Monitoring the Characteristics of Ecological Cumulative Effect Due to Mining Disturbance Utilizing Remote Sensing

: This study conducted land cover classiﬁcation and inversion analysis to estimate land surface temperature, soil moisture, speciﬁc humidity, atmospheric water vapor density, and relative humidity using remote sensing and multi-source mining data. Using 1990–2020 data from the Shendong mining area in Inner Mongolia, China, the eco-environmental evolution and the ecological cumulative effects (ECE) of mining operations were characterized and analyzed at a long-term scale. The results show that while the eco-environment was generally stable, mining activities affected the eco-environment at the initial stage (1990–2000) to a certain degree. During the rapid development stage of coal mining, the eco-environment was severely damaged, and the ECE were signiﬁcant at the temporal scale. The absolute value of the change rate of ecological parameters was increasing. Due to an increased focus on ecological restoration, starting in 2010, the environmental indicators gradually stabilized and the eco-environment improved considerably, ushering in a period of stability for coal mining activities. The absolute value of the change rate of ecological parameters became stable. Analysis of the change in eco-environmental indicators with distance and comparison to the contrast area showed the ECE characteristics from mining disturbance at the spatial scale. This study shows that remote sensing technology can be used to characterize the ECE from mining operations and analyze eco-environmental indicators, providing crucial information in support of ecological protection and restoration, particularly in coal mining areas. specific relative atmospheric water using remote sensing and multi-source and the characteristics of cumulative effect coal mining were analyzed on a long-time the comparative from


Introduction
Ecological cumulative effect (ECE) refers to the cumulative impact on the eco-environment caused by activities from the past to the future. This metric is particularly useful for understanding that, while the individual effects of some operations may be small or subtle, the combined effects could be significant [1]. Characterized by long-term lagging, accumulation, and interaction [2], the ECE consists of temporal and spatial aspects. The temporal cumulative effect refers to the accumulation generated when the time interval between two disturbances is less than that required by environmental restoration. The spatial cumulative effect refers to the accumulative phenomenon generated when the distance between two disturbances is less than that required by the attenuation [3].
Understanding the cumulative effects on the environment of policies, industries, and other economic activities is extremely important when evaluating choices and addressing the long-term effects caused by various anthropogenic endeavors. One such industry where the holistic comprehension of environmental effects is crucial is coal mining. Although coal mining has brought huge economic benefits, it also has resulted in the continuous decline in the quality of soil, water, vegetation, and atmosphere in the mining area, and has caused long-term adverse effects on the ecological environment [4][5][6]. Given its small object and long-duration features, the eco-environmental change caused by coal mining activities 2 of 28 shows typical cumulative effect characteristics, which could be reflected by the superposition of the damage degree at the temporal scale and the attenuation with distance at the spatial scale. The cumulative effect from coal mining activities has accelerated the degradation of eco-environmental quality. Therefore, analyzing the cumulative effect of coal mining would be extremely useful for ecological restoration and sustainable development.
The impact of mining and restoration activities in coal mining areas is a continuous and intense process. Environmental effects vary considerably at the different stages of mining (i.e., initial stage, intense phase, stable period, reduction phase, and closure stage), as well as before, during, and after ecological restoration. Generally speaking, there is a coupling relationship between mining and restoration activities and various ecoenvironmental factors. This means that monitoring and evaluating the eco-environment in coal mining areas require specialized observational techniques involving large-scale, high-frequency, continuous, long-term, and comprehensive observations and quantitative analyses. However, traditional approaches, such as field surveys, ground monitoring stations, and sensor networks are unable to meet these requirements.
In recent years, remote sensing has been used to monitor and evaluate the ecoenvironment in mining areas, given its ability to provide rapid synchronous observations in large areas [7][8][9][10][11]. The combination of remotely sensed images and ground data can generate land cover information and physical-chemical parameters of various ecological elements, which are essential in estimating the ECE of mining activities [12]. Numerous studies have employed and developed remote sensing techniques to measure ecological parameters in mining areas. For example, in terms of land cover, Balaniuk et al. [13] used Sentinel-2 satellite images to conduct nationwide automatic identification and classification of open-pit mines and tailings dams in Brazil, with an accuracy of more than 95%. A systematic image analysis method based on geographical objects (GEOBIA) was proposed by Nascimento et al. [14], and land-use changes of open-pit mines were quantified using high-resolution satellite images from different sensors, with an accuracy of more than 90%. In terms of ecological parameters, Wu et al. [15] adopted the BFAST algorithm to detect abrupt changes in vegetation cover from open-pit mining using Landsat time series. This algorithm can automatically detect the time of initial mining and evaluate the spatial distribution of vegetation destroyed by mining. Fu et al. [16] monitored NDVI (Normalized Difference Vegetation Index) in Xilinhot using the rate of change in greenness (RCG) and coefficient of variation (CV) as indicators. They used correlation analysis and stepwise regression to analyze the driving factors of vegetation and the impact of mining activities. Using SPOT 5/6 and WorldView-2 data, Liu et al. [17] analyzed soil moisture (SM) in the Dariuta Coal Mine through the Scaled SM Monitoring Index (S-SMMI). Cao et al. [18] analyzed land use, vegetation cover, and land surface temperature (LST) of the Jixi mining area in Heilongjiang with the Landsat 8 Operational Land Imager (OLI) data and used correlation analysis to explore the influence of land-use types, vegetation cover and coal mining activities on LST. García Millán et al. [19] detected water distribution change through multi-temporal analysis of multispectral data and extracted mine-related flood zones using high-resolution hyperspectral data. Using Hyperion hyperspectral data and Landsat data, Kayet et al. [20] conducted indirect monitoring of dust in opencast iron ore mines using eight vegetation indices. When analyzed and compared with field measurements, the remote sensing estimates were found to be reliable. In terms of cumulative effect [21][22][23][24], a pixel-based Ecosystem Service Value (ESV) time series model was proposed that quantified the ECE of the Yanzhou Coalfield over the past 30 years. However, to the best of our knowledge, there has been no study estimating the ECE of coal mining operations and analyzing its temporal and spatial dimensions using remote sensing.
Using long-term remote sensing and multi-source spatial data from 1990 to 2019, this study performed land cover classification and change detection and quantitatively estimated land surface (LST and SM) and atmospheric parameters (specific humidity, atmospheric water vapor density, and relative humidity). The characteristics of ECE due to coal mining disturbance were also analyzed temporally and spatially. atmospheric water vapor density, and relative humidity). The characteristics of ECE d to coal mining disturbance were also analyzed temporally and spatially.

Study Area
The study area is located in the Ejin Horo Banner, Inner Mongolia Autonomous R gion, and the Shenmu, Shaanxi Province, China ( Figure 1). Ejin Horo Banner is located in the southeastern part of Ordos, Inner Mongolia, Chin The southern part comprises the northern edge of the Mu Us Desert. The area has an a erage annual temperature of ~6.9 °C, and the terrain is high in the northwest and low the southeast. Having a typical temperate continental climate, the region is dry, expe ences little rainfall, and has annual precipitation less than evaporation. The vegetation mostly herbaceous and small shrubs. Ejin Horo Banner is rich in mineral resources, t third-largest coal-producing county in China, and is considered a critical national ener strategic base [25], with coal reserves of 56 billion tons and an annual output of about 2 million tons. Shenmu, found in Yulin, Shaanxi, is a transitional zone from grassland, f est, and hill to steppe and desert. The area has a typical temperate continental climate a has an annual average temperature of ~8.9 °C. Shenmu is located in the southeastern p of the Mu Us Desert and has proven coal reserves of over 50 billion tons, with stable co occurrence and excellent mining conditions [26,27].
The Shendong mining area across the Ejin Horo Banner and Shenmu is one of t most famous coal fields and mining bases with the largest proven reserves in China. A cording to the statistical data of raw coal output in Ordos and Yulin, from 1990 to 20 [28], the study area underwent three stages of development based on the intensity of co mining activities: initial stage  Ejin Horo Banner is located in the southeastern part of Ordos, Inner Mongolia, China. The southern part comprises the northern edge of the Mu Us Desert. The area has an average annual temperature of~6.9 • C, and the terrain is high in the northwest and low in the southeast. Having a typical temperate continental climate, the region is dry, experiences little rainfall, and has annual precipitation less than evaporation. The vegetation is mostly herbaceous and small shrubs. Ejin Horo Banner is rich in mineral resources, the thirdlargest coal-producing county in China, and is considered a critical national energy strategic base [25], with coal reserves of 56 billion tons and an annual output of about 200 million tons. Shenmu, found in Yulin, Shaanxi, is a transitional zone from grassland, forest, and hill to steppe and desert. The area has a typical temperate continental climate and has an annual average temperature of~8.9 • C. Shenmu is located in the southeastern part of the Mu Us Desert and has proven coal reserves of over 50 billion tons, with stable coal occurrence and excellent mining conditions [26,27].

Datasets
The Landsat data used in this paper have been processed by cloud removing, radiometric calibration, atmospheric correction, mosaicking and clipping. The time range from 1 July to 1 September during 1990-2020 was selected as the time range of the image. Images with less than 5% cloud cover were utilized. All of the steps were completed on Google Earth Engine's remote sensing cloud computing platform. In this study, Landsat images were used for retrieving the ecological parameters ( Figure 2).
In addition, ASTER-GDEM (Advanced Spaceborne Thermal Emission and Reflection Radiometer-Global Digital Elevation Model) images were used in land cover classification. FLDAS (Famine Early Warning Systems Network Land Data Assimilation System), Chinese farmland soil moisture ten-day dataset, and other soil moisture products were used for estimating soil moisture, while the Atmospheric reanalysis ERA5 dataset was used for determining the atmospheric parameters.

Methodology
The main methods used in the study are as follows: multi-source remote sensing data processing, remote sensing classification of land cover in mining areas, inversion of land In addition, ASTER-GDEM (Advanced Spaceborne Thermal Emission and Reflection Radiometer-Global Digital Elevation Model) images were used in land cover classification. FLDAS (Famine Early Warning Systems Network Land Data Assimilation System), Chinese farmland soil moisture ten-day dataset, and other soil moisture products were used for estimating soil moisture, while the Atmospheric reanalysis ERA5 dataset was used for determining the atmospheric parameters.

Methodology
The main methods used in the study are as follows: multi-source remote sensing data processing, remote sensing classification of land cover in mining areas, inversion of land surface ecological parameters, reanalysis of atmospheric parameters, and analysis of impacts on the ecological environment. The corresponding flowchart of the technology framework is presented in Figure 3.
Remote Sens. 2021, 13, 5034 5 of 28 surface ecological parameters, reanalysis of atmospheric parameters, and analysis of impacts on the ecological environment. The corresponding flowchart of the technology framework is presented in Figure 3.

Land Cover Classification
The study area was classified into six land cover types: vegetation, water area, cultivated land, bare land, mining land, and artificial land (i.e., urban and construction lands).

Land Cover Classification
The study area was classified into six land cover types: vegetation, water area, cultivated land, bare land, mining land, and artificial land (i.e., urban and construction lands). Using high spatial resolution historical images, samples for each land cover type were selected using manual interpretation and were then used for model training and result verification. The Random Forest (RF) [29] classification algorithm, an integrated learning algorithm based on multiple decision trees, was adopted to generate land cover classification for the mining areas using remote sensing images. Figure 4 presents the schematic diagram of the RF algorithm.
Remote Sens. 2021, 13, 5034 6 of 28 verification. The Random Forest (RF) [29] classification algorithm, an integrated learning algorithm based on multiple decision trees, was adopted to generate land cover classification for the mining areas using remote sensing images. Figure 4 presents the schematic diagram of the RF algorithm.

Inversion of LST
The method proposed by Xu [30] was used in the inversion analysis of the land surface temperature (LST). After acquiring the vegetation coverage using the pixel dichotomy method [31][32][33] and empirically determining the surface emissivity of different ground objects [34], LST can be obtained using the formula: where T is the temperature value at the sensor, which can be obtained by radiation calibration; λ is the center wavelength of the thermal infrared band of the sensor; ρ = 1.438 × 10 −2 mK; and ε is the emissivity of the surface.

Inversion of SM
The Temperature Vegetation Dryness Index (TVDI) [35] was used for the inversion of soil volumetric moisture (unit: m 3 /m 3 ). The Normalized Difference Vegetation Index (NDVI) is calculated by the formula: NIR R NDVI NIR+R − = (2) where NIR represents the reflectivity in the near-infrared waveband, and R represents the reflectivity in the red-light waveband.
The TVDI is determined by constructing the feature space of LST (Ts) and NDVI (Figure 5) and calculated using the following equations:

Inversion of LST
The method proposed by Xu [30] was used in the inversion analysis of the land surface temperature (LST). After acquiring the vegetation coverage using the pixel dichotomy method [31][32][33] and empirically determining the surface emissivity of different ground objects [34], LST can be obtained using the formula: where T is the temperature value at the sensor, which can be obtained by radiation calibration; λ is the center wavelength of the thermal infrared band of the sensor; ρ = 1.438 × 10 −2 mK; and ε is the emissivity of the surface.

Inversion of SM
The Temperature Vegetation Dryness Index (TVDI) [35] was used for the inversion of soil volumetric moisture (unit: m 3 /m 3 ). The Normalized Difference Vegetation Index (NDVI) is calculated by the formula: where NIR represents the reflectivity in the near-infrared waveband, and R represents the reflectivity in the red-light waveband. The TVDI is determined by constructing the feature space of LST (Ts) and NDVI ( Figure 5) and calculated using the following equations: where a 1 , b 1 , a 2 , and b 2 are the undetermined coefficients of the dry-wet edge, obtained by fitting the dry-wet edge equations using the least square method [36]. Equation (6) is obtained by substituting Equations (4) and (5) into Equation (3).
After scale conversion, the FLDAS dataset and the Chinese farmland soil moisture ten-day dataset were used to construct the remote sensing regression model of SM with the TVDI.

Atmospheric Parameter Reanalysis
The ERA5 reanalysis data provide temperature, pressure, relative humidity, and spe- After scale conversion, the FLDAS dataset and the Chinese farmland soil moisture ten-day dataset were used to construct the remote sensing regression model of SM with the TVDI.

Atmospheric Parameter Reanalysis
The ERA5 reanalysis data provide temperature, pressure, relative humidity, and specific humidity values using 37 pressure levels. Note that the geopotential height should be converted into geometric height using the formula [37]): where h is the ellipsoidal height (km), H is the geopotential height (km), ϕ is the latitude, g 0 = 9.80665 ms −2 , R e (ϕ) is the radius of curvature of the Earth at latitude ϕ, and g(ϕ) represents the gravity on the geoid. The converted meteorological parameters were then interpolated to obtain the atmospheric relative humidity and the atmospheric specific humidity at the Earth's surface. The atmospheric water vapor density ρ v can be obtained using the atmospheric equation of state: where T is the temperature in Kelvin; R v is the gas constant of water vapor equal to 461.5 J/kg; and e is the water vapor pressure, calculated from the relative humidity and temperature provided by ERA5 using the Magnus formula [38]. The water vapor humidity factors (i.e., density, relative humidity, and specific humidity) at the Earth's surface can then be obtained. Time series analysis for these factors was conducted for the different production periods, and the change rates in the different regions were fitted by the least square method.

Analysis of Land Cover Change
The land cover classification results using remote sensing are shown in Figure 6. To further explore the evolution of the land cover changes in the past 30 years, changes in each land cover type were calculated for the different time periods. The results are presented in Figure 7, and the land cover transfer matrix is presented in Table  The atmospheric water vapor density ρv can be obtained using the atmospheric equation of state: v where T is the temperature in Kelvin; Rv is the gas constant of water vapor equal to 461.5 J/kg; and e is the water vapor pressure, calculated from the relative humidity and temperature provided by ERA5 using the Magnus formula [38]. The water vapor humidity factors (i.e., density, relative humidity, and specific humidity) at the Earth's surface can then be obtained. Time series analysis for these factors was conducted for the different production periods, and the change rates in the different regions were fitted by the least square method.

Analysis of Land Cover Change
The land cover classification results using remote sensing are shown in Figure 6. To further explore the evolution of the land cover changes in the past 30 years, changes in each land cover type were calculated for the different time periods. The results are presented in Figure 7, and the land cover transfer matrix is presented in Table

. Analysis of the LST Inversion Results
The results of the LST inversion using remote sensing images (mean values from June to August in summer) are shown in Figure 8, while the mean LST values for summer at different years are presented in Figure 9. The results show the mean LST value increased annually from 1990, reaching its peak in 2005 with the value at over 40 °C. The LST decreased rapidly from 2005 to 2010, slightly increased in 2010-2015, and stabilized from 2015 to 2019. In particular, as shown in Figure 8, the LST has a more pronounced increase at the northeastern section of the central image from 1990 to 2005. This is because many coal mining areas are located in this region and have increased significantly over the years.
The LST change rates varied considerably during the 30-year study period. The increase in LST was accelerating from 1990 to 2005, which was the initial and rapid development stage of coal mining activities. The increased influence of mining activities accelerated the upsurge in LST, showing a significant ECE at the temporal scale. This was because more attention was given to ecological restoration and green mine construction, and the impacts caused by mining activities were artificially improved. When coal mining

Analysis of the LST Inversion Results
The results of the LST inversion using remote sensing images (mean values from June to August in summer) are shown in Figure 8, while the mean LST values for summer at different years are presented in Figure 9. The results show the mean LST value increased annually from 1990, reaching its peak in 2005 with the value at over 40 • C. The LST decreased rapidly from 2005 to 2010, slightly increased in 2010-2015, and stabilized from 2015 to 2019. In particular, as shown in Figure 8, the LST has a more pronounced increase at the northeastern section of the central image from 1990 to 2005. This is because many coal mining areas are located in this region and have increased significantly over the years.

Analysis of the SM Inversion Results
The inversion results using remote sensing images (mean values from June to August in summer) are shown in Figure 10 The rate of increase in mean SM in 1990-1995 was comparable to its rate of decrease in 1995-2000. Note that the rate of decline in SM decelerated during the rapid development of coal mining activities (from 2000 to 2005). The reason may be that while there may have been increased mining activities during this period, people have started to pay greater attention to ecological restoration, environmental protection, and sustainability. Since 2005, the SM has gradually increased, particularly in 2015-2019, exhibiting apparent activities became stable after 2010, the LST change rate decelerated, and the im mining activities gradually stabilized.

Analysis of the SM Inversion Results
The inversion results using remote sensing images (mean values from June to in summer) are shown in Figure 10, while the mean SM values for summer (Juneat different years are presented in Figure 11  The LST change rates varied considerably during the 30-year study period. The increase in LST was accelerating from 1990 to 2005, which was the initial and rapid development stage of coal mining activities. The increased influence of mining activities accelerated the upsurge in LST, showing a significant ECE at the temporal scale. This was because more attention was given to ecological restoration and green mine construction, and the impacts caused by mining activities were artificially improved. When coal mining activities became stable after 2010, the LST change rate decelerated, and the impact of mining activities gradually stabilized.

Analysis of the SM Inversion Results
The inversion results using remote sensing images (mean values from June to August in summer) are shown in Figure 10

Analysis of Atmospheric Parameters
ERA5 reanalysis data were used to obtain mean values for relative humidity, specific humidity, and atmospheric water vapor density (June to August). Figures 12-14 show the spatial distribution of these parameters for 1990, 2005, and 2019. Specific humidity and atmospheric water vapor density exhibited pronounced spatial variations in each interval. In terms of spatial distribution, the parameter values were higher in the southern region than in the north. The three atmospheric parameters were also found to have similar trends

Analysis of Atmospheric Parameters
ERA5 reanalysis data were used to obtain mean values for relative humidity, sp humidity, and atmospheric water vapor density (June to August).  The rate of increase in mean SM in 1990-1995 was comparable to its rate of decrease in 1995-2000. Note that the rate of decline in SM decelerated during the rapid development of coal mining activities (from 2000 to 2005). The reason may be that while there may have been increased mining activities during this period, people have started to pay greater attention to ecological restoration, environmental protection, and sustainability. Since 2005, the SM has gradually increased, particularly in 2015-2019, exhibiting apparent ecological cumulative effects related to progress in sustainable mining practices and focus on environmental protection.

Analysis of Atmospheric Parameters
ERA5 reanalysis data were used to obtain mean values for relative humidity, specific humidity, and atmospheric water vapor density (June to August). Figures 12-14 show the spatial distribution of these parameters for 1990, 2005, and 2019. Specific humidity and atmospheric water vapor density exhibited pronounced spatial variations in each interval. In terms of spatial distribution, the parameter values were higher in the southern region than in the north. The three atmospheric parameters were also found to have similar trends

Monitoring of Key Mining Area-Shangwan Coal Mine
Shangwan coal mine (see Figure 17), located in Wulanmulun Town, Ejin Ho ner, Ordos City, China, is one of the ultra-large modern backbone mines of S Shendong Coal Group Co., Ltd. [39]. The mine is a subterranean coal mine (pit dustry coal mine) with an area of 61.8 km 2 and a geological reserve of 1.23 bill [40]. It started its operation in 2000, with an approved production capacity of 14 tons per year.

Monitoring of Key Mining Area-Shangwan Coal Mine
Shangwan coal mine (see Figure 17), located in Wulanmulun Town, Ejin Ho ner, Ordos City, China, is one of the ultra-large modern backbone mines of S Shendong Coal Group Co., Ltd. [39]. The mine is a subterranean coal mine (pith dustry coal mine) with an area of 61.8 km 2 and a geological reserve of 1.23 bill [40]. It started its operation in 2000, with an approved production capacity of 14 tons per year.

Monitoring of Key Mining Area-Shangwan Coal Mine
Shangwan coal mine (see Figure 17), located in Wulanmulun Town, Ejin Horo Banner, Ordos City, China, is one of the ultra-large modern backbone mines of Shenhua Shendong Coal Group Co., Ltd. [39]. The mine is a subterranean coal mine (pithead industry coal mine) with an area of 61.8 km 2 and a geological reserve of 1.23 billion tons [40]. It started its operation in 2000, with an approved production capacity of 14 million tons per year.

Land Cover Classification of the Shangwan Coal Mine
The land cover classification results for the Shangwan coal mining area at d years are shown in Figure 18, and the tabulated results for each cover type are rized in Table 2. In 1990, almost no mining land and artificial land was foun Shangwan coal mining area; the main cover types were vegetation and bare lands. the main cover types were still vegetation and bare land. Patches of cultivated l mining land were observed, but there was still no artificial land. In 2019, most of had vegetation cover type; bare land decreased, while mining land and artificial creased significantly. In the past 30 years, changes in the land cover types in the area have been pronounced. The primary land cover type was vegetation. Bare l water had decreased considerably, while mining land and artificial land have in year by year.

Land Cover Classification of the Shangwan Coal Mine
The land cover classification results for the Shangwan coal mining area at different years are shown in Figure 18, and the tabulated results for each cover type are summarized in Table 2. In 1990, almost no mining land and artificial land was found in the Shangwan coal mining area; the main cover types were vegetation and bare lands. In 2005, the main cover types were still vegetation and bare land. Patches of cultivated land and mining land were observed, but there was still no artificial land. In 2019, most of the land had vegetation cover type; bare land decreased, while mining land and artificial land increased significantly. In the past 30 years, changes in the land cover types in the mining area have been pronounced. The primary land cover type was vegetation. Bare land and water had decreased considerably, while mining land and artificial land have increased year by year.

Land Cover Classification of the Shangwan Coal Mine
The land cover classification results for the Shangwan coal mining area at different years are shown in Figure 18, and the tabulated results for each cover type are summarized in Table 2. In 1990, almost no mining land and artificial land was found in the Shangwan coal mining area; the main cover types were vegetation and bare lands. In 2005, the main cover types were still vegetation and bare land. Patches of cultivated land and mining land were observed, but there was still no artificial land. In 2019, most of the land had vegetation cover type; bare land decreased, while mining land and artificial land increased significantly. In the past 30 years, changes in the land cover types in the mining area have been pronounced. The primary land cover type was vegetation. Bare land and water had decreased considerably, while mining land and artificial land have increased year by year.    In Figure 19, the vegetation area in the Shangwan coal mining area showed an accelerated growth trend from 1990 to 2005. The effects of mining activities on the vegetation were minimal at this time, given that it was still the initial and early stages of development for mining activities. During the period of rapid development of coal mining activities (from 2005 to 2010), vegetation showed pronounced ECE, which gradually increased due to its cumulative effects year by year. In Figure 19, the vegetation area in the Shangwan coal mining area showed an accelerated growth trend from 1990 to 2005. The effects of mining activities on the vegetation were minimal at this time, given that it was still the initial and early stages of development for mining activities. During the period of rapid development of coal mining activities (from 2005 to 2010), vegetation showed pronounced ECE, which gradually increased due to its cumulative effects year by year. There was a slow decline in vegetation from 2010 to 2019. Compared with the previous period, vegetation in the Shangwan coal mine had increased overall, while water and bare lands had downward trends. In 2000, lands classified as water were more extensive than usual, possibly due to the heavy precipitation occurring in that year. The expansion of artificial lands accelerated in 2010-2015; coal mining processing plants (screening fields) and other artificial lands were built in the eastern part of the study area, given increased mining activities in the region. Cultivated lands had increased rapidly in 1990-2000 and fluctuated in the following decades. These changes are closely related to changes in human activities.

Analysis of LST and SM
The LST inversion results for the Shangwan coal mining area are shown in Figure 20, and the mean LST summer values (June to August) at different years are presented in Figure 21. The LST exhibited an overall upward trend in 1990-2005, reaching its peak in 2005 before declining in subsequent years. As shown in Figures 11 and 21, the change trend in LST for the Shangwan coal mining area is comparable to that of the entire study area. The LST in the mining area gradually surged in 1990-2000 because the impact from mining on the surrounding environmental factors was weak at the initial stage of coal mining operations. From 2000 to 2010, the LST initially increased and then declined. To a certain extent, these changes were influenced by the cumulative effects of mining activities, showing particular ECE characteristics. The LST in the mining area increased gradually in 2010-2019, given the plateauing of coal mining activities, and the impact on the LST diminished.             Figures 26-28 show the change rates in atmospheric specific humidity, atmospheric relative humidity, and atmospheric water vapor density. The atmospheric specific humidity increased from 1990 to 2000. However, the increase rate showed a decreasing trend from the northwest to the southeast, reflecting the variations in physical geography. The atmospheric specific humidity had a downward trend for 2000-2010, and the decrease rate near the mining area was significantly higher than that in other regions. From 2010 to 2020, the atmospheric specific humidity again surged, with a large increase rate and a significant ecological cumulative effect. The spatial and temporal distribution patterns of water vapor density and relative humidity change rate were consistent with that of atmospheric specific humidity.
For 1990-2005, the SM in the study area decreased overall. SM mainly increased in 1990-1995 and then declined in the following decade. From 2005 to 2010, the SM increased considerably in the eastern part of the study area and slightly increased in other areas. From 2010 to 2015, the SM was comparable to the 2005-2010 levels, with the main difference being that in the east-central area, the SM decreased slightly. From 2015 to 2019, the SM of the entire study area had drastically changed. Figures 26-28 show the change rates in atmospheric specific humidity, atmospheric relative humidity, and atmospheric water vapor density. The atmospheric specific humidity increased from 1990 to 2000. However, the increase rate showed a decreasing trend from the northwest to the southeast, reflecting the variations in physical geography. The atmospheric specific humidity had a downward trend for 2000-2010, and the decrease rate near the mining area was significantly higher than that in other regions. From 2010 to 2020, the atmospheric specific humidity again surged, with a large increase rate and a significant ecological cumulative effect. The spatial and temporal distribution patterns of water vapor density and relative humidity change rate were consistent with that of atmospheric specific humidity. Figures 26-28 show the change rates in atmospheric specific humidity, atmospheric relative humidity, and atmospheric water vapor density. The atmospheric specific humidity increased from 1990 to 2000. However, the increase rate showed a decreasing trend from the northwest to the southeast, reflecting the variations in physical geography. The atmospheric specific humidity had a downward trend for 2000-2010, and the decrease rate near the mining area was significantly higher than that in other regions. From 2010 to 2020, the atmospheric specific humidity again surged, with a large increase rate and a significant ecological cumulative effect. The spatial and temporal distribution patterns of water vapor density and relative humidity change rate were consistent with that of atmospheric specific humidity.   The mean LST and SM change rates (in absolute value) at each time period are summarized in Figure 29. In the first four intervals (1990-1995, 1995-2000, 2000-2005, and 2005-2010), the mean LST change rate (in absolute value) showed an upward trend, indicating that the LST increase was accelerating. In 1990-2000, mining operations were still in the initial stage of development, and the succeeding decade (2000-2010) was the period for rapid development. During these years, with the gradual strengthening of mining activities, the influence of ecological factors also expanded, as indicated by the accelerated changes in LST.  The mean LST and SM change rates (in absolute value) at each time period are summarized in Figure 29. In the first four intervals (1990-1995, 1995-2000, 2000-2005, and 2005-2010), the mean LST change rate (in absolute value) showed an upward trend, indicating that the LST increase was accelerating. In 1990-2000, mining operations were still in the initial stage of development, and the succeeding decade (2000-2010) was the period for rapid development. During these years, with the gradual strengthening of mining activities, the influence of ecological factors also expanded, as indicated by the accelerated changes in LST. For 2010-2019, the values were in a downward trajectory, indicating a deceleration in LST change rates. At this period (2010-2019), coal mining activities were at the stage of steady development. Vigorous implementation of environmental protection and ecological restoration measures has gradually stabilized the ecological elements, as manifested by the decelerating changes in LST. The mean LST and SM change rates (in absolute value) at each time period are summarized in Figure 29. In the first four intervals (1990-1995, 1995-2000, 2000-2005, and 2005-2010), the mean LST change rate (in absolute value) showed an upward trend, indicating that the LST increase was accelerating. In 1990-2000, mining operations were still in the initial stage of development, and the succeeding decade (2000-2010) was the period for rapid development. During these years, with the gradual strengthening of mining activities, the influence of ecological factors also expanded, as indicated by the accelerated changes in LST. Time accumulation is the cumulative phenomenon that occurs on t when the time interval between two disturbances is less than the time requ ronmental restoration. The external manifestation is an increase in the rate the time scale. Therefore, the accelerated changes in LST and SM values sho ical cumulative effect of mining disturbance on the time scale.

Features on Spatial Scale
As shown in Figure 30, seven buffer zones were set up, with the Shangw as center and a 300 m interval between two adjacent buffers. The mean LST a for each buffer zone in 2019 were then calculated, and the summary of resul in Figure 31.

Features on Spatial Scale
As shown in Figure 30, seven buffer zones were set up, with the Shangwan Coal Mine as center and a 300 m interval between two adjacent buffers. The mean LST and SM values for each buffer zone in 2019 were then calculated, and the summary of results is presented in Figure 31. Time accumulation is the cumulative phenomenon that occurs on when the time interval between two disturbances is less than the time req ronmental restoration. The external manifestation is an increase in the rat the time scale. Therefore, the accelerated changes in LST and SM values sh ical cumulative effect of mining disturbance on the time scale.

Features on Spatial Scale
As shown in Figure 30, seven buffer zones were set up, with the Shangw as center and a 300 m interval between two adjacent buffers. The mean LST for each buffer zone in 2019 were then calculated, and the summary of resul in Figure 31.   The first buffer zone (>300 m) had the highest mean LST and the lowest average SM values. The mean LST gradually declines as the distance increases, while the mean SM exhibits a gradual upward trend. Spatial accumulation refers to the accumulative phenomenon generated on the spatial scale when the spatial proximity between adjacent disturbance factors is less than the distance required to remove each disturbance. Externally, the influence of disturbance on the spatial scale attenuates with the distance. Therefore, the variations in LST and SM with regard to distance show pronounced ECE characteristics on the spatial scale. Based on the temporal and spatial cumulative effect characteristics (Wang et al., 2010), the source of the cumulative effect may be attributed to mining disturbance by way of mining subsidence and the land-use change.
To better analyze the environmental evolution characteristics, we chose an area unaffected by human activities to compare with the Shangwan coal mining area. The contrast area is rectangular, located two kilometers from the western boundary of the Shangwan coal mining area, and is similarly sized as the study site. The high-resolution image of the contrast area is shown in Figure 32.  The first buffer zone (>300 m) had the highest mean LST and the lowest average SM values. The mean LST gradually declines as the distance increases, while the mean SM exhibits a gradual upward trend. Spatial accumulation refers to the accumulative phenomenon generated on the spatial scale when the spatial proximity between adjacent disturbance factors is less than the distance required to remove each disturbance. Externally, the influence of disturbance on the spatial scale attenuates with the distance. Therefore, the variations in LST and SM with regard to distance show pronounced ECE characteristics on the spatial scale. Based on the temporal and spatial cumulative effect characteristics (Wang et al., 2010), the source of the cumulative effect may be attributed to mining disturbance by way of mining subsidence and the land-use change.
To better analyze the environmental evolution characteristics, we chose an area unaffected by human activities to compare with the Shangwan coal mining area. The contrast area is rectangular, located two kilometers from the western boundary of the Shangwan coal mining area, and is similarly sized as the study site. The high-resolution image of the contrast area is shown in Figure 32. The first buffer zone (>300 m) had the highest mean LST and the lowest average SM values. The mean LST gradually declines as the distance increases, while the mean SM exhibits a gradual upward trend. Spatial accumulation refers to the accumulative phenomenon generated on the spatial scale when the spatial proximity between adjacent disturbance factors is less than the distance required to remove each disturbance. Externally, the influence of disturbance on the spatial scale attenuates with the distance. Therefore, the variations in LST and SM with regard to distance show pronounced ECE characteristics on the spatial scale. Based on the temporal and spatial cumulative effect characteristics (Wang et al., 2010), the source of the cumulative effect may be attributed to mining disturbance by way of mining subsidence and the land-use change.
To better analyze the environmental evolution characteristics, we chose an area unaffected by human activities to compare with the Shangwan coal mining area. The contrast area is rectangular, located two kilometers from the western boundary of the Shangwan coal mining area, and is similarly sized as the study site. The high-resolution image of the contrast area is shown in Figure 32.  The land cover results of the contrast area at different years are shown in Figure 33, and the area statistics are summarized in Table 3. There was almost no mining or artificial land in the contrast area in 1990, and the main coverage types were vegetation (79%) and bare land (18%). In 2005, vegetation cover (90%) was still the primary coverage type, while barren land decreased to 8%. In 2019, vegetation (53%) decreased considerably, while bare land disappeared. Along with vegetation, cultivated land (47%) became a major land cover type, and new artificial lands (1%) were established. Results from the cover classification analysis reveal that the land cover in the contrast area has significantly changed; in particular, large portions of vegetation and bare lands were converted into cultivated land. The land cover results of the contrast area at different years are shown in Figure 33, and the area statistics are summarized in Table 3. There was almost no mining or artificial land in the contrast area in 1990, and the main coverage types were vegetation (79%) and bare land (18%). In 2005, vegetation cover (90%) was still the primary coverage type, while barren land decreased to 8%. In 2019, vegetation (53%) decreased considerably, while bare land disappeared. Along with vegetation, cultivated land (47%) became a major land cover type, and new artificial lands (1%) were established. Results from the cover classification analysis reveal that the land cover in the contrast area has significantly changed; in particular, large portions of vegetation and bare lands were converted into cultivated land.   Figure 34 shows the various land cover changes in the Shangwan coal mining area and the contrast area for the last 30 years. The vegetation cover in these two areas exhibited a growing trend in the first 20 years but has since trended downwards in the last decade. Vegetation in the contrast area decreased further and faster than in the mining area. The change trends for water, bare land, and cultivated lands in the two areas are comparable, but cultivated lands have expanded more rapidly in the contrast area in the past five years. Both mining and artificial lands had increased in the mining zones.
The LST inversion results for the contrast area are shown in Figure 35. The overall LST increased from 1990 until its peak in 2005 and then declined. The SM values initially decreased, reached the lowest value in 2005, and then increased to the highest point in 2019.   Figure 34 shows the various land cover changes in the Shangwan coal mining area and the contrast area for the last 30 years. The vegetation cover in these two areas exhibited a growing trend in the first 20 years but has since trended downwards in the last decade. Vegetation in the contrast area decreased further and faster than in the mining area. The change trends for water, bare land, and cultivated lands in the two areas are comparable, but cultivated lands have expanded more rapidly in the contrast area in the past five years. Both mining and artificial lands had increased in the mining zones.
The LST inversion results for the contrast area are shown in Figure 35.   Figure 36 shows the LST box plot for the Shangwan coal mining area and the contrast area. Comparing Figures 20 and 35, the LST in the two regions showed a gradually increasing trend from 1990 to 2005. However, due to the emergence of mining land and artificial land in the Shangwan coal mining area in 2019, the coal mine processing plant and its surroundings yielded high LST values, while in other land cover types, the LST values were considerably lower. In the contrast area, the LST decreased overall. As shown in Figure 36, the Shangwan coal mining area had more concentrated LST values and a higher mean LST than the contrast area.  Figure 36 shows the LST box plot for the Shangwan coal mining area and the contrast area. Comparing Figures 20 and 35, the LST in the two regions showed a gradually increasing trend from 1990 to 2005. However, due to the emergence of mining land and artificial land in the Shangwan coal mining area in 2019, the coal mine processing plant and its surroundings yielded high LST values, while in other land cover types, the LST values were considerably lower. In the contrast area, the LST decreased overall. As shown in Figure 36, the Shangwan coal mining area had more concentrated LST values and a higher mean LST than the contrast area.  Figure 36 shows the LST box plot for the Shangwan coal mining area and the contrast area. Comparing Figures 20 and 35, the LST in the two regions showed a gradually increasing trend from 1990 to 2005. However, due to the emergence of mining land and artificial land in the Shangwan coal mining area in 2019, the coal mine processing plant and its surroundings yielded high LST values, while in other land cover types, the LST values were considerably lower. In the contrast area, the LST decreased overall. As shown in Figure 36, the Shangwan coal mining area had more concentrated LST values and a higher mean LST than the contrast area.
The SM inversion results for the contrast area are shown in Figure 37. The SM box plot between the Shangwan coal mining area and the contrast area is shown in Figure 38. Comparing Figures 22 and 37, the two regions had similar SM change trends, decreasing initially and then increasing. As shown in Figure 38, the distribution of the SM value for the two regions is relatively consistent and concentrated. However, the mining area has lower mean SM than the contrast area, which indicates that mining activities affect SM and that the impact is highly concentrated. The SM inversion results for the contrast area are shown in Figure 37. The SM box plot between the Shangwan coal mining area and the contrast area is shown in Figure 38. Comparing Figures 22 and 37, the two regions had similar SM change trends, decreasing initially and then increasing. As shown in Figure 38, the distribution of the SM value for the two regions is relatively consistent and concentrated. However, the mining area has lower mean SM than the contrast area, which indicates that mining activities affect SM and that the impact is highly concentrated.   The SM inversion results for the contrast area are shown in Figure 37. The SM box plot between the Shangwan coal mining area and the contrast area is shown in Figure 38. Comparing Figures 22 and 37, the two regions had similar SM change trends, decreasing initially and then increasing. As shown in Figure 38, the distribution of the SM value for the two regions is relatively consistent and concentrated. However, the mining area has lower mean SM than the contrast area, which indicates that mining activities affect SM and that the impact is highly concentrated.

Conclusions
Land cover classification and change analysis were carried out to retrieve land surface temperature, soil moisture, specific humidity, relative humidity, and atmospheric water vapor density using remote sensing and multi-source spatial data. The eco-environment change and the characteristics of ecological cumulative effect from coal mining operations were analyzed on a long-time scale. The Shangwan Coal mine was used as example in the comparative analysis between the mining zone and the contrast area. The main

Conclusions
Land cover classification and change analysis were carried out to retrieve land surface temperature, soil moisture, specific humidity, relative humidity, and atmospheric water vapor density using remote sensing and multi-source spatial data. The eco-environment change and the characteristics of ecological cumulative effect from coal mining operations were analyzed on a long-time scale. The Shangwan Coal mine was used as example in the comparative analysis between the mining zone and the contrast area. The main highlights of the study are as follows: (1) In the initial stage of coal mining activities (1990)(1991)(1992)(1993)(1994)(1995)(1996)(1997)(1998)(1999)(2000), the eco-environment was generally stable, but mining activities had impacted the eco-environment to a certain degree. (2) During the rapid development stage of coal mining operations, the eco-environment was damaged severely, showing significant ECE, including temporal and spatial cumulative effects. The change rates for the eco-environmental parameters accelerated and showed pronounced ecological cumulative effects at the temporal scale. In 2010, coal mining activities entered a period of relative stability, and ecological restoration started to receive greater attention. The eco-environmental parameters gradually recovered, and the eco-environment generally improved. Results from the land surface temperature and soil moisture analyses and the spatial comparison with the contrast area show ECE characteristics due to mining disturbance at the spatial scale.
Future studies can comprehensively utilize remote sensing, ground investigation, and statistical data to conduct quantitative and high-frequency observations for more parameters at long-term scales in mining areas. The results would help support understanding the mechanisms and characteristics of the ecological cumulative effects caused by mining operations.