspect categorized into natural and social factors. Firstly, research indicates that technological factors, such as innovation and the digital economy [26,27], can bolster ecological resilience. Secondly, at the environmental level, scholars have scrutinized the influence of environmental regulations [28], industrial structure [29], and regional climate and natural resources on ecological resilience [30]. In terms of research methodology, traditional mainstream econometric models, including multiple linear regression models, geo-detectors, and correlation analyses, have been commonly employed.

The above-mentioned literature provides valuable references and insights for our study, but it also exhibits several flaws. First, in constructing indicator systems, the existing literature often develops evaluation frameworks based on the connotations or characteristics of ecological resilience systems. While these studies explore the causal relationships between systems influencing ecological resilience, they frequently overlook structural analyses within these systems. This oversight results in incomplete dimensional construction and the selection of inadequate indicators. Second, in analyzing spatiotemporal characteristics, studies focusing on provinces often neglect the unique characteristics of individual cities within each province. Research concentrating on urban agglomerations or specific river basins fails to fully capture the current state of ecological resilience in China. Moreover, adopting an east–central–west perspective without analyzing the crucial north–south dimension further compromises the accuracy and comprehensiveness of the findings. Finally, regarding the analysis of influencing factors, the literature identifies various factors affecting ecological resilience. However, the selection of variables often appears disjointed, lacking an integrated analytical framework to advance ecological resilience at a theoretical level. Additionally, research methods frequently rely on geographical detectors, linear regression models, and geographically weighted models, which only consider the isolated effects of individual variables, disregarding interactions among multiple factors. This methodological approach inadequately explores the intricate causal relationships influencing ecological resilience.

Therefore, this paper makes the following marginal contributions. Firstly, by synthesizing existing research, it combines the “Driving Force–Pressure–State–Influence–Response” (DPSIR) and “Environment–Economy–Society” (EES) models based on the dynamic cyclic process of ecological resilience. This approach considers the causal relationships between ecological resilience systems and examines their internal structures, providing a clearer reflection of element interactions and improving the quantitative evaluation framework of urban ecological resilience. Secondly, in exploring the driving mechanisms, adopting a configurational perspective and integrating the TOE theoretical framework suited for analyzing urban ecological resilience in real-world scenarios, we have developed a comprehensive framework for analyzing the multifaceted factors influencing ecological resilience, and the fsQCA method is employed to investigate how different combinations of antecedent variables impact the enhancement in urban ecological resilience. Three configurations that promote high urban ecological resilience are identified, offering various implementation options to improve ecological resilience across different cities, thereby strengthening the practicality and depth of policy recommendations.

The remainder of this paper is organized as follows. The materials and methods are presented in Section 2. Section 3 explains the results. The dynamic evolution of ERI is reported in Section 4 and regional difference decomposition is shown in Section 5. Section 6 analyzes the factors affecting ecological resilience. Finally, Section 7 discusses the conclusions, implications, limitations, and directions for future research.

2. Materials and Methods

2.1. Study Area

In order to ensure the availability and accuracy of the data, this paper takes 280 cities in China as the research object, including four municipalities directly under the central government, 25 provincial capitals, and 251 prefecture-level cities. Based on the research by Nie et al., the provinces in the eastern, central, and western regions were delineated according to the commonly adopted 11:8:12 division standard in National Bureau of Statistics. For the north–south direction, provincial divisions were established using Qinling Mountains—Huaihe River [31]. Subsequently, the regional divisions of cities within each province were determined accordingly.

This study employs two principal types of data. On the one hand, the data used to measure the ERI of each city are that which span the period from 2011 to 2021. With the exception of the data on night-time light intensity for each city derived from National Earth System Science Data Center, National Science and Technology Infrastructure of China “http://www.geodata.cn (accessed on 17 March 2024)”, and the proportion of days with good air quality derived from China National Environment Monitoring Centre “https://www.cnemc.cn/ (accessed on 17 March 2024)”, the remaining data were sourced from China’s Urban Statistical Yearbook, China’s Regional Economic Statistics Yearbook, China’s Urban and Rural Construction Yearbook, and provincial statistical yearbooks. On the other hand, the data used to study the mechanisms driving urban ecological resilience were for 2021. The number of green patents was sourced from the CNRDS database, the informatization level was derived from the Peking University Digital Financial Inclusion Index of China, the total number of word frequencies of green development in government documents was obtained from the government work report of each city, the data on the number of enterprises above the scale and the industrial structure were extracted from China’s Urban Statistical Yearbook, and the data on the level of social concern were obtained from the Baidu index of Baidu Inc., (Beijing, China). “https://index.baidu.com/ (accessed on 17 March 2024)”.

2.2. Evaluation System Based on “DPSIR-EES” Model

The DPSIR model, proposed by the European Environment Agency (EEA) in 1999, delineates causal relationships between economic and environmental systems using a “why–how–what” logic [32], aligning with resilience phases within the social–ecological resilience perspective [25,33]. As shown in Figure 1, in the DPSIR model, driving forces (D) and pressures (P) represent explicit or implicit elements influencing changes in the state (S) of urban ecological resilience; the state (S) and influence (I) characterize the ability of the urban ecological resilience to absorb, adapt, and recover to the pre-disturbance resilience level; response (R) portrays how human societies can bolster ecological resilience through learning and innovation. The five dimensions of the DPSIR model intricately interconnect, shaping the internal mechanisms and evolutionary processes of urban ecological resilience systems. Combining the progressive logic of the DPSIR model systems with the development stages of urban ecological resilience extends and innovates the evaluation system based on the concept of resilience. However, it is important to acknowledge that the model does not capture causal relationships within each system.

The EES model, introduced by Shijun Ma in 1984, presents a composite system model capable of capturing the three-dimensional complexity inherent within systems [34]. Specifically, it pertains to the tripartite attributes encompassing economic, environmental, and social dimensions of human activities. The EES model effectively captures the composite attributes within each ecological resilience system to analyze its internal structure, yet it struggles to depict causal relationships between these systems. Therefore, by combining the DPSIR and EES models, a scientific and systematic approach emerges for evaluating the efficacy of ecological resilience building. This integrated approach underscores causal relationships between systems while emphasizing their composite attributes.

Using the DPSIR-EES model as a framework and adhering to the principles of scientific rigor, comprehensiveness, and data availability, we decompose ecological resilience into five system levels: driving forces, pressures, states, influences, and responses. Within each level, dimensions are categorized into economic, social, and environmental aspects to guide indicator selection. This methodology resulted in the development of a city ecological resilience evaluation system comprising 25 indicators, as detailed in

Table 1. Ecological resilience evaluation indicator system in the DPSIR-EES framework.

Component	Dimension	Indicators	Calculation	Type
Driving forces	Economy	Urban economic output	GDP per capita (10,000 CNY)	+ *
		Urban per capita income	Average wages of employees (CNY)	+
	Society	Population growth trends	Natural population growth rate (‰)	$-$
		Urban vitality	Night-time light intensity	+
	Environment	Ecosystem state	Area of green space/total administrative area (%)	+
Pressures	Economy	Pressure for economic growth	GDP growth rate (%)	$-$
	Society	Pressure for population agglomeration	Population density (persons/km2)	$-$
		Pressure for social employment	Urban unemployment rate (%)	$-$
	Environment	Pollution emission intensity	Wastewater, waste gas, and waste solid emission/GDP (t/CNY)	$-$
State	Economy	Urban economic density	GDP per unit area (10,000/CNY)	+
		Urban investment	Per capita investment in fixed assets (10,000 CNY/km2)	+
	Society	Urban consumer dynamism	The total retail sales of consumer goods per capita (10,000 CNY)	+
		Urban residential carrying capacity	Residential land area per capita (m2/10,000 persons)	+
	Environment	Urban water resources	Total water supply per capita (10,000 m3)	+
		Urban livability	Greening coverage of urban built-up areas (%)	+
Influences	Economy	Urban economic development	Local fiscal revenues (10,000 CNY)	+
	Society	Population and economic growth elasticity	Natural population growth rate/GDP growth rate	$-$
		Urban land intensification capacity	Area of construction land per unit of GDP (m2/GDP)	$-$
	Environment	Urban ecological quality	Proportion of days with good air quality (%)	+
Responses	Economy	Degree of optimization of economic structure	Proportion of added value of the tertiary industry (%)	+
	Society	Expenditure on urban science and technology	Science and technology expenditure/total fiscal expenditure (%)	+
		Expenditure on urban education	Education expenditure/total fiscal expenditure (%)	+
	Environment	Urban life quality	Harmless treatment rate of domestic waste (%)	+
		Urban environmental governance	Comprehensive utilization rate of general industrial solid waste (%)	+
		Expenditure on urban environmental protection	Landscaping investment expenditure/total fiscal expenditure (%)	+

* “+” means positive type, “$-$” means negative type.

. The specific explanation is as follows.

• Driving Force Indicator Set: Driving forces are fundamental factors that influence changes in ecological resilience. Existing research identifies economic factors, social development, and ecological construction as key drivers of urban ecological resilience [26,29]. This paper, building on relevant studies, selects GDP per capita and employee wage levels as indicators for economic driving forces, reflecting the impact of regional economic development on ecological resilience. Population size and night-time light intensity are chosen as indicators for social driving forces, demonstrating how population growth and urbanization affect ecological resilience. The proportion of urban green space is used as an indicator for environmental driving forces, highlighting the role of greening activities, such as afforestation, in enhancing ecological resilience. Given that increasing population size results in a corresponding rise in waste and pollutants, population size is considered a negative indicator.
• Pressure Indicator Set: Pressure represents the negative factors that influence ecological resilience as a result of driving forces. This paper selects the GDP growth rate as an indicator of economic pressure. The longstanding competition to increase GDP in China has significantly expanded economic output but has also incurred substantial environmental costs [35], leading to considerable ecological risks in urban development. Therefore, the GDP growth rate is treated as a negative indicator. Additionally, this paper uses the population density and urban unemployment rate as indicators of social pressure, both of which have detrimental effects on ecological resilience. Population density signifies the environmental stress associated with urban population concentration, while the urban unemployment rate illustrates the trade-off between job creation and environmental protection. Increased urban unemployment is likely to result in a compromise of green development goals for economic growth [36]. Finally, industrial emissions of air, water, and solid waste are used as indicators of environmental pressure.
• State Indicator Set: The state refers to the current development status of economic, social, and environmental subsystems under the combined influence of driving forces and pressures. Unlike driving forces, which function as “conditions” for the evolution of ecological resilience, the state primarily represents the “outcome” of this process. Based on existing research [37,38], this paper uses GDP per unit area and fixed asset investment per unit area as indicators of the economic state, reflecting current economic density and investment activity. For the social state, indicators include per capita retail consumption and per capita residential land area, which reflect current consumption levels and living standards. For the environmental state, per capita total water supply and green coverage rate in built-up areas are chosen as indicators, representing the current capacity for resource and environmental support.
• Influence Indicator Set: Influence reflects the direct effects of changes in ecological resilience on economic, social, and environmental subsystems. Unlike state indicators, which represent static conditions, influence indicators capture dynamic changes. Based on existing research [37,39,40], this paper uses local fiscal revenue to represent economic influence, reflecting how driving forces and pressures influence regional economic development. For social influence, the indicators are the elasticity of human resources to economic growth and construction area per unit of GDP, which reflect the interplay between human development, economic growth, and land use. For environmental influence, the proportion of days with good air quality is selected as the indicator.
• Response Indicator Set: Response refers to the management measures and policies implemented to enhance ecological resilience. Firstly, regarding economic response, the tertiary sector is more environmentally friendly than the primary and secondary sectors due to its lower resource consumption and pollution emissions. Therefore, this paper uses the value added by the tertiary sector to gauge the extent of economic structural optimization. Secondly, financial investments in education and technology are used as indicators of social response. Education and technology, fundamental to sustainable development [37], enhance ecological resilience by increasing environmental awareness and advancing high-tech applications. Lastly, this paper employs the rates of harmless treatment of household waste, comprehensive utilization of industrial solid waste, and investment in landscaping as indicators of environmental response.

2.3. Methods

2.3.1. Entropy-TOPSIS Method

The entropy-TOPSIS method, as a multi-attribute decision-making approach with finite schemes, offers several advantages, including intuitive geometric interpretation, minimal information loss, and flexible calculation. This method holds particular significance in a multi-objective decision analysis [41]. The methodology entails initially utilizing the entropy weight method to identify each index’s weight, thus mitigating subjectivity. Subsequently, the approximation ideal solution technique is employed to rank the advantages and disadvantages of the evaluation objects [42].

2.3.2. Methods for Dynamic Evolution Process

• Kernel density estimation, as a quantization tool with low dependence on the model and great robustness, is widely used in the evaluation of spatial disequilibrium, and the continuous density curve is used to unveil the changes in spatial absolute differences. This paper uses traditional kernel density estimation to examine the overall dynamic distribution of the development level of urban ecological resilience in China. Equations (1) and (2) are the density distribution functions of ERI:
  $$f\left(x\right)={\displaystyle \frac{1}{Nh}}{\sum }_{i=1}^{N}K({\displaystyle \frac{X_i-x}{h}}),$$

  (1)

  $$K\left(x\right)={\displaystyle \frac{1}{\sqrt{2\pi }}}{e}^{(-{\displaystyle \frac{x^2}{2}})}$$

  (2)
  In Equation (1), where N denotes the number of cities in each region, X_i represents the independent identically distributed observations, which is ERI of each city, and x and K represent the average value and the kernel density, respectively. h represents the bandwidth; the smaller the bandwidth, the higher the accuracy of the estimation, but the smoothness of the curve is correspondingly lower. Conversely, the larger the bandwidth, the smoother the curve, but the lower the estimation accuracy.
• The traditional Markov chains are used to study the problem of stochastic transfer of time and state in the absence of an aftereffect [43]. By constructing a Markov transition probability matrix, this method addresses the limitations of kernel density estimation in revealing internal transition information and predicting long-term trends. Therefore, this paper uses the traditional Markov chain to characterize the internal dynamic evolution law of ERI in China.
  $$P\left\{X\left(t\right)=j|X\left(t-1\right)=i,\cdots ,X\left(0\right)=i_0\right\}=\{X\left(t\right)=j|X\left(t-1\right)=i\},$$

  (3)

  $$p_{ij}={\displaystyle \frac{n_{ij}}{n_j}}$$

  (4)
  In Equation (3), it shows that the probability of ERI X in the current state j period t depends on its ERI in the previous period, t − 1. Equation (4) shows the probability that ERI will transfer from state i in the current period to state j in the next period. By arranging all types of transfer probabilities in the form of a matrix, the transfer probability matrix p_ij of China’s urban ecological resilience is formed.

2.3.3. Dagum Gini Coefficient Decomposition Method

While the kernel density estimation method reveals the evolution of ecological resilience difference from an absolute perspective, the Dagum Gini coefficient decomposition method uncovers regional disparities’ evolution from a relative perspective. This study further explores regional disparities in ERI using the Dagum Gini coefficient decomposition method, elucidating the underlying causes of variation across regions in China. Unlike regional difference quantification methods such as the Theil index, which overlooks the specific distribution within subsamples, the Dagum Gini coefficient method precisely reflects the magnitude and sources of relative differences. According to this way, the overall Gini coefficient (G) can be broken down into three components: intra-regional difference (G_w), inter-regional super variable net value difference (G_nb), and inter-regional transvariation density (G_t). In general, smaller overall Gini coefficients denote lesser regional disparities, while larger coefficients indicate weaker inter-regional synergy [44]. Hence, the calculation equation for the Dagum Gini coefficient is as follows [45].

$$G={\displaystyle \frac{{\sum }_{i=1}^k{\sum }_{m=1}^k{\sum }_{j=1}^{n_i}{\sum }_{r=1}^{n_m}|y_{ij}-y_{mr}|}{2n^2\mu }},{\mu }_m\le {\mu }_i\le \cdots \le {\mu }_k,$$

(5)

$$G_{ii}={\displaystyle \frac{{\sum }_{j=1}^{n_i}{\sum }_{r=1}^{n_i}|y_{ij}-y_{ir}|}{2n_i^2{\mu }_i}},$$

(6)

$$G_w={\sum }_{i=1}^k{G}_{ii}{p}_i{s}_i,$$

(7)

$$G_{im}={\displaystyle \frac{{\sum }_{j=1}^{n_i}{\sum }_{r=1}^{n_m}|y_{ij}-y_{mr}|}{n_i{n}_m({\mu }_i+{\mu }_m)}},$$

(8)

$$G_{nb}={\sum }_{i=2}^k{\sum }_{m=1}^{i-1}{G}_{im}(p_i{s}_m+p_m{s}_i)D_{im},$$

(9)

$$G_t={\sum }_{i=2}^k{\sum }_{m=1}^{i-1}{G}_{im}(p_i{s}_m+p_m{s}_i)(1-D_{im}),$$

(10)

$$D_{im}={\displaystyle \frac{d_{im}-p_{im}}{d_{im}+p_{im}}},$$

(11)

$$d_{im}={\int }_0^{\infty }{dF}_i\left(y\right){\int }_0^{y}y-x{dF}_m\left(x\right),$$

(12)

$$p_{im}={\int }_0^{\infty }{dF}_m\left(y\right){\int }_0^{y}y-x{dF}_i\left(x\right),$$

(13)

In Equation (5), where k represents the number of subregions, n represents 280 cities in China. y_ij (y_mr) denotes the ERI of city j (r) in region i (m), respectively; n_i (n_m) represents the number of cities in region i (m); and μ_i (μ_m) represents the average ERI in region i (m). Equations (6) and (8) calculate the Gini coefficient (G_ii) for region i, and the Gini coefficient (G_im) between regions i and m, respectively; Equations (7), (9), and (10) calculate the G_w, G_t, and G_nb of ERI in China, respectively. Equations (11)–(13) calculate d_im and p_im.

2.3.4. Focused Stepwise Quantitative Case Analysis

The focused stepwise quantitative case analysis (fsQCA) is an analytical method that integrates the benefits of both qualitative and quantitative analyses, facilitating their effective integration [46]. Two primary rationales justify the selection of fsQCA in this study. Firstly, ecological resilience factors may exhibit interdependence, making it more suitable to examine their synergistic mechanisms holistically. Secondly, regional and urban disparities in ecological resilience exist, and fsQCA offers an effective means to elucidate the varying impact mechanisms of influencing factors.

3. Results

3.1. The National Level Analysis of ERI

In general, as depicted in Figure 2, the overall evaluation results of ERI during the study period exhibited a tendency of moderate growth in variability with an annual growth rate of 0.76%, and the overall average level increased from 0.4794 in 2011 to 0.5157 in 2021, an increase of 7.5% year on year. These findings suggest that the ecological resilience of Chinese cities maintained a moderate level during the sample period, with a favorable ongoing development trend. In terms of dimensions, the trends in the system indices of the DPSIR model exhibit consistency, featuring a distinct stepped distribution overall. Notably, the influence component shows the highest system-level values consistently ranging between 0.6 and 0.7, showing a steady upward trend in recent years, indicating that the dual effects of driving forces and pressures are increasingly influencing urban ecological resilience. The pressure system index ranges between 0.5 and 0.6, experiencing fluctuations of “increase–decrease–increase–decrease” during the sample period, with the fluctuation cycles shortening, suggesting that the risks and challenges faced by urban ecological resilience systems are becoming more frequent. The response system index ranges between 0.45 and 0.55, consistently showing an upward trend, indicating that cities have responded positively to enhance ecological resilience, leading to a gradual strengthening of urban ecological resilience. The driving force system index ranges between 0.4 and 0.5, with an accelerated growth rate since 2016, mainly due to China’s shift towards high-quality development, with greater emphasis on development quality, providing the primary impetus for enhancing urban ecological resilience. The state system index ranges between 0.35 and 0.4; although the annual growth rate is low, it has consistently shown an upward trend during the sample period, which aligns with the inherent dynamic game of multiple systems within the ecological resilience system, indicating a positive transformation in the ecological resilience of Chinese cities.

3.2. The Regional Level Analysis of ERI

Figure 3 illustrates the dynamic changes in regional levels of ERI during the study period. It is evident that ERI in all regions experienced a fluctuating upward trend from 2011 to 2021. Regarding the east–west direction, the gradient distribution across the east, center, and west is pronounced. The ERI in the east notably surpasses that of the center and west, with average annual growth rates of 0.7%, 0.76%, and 0.77%, respectively. From a north–south perspective, the average annual growth rate of ERI in the south is 0.79%, which exceeds that of the north at 0.67%. Notably, firstly, economically developed areas exhibit higher ERI compared to underdeveloped regions, attributable to differences in development concepts, environmental governance capabilities, and capital investment. Secondly, areas with optimized industrial structures demonstrate higher ERI, as restrictions on industries with high energy consumption and pollution have been effectively implemented. Lastly, regions with robust ecological foundations display higher ERI.

3.3. The City-Level Analysis of ERI

Figure 4 presents the spatial distribution of ERI across the sample period, with Beijing, Shanghai, Shenzhen, Guangzhou, Nanjing, and Suzhou ranking among the top 280 cities. Conversely, cities such as Yichun, Jingzhou, Yuncheng, Zhaotong, and Jixi consistently lagged behind in terms of ERI. Notably, cities like Xi’an, Chongqing, Zhengzhou, Chengdu, and Urumqi have significantly improved their rankings. It is evident that cities with higher administrative levels tend to exhibit higher ERI, with most top-ranking cities being municipalities directly under the central government, sub-provincial cities, or regional centers. This association may stem from the correlation between a city’s administrative level and its foundation, capability, and policy implementation for achieving better ecological resilience. Conversely, cities at the bottom of the ranking typically adhere to traditional development models characterized by high energy consumption, low efficiency, and high pollution, exacerbating environmental degradation and hindering transformation efforts. Moreover, central and western hub cities, such as Chongqing, have seized historic opportunities for comprehensive transformation and upgrading, while other cities in these regions still remain at a lower development level but hold significant potential for growth. Undertaking the transfer of industries from the east, establishing national independent innovation demonstration zones, and fostering national urban agglomerations have contributed to the rising trend in ERI in these regions.

4. The Dynamic Evolution of ERI

4.1. ERI’s Kernel Density Estimation

Figure 5 illustrates the dynamic evolution of China’s ERI distribution. The kernel density estimation curves provide information on the position, shape, and extent of the variable distribution.

In terms of the distribution pattern, often utilized to analyze geographic differences and the degree of polarization, notable shifts were observed during the study period. The kernel density curves, representing both the entire country and individual regions, notably shifted to the right over time, particularly after 2015. This shift signifies significant improvements in ecological resilience not only nationally but also regionally, underscoring the effectiveness of China’s long-term ecological and environmental protection efforts. Notably, the largest rightward shift occurred in the central and western regions, indicating rapid progress in sustainable development in these areas.

The peak height and curve width serve as indicators of differences. Overall, the main peak height of ERI among Chinese cities during the study period initially rises before declining, while the peak width tightens before widening. This trend suggests an expanding absolute spatial gap in China’s ERI in recent years, with limited inter-regional synergy. On a subregional level, except for the southern region where the main peak height of the curve has increased to some extent and narrowed in width, the distribution pattern in other regions largely mirrors the overall trend. Here, the peak height initially rises before falling, and the width tightens before widening, signifying a narrowing spatial gap in the south, while ERI dispersion in other regions is widening.

The distribution extension serves to describe spatial and polarization differences within regions. All curves exhibit both left and right trails, reflecting not only cities with high ERI but also those with low ERI, both nationally and regionally. However, from a dynamic perspective, the left trail periods consistently diminish, suggesting a convergence in inter-regional differences among cities. Furthermore, the right trails of the central, northern, and southern regions show a widening trend, indicating ongoing momentum in upgrading cities with high ERI within these regions, while cities with lower ERI have yet to fully catch up.

The presence of multi-peaks in the distribution suggests a tendency towards polarization. At the national level, this polarization has intensified over time, as evidenced by the curve transitioning from a single-peaked to a multi-peaked state. Specifically, in the east, center, and north regions, the kernel density curves shifted from multi-peaked to single-peaked, indicating a reduction in polarization within these areas, with less developed cities catching up to their counterparts. Conversely, the distribution in the west and south has shifted from a single peak to double peak, indicating an increased degree of polarization within these regions.

4.2. The ERI’s Probability of Transition

The ERI was categorized into four distinct categories in the city-level analysis. Building upon this classification, we examined the dynamic evolution characteristics of ERI in China using a Markov chain analysis. To establish the time frame, the time span was set as 1, and we defined the four categories as low, medium–low, medium–high, and high. The transition probability matrix, generated through a traditional Markov chain analysis, is presented in

Table 2. Matrix of Markov transfer probability of ERI in China and its five regions.

t + 1	Type	Low	Medium–Low	Medium–High	High
Overall	Low	0.6214	0.3100	0.0629	0.0057
	Medium–low	0.1314	0.4714	0.3714	0.0257
	Medium–high	0.0171	0.1286	0.5914	0.2629
	High	0.0014	0.0057	0.0600	0.9329
Eastern	Low	0.7067	0.2667	0.0267	0.0000
	Medium–low	0.1333	0.5800	0.2867	0.0000
	Medium–high	0.0200	0.0833	0.7167	0.1800
	High	0.0000	0.0000	0.0414	0.9586
Center	Low	0.6300	0.2450	0.1150	0.0100
	Medium–low	0.1100	0.3750	0.4500	0.0650
	Medium–high	0.0350	0.0900	0.5050	0.3700
	High	0.0000	0.0053	0.0947	0.9000
Western	Low	0.5550	0.2950	0.1000	0.0500
	Medium–low	0.0650	0.4500	0.3750	0.1100
	Medium–high	0.0350	0.1300	0.3650	0.4700
	High	0.0091	0.0364	0.1000	0.8545
Northern	Low	0.6242	0.2818	0.0788	0.0152
	Medium–low	0.1152	0.4758	0.3485	0.0606
	Medium–high	0.0273	0.1364	0.5061	0.3303
	High	0.0061	0.0242	0.0970	0.8727
Southern	Low	0.5892	0.3568	0.0541	0.0000
	Medium–low	0.1162	0.4892	0.3919	0.0027
	Medium–high	0.0108	0.0865	0.6973	0.2054
	High	0.0000	0.0000	0.0378	0.9622

.

At the national level, the annual probability of maintaining the original level at low, medium–low, medium–high, and high is 62.14%, 47.14%, 59.14%, and 93.29%, respectively. The probabilities of transitioning to the next level are 31%, 37.14%, and 26.29% for low, medium–low, and medium–high levels, respectively. Moving from low to medium–high and high has a probability of 6.29% and 0.57%, respectively. Similarly, the probability of transitioning from medium–low to low is 13.14%, whereas the probability of transitioning to the high level is only 2.57%. Moreover, the probabilities of moving from medium–high to low and from low to medium–low are 1.71% and 12.86%, respectively, while the probabilities of transitioning from high to low, medium–low, and medium–high are 0.14%, 0.57%, and 6%. Additionally, the probabilities on the diagonal of the overall ERI Markov transition matrix are 62.14%, 47.14%, 59.14%, and 93.29%, which are all higher than those off the diagonal, indicating strong stability across different levels, with pronounced solidity at both high and low levels. This also further indicates the presence of four types of club convergence phenomena, with the probabilities of high and low ecological resilience convergence being greater than those of mid-high and mid-low ecological resilience convergence. Furthermore, compared to the probability of downward transition, upward transition proves to be more challenging, with a low likelihood of cross-level transitions.

Meanwhile, the eastern, northern, and southern regions exhibit an internal dynamic evolution similar to the overall level of ERI, demonstrating higher probabilities on the diagonal of the transfer matrix than on the off-diagonal elements. Moreover, convergence at high ecological resilience levels is significantly greater than convergence at other levels. Specifically, in the eastern region, the diagonal probabilities of the Markov transition matrix are 70.67%, 58.00%, 71.67%, and 95.86%; for the northern region, they are 62.42%, 47.58%, 50.61%, and 87.27%; and for the southern region, they are 58.92%, 48.92%, 69.73%, and 96.22%. This suggests robust internal stability within these regions. For the above regions, the lowest probabilities on the diagonal are 58%, 47.58%, and 48.92%, respectively, while their highest probabilities on the off-diagonal are 28.67%, 34.85%, and 39.19%, respectively. These figures are lower than the lowest probabilities for ERI to remain unchanged within the regions, indicating a more stable evolutionary pattern within that is difficult to disrupt within these regions in the short term. In contrast, for the central and western regions, the probabilities of maintaining the original level year after year are as follows: low (63% and 55.5%), medium–low (37.5% and 45%), medium–high (50.5% and 36.5%), and high (90% and 85.45%). Notably, the probability of transitioning from the medium–low to the medium–high in the center is 45%, significantly higher than the 37.5% probability of remaining at the medium–low level. Similarly, the 36.5% probability of staying at medium–high in the west is notably lower than the 47% probability of transitioning from the medium–high to the high level. The findings indicate that the ERI in the central and western regions is stable and shows an upward trend in optimization with less risk of decline.

In conclusion, the following findings emerge: Firstly, both nationally and within each region, there is a notable tendency for high-level stability in ecological resilience. Secondly, the likelihood of an upward shift in ERI outweighs that of a downward shift, implying a relative long-term growth trajectory for China’s cities. Thirdly, positive transitions predominantly occur within adjacent levels, with minimal cross-level shifts, indicating that ERI enhancement exhibits a distinct stage characteristic, making rapid leaps in improvement challenging to achieve in the short term.

5. Regional Difference Decomposition

5.1. Total Regional Differences and Intra-Regional Differences

The trend of China’s overall Gini coefficient in relation to the ERI is shown in Figure 6. Throughout the sample period, the average Gini coefficient for ERI stands at 0.025, initially displaying a consistent decline followed by subsequent fluctuations. Precisely, before 2015, the overall Gini coefficient demonstrated a steady downward trajectory, declining annually by an average of 5.4%. However, after 2015, a pattern of frequent “rise–decline” fluctuations emerged, with an average annual increase of 2.06%. This challenge necessitates the urgent formulation and implementation of coordinated governance and integrated development policies to effectively bridge the gap in ERI between regions.

The decomposition of the Gini coefficient employs two classification methods: one divides regions into eastern, central, and western along the east–west axis, while the other classifies them into northern and southern along the north–south axis. Figure 6 also depicts the intra-regional Gini coefficients across these five regions. It is evident that intra-regional Gini coefficients are higher in the east and south, with average values of 0.03 and 0.026, respectively. This disparity primarily stems from significantly higher ERI in cities like Beijing, Shanghai, and Shenzhen compared to others in the eastern and southern regions. Notably, cities such as Heyuan, Zhangjiakou, and Shijiazhuang have consistently exhibited low ERI, contributing to pronounced disparities within these regions. Conversely, the Gini coefficients for the central and western regions are relatively small, indicating minimal gaps in ERI among cities within these regions and a more balanced development. From a dynamic standpoint, intra-group differences in all five regions initially exhibit a trend of steady decline. However, since 2016, except for the southern region, where intra-regional disparities have tended to fluctuate downwards, the remaining regions have experienced an upward fluctuation, highlighting the pressing need to address intra-regional disparities as a core priority for regional coordinated and sustainable development.

5.2. Inter-Regional Differences

Figure 7 reports the trends of the inter-regional Gini coefficient of ERI in 2011–2021. From the “east–central–west” perspective, the Gini coefficients vary as follows: 0.027–0.032 for east–center, 0.025–0.033 for east–west, and 0.016–0.022 for center–west. The Gini coefficient ranks these regions in descending order: east–west is roughly equal to east–center but notably higher than center–west. Similarly, from a “north–south” viewpoint, the Gini coefficient for south–north ranges from 0.023 to 0.030. Figure 7 illustrates a “V”-shaped trend in the Gini coefficients of these regional pairs over time, indicating an increasing urban ecological resilience gap between east–west and north–south regions in recent years. This dual non-equilibrium pattern of east–west and north–south disparities is consistent with findings from the regional ERI analysis in the preceding section. The widening gap in urban ecological resilience between east and west regions is primarily linked to differing economic development models. Eastern cities have increasingly transitioned to an intensive economic growth mode characterized by “low consumption—low pollution—high efficiency”, thereby enhancing the resilience system’s defense, absorption, adaptation, and learning capacities. Conversely, less economically developed central and western regions continue to rely on more rudimentary development approaches aimed at catching up, posing significant challenges to their urban ecological resilience systems. Regarding the north–south disparity in urban ecological resilience, two factors stand out: firstly, the south benefits from a stronger ecological foundation and inherent resilience capacity; secondly, northern cities, dominated by industrial structures, face heightened pressure on their resilience systems.

5.3. Sources of Regional Differences and Contribution Rate

Regarding the sources and contribution rates of regional disparities, Figure 8a illustrates variations in the east–west direction. Notably, intra-regional differences exhibit consistent fluctuation, with the contribution rate of differentiation stabilizing at 33%. Inter-regional differences initially ascend, reaching a peak of 42.92% in 2016, before declining. Transvariation density demonstrates a fluctuating pattern, decreasing initially to 24.81% in 2016, then peaking at 36.53% in 2021. This evidence yields two conclusions. Firstly, the relatively minor fluctuations in intra-regional differences’ contribution rate indicate smooth changes in ERI within regions, suggesting potential for spatial synergistic governance. Secondly, there is an inverse relationship between the contribution rate of inter-regional differences and transvariation density, with transvariation density increasingly superseding inter-regional differences as the predominant source of spatial ERI disparities among the eastern, central, and western regions in recent years. This shift underscores a rising overlap in ERI across regions, emphasizing the need to address ERI disparities in underdeveloped cities within developed regions.

Figure 8b illustrates the contribution rates and the sources of spatial differences in the north–south direction. It is evident that intra-regional differences consistently constitute the largest proportion of regional disparities throughout the study period, stabilizing at approximately 45%. Over the course of evolution, it can be seen that there is an inverse relationship between the contribution rate of inter-regional differences and transvariation density. Although the contribution of transvariation density has diminished in recent years, the inter-regional disparities’ contribution rate has steadily increased, converging with the share of intra-regional disparities and peaking at 39.41% in 2020. These findings indicate that intra-regional differences predominantly drive regional differences between the north and south. Therefore, expediting the establishment of mechanisms for coordinated evolution within the region is vital to foster ecological resilience of the north–south divide. Moreover, the persistent widening of the disparity between the northern and southern regions underscores the importance of narrowing this gap to prevent a further exacerbation of disparities in China’s ERI.

6. Analysis on Factors Affecting Ecological Resilience

6.1. Research Design and Variable Selection

The TOE framework, proposed by Tornatzky and Fleischer in 1990, is a comprehensive analytical tool with significant explanatory power and applicability [47]. It emphasizes the influence of factors across technology, organization, and environment dimensions on organizations’ technology integration and adoption behaviors [48]. Given that the TOE framework, which takes technology application as its essence, is highly compatible with the ecological resilience of basic needs [49], which is taken as its core driving force, this paper embeds the TOE framework, which is a synergistic development of technology–organization–environment, into the scenarios of Chinese ecological resilience, and integrates the driving factors in the existing literature, which is of strong practical significance in the search for a realistic pathway of achieving high ERI.

The theoretical model of TOE is shown in Figure 9. The first aspect, technological conditions, involves analyzing technology characteristics and associated elements, including two secondary factors: the level of green innovation and informatization. Green technological innovation drives ERI by enhancing production efficiency, reducing costs, increasing resource productivity, curbing pollution, and promoting waste utilization [49,50]. Additionally, through the provision of data support for monitoring and management, upgrading of technical means to achieve efficiency and resource conservation, media publicity to guide green production and consumption, etc., leveraging informatization empowers the construction of ecological civilization and enhances ecological resilience [51]. The second aspect concerns organizational conditions, which entails analyzing features aligning with technological conditions, encompassing institutions, mechanisms, financial inputs, and other factors, with a primary emphasis on government support. Particularly in the Chinese context, government backing has emerged as an essential strategic resource and a pivotal organizational condition for ecological protection and governance [52]. Governments possess numerous scarce resources crucial for industry development, such as land, capital, and project opportunities, while also overseeing and facilitating various business activities essential for enterprise growth. Consequently, government policy support is paramount. The third aspect pertains to environmental conditions, which encompass resources, demand, infrastructure, and other environmental factors influencing technological capabilities. This includes three secondary conditions: the level of market competition, industrial structure, and societal concerns. Firstly, intense market competition increases the likelihood of imitation, prompting enterprises to innovate technologically for green production [53]. Secondly, a sound industrial structure is pivotal for regional economic development, reducing pollution and carbon emissions and fostering city resilience. Finally, societal engagement is crucial for realizing high ERI, reflecting local residents’ environmental demands and enhancing public lifestyle and consumption patterns to effectively promote sustainability.

The green innovation capacity is gauged by the number of patent applications for the green technology field, reflecting a higher patent technology threshold while minimizing time delays [54]. Meanwhile, informatization is assessed using the informatization level from the Peking University Digital Financial Inclusion Index of China. Analyzing the government work report of each city, this paper tallies ecosystem-related keywords to gauge each city’s support for ecological resilience based on government word frequency totals. Market competition is partly associated with the number of enterprises in the market; hence, this paper employs the count of scale-above enterprise units to measure market competitiveness. Industrial structure, a crucial factor for ecological resilience, is characterized by the ratio of tertiary industry value-added to that in the secondary industry. Additionally, Baidu, China’s leading search engine, provides a Baidu index reflecting public search behavior, facilitating the measurement of social concern. This paper tracks keywords such as “green life” and “ecosystem recovery capacity” to calculate daily average search values for each city in 2021, reflecting social concern levels.

6.2. Data Calibration

Most existing studies use the quartile method for direct calibration to calibrate the variable into three anchor points [55].

Table 3. Descriptive statistics and calibration points.

Type	Variables	Descriptive Statistics				Calibration
		Mean	SD	Min	Max	0.75	0.5	0.25
outcome variable	Ecological resilience	0.516	0.025	0.456	0.607	0.526	0.512	0.499
condition variables	Green innovation	583.954	1564.599	3	16,215	392.750	123	54
	Informatization level	303.108	12.220	271.765	334.501	312.569	303.149	293.924
	Government support	0.003	0.001	0.001	0.010	0.004	0.003	0.0025
	Market competition	1521.918	2036.471	54	13,027	1723.5	847	418
	Industrial structure	1.345	0.648	0.386	5.072	1.511	1.194	0.988
	Social concerns	55.185	62.507	1.450	511.337	61.531	33.804	17.690

displays the results.

6.3. Empirical Analysis

6.3.1. Necessity Test

The necessity test assesses whether the outcome condition relies on a single variable.

Table 4. Necessity test.

Causal Conditions	High Resilience		Non-High Resilience
	Consistency	Coverage	Consistency	Coverage
Green innovation	0.647	0.691	0.410	0.424
~Green innovation	0.460	0.446	0.701	0.658
Informatization level	0.684	0.707	0.390	0.390
~Informatization level	0.410	0.409	0.707	0.684
Government support	0.500	0.512	0.584	0.579
~Government support	0.589	0.594	0.508	0.496
Market competition	0.685	0.716	0.383	0.388
~Market competition	0.414	0.410	0.720	0.689
Industrial structure	0.521	0.541	0.551	0.554
~Industrial structure	0.570	0.567	0.543	0.524
Social concerns	0.654	0.674	0.426	0.425
~Social concerns	0.443	0.443	0.674	0.654

indicates that for the outcome variable of high and non-high resilience, none of the antecedent conditions’ consistency levels exceed 0.9. Consequently, the influence of an antecedent condition on ecological resilience appears to be weak.

6.3.2. Configuration Analysis

A configuration analysis aims to elucidate the adequacy of the outcome condition through distinct groupings of antecedent conditions. Initially, 0.8 is set as the threshold for consistency based on data characteristics [56]. Secondly, given the study’s sample size of 280 cities, the case threshold is established at 2 [46,57]. Lastly, to mitigate potential inconsistencies in configurations, the consistency threshold for the PRI is set to 0.6 [58]. The configuration paths for high ecological resilience are shown in

Table 5. Configuration paths for high ecological resilience.

Causal Conditions	High Ecological Resilience
	H1	H2a	H2b	H3
Green innovation	● *	•		⊗
Informatization level	●	●	●	●
Government support		●	●	⊗
Market competition	●	●	●	⊗
Industrial structure		⊗	⊗	⊗
Social concerns	●		•	⊗
Raw coverage rate	0.476	0.205	0.206	0.081
Unique coverage rate	0.281	0.015	0.020	0.031
Consistency	0.810	0.828	0.818	0.826
Total coverage	0.547
Total consistency	0.813

* ● or • means the existence of a condition; ⊗ means the lack of a condition; ● or ⊗ represents a core condition; • represents an auxiliary condition.

.

• Driven by technology and environment type;

Configuration H1 indicates that a combination of high levels of green innovation, informatization, intense market competition, and social concern serves as core conditions for fostering high ecological resilience. The results reveal that configuration H1 achieves a consistency of 0.81 and a raw coverage rate of 0.476, explaining 47.6% of the urban situation, typified by cities like Beijing, Shanghai, and Nanjing. For instance, Shanghai, as China’s economic, financial, and technological innovation hub, has made notable strides in emerging technology innovation, optimizing industrial structure, and ecological preservation by integrating “digital” elements into green development. It has spearheaded the coordinated development of the economy and ecology. Moreover, Shanghai residents exhibit significant concern for the region’s ecological environment and green development, catalyzing a societal “self-revolution” in production and lifestyle, thereby enhancing the city’s ecological resilience. Additionally, as a first-tier city, Shanghai hosts numerous startups and traditional companies, fostering a fierce market environment that drives innovation and steers toward high ERI.

2.

Driven by technology, organization, and environment type;

Configuration H2 focuses on the informatization level, government support, and market competition as core conditions, excluding industrial structure. The corresponding configurations are H2a and H2b. These configurations imply that in cities reliant on traditional high-energy-consuming and high-polluting industries with low ERI, government support should be increased. This involves empowering traditional industries with the digital economy, attracting new industries, and continuously enhancing urban resilience. H2a features green innovation as the auxiliary condition, whereas H2b emphasizes social concern. H2a covers 20.5% of cities like Suzhou, Quanzhou, and Changzhou, while H2b encompasses 20.6% of cities including Shaoxing, Handan, and Zibo. Suzhou, for instance, has rapidly ascended to new first-tier city status, driven by government-backed initiatives attracting new technologies, industries, and top talent, intensifying market competition, and bolstering green innovation. Despite transitioning from old to new energy, Suzhou’s industrial structure evolves systematically, striving for industrial excellence and urban ecological resilience. Conversely, Handan’s heavy industry-centric structure has disproportionately impacted its ecological resilience, prompting heightened local demand for green environmental protection and ecological preservation.

3.

Driven by informatization level type;

Configuration H3 suggests that a combination of weak green innovation capacity, government support, market competition, the rationalization of industrial structure, and public attention alongside a high informatization level leads to increased ecological resilience. Here, when the informatization level serves as the core condition, the presence or absence of other conditions becomes irrelevant to ecological resilience. Configuration H3 encompasses 8.1% of cities, exemplified by Xinyu and Tongling. For example, Xinyu City’s economic development lags behind, characterized by a relatively simple industrial structure. The economy is heavily dependent on traditional industries, such as steel and non-ferrous metals, and traditional factor-driven growth. A shortage of specialized technical talent and a low willingness for green innovation have resulted in a long-term insufficiency in green innovation capabilities [59]. Informatization serves as a strategic engine for modernization, driving technological, production, and service mode changes, and facilitating strategic shifts toward leapfrog development. Since the 13th Five-Year Plan, Xinyu City has introduced several policy documents focusing on “informationization” and “digitalization,” including the “14th Five-Year Plan for the Development of Big Data Industry and New-Type Smart City Construction in Xinyu City,” the “2023 Work Plan for Xinyu City Big Data Center,” and the “14th Five-Year Plan for Digital Economy Development in Xinyu City.” These policies emphasize development goals centered around information technology. In practice, Xinyu City has made notable progress in constructing information infrastructure, establishing information industry parks, and attracting information technology enterprises, greatly enhancing the city’s ecological resilience through optimized resource management and strengthened monitoring and prevention measures.

6.3.3. Robustness Test

To ensure the reliability of the findings, this paper conducts a robust test of the antecedent grouping patterns that produce high ecological resilience [60]. The specific results are shown in

Table 6. Robustness test results.

Causal Conditions	High Ecological Resilience
	H1	H2a	H2b	H3
Green innovation	● *	•		⊗
Informatization level	●	●	●	●
Government support		●	●	⊗
Market competition	●	●	●	⊗
Industrial structure		⊗	⊗	⊗
Social concerns	●		•	⊗
Raw coverage rate	0.476	0.205	0.206	0.081
Unique coverage rate	0.281	0.015	0.020	0.031
Consistency	0.810	0.828	0.819	0.826
Total coverage	0.547 0.813
Total consistency

* ● or • means the existence of a condition; ⊗ means the lack of a condition; ● or ⊗ represents a core condition; • represents an auxiliary condition.

. Based on the existing literature [61], this paper proposes to raise the consistency threshold from 0.8 to 0.85. The measurements were subjected to a more stringent consistency threshold, and the results before and after adjusting the threshold remained highly consistent, demonstrating strong robustness.

7. Discussion

7.1. Conclusions

Using the data of 280 cities in China from 2011 to 2021, this paper builds an evaluation system of ecological resilience based on the “DPSIR-EES” model, and dynamically measures China’s ERI by the method of entropy TOPSIS. The kernel density estimation and the Markov chain method were used to study the dynamic evolution of ERI. The Dagum Gini coefficient method was examined to investigate the spatial–regional differences of ERI from the east to west and north to south directions. The fsQCA approach was used to find the effective driving pathways to achieve high ERI in a given context. The main conclusions are as follows.

First, China’s ERI shows moderate growth in variability, with regional ERI exhibiting a fluctuating upward trend and a clear gradient distribution. Similarly, within the DPSIR model, the system indices display a distinct gradient, with the influence system index being the highest and the state system index being the lowest. Additionally, higher administrative levels correspond to higher ERI. Second, the kernel density curves for both overall and subregional ERI are similar in terms of distribution, and exhibit a clear tendency of right shifting, and the height of the main peak falls and then increases with broadened width. All of the characteristics mentioned above indicate that China’s ERI will continue to improve, and an increasing trend in the absolute difference can be seen, followed by a decrease. While upward transitions in ERI become less challenging with time, cross-level transitions remain unlikely. Third, spatial differences in China’s ERI show a slowly rising trend despite swings, with the dual non-equilibrium state of east–west and north–south directions gradually emerging. Transvariation density gradually replaces inter-regional differences as the primary source of spatial differences in the east–west direction, and the intra-regional differences are the main source between the north and south. Forth, achieving high ERI is influenced by a combination of antecedent conditions and there are no necessary conditions. Also, three effective pathways to activate high ERI are identified, with a further analysis revealing that information technology plays a central role in all three pathways.

7.2. Theoretical Implications

The theoretical implications of this paper mainly have two aspects. First, existing methods for constructing ecological resilience evaluation systems can be broadly categorized as following a “theme-hierarchy” approach. This approach involves identifying secondary themes from a primary theme and selecting specific indicators under these secondary themes. For instance, many scholars categorize the evaluation dimensions of ecological resilience into resistance, adaptability, and recovery, and then gather indicators that represent these dimensions [9,15,16,24]. While the “theme-hierarchy” approach is intuitive, it often results in a complex and fragmented indicator system with weak systematicity and logic between indicators. To address these issues, this paper integrates the strengths of the DPSIR and EES models. This combined approach not only effectively identifies the causal relationships between subsystems of ecological resilience but also reveals the internal structure of each subsystem. Consequently, it improves the expression of the interrelationships and importance of the indicators, enhancing the scientific rigor of the evaluation results.

Second, the evolution of ecological resilience is a complex and dynamic process, making the analysis of its driving factors crucial for identifying pathways to enhance resilience. However, existing research may have two limitations. First, most studies primarily operate under linear assumptions, focusing on the net effects of driving factors while overlooking their synergies [26,27]. Second, existing studies suggest that numerous factors influence ecological resilience, leading to a dispersed selection of variables without constructing a theoretical framework that integrates antecedent factors for enhancing resilience. To address these gaps, this paper introduces a configurational perspective into the study of ecological resilience. By employing the classic TOE framework for an integrated analysis of key factors and using the fsQCA method to explore the complex impact pathways of combinations of antecedent variables, this study supplements the prevailing linear assumptions in existing research.

7.3. Practical Implications

Considering the conclusions above, the potential policy recommendations are proposed. Firstly, it is essential to construct an urban ecological resilience development system within a comprehensive framework and to strive for the holistic improvement in urban ecological resilience. The research findings indicate that the influence system index is the highest, the state system index is the lowest, and the indices of each system are clearly distributed in a tiered pattern. This suggests that the development of urban ecological resilience should prioritize a systemic, holistic, and coordinated approach, focusing on the improvement in and optimization of lagging subsystems. Additionally, comprehensive and stage-by-stage monitoring and evaluation should be strengthened to achieve balanced development of each subsystem and enhance urban ecological resilience. Secondly, the objective is to promote regional coordinated development strategies and gradually reduce disparities in regional ecological resilience. According to Dagum’s Gini coefficient results, the current regional development of urban ecological resilience in China exhibits a double imbalance pattern of east–west and north–south differences. To address this, cities should guide the orderly transfer, flow, and diffusion of technology, capital, information, and other elements from cities with high ecological resilience to those with low ecological resilience by improving the factor market circulation mechanism. This approach aims to fundamentally enhance ecological resilience. Additionally, cities should strengthen cooperation and communication, learning from advanced experiences through counterpart support and collaborative assistance. Thirdly, it is essential to understand the characteristics of urban development and enhance ecological resilience according to local conditions. According to fsQCA analysis results, there are three main configuration paths to improve urban ecological resilience in China, all of which require a high level of informatization. Therefore, under government guidance, it is crucial to fully utilize existing economic and technological resources to promote informatization. Additionally, efforts should be accelerated in industrial structure adjustment, clean energy development, and other areas based on existing resource endowments. By employing a comprehensive strategy, the overall enhancement in urban ecological resilience can be achieved.

7.4. Limitations and Future Research

Although we tried our best to ensure the reliability of this paper, it is not without limitations. Firstly, when analyzing urban ecological resilience measurement and its driving mechanisms, this paper primarily uses data from China’s official statistics, with a limited application of multi-source data such as geographic remote sensing, government platform, and public opinion data. Official statistics may be influenced by human factors, potentially introducing bias into the research findings. As data acquisition channels and methods continue to expand, future research should integrate and apply multi-source data to obtain a more comprehensive, accurate, and objective understanding of ecological resilience trends and enhancement pathways. Furthermore, enhancing ecological resilience is a dynamic process, with the composition of groupings and interactions of elements changing over time. The fsQCA method used in this paper does not account for the influence of time on conditional groupings, possibly leading to biased results. Future studies should conduct group analyses across multiple time periods to improve the robustness and generalizability of the conclusions.