Introduction

Shanxi Province is one of the most coal-rich regions in China, holding significant strategic importance for ensuring national energy security and supporting economic growth1. However, large-scale coal mining activities2,3 have caused substantial ecological degradation4. Therefore, the protection and restoration of ecosystems in Shanxi and its designated mining areas are essential. These efforts not only contribute to improving local environmental quality but also help mitigate the broader impacts of global climate change5. Conducting research and continuous monitoring of ecological quality in these areas6 is of great importance for regional ecological conservation and the effective implementation of sustainable development strategies7.

Ecological quality is intricately linked to human life, making the rational assessment of regional ecological environment quality particularly significant. The rapid advancement of remote sensing technology8 has opened up new possibilities for large-scale, long-term monitoring of dynamic changes in ecological environment quality. By analyzing remote sensing data, real-time monitoring of ecosystem changes can be achieved, allowing for the evaluation of environmental quality and providing support for informed decision-making. For instance, Zhong et al. utilized the Normalized Difference Vegetation Index (NDVI) to assess regional vegetation coverage9; Liu et al. identified landscape drivers of urban heat island effects using land surface temperature (LST)10; and Li et al. investigated ecological changes in the Mu Us Sandy Land by combining land cover trends with NDVI data11. Ye et al. quantified ecological quality in the Loess Plateau through soil moisture (SM) measurements12. While these remote sensing indices provide valuable insights, the assessment based on single indicators is often one-dimensional and insufficient for comprehensive ecological evaluations. As a result, the trend toward utilizing multiple integrated remote sensing-based indicators for ecological quality assessment has become a necessity. Xu et al. proposed the Remote Sensing Ecological Index (RSEI), which integrates greenness, wetness, heat, and dryness13. Through statistical methods such as principal component analysis, RSEI can ascertain the weights of each indicator’s impact on ecological quality, leading to comprehensive ecological quality assessments. Currently, The RSEI has gained widespread application in ecological assessments across various landscapes, including urban agglomerations14, islands10, and mountainous regions15. This established methodology makes RSEI an appropriate and reliable tool for evaluating ecological quality in Shanxi Province and its designated mining areas. While previous studies have commonly employed the NDVI to characterize vegetation greenness, the Kernel Normalized Difference Vegetation Index (kNDVI) offers distinct advantages. The kernel-based approach16 of kNDVI effectively mitigates saturation effects, accounts for complex phenological cycles, and better accommodates seasonal variations. Furthermore, kNDVI demonstrates superior performance in resolving mixed pixel issues and exhibits enhanced sensitivity to vegetation dynamics. These technical advantages enable kNDVI to provide more accurate vegetation coverage and health data across diverse ecological conditions, consequently improving the robustness of ecological assessments. Therefore, incorporating kNDVI as the greenness component in RSEI construction for monitoring ecological changes in Shanxi Province represents both a methodologically sound and scientifically rigorous approach.

Ecological quality, as a key indicator of ecosystem health, stability, and resilience17, is closely linked to human well-being and socio-economic development18. In the context of globalization, various factors—including topography, climate change, and population density—exert substantial influence on ecological environmental quality19. Specifically, moderate wind speeds and adequate precipitation contribute to enhanced ecosystem stability and productivity, while elevation and slope affect ecological conditions by altering local climate regimes, soil characteristics20, and light availability. Areas with high population density21 often face intensified land development and resource consumption, leading to the degradation of ecosystem services22 and overall ecological quality. Identifying and quantifying the factors that influence ecological quality is essential for implementing targeted measures to prevent environmental pollution and mitigate ecosystem degradation. Such efforts are critical for advancing ecological protection and restoration23. Therefore, when evaluating the driving forces behind ecological quality, it is imperative to take a comprehensive approach that accounts for the complex interplay among multiple influencing factors, as well as the effects of spatial heterogeneity.

When analyzing the driving factors behind changes in ecological quality, traditional regression methods, correlation analyses, and residual analyses exhibit certain limitations, particularly as these approaches often assume a linear relationship between factors and ecological quality. This assumption may not adequately capture the complex coupling effects among multiple variables. In contrast to conventional analytical techniques, CatBoost, a machine learning algorithm based on gradient boosting decision trees24, effectively addresses classification and regression challenges while learning intricate nonlinear relationships between input and output variables. Furthermore, the application of SHAP (SHapley Additive exPlanations) tools allows for the interpretation of CatBoost model predictions25, quantifying the impact of different features on these predictions and thereby providing deeper insights into large-scale ecological influencing factors. In addition, traditional methods often fail to account for the spatial heterogeneity inherent in geospatial data, which can significantly influence the assessment and prediction of ecological quality26. The Geographically Weighted Regression (GWR) model is specifically designed to address spatial data27,28, allowing it to capture spatial heterogeneity and reveal localized relationships between ecological variables and the RSEI across different regions. Applying the GWR model to analyze the driving forces behind ecological change is essential for understanding the complex spatial dynamics of ecosystems29. While CatBoost excels at modeling nonlinear relationships and handling complex variable interactions, GWR is particularly effective in detecting spatial variation and regional disparities. The combined use of these two methods enables a more comprehensive and accurate analysis of ecological quality changes and their underlying drivers30.

This study employs the RSEI to evaluate the variations in ecological quality across Shanxi Province and its planned mining areas. By calculating the kNDVI, humidity (WET), land surface temperature (LST), and dryness (NDBSI), the RSEI is constructed through PCA to provide a comprehensive assessment of the ecological environment quality in the region. Subsequently, utilizing the CatBoost model and GWR, this research reveals the influence of different factors on the ecological environment, thereby enhancing the reliability of the study’s findings. The objectives of this research are to: (1) investigate the dynamic changes in ecological quality (RSEI) in Shanxi Province and its planned mining areas from 2000 to 2020; (2) analyze the trends in RSEI variations within the region; (3) explore the potential factors contributing to changes in ecological quality in Shanxi Province and its mining areas; and (4) reveal the spatial heterogeneity of ecological quality and examine its specific relationships with underlying factors.

Materials and methods

Study area

Shanxi Province (34°34′-40°44′N, 110°14′-114°33′E) is located in northern China, in the eastern part of the Loess Plateau, with a total area of 156,700 square kilometers (Fig. 1). The province encompasses twelve planned mining areas, featuring a wealth of diverse mineral resources and a complex geological environment. The climate is classified as a temperate continental monsoon climate, characterized by significant temperature variations across regions, with temperatures decreasing from south to north and from plains to mountainous areas. The province’s geomorphology is predominantly mountainous and hilly, comprising approximately 80% of the total area. The study area includes critical regions such as the Taihang and Lüliang mountains, which are rich in resources but face numerous challenges, including resource depletion and ecological restoration.

Fig. 1
figure 1

Location map of the study area. (This image was created using ArcGIS 10.8 software, which is an Esri product and is publicly available at the URL (https://www.esri.com/). The base map is based on the standard map (GS2024, No. 0650), and no modifications have been made to the map boundaries.)

Full size image

Technical route

The detailed workflow of this study is illustrated in Fig. 2: (1) RSEI was constructed based on the GEE platform to invert the ecological quality changes in Shanxi Province and its planned mining areas from 2000 to 2020; (2) Theil-Sen analysis and Mann-Kendall trend tests were employed to analyze the RSEI trend; (3) Moran’s I was utilized to assess the spatial correlation of RSEI in Shanxi Province; (4) The interactions between RSEI and potential driving factors were analyzed using CatBoost and GWR models.

Fig. 2
figure 2

Technology roadmap. (This maps was created using ArcGIS 10.8 software, which is an Esri product, generated using our dataset, and is publicly available at the URL (https://www.esri.com/).)

Full size image

Datasets and preprocessing

This study employed the MODIS dataset to construct the ecological quality assessment index. The data sources included the MOD09A1 (2000–2023, 500 m resolution, 8-day interval) and MOD11A2 (2000–2023, 500 m resolution, 8-day interval) from the MODIS collection. The JRC annual water classification historical data was utilized for water body masking. Additional datasets were leveraged to extract potential factors that may influence ecological quality in the study area, including Terra Climate (2000–2020, providing climate factor data), NASA’s SRTM Digital Elevation data (90 m resolution), population density data provided by East View Cartographic, and MYD17A3HGF (2000–2020, net primary productivity data). All data were acquired from the GEE cloud platform, with all datasets resampled to a resolution of 500 m. Atmospheric correction and relevant preprocessing techniques were applied to enhance data quality. To account for seasonal variation and meteorological conditions, the data time window was set from June 15 to September 15 each year, ensuring consistency in vegetation status and ecological outcomes.

Methods

RSEI indicator construction

RSEI leverages ground information and ecological data obtained through remote sensing technologies, employing mathematical and statistical methods for processing and analysis, thereby facilitating rapid assessment and monitoring of the health status and environmental quality of specific regions31. Given that greenness, humidity, heat, and dryness are crucial components of the ecological environment, this study selects these four ecological factors to construct the RSEI index32.

$$ \:RSEI = f(Greenness,Wetness,Heat,Dryness) $$
(1)

where f is the set of functions used to perform PCA.

kNDVI applies a weighted average of NDVI for each pixel and its surrounding pixels to more accurately capture the details and variations in vegetation coverage. Thus, kNDVI serves as a representation of greenness. The calculation formula is as follows:

$$\:NDVI=\frac{NIR-Red}{NIR+Red}$$
(2)
$$ \:kNDVI = tanh\left( {\left( {\frac{{NIR - Red}}{{2\sigma \:}}} \right)^{2} } \right) = tanh\left( {\left( {\frac{{NDVI}}{{2\sigma \:}}} \right)^{2} } \right) $$
(3)

where NIR and RED represent the reflectances of the near-infrared (NIR1) and red bands of MOD09A1, respectively. The symbol σ denotes a length scale that is proportional to the average reflectance of the near-infrared and red bands, which can be adjusted. When σ is set to 0.5(NIR + RED), the calculation formula is as follows:

$$\:kNDVI=tanh\left({NDVI}^{2}\right)$$
(4)

Humidity (WET) is calculated using the formula:

$$ \begin{gathered} \:WET = 0.1147 \times \:\rho \:_{1} + 0.2489 \times \:\rho \:_{2} + 0.2408 \times \:\rho \:_{3} + 0.3132 \hfill \\ \quad \quad \times \:\rho \:_{4} - 0.3122 \times \:\rho \:_{5} - 0.6416 \times \:\rho \:_{6} - 0.5087 \times \:\rho \:_{7} \hfill \\ \end{gathered} $$
(5)

where \(\:{\rho\:}_{i}\) (i = 1, 2, …, 7) represents the reflectance of the red, near-infrared 1, blue, green, near-infrared 2, short-wavelength infrared 1, and short-wavelength infrared 2 bands of the MOD09A1 image, respectively.

Dryness consists of the Index of building (IBI) and the index of bare Soil (SI). The formula for calculation is:

$$\:NDBSI=\left(IBI+SI\right)/2$$
(6)
$$\:IBI=\frac{2{\rho\:}_{6}/({\rho\:}_{6}+{\rho\:}_{2})-{\rho\:}_{2}/({\rho\:}_{2}+{\rho\:}_{1})+{\rho\:}_{4}/({\rho\:}_{4}+{\rho\:}_{6})}{2{\rho\:}_{6}/({\rho\:}_{6}+{\rho\:}_{2})+{\rho\:}_{2}/({\rho\:}_{2}+{\rho\:}_{1})+{\rho\:}_{4}/({\rho\:}_{4}+{\rho\:}_{6})}$$
(7)
$$\:SI=\frac{\left({\rho\:}_{6}+{\rho\:}_{1}\right)-\left({\rho\:}_{2}+{\rho\:}_{3}\right)}{\left({\rho\:}_{6}+{\rho\:}_{1}\right)+\left({\rho\:}_{2}+{\rho\:}_{3}\right)}$$
(8)

where \(\:{\rho\:}_{i}\) (i = 1, 2, …, 6) denotes the reflectance of the red, near-infrared 1, blue, green, near-infrared 2, and short-wavelength infrared 1 bands of the MOD09A1 image, respectively.

Heat is expressed in terms of land surface temperature (LST). MOD11A2 LST products providing the necessary data. The raw LST data (LST0) requires unit conversion from Kelvin (K) to degrees Celsius (°C). The formula for this conversion is:

$$\:LST=0.02LS{T}_{0}-273.15$$
(9)

In this paper, the RSEI is calculated using the following formula:

\(\:RSEI=f(kNDVI,Wet,NDBSI,LST\))(10).

After obtaining the results of the four ecological factors, it is essential to linearly map each indicator’s values to the range of [0, 1] for normalization. This process eliminates the impacts caused by differing units and value ranges, allowing for PCA to construct the RSEI.

Furthermore, RSEI is subjected to additional normalization to facilitate comparison and measurement, as shown in the following formula:

$$\:RSEI=\frac{{RSEI}_{0}-{RSEI}_{min}}{{RSEI}_{max}-{RSEI}_{min}}$$
(11)

where \(\:{RSEI}_{0}\)is the initial RSEI, and \(\:{RSEI}_{min}\) and \(\:{RSEI}_{max}\) represent the minimum and maximum values of \(\:{RSEI}_{0}\), respectively. A higher RSEI value, approaching 1, indicates better ecological environmental quality, while a lower RSEI value signifies poorer ecological environmental quality.

To enhance the assessment and comparison of ecological conditions, the RSEI values were categorized into five levels from high to low: excellent (0.8–1), good (0.6–0.8), moderate (0.4–0.6), fair (0.2–0.4), and poor (0–0.2). This classification provides a clearer understanding of ecological quality, with values closer to 1 indicating better ecological environment health.

Sen’s slope estimator and Mann–Kendall statistical test

Sen’s Slope Estimator can accurately estimate the trend slope in time series data, encompassing both linear and nonlinear trends. Its robustness allows it to effectively handle outliers and data with significant fluctuations. By employing Sen’s Slope Estimator, the strength of the trend in each time series can be determined, thereby identifying the rate and magnitude of ecosystem changes. The calculation formula is as follows:

$$\:\beta\:=Median\left(\frac{{X}_{j}-{X}_{i}}{j-i}\right),\forall\:j>i$$
(12)

where \(\:\beta\:\) denotes the slope and \(\:{X}_{i}\:\)and \(\:{X}_{j}\:\)denote the data values for year i and year j, respectively.

The MK test is employed to detect the presence and significance of trends in time series data33. It does not rely on assumptions about the data distribution, making it applicable to various types of ecological data. While Sen’s Slope Estimator provides a quantitative assessment of the trends, the Mann–Kendall test confirms the significance and statistical stability of these trends34. The calculation formulas for the MK test are as follows:

$$\:S={\sum\:}_{i=1}^{n-1}{\sum\:}_{j=i+1}^{n}sgn\left({x}_{j}-{x}_{i}\right)$$
(13)
$$\:sgn\left({x}_{j}-{x}_{i}\right)=\left\{\begin{array}{c}1,\:{x}_{j}-{x}_{i}>0\\\:0{,x}_{j}-{x}_{i}=0\\\:-1{,x}_{j}-{x}_{i}<0\end{array}\right.$$
(14)
$$\:V=n\left(n-1\right)\left(2n+5\right)/18$$
(15)
$$\:Z=\left\{\begin{array}{c}\frac{S-1}{\sqrt{V}},S>0\\\:0,\:S=0\\\:\frac{S+1}{\sqrt{V}},S<0\end{array}\right.$$
(16)

where n denotes the time series length; sgn is the sign function; S is the statistic; and Z is the trend significance test statistic.

In the MK test, the significance of RSEI changes is determined based on a set significance level α. Typically, α is set at 0.05, which corresponds to a Z value of ± 1.96. Therefore, when the absolute value of the calculated Z from the MK test exceeds 1.96, the change passes the significance test at a 95% confidence level, indicating that the trend change is significant; conversely, the opposite holds true. The results are categorized into five classes, as shown in the table below(Table 1).

Table 1 RSEI trend classification.
Full size table

Hurst exponent

The Hurst exponent is an indicator used to measure the long-term memory in time series or fractals, aiding in understanding the fundamental dynamics of the series and providing a degree of predictability for changes in ecological quality35. The formula for calculating the Hurst exponent is as follows:

$$\:H=log(R/S)/log\left(n\right)$$
(17)

where H is the Hurst exponent. In the actual computation, it is typically necessary to perform multiple decompositions and calculate the R/S values at different scales, followed by averaging these R/S values to obtain the final estimate of the Hurst exponent.

The Hurst exponent is commonly employed to analyze the persistence or anti-persistence of time series, with values ranging from 0 to 1. Specifically, when 0 < H < 0.5, it indicates anti-persistent change; the closer H is to 0, the stronger the anti-persistence. When 0.5 < H < 1, it suggests persistent change; the closer H is to 1, the stronger the persistence. When H equals 0.5, it denotes that the time series is independent and random, exhibiting uncertainty with no correlation between past and future trends.

Global moran’s

Global Moran’s I is a commonly used method for spatial autocorrelation analysis, employed to measure the spatial correlation among observations within a spatial dataset and to reveal the spatial relationships between reference units and their neighboring units26. The Moran’s I statistic ranges from − 1 to + 1; a value close to + 1 indicates that the observations of neighboring geographic units tend to be similar, while a value close to -1 suggests that the observations are inclined to be opposite. A Moran’s I value near 0 indicates that the data are randomly distributed in space, showing no significant spatial autocorrelation. The formula for calculating Moran’s I is as follows:

$$\:Moran{\prime\:}s\:I=\frac{m*{\sum\:}_{i=1}^{m}{\sum\:}_{j}^{m}{w}_{ij}({x}_{i}-\stackrel{-}{x})({x}_{j}-\stackrel{-}{x})}{{\sum\:}_{i=1}^{m}{\sum\:}_{j}^{m}{w}_{ij}{({x}_{i}-\stackrel{-}{x})}^{2}}$$
(18)

where m represents the total number of samples, \(\:{x}_{i}\)and \(\:{x}_{j}\)denote the RSEI values at positions i and j, \(\:{w}_{ij}\)indicates the spatial weight value between i and j, and \(\:\stackrel{-}{x}\) is the mean value of the RSEI.

CatBoost

CatBoost is a machine learning algorithm based on gradient-boosted decision trees, demonstrating exceptional performance in handling classification and regression problem36. One notable advantage of CatBoost is its ability to efficiently manage categorical features while achieving high prediction accuracy. Additionally, CatBoost employs regularization techniques to mitigate overfitting, enhancing the robustness of the model when dealing with complex data. The fundamental steps for training a CatBoost model are illustrated in Fig. 3: first, create a CatBoost model and train it using the training dataset; once the model is trained, it can be utilized for predicting new data. Throughout the training and prediction processes, CatBoost simplifies the complexities of data preprocessing and offers an efficient gradient boosting algorithm, assisting users in constructing more accurate models.

Fig. 3
figure 3

CatBoost implementation process. In the figure X_traint ,Y traint is the training set, X_test, Y _test is the test set, pt is the prediction of the t-th sample, a is the weight of the t-th sample, and Y is the final prediction obtained by weighting the prediction of each learner.

Full size image

Geographically weighted regression (GWR)

In spatial research, particularly at large scales, conventional multivariate linear regression models may fail to accurately capture the variations and heterogeneity in geographic space. The GWR model, on the other hand, is adept at exploring the relationships between variables in spatial data, effectively reflecting the correlations between dependent and independent variables across geographic space37. Moreover, the GWR model estimates parameters through independent linear regressions within each local area, utilizing local parameter estimates instead of global ones. The kernel type is “bisquare”and the number of neighbors is determined by the optimal bandwidth (gwr_bw).The formula for the GWR model is as follows:

$$\:{\varvec{y}}_{\varvec{i}}={\varvec{\beta\:}}_{0}({\varvec{u}}_{\varvec{i}},{\varvec{v}}_{\varvec{i}})+{\sum\:}_{\varvec{k}}{\varvec{\beta\:}}_{\varvec{k}}({\varvec{u}}_{\varvec{i}},{\varvec{v}}_{\varvec{i}}){\varvec{x}}_{\varvec{k},\varvec{i}}+{\varvec{\epsilon\:}}_{\varvec{i}}$$
(19)

where \(\:{y}_{i}\) is the dependent variable for sample i. \(\:({u}_{i},{v}_{i})\) are the coordinates of sample i. \(\:{\beta\:}_{0}({u}_{i},{v}_{i})\)is the intercept term for sample i. \(\:{\beta\:}_{k}({u}_{i},{v}_{i})\) is the k-th regression parameter for sample i. \(\:{x}_{k,i}\)​ is the k-th independent variable for sample i. \(\:{\epsilon\:}_{i}\) represents the random error.

Model evaluation

To evaluate the model’s estimation performance, the dataset is divided into 80% training set and 20% testing set. Two evaluation metrics are used: the coefficient of determination (R2) and the root mean square error (RMSE). Additionally, to assess the model’s generalization ability, we use the 5-fold cross-validation method. Specifically, the data is randomly divided into 5 subsets, and the model is trained and tested on each subset. The final model performance is the average of the results from each fold. The calculation formulas are as follows:

$$\:{R}^{2}=1-\frac{{\sum\:}_{i=1}^{n}{\left({x}_{i}-{y}_{i}\right)}^{2}}{{\sum\:}_{i=1}^{n}{\left({x}_{i}-\stackrel{-}{{y}_{i}}\right)}^{2}}$$
(20)
$$\:RMSE=\:\sqrt{\frac{1}{m}{\sum\:}_{i=1}^{n}{\left({x}_{i}-{y}_{i}\right)}^{2}}$$
(21)

Where \(\:{x}_{i}\) is the true value of RSEI, \(\:{y}_{i}\) is the estimated value of RSEI, \(\:{\stackrel{-}{y}}_{i}\) is the mean of the true RSEI values, and n is the number of samples in the test set.

Results

RSEI-based evaluation of ecological environment quality

The average RSEI values and PCA results for Shanxi Province from 2000 to 2020 are presented in Table 2. The contribution of the first principal component (PC1) reached a maximum of 83.13% and a minimum of 65.70%, with a multi-year average of 76.14%. This indicates that PC1 concentrates the primary characteristics of each indicator, confirming the scientific validity of using kNDVI, WET, LST, and NDBSI in constructing the RSEI. Furthermore, pixel-by-pixel correlation analysis of RSEI and kNDVI from 2000 to 2020 (Figure S1) revealed a high correlation between the two (R2 = 0.90), which to a certain extent verified the reliability and representativeness of RSEI.

Table 2 Mean and PC1 results of RSEI in Shanxi Province (2000–2020).
Full size table

The time-series variation of average RSEI values in Shanxi Province and its planned mining areas from 2000 to 2020 is presented in Fig. 4. During this period, RSEI values in Shanxi ranged from 0.42 to 0.58, with an average annual increase of 1.44%. The lowest value occurred in 2000, while the highest was observed in 2020. Throughout the study period, the average ecological grade consistently remained at a moderate level, indicating a relatively stable and favorable ecological condition in the region. To better illustrate the long-term trend of RSEI averages, the Savitzky-Golay (SG) filter was applied to smooth the time-series curve. The smoothed curve indicates an upward trend from 2000 to 2008 and from 2011 to 2014, while fluctuations or downward trends occurred during other periods. The research found that starting in 2000, ecological restoration projects were initiated in Shanxi Province. The significant rise observed from 2010 to 2014 may be attributed to the comprehensive launch of ecological province construction in 2011 and the in-depth implementation of ecological environment governance and restoration projects, resulting in noticeable improvements in the ecological environment during this period. Overall, from 2000 to 2020, the ecological environment quality in Shanxi Province showed a clear improvement trend, indicating the effectiveness of ecological protection measures.

Fig. 4
figure 4

Time-series changes of the mean RSEI value in Shanxi Province from 2000 to 2020.

Full size image

Figure 5 reveals the spatiotemporal distribution of RSEI values in Shanxi Province and the planned mining areas. From 2000 to 2020, the overall ecological quality in the eastern region of the study area was superior to that of the western region, with areas of better ecological quality concentrated on both sides of the Lüliang and Taihang mountain ranges. Notably, the ecological quality in the Jincheng mining area, Xiangning mining area, and the Xishan gujiao mining area was relatively high, situated in the central and southern parts of Shanxi Province. Conversely, the ecological quality in the Datong mining area, Liulin mining area, Lishi mining area, and Hebao mining area was poorer, predominantly located in the western part of Shanxi Province.

Fig. 5
figure 5

(a) to (e) represent the spatial and temporal distribution of RSEI in Shanxi Province and planned mining areas in 2000, 2005, 2010, 2015 and 2020, respectively. (This image was created using ArcGIS 10.8 software, which is an Esri product, generated using our dataset, and is publicly available at the URL (https://www.esri.com/).)

Full size image

Analyzing the changes in the proportion of areas with different RSEI grades in Shanxi Province, as shown in Fig. 6, reveals that regions with moderate RSEI levels accounted for the highest share, reaching approximately 40% ± 3%. The proportions of areas classified as fair and poor showed a downward trend, with the poor grade declining from 8% in 2000 to 0% in 2020, and the fair grade decreasing from 41 to 13%. Conversely, the proportions of areas with good and excellent grades exhibited a steady upward trend, with good areas increasing from 14 to 38% and excellent areas rising from 0 to 7%. These results indicate that the ecological quality in Shanxi Province in 2020 is significantly better than in 2000, demonstrating a steady improvement in ecological quality over the past 20 years.

Fig. 6
figure 6

Percentage change of area occupied by different classes of RSEI in Shanxi Province.

Full size image

Analysis of the annual mean RSEI box plots for the twelve planned mining areas (Fig. 7) reveals that the Jincheng mining area demonstrates relatively better ecological quality with a mean RSEI value of 0.32 compared to other mining areas. In contrast, the Liulin mining area exhibits poorer ecological conditions, showing the lowest mean RSEI value of 0.16. The overall ecological assessment indicates that the average RSEI grades across all twelve planned mining areas fall within the “fair” to “poor” categories, highlighting the necessity for implementing ecological protection measures in these regions.

To better understand the trends in ecosystem changes, an assessment and comparison of the ecological conditions across different mining areas were conducted, categorizing the annual growth rates into three Levels based on their values: Level 1 for annual growth rates exceeding 2.5%; Level 2 for rates between 1.5% and 2.5%; and Level 3 for rates ranging from 0 to 1.5%. Among these, the Xiangning, Liulin, and Huozhou mining areas were classified as Level 1, with growth rates of 2.74% and 3.35%. The Pingshuo Shuo’nan, Hebao Piao, Lu’an, and Lishi mining areas fell into Level 2, with growth rates of 1.55%, 2.36%, 1.80%, and 2.29%, respectively. The Datong, Jincheng, Qinyuan, Xishan Gujiao, and Yangquan mining areas were categorized as Level 3, with growth rates of 1.48%, 1.17%, 1.42%, 1.17%, and 1.34%. Overall, the annual average growth rates across the planned mining areas indicate that from 2000 to 2020, the RSEI values in all mining areas displayed an upward trend, reflecting improvements in their ecological conditions.

Fig. 7
figure 7

Annual average change of RSEI in the planned mining area.

Full size image

Spatial and Temporal trends in RSEI

The RSEI values in Shanxi Province exhibit significant spatial variability, with the analysis utilizing Sen’s slope and the MK test to explore the temporal and spatial trends of RSEI, resulting in five distinct groups. As illustrated in Fig. 8, from 2000 to 2010, the overall RSEI in Shanxi Province demonstrated an upward trend, with the area of increase accounting for 79.47% of the total area. Notably, the majority of this increase was no significant rise, comprising 59.85%. From 2010 to 2020, the area of increase constituted 62.73%, with a significant rise of 9.48% primarily concentrated in the northern part of Shanxi Province. Over the entire period from 2000 to 2020, the area exhibiting an upward trend accounted for 88.3%, while the area showing a downward trend comprised 8.49%, primarily distributed in the southern and eastern regions. These results indicate a general upward trajectory in ecological quality across Shanxi Province, reflecting the significant effectiveness of ecological governance measures implemented in the region.

Fig. 8
figure 8

(a–c) Spatial distribution and area proportions of RSEI trends for 2000–2010, 2010–2020, and 2000–2020, respectively. (This image was created using ArcGIS 10.8 software, which is an Esri product, generated using our dataset, and is publicly available at the URL (https://www.esri.com/).)

Full size image

From 2000 to 2020, the area with increased RSEI in the 12 planned mining areas accounted for 87.08%, while the area that relative stable was only 2.89%, and the area with decreased RSEI for 10.03%. The spatiotemporal trends of RSEI in the planned mining areas align closely with the overall trends observed in Shanxi Province. The varying spatial trends of RSEI across different grades of mining areas and their respective proportions are detailed in Table 3. Overall, all grades of mining areas in Shanxi Province demonstrate an upward trend in RSEI, indicating a reduction in ecological vulnerability and an improvement in ecological conditions. Notably, the ecological improvement in Level 1 mining areas is particularly pronounced, significantly surpassing that of Level 2 and Level 3 areas. This suggests that ecological restoration efforts in Level 1 mining areas, such as those in Xiangning, Liulin, and Huozhou, have achieved more substantial results. This trend corresponds with the broader improvement in ecological quality throughout Shanxi Province, reflecting the positive outcomes of environmental protection and restoration initiatives in the region. Since 2007, Shanxi Province has implemented a special plan for the comprehensive management of subsidence areas, gradually restoring the ecological functions of the land in these areas through integrated measures such as surface backfilling, land reclamation, and ecological reconstruction. Additionally, between 2000 and 2020, the province advanced ecological restoration efforts for abandoned open-pit mines in key areas of the Yellow River Basin.

Table 3 Area of different Spatial trends in RSEI at various levels of the mine and their percentages.
Full size table

Spatial autocorrelation analysis of the RSEI

Spatial autocorrelation analysis was employed to explore the ecological quality in Shanxi Province by sampling RSEI into a 5 km × 5 km grid. The results, validated at a 1% significance level, revealed significant spatial autocorrelation of RSEI at the 99% confidence level, indicating a clear spatial correlation in ecological quality across various regions of Shanxi Province. The Moran’s I values for the years 2000, 2005, 2010, 2015, and 2020 were 0.78, 0.75, 0.74, 0.74, and 0.70, respectively (Fig. 9), demonstrating a gradual weakening of spatial autocorrelation in ecological quality, but the values remain relatively high, suggesting a strong spatial correlation still exists. The Moran scatter plot delineates spatial clustering or dispersion patterns across four quadrants: HH (high-high), LH (low-high), LL (low-low), and HL (high-low). The results indicate that RSEI’s Moran’s I scatter plot is predominantly concentrated in the HH and LL quadrants, highlighting a substantial positive spatial correlation in ecological quality across Shanxi Province and suggesting that the spatial distribution of RSEI exhibits a degree of aggregation.

Fig. 9
figure 9

2000, 2005, 2010, 2015 and 2020 Moran I scatter plots for the study area.

Full size image

To analyze the spatiotemporal variation in the distribution of RSEI within the study area, a Local Indicators of Spatial Association (LISA) cluster map was generated. The LISA clustering results (Fig. 10) demonstrate a clear spatial correlation between ecological quality patterns and mining activities, climatic conditions, and topographic characteristics in Shanxi Province. H-H clusters are predominantly distributed in the central and southeastern regions (e.g., Yuncheng and Jincheng), where higher annual precipitation and the ecological barrier effects of the Taihang and Lüliang Mountains limit mining disturbances. These areas also benefit from fertile river valleys such as the Fenhe and Qinhe, which support vegetation growth.

In contrast, L-L clusters are concentrated in northern coal mining hubs (Datong, Shuozhou) and western mining zones (Lüliang, Liulin, Hebao Pian), overlapping significantly with the open-pit mining areas outlined in the Shanxi Mineral Resources Plan. The temporal analysis from 2000 to 2020 reveals that H-H clusters have expanded and spatially consolidated, while L-L clusters have shifted northward and decreased in area. This trend corresponds with the implementation of ecological restoration projects in the Yellow River Basin, including vegetation planting and mining land rehabilitation. Notably, western mining areas like Liulin and Hebao Pian show reduced L-L clusters, whereas regions such as Xiangning and Jincheng exhibit increased H-H clusters, reflecting the combined effects of mining regulation, climate advantages, and terrain protection.

Fig. 10
figure 10

(a–e) shows the LISA clusters for the study area for the years 2000, 2005, 2010, 2015, and 2020. (This image was created using ArcGIS 10.8 software, which is an Esri product, generated using our dataset, and is publicly available at the URL (https://www.esri.com/).)

Full size image

Potential drivers of change affecting ecological quality

To comprehensively and systematically investigate the driving factors influencing RSEI, various potential drivers were treated as independent variables, with RSEI as the dependent variable. Due to data availability constraints—some datasets being updated only through 2020—five representative years (2000, 2005, 2010, 2015, and 2020) were selected for analysis. The Pearson correlation coefficient method was used to evaluate the linear relationships between RSEI and the selected driving factors for each of these years. The results, validated through significance testing, identify the key factors influencing RSEI, as presented in Table 4. At the same time, taking 2020 as an example, we analyzed the statistical characteristics of each variable (Table S2). The changes in ecological quality within mining areas are also affected by factors such as elevation, wind speed (Ws), Precipitation accumulation (Pa), Vapor pressure (Vp) and NPP. However, in studies of ecological quality changes in mining areas, human activities—particularly coal mining—have a more pronounced impact on the ecological quality of these regions. Therefore, when investigating the potential influencing factors of ecological quality changes, it is essential to consider the overall context of Shanxi Province. A systematic study of Shanxi Province can enhance the understanding of the impacts of coal mining and provide effective strategies for ecological restoration in mining areas, ultimately achieving more efficient results.

Table 4 Significance test results of the main drivers of RSEI.
Full size table

Results of driver regression based on catboost and GWR

Table 5 presents the average results of the CatBoost model using 5-fold cross-validation. The model exhibited a high coefficient of determination (R2 > 0.670) and a low root mean square error (RMSE < 0.083), highlighting CatBoost’s ability to efficiently and accurately predict RSEI evaluation metrics. These results confirm the model’s effectiveness in handling the complex relationships between ecological factors and RSEI.

Table 5 Evaluation results of catboost regression model.
Full size table

The analysis of the CatBoost model’s prediction results using the SHAP tool (Fig. 11) quantitatively assessed the contribution of each feature to the model’s outputs. While traditional models can indicate which features are important, they often fail to explain how these features influence the predictions. The key advantage of the SHAP approach lies in its ability to reveal the impact of each feature on a per-sample basis, including whether the effect is positive or negative. The results show that population density exerts a negative influence on RSEI, whereas the other variables have positive effects. Moreover, the analysis highlights the varying degrees to which different features impact RSEI: NPP has a substantial effect, with lower NPP values resulting in smaller model outputs, while higher NPP values lead to larger outputs. Parameters such as Pa, elevation, wind speed, population density, and vapor pressure have moderate impacts on RSEI, whereas slope has a relatively minor effect.

Fig. 11
figure 11

Scatter plot of SHAP feature density.

Full size image

To capture the temporal variation and spatial heterogeneity in the effects of various factors on RSEI, both GWR and OLS models were constructed, and the fitted parameter results for the years 2000, 2005, 2010, 2015, and 2020 were compared and a map of localized R2 coefficients (Table 6). The results indicated that the GWR model significantly outperformed both the OLS and CatBoost models in terms of RMSE and R2 metrics. Furthermore, the AICc value for the GWR model was notably lower than that of the OLS model, suggesting that the GWR model can more accurately fit the data when addressing spatially heterogeneous data. By comparing the global R2 and local R2 coefficients (Fig. 12), while the global R2 is usually higher, indicating a better overall model fit, the variation in the local R2 reveals differences in fit across different regions or sample points. This is crucial for understanding the spatial heterogeneity of the data. This reflects its superior sensitivity and adaptability to spatial variations, further validating its unique advantage in predicting RSEI within the context of spatial correlation.

Fig. 12
figure 12

Local R2 Coefficient Map. (This image was created using ArcGIS 10.8 software, which is an Esri product, generated using our dataset, and is publicly available at the URL (https://www.esri.com/).)

Full size image
Table 6 Model evaluation results of OLS and GWR.
Full size table

To examine the spatial and temporal differences in the coefficients between independent and dependent variables in the GWR model, the average coefficients for different years are presented in Table 7. By analyzing these average coefficients, we can reveal the temporal and spatial trends of various environmental factors in relation to RSEI. The positive and negative values of the coefficients indicate the nature of the influence that each driving factor has on RSEI. In all years, elevation, Ws, Vp, and NPP showed a positive correlation with RSEI, while population density exhibited a negative correlation. The absolute values of the coefficients reflect the magnitude of the influence each driving factor has on RSEI, indicating that NPP and population density have a significant impact, whereas slope has a relatively minor effect.

Table 7 Mean values of regression coefficients of independent variables in the GWR model.
Full size table

The spatial distribution of the mean regression coefficients for various factors across different years is depicted in Fig. 13. It is evident from the figure that the relationships between elevation, slope, Ws, Pa, Vp, NPP, and population density with RSEI exhibit both temporal and spatial variations. The levels of influence these factors have on RSEI and their corresponding areas of impact differ, with the distribution of regression coefficients not appearing random. Notably, the coefficients for NPP and population density demonstrate a pronounced clustering effect over large areas, indicating that the influence of environmental factors on RSEI is related to spatial location. The GWR model further confirms that the impact of different factors on the ecological environment varies spatially, while within certain ranges, the same influencing factors consistently affect RSEI.

Fig. 13
figure 13

Spatial distribution of regression coefficients of independent variables of GWR model. (This image was created using ArcGIS 10.8 software, which is an Esri product, generated using our dataset, and is publicly available at the URL (https://www.esri.com/).)

Full size image

Discussion

Changes in ecological environment quality

Based on long-term RSEI monitoring (2000–2020), the ecological quality of Shanxi Province showed a significant upward trend (38% increase in mean RSEI), with particularly prominent improvements in planned mining areas. This conclusion aligns with studies by Zhao et al. (2019)38, Gong et al.39, and Zhang et al.40, all indicating that ecological restoration policies—including mine rehabilitation, soil erosion control, and afforestation—have driven quality improvements in Shanxi. For example, Gong et al. reported that post-2000 reclamation projects increased vegetation coverage by approximately 21% in mining areas, consistent with the sustained RSEI improvement observed in this study (87.08% of mining areas showed upward trends)39. However, Xu et al. emphasized a 5–8-year lag in ecological recovery within mining areas, contrasting with the long-term positive trends identified here41. This discrepancy may stem from differences in study duration and methodology: Xu et al. relied on short-term observations (< 10 years), reflecting initial restoration fluctuations, whereas this study captured cumulative recovery effects through 20-year temporal analysis.

To forecast future RSEI trends, the Hurst exponent was employed, yielding values between 0.14 and 0.89 (mean: 0.46) for Shanxi (Fig. 14a), indicating a likelihood of trend reversals. The overlay analysis of RSEI spatial trends and Hurst index (Fig. 14b) demonstrates that approximately 63.12% of Shanxi Province is likely to experience “reverse rise” (early degradation followed by future improvement) according to statistical results (Table 8), a phenomenon consistent with the “delayed recovery” pattern observed in global mining restoration studies42. Despite the overall upward trend in ecological quality, localized degradation remains evident in areas of intensive mining, such as Pingshuo-Shuonan (18.43% showing a declining trend) and Jincheng (19.73%), consistent with findings by Li et al. in northern China30. Validation using RSEI data from 2021 to 2023 showed over 87% consistency with Hurst-based predictions, supporting the reliability of this method and aligning with similar applications by Kang et al.43. on the Loess Plateau. These findings highlight the importance of balancing coal extraction with ecological management to ensure sustained long-term environmental quality.

Fig. 14
figure 14

Future RSEI Changes in Shanxi Province. (This image was created using ArcGIS 10.8 software, which is an Esri product, generated using our dataset, and is publicly available at the URL (https://www.esri.com/).)

Full size image
Table 8 Changes in RSEI sustainability grading for Shanxi Province and its planned mining areas.
Full size table

Figure 14c illustrates the RSEI mean distribution from 2021 to 2023, using the average RSEI values from these three years to validate the accuracy of future trend changes against the averages from 2018 to 2020. This method visually represents the variations between the two time periods and effectively mitigates the influence of short-term fluctuations in long-term trend analysis. The results, as shown in Table 9, reveal that RSEI values for poor, fair, and moderate grades declined by 0.39%, 2.35%, and 0.63%, respectively, while RSEI for good grades increased by 3.99% and for excellent grades decreased by 0.65%. Overall, the RSEI exhibits an upward trend, consistent with the predictions made by the Hurst index.

Table 9 Change in average RSEI, 2021–2023 vs. average RSEI, 2018–2020.
Full size table

Driving factors of ecological quality

As a major coal-producing province, Shanxi’s ecological environment is highly sensitive to human activities and vegetation dynamics. The CatBoost model identified NPP as the primary driver of RSEI improvement (32.7% contribution), consistent with Gou et al.’s findings in the Loess Plateau44, which highlight vegetation restoration as critical in arid/semi-arid regions. However, this study found that the negative effect of population density on RSEI is more significan than that reported by Xu et al.45. in the Qinghai-Tibet–Loess Plateau transition zone (with a difference in SHAP values reaching 1.8 times). This divergence reflects Shanxi’s unique resource-dependent context: intensive mining and topographic constraints (e.g., > 70% mountainous terrain) likely amplify population pressure on ecosystems46.

The combination of the CatBoost model and SHAP tool effectively quantified the influence of various factors on RSEI and intuitively illustrated how each feature affects RSEI, thereby overcoming the limitations of traditional models. Although the CatBoost model demonstrated excellent predictive performance, it did not adequately account for spatial data and geographic relationships, overlooking the spatial autocorrelation present in Shanxi Province. To address the fitting errors caused by spatial autocorrelation, this study introduced the GWR model to analyze the impacts of potential ecological factors on RSEI in Shanxi Province. The GWR model not only confirmed the magnitude and direction of each factor’s influence on RSEI but also clearly revealed the spatial heterogeneity of regression coefficients through the spatial distribution maps of GWR coefficients (Fig. 13). The results of the GWR model further confirmed that the effects of ecological factors on RSEI vary across different regions. Moreover, the GWR results showed that the NPP coefficients were relatively low in the southern part of Shanxi Province and the Lüliang Mountain area. This is mainly because the southern region of Shanxi has experienced a faster pace of urbanization and has relatively lower vegetation coverage, making NPP primarily influenced by vegetation cover and topography40. This result is consistent with actual conditions and verifies the relationship between NPP and RSEI. The combined findings from the CatBoost and GWR models can assist policymakers in formulating targeted policy measures that take into account the specific needs and conditions of different regions within Shanxi Province, thereby promoting more effective ecological management and restoration efforts.

Limitations and future prospects

This study utilized remote sensing data to evaluate ecological quality trends and drivers in Shanxi, providing a scientific basis for environmental assessment. However, compared to high-precision monitoring studies (e.g47,48), limitations exist. First, reliance on MODIS imagery—despite its high temporal resolution—limited spatial detail, potentially underestimating micro-scale changes in mining areas. Second, in future studies, we plan to introduce spatial downscaling models and integrate high-resolution data such as Landsat to improve the representation of localized regions. While this investigation considered seven primary influencing factors (elevation, slope, Ws, Pa, Vp, NPP, and population density), ecological quality is fundamentally determined by a more complex interplay of numerous elements. Future studies would benefit from incorporating additional scientifically relevant variables, such as GDP, industrial activity indicators, and energy consumption patterns40, to enable a more comprehensive evaluation of their synergistic effects on ecosystem health. As a major coal resource province, incorporating coal mining resource data into ecological evaluation metrics can provide new perspectives. Specifically, the study can reveal the relationship between mining activities and ecological quality by assessing the direct impacts of coal mining on the surrounding ecological environment, such as changes in vegetation coverage and increased soil erosion. Furthermore, by examining the intensity and distribution characteristics of coal mining resources, a more precise analysis of their long-term impacts on regional ecological environments can be achieved, particularly in the context of ecological restoration and management, evaluating the balance between mining and ecological recovery.

Conclusion

This study systematically evaluated the spatiotemporal changes in ecological environment quality and their driving mechanisms in Shanxi Province and its planned mining areas from 2000 to 2020 by constructing a RSEI, integrated with trend analysis, spatial autocorrelation, and machine learning models. The results demonstrate a significant improvement in the ecological quality of Shanxi Province, with the mean RSEI increasing from 0.42 in 2000 to 0.58 in 2020, representing a 38% enhancement—significantly higher than the national average vegetation restoration rate during the same period. Notably, 88.3% of the study area exhibited sustained ecological improvement, while only 8.49% showed degradation, highlighting the province’s success in balancing resource exploitation and ecological protection. The planned mining areas also exhibited significant ecological recovery, with 87.08% of the regions showing upward trends, thereby validating the effectiveness of ecological restoration measures implemented in these zones. To further assess the persistence of ecological quality changes, this study integrated Hurst index analysis with validation using recent observational data (2021–2023), confirming a high level of consistency (> 87%) with projected improvement trends. Spatial autocorrelation analysis, as indicated by a Global Moran’s I exceeding 0.7, verified the presence of strong spatial clustering in ecological quality, thereby supporting subsequent spatially explicit analyses of driving factors. In examining the underlying mechanisms, the CatBoost model identified NPP as the most influential factor contributing to RSEI improvements (contribution rate: 32.7%). In contrast, population density showed a significant negative correlation with RSEI, highlighting the persistent anthropogenic pressure on regional ecosystems. GWR further revealed significant spatial heterogeneity in the ecological effects of different drivers, offering scientific guidance for region-specific governance strategies. The proposed framework—integrating multi-source data with machine learning-driven interpretation in this study is not only applicable to ecological monitoring and management in resource-dependent regions but can also be extended to sustainability assessments of similar mining areas globally, such as Australia’s Bowen Basin and Inner Mongolia’s Shendong mining area. These findings provide quantitative support for optimizing the synergy between resource development and ecological conservation, aiding in the formulation of targeted ecological restoration policies.