The Impact of Urbanization Level on Urban Ecological Resilience and Its Role Mechanisms: A Case Study of Resource-Based Cities in China

Abstract

Against the backdrop of accelerating global urbanization and intensifying ecological pressures, investigating the relationship between urbanization levels and ecological resilience in resource-based cities has become crucial for nations striving to achieve both sustainable development and ecological conservation. Utilizing panel data from 114 resource-based cities in China between 2010 and 2023, this study innovatively employs a composite nighttime light index to measure urbanization levels and constructs a comprehensive ecological resilience index using the entropy method. By applying a double machine learning model, this study thoroughly examines the impact, mechanisms, and heterogeneity of urbanization on ecological resilience in these cities. The findings reveal a gradual increase in ecological resilience among China’s resource-based cities, with the majority reaching high resilience levels by 2023. Spatial aggregation centers are identified in eastern China, the Yangtze River Delta, and the Pearl River Delta. Moreover, urbanization demonstrates a significant positive correlation with ecological resilience, a conclusion reinforced through robustness tests. Mechanism analysis reveals that industrial structure upgrading, green technology innovation, and energy efficiency improvement serve as key transmission channels. Heterogeneity analysis indicates that urbanization exerts a more pronounced effect on enhancing ecological resilience in regenerative resource-based cities as well as those located in eastern and central regions, while its impact is relatively weaker in declining resource-based cities and those in western and northeastern regions. Finally, this study proposes policy recommendations focusing on advancing industrial structure sophistication, constructing a green technology innovation ecosystem, implementing an energy efficiency enhancement initiative, deepening region-specific governance, and adopting targeted policy interventions. These findings provide theoretical support for precise policy formulation in resource-based cities and contribute to advancing academic understanding of the relationship between sustainable development and ecological resilience in such regions.

1. Introduction

The sustainable development transformation of resource-based cities represents a critical frontier in urban and regional studies, with their complex challenges transcending national boundaries. From Germany’s Ruhr region to the Rust Belt in the United States, the developmental trajectory of resource-based cities often demonstrates a common pattern characterized by economic stagnation, social vulnerability, and ecological degradation following resource depletion [1,2]. In China’s economic landscape, resource-based cities hold significant importance. Their economic development relies on the exploitation of natural resources, providing essential material support for the nation’s industrialization and urbanization processes. However, the long-term dependence on a resource-oriented development model in these cities has led to issues such as resource depletion, environmental degradation, and a monolithic industrial structure. Consequently, this has triggered frequent ecological problems and continuously intensified environmental pressures, which not only pose serious threats to urban and regional ecological security but also significantly undermine urban ecological resilience [3,4]. Since the 18th National Congress of the Communist Party of China on 8 November 2012, China has placed ecological civilization construction in a prominent position in its overall work, proposing to firmly establish and practice the concept that “lucid waters and lush mountains are invaluable assets,” intensify efforts to prevent and control pollution, safeguard the boundaries of natural ecological security, and strengthen the green foundation of high-quality development. In 2013, China promulgated the “National Sustainable Development Plan for Resource-Based Cities.” This policy document aims to help resource-based cities break away from the constraints of traditional development models by regulating resource extraction practices, establishing a sustainable development indicator system, and promoting low-carbon development pathways [5]. Therefore, reducing ecological risks and enhancing ecological resilience have become core issues in advancing the green and sustainable development of resource-based cities, which is not only an intrinsic requirement of the harmonious coexistence of ecology and economy but also a profound interpretation and practical reflection of the concept of ecological civilization construction.

Due to excessive exploitation and lagging upgrading of traditional industries, the ecological vulnerability of resource-based cities has intensified, leading to increasingly severe ecological security challenges. Ecological resilience in these regions is under immense pressure, becoming a major obstacle to high-quality economic development [6]. In 1968, the German government enacted the “Ruhr Development Program,” which enabled the Ruhr region to successfully transition from “black coal and steel” to “green technology” through systematic industrial restructuring. In 2019, President von der Leyen launched the “European Green Deal,” aiming to transform the EU into a modern, resource-efficient, and competitive economy. In response, the Chinese government has fully recognized the importance and urgency of enhancing ecological resilience. It has actively addressed these challenges by implementing a series of measures to promote sustainable urban development, striving to achieve economic growth while protecting the ecological environment, thereby laying a solid foundation for long-term urban development [7]. In October 2021, the Chinese government issued the “Opinions on Promoting Green Development in Urban and Rural Construction,” which proposed implementing “coordinated control by hierarchy and region” for cities based on resource and environmental carrying capacity. It also advocated for building a continuous and comprehensive ecological infrastructure system through projects such as urban ecological restoration, sponge cities, green communities, and “zero-waste cities”, indicating that, in the early stages of urbanization, the Chinese government emphasizes using ecological carrying capacity as a constraint to avoid excessive urban expansion. As urbanization continues, China has consistently stressed ecological restoration through green transformation. In May 2025, the “Opinions on Continuously Promoting Urban Renewal Actions” further listed “restoring urban ecosystems” as one of its eight key tasks. It required the establishment of an urban inspection and assessment system that incorporates green development indicators into annual inspections and quinquennial evaluations to ensure the simultaneous advancement of urban renewal and ecological resilience [8]. These policy initiatives reflect the differential impacts of various urbanization stages on ecological resilience and provide a policy backdrop and practical basis for in-depth research on the relationship between urbanization levels and ecological resilience. However, during a period when resource-based cities face multiple challenges, such as environmental pollution, ecosystem degradation, and resource depletion, how exactly does urbanization construction affect their ecological resilience? What is the underlying mechanism of this impact? And does this effect exhibit heterogeneity? Existing research has yet to explore these questions in depth. Therefore, this paper takes the urbanization of resource-based cities as its starting point to systematically analyze the impact and mechanisms of urbanization levels on ecological resilience. This study is of significant theoretical and practical importance for comprehensively understanding the value of urban construction in resource-based cities, broadening the pathways to enhance urban ecological resilience, and providing references for achieving sustainable development in these cities.

2. Literature Review

2.1. Research Progress

Research relevant to this study primarily falls into three interconnected domains: urban ecological resilience, urbanization, and the interplay between urbanization and ecological resilience. To ensure a comprehensive and transparent synthesis of the existing evidence, the literature search strategy for this review was designed in accordance with the principles of systematic reviews. A systematic search was conducted in August 2025 across major academic databases, including the Web of Science (WoS) Core Collection, Scopus, and the China National Knowledge Infrastructure (CNKI). The search utilized the following key terms: (“urbanization” OR “urban expansion”) AND (“ecological resilience” OR “urban resilience”) AND (“resource-based city” OR “mining city” OR “industrial city”). The timeframe covered publications from the inception of each database until August 2025, and only publications in English were considered. After removing duplicates and screening titles and abstracts, the most relevant and influential studies were selected for full-text review and form the basis of this synthesis.

2.1.1. Studies on Ecological Resilience

The term “resilience” originates from Latin, meaning “the ability to rebound to a stable state”. This concept was initially applied in the field of ecology, where Holling [9] defined it as the capacity of an ecosystem to rapidly recover its original state or form a new equilibrium while maintaining its structure and function when subjected to external shocks. As scholarly research on “resilience” expanded, related studies gradually extended from natural ecology to human ecology and, further, to urban ecology, forming several distinct perspectives. Meerow et al. [10] contend that urban ecological resilience refers to the ability of an urban ecosystem to maintain resistance and recover swiftly when confronted with impacts or disturbances from uncertain factors. Balland and Rigby [11] place greater emphasis on the urban ecosystem’s capacity to absorb disturbances, achieve reorganization, and sustain development. A synthesis of existing research reveals a relatively unified conceptual framework within academia regarding urban resilience, primarily encompassing two core dimensions: first, the endogenous self-organizing and adaptive capacity within urban systems; and second, the ability to resist disturbances stemming from external environmental uncertainties and sudden risks. This consensus understanding emphasizes that cities, as complex adaptive systems, must simultaneously possess the dual characteristics of maintaining structural and functional stability and achieving dynamic transformation [6]. In recent years, research on urban ecological resilience has expanded from theoretical exploration to quantitative assessment. Some scholars have constructed process-oriented evaluation frameworks for urban ecological resilience [12], such as the “Driver-Pressure-State-Impact-Response” (DPSIR) model and the “Risk-Connectivity-Potential” framework, based on the ontological theory of resilience. Other scholars have developed evaluation index systems that integrate structure and function from an urban spatial perspective, such as the “Scale-Density-Coordination” model [13]. Concurrently, methods commonly employed for measuring urban ecological resilience include the entropy weight method [14], analytic hierarchy process (AHP) [15], and linear weighting method [16]. In terms of influencing factors of urban ecological resilience, numerous scholars have conducted in-depth explorations from various perspectives. Some Chinese researchers have identified government innovation preferences [17], population density [18], land-use changes [19], and urbanization [20] as significant factors enhancing ecological resilience. Other international studies suggest that alternative urban models (including urban morphology, land-use distribution, and connectivity) [21], avian biodiversity [22], and foreign direct investment [16] play crucial roles in promoting urban ecological resilience.

2.1.2. Studies on Urbanization

Urbanization serves as a significant indicator of urban development levels. Current research on urbanization has reached relative maturity. Tisdale [23] conceptualized urbanization as a process of population concentration, involving both the increase in number and expansion of cities, while examining its relationships with political, economic, and cultural dimensions [24,25,26]. Aslam et al. analyzed the causes, impacts, and policy implications of urbanization, examining its characteristics and effects from diverse theoretical perspectives such as liberalism, socialism, capitalism, and environmentalism [27]. McGranahan and Satterthwaite proposed that “urbanization involves not only population concentration but also a transformation of socioeconomic and governance structures,” emphasizing the critical role of service accessibility and governance frameworks in defining urbanization [28]. Research on urbanization measurement primarily focuses on constructing multidimensional characterization systems, with most scholars evaluating urbanization through demographic, land use, economic, and social dimensions [29,30], while others utilize single metrics such as nighttime light data [31,32]. In terms of research scale, studies predominantly concentrate on urban agglomerations [12] and economic zones [33,34]. The relationship between urbanization and ecological environment has been extensively investigated in academic circles. The existing literature typically employs spatial units at national, provincial, urban agglomeration, or individual city levels [35], utilizing econometric methods including BP neural networks, coupling coordination analysis, Tobit regression, and grey relational analysis to conduct empirical examinations of this relationship [36,37,38,39]. Sun et al. employed a geographically and temporally weighted regression model integrated with spatial econometric methods to analyze the relationship between urbanization and air pollution in China’s Yangtze River Delta region. Their study revealed that new-type urbanization effectively enhanced regional eco-environmental quality by optimizing resource allocation, promoting the application of green technologies, and facilitating pollution reduction [40]. Conversely, Bai et al. [41] systematically demonstrated that urbanization leads to ecosystem degradation through land expansion and surging energy consumption. Sun et al. [42] characterized the urbanization–ecological environment relationship as a complex dialectical unity. This relationship exhibits both symbiotic mutual reinforcement, where technological advancements and improved resource efficiency during urbanization contribute to ecological improvement, and fundamental contradictions, where environmental pollution and ecological damage pose significant threats to ecological systems.

2.1.3. Research Gaps and Contribution

Existing research findings provide both a theoretical foundation and empirical support for this study’s investigation into the impact of urbanization levels on the ecological resilience of resource-based cities. However, several research gaps remain: First, there is insufficient focus on the research subject. Few studies have examined the relationship between urbanization and “ecological resilience,” particularly regarding resource-based cities. Second, the comprehensiveness of the research content requires enhancement. Rarely have studies explored the mechanisms through which urbanization affects the ecological resilience of resource-based cities from the perspectives of industrial structure, green technology innovation, and energy utilization efficiency. Third, methodological limitations persist. Conventional econometric approaches suffer from statistical issues, including functional form misspecification, the curse of dimensionality, and multicollinearity, which compromise the accuracy and reliability of models. This study addresses these limitations through three principal innovations: First, it focuses specifically on resource-based cities to examine how urbanization affects urban ecological resilience, thereby guiding sustainable development in such cities. Second, it investigates the mechanisms through which urbanization influences ecological resilience via industrial structure optimization, green technology innovation, and energy efficiency improvement. Third, it employs a double machine learning model for regression analysis, effectively overcoming the limitations of traditional causal inference methods and significantly enhancing the precision of the estimation results.

2.2. Theoretical Analysis and Research Hypotheses

2.2.1. Direct Impact of Urbanization Level on Urban Ecological Resilience

Urbanization generally refers to the dynamic process of rural-to-urban migration coupled with concomitant agglomeration of non-agricultural industries. Its essence is driven by the transformation from a monostructure dominated by resource extraction and primary processing to a diversified structure characterized by advanced manufacturing, modern services, and green technology [43]. This process signifies not merely the reallocation of population, land, and capital but also a systematic reshaping of energy utilization patterns, environmental governance models, and risk management systems. Based on this understanding, the impact of urbanization on the ecological resilience of resource-based cities can be interpreted through three theoretical frameworks: ecological modernization theory, urban ecological transition theory, and resilient city theory.

Ecological modernization theory posits that when urbanization transitions from being driven by “resources and scale” to being driven by “technology and institutions,” cities can leverage ecological advantages to foster modernization, achieving both economic prosperity and ecological conservation [44]. For resource-based cities, the continuous advancement of urbanization induces green technological innovation, industrial chain extension, and enhanced environmental regulations, thereby significantly improving the ecosystem’s adaptive capacity to resource depletion, price fluctuations, and climate shocks [45,46]. Urban ecological transition theory further indicates that while urbanization intensifies ecosystem pressure through expanded production and consumption, it also drives the evolution of urban governance systems toward greater adaptability due to rising incomes, shifting values, and increasing environmental demands [47]. For resource-based cities, the evolutionary pathway of ecological resilience is ultimately determined by the dynamic interplay between the accumulated ecological vulnerability from the expansion of resource-dependent industries and the growing demand for enhanced environmental governance during urban development. Resilient city theory emphasizes the redundancy, diversity, and self-organizing capacity of urban systems [48]. Through the appropriate concentration of population and industries, urbanization facilitates infrastructure sharing, economies of scale in environmental governance, and rapid dissemination of risk information, thereby strengthening the recovery and adaptive capacities of resource-based cities when confronted with resource exhaustion, market volatility, and extreme events [10]. Accordingly, this study proposes the following research hypothesis:

Hypothesis 1:

The urbanization level exerts a significantly positive influence on the ecological resilience of resource-based cities.

2.2.2. Indirect Effects of Urbanization Level on Urban Ecological Resilience

Drawing upon economic logic and the existing literature, this study posits that the impact mechanism of urbanization on the ecological resilience of resource-based cities may manifest through the following pathways: First, the industrial structure upgrading pathway: Urbanization drives resource-based cities’ transition from singular extraction or primary processing toward advanced manufacturing and producer services through spatial reallocation of resources, production factors, and markets. Service-oriented industrial structures exhibit characteristics of low energy consumption, low emissions, and high value-added output, thereby reducing the intensity of natural capital utilization while enhancing ecosystem redundancy and stability [49,50]. Furthermore, increased product complexity resulting from industrial chain extension and value chain upgrading helps mitigate risks associated with resource price fluctuations, thereby strengthening the economic subsystem’s capacity to withstand shocks. However, when urbanization excessively relies on the scale expansion of resource-intensive industries, it reinforces path dependency in economic development and triggers an industrial “lock-in effect.” This effect significantly inhibits the development and enhancement of urban ecological resilience by constraining industrial transformation space and crowding out ecological governance resources [51]. Second, the green technology innovation pathway: Knowledge spillovers, industry–university–research clustering, and enhanced environmental regulation intensity induced by urbanization create economies of scale and scope conditions for green technology’s development and diffusion [52]. Widespread adoption of green process innovations and environmental governance technologies can reduce the ecological footprint per unit output while restoring damaged ecosystem functions [53]. Concurrently, technologies such as digital monitoring, smart water management, and ecological IoT enhance cities’ capacity to perceive and respond to sudden environmental incidents, strengthening the learning and adaptive capabilities of ecosystems [54]. Third, the energy efficiency pathway: Urbanization significantly improves energy utilization efficiency and optimizes the energy structure through infrastructure sharing, public transportation network development, and distributed energy system deployment [55]. Specifically, a compact urban form and economies of scale contribute to improved energy efficiency, which helps mitigate the pressures associated with high per capita energy consumption and significantly reduces energy use per unit of GDP. Meanwhile, an increasing share of renewable energy, coupled with a declining proportion of coal consumption, will drive the transition toward a cleaner energy structure, thereby reducing the environmental impact of energy use [56]. Additionally, enhanced environmental awareness and willingness to pay among urban residents further create social demand foundations for efficient and clean energy investments [57]. Based on these integrated mechanisms, this study proposes the following research hypothesis:

Hypothesis 2:

Urbanization level exerts positive promoting effects on the ecological resilience of resource-based cities through industrial structure advancement, enhancement of green technology innovation, and improvement of energy utilization efficiency.

Synthesizing the preceding theoretical foundations, Figure 1 graphically depicts the conceptual framework constructed in this study:

3. Methodology and Analysis

3.1. Study Area

China’s resource-based cities, as a distinct urban category formed during the rapid industrialization process, emerged during the large-scale industrialization period after the founding of the People’s Republic of China (1949–1957). Primarily located in regions rich in resources such as minerals and forests—including northeast China and northwest China—these cities developed around resource extraction and processing activities, with their spatial layout and industrial systems structured to support these functions. Under the planned economy system at the time, resource-based cities played a critical role in supplying energy and raw materials for national development, making their growth intrinsically linked to resource exploitation. However, with the gradual depletion of natural resources, these cities now face multiple challenges, such as sluggish economic growth, a monolithic industrial structure, and ecological degradation. As one of the pioneers of rapid urbanization, exploring pathways for sustainable urban construction and enhancing urban ecological resilience have become urgent issues for resource-based cities to achieve economic transformation and promote high-quality development.

China’s “National Sustainable Development Plan for Resource-Based Cities (2013–2020)” identified 262 resource-based cities in 2013. From this comprehensive inventory, 114 prefecture-level cities were selected as research samples based on complete data availability, consistent administrative levels, and sustained resource-based characteristics, following the exclusion of regions with severe data deficiencies. The final sample encompasses diverse resource types and regional variations. According to the established policy classification framework, these cities are categorized into four developmental types: growing, mature, declining, and regenerative resource-based cities. The spatial distribution and categorical classification are presented in Figure 2 and

Table 1. Classification of resource-based cities in China.

Resource-Based City Type	Cities
Growing Type (15)	Shuozhou, Hulunbuir, Ordos, Songyuan, Hezhou, Nanchong, Liupanshui, Bijie, Zhaotong, Yan’an, Xianyang, Yulin, Wuwei, Qingyang, Longnan
Mature Type (61)	Zhangjiakou, Chengde, Xingtai, Handan, Datong, Yangquan, Changzhi, Jincheng, Xinzhou, Jinzhong, Linfen, Yuncheng, Lüliang, Chifeng, Benxi, Jilin, Heihe, Daqing, Jixi, Mudanjiang, Huzhou, Suzhou, Bozhou, Huainan, Chuzhou, Chizhou, Xuancheng, Nanping, Sanming, Longyan, Ganzhou, Yichun, Dongying, Jining, Tai’an, Laiwu, Sanmenxia, Hebi, Pingdingshan, Ezhou, Hengyang, Chenzhou, Shaoyang, Loudi, Yunfu, Baise, Hechi, Guangyuan, Guang’an, Zigong, Panzhihua, Dazhou, Ya’an, Anshun, Qujing, Baoshan, Pu’er, Lincang, Weinan, Baoji, Jinchang, Pingliang
Declining Type (23)	Wuhai, Fuxin, Fushun, Liaoyuan, Baishan, Yichun, Hegang, Shuangyashan, Qitaihe, Huaibei, Tongling, Jingdezhen, Xinyu, Pingxiang, Zaozhuang, Jiaozuo, Puyang, Huangshi, Shaoguan, Luzhou, Tongchuan, Baiyin, Shizuishan
Regenerating Type (15)	Tangshan, Baotou, Anshan, Panjin, Huludao, Tonghua, Xuzhou, Suqian, Ma’anshan, Zibo, Linyi, Luoyang, Nanyang, Lijiang, Zhangye

Note: The city classifications in Figure 2 and Table 1 are derived from China’s “National Sustainable Development Plan for Resource-Based Cities” (2013–2020).

. Cities in the growing stage exhibit expanding resource exploitation and rapid economic growth, while demonstrating the initial emergence of social functionality and ecological concerns. Mature cities maintain peak resource development and economic stability, yet they confront accumulating social conflicts and progressive ecological degradation. Declining cities experience resource depletion and economic recession, accompanied by exacerbated social problems and severe ecological damage. Regenerative cities achieve successful economic transition through diversification, accompanied by gradual ecological rehabilitation and social restoration.

China’s “National Sustainable Development Plan for Resource-Based Cities (2013–2020)” formally defines resource-based cities as urban entities whose core industrial activities center on the extraction and processing systems of local natural resources, including minerals and forests. The document further emphasizes that these cities serve not only as crucial safeguard zones for China’s energy and strategic resource supply but also as pivotal supports for maintaining stable national economic operation. Their sustainable transformation carries significance that extends beyond the fundamental restructuring of economic development patterns and the realization of modernization goals. Furthermore, it embodies multifaceted strategic importance, including regional development coordination, synchronized advancement of new industrialization and urbanization, social stability maintenance, and ecological civilization construction.

3.2. Model Specification

(1)

Double Machine Learning (DML) Model

This study investigates the impact of urbanization levels on the ecological resilience of resource-based cities. While existing research predominantly employs traditional causal inference methods for parameter estimation, such approaches often suffer from model specification bias, the curse of dimensionality, and multicollinearity, potentially leading to biased estimates. To address these limitations, this study adopts the DML model proposed by Chernozhukov et al. [58] for causal inference. As a semiparametric method based on Neyman orthogonality, DML effectively separates the estimation of high-dimensional control variables from that of the target parameter. Specifically, the model first employs machine learning algorithms to flexibly estimate the relationships between the treatment and outcome variables and the covariates. It then constructs an unbiased estimator using the orthogonalized residuals. This design enables DML to effectively avoid the estimation bias caused by model misspecification in traditional parametric methods while maintaining strong adaptability to high-dimensional data [59]. However, the validity of DML relies on several key assumptions. First, the machine learning algorithm must accurately identify covariates related to the treatment variable from the data, and these covariates should exhibit sufficiently strong correlation with the endogenous treatment variable to ensure effective prediction. Second, these covariates must satisfy the exogeneity condition, meaning that they are uncorrelated with potential unobserved confounders. This ensures that variations in the treatment variable can be identified independently of other unobserved factors, thereby preserving the unbiasedness of the causal effect estimation [60].

In terms of model implementation, this study employs cross-fitting to balance sample utilization efficiency and estimation unbiasedness. Specifically, the original sample is randomly partitioned into five subsets. Iteratively, one subset is designated as the validation sample, while the remaining four serve as the training sample. Both the treatment model and the outcome model are fitted separately in each iteration, and the treatment effect estimator is ultimately constructed through orthogonalization. This iterative process not only fully utilizes sample information but also effectively mitigates the risk of overfitting, thereby providing a reliable foundation for subsequent estimation. Furthermore, to address unobserved heterogeneity and potential endogeneity concerns, DML constructs the estimation equation using an orthogonal score function. This approach ensures consistent estimation even in the presence of high-dimensional control variables, thereby partially alleviating bias caused by omitted variables. Additionally, DML automatically selects control variables by employing machine learning algorithms to model and adjust for potential confounders. In the first stage, the algorithm evaluates the predictive contribution of each covariate through variable importance measures, identifying key control variables for use in the second-stage estimation of the causal effect of the treatment variable on the outcome variable [58].

Based on the data characteristics and model requirements, this study employs Random Forest as the core machine learning algorithm [61]. This method constructs multiple decision trees through bootstrap sampling and random feature selection, and then integrates their predictions to effectively capture complex nonlinear relationships and interaction effects among variables. To further enhance the model performance, a combination of grid search and cross-validation is adopted to optimize the hyperparameters of the Random Forest algorithm, including the number of trees and the maximum tree depth. The optimal parameter combination is determined by minimizing the mean squared error. The specific models constructed based on the above analysis are presented in Equations (1) and (2):

URBAN_i,t = θ₀ECR_i,t + g(X_i,t) + U_i,t

(1)

E(U_i,t|ECR_i,t,X_i,t) = 0

(2)

where i represents the city, t represents the year, URBAN_i,t represents the urbanization level of city i in year t, ECR_i,t represents the ecological resilience of city i in year t, θ₀ represents the coefficient of the key disposal variable, X_i,t represents a set of control variables, and the specific form g(X_i,t) must be estimated using machine learning algorithms. U_i,t represents the error term.

If Equations (1) and (2) are used directly for estimation, the estimated values $\widehat{\mathsf{\theta }}$₀ will be biased. This is because in high-dimensional or complex model settings, machine learning must introduce regularization terms to reduce the dimensionality. Although this avoids excessive variance in the estimate, it also introduces regularization bias in the function, making it difficult for $\widehat{\mathsf{\theta }}$ to converge to θ₀. Therefore, this paper constructs an auxiliary regression for parameter estimation, as shown in Equations (3) and (4):

ECR_i,t = m(X_i,t) + V_i,t

(3)

E(V_i,t|X_i,t) = 0

(4)

where m(X_i,t) is the regression function of the response variable on high-dimensional control variables, which is obtained through machine learning algorithms. The specific method is as follows: First, use DML to estimate m(X_i,t), obtaining the estimator $\widehat{\mathrm{m}}\left({\mathrm{X}}_{\mathrm{i},\mathrm{t}}\right)$. Then, calculate the residual term V_i,t as $\widehat{\mathrm{V}}$_it = URBAN_i,t − $\widehat{\mathrm{m}}$(X_i,t). Next, use $\widehat{\mathrm{V}}$_i,t as an instrumental variable to estimate ECR_i,t. Finally, continue using machine learning algorithms to estimate the function g(X_i,t), obtaining the estimator $\widehat{\mathrm{g}}\left({\mathrm{X}}_{\mathrm{i},\mathrm{t}}\right)$, and derive the unbiased estimate ${{\widehat{\mathsf{\theta }}}_0 = \left({\displaystyle \frac{1}{\mathrm{n}}}\sum _{\mathrm{i}\in 1,\mathrm{t}\in \mathrm{T}}{\widehat{\mathrm{V}}}_{\mathrm{i},\mathrm{t}}{\mathrm{ECR}}_{\mathrm{i},\mathrm{t}}\right)}^{-1}{\displaystyle \frac{1}{\mathrm{n}}}\sum _{\mathrm{i}\in 1,\mathrm{t}\in \mathrm{T}}{\widehat{\mathrm{V}}}_{\mathrm{i},\mathrm{t}}\left({\mathrm{URBAN}}_{\mathrm{i},\mathrm{t}}-\widehat{\mathrm{g}}\left({\mathrm{X}}_{\mathrm{i},\mathrm{t}}\right)\right)$.

(2)

Instrumental Variables Model Based on DML

In the benchmark regression, although this study mitigates the interference of endogeneity issues on the estimation results by controlling for high-dimensional fixed effects and adding as many variables as possible that affect urban ecological resilience, endogeneity issues may still arise due to omitted variables, measurement errors, and other factors. Therefore, this paper draws on the research of Chernozhukov et al. [58] to construct a partial linear instrumental variable model based on DML to eliminate potential endogeneity risks. The model settings are shown in Equations (5) and (6):

URBAN_i,t = θ₀ECR_i,t + g(X_i,t) + U_i,t

(5)

IV_i,t = m(X_i,t) + V_i,t

(6)

where IV_i,t is the instrumental variable for ECR_i,t.

(3)

Mechanism Testing Model Based on DML

To further test how urbanization levels affect urban ecological resilience through mechanism variables, this paper constructs the following mechanism testing model based on DML:

M_i,t = γ₀ ECR_i,t +g(X_i,t)+ U_i,t

(7)

E(U_i,t|ECR_i,t, X_i,t) = 0

(8)

ECR_i,t = m(X_i,t)+ V_i,t

(9)

E(V_i,t|X_i,t) = 0

(10)

where γ₀ reflects the impact of urbanization levels on the mechanism variable M. Its significant coefficient indicates that urbanization levels can influence urban ecological resilience through the mechanism variable.

3.3. Variable Selection

3.3.1. Dependent Variable: Urban Ecological Resilience (ECR)

Urban ecological resilience reflects the capacity of an ecosystem to promptly respond to external disturbances or shocks while achieving systemic restructuring and sustainable development. Drawing on methodologies from Yuan et al. [62] and Li et al. [63], this study constructs an evaluation index system based on three attribute dimensions: state, pressure, and response. Pressure denotes the burden imposed on ecosystems through the interaction of human activities and natural processes, state describes the equilibrium characteristics of the natural environment under external disturbances, and response refers to the ability of human societies to restore or optimize urban system functions through prevention, adaptation, or improvement strategies when confronting shocks from natural and social factors [10]. To mitigate potential weight distortion issues associated with the entropy weight method, the entropy-weighted TOPSIS [64] method is employed to quantify the ecological resilience levels of resource-based cities. The specific indicator system is detailed in

Table 2. Evaluation index system for urban ecological resilience.

Primary Indicator	Secondary Indicator	Tertiary Indicator (Direction)	Unit	Source
Ecological Resilience	Status	Per capita water resources (+)	m3/person	Statistical yearbooks of various cities
		Urban green space coverage rate (+)	%	Statistical yearbooks of various cities
		Per capita park green space area (+)	Hectares/10,000 people	Statistical yearbooks of various cities
		Per capita urban area (+)	km2/10,000 people	Statistical yearbooks of various cities
	Pressure	Per capita industrial wastewater discharge (−)	Tons/person	Statistical yearbooks of various cities
		Per capita industrial sulfur dioxide discharge (−)	Tons/person	Statistical yearbooks of various cities
		Per capita industrial smoke and dust discharge (−)	Tons/person	Statistical yearbooks of various cities
		Per capita industrial nitrogen oxides discharge (−)	Tons/person	Statistical yearbooks of various cities
		Annual average concentration of PM2.5 (−)	μg/m3	Statistical yearbooks of various cities
	Response	Industrial sulfur dioxide removal (+)	Tons	Statistical yearbooks of various cities
		Industrial smoke and dust removal (+)	Tons	Statistical yearbooks of various cities
		Harmless treatment rate of domestic waste (+)	%	Statistical yearbooks of various cities
		Centralized treatment rate of sewage treatment plants (+)	%	Statistical yearbooks of various cities
		Comprehensive utilization rate of industrial solid waste (+)	%	Statistical yearbooks of various cities

.

3.3.2. Core Explanatory Variable: Urbanization Level (URBAN)

The most commonly used indicator for measuring urbanization levels is the officially published proportion of the permanent urban population. However, this metric fails to capture the spatial agglomeration intensity of population, potentially overlooking essential urban characteristics [65]. Consequently, relying solely on urban population proportion may inadequately reflect actual urbanization levels, thereby compromising the credibility of research findings. This study employs nighttime light data to quantify urbanization, encompassing both depth and breadth dimensions: the former reflects urban development intensity, while the latter indicates spatial expansion extent. Using stable light density values alone would only capture development intensity, without accounting for spatial propagation. Thus, drawing on methodologies from Yi et al. [66], this study uses nighttime light data released by the National Oceanic and Atmospheric Administration National Geophysical Data Center (NOAA) to calculate indicators that more comprehensively reflect the level of urban development. The specific calculation formula is as follows:

$${\mathrm{URBAN}}_{\mathrm{i}} = \mathsf{\phi }\mathrm{URBAN}{1}_{\mathrm{i}}+\left(1-\mathsf{\phi }\right)\mathrm{URBAN}{2}_{\mathrm{i}}$$

(11)

$$\mathrm{URBAN}{1}_{\mathrm{i}}={\sum }_{\mathrm{j}=1}^{63}\mathrm{D}{\mathrm{N}}_{\mathrm{j}}\times {\displaystyle \frac{{\mathrm{n}}_{\mathrm{j}}}{\mathrm{N}\times 63}}$$

(12)

$$\mathrm{URBAN}{2}_{\mathrm{i}}={\displaystyle \frac{{\mathrm{S}}_{\mathrm{N}}}{\mathrm{S}}}$$

(13)

where URBAN1 represents the urbanization depth measured by average light intensity, URBAN2 denotes the urbanization breadth quantified by the spatial attributes of regional lighting, and $\mathsf{\phi }$ (0 < $\mathsf{\phi }$ < 1) indicates the weight assigned to urban1. This study considers both urbanization depth and breadth to be equally important, thus setting $\mathsf{\phi }$= 0.5 in baseline regressions. DN_j represents the j-th digital number (DN) value (ranging from 1 to 63, with 63 being the maximum DN value for stable light data) in region i, n_j denotes the pixel count at the j-th DN level, and N indicates the total number of light-emitting pixels in region i. URBAN1 characterizes the ratio between actual light intensity and maximum potential intensity in region i, while URBAN2 is defined as the ratio of the total area of lit pixels (S_N) to the region’s total area (S), measuring the spatial proportion of light coverage to reflect urban spatial expansion characteristics. Consequently, this study employs both URBAN1 and URBAN2 to measure urbanization from the depth and breadth dimensions, respectively.

3.3.3. Control Variables

Other variables may also influence the effectiveness of urbanization levels. Beyond examining the impact of urbanization on urban ecological resilience, it is essential to control for additional factors affecting ecological resilience. Drawing on the literature from Liu et al. [67] and Xu et al. [68], the following control variables were selected: (1) Environmental Governance Capacity (ENV): Effective environmental governance mitigates environmental degradation and protects ecosystems, providing a stable foundation and dynamic impetus for urban ecological resilience, measured by the proportion of environmental protection investment to GDP. (2) Economic Growth (GDP): Economic development serves as crucial support for enhancing urban ecological resilience. Higher economic development levels expand available means and strengthen capacities for improving ecological resilience, measured by per capita GDP. (3) Population Density (POP): Population density affects resource demands and environmental pressures, driving societies to explore more efficient green technologies and management innovations, thereby potentially increasing urban ecological resilience, measured by the ratio of year-end population to urban area. (4) Openness Level (OPEN): Openness represents a key indicator of urban development potential. While foreign-invested resource-intensive enterprises contribute to economic growth, they may simultaneously exert environmental stress effects, measured by the share of foreign direct investment in GDP. (5) Land-use intensity (LU): Land use constitutes the most direct expression of human–nature interaction, where varying degrees of development activities across different land types yield corresponding ecological consequences. During urbanization, land types experiencing intensive human activity frequently exhibit landscape fragmentation to varying extents, thereby influencing urban ecological resilience, measured by the proportion of construction land area to total urban area. (6) Infrastructure Development Level (INF): Infrastructure embodies the physical manifestation of urban public services. More comprehensive transportation networks enhance urban circulatory capacity, measured by per capita road area.

3.3.4. Mechanism Variables

Based on the preceding mechanistic analysis, this study selects industrial structure (SEC), green technology innovation level (TEC), and energy utilization efficiency (ENE) as mechanism variables to investigate how urbanization influences the ecological resilience of resource-based cities. (1) Industrial Structure (SEC): Industrial structure serves as a critical driver for enhancing urban ecological resilience. By revolutionizing production modes and adjusting industrial proportions, it promotes rationalization and advanced transformation of industries, directly contributing to energy conservation and emission reduction. The output value ratio of secondary and tertiary industries measures this variable. (2) Green Technology Innovation Level (TEC) [69]: Green technological innovation enhances resource utilization efficiency, reduces pollution emissions, and facilitates industrial green transition. It significantly strengthens the resistance, adaptation, and recovery capacities of urban ecosystems when confronting climate change and extreme events [70]. The number of authorized green invention patents quantifies this variable. (3) Energy Utilization Efficiency (ENE) [71]: Improved energy efficiency reduces energy consumption and emissions per unit of urban output, alleviating pressure on ecosystems. Through technological upgrades and policy coordination, it enhances cities’ adaptation, recovery, and transformation capabilities against climate shocks, thereby significantly boosting urban ecological resilience. This variable is measured by GDP per unit of energy consumption.

3.4. Data Sources

Urbanization level data were obtained from global nighttime light data (2010–2023) published by the National Geophysical Data Center of the National Oceanic and Atmospheric Administration (NOAA). Data for urban ecological resilience and other variables were primarily sourced from the China City Statistical Yearbook, China Regional Economic Statistical Yearbook, China Urban Construction Statistical Yearbook, municipal statistical yearbooks, and statistical bulletins on national economic and social development for each study unit (2010–2023). Missing data were supplemented using linear interpolation. Descriptive statistics for all variables are presented in

Table 3. Descriptive statistics.

Variable	N	Mean	SD	Min	Max
URBAN	1596	4.607	4.322	0.337	25.62
ECR	1596	0.312	0.025	0.124	0.361
SEC	1596	0.464	0.122	0.000	0.822
TEC	1596	17.22	32.88	0.000	393
ENE	1596	0.876	0.495	0.193	4.132
ENV	1596	42.05	16.05	10.31	100.3
GDP	1596	10.65	0.554	8.773	12.49
POP	1596	6.177	1.053	2.503	15.47
OPEN	1596	0.081	0.115	0.000	1.458
LU	1596	0.121	2.039	0.000	81.38
INF	1596	18.27	8.682	1.370	61.41

Note: All variables are defined in detail in Supplementary Table S1.

.

4. Results

4.1. Spatial Distribution Characteristics of Urban Ecological Resilience

Using the natural breaks classification method, this study categorizes the ecological resilience of resource-based cities into five levels: low, relatively low, medium, relatively high, and high. Spatial visualization was conducted using ArcGIS 10.8, as illustrated in Figure 3.

(1)

In 2010, China’s resource-based cities were predominantly characterized by medium-level ecological resilience. Over 70% of these cities fell within the medium or lower resilience categories, forming a C-shaped spatial pattern. During this period, the mining industry contributed significantly to GDP in most resource-based cities, resulting in high ecological disturbance intensity. In the aftermath of the financial crisis, local governments prioritized economic growth, leading to reduced environmental governance expenditures. Additionally, China’s deposit system for mine geological environment restoration was still in its pilot phase, limiting the scale of ecological remediation efforts.

(2)

In 2014, a gradient in ecological resilience began to emerge. The medium-resilience zone expanded notably northward and eastward, forming a belt of relatively high ecological resilience spanning Shandong, Jiangsu, and Anhui provinces. The implementation of the National Sustainable Development Plan for Resource-Based Cities (2013–2020), which designated 262 resource-based cities, facilitated fiscal transfers and subsidies for resource-exhausted cities. These measures promoted ecological governance in central China. Eastern coastal cities such as Xuzhou and Zaozhuang were integrated into the Yangtze River Delta’s industrial division through high-speed rail networks, increasing their tertiary industry share and gradually reducing ecological pressure.

(3)

In 2018, ecological resilience exhibited overall improvement. High-resilience areas emerged for the first time in the Yangtze River Delta, Pearl River Delta peripheries, and Shandong Peninsula, while a relatively high-resilience zone formed around China’s Yellow River Horseshoe Bend. In contrast, southwestern resource-based cities displayed lower resilience levels. Stringent policies, including the Soil Pollution Prevention and Action Plan and central environmental inspections, accelerated the adoption of Public–Private Partnership (PPP) models for ecological restoration. Eastern cities reduced their industrial land use, shifting toward the service and commercial sectors, thereby enhancing their ecological resilience. Western cities, however, resumed resource-dependent development due to rising resource prices, resulting in delayed ecological restoration.

(4)

In 2023, spatial patterns of ecological resilience became consolidated. Driven by China’s dual-carbon goals (carbon peak and neutrality) and digital transformation, resilience distribution converged toward a “high-level stabilization with eastern concentration and western dispersion” pattern. Regenerative cities in eastern China and the peripheries of the Yangtze River Delta and Pearl River Delta predominantly achieved high resilience, emerging as leaders under the dual-carbon framework. Mature cities in central China transitioned to relatively high resilience levels. Growing cities in western China developed medium-resilience clusters, supported by the Integrated Protection and Restoration Project of Mountains, Waters, Forests, Farmlands, Lakes, Grasslands, and Deserts and investments in new energy. However, persistent resource dependence led to polarized intra-regional disparities. Overall, the spatial resilience pattern stabilized in 2023, with regional differences remaining the primary source of overall variation.

4.2. Empirical Analysis

4.2.1. Baseline Regression

This study employed DML to examine the impact of urbanization on the ecological resilience of resource-based cities. The sample was split in a 1:4 ratio, and the Random Forest algorithm was applied to solve the primary and auxiliary regressions. The regression results are presented in

Table 4. Baseline regression results.

Variant	(1)	(2)	(2)	(3)
	ECR	ECR	ECR	ECR
URBAN	0.304 ***	0.115 ***	0.225 ***	0.174 ***
	(0.083)	(0.019)	(0.062)	(0.010)
Controls	Yes	Yes	Yes	Yes
Controls-Squared	Yes	Yes	Yes	Yes
Time FE	No	Yes	No	Yes
City FE	No	No	Yes	Yes
N	1596	1596	1596	1596

Note: *** indicate significance levels of 1%; the values in parentheses below the coefficients are their standard errors; FE denotes fixed effects; the following tables have the same interpretation. All variables are defined in detail in Supplementary Table S1.

. Columns (1) to (4) report the results without fixed effects, with only time fixed effects, with only city fixed effects, and with both time and city fixed effects, respectively. The estimated effects are 0.304, 0.115, 0.225, and 0.174, respectively, all statistically significant at the 1% level, indicating that urbanization significantly enhances the ecological resilience of resource-based cities. These empirical findings align with the theoretical analysis, validating Hypothesis H1.

On the one hand, urbanization facilitates the concentration of population, capital, and technology in urban areas, enabling large-scale investments in green infrastructure and improving the coverage and efficiency of public services such as wastewater treatment, solid waste recycling, and mine restoration. On the other hand, urbanization promotes industrial diversification and value chain extension, gradually reducing reliance on highly polluting and energy-intensive resource-based industries. This shift drives the transformation of traditional mining sectors toward green manufacturing and circular economy practices, thereby mitigating ecological pressures at their source. Furthermore, higher levels of urbanization are often accompanied by stricter environmental regulations and advanced governance tools (emissions trading systems, ecological compensation mechanisms, and smart environmental platforms). Simultaneous enhancements in governmental fiscal capacity and public demand for environmental protection create a sustained positive feedback mechanism.

4.2.2. Robustness Tests

To ensure the robustness of the baseline regression results, this study conducted additional tests by replacing the explained variable, excluding outliers, and modifying the machine learning model specifications. First, this study employed alternative measures for the explained variable. Following methodologies reported by Li et al. [72], this study substituted the urbanization level with the ratio of urban permanent population to the total permanent population. The results in Column (1) of

Table 5. Robustness test results.

Variant	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
	Replace Variables	Partial Samples	1% Tailing	5% Tailing	1:2	1:7	LassoCV	GBDT
URBAN	2.230 ***	0.255 ***	0.127 **	0.156 ***	0.136 ***	0.125 ***	0.150 ***	0.148 ***
	(0.047)	(0.074)	(0.058)	(0.049)	(0.019)	(0.060)	(0.043)	(0.032)
Controls	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Controls-Squared	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Time FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
City FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
N	1596	1140	1596	1596	1596	1596	1596	1596

Note: ***, ** indicate significance levels of 1%, 5%, respectively. All variables are defined in detail in Supplementary Table S1.

confirm the robustness of the baseline findings. Second, to mitigate potential distortions from the COVID-19 pandemic, this study excluded post-2019 samples and restricted the analysis to the 2010–2019 period. As shown in Column (2) of Table 5, the regression coefficient for urbanization remained statistically significant at the 5% level, supporting the original conclusion. Third, this study addressed outliers. Given the substantial disparities among resource-based cities in economic scale and geographical attributes, extreme values may bias the results. To minimize such interference, this study winsorized all variables at the 1% and 5% levels. Columns (3)–(4) of Table 5 demonstrate that the core findings remain robust after outlier treatment. Fourth, this study reconfigured the machine learning model. To control for potential estimation biases arising from model specification, this study adjusted the original 1:4 sample splitting ratio to 1:2 and 1:7, and it sequentially replaced the Random Forest algorithm with Lasso Regression (LassoCV) and Gradient-Boosting Decision Trees (GBDT). Columns (5)–(8) of Table 5 show that the regression coefficients for urbanization remain significantly positive at the 1% level under altered sample splits and machine learning methods, further validating the robustness of the baseline results.

4.2.3. Endogeneity Tests

Given that urbanization levels are not randomly assigned, potential endogeneity concerns may arise in estimating their relationship with ecological resilience in resource-based cities. Specifically, endogeneity primarily stems from two sources: first, reverse causality, whereby cities with higher ecological resilience may exhibit greater attractiveness to population and industries, thereby further driving urbanization; second, omitted-variable bias, as although the model controls for observable factors such as economic scale, population density, and openness, it may still fail to fully capture unobserved confounders like institutional quality and public environmental awareness.

To address endogeneity, this study employs the relief degree of land surface (REL) as an instrumental variable. As REL is cross-sectional data without time-varying characteristics, it cannot be directly applied to panel data analysis. Following Henderson’s approach [73], this study combines REL with a time trend term to construct an instrumental variable for regression analysis. The validity of the instrumental variable relies on relevance and exogeneity. Regarding relevance, terrain roughness constrains urbanization by increasing construction costs, limiting the developable land area, and inhibiting population and industrial agglomeration. In terms of exogeneity, as a natural geographical factor, REL remains invariant to short-term socioeconomic changes, does not directly affect ecological resilience, and exhibits no systematic correlation with unobserved omitted variables. Additionally, drawing on methodologies from Arellano et al. [74] and Blundell et al. [75], this study re-estimates the model using the first-order (L1.URBAN) and second-order (L2.URBAN) lags of urbanization as instrumental variables. Urbanization exhibits significant temporal persistence, meaning that current urbanization levels largely depend on historical levels. Thus, lagged terms effectively capture variations in contemporary urbanization, satisfying the relevance condition. In panel data settings, lagged urbanization terms are pre-determined and unaffected by contemporaneous error terms. Assuming no higher-order serial correlation in disturbances, these lagged variables serve as valid instruments meeting exclusion restrictions. The endogeneity test results, obtained using the previously specified DML partial linear instrumental variable model, are presented in Columns (1)–(3) of

Table 6. Instrumental variable regression results.

Variant	(1)	(2)	(3)
	IV = REL	IV = L1.URBAN	IV = L2.URBAN
URBAN	0.174 ***	1.101 ***	0.217 ***
	(0.035)	(0.056)	(0.025)
Controls	Yes	Yes	Yes
Controls-Squared	Yes	Yes	Yes
Time FE	Yes	Yes	Yes
City FE	Yes	Yes	Yes
N	1596	1596	1596

Note: *** indicate significance levels of 1%. All variables are defined in detail in Supplementary Table S1.

. The coefficients remain positive and statistically significant, consistent with the baseline regression results.

4.2.4. Mechanism Analysis

Theoretical analysis suggests that urbanization may influence the ecological resilience of resource-based cities through three pathways: industrial structure (SEC), green technology innovation (TEC), and energy utilization efficiency (ENE). This study empirically examines these transmission mechanisms using the previously constructed DML model, with detailed results presented in

Table 7. Mechanism test results.

Variant	(1)	(2)	(3)	(4)	(5)	(6)
	SEC	ECR	TEC	ECR	ENE	ECR
URBAN	0.198 ***		1.833 ***		0.276 ***
	(0.001)		(0.012)		(0.004)
SEC		1.485 ***
		(0.020)
TEC				0.330 *
				(0.175)
ENE						1.508 ***
						(0.015)
Controls	Yes	Yes	Yes	Yes	Yes	Yes
Controls-Squared	Yes	Yes	Yes	Yes	Yes	Yes
Time FE	Yes	Yes	Yes	Yes	Yes	Yes
City FE	Yes	Yes	Yes	Yes	Yes	Yes
N	1596	1596	1596	1596	1596	1596

Note: ***, * indicate significance levels of 1%, 10%, respectively. All variables are defined in detail in Supplementary Table S1.

.

(1)

Industrial Structure

Theoretical insights indicate that urbanization can enhance urban ecological resilience by elevating industrial structure. To validate this mechanism, this study separately tested the effect of urbanization on industrial structure and the effect of industrial structure on ecological resilience. Columns (1) and (2) of Table 7 report the results. The estimated coefficients for both urbanization and industrial structure are significantly positive at the 1% level, confirming that urbanization improves industrial structure, which, in turn, promotes ecological resilience. These findings align with theoretical expectations, supporting Hypothesis 2. The underlying rationale is that urbanization facilitates industrial optimization and upgrading, leading to more efficient resource utilization and rationalized industrial structure, thereby reducing ecological damage. Simultaneously, industrial advancement drives transitions toward high-end, intelligent, and green development, enhancing value added and competitiveness while increasing corporate investment in and awareness of ecological protection. The synergistic interaction between urbanization and industrial upgrading collectively fosters ecological resilience.

(2)

Green Technology Innovation

To examine whether urbanization enhances ecological resilience through green technology innovation, this study applied the same mechanistic testing framework. Columns (3) and (4) of Table 7 show that the coefficients for urbanization and green technology innovation are significantly positive at the 1% level, verifying this mechanism and supporting Hypothesis 2. Urbanization provides resources and incentives for green technology innovation, which subsequently strengthens the stability and restorative capacity of urban ecosystems, thereby contributing to ecological resilience.

(3)

Energy Utilization Efficiency

The role of energy utilization efficiency as a mechanistic pathway was similarly tested. As shown in Columns (5) and (6) of Table 7, both urbanization and energy efficiency exhibit significantly positive coefficients at the 1% level, confirming the mechanism and validating Hypothesis 2. Urbanization improves infrastructure and promotes industrial agglomeration in resource-based cities, creating a foundation for enhanced energy efficiency. Improved energy efficiency reduces pollutant emissions, optimizes resource allocation, and strengthens ecological resilience. The significant positive relationship confirms that energy utilization efficiency serves as a functional channel through which urbanization fosters ecological resilience.

4.2.5. Heterogeneity Analysis

(1)

Heterogeneity Across Resource-Based City Types

Significant variations in resource endowment, industrial structure, and institutional environments among resource-based cities may obscure stage-specific and regional differences in the effect of urbanization on ecological resilience when using aggregate estimates. Following the classification framework of the National Sustainable Development Plan for Resource-Based Cities (2013–2020), this study categorizes the 114 sample cities into four types: growing, mature, declining, and regenerative resource-based cities (see Table 1 and Figure 2 for classifications). The results from the heterogeneity tests in

Table 8. Heterogeneity test results by resource-based city type.

Variant	Growing	Mature	Declining	Renewable
URBAN	0.375 ***	0.451 ***	0.419	1.162 ***
	(0.085)	(0.076)	(0.322)	(0.049)
Controls	Yes	Yes	Yes	Yes
Controls-Squared	Yes	Yes	Yes	Yes
Time FE	Yes	Yes	Yes	Yes
City FE	Yes	Yes	Yes	Yes
N	210	854	322	210

Note: *** indicate significance levels of 1%. All variables are defined in detail in Supplementary Table S1.

reveal divergent effects of urbanization on ecological resilience across city types. Urbanization significantly enhances ecological resilience in growing, mature, and regenerative cities, with the strongest effect observed in regenerative cities, followed by growing and mature cities. In contrast, urbanization shows no statistically significant effect on ecological resilience in declining resource-based cities.

Regenerative cities have typically overcome dependence on depleted resources and achieved initial industrial diversification, resulting in more resilient economic structures. In this context, urbanization advances are accompanied by improved infrastructure, more efficient environmental governance systems, and enhanced technological innovation and diffusion capabilities. These factors collectively increase resource-use efficiency, strengthen pollution control, and boost ecosystem recovery capacity, thereby maximizing the positive effects of urbanization on ecological resilience. In comparison, growing and mature resource-based cities remain highly dependent on mineral and other natural resources. Although urbanization may bring some technological and managerial advancements, their economic structures remain centered on resource extraction and primary processing. This resource dependence implies that urbanization may simultaneously intensify resource consumption and environmental pollution pressures. Moreover, resource rent effects may dampen incentives for transitioning toward cleaner and more efficient industrial structures. Consequently, while urbanization exerts a positive influence on ecological resilience, this effect is partially offset by environmental negative externalities stemming from resource dependence, resulting in smaller net gains compared to regenerative cities. Declining resource-based cities face structural issues driven by resource depletion. These cities often grapple with sharply reduced fiscal revenues, high unemployment, aging infrastructure, social instability, and substantial historical environmental liabilities. Under these conditions, urbanization advances rarely translate into sufficient financial, technological, or managerial capacity for effective environmental governance and ecological restoration. With urban systems in overall decline, lacking transformational momentum and institutional efficacy, the potential positive factors associated with urbanization fail to materialize into significant improvements in ecological resilience.

(2)

Heterogeneity Test: Geographical Location

Variations in resource endowment, economic structure, and socio-cultural factors across regions may lead to divergent pathways and effectiveness through which urbanization influences the ecological resilience of resource-based cities. Following the economic zoning standards established by the Statistical Data Management Center of China’s National Bureau of Statistics, the 114 resource-based cities in this study were categorized into four regions based on geographic location: western (37 cities), central (41 cities), eastern (17 cities), and northeastern (19 cities). The spatial distribution and classification are detailed in Figure 4 and

Table 9. Urban geographic location table.

Region	Cities
Western (37)	Baiyin, Baise, Baotou, Baoji, Baoshan, Chifeng, Dazhou, Ordos, Guang’an, Guangyuan, Hechi, Hezhou, Hulunbuir, Jinchang, Lijiang, Lincang, Liupanshui, Longnan, Luzhou, Nanchong, Panzhihua, Pingliang, Pu’er, Qingyang, Qujing, Shizuishan, Tongchuan, Weinan, Wuhai, Wuwei, Xianyang, Ya’an, Yan’an, Yulin, Zhangye, Zhaotong, Ziyang
Central (41)	Anshun, Bijie, Bozhou, Chenzhou, Chizhou, Chuzhou, Datong, Ezhou, Ganzhou, Handan, Hebi, Hengyang, Huaibei, Huainan, Huangshi, Jiaozuo, Jincheng, Jinzhong, Jingdezhen, Linfen, Loudi, Luoyang, Lvliang, Ma’anshan, Nanyang, Pingdingshan, Pingxiang, Puyang, Sanmenxia, Shaoyang, Shuozhou, Suzhou, Tongling, Xinzhou, Xinyu, Xingtai, Xuancheng, Yangquan, Yichun, Yuncheng, Zhangzhou
Eastern (17)	Chengde, Dongying, Huzhou, Jining, Linyi, Longyan, Nanping, Sanming, Shaoguan, Suqian, Tai’an, Tangshan, Xuzhou, Yunfu, Zaozhuang, Zhangjiakou, Zibo
Northeast (19)	Anshan, Baishan, Benxi, Daqing, Fushun, Fuxin, Hegang, Heihe, Huludao, Jixi, Jilin, Liaoyuan, Mudanjiang, Panjin, Shuangyashan, Songyuan, Tonghua, Yichun, Qitaihe

Note: The regional classifications in Figure 4 and Table 9 are sourced from the Statistical Data Management Center of China’s National Bureau of Statistics.

. Columns (1) to (4) in

Table 10. Results of heterogeneity tests for regions.

Variant	(1)	(2)	(3)	(4)
	Western	Central	Eastern	Northeast
URBAN	0.106 ***	0.187 ***	0.235 ***	0.0992 ***
	(0.019)	(0.039)	(0.055)	(0.024)
Controls	Yes	Yes	Yes	Yes
Controls-Squared	Yes	Yes	Yes	Yes
Time FE	Yes	Yes	Yes	Yes
City FE	Yes	Yes	Yes	Yes
N	518	574	238	266

Note: *** indicate significance levels of 1%. All variables are defined in detail in Supplementary Table S1.

present the effects of urbanization on ecological resilience across these regions. The regression coefficients for urbanization are significantly positive in all four regions, indicating that urbanization consistently enhances ecological resilience in resource-based cities regardless of location. However, the magnitude of this effect varies: it is strongest in the eastern region, followed by the central and western regions, and weakest in the northeastern region.

Resource-based cities in the eastern region benefit from well-developed economies and early achievements in green industrial upgrading, enabling high-intensity investments in capital and technology during urbanization. Combined with substantial ecological construction efforts and policy support, these cities exhibit systematic enhancement in ecological resilience, highlighting the effectiveness of their green development pathway. The central region, while experiencing urbanization driven by industrial relocation from the eastern region, faces pollution challenges that moderate its ecological resilience improvement. The western region, dominated by resource-dependent cities, focuses on resource exploitation during urbanization, with limited investment in environmental governance, thus constraining resilience growth. The northeastern region, characterized by resource depletion and a monolithic industrial structure, shows the weakest promotive effect of urbanization on ecological resilience. These disparities reflect differences in resource allocation, industrial structure, and environmental governance strategies across regions during urbanization, ultimately influencing the efficiency of ecological resilience enhancement.

5. Discussion

5.1. Discussion on Research Results

Based on the empirical analysis above, the findings of this study elucidate several key issues. The urbanization level exhibits a significant positive driving effect on the ecological resilience of resource-based cities, confirming the critical role of urbanization in enhancing ecological resilience. This result aligns with the perspective of Huber et al. [44]. Furthermore, this study reveals the mediating roles of green technological innovation, energy efficiency improvement, and industrial structure upgrading. Among these, green technological innovation demonstrates the strongest mediating effect, which is consistent with Chen et al.’s [76] view on “carbon unlocking.” Improved energy efficiency directly reduces environmental pressure in high-energy-consumption cities, a finding that complements the work of Wang et al. [77]. Although the direct contribution of industrial structure upgrading is relatively limited, its long-term role in green transformation should not be overlooked, supporting Huang et al.’s conclusions [78]. Heterogeneity analysis indicates that the positive effects are more pronounced in regenerative cities and resource-based cities in eastern regions, while declining cities show weaker effects due to institutional and technological constraints. This finding corroborates the “urban life cycle hypothesis” proposed by Cottineau et al. [79]. By systematically examining the impact mechanisms of urbanization on the ecological resilience of resource-based cities through three novel perspectives—green technological innovation, energy efficiency, and industrial structure upgrading—this study provides theoretical and empirical support for formulating differentiated ecological resilience policies.

5.2. Policy Suggestions

To further enhance the ecological resilience of resource-based cities and advance their long-term sustainable development, the following policy recommendations are proposed: (1) Promoting industrial structure advancement is essential. Regenerative cities should prioritize the development of high-tech industries and services, while declining cities need to establish ecological funds for mine site restoration and attract alternative industries such as new energy. All regions should encourage the formation of cross-regional industrial alliances to facilitate technology exchange and collaborative upgrading. (2) Establishing a green technology innovation system requires concerted efforts. At the national level, increased R&D investment and major specialized projects should be implemented. Locally, demonstration zones and joint R&D centers should be established, alongside raising the super-deduction ratio for R&D expenses to reduce innovation costs for enterprises. (3) Enhancing energy utilization efficiency demands regional strategies. Eastern areas should advance intelligent energy management systems, and western regions should develop new energy infrastructure, while northeastern and central areas should promote waste heat recovery and energy-saving technologies. Specialized loans and other mechanisms can further incentivize efficiency gains. (4) Implementing differentiated regional governance is crucial. Eastern cities should enforce stringent environmental standards to create demonstration zones, western cities should strengthen fiscal transfers to advance ecological restoration, central regions should accelerate technology diffusion, and northeastern areas should develop circular economies complemented by retraining programs. (5) Policy interventions should align with urban lifecycle stages. Regenerative cities can establish comprehensive green transition demonstration zones. Growing and mature cities must enforce strict environmental access controls and diversify their industrial base early. Declining cities should prioritize ecological rehabilitation and employment resettlement.

5.3. Limitations and Future Research Directions

While this study has yielded several findings, certain aspects warrant further investigation. First, nighttime light data have a limited capacity to capture the nuanced socioeconomic disparities within urban areas. Future research could refine urbanization metrics by integrating multi-source data. Second, constrained by data availability, subsequent studies could employ longer time-series data to capture the long-term evolutionary characteristics of ecological resilience and potential dynamic threshold effects. Third, this study did not fully disentangle the nonlinear interactions among mechanisms such as innovation agglomeration and industrial upgrading. More flexible modeling approaches could be adopted to elucidate the complex interplay of multiple mediating pathways. Furthermore, the empirical analysis did not explicitly incorporate resource dependence as a moderating variable. Future work should include relevant indicators to control for heterogeneity. Finally, expanding the research scope to examine urbanization’s impact on other sustainable development dimensions—such as economic resilience and resident well-being—would contribute to a more comprehensive assessment framework.

6. Conclusions

This study employs panel data from 114 resource-based cities spanning 2010 to 2023. A composite index derived from nighttime light data is used to measure urbanization levels, while urban ecological resilience is calculated using the entropy method. The double machine learning model is applied to examine the impact, mechanisms, and heterogeneity of urbanization on ecological resilience. The main findings are as follows: (1) The level of urbanization exerted a significant positive driving effect on the ecological resilience of resource-based cities during the study period, indicating that urbanization has become a key force in promoting ecological resilience in China’s resource-based cities. This conclusion demonstrates high universality and stability across multiple robustness tests, fully confirming the reliability of the findings. (2) Mechanism analysis reveals that urbanization enhances the ecological resilience of resource-based cities through three pathways: elevating the level of industrial structure, promoting green technology innovation, and improving energy-use efficiency, thereby providing strong support for sustainable urban development. (3) Heterogeneity analysis indicates that the positive effect of urbanization on ecological resilience is more pronounced in regenerative resource-based cities compared to growing, mature, and declining resource-based cities. Furthermore, the promoting effect of urbanization is more significant in resource-based cities located in the eastern region relative to those in the western, central, and northeastern regions.