Village-Level Spatio-Temporal Patterns and Key Drivers of Social-Ecological Vulnerability in a Resource-Exhausted Mining City: A Case Study of Xintai, China

Abstract

Evaluation of socio-ecological vulnerability is crucial for sustainable management in mining cities. This study selected Xintai City, China, as a case and constructed a comprehensive vulnerability assessment framework based on 2010–2020 multi-source data. By integrating the Geodetector, spatial autocorrelation analysis, and ordered weighted averaging (OWA), we systematically explored the spatio-temporal patterns and driving mechanisms of socio-ecological vulnerability. The Theil index at the village level revealed finer spatial heterogeneity than large-scale analyses. The results show the following: (1) Socio-ecological vulnerability in Xintai City is generally moderate, with high-vulnerability areas concentrated in the urban center and former coal mining zones. Over the past decade, high—vulnerability levels in these areas have improved, whereas the urban-rural fringe has experienced a significant increase in vulnerability, primarily driven by industrial transfer and uneven resource allocation. (2) Geodetector analysis indicated a shift in dominant drivers from natural to socio-economic factors, with population density and construction land proportion surpassing natural conditions such as average annual rainfall by 2020. Additionally, mining land proportion, land use change, and the spatial distribution of social services played key roles in shaping vulnerability patterns, while ecological and public service factors showed weaker explanatory power. (3) Scenario simulation based on OWA demonstrated that an economic-priority pathway leads to the outward expansion of vulnerable areas into mountainous regions, while an ecological-priority approach promotes spatial contraction and optimization of vulnerability zones. These findings provide scientific guidance for identifying key vulnerable areas and formulating differentiated management strategies, offering reference value for the sustainable development of resource-exhausted mining cities.

1. Introduction

With the ongoing progress of industrialization and urban expansion worldwide, cities dependent on natural resources are facing an escalating array of challenges. These include resource depletion, declining ecological services, and excessive reliance on single industries. The National Sustainable Development Plan for Resource-based Cities (2013–2020) highlights the transition of approximately 79.39% of China’s 262 resource-oriented cities into stages defined by maturity or decline, now grappling with severe risks such as ground subsidence, water pollution, and population loss. These issues are not unique to China; they represent global phenomena that impede balanced regional development and threaten long-term sustainability.

The concept of social-ecological vulnerability integrates human governance with natural resource utilization, revealing the dynamic interaction of ecological systems and socio-economic conditions while providing a standardized framework for risk assessment and mitigation [1,2]. Given the complexity, nonlinearity, uncertainty, and multi-layered interactions inherent to social-ecological systems—marked by multi-objective and multi-scale linkages—this perspective is essential for addressing ecological challenges and achieving sustainability goals [3,4]. Major theoretical frameworks in this field include the coupled human–environment systems (CHANS) vulnerability framework, urban metabolism theory, and complex adaptive systems theory [5].

Vulnerability remains a central topic under the conditions of worldwide environmental transformation and sustainable development pathways. Recent research has focused on sustainable livelihoods, disaster responses, and vulnerability in addressing the impacts of climate change, consistently underscoring the importance of acknowledging multiple risks in both policy and practice [6,7,8,9,10,11,12,13]. Studies investigating climate-related vulnerabilities, such as those arising from droughts, heatwaves, and floods, have identified social poverty, resource management efficiency, and government performance as critical driving factors. In resource-based regions, the impacts of socioeconomic development on ecosystems are increasingly evident, prompting investigations of social-ecological vulnerability in petroleum cities, coal mining areas, ecologically sensitive zones, karst landscapes, and rocky desertification regions. These studies collectively enrich the global case base for social-ecological vulnerability research.

First, while most previous studies assessed social-ecological vulnerability at broader spatial scales, including national, basin, provincial, municipal, county, and township levels [13,14,15,16,17,18,19,20,21,22,23], few have examined vulnerability at the village level, which is crucial for two reasons. Village-level analysis captures heterogeneity within townships and reveals high-risk areas often masked by “average effects” at larger scales. Processing village-level data allows precise representation of micro-level characteristics, such as local topography, land-use patterns, population density, and availability of social services. For example, differences in terrain, cultivated land structure, population distribution, and local infrastructure among villages can be accurately quantified. Moreover, this approach enables the identification of most vulnerable units, highlighting resource-poor villages or areas with inadequate public services, providing a scientific basis for targeted resource allocation and effective rural governance.

Second, although village-level vulnerability has received some attention, research on its dynamic evolution remains limited. Most existing studies rely on static assessments at a single time point, rarely exploring temporal changes within small-scale units [16,22,24,25,26,27,28]. Villages, as fundamental management units, are often overlooked in longitudinal analyses, and village-level temporal data are scarce. To address this gap, our study uses data from 2010 and 2020 to examine the spatio-temporal dynamics of social-ecological vulnerability over a decade. This approach provides quantitative and visual insights into improvement or deterioration trends in each village, identifies emerging risk points, and evaluates the effectiveness of policy interventions such as ecological restoration and infrastructure upgrades. By integrating fine-scale spatial and temporal perspectives, this study contributes methodological innovation and informs adaptive policy design in rural social-ecological systems.

Third, scenario simulation has gained attention in vulnerability research, but its application at the village level remains limited [29,30,31,32,33]. Existing studies often focus on static analysis or do not fully integrate uncertainty and decision-making preferences. For instance, Dasgupta et al. assessed adaptability and vulnerability in Indian Himalayan villages [32], while Chen et al. simulated changes in social-ecological vulnerability in Yulin City’s semi-arid region [30]. Our study introduces scenario simulation at the village level, enabling prediction of future vulnerability trends and spatial patterns under multiple external drivers. This method allows identification of high-risk villages, quantification of key drivers, and provides a scientific foundation for proactive governance, risk early warning, and sustainable management of village-level social-ecological systems.

Take Xintai as a case study, this research integrates a vulnerability analysis framework to create an index system. We explored the dynamics evolution patterns and spatio-temporal heterogeneity of social-ecological vulnerability in Xintai during the resource transformation period from 2010 to 2020. The objective is to generate in-depth insights that can contribute to the sustainable transformation of resource-dependent regions worldwide. This research aims to achieve the following specific goals: ① Assess the spatial differentiation and temporal trends of social-ecological vulnerability in Xintai from 2010 to 2020. ② Examine the determinants influencing social-ecological vulnerability. ③ Simulate the evolutionary paths of social-ecological vulnerability and differentiated regulatory schemes under various policy scenarios.

2. Materials and Methods

2.1. Study Area

Xintai, an administrative unit at the county level within Shandong Province, lies in the southeast of Tai’an and is administered by it. It serves as a major coal-based city in East China, with its economic development heavily reliant on coal resource exploitation. The mining industry has long been a pillar of the local economy, exhibiting significant characteristics of a mining-dominated economy.

In 2011, Xintai was designated as one of the third group of pilot areas for the transformation of resource-depleted cities. The city is home to cumulative proven coal reserves estimated at approximately 1.87 billion tons. However, extensive mining activities have resulted in a subsidence zone that spans 129 km², accounting for 7.22% of Xintai’s total land area. This situation has exacerbated significant conflicts regarding the relocation of coal villages and has given rise to complex issues such as farmland degradation, surface cracking, and ecological deterioration, all contributing to persistent socio-ecological vulnerability.

Xintai covers approximately 1787 km² and has a population of 1.23 million, with an urbanization rate of 62.14% in 2023. A comparison of the second and third surveys from 2011 shows that mining land decreased from 22.46 km² to 10.44 km², while construction land increased from 311.36 km² to 319.18 km². This indicates that Xintai City is undergoing a transition in its social-ecological system., highlighting the importance of studying resource-based city transformations through the lens of social-ecological integration (Figure 1).

2.2. Data Source

This study utilized multi-source datasets covering five major categories: remote sensing, meteorological, soil, land use, and socio-economic data. To ensure consistency across analyses, all raster datasets were resampled to a uniform spatial resolution of 30 m using the Resample tool in ArcGIS 10.2. A summary of the datasets, including sources, key parameters, and resolutions, is provided in

Table 1. Data Sources.

Data Type		Data Source	Key Parameters/Description
Remote sensing data	DEM	Geospatial Data Cloud (http://www.gscloud.cn/, accessed on 18 March 2024)	Elevation, 30 m resolution
	NDVI	NASA LAADS DAAC (https://ladsweb.modaps.eosdis.nasa.gov/, accessed on 18 March 2024)	MOD13Q1 product (NASA LP DAAC, Sioux Falls, SD, USA), 250 m resolution
Meteorological data	Precipitation/Temperature	National Earth System Science Data Center (https://www.geodata.cn/, accessed on 18 March 2024)	1 km resolution annual precipitation and mean annual temperature data
Soil data	Soil properties	Resource and Environmental Sciences Data Center of the Chinese Academy of Sciences (https://www.resdc.cn/, accessed on 19 March 2024)	Soil texture (sand, silt, clay content); soil erosion intensity (light, moderate, severe)
Land use data	Land use data	Resource and Environmental Sciences Data Center of the Chinese Academy of Sciences (https://www.resdc.cn/, accessed on 20 March 2024)	30 m resolution Land use types include built-up (construction) land, grassland, cropland, forestland, water area, unused land
	Land survey data	Ministry of Natural Resources	Extracts change information for forest, grassland, garden land, and mining land areas
Socio-economic data	Nighttime light	Resource and Environmental Sciences Data Center of the Chinese Academy of Sciences (https://www.resdc.cn/, accessed on 20 March 2024)	DMSP/OLS (2010) (NOAA, Washington, DC, USA), NPP/VIIRS (2020) (NASA EOSDIS, Greenbelt, MD, USA), 500 m resolution
	GDP/Population density data	Resource and Environmental Sciences Data Center of the Chinese Academy of Sciences (https://www.resdc.cn/, accessed on 20 March 2024)	GDP and population density, 1 km resolution
	Statistical Yearbooks	Tai’an Municipal Statistics Bureau	GDP, industrial structure, urbanization rate
	POI data	Amap API (2012; 2020)	Locations of industrial enterprises and public services
	Census Data	Xintai Municipal Bureau of Statistics	Administrative village population density, age structure

for clarity and reproducibility.

2.3. Framework

Social-ecological vulnerability integrates human society and ecosystems as a complex system [34]. The Vulnerability Scoping Diagram (VSD) offers a standardized framework for evaluating this vulnerability by focusing on key dimensions of exposure, sensitivity, and adaptive capacity to inform mitigation and adaptation strategies for global change [1,2]. Social-ecological systems are characterized by complexity, nonlinearity, uncertainty, and multi-objective interactions across various scales. Addressing these issues is crucial for achieving sustainable development [3,4,5]. The framework comprises four key steps (Figure 2). First, a comprehensive spatiotemporal database is created by integrating geospatial and socioeconomic data. Second, guided by the VSD theoretical framework, we identify 19 evaluation indicators to construct the social-ecological vulnerability evaluation system. Third, we analyze patterns and types of social-ecological vulnerability in Xintai. Finally, we use the GeoDetector method to explore driving mechanisms of social-ecological vulnerability and apply the Ordered Weighted Averaging (OWA) method to simulate spatial evolution trends based on decision-making preferences.

2.4. Study Methods

2.4.1. Establishing an Assessment Index System

The evaluation of social-ecological vulnerability relies on the careful selection of indicators that accurately represent the system’s current state and respond to changes. Currently, there is no universally accepted standard for the number or types of parameters needed to effectively capture social-ecological vulnerability attributes. This study builds on previous research [15,35,36] and considers regional characteristics to determine the drivers of vulnerability within the study region. The indicator system developed in this research consists of three categories: exposure, sensitivity, and adaptive capacity. Altogether, 19 indicators were chosen to form the social-ecological vulnerability index (

Table 2. Assessment indicator system of Social-Ecological Vulnerability in Xintai.

Criterion Layer	Element Layer	Indicator Layer	Indicator Properties	2010 Weight	2020 Weight
Exposure	Natural systems	Average elevation	+	0.0740	0.0731
		Average slope	+	0.1705	0.1683
		Terrain relief	+	0.1290	0.1274
		Annual precipitation	+	0.0428	0.0483
		Annual temperature	+	0.0158	0.0138
		Soil texture	−	0.2543	0.1542
		Soil erosion intensity	+	0.1293	0.2510
	Human Activities	Proportion of construction land area	+	0.0280	0.1336
		Population density	+	0.1563	0.0304
Sensitivity	Land Use	Proportion of mining land area	+	0.5715	0.6834
		Proportion of river area	+	0.3641	0.2433
		Proportion of cultivated land area	+	0.0300	0.0359
	Ecological	NDVI	−	0.0234	0.0219
		Proportion of forest, grassland, and garden land area	−	0.0110	0.0155
Adaptive Capacity	Economic	Kindergarten density	−	0.0184	0.0322
		Primary and secondary school density	−	0.0190	0.0390
		Hospital density	−	0.0201	0.0351
	Social	GDP	−	0.3477	0.8085
		Nighttime light intensity	−	0.5948	0.0851

Note: 1. ‘+’ represents a positive indicator (higher values indicate higher vulnerability), and ‘−’ represents a negative indicator (higher values indicate lower vulnerability). 2. Weights were calculated using the entropy method, and several indicators showed substantial changes between 2010 and 2020. These changes reflect both the sensitivity of the entropy method to variations in indicator values and spatial distribution, as well as actual socio-economic and ecological changes in Xintai City. Specifically, rapid economic growth increased differences in GDP across townships, resulting in a higher weight for GDP. The relative importance of nighttime light intensity decreased as urban lighting patterns stabilized and spatial variation diminished. The weight of construction land proportion increased due to urban expansion and the resulting greater land use variation. Population density weight decreased because residents concentrated in central urban areas, reducing spatial variation at the county scale. In contrast, intensified mining activities, legacy impacts of abandoned mines, and agricultural practices increased spatial differences in soil erosion, leading to a higher weight for soil erosion intensity.

). The reasons for their selection are provided below:

Exposure indicators assess the intensity of stress on the system from natural disturbances and human activities, incorporating both ecological conditions and societal factors. The nine selected indicators are average elevation, average slope, terrain relief, annual precipitation, annual temperature, soil texture, soil erosion intensity, proportion of construction land area, and population density. These indicators have been validated for their correlation with exposure.

Sensitivity indicators evaluate the system’s stability and its likelihood of destabilization due to disturbances, focusing on land use and ecological structure. The five indicators include the proportion of mining land, river area, cultivated land, the Normalized Difference Vegetation Index (NDVI), and the proportion of forest, grassland, and garden land. The increase in mining land raises the risk of ecosystem degradation, resulting in higher sensitivity. Conversely, changes in NDVI reflect ecosystem stability and resilience, higher NDVI values indicate lower sensitivity.

Adaptive capacity indicators describe the capability of the system to withstand and adapt to outside disruptions, The five indicators include kindergarten density, primary and secondary school density, hospital density, GDP, and nighttime light intensity. Nighttime lights indicate economic activity; higher economic activity correlates with greater adaptability. The density of hospitals, schools, and kindergartens reflects the spatial concentration of public service facilities, higher density suggests weaker capacity to respond. All indicators align with village-level boundaries for accurate assessments at the village level.

2.4.2. Entropy Value Method

The range standardization was applied to reduce the impact arising from differences in indicator dimensions within the raw indicator data, which can positively or negatively impact the social-ecological vulnerability. Formulas (1) and (2) present the standardization methods applied to positive and negative indicators [37,38,39].

$${\mathrm{P}\mathrm{o}\mathrm{s}\mathrm{i}\mathrm{t}\mathrm{i}\mathrm{v}\mathrm{e} \mathrm{i}\mathrm{n}\mathrm{d}\mathrm{i}\mathrm{c}\mathrm{a}\mathrm{t}\mathrm{o}\mathrm{r}\mathrm{s}: Y}_{ij}={\displaystyle \frac{\left(X_{ij}-X_{jmin}\right)}{\left(X_{jmax}-X_{jmin}\right)}}$$

(1)

$$\mathrm{N}\mathrm{e}\mathrm{g}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{v}\mathrm{e} \mathrm{i}\mathrm{n}\mathrm{d}\mathrm{i}\mathrm{c}\mathrm{a}\mathrm{t}\mathrm{o}\mathrm{r}\mathrm{s}: Y_{ij}={\displaystyle \frac{\left(X_{jmax}-X_{ij}\right)}{\left(X_{jmax}-X_{jmin}\right)}}$$

(2)

where ${\mathrm{Y}}_{\mathrm{i}\mathrm{j}}$ represents the normalized value located at row $i$, column $j$; $X_{ij}$ is the raw data in row $i$, column $j$; and $X_{jmax}$, $X_{jmin}$ indicate the maximum and minimum values in column $j$, respectively.

The entropy value method was applied to determine indicator weights. The detailed calculation procedure is provided below [40].

• Based on the normalized value $Y_{ij}$, compute the proportion $X_{ij}^{\prime }$ for each unit $\mathrm{i}$ under index $j$:

  $$X_{ij}^{\prime }={\displaystyle \frac{Y_{ij}}{\sum _{i=1}^m{Y}_{ij}}}\left(i=\mathrm{1,2},3,\dots ,n; j=\mathrm{1,2},3,\dots ,m\right)$$

  (3)
• Calculate the entropy ($e_j$):

  $$e_j=-k{\displaystyle {\sum }_{i=1}^m}\left(X_{ij}^{\prime }\times ln{X}_{ij}^{\prime }\right), k={\displaystyle \frac{1}{lnm}},0\le e_j\le 1$$

  (4)
• Calculate the difference coefficient ($d_j$):

  $$d_j=1-e_j$$

  (5)
• Normalize the divergence coefficient ($d_j$) to obtain the final weight ($W_j$):

  $$W_j={\displaystyle \frac{d_j}{{\displaystyle {\sum }_{j=1}^n}{d}_j}}$$

  (6)

  where $m$ is the number of research units, and $n$ is the number of indicators.

2.4.3. Vulnerability Functional Model

The IPCC has provided a more widely accepted definition of vulnerability, which is generally regarded as a function of exposure, sensitivity, and adaptive capacity [41]. In this study, the vulnerability function model integrates the VSD framework to quantitatively express vulnerability as the result of the interaction between these three dimensions [1]. The formula below is employed to compute the social-ecological vulnerability index [26,35].

$$V=E+S-A$$

(7)

where $V$ denotes the overall vulnerability, $E$ represents exposure, $S$ denotes sensitivity, and $A$ is adaptive capacity. To categorize the calculated values of $V$, $E$, $S$, and $A$, this research applies the Jenks Natural Breaks in ArcGIS 10.2. The results are classified into five distinct levels: low, moderately low, moderate, moderately high, and high.

2.4.4. Spatial Autocorrelation Analysis

Spatial autocorrelation analysis is utilized to assess whether spatial units exhibit interdependence or similarity based on their geographic proximity. In general, spatial autocorrelation can be classified as either global or local. The global spatial autocorrelation captures the overall spatial association or spatial agglomeration across the entire study area, and is commonly evaluated using the Moran’s I index. Local spatial autocorrelation assesses the spatial correlation between individual spatial units and their neighboring units within the study area, revealing the distribution characteristics of spatial heterogeneity in local regions, often characterized by local Moran’s I. Based on the local Moran’s I index, spatial units are classified into five categories: high and high clusters, low and low clusters, high and low outliers, low and high outliers, and areas with no significant spatial association.

To quantify this spatial correlation, this study employs Moran’s I index and global Moran’s I index to describe the spatial correlation and distribution differences in the social-ecological vulnerability in 2010 and 2020. The computation follows the formulas below [11,37]:

$$Mora{n}^{\prime }s I={\displaystyle \frac{n\sum _i\sum _j{W}_{ij}\left(x_i-\overline{x}\right)\left(x_j-\overline{x}\right)}{\left(\sum _i\sum _j{W}_{ij}\right)\sum _i{\left(x_j-\overline{x}\right)}^2}}$$

(8)

$$Local Mora{n}^{\prime }s I={\displaystyle \frac{n\left(x_i-\overline{x}\right){\sum }_j{W}_{ij}\left(x_j-\overline{x}\right)}{{\sum }_i{\left(x_i-\overline{x}\right)}^2}}$$

(9)

where $W_{ij}$ denotes spatial weight matrix. $x_i$ and $x_j$ represent the vulnerability index values for samples $i$ and $j$, respectively. n is the total number of spatial units. $\overline{x}$ is the mean vulnerability value. A positive Moran’s I (>0) indicates spatial clustering of similar values (positive spatial correlation); a negative Moran’s I (<0) suggests dispersion (negative spatial correlation) and Moran’s I equal to 0 suggests a random spatial pattern with no significant spatial dependence.

2.4.5. Geographic Detector

As a spatial statistical technique, GeoDetector is used to quantify spatial heterogeneity and reveal key drivers behind geographic processes. We employ factor and interaction detector to measure the driving mechanisms behind the social-ecological vulnerability, focusing on how each factor contributes to observed vulnerability patterns. The factor detection process is represented by the following formula [19,36,42]:

$$\mathrm{q}=1-{\displaystyle \frac{1}{N{\sigma }^2}}\sum _{h=1}^L{N}_h{{\sigma }_h}^2$$

(10)

where $q$ denotes how much a factor explains vulnerability, ranging from 0 to 1, where larger values indicate a greater influence on social-ecological vulnerability. $L$ refers to the number of categories of the influencing factor. $N_h$ and $N$ denote the unit counts for each zone and for the study area as a whole. ${\sigma }_h^2$ and ${\sigma }^2$ are the variances within each zone and the study area as a whole, respectively.

To evaluate the combined effects of two influencing factors, the interaction detector was used to determine whether their interaction strengthens or diminishes their separate contributions to social-ecological vulnerability. The interpretation categories of interaction outcomes are summarized in

Table 3. Interaction types.

Criteria	Interaction Type
$q\left(X_1\cap X_2\right)<Min\left[q\left(X_1\right),q\left(X_2\right)\right]$	Nonlinear attenuation
$Min\left[q\left(X_1\right),q\left(X_2\right)\right]<q\left(X_1\cap X_2\right)<Max\left[q\left(X_1\right),q\left(X_2\right)\right]$	Single-factor nonlinear attenuation
$q\left(X_1\cap X_2\right)>Max\left[q\left(X_1\right),q\left(X_2\right)\right]$	Bifactorial enhancement
$q\left(X_1\cap X_2\right)=q\left(X_1\right)+q\left(X_2\right)$	Mutual independence
$q\left(X_1\cap X_2\right)>q\left(X_1\right)+q\left(X_2\right)$	Nonlinear enhancement

[43].

2.4.6. Ordered Weighted Averaging (OWA)

Ordered Weighted Averaging (OWA) operators integrate criterion weights with order weights of spatial indicators, enabling the simulation of evaluation outcomes under varying decision-maker preferences. We use the OWA to simulates vulnerability evolution trends. The formula is as follows [36,44,45,46]:

$${OWA}_i=\sum _{j=1}^n\left({\displaystyle \frac{u_j{v}_j}{{\sum }_{i=1}^n{u}_j{v}_j}}\right)Z_{ij}$$

(11)

where $Z_{ij}$ represents the standardized score of indicators $j$ for unit $i$, $u_j$ and $v_j$ represents is the criterion weight and order weight, respectively. In this study, the entropy method was applied to determine the criterion weights. The formula for order weight is as follows:

$$v_j={\left({\displaystyle {\sum }_{k=1}^j}{w}_k\right)}^a-{\left({\displaystyle {\sum }_{k=1}^{j-1}}{w}_k\right)}^a, a\in (0,\infty )$$

(12)

where $a$ is the risk coefficient. When $a<1$, it reflects optimistic decision-making, indicating that regional vulnerability risks are within acceptable or controllable thresholds and will not compromise the stable development of the social-ecological system; when $a=1$, the model adopts equal-probability decision-making, assuming current vulnerability trends persist with order weights assigned equally; when $a>1$, it represents pessimistic decision-making, highlighting scenarios where heightened exposure or sensitivity risks significantly escalate vulnerability, thereby undermining the sustainable development of the social-ecological system.

$W_k$ denotes the weight assigned to the index and is determined by the formula below:

$$w_k={\displaystyle \frac{n-r_k+1}{{\displaystyle {\sum }_{I=1}^k}\left(n-r_k+1\right)}}(k=\mathrm{1,2},3,\dots ,n)$$

(13)

where $n$ represents the total count of indicators. $r_k$ indicates the rank of each indicator according to its importance. The most important indicator is ranked 1, the second most important is 2, and so on, with the least important ranked $n$.

2.4.7. Theil Index

The Theil index is used to measure the degree of vulnerability imbalance at the town scale. Based on the 20 towns in Xintai City, the Theil index of the vulnerability index is decomposed into inter-town differences ($T_b$) and intra-town differences ($T_w$). The intra-town difference reflects the vulnerability differences among villages within the same town. The larger the intra-town difference $T_w$, the greater the vulnerability differences among villages within the same town. The specific calculation formula is as follows:

$$T=T_b+T_w=\sum _{i=1}^n\left({\displaystyle \frac{V_i}{V_{total}}}\right)ln\left({\displaystyle \frac{V_i/V_{total}}{P_i/P_{total}}}\right)$$

(14)

$$T_b=\sum _{k=1}^K\left({\displaystyle \frac{V_k}{V_{total}}}\right)ln\left({\displaystyle \frac{V_k/V_{total}}{P_k/P_{total}}}\right)$$

(15)

$$T_w=\sum _{k=1}^k\left({\displaystyle \frac{V_k}{V_{total}}}\right)T_k$$

(16)

$$T_k=\sum _{i\in k}\left({\displaystyle \frac{V_i}{V_k}}\right)ln\left({\displaystyle \frac{V_i/V_k}{P_i/P_k}}\right)$$

(17)

where $T$ represents the overall Theil index; $T_b$ represents the inter-town disparity; $T_w$ represents the intra-town disparity; $n$ represents the total count of administrative villages; $k$ indicates the total count of towns/streets; $P_i$ is the population of the $i$ administrative village; $P_k$ is the total population of the $k$ town; $P_{total}$ is the total population of Xintai City; $V_i$ is the social-ecological vulnerability index of the $i$ village; $V_k$ is the total vulnerability of the $k$ town; and $V_{total}$ is the total vulnerability of Xintai City.

3. Results

3.1. Spatio-Temporal Patterns of Social-Ecological Vulnerability

3.1.1. Spatio-Temporal Patterns

From 2010 to 2020, exposure maintained a consistent level, with its spatial differentiation pattern closely aligned with the topographic conditions of Xintai (Figure 3). The mean exposure level increased from 0.2761 to 0.2792. High and moderately high exposure areas were mainly distributed across the northern Lianhua Mountain-Phoenix Mountain region and the southern hilly area of Xintai City (such as Shilai Town and Wennan Town) (Figure 4). These areas were constrained by natural factors such as significant topographic fluctuations and high soil erosion intensity. In contrast, low exposure areas were stably distributed in the northwestern plain region (such as Xizhangzhuang Town, the southern part of Yangliu Town, and Liudu Town), maintaining lower exposure levels due to their flat terrain, fertile soil, and minimal human disturbance. Exposure levels were primarily influenced by natural spatial patterns, making it difficult to alter this pattern in the short term.

Sensitivity exhibited a declining trend, with the average sensitivity decreasing from 0.0867 to 0.0688. The proportion of highly sensitive areas significantly reduced from 1.92% in 2010 to 0.48% (

Table 4. The number and proportion of different social-ecological vulnerability levels in 2010 and 2020.

		Exposure		Sensitivity		Adaptive Capacity		Vulnerability
		2010	2020	2010	2020	2010	2020	2010	2020
Low	Number	158	173	512	346	82	78	165	143
	Proportion	18.92%	20.72%	61.32%	41.44%	9.82%	9.34%	19.76%	17.13%
Moderately low	Number	239	229	196	382	438	152	269	233
	Proportion	28.62%	27.43%	23.47%	45.75%	52.46%	18.20%	32.22%	27.90%
Medium	Number	180	173	78	71	180	507	220	259
	Proportion	21.56%	20.72%	9.34%	8.50%	21.56%	60.72%	26.35%	31.02%
Moderately high	Number	191	191	33	32	79	59	126	150
	Proportion	22.87%	22.87%	3.95%	3.83%	9.46%	7.07%	15.09%	17.96%
High	Number	67	69	16	4	56	39	55	50
	Proportion	8.02%	8.26%	1.92%	0.48%	6.71%	4.67%	6.59%	5.99%

), mainly concentrated in industrial and mining areas (such as Xiaoxie Town and Dongdu Town) and water source protection zones (such as the Chaiwen River Basin and reservoirs) (Figure 4). Through the restructuring and upgrading of the coal industry, ecological restoration, and environmental management measures, a comparison of data from the second and third surveys revealed that the mining land area in Xintai City decreased from 22.46 km² to 10.44 km², significantly reducing sensitivity in these regions. However, the proportion of low-sensitivity areas increased from 23.47% in 2010 to 45.75% in 2020, mainly concentrated in the southern region of Xintai. This region has experienced noticeable population outflow, leading to an increasingly prominent trend of afforestation on cultivated lands, with the area of farmland decreasing from 964.18 km² in 2010 to 776.52 km² in 2020.

The proportion of areas with median adaptability increased from 21.56% in 2010 to 60.72% (Table 4), while low adaptability areas decreased from 52.46% to 18.2%, indicating an overall improvement in adaptability. High adaptability is concentrated in county towns (e.g., Qingyun Street and Xinfu Street) (Figure 4), supported by quality public services and economic agglomeration effects. Relatively high adaptability regions are near industrial and mining areas (e.g., Zhai Town and Xiaoxie Town), whereas medium, relatively low, and low adaptability regions located in mountain district. Nighttime lights and GDP serve as key indicators of adaptability, reflecting the positive impact of national policies on industrial transformation and the equalization of public service facilities in recent years.

Between 2010 and 2020, the proportion of areas with high socio-ecological vulnerability decreased from 6.59% to 5.99%, while those with relatively high vulnerability increased from 15.09% to 17.96% (Table 4). Conversely, low vulnerability areas dropped from 19.76% to 17.13%. The most vulnerable areas are primarily located in county towns, such as Xinwen Street and Qingyun Street (Figure 4), with Xinwen having once been a major coal-producing area in Xintai City. Recent transformation efforts in resource-based cities, including closures and the implementation of smart technologies, have helped reduce vulnerability in these high-risk regions. The western region, known as Xiyangguo (which includes Xizhangzhuang Town, Yangliu Town, and Guodu Town), is heavily industrialized, featuring cranes and hoisting devices that increase its vulnerability. In the southern part of the area (including Liudu Town, Yuejiazhuang Town, Shilai Town, and Fangcheng Town), the combination of low mountainous terrain and agricultural zones has led to significant population outflow, resulting in trends of afforestation and farmland abandonment, which further increases their vulnerability.

3.1.2. Change Trends of Social-Ecological Vulnerability

The graded assessment of social-ecological vulnerability changes was conducted based on the difference method (Figure 5), with the magnitude of change categorized into six levels: values ≤ −0.2000 were classified as significant decrease, −0.2000 to −0.1000 as stable decrease, −0.1000 to 0 as slight decrease, 0 to 0.1000 as slight increase, 0.1000 to 0.2000 as stable increase, and ≥0.2000 as significant increase.

The results show that from 2010 to 2020, social-ecological vulnerability in the peripheral areas of Xintai increased due to arable land prevalence, low hills, farmland afforestation, and population outflow. In contrast, vulnerability in the central urban area decreased as several coal mining enterprises closed and ecological restoration efforts were made in existing mines. However, rapid heavy industry expansion in the “Xiyangguo” sub-center has encroached on ecological space, leading to an overall increase in vulnerability.

For the period 2010–2020, approximately 917.41 km² of the study area experienced variations in socio-ecological vulnerability levels, representing 51.33% of the region’s total area. Among these variations, 655.92 km² (36.70%) experienced an upward shift in vulnerability levels, while 261.49 km² (14.63%) saw a downward shift. The area that remained in the same vulnerability level totaled 869.74 km² (48.67%) (

Table 5. Transition matrices of each vulnerability level from 2010 to 2020.

2020	Low		Moderately Low		Medium		Moderately High		High
2010	Area (km2)	Proportion (%)	Area (km2)	Proportion (%)	Area (km2)	Proportion (%)	Area (km2)	Proportion (%)	Area (km2)	Proportion (%)
low	152.48	8.53	113.15	6.33	20.36	1.14	0.52	0.03	0.00	0.00
moderately low	43.18	2.42	237.82	13.31	240.38	13.45	35.99	2.01	1.20	0.07
medium	9.24	0.52	35.62	1.99	252.52	14.13	207.16	11.59	9.54	0.53
moderately high	2.11	0.12	43.30	2.42	71.66	4.01	144.56	8.09	27.63	1.55
high	0.39	0.02	1.81	0.10	24.07	1.35	30.11	1.68	82.36	4.61

).

The transitions among the different socio-ecological vulnerability levels primarily occurred between adjacent classes (Figure 6), with notable transfer types being “moderate → higher,” “lower → moderate,” and “low → lower.” Specifically, the area transitioning from moderate to high vulnerability was 207.16 km² (11.59%), while 240.38 km² (13.45%) shifted from lower to moderate vulnerability, and 113.15 km² (6.33%) went from low to lower vulnerability. This indicates a trend of vulnerability centers moving towards moderate risk. The proportion of high vulnerability areas that downgraded was only 3.15%, suggesting that ecological restoration efforts are predominantly located within industrial and mining areas.

3.1.3. Spatial Aggregation of Social-Ecological Vulnerability

The social-ecological vulnerability in Xintai shows significant spatial dependence. The global Moran’s I index declined from 0.7798 to 0.4645 between 2010 and 2020, showing a statistically significant (p < 0.01) positive spatial autocorrelation in system vulnerability, accompanied by a weakening of clustering intensity. LISA cluster analysis demonstrates that vulnerability exhibits spatial clustering, primarily as high-high and low-low groupings (Figure 7). In 2010, High-high clusters were mainly situated in the urban center and mining zones, significantly overlapping with high vulnerability areas influenced by variables including land use, the proportion of mining land, and population density. Conversely, L-L clusters were found in the northwestern plains, which had lower vulnerability. By 2020, H-H clusters had contracted and become more dispersed due to coal mine closures and industrial upgrades; however, they remained in the southern parts of the county, as well as in eastern Xiexie Town and Dongdu Town. Compared to 2010, L-L clusters dramatically decreased by 2020, while H-L and L-H clusters became more sporadic.

3.1.4. Inequality of Social-Ecological Vulnerability

The Theil index was used to measure intra-town differences in social-ecological vulnerability in Xintai City from 2010 to 2020 (

Table 6. The Theil Index and Its Decomposition of Vulnerability Regional Disparities among Townships in Xintai City.

Town/Street	2010 Tw	2020 Tw
Theil index	0.3639	0.2777
Between Towns/Streets (contribution)	0.0676 (18.58%)	0.0264 (9.51%)
Within Towns/Streets (contribution)	0.2963 (81.42%)	0.2513 (90.49%)
Qingyun Street	0.0729	0.0516
Xinwen Street	0.0118	0.0047
Xinfu Street	0.0226	0.0123
Dongdu Town	0.0082	0.0093
Xiaoxie Town	0.0050	0.0049
Zhai Town	0.0169	0.0094
Quangou Town	0.0116	0.0127
Yangliu Town	0.0167	0.0187
Guodu Town	0.0054	0.0045
Xizhangzhuang Town	0.0033	0.0050
Loude Town	0.0040	0.0042
Yucun Town	0.0089	0.0088
Gongli Town	0.0069	0.0054
Guli Town	0.0202	0.0181
Shila Town	0.0148	0.0158
Fangcheng Town	0.0042	0.0048
Liudu Town	0.0086	0.0078
Wen’nan Town	0.0356	0.0385
Longting Town	0.0164	0.0126
Yuejiazhuang Town	0.0021	0.0024

). The Theil index values for social-ecological vulnerability were 0.3639 in 2010 and 0.2777 in 2020, indicating a gradual decrease over the past decade and a more balanced spatial distribution of vulnerability, which can be partly attributed to significant changes in the industrial structure. Xintai City was divided into 20 towns, and the Theil index was analyzed for both periods. In 2020, there were still notable imbalances among towns. For instance, Qingyun Street, Wen’nan Town, and Guli Town had relatively high ${ T}_w$ values of 0.0516, 0.0385, and 0.0181, respectively, indicating pronounced internal vulnerability. Conversely, Xinwen Street, Guodu Town, and Yuejiazhuang Town showed lower ${ T}_w$ values of 0.0047, 0.0045, and 0.0024, suggesting more homogeneity within these towns. Overall, Vulnerability was spatially distributed following a pattern of “high–low differentiation”, with considerable intra-town disparities persisting in specific areas.

A temporal comparison of $T_w$ values between 2010 and 2020 revealed that most towns experienced a downward trend in internal differences, Qingyun Street decreased from 0.0729 to 0.0516, and Xinfu from 0.0226 to 0.0123, implying increased internal balance. However, some towns, such as Wen’nan and Yangliu, saw slight increases in their ${ T}_w$ values, reflecting growing internal disparities. This divergence suggests that, although the overall vulnerability across Xintai City is converging, imbalances in development levels and resource allocation remain at the intra-town scale.

Further decomposition of the Theil index reveals deeper structural dynamics. In 2010, the contribution of between-town disparities to overall inequality was relatively high at 18.58%, but intra-town differences remained the dominant contributor at 81.42%. Notably, the internal disparity within Qingyun Street alone (0.0729) exceeded the total disparity between all towns (0.0676), underscoring that county-level inequality is primarily driven by disparities within towns. By 2020, the contribution of inter-town disparities had nearly halved, falling to 9.51%, while intra-town disparities increased to 90.49%, suggesting a growing importance of village-level heterogeneity. Although the between-town component dropped significantly from 0.0676 to 0.0264, the intra-town component only slightly declined from 0.2963 to 0.2513. This pattern implies that assessments conducted solely at the town level may substantially underestimate the long-term persistence of internal inequality and risk misleading policymakers into unwarranted optimism. Consequently, incorporating fine-scale, village-level analysis is essential to fully capture the evolving patterns of social-ecological vulnerability and to support more targeted and equitable policy interventions aimed at sustainable regional development.

3.2. Analysis of Driving Mechanisms for Rural Social-Ecological Vulnerability

With the social-ecological vulnerability index serving as the dependent variable and 19 assessment indicators as independent variables, we categorized the data with the natural breakpoint method. We applied Geodector for driving force analysis.

3.2.1. Factor Detection

The factor analysis results show that the contributions of key factors driving social-ecological vulnerability in Xintai City change over time (

Table 7. Importance of variables in explaining social-ecological vulnerability.

Indicator Factor	2010			2020
	q-Value	p-Value	q Rank	q-Value	p-Value	q Rank
Average elevation (X1)	0.1162	0.000	5	0.1694	0.000	6
Average slope (X2)	0.0534	0.000	12	0.1114	0.000	7
Terrain relief (X3)	0.0647	0.000	10	0.1002	0.000	9
Annual precipitation (X4)	0.2572	0.000	1	0.2313	0.000	3
Annual temperature (X5)	0.0659	0.000	9	0.0936	0.000	10
Soil texture (X6)	0.0981	0.000	7	0.2137	0.000	4
Soil erosion intensity (X7)	0.0144	0.058	16	0.1008	0.000	8
Proportion of construction land area (X8)	0.2427	0.000	2	0.2745	0.000	2
Population density (X9)	0.1781	0.000	3	0.3334	0.000	1
Proportion of mining land area (X10)	0.1439	0.000	4	0.1803	0.000	5
Proportion of river area (X11)	0.0119	0.364	17	0.0243	0.179	17
NDVI (X12)	0.0106	0.098	18	0.0068	0.273	18
Proportion of cultivated land area (X13)	0.1037	0.000	6	0.0351	0.000	16
Proportion of forest, grassland, and garden land area (X14)	0.0066	0.274	19	0.0064	0.327	19
Kindergarten density (X15)	0.0460	0.005	13	0.0607	0.000	14
Primary and secondary school density (X16)	0.0297	0.054	15	0.0385	0.031	15
Hospital density (X17)	0.0431	0.014	14	0.0622	0.000	13
GDP (X18)	0.0605	0.000	11	0.0851	0.000	11
Nighttime light intensity (X19)	0.0894	0.000	8	0.0782	0.000	12

Note: p-values less than 0.05 are considered statistically significant.

). In 2010, the top five factors were: annual average rainfall (0.2572), proportion of construction land area (0.2427), population density (0.1781), proportion of mining land (0.1439), and average elevation (0.1162). These were mainly related to the natural environment. For instance, rainfall and elevation directly determined ecological carrying capacity, influencing agricultural productivity and the stability of water and soil systems. The proportion of mining land reflected the heavy dependence on coal resource extraction, which increased ecological disturbance and environmental risks. Meanwhile, limited construction land and rising population density in mining towns created additional social stress.

By 2020, population density (0.3334) became the most significant factor, followed by proportion of construction land area (0.2745), annual average rainfall (0.2313), soil texture (0.2137), and mining land proportion (0.1803). This shift indicates a growing influence of anthropogenic factors due to urbanization, which has notably affected vulnerability through increased population concentration and expanded built-up areas. Meanwhile, rainfall and soil texture continued to influence ecological resilience.

3.2.2. Interaction Detection

The two-factor interaction analysis indicates that the combined effect of any two variables surpasses that of each individual variable (Figure 8). The spatial differentiation of social-ecological vulnerability in Xintai arise due to the interplay of various driving factors rather than any individual factor. In 2010, the most significant interactive effects occurred between annual average rainfall ($X_4$) and construction land proportion ($X_8$), as well as between annual average rainfall ($X_4$) and population density ($X_9$), with explanatory power values (q) of 0.5223 and 0.4449, respectively, explaining about 50% of the spatial pattern of social-ecological vulnerability in Xintai City. In 2020, the strongest interactions involved soil texture ($X_6$) with population density ($X_9$), soil erosion degree ($X_7$) with population density ($X_9$), and average slope ($X_2$) with population density ($X_9$), yielding explanatory power values (q) of 0.5973, 0.5871, and 0.5823, respectively, accounting for approximately 59%. These findings indicate that construction land proportion, population density, and mining land proportion play a major role in influencing social-ecological vulnerability within Xintai City. In 2010, interactions involving construction land proportion ($X_8$) dominated alongside other factors like average elevation ($X_1$), average slope ($X_2$), terrain relief ($X_3$), annual precipitation ($X_4$), and mean temperature ($X_5$). By contrast, in 2020, interaction intensity increased between population density ($X_9$) and these same factors, suggesting a shift in dominant influencing forces from “construction land growth” to “population agglomeration.” For instance, the area of construction land expanded from 204.21 km² to 311.36 km² from 2000 to 2010, with a growth rate of 52.47%. From 2010 to 2020, the area of construction land rose from 204.21 km² to 311.36 km², with a growth rate of 2.51%. It can be seen that the period from 2010 to 2020 was a period of rapid growth in construction land. Similarly, the total population grew from 1.3377 million to 1.392 million from 2000 to 2010, with a growth rate of 4.06%. From 2010 to 2019, the total population of Xintai City reduced from 1.392 million to 1.3716 million, with a decline rate of 8.57%.

In 2010, there was a strong interaction among annual precipitation (X4), the proportion of construction land area (X8), and population density (X9), and the mutual influence of these factors significantly affected ecological vulnerability. This indicates that the potential threats to the ecological environment from climate change, urbanization, and human activities cannot be ignored. Entering 2020, factors such as soil texture (X6), the proportion of construction land area (X8), population density (X9), and the proportion of mining land area (X10) have shown more significant interactions, indicating that the roles these factors play in the ecosystem are becoming increasingly important. This change reflects the growing ecological pressure brought about by resource development and urbanization. Therefore, annual precipitation, the level of urbanization, and mining area development have become key determinants of ecological vulnerability. To address these challenges, enhancing land-use governance and planning is essential, alongside promoting synergy between economic growth and ecological protection, so as to support progress toward sustainability targets and improve ecosystem resilience.

3.3. Scenario Simulation of Social-Ecological Vulnerability

3.3.1. Scenario Settings

Based on the Multi-Criteria Evaluation (MCE) module of IDRISI 17.0 software, the OWA operator is employed for multi-scenario simulations of social-ecological vulnerability. The decision-making risk coefficient (α ∈ (0, ∞)) reflects how decision-makers balance risk aversion and opportunity (

Table 8. Results of the ordered weight.

Decision Risk Coefficient (α)		a = 0.001	a = 0.5	a = 0.8	a = 1	a = 1.2	a = 2	a = 1000
Scenario		-	-	1	2	3	-	-
criterion weights	X5	0.9977	0.3162	0.1585	0.0526	0.0631	0.0100	0.0000
	X14	0.0007	0.1251	0.1116	0.0526	0.0773	0.0279	0.0000
	X13	0.0004	0.0918	0.0954	0.0526	0.0806	0.0429	0.0000
	X9	0.0003	0.0739	0.0843	0.0526	0.0807	0.0550	0.0000
	X15	0.0002	0.0619	0.0756	0.0526	0.0792	0.0644	0.0000
	X17	0.0002	0.0530	0.0682	0.0526	0.0765	0.0714	0.0000
	X12	0.0001	0.0459	0.0616	0.0526	0.0730	0.0760	0.0000
	X16	0.0001	0.0401	0.0556	0.0526	0.0689	0.0784	0.0000
	X4	0.0001	0.0351	0.0500	0.0526	0.0643	0.0789	0.0000
	X1	0.0001	0.0307	0.0448	0.0526	0.0594	0.0776	0.0000
	X19	0.0001	0.0267	0.0398	0.0526	0.0542	0.0745	0.0000
	X3	0.0001	0.0231	0.0350	0.0526	0.0487	0.0700	0.0000
	X8	0.0000	0.0197	0.0303	0.0526	0.0430	0.0642	0.0000
	X6	0.0000	0.0166	0.0258	0.0526	0.0371	0.0572	0.0000
	X2	0.0000	0.0136	0.0213	0.0526	0.0312	0.0492	0.0000
	X11	0.0000	0.0108	0.0170	0.0526	0.0250	0.0403	0.0000
	X7	0.0000	0.0080	0.0127	0.0526	0.0189	0.0308	0.0000
	X10	0.0000	0.0053	0.0084	0.0526	0.0126	0.0208	0.0051
	X18	0.0000	0.0026	0.0042	0.0526	0.0063	0.0105	0.9949

Note: The table presents the criterion weights of 19 social-ecological vulnerability indicators calculated under different decision risk coefficients (α = 0.001, 0.5, 0.8, 1, 1.2, 2, and 1000). The analysis focuses on three scenarios: ecological priority (α = 0.8), maintain the status quo (α = 1), and economic priority (α = 1.2).

). According to the parameterized criterion proposed by Yager, α ranges from 0.001 to 1000, indicating a shift in decision-makers’ attitudes from extremely positive to extremely negative [44]. Referring to the analysis conclusion of Malczewski, the interval of α ∈ (0.5, 2) can effectively reflect the fuzziness and uncertainty of real-world decision-making [45]. Therefore, in light of the typical contradictions in Xintai City, the three scenarios of α = 0.8, 1, and 1.2 are analyzed in detail.

Ecological priority scenario: α = 0.8 indicates that decision-makers prioritize ecological protection by enhancing governance and investment while prohibiting disorderly development in protected areas. Maintain the status quo scenario: α = 1 indicates that decision-makers have no bias towards any attitude and remain neutral. It represents the normal operation of the social ecosystem and belongs to a state of natural evolution. Economic priority scenario: α = 1.2 indicates that decision-makers attach greater prioritize economic development, with the development and expansion of construction land as the main line, while neglecting ecological protection and natural disaster governance.

3.3.2. Scenario Simulation Results Analysis

Ecological priority scenario (Figure 9): Low vulnerability areas dominate at 54.70%, followed by areas of medium vulnerability at 39.64%. High vulnerability occupies the smallest proportion at 5.65%, indicating significant protection of the local ecological conditions. High vulnerability zones are located in the urban center and Xiyangguo sub-center, but their patch sizes are relatively small. Low vulnerability areas primarily surround the county town. This simulation suggests that under the three red lines of territorial space and the Xintai Territorial Space Master Plan (2021–2035), decision-makers should prioritize ecological protection, enhance policy implementation, and invest more in regional ecological risk management. While safeguarding the ecological environment, they should also increase social and economic investments to promote balanced development.

Maintain the status quo scenario (Figure 9): Medium vulnerability has the largest area proportion at 43.73%, followed by higher and lower vulnerabilities at 5.91% and 50.36%, respectively. This suggests that Xintai’s overall vulnerability is troubling and underscores significant risks in the current management practices. High vulnerability areas cluster in urban center and Xiyangguo sub-center with medium-sized patches, while low vulnerability zones mainly surround the county town. Compared to the ecological priority scenario, decision-makers here show no clear preference; their approach is critical regarding environmental management.

Economic priority scenario (Figure 9): The proportion of highly vulnerable areas reached 7.50%, which was higher than that in both the ecological priority and the status quo scenarios. Compared with the status quo scenario, the area of highly vulnerable regions expanded by 26.85%, mainly extending towards the northern mountainous areas. Low and medium vulnerability regions account for 67.92% and 24.58% of the total study area, respectively. This pattern of vulnerability distribution reveals the significant pressure that the ecological environment in Xintai faces in the pursuit of economic development. Under this scenario, mineral development activities in Xintai are typically prioritized, with decision-makers often placing economic interests above ecological protection. This tendency may lead to excessive resource exploitation, exacerbating environmental issues including soil erosion, ecosystem destruction, and reduced biological diversity. In light of the increasingly prominent vulnerability issues, decision-makers must strive to find a reasonable balance between mineral development and environmental protection to ensure regional sustainable development.

4. Discussion

4.1. Fine Assessment at the Village Level

During recent times, there has been an intensified focus on the importance of fine-scale, village-level assessments in social-ecological vulnerability studies. Unlike coarse, macro-scale analyses, village-level research captures the spatial heterogeneity, local drivers, and community-level nuances that are often masked at broader scales. For example, Shukla et al. established an indicator-driven framework for assessing both social and biophysical vulnerabilities of agricultural communities in Uttarakhand, India, revealing pronounced differences in vulnerability even among neighboring villages [47]. Similarly, Gathongo and Tran applied the analytic hierarchy process to villages in the Mt. Kasigau region of Kenya, demonstrating that micro-scale assessments can identify unique local vulnerability patterns shaped by both environmental and socio-economic factors [48]. These studies highlight that village-level investigations are vital for detecting vulnerability hotspots, tailoring interventions, and informing targeted policy measures.

The significance of spatial heterogeneity in vulnerability is further corroborated by research in diverse contexts. For instance, Cao et al. conducted a hierarchical linear model analysis in Guizhou Province, China, to examine the drivers of household poverty vulnerability, finding that village-level conditions and resources exert a significant influence on household outcomes [49]. Wang et al. developed a multidimensional poverty measurement at the village level in China, illustrating how fine-resolution data can reveal the distribution and determinants of rural poverty with greater precision [50]. Yu et al. evaluated the vulnerability of villages in the Northeast China Plain using a suite of social-ecological indicators, proposing tailored improvement strategies based on village-specific characteristics [23]. These findings collectively support the view that village-level analysis enables a more nuanced understanding of vulnerability and a stronger empirical basis for rural policy design.

In addition to spatial heterogeneity, recent studies underscore the relevance of village-level research for understanding dynamic and multi-level processes. Shi et al. analyzed the rising vulnerability of heritage sites across 123 villages of Aba Prefecture, Sichuan, China, identifying how cultural, environmental, and institutional factors interact at the local level [51]. Ur Rahman et al. assessed household socio-economic conditions and resilience based on village-level data from Sichuan Province under COVID-19, demonstrating how community-level social capital and infrastructure mediate vulnerability during crises [52]. Wang et al. applied GIS and hierarchical linear modeling to detect differences in village-level regional development, further validating the necessity of micro-level approaches for capturing development disparities [53]. Zhang et al. utilized a multi-level analysis to explore the determinants of cropland abandonment in China’s mountainous regions, finding that parcel, household, and village-level factors jointly shape land use decisions [54]. Sun et al. designed an indicator framework to assess rural resilience in a coastal village of Zhejiang Province, emphasizing the value of composite, village-specific metrics for promoting sustainable and equitable rural development [55].

The collective evidence from these studies suggests that village-level, fine-grained research not only enhances the accuracy of vulnerability assessments but also improves the effectiveness of local governance and resource allocation. By integrating environmental, social, economic, and institutional dimensions at the community level, such research provides a robust scientific foundation for place-based adaptation strategies and rural revitalization.

4.2. Multiple Decision-Making Attitude Scenario Analysis

Scenario simulation has become a widely adopted approach in social-ecological vulnerability research, allowing for the assessment of system trajectories under varying assumptions about future social, economic, and environmental conditions. Unlike static evaluations, scenario-based analyses provide the advantage of simulating vulnerability dynamics across multiple decision-making attitudes, policy options, and uncertainties, thereby delivering more robust and actionable insights for risk management and adaptation planning. For instance, Asadi et al. employed a GIS-based OWA multi-criteria decision-making method to model earthquake vulnerability in Tehran, Iran, enabling spatial simulations under diverse risk scenarios [33]. Similarly, Chen et al. applied scenario simulation in southern China’s karst rocky desertification regions to identify the spatio-temporal patterns of social-ecological vulnerability and its underlying drivers, while Feng et al. used spatial analysis and factor identification to highlight scenario differences in ecological vulnerability in the Yellow River Basin [20,36].

Further supporting the significance of scenario simulation, Huang et al. assessed ecological vulnerability in Fujian Province by integrating the SRP-SES framework and OWA operator, explicitly addressing the uncertainty of multiple decision-making attitudes [46]. Malczewski introduced the OWA operator in GIS-based multi-criteria evaluation, providing a flexible tool for scenario simulation and land use suitability analysis [45]. Pirasteh et al. utilized spatially explicit modeling and multi-scenario assessment to examine vulnerability to multiple environmental hazards in mountainous regions, while Zhang et al. (2025) simulated the spatio-temporal patterns of ecological vulnerability on the Loess Plateau under different scenarios, revealing shifting vulnerability trends [25,28].

Additionally, scenario-based approaches have been applied to diverse regions and contexts: Zhong et al. analyzed vulnerability shifts in the Yanhe River Basin under varied urban expansion scenarios [22], Chen et al. simulated vulnerability patterns and trends in Yulin City [30], and Zhang et al. used the DPSIR-OWA-GIS method for multi-scenario risk prediction in Guangxi-Beibu Gulf [31]. Studies such as Dasgupta et al. (2022), Gupta et al. (2020), and He et al. (2021) have further demonstrated that scenario simulation is crucial for evaluating system adaptability and supporting flexible policy formulation under future climate and socio-economic change [32,56,57].

4.3. Social-Ecological Vulnerability of Resource-Based Cities

The social-ecological vulnerability of resource-based cities exhibits distinct characteristics driven by the legacy of intensive resource extraction, ongoing industrial restructuring, and complex patterns of population movement. Previous studies have consistently shown that these regions are subject to unique challenges, including industrial decline, resource depletion, and the challenges of ecological restoration [30,31,32,36]. For instance, Berrouet et al. proposed a conceptual framework emphasizing multi-dimensional pressures in resource-dependent areas [7], while Chen et al. provided a detailed analysis of the spatio-temporal patterns and driving forces of vulnerability in the core karst rocky desertification zone, identifying mining development and population mobility as primary factors [36]. Dai et al. further explored the evolutionary process and drivers of ecological vulnerability in Panzhihua, a prototypical resource-development city [14].

In addition, recent research has highlighted the crucial roles of land use change and population density as dominant social drivers of vulnerability, reflecting the profound impacts of rapid urbanization and ongoing industrial transformation in resource-based regions [17,19,37,40,58]. These findings are reinforced by empirical evidence from studies such as Li et al., who demonstrated the mitigation potential of targeted ecological protection and restoration in resource-intensive zones [15], and Chen et al., who revealed the multi-dimensional impacts of mining development and industrial restructuring in Yulin, China [30]. Moreover, the complex interplay between resource development, land use, and ecological risk has been systematically analyzed using advanced modeling methods, as demonstrated in the work of Zhang et al. and Cao et al. [31,49].

Collectively, these studies underscore the necessity for differentiated, context-specific frameworks and policies that address the intricate vulnerability profiles of resource-based cities at the grassroots level.

4.4. International Case Comparisons

The spatio-temporal patterns and key drivers of socio-ecological vulnerability in Xintai can be contextualized with experiences from other resource-exhausted regions. In Germany’s Ruhr region, industrial decline was mitigated through economic diversification, green infrastructure, and multi-scalar governance, including bicycle networks and city branding initiatives [59,60,61]. In the U.S. Rust Belt, including Detroit, proactive industrial policies and migration-driven adaptation strategies alleviated economic and social vulnerabilities, whereas insufficient intervention in Detroit led to severe urban decay [62,63]. Johannesburg, South Africa, demonstrates how governance-led spatial projects such as “Corridors of Freedom” address both resource decline and entrenched social inequality [64].

Xintai villages share common drivers with these cases, including industrial monostructure and environmental stress, but differ in scale and governance. Unlike the Ruhr and Rust Belt, transformation is largely state-led at the village level, with limited bottom-up initiatives. Compared to Johannesburg, social inequality is less structurally entrenched, yet vulnerability emerges from interactions among environmental stressors, social networks, and economic restructuring. These comparisons highlight the universality of economic, social, and environmental drivers in resource-exhausted contexts while emphasizing the distinctive Chinese pathway of state-guided resilience, which effectively averts worst-case outcomes but could benefit from integrating community-driven and culturally sensitive strategies to enhance long-term sustainability.

5. Conclusions

This research focuses on Xintai, a representative resource-based city, and establishes a village-level social-ecological vulnerability assessment framework. It examines the spatio-temporal patterns of social-ecological vulnerability from 2010 to 2020. The geographical detector method is used to explore driving mechanisms and scenario simulations for this vulnerability. Key findings include the following:

(1)

Temporal trends in social-ecological vulnerability: By analyzing the social-ecological vulnerability index (VI) from 2010 to 2020 shows that high-vulnerability areas have decreased over the past decade, with particularly notable improvements in the urban center and former coal mining zones. Exposure changes were minor, indicating no substantial increase in external threats. Adaptability has generally improved, showcasing the positive outcomes achieved by Xintai in strengthening the resilience of its social-ecological system. This trend highlights the effectiveness of traditional industry upgrades and the development of emerging industries in reducing social-ecological vulnerability.

(2)

Spatial heterogeneity of social-ecological vulnerability: From 2010 to 2020, urban center of Xintai experienced a general decline in social-ecological vulnerability, whereas peripheral regions exhibited an increase in vulnerability. This phenomenon is primarily attributed to the peripheral areas being predominantly farmland and low mountains, where recent conversions of farmland into forests and population outflows have contributed to heightened vulnerability. Over the last decade, approximately 917.41 km² of the region underwent changes in vulnerability levels, representing 51.33% of the overall study area. Among these, 36.7% showed increased vulnerability, while 14.63% demonstrated reduced vulnerability. The global Moran’s I index declined from 0.7798 to 0.4645 between 2010 and 2020, showing a statistically significant (p < 0.01) positive spatial autocorrelation in vulnerability, but with diminishing clustering intensity. The spatial association pattern of social-ecological vulnerability is mainly manifested as high-high and low-low clusters.

(3)

Driving mechanisms of social-ecological vulnerability: The proportion of construction land (0.2586), population density (0.2558), annual average rainfall (0.2443), soil texture (0.2137), and the proportion of mining land (0.1621) act as key determinants contributing to spatial heterogeneity in vulnerability. Social factors contribute significantly more than natural factors. Double-factor interaction detection further reveals that the interaction between the proportion of construction land and population density with other factors is most pronounced. The combined influence of paired variables exceeds the impact of an individual driver, indicating that the spatial variation in social-ecological vulnerability in Xintai results from the interplay of multiple driving forces rather than being determined by a single factor. This underscores the combined influence of natural and human drivers on vulnerability differentiation.

(4)

Simulation of social-ecological vulnerability: As the decision-making risk coefficient α increases, decision-making attitudes shift from optimism to pessimism, leading to a gradual rise in social-ecological vulnerability in Xintai. In the ecological priority scenario (α = 0.8), low vulnerability dominates, followed by low-medium vulnerability, with relatively small patches of high vulnerability. In the maintain-the-status quo scenario (α = 1), medium vulnerability occupies the largest area, followed by high and low-medium vulnerability, with medium-sized patches of high vulnerability. Under the economic priority scenario (α = 1.2), high vulnerability covers the largest area. In this scenario, decision-makers prioritize economic benefits over ecological protection, emphasizing the importance of regional policies that harmonize economic and social development, and ecological preservation. Decision-makers should flexibly adjust strategies within a reasonable range based on actual conditions, goals, and regional policies, implementing tailored management measures at different development stages.

Based on the findings of this study, four key policy implications are proposed for Xintai and similar resource-exhausted cities. First, for high-vulnerability villages in central and northern mining subsidence areas, policy should prioritize managed retreat and ecological restoration. Residents can be relocated through Ecological Migration Programs with financial incentives and vocational training, while essential services are maintained for remaining populations. Abandoned mining lands should be converted into ecological zones via reforestation or wetland restoration, potentially generating carbon credits. Second, for medium- and low-vulnerability villages near the urban core, policy should focus on smart growth and economic diversification. Upgrading transportation and digital infrastructure, attracting light manufacturing and agri-businesses, and promoting cultural heritage tourism can strengthen local economies and integrate these villages into the urban system. Third, at the city-wide level, coordinated development between urban centers and peripheral regions should be enhanced. Policies should support peripheral regions with infrastructure development, sustainable land management, and resource allocation to reduce regional disparities, while social factors such as community services, population density, and construction land use should be optimized to strengthen resilience. Fourth, industrial upgrading and fostering emerging sectors, including renewable energy and green technologies, are essential to reduce social-ecological vulnerability. Supporting industrial restructuring, technological innovation, and dynamic management strategies can balance economic growth with ecological preservation, improve production efficiency, and enhance overall system resilience.

Nonetheless, this study has certain limitations. For instance, the scenario simulations were conducted based on historical data and did not fully capture the cascading effects of uncertain factors such as extreme climatic events including increased frequency of heavy rainfall or abrupt policy shifts such as accelerated carbon neutrality policies. In future research, more sophisticated approaches such as land use and land cover modeling with CA -ANN or the incorporation of DEM data derived from InSAR analysis could be employed to forecast socio-ecological vulnerability over the coming decades with greater accuracy.