Quantitative Analysis of Vegetation Dynamics and Driving Factors in the Shendong Mining Area under the Background of Coal Mining

Abstract

Elucidating the response mechanism of vegetation change trends is of great value for environmental resource management, especially in coal mining areas where climate fluctuations and human activities are intense. Taking the Shendong mining area as an example, based on the Google Earth Engine cloud platform, this study used the kernel Normalized Vegetation Index (kNDVI) to study the spatiotemporal change characteristics of vegetation cover during 1994–2022. Then, it carried out an attribution analysis through the partial derivative analysis method to explore the driving mechanism behind vegetation greening. The results showed that (1) the growth rate of vegetation cover change from 1994 to 2022 was 0.0052/a. The area with an upward trend of kNDVI accounted for 94.11% of the total area of the study area. The greening effect was obvious, and the kNDVI change would continue to rise. (2) Under the scenario of regional climate warming and humidifying, kNDVI responds slightly differently to different climatic factors, and kNDVI is positively correlated with temperature and precipitation in 85.20% of the mining area. The average contribution of precipitation, temperature, and human activities to kNDVI change in the Shendong mining area were 0.00094/a, 0.00066/a, and 0.0036/a, respectively. The relative contribution rates of human activities and climate change were 69.23% and 30.77%, respectively. Thus, human activities are the main driving factor for the changing of vegetation cover in this mining area, and climate change is the secondary driving factor. (3) The dynamic change in land use presents an increase in forest area under the ecological restoration project. The results of this study can provide a scientific basis for the future ecological construction of the Shendong mining area and help in the realization of regional green sustainable development goals.

1. Introduction

Coal is an essential primary energy source that plays a critical role on a worldwide scale. It is responsible for providing 37% of the world’s electricity and over 70% of its steel production [1,2]. Coal mining activities frequently cause significant harm to natural ecosystems and considerable land degradation globally [3,4,5]. To reconcile the inherent conflict between coal production and ecological preservation, China has established a range of measures designed to alleviate the negative impacts of mining activities. The implementation of land reclamation, ecological restoration, and afforestation techniques has made a substantial contribution to enhancing the overall ecological conditions in mining areas [6,7]. However, appropriately evaluating and assigning ecological changes and restoration activities in these areas continues to be a significant obstacle.

Vegetation plays a crucial role in the terrestrial ecosystem, regulating the exchange of matter and energy between the land and the atmosphere [8,9]. It is an important indicator of the change in the ecological environment, sensitive to the comprehensive impact of climate change and human activities, and it is an important parameter reflecting the dynamic change in the ecological environment [10,11]. Temperature and precipitation are the main factors affecting vegetation growth [12,13] and promote the overall greening of global vegetation. Recent studies have shown that global warming and increased precipitation have led to significant changes in vegetation cover, most notably in the high latitudes of the Northern Hemisphere [14,15]. In China, different climatic factors have different impacts on vegetation, and the combined effects of temperature and precipitation show strong spatial heterogeneity [16]. For example, in wet and cold regions, temperature is the main constraint factor for vegetation growth [17]. In arid and semi-arid areas, precipitation dominates vegetation growth, and there is a significant positive correlation between rainfall and vegetation growth [18,19]. A number of studies have revealed the effects of temperature and precipitation changes on vegetation dynamics. However, there are still deficiencies in quantifying the specific effects of temperature and precipitation on vegetation change. In addition, human activities, including the implementation of ecological restoration projects and grazing, have also contributed to changes in vegetation cover [20,21]. Previous studies have shown that human activities also have an important impact on vegetation dynamics, which can be either positive or negative [22,23]. In recent decades, human activities in the Shendong mining area have been dominated by mining [24]. With the development of the economy, the local ecological environment in the Shendong Mining area has further deteriorated due to large-scale coal mining [25,26]. To stop the degradation of the local ecological environment, the Chinese government has initiated a series of ecological restoration projects in the Shendong mining area. Since 1994, a combination of engineering and biological measures has been adopted. Comprehensive soil and water conservation management should be carried out in small watersheds around the mining area [27]. Since implementing these, the ecological environment of the mining area has been restored by afforestation and protection of local natural forests. In recent decades, ecological engineering projects have not only greatly promoted vegetation restoration in the semi-arid and sub-humid areas of northern China [28] but also played a huge role in the vegetation restoration of mining areas. However, the impact of human activities on the change in vegetation cover in mining areas is still unknown. Therefore, a clear understanding of the impact of climate change and human activities on vegetation change can effectively support the formulation of reasonable ecological environment restoration policies.

The monitoring and attribution of vegetation change have long attracted much attention. At present, the most common method for monitoring vegetation change is to use remote sensing technology combined with a vegetation index to characterize the dynamics and growth status of ground vegetation [29,30]. The normalized difference vegetation index (NDVI) is the most widely used [31], and research on this index has achieved certain results in characterizing the spatiotemporal changes of vegetation cover in China and even the world [32,33]. However, NDVI has the limitation that it is nonlinear and saturated with green biomass and also has errors in processing atmospheric noise, soil background, and saturation [34]. Some scholars have tried to make use of information from other bands to make up for the existing problems in NDVI, but the problem has not been well solved [35,36]. Therefore, Camps Valls proposed the kernel Normalized Differential Vegetation Index (kNDVI) in 2021, which is a new vegetation index proposed by applying the kernel method theory to the extraction and calculation of NDVI based on the advantages of machine learning principles [37]. The sensitivity of the index to vegetation change was improved by introducing a correction factor. It is generally more suitable for monitoring vegetation and biomass changes than EVI, especially in areas with low vegetation cover or complex terrain, where kNDVI can provide more accurate information. To clarify the extent of the impacts of climate change and human activities on vegetation change, unlike some studies using residual analysis, a partial derivative method can quantify the respective contributions of each component. At present, the partial derivative method is widely used to study the response of hydrological dynamics and vegetation development to impact factors [38,39,40]. Taking the Yangtze River Basin as an example, the partial derivative method is used to conclude that human activities are the most significant driving factor affecting vegetation distribution [22]. There are various methods used to analyze the attribution of climate change and human activities to vegetation, such as statistical methods, partial correlation methods, multiple regression residual analysis methods, and the combination of statistical methods and partial correlation analysis methods [41,42,43], which can explore the factors affecting vegetation change [44,45]. In addition, multiple regression residual analysis has been used in many studies to analyze the influence of vegetation dynamics and has achieved certain results. However, so far, the residual analysis method usually only distinguishes various driving factors and cannot accurately explore the degree of impact of climatic and non-climatic driving factors on vegetation change [46,47].

The Shendong mining area, located in China, is positioned in the transitional region between the Mu Us Desert and the Loess Plateau. It is one of the country’s 13 significant coal bases. This region is characterized by severe natural geographical conditions, such as intense winds, sandstorms, and scarce rainfall. It symbolizes a fragile ecological equilibrium between farming and animal husbandry. As a result, it demonstrates intrinsic vulnerability in terms of the structure and function of the ecosystem [17]. Moreover, the considerable coal mining operations in this region have unavoidably caused significant harm to its ecological environment [26,48]. In the Shendong mining area, there have been notable alterations in vegetation cover in recent decades due to the implementation of the “mining while controlling” concept. Studies suggest that these alterations are impacted by multiple variables, including severe weather phenomena (such as temperature fluctuations and precipitation patterns) and anthropogenic activities like mining operations and ecological restoration initiatives [49,50]. Hence, investigating the dynamic changes in plant cover is crucial for comprehending the long-term patterns in environmental conditions within protected areas and developing complete management approaches for ecological habitats in mining zones.

Essentially, this work employed Landsat remote sensing photos and meteorological data to create a chronological record of alterations in vegetation cover using the GEE cloud platform. The study conducted a quantitative analysis of the spatial and temporal aspects of changes in plant cover in the Shendong mining area between 1994 and 2022. Moreover, meteorological data on temperature and precipitation were included to show the spatial heterogeneity and association between vegetation dynamics, climate change, and human activities. The method of partial derivatives was used to determine the rate at which each important element contributes to changes in vegetation. The research findings deepen our comprehension of the diverse driving forces that impact vegetation change and offer useful insights for ecological restoration endeavors in mining regions.

2. Materials and Methods

2.1. Study Area

The Shendong mine region [27] is located in the interlace zone between Ordos in Inner Mongolia and north-western Shaanxi. It is the largest coal base in China and covers the northern side of the Loess Plateau and the eastern part of the transitional zone of the Mu Us Desert. The geographical coordinates span from 109.83° to 110.34° east longitude and 39.56° to 39.19° north latitude. The Shendong Coal Group operates ten mining areas in the Inner Mongolia Autonomous Region, including Ulanmulun Coal Mine, Daliuta Coal Mine, Cuncaota Coal Mine, Buertai Coal Mine, Bulianta Coal Mine, Shangwan Coal Mine, Shigetai Coal Mine, Halagou Coal Mine, and Daliuta Coal Mine (including Huojitu well). These mining areas cover a total area of approximately 900 km² (Figure 1A). The mean annual temperature is documented as 8.03 °C, with annual rainfall varying between 300 and 400 mm, predominantly falling between June and September annually. The vegetation growth season is brief and characterized by a lengthy period of dormancy. Evaporation typically surpasses precipitation by a factor of four to five, indicating a typical arid and semi-arid plateau continental climate [24]. The topography demonstrates an upward slope in the northwest direction, gradually declining towards the southeast, with elevations ranging from 1054 m to 1434 m (Figure 1B). The northern and southeastern sections are comprised of loess hills with gullies. These places are known for their undulating ridges and fractured surface topography. The primary plant species found on these surfaces are sandy xerophytes, which are well-adapted to withstand droughts and freezing temperatures. The coal seam in this mining location features consistent occurrence patterns and uncomplicated structural properties that are well-suited for mechanized mining methods, particularly those that utilize well-mining techniques and employ caving toward longwall extraction operations.

2.2. Data Source and Preprocessing

2.2.1. kNDVI Datasets

The kNDVI is a normalized vegetation index that incorporates kernel functions, which are machine learning techniques, to improve the accuracy and performance of the NDVI index. The objective of this index is to tackle the problems related to scale conversion and nonlinearity that are found in conventional NDVI computations. By introducing a mathematical function, it is able to map the data from the original feature space to a new feature space, which is easier to analyze and process, thus enabling kNDVI to provide more robust and accurate vegetation information at different scales and with nonlinear changes [37]. The study utilized Landsat T1_L2 products obtained from the United States Geological Survey (USGS) to cover the time span from 1994 to 2022. The Google Earth Engine (GEE) platform was used to acquire annual NDVI data, which had an image resolution of 30 m and a time resolution of 16 days. In order to guarantee the accuracy of the data, the amount of cloud cover throughout the local vegetation growing season (June–September) was restricted to below 20%. Additionally, pre-processing procedures such as atmospheric correction, radiometric calibration, and cloud removal were implemented (

Table 1. Sources of data used in this study.

Dataset		Type	Image Usability Analysis	Spatial Resolution/m	Time Resolution/Year	Data Source
Image data	Landsat 5 T1	Raster	114 scenes	30	1994–2011	United States Geological Survey https://www.usgs.gov/ (accessed on 2 November 2023).
	Landsat 7 T1	Raster	142 scenes	30	1999–2022	United States Geological Survey https://www.usgs.gov/ (accessed on 3 November 2023).
	Landsat 8 T1	Raster	91 scenes	30	2013–2022	United States Geological Survey https://www.usgs.gov/ (accessed on 4 November 2023).
Basic data	Temperature	Raster	/	1000	1994—2022 (monthly)	National Tibetan Plateau Data Center https://data.tpdc.ac.cn/ (accessed on 25 November 2023).
	Precipitation	Raster	/	1000	1994—2022 (monthly)	National Tibetan Plateau Data Center https://data.tpdc.ac.cn/ (accessed on 24 November 2023).
	DEM	Raster	/	30	2019—2021	China geospatial data cloud https://www.gscloud.cn/ (accessed on 26 November 2023).
	Land use type	Raster	5 scenes	30	1994, 2022	/

). According to CampsValls et al. [37]. (Equations (1)–(3)), we derived the kernel normalized differential Vegetation Index (kNDVI) by utilizing the normalized vegetation index (NDVI) on the cloud platform.

$$\mathrm{k}NDVI=tanh\left({\left(\frac{NIR-Red}{2\sigma }\right)}^2\right)=tanh\left({\left(\frac{NDVI}{2\tau }\right)}^2\right)$$

(1)

This equation incorporates a length scale parameter $\sigma$ directly proportional to the average reflectance of near-infrared light and red light in the acquired remote sensing image. When $\tau$ = 0.5, there exists an optimal trade-off between accuracy and simplicity [51]. Based on the equation $\sigma =\tau \left(NIR+Red\right)$, its derivative can be obtained as follows:

$$\frac{d\mathrm{k}NDVI}{dNDVI}=\frac{1}{{2\tau }^2}\left(1-{\mathrm{k}NDVI}^2\right)NDVI$$

(2)

$$\mathrm{k}NDVI=tanh\left({NDVI}^2\right)$$

(3)

The $\mathrm{k}NDVI$ data for a span of 29 years (1994~2022) was computed using NDVI data and was subsequently analyzed to investigate its spatiotemporal variations.

2.2.2. Driving Factor Datasets

The Shendong mining area’s annual mean temperature and precipitation meteorological datasets from 1994 to 2022 were acquired from the National Qinghai-Tibet Plateau Data Center Platform (https://data.tpdc.ac.cn/, accessed on 24 November 2023). The datasets have a spatial resolution of 1 km. Peng et al. [52] utilized Delta downscaling methodology to combine Climate Research Unit (CRU) data and WorldClim data and verified the precision and dependability of this approach by comparing it to observation data from 496 meteorological stations in China. The digital Elevation Model (DEM) data were obtained from the geospatial data cloud platform (http://www.gscloud.cn/, accessed on 26 November 2023) (Table 1). The Chinese Digital Elevation Model (DEM) dataset was trimmed according to the boundaries of the study region in order to provide an elevation distribution specific to the research area.

The ecosystem type was determined by analyzing Landsat series remote sensing imageries of mining sites from 1994 and 2022 through computer automatic classification results combined with manual judgment. The ecosystem types in the mining region were classified as farmland, forest land, grassland, water area, unused land, and built-up land, based on the land cover classification system used by the Ministry of Ecology and Environment and the actual ecosystem features. The images underwent radiometric, geometric, and atmospheric adjustment using ENVI5.3 software during the preprocessing stage. Afterward, a human–computer interactive interpretation was performed on the data, which was then merged with field investigation results and high-definition remote sensing images. This process obtained two periods of ecosystem type data, describing the ecosystem types of the Shendong mining area in 1994 and 2022 respectively.

2.3. Methods

2.3.1. Variation Trends Analysis

The linear regression method [53] was employed in this study to analyze the temporal trend of annual vegetation cover change in the Shendong mining area over a span of 29 years, on a pixel-by-pixel basis. The calculation equation is presented in Equation (4):

$$slope=\frac{n\times \sum _{i=1}^{n}i\times {\mathrm{k}NDVI}_i-\sum _{i=1}^{n}i\times \sum _{i=1}^n{\mathrm{k}NDVI}_i}{n\times \sum _{i=1}^n{i}^2-{\left(\sum _{i=1}^{n}i\right)}^2}$$

(4)

The slope in the linear regression equation represents the rate of change, while n denotes the number of years (1994–2022, $n$ = 29). A positive slope indicates an increasing trend in vegetation cover, whereas a negative slope suggests a decreasing trend.

The Coefficient of Variation ($CV$ is a statistical method used to quantify the dispersion of values in a dataset relative to the mean. It is calculated by dividing the standard deviation by the mean [54]. In this study, we computed the coefficient of variation of $\mathrm{k}NDVI$ d at a pixel level in the Shendong mining area from 1994 to 2022 to assess the long-term stability of $\mathrm{k}NDVI$ d changes. The calculation equation is presented in Equation (5):

$$CV=\frac{\delta }{\gamma }$$

(5)

where $CV$ is the $\mathrm{k}NDVI$ d coefficient of variation, $\delta$ is the $\mathrm{k}NDVI$ d standard deviation, and γ is the $\mathrm{k}NDVI$ d arithmetic mean. A lower $CV$ value corresponds to a more concentrated data distribution and less variance in $\mathrm{k}NDVI$ d between years. Conversely, a higher $CV$ value suggests a more scattered data distribution and greater variation in $\mathrm{k}NDVI$ d between years.

2.3.2. Future Vegetation Dynamics Trend Analysis

The Hurst index is a commonly used approach developed by hydrologists to measure the long-term correlation of time series based on the re-scaled range (R/S) analysis technique. Presently, it is widely employed in evaluating the dynamic stability of vegetative cover [55], and the computation process is as outlined below:

The mean sequence is defined for a given time series {$\mathrm{k}NDVI$ d ($t$), 1, 2, …, $n$}, as shown in Equation (6):

$${\overline{\mathrm{k}NDVI}}_{\left(T\right)}=\frac{1}{T}\sum _{t=1}^T{\mathrm{k}NDVI}_{\left(T\right)}\hspace{1em}T=\mathrm{1,2},\cdots ,n$$

(6)

Equation (7) is utilized to compute the cumulative deviation:

$$X_{\left(t,T\right)}=\sum _{t=1}^t\left({\mathrm{k}NDVI}_{\left(t\right)}-\overline{{\mathrm{k}NDVI}_{\left(T\right)}}\right)\hspace{1em}1\le t\le T$$

(7)

The range calculation is illustrated by Equation (8) as:

$$R_{\left(T\right)}={maxX}_{\left(t,T\right)}-{minX}_{\left(t,T\right)}\hspace{1em}T=\mathrm{1,2},\cdots ,n$$

(8)

Equation (9) is used to calculate the standard deviation as:

$$S_{\left(T\right)}={\left[\frac{1}{T}\sum _t^T{\left({\mathrm{k}NDVI}_{\left(t\right)}-{\mathrm{k}NDVI}_{\left(T\right)}\right)}^2\right]}^{\frac{1}{2}}\hspace{1em}T=\mathrm{1,2}\cdots ,n$$

(9)

Equation (10) is derived based on the aforementioned information as:

$$\frac{R_{\left(T\right)}}{S_{\left(T\right)}}\cong \frac{R}{S}$$

(10)

The presence of the Hurst phenomenon can be inferred if there is a proportional relationship between $R/S$∝TH in the sequence under analysis. The $H$ value is derived by fitting log($R/S$)$n$ = α + H × log(n), which represents the dynamic change trend of vegetation within a range of 0–1. This value can be categorized into three groups: $H$ = 0.5, >0.5, and <0.5, respectively, indicating randomness, consistency, and opposition to the kNDVI sequence from 1994 to 2022.

2.3.3. Relationships between Vegetation and Influencing Factors

A correlation analysis is meant to investigate the association between two variables. Nevertheless, in an empirical study, the inclusion of a third variable frequently introduces confounding effects that hinder the correlation coefficient from precisely reflecting the degree of a linear relationship [56]. Hence, it is imperative to utilize a partial correlation analysis in situations where there are potential factors that may influence the correlation between two variables. Equation (11) provides the calculation procedure for partial correlation analysis as:

$$r_{xy.z_1{z}_2\cdots z_g}=\frac{r_{xy.z_1{z}_2\cdots z_{g-1}}-r_{xzg.z_1{z}_2\cdots z_{g-1}}{r}_{yzg.z_1{z}_2\cdots z_{g-1}}}{\sqrt{\left(1-r_{xzg.z_1{z}_2\cdots z_{g-1}}^2\right)\left(1-r_{yzg.z_1{z}_2\cdots z_{g-1}}^2\right)}}$$

(11)

The partial correlation coefficient $r_{xy.z_1{z}_2\cdots z_g}$ represents the influence of $z_1{z}_2\cdots z_g$ on kNDVI interannual change, considering control variables x and y.

In order to gain a more comprehensive understanding of how vegetation changes are influenced by both natural forces and human activities, we utilized the partial derivative analysis approach to isolate and examine the specific effects of these driving elements in the Shendong mining area. This enabled us to ascertain the contribution of each element to the alteration of vegetation. To obtain detailed calculations, please refer to [57]. Equation (12) demonstrates the sequential steps of the calculation as:

$$\begin{array}{c}\frac{d\mathrm{k}NDVI}{dt}=\frac{\partial \mathrm{k}NDVI}{\partial Tem}\times \frac{dTem}{dt}+\frac{\partial \mathrm{k}NDVI}{\partial Pre}\times \frac{dPre}{dt}+UF\\ \\ ={\mathrm{k}NDVI}_{Tem}+{\mathrm{k}NDVI}_{Pre}+UF\end{array}$$

(12)

where ${\mathrm{k}NDVI}_{Tem}$ and ${\mathrm{k}NDVI}_{Pre}$ represent the respective contributions of temperature and precipitation to the variability of kNDVI. Meanwhile, $\mathrm{k}NDVI/dt$, $dTem/dt$, and $dPre/dt$ denote the interannual rates of change for kNDVI, Tem, and Pre with respect to time variable t. The calculation equation is presented in Equation (1). $\partial \mathrm{k}NDVI/\partial Tem$ and $\partial \mathrm{k}NDVI/\partial Pre$ correspond to the slopes of the linear regression line between kNDVI and temperature/precipitation, representing their functional relationship. UF represents the residual term in this equation accounting for other drivers influencing interannual variations in kNDVI apart from precipitation and temperature, such as anthropogenic factors or uncertainties.

3. Results

3.1. Spatiotemporal Dynamics and Vegetation Distribution Pattern

3.1.1. Temporal Variation in Vegetation Dynamics

To examine the changes in kNDVI over time in the Shendong mining area, we used the average pixel value of kNDVI from 1994 to 2022 as a typical indicator of the overall vegetation condition during those years. Figure 2a shows the interannual variations in kNDVI in the Shendong mining area from 1994 to 2022, shown by a dashed line. The kNDVI value demonstrates a periodic increase with a growth rate of 0.0052 per annum. Significant fluctuations are seen in the annual kNDVI measurements, with a peak value of 0.209 documented in 2018 and a nadir value of 0.024 detected in 2001. The Shendong mining area experiences an annual growth rate of kNDVI that varies between −0.0017 and 0.028 per year, as shown in Figure 2b and c respectively. Notably, areas with significant growth rates (at a significance level of p < 0.01) are mainly found on both the southeast and west sides of the town, which are adjacent to the mining site. In general, there has been a consistent rise in the amount of vegetation in the Shendong mining area from 1994 to 2022.

An analysis of the spatial distribution trend of kNDVI change in the Shendong mining area from 1994 to 2022 was achieved using slope analysis and the p-value significance test. The pixel-scale trend of the normalized difference vegetation index (kNDVI) was established in the Shendong mining area. As shown in Figure 2d, the results were divided into five categories and the percentage of each category’s area was computed. Regions exhibiting an increasing tendency constituted 94.11% of the total mining area, whereas regions with a notably increasing trend accounted for 90.44% (p < 0.01) (Figure 2c). The concentration of these places was predominantly in the southeastern and northwestern parts of the mining area. Out of all the pixel regions, 2.53% displayed a declining trend, with 1.13% of them showing a significant reduction (p < 0.01) (Figure 2c). The pixels were predominantly concentrated in the middle region of the entire mining area, including settlements and open pit sites.

3.1.2. Spatial Pattern of Vegetation Dynamics

Figure 3a illustrates the regional distribution patterns of the average kNDVI value during 29 years in the Shendong mining area. In order to evaluate the extent of plant growth in the mining region between 1994 and 2022, we computed the yearly mean kNDVI values. The chart illustrates that the kNDVI values span from 0 to 0.45 throughout the entire mining region, with an average value of 0.096. The southeast regions exhibit a greater extent of vegetation coverage compared to the northwest regions, which have a relatively lesser amount of vegetation. Specific mining sites, such as the Daliuta and Ulanmulun mines, have low levels of vegetation. The Shendong mining area exhibits improved vegetation covering along its eastern and western limits, which aligns with its designation as a significant implementation location for China’s ecological protection initiatives, specifically afforestation and reforestation. Human activities have played a substantial role in promoting the establishment of vegetation cover [54]. Figure 3b demonstrates a progressive increase in vegetation cover as elevation rises in the Shendong mining area. Additionally, it shows that high kNDVI values are maintained within the height range of 1175 to 1425 m.

To conduct a more thorough analysis of the evolving vegetation, we computed the variations in vegetation coverage within the Shendong mining region. We utilized the natural breakpoint approach to classify the kNDVI variation coefficients into five distinct groups, taking into account the pixel scale. Subsequently, we performed a comprehensive statistical analysis. According to the data in

Table 2. Coefficient of variation of kNDVI for the Shendong coal mine.

CVkNDVI	Fluctuation Degree	Area Percentage (%)
CVkNDVI < 0.55	Low fluctuation	23.95
0.55 ≤ CVkNDVI < 0.71	Relatively low fluctuation	47.70
0.71 ≤ CVkNDVI < 0.92	Moderate fluctuation	21.85
0.92 ≤ CVkNDVI < 1.31	Relatively high fluctuation	5.70
CVkNDVI ≥ 1.31	High fluctuation	0.80

, approximately 71.65% of the whole mining area consists of regions with low and relatively low variations, whereas areas with large and relatively large fluctuations only make up 6.5% of the entire mining area. Furthermore, as illustrated in Figure 3c, areas with low and relatively stable fluctuations are predominantly found along the periphery of the mining area. Conversely, regions with large fluctuations and relatively large fluctuations are primarily concentrated in the central part of the mining area and in proximity to urbanized zones, resulting from the conversion of farmland to forest and the impacts of urbanization. Furthermore, Figure 3d demonstrates a significant decline in the coefficient of variation at various altitudes throughout the whole mining region. Regions with greater coefficients of variation are primarily located at altitudes below 1125 m, where dense population concentration and rapid urbanization contribute to more pronounced alterations in vegetation cover.

3.2. Future Trends in Vegetation Dynamics

The Hurst average value of the normalized difference vegetation index (kNDVI) in the Shendong mining area is 0.518, as shown in Figure 4a. The regions with a Hurst value below 0.5 are mainly situated in the northwest of the mining area, namely covering the Cuncaota and Buertai mines. These regions make up 49.66% of the overall area. In contrast, areas with a Hurst value greater than 0.5 are mostly found in the Bulianta, Shangwan, and Daliuta mines located in the southeastern part of the mining area, making up 50.34% of the entire area. The overall analysis indicates that the sustainability of kNDVI within the Shendong mining area follows an aggregated distribution pattern and is characterized by instability.

This work used a superimposition method to integrate the overall change trend of kNDVI with the Hurst index, producing coupling information data to improve the clarity of demonstrating vegetation change trends and sustainability. The findings are classified into seven unique groups, as seen in Figure 4b and summarized in

Table 3. Classification of sustainability of vegetation variations.

SlopekNDVI	Z Value	H Value	Variation Types	Area Percentage (%)
≥0.001	≥1.96	>0.5	Consistent and significant improvement	53.54
0.0001–0.001	−1.96–1.96	>0.5	Consistent and slight improvement	2.25
≥0.0001	−1.96–1.96	<0.5	Inconsistent and improvement	40.02
−0.0001–0.0001	−1.96–1.96	-	Stable area	0.54
≤−0.0001	−1.96–1.96	<0.5	Inconsistent and degradation	2.37
−0.001–(−0.0001)	−1.96–1.96	>0.5	Consistent and slight degradation	1.00
<−0.001	≤−1.96	>0.5	Consistent and significant degradation	0.28

. The results indicate that a significant portion of the Shendong mining area, roughly 474.90 km² or 53.54% of the overall area, demonstrates sustainable vegetation growth. These locations are mostly located in Bulianta, Shangwan, and the Halagou coal mine, among other places. The present trajectory shows an increase in vegetation levels, but future forecasts imply a decrease mostly focused on the Cuncaota and Buertai coal mine areas, which encompass around 40.02% of the total mining area. However, field investigations have shown that the use of photovoltaic panel planting technology at the Cuncaota and Buertai coal mines, along with remote sensing monitoring methods, may not provide an accurate representation of vegetation decline in the Buertai Coal mine area. Only a small percentage (1.28%) of the mining region experiences a consistent decrease in dynamic sustainable vegetation. This drop is mostly observed in open pit mines and the nearby town areas due to human activities.

3.3. Characteristics of Climate Change and Its Partial Correlation with Vegetation Dynamics

3.3.1. Characteristics of Climate Change

The average annual temperature and total annual precipitation in the Shendong mining area were used as indications of the general climatic conditions throughout the study period. The regional distribution characteristics of climate in this area are depicted by the average temperature and precipitation recorded between 1994 and 2022. Figure 5a–c illustrates that the annual precipitation in the Shendong mining area varies between 368 and 418 mm, with an average of 386 mm. Geographically, regions with abundant rainfall are primarily situated in the southern portion of the mining area. Regarding temporal patterns, precipitation demonstrated a moderate upward tendency with periodic variations. The annual cumulative precipitation experienced a growth rate of 1.6689 mm per annum and exhibited a spatial pattern consistent with its distribution throughout the area. Figure 5d–f shows that the average annual temperature in the Shendong mining area ranged from 7 to 8.91 °C. It also indicates a consistent increase over time, with a rate of 0.0274 °C per year. Concerning spatial distribution, there is a notable disparity in temperature between the northern and southern directions, while the rates of change differ greatly along the east–west directions. In general, both precipitation and temperature display clear regional distribution patterns and fluctuations over time within the Shendong mining Area from 1994 to 2022.

3.3.2. Partial Correlation between Vegetation Dynamics and Climate Factors

The correlation between temperature and precipitation has a substantial impact on the fluctuating patterns of vegetation. A pixel-scale partial correlation analysis was performed to investigate the connection between changes in vegetation cover and climatic variables in order to analyze the link between kNDVI and temperature as well as precipitation. Figure 6a,b illustrates the partial correlation coefficients between kNDVI and temperature, as well as precipitation, from 1994 to 2022. The correlation coefficient for kNDVI and temperature was 0.377, while the correlation coefficient for kNDVI and precipitation was 0.398. The data clearly show that kNDVI has positive associations with both precipitation and temperature. However, precipitation has a stronger influence on vegetation changes compared to temperature. The data presented in Figure 6c clearly demonstrates a significant positive connection (87.81%) between kNDVI and precipitation, with temperature explaining 82.58% of the relationship. The locations with strong relationships between temperature and precipitation are mostly focused on the eastern and western portions of the mining region. Due to the arid to semi-arid climate in the Northern Hemisphere, high temperatures cause water to evaporate more quickly, resulting in a greater loss of soil moisture, which puts additional strain on vegetation [58]. Additionally, insufficient and unpredictable rainfall can be restored by an increase in precipitation, thereby supplying the necessary water resources needed for plant growth. This, in turn, supports the restoration efforts in the mining region.

3.4. Contributions of Climate Change and Human Activities to Vegetation Dynamics

3.4.1. Quantifying the Effects of Climate Change on Vegetation Dynamics

Various climate elements have a significant influence on changes in vegetation, with temperature and precipitation being the primary drivers for vegetation growth [59]. However, the partial correlation analysis mentioned above only offers a limited understanding of the possible connection between kNDVI and temperature, as well as precipitation, from a single viewpoint. It does not effectively measure the extent to which these two factors contribute to year-to-year fluctuations in kNDVI within the Shendong mining area. Thus, this study performed a partial derivative analysis to understand the impact of temperature and precipitation on kNDVI in this area. The analysis focused on interannual fluctuations in kNDVI and its relationship with climate parameters, specifically mean temperature and precipitation (Figure 7). The relationship between interannual variation trends of NDVI and other climatic parameters is clear. The partial derivative attribution results show that the rates of impact of temperature and precipitation on interannual changes in kNDVI are 0.00066/a and 0.00094/a, respectively. Precipitation has a comparatively more significant impact (18.08%) on variations in vegetation cover compared to temperature (12.69%).

Precipitation is the main factor driving the positive interannual trend of kNDVI in the Shendong mining area, accounting for 95.12% of the region. The regions most affected by precipitation are mainly situated in the southeastern and northwestern sections of the mining area, as well as along its diagonal axis. The impact of air temperature on vegetation development is predominantly noticeable in the southern and northeastern parts of the mining area, displaying notable spatial heterogeneity in contrast to the effects of precipitation. The temperature effect contributes to around 89.52% of the total area. The vegetation growth in Shendong’s arid and semi-arid zones is mostly determined by the interplay between precipitation and temperature.

3.4.2. The Contribution of Human Activities to Vegetation Dynamics

In the Shendong mining area, land use change caused by the afforestation project is the most significant factor of human activities. Based on Landsat remote sensing images of the Shendong mining area in 1994 and 2022, we obtained the land use classification of the Shendong Mining area in 1994 and 2022 by using the “supervised classification” method of ENVI5.3 software (Figure 8a,b). The changes in grassland and forest land in the Shendong mining area from 1994 to 2022 are the most significant (Figure 8c). From 1994 to 2022, the change in forest land is 301.28 km², accounting for 33.74% of the total area of the Shendong mining area. The change in grassland area was 326.70 km². The variation in both accounts for 70.32% of the total mining area. The forest land increased from 21.48 km² in 1994 to 309.31 km² in 2022, accounting for 32.98% of the total area of the mining area. The increase in forest land and ecological area is the main reason for the increase in kNDVI. On the other hand, due to the degradation of a large amount of cultivated land and grassland caused by intense mining activities, kNDVI in specific areas decreases. In addition, with the development of the regional economy and the acceleration of urbanization, the expansion of construction land occupies forest land, grassland, and cultivated land, which inhibits the restoration of vegetation.

On this basis, the relative contributions of climate change and human activities to vegetation change in the Shendong mining area were further explored through partial derivative analysis. According to the results of partial derivative calculation, the kNDVI interannual variation rate of the Shendong mining area from 1994 to 2022 is 0.0052/a. The influence rates of climate and human activities on kNDVI were 0.0016/a and 0.0036/a, respectively. On the whole, the relative contribution rate of climate change to vegetation change is 30.77%, while the relative contribution rate of human activities is 69.23%. The results show that human activity is the main factor affecting vegetation activity compared to the contribution of climate change. Meanwhile, dominated by climate change, the area with a positive effect on vegetation dynamics accounted for 94.58% (Figure 9a), mainly on the east and west sides of the mining area. The area dominated by human activities and having a positive effect on vegetation dynamics accounted for 97.23% (Figure 9b), and its spatial distribution was consistent with the area affected by climate change on vegetation dynamics. The reason for this change is that the local mining activities and the implementation of ecological restoration projects have changed the vegetation distribution in these areas. In general, the positive effect of human activities on vegetation change is much greater than the negative effect.

4. Discussion

4.1. Temporal and Spatial Characteristics of Vegetation Dynamics

The results indicate that between 1994 and 2022, there was a varying but mostly increasing pattern in kNDVI in the Shendong mining Area, with an annual growth rate of 0.0052. The recent research conducted in [60] supports the notion of a worldwide rise in vegetation leaf area between 2000 and 2017, with a specific emphasis on China and India, as highlighted by [61]. Vegetation recovery is influenced by several factors, including human land use management, climate change, CO₂ fertilization, and nitrogen deposition. Climate change and CO₂ fertilization appear to be the main catalysts for the global increase in vegetation [61]. Nevertheless, this study demonstrates that human forces have exerted a substantial beneficial influence on the vegetation trend in the Shendong mining area, surpassing the impact of natural variables, from 1994 to 2022. There has been a growing disparity in kNDVI values between various regions within the mining area. Under ideal circumstances, if a place remains untouched by human activity and possesses comparable natural conditions, there should be no substantial disparities in the process of vegetation restoration. Therefore, the noticeable variations in kNDVI seen in this study can be partially attributable to human intervention [62], thereby proving the substantial influence of human impacts on vegetation restoration in the mining region. Furthermore, starting in 1994, the Shendong mining Area has employed a combination of engineering and biological methods to effectively oversee soil and water conservation in the nearby small watersheds. Several vegetation restoration models have been created taking into account the distinct ecological circumstances seen in various regions. Due to extensive ecological restoration efforts, the vegetation cover in the Shendong mining Area has undergone a substantial improvement, rising from an initial 3% to a current level exceeding 64% [63]. Nevertheless, the considerable augmentation in vegetation cover inside the mining area cannot be exclusively ascribed to artificial ecological restoration, since natural causes also significantly contribute to the promotion of plant development [64]. Overall, the impressive resurgence of vegetation cover in the mining region can be attributed to a combination of local reforestation efforts and the influence of climate conditions.

4.2. Effects of Climate on Vegetation Dynamics Change

Human activities and climate-related variables are the main causes of alterations in the structure and composition of vegetation worldwide. Nevertheless, the Shendong mining area, which is prone to aridity and semi-aridity, has failed to adequately address the impact of human activities and climate change. This study investigates the effects of climatic change and changes in environmental policies on the geographical distribution and arrangement of vegetation cover in the Shendong mining area during the past 29 years. An analysis is conducted to examine the correlation between vegetation cover and climate change. This analysis utilizes a satellite-derived vegetation index (specifically, the kernel normalized vegetation index), as well as annual precipitation and temperature data. Partial derivative analysis provides a clearer understanding of how climate change and environmental policy impact changes in vegetation cover over time. The results indicate that there was no substantial causal connection between changes in vegetation, precipitation, and temperature during the course of the 29-year research. The study determined that precipitation and temperature collectively accounted for 30.77% of the variation in kNDVI. Significantly, there is a clear and direct relationship between the environmental policies that were put into effect during this time period and the alterations in vegetation cover. In summary, this research indicates that environmental regulations have a stronger impact on vegetation compared to meteorological conditions.

Previous studies have shown that vegetation growth in arid and semi-arid areas is more vulnerable to thermal effects (solar radiation and temperature changes) [65,66]. Due to the influence of global warming, vegetation in the Shendong mining area will show an increasing trend in the future, and water conditions will surpass thermal factors such as temperature and become the most important climate factor affecting vegetation growth [67]. This is also consistent with the fact that the partial correlation between rainfall change and vegetation dynamics is greater than that between temperature and vegetation dynamics in this study. At the same time, in terms of water limitations, the climate of semi-arid areas is characterized by water scarcity, and vegetation growth is limited by water supply [68]. Precipitation is the key factor of vegetation growth, which directly affects the type, growth status and distribution of vegetation. Therefore, the change in precipitation will directly affect the dynamic change in vegetation. In terms of temperature effects, temperature has a direct influence on the growth and metabolism of vegetation. In semi-arid areas, suitable temperatures can promote plant growth and improve vegetation cover. Too high or too low a temperature may limit the growth of vegetation and reduce the vegetation coverage [69]. Vegetation in semi-arid areas usually has some adaptability to water and temperature, but this adaptability has a certain range. Environmental changes beyond the adaptive range of vegetation will significantly affect the growth and cover of vegetation [70]. The improvement in vegetation status is conducive to soil and water conservation and can promote soil humidification. Soil surface water storage, water storage at different levels, radiation and accumulated temperature may also be important factors affecting vegetation growth in this region in the future [71,72] and need further in-depth analysis.

4.3. Effects of Human Activities on Vegetation Dynamics Change

In the Shendong mining area, the human factor is the main reason for the significant increase in kNDVI in the mining area [73]. Prior research has found vegetation dynamics to be more susceptible to alterations caused by human activities than climate change. However, it is important to recognize the intricate ways in which these variables interact with one another and the various types of land use [74,75]. The efficacy of ecological restoration policies and programs in mitigating environmental degradation and encouraging vegetation regeneration, specifically in ecologically vulnerable regions, has been widely acknowledged [76,77]. Prior studies have established that human activities, particularly afforestation initiatives, are of paramount importance in promoting the restoration of vegetation in China. For instance, Zhu et al. [78] emphasized that the regional greening trends observed in southeast China were primarily attributable to land use changes resulting from afforestation initiatives. Tree-planting campaigns have been recognized as substantial agents of vegetation cover expansion through the conversion of grasslands and farmlands to forests, thus promoting the greening of regions [15]. The effectiveness of ecological engineering practices is further substantiated by the increase in vegetation coverage, which contributes to the improvement in the regional ecological environment and vegetation restoration [10,79].

Wide-ranging tree planting is being executed on a national level as a crucial strategy to enhance ecological construction. Notably, the Shendong mining area is being designated as a pivotal region for the development of green mines. Vegetation restoration has been undertaken in the vicinity of the Shendong coal mine since its inception in 1985, in response to the severe soil erosion and frequent sandstorms that are defining features of the mining region. During the initial phases of vegetation restoration, techniques like grid sand fixation were implemented to stabilize the mobile sand. Following this, diverse vegetation restoration models were formulated in accordance with the unique site conditions observed in various regions (see Figure 10a). Economic forestry, ecological forestry, industrial–ecological integration, and sediment control models are among these. As illustrated in Figure 10b, the economic forestry and ecological forestry models account for the majority of afforestation programs between 1994 and 2022, with an increase of 112.940 km² and 90.01 km², respectively, in forested land. A restricted number of trees, shrubs, and vegetative plants are cultivated in accordance with the economic forestry model, predominantly in the Shigetai Coal Mine, Halagou Coal Mine, and Daliuta Coal Mine situated on the eastern side of the mining region. The ecological forestry restoration model is applied to the Bulianta Coal Mine, Shangwan Coal Mine, and Huojitu Coal Mine on the western side of the mining area. Its primary objective is the establishment of tall deciduous trees, including Pinus tabuliformis, Picea meyeri, and Populus simonii. The industrial–ecological integration model entails the establishment of windbreak and sand fixation ecological belts through planting fruit trees and Pinus tabuliformis and Picea meyeri in areas with favorable conditions adjacent to the photovoltaic industry, as well as forage and economic shrubs beneath the panels. The area of forested land has expanded by 61.37 km² over time, predominantly in the northern portion of the mining area encompassing the Cuncaota Coal Mine and Buertai Coal Mine, under this model. On the contrary, the sand control model, characterized by the predominant utilization of robust and rapid-growing sand-fixing shrubs like Artemisia ordosica and Salix psammophila, results in a comparatively modest expansion of forested land area (11.72 km²). Furthermore, this model is primarily implemented in the northern region of the mining area encompassing the Daliuta Coal Mine and Ulanmulun Coal Mine on the eastern side. The area of forested land in the mining region has experienced growth as a result of the application of various ecological restoration models; vegetation coverage has increased from 3% to 11% to over 64% at present.

Nonetheless, the beneficial effects of ecological restoration policies may be compromised due to the complex interplay between climate change and uninformed human operations. Hence, it is crucial to conduct thorough and additional rigorous investigation in this particular domain. Thus, in accordance with the aforementioned studies, our research confirms that anthropogenic actions exert a measurable impact on the dynamics of vegetation in the study region. Therefore, it is imperative to develop efficient strategies and policies for rehabilitation that advocate for sustainable land administration in mining regions while concurrently confronting the obstacles presented by climate change.

4.4. Limitations and Prospects

Demonstrating the precise rate at which vegetation restoration is influenced by human activities, specifically ecological planning, has consistently posed a formidable challenge. At present, statistical methods continue to be extensively utilized in approaches to determining attributive contribution rates; nevertheless, these methods are devoid of a robust physical foundation [80,81]. The present investigation employed the partial derivative analysis technique to differentiate the pixel-scale impacts of climatic and non-climatic variables on vegetation restoration. While the two climate factors chosen for this scale are the most representative, they are not absolute substitutes for all climatic variables and should be supplemented with additional climate factors. The findings at the pixel scale suggest that human activities have the greatest influence on vegetation restoration, whereas the effects of climate factors are relatively insignificant. However, as previous research has shown [61,82] that temperature and precipitation alone do not fully capture the effects of climate on vegetation change, it is imperative to utilize more efficient methodologies to quantify the relative contributions of ecological planning and climate change. Subsequent investigations ought to incorporate additional environmental variables that collectively impact the dynamics of vegetation, as well as construct an integrated framework for quantifying diverse human activities.

5. Conclusions

In this study, the kernel Normalized Vegetation Index (kNDVI) was calculated based on the GEE platform. The kNDVI data from 1994 to 2022 were used to evaluate the overall vegetation status in the Shendong mining area. Combined with time series data such as temperature and precipitation, the relative contributions of climate and human activities to vegetation change were quantified and distinguished. By analyzing the change in vegetation in the mining area, the effect of the ecological restoration project is evaluated scientifically, which can provide reference for ecological restoration in other mining areas. The results show that:

(1)

kNDVI showed a fluctuating upward trend in the last 29 years, with an increase rate of 0.0052/a. With the contribution of the ecological restoration project, vegetation kNDVI in mining areas showed a significant increasing trend in the whole region. However, according to the Hurst index analysis, the development trend of sustainable growth ability of vegetation will not continue to increase, and the spatial distribution shows that 40.02% of the area will show a downward trend in the future.

(2)

Precipitation and temperature are the key climatic factors that affect vegetation growth. The partial correlation coefficients between kNDVI and precipitation, and temperature in 29 years were 0.377, and 0.398, respectively. Compared with temperature, precipitation has a greater positive effect on kNDVI.

(3)

The results of attribution analysis show that climate change and human activities have dual effects on vegetation change, but the positive effects are dominant overall. In addition, compared with climate (30.77%), human activities (69.23%) were the main driving factors affecting vegetation change in the Shendong mining area. In arid and semi-arid areas, human activities (ecological restoration projects) have promoted the restoration of vegetation in mining areas, and afforestation has made outstanding contributions to the greening of mining areas.