Tempo-Spatial Variation of Vegetation Coverage and Influencing Factors of Large-Scale Mining Areas in Eastern Inner Mongolia, China

Vegetation in eastern Inner Mongolia grasslands plays an important role in preventing desertification, but mineral exploration has negative effects on the vegetation of these regions. In this study, the changing trend types of vegetation in eastern Inner Mongolia were analyzed using the normalized difference vegetation index (NDVI) time series from the Global Inventory Modeling and Mapping Studies (GIMMS) NDVI 3g dataset from 1982 to 2015. Meanwhile, changing trend and influencing factors of 25 large-scale mining areas before and after mining were explored with the methods of trend line, residual calculation, and correlation analysis. The vegetation coverage towards increasing in eastern Inner Mongolia decreased in the order of Tongliao > Hinggan League > Chifeng > Hulunbuir > Xilingol over the past 34 years. Vegetation showed a decreasing tendency in 40% mining areas, but an increasing tendency in 60% mining areas after mining. Vegetation change in Shengli No. 1 had a significant correlation with precipitation and human activities after mining. Except Shengli No. 1, an obvious correlation was found between vegetation change and precipitation in 45.83% mining areas after mining. Human activities had significant positive effects on vegetation growth in 25% mining areas. Significant negative effects of human activities were found in 8.34% mining areas, causing the vegetation degradation. However, there were 20.83% mining areas with vegetation changes not affected by precipitation and human activities.


Introduction
Grasslands in the world are mainly located in Eurasian Steppe, North American Steppe, and South American Steppe [1]. Meanwhile, there are a lot of coal resources in these regions. For instance, coal reserves are estimated at 122.4 million tons in the Powder River Basin, Wyoming [2]. In China, the proved coal reserves are more than 800 million tons, forming two large coal-electricity bases in eastern Inner Mongolia [3]. However, the exploitation and utilization of mineral resources have changed the material cycle and energy flow of mining area ecosystem, resulting in serious vegetation degradation and environmental pollution [4]. Therefore, there has been increasing attention given to explore vegetation change in grassland mining areas [5,6].
Grassland vegetation plays a pivotal role in the Earth's material and energy exchange, which is the most sensitive part of ecosystems to climate change in arid and semi-arid environments [7].

Study Area
The regions of eastern Inner Mongolia include Hulunbuir, Xilingol, Chifeng, Tongliao, and Hinggan League (latitude 111.2 • -126 • N, longitude 41.3 • -53.3 • E) ( Figure 1) [24]. There is a semi-arid and arid climate, and precipitation is in the range of 200-400 mm p.a. The tempo-spatial distribution of precipitation is extremely uneven, mostly concentrated in summer, and it is easy to form floods and soil erosion. The area of grassland and forest are 1.98 × 10 7 and 3.43 × 10 7 hm 2 , covering approximately 30.31% and 50.66% of eastern Inner Mongolia, respectively [25]. The notice on adjusting the scale standards for production and construction of some mines (2004, No. 208 [44] Open-pit Holingola, Tongliao 1800 1981 Zhahanao'er [45] Open-pit Jarud Banner, Tongliao 1800 1999 Jinyuanli [46] Underground mining Holingola, Tongliao 120 2008 Yuanbaoshan [47] Open This study used the Global Inventory Modeling and Mapping Studies (GIMMS) NDVI 3g. v1 (third generation version 1) bimonthly products with spatial resolution of 8 km × 8 km [51]. They were downloaded from https://ecocast.arc.nasa.gov/data/pub/gimms/3g.v1/00FILE-LIST.txt and covered a period from July 1981 to December 2015. These NDVI datasets have been corrected for calibration, viewing geometry, volcanic aerosols, and other effects that are not related to vegetation variation. These data contain global geographical projections (Geographic, WGS 1984). The GIMMS products are 16-day maximum value composite (MVC) bimonthly global NDVI product generated from AVHRR data. The datasets are in NetCDF format. We translated these datasets into GeoTIFF format using MatlabR2014a. NDVI datasets from 1982 to 2015 were selected in this study.

Vector Boundary Data
The city boundaries were from the 1:100 million vector map data of the National Catalogue Service for Geographic Information downloaded from http://www.webmap.cn/commres.do?method= result100W. The mining area boundary was obtained by its inflection points, geographic coordinates, and Google Earth images, which showed in the references of . The buffer boundary (e.g., 10, 20, 30, 40, and 50 km) was completed based on their mining area boundary by buffer analysis of ArcGIS software. GIMMS NDVI 3g data were clipped using city and mining area boundaries.

Climate Observation Data
Daily average temperature ( • C) and rainfall (mm) data from July 1982 to December 2015 were provided by the National Oceanic and Atmospheric Administration (NOAA) downloaded from https://gis.ncdc.noaa.gov/maps/ncei/cdo/daily. There are 16 monitoring sites in eastern Inner Mongolia, consisting of three in Hulunbuir, seven in Xilingol, three in Chifeng, two in Tongliao, and one in Hinggan League. The annual average temperature and precipitation of growing season at each monitoring site from 1982 to 2015 were calculated by its daily average temperature and rainfall from April to October each year. The space distribution of temperature and precipitation each year in eastern Inner Mongolia was realized by Kriging interpolation of ArcGIS software.

Maximum Value Composites
Considering the demand of vegetation growth for temperature and precipitation, we selected April to October per year as the vegetation growing season. The Maximum Value Composite (MVC) method was used to obtain the max monthly NDVI (MNDVI). In addition, the max NDVI of growing season each year (GNDVI) was calculated based on the MNDVI from April to October [22]. In this study, GNDVI was used to reflect the vegetation coverage. MNDVI and GNDVI were computed using the following equations: where i is the series number of year (i = 1, 2, · · · , 34), j is the series number of month ( j = 1, 2, · · · , 12), MNDVI ij is the max NDVI for the jth month of the ith year, NDVI ij1 is max NDVI for the first half of the jth month of the ith year, NDVI ij2 is max NDVI for the second half of the jth month of the ith year, and GNDVI i is the max NDVI for the ith year.

Trend Line Analysis
We adopted the ordinary least-squares (OLS) approach to determine the changing trend of GNDVI, precipitation, temperature, and residual from 1982 to 2015 [22]. The slope was calculated using the following equations: where n represents the total number of years (n =34), i is the series number of year (i = 1, 2, · · · , 34), and x i is the value (GNDVI, precipitation, temperature, and residual) of year i. A slope greater than 0 indicates an increasing trend, and one less than 0 shows a decreasing trend. The significance degree of changing trend is determined by the correlation coefficient r. r was calculated using the following equations: r greater than 0 indicates an increasing trend, and r less than 0 shows a decreasing trend. In addition, the significance level of change trend was assessed using the t-test, and the p-value was adopted to reflect the significance level. The changing types and significance levels are shown in Table 2 [52]. t was calculated using the following equations:

Correlation Analysis
Correlation coefficients were calculated between mining area and its buffer zones for vegetation variation. In addition, we applied correlation analysis to explore influencing factors of vegetation variation. Pearson correlation coefficients were calculated among the GNDVI and precipitation and temperature from 1982 to 2015.

Residual Analysis
We adopted pixel-based residual trend analysis to distinguish the human-induced GNDVI changes from the changes induced by climate factors. We defined the residual as the difference between the observed GNDVI and the predicted GNDVI [21]. The equation is shown as follows: where GNDVI P is the observed GNDVI, a and b are the regression coefficients, c is constant, and T and P are the temperature and precipitation, respectively. GNDVI r is the GNDVI residuals. GNDVI O is the predicted GNDVI.

Statistical Analysis and Spatial Distribution Maps
GNDVI in mining areas and buffer zones and GNDVI variances were achieved by using Excel software (Version 14, Microsoft, Inc., Redmond, Washington, DC, USA). In this study, the weight of GNDVI in mining area is the ratio of mining area to the area of all pixels covered. In addition, GNDVI in buffer zones are based on their corresponding ratios. Ordinary kriging (OK) interpolation was performed on the climate factors and the trend line analysis using ArcGIS software (Version 10.2, ESRI, Inc., Redlands, CA, USA). Then, correlation analysis was achieved by using IBM SPSS statistics software (Version 19.0, Armonk, NY, USA).

Tempo-Spatial Change of Vegetation Coverage in eastern Inner Mongolia
The GNDVI slope in eastern Inner Mongolia is shown in Figure 2. Overall, GNDVI slope values were in the range of −0.009 to 0.021, exhibiting an obvious tempo-spatial difference from 1982 to 2015 ( Figure 2a). Changing trend types and their proportions are shown in Figure 2b and Table 3, respectively. Vegetation coverage of 9.97% pixels mainly located in central Hulunbuir, northern Chifeng, and northwest Tongliao presented an extremely significant decreasing trend. An extremely significant increasing trend was found in vegetation coverage of 14.22% pixels distributed in western Hulunbuir, southeast Chifeng and Tongliao, and eastern Hinggan League. There were 62.59% pixels in eastern Inner Mongolia and their vegetation coverage remained unchanged over the past 34 years.
In Hulunbuir, GNDVI slope values ranged from −0.008 to 0.021. Vegetation coverage of 61.56% pixels has no significant change. A decreasing trend was found in vegetation coverage of 19.71% pixels in Ewenki Autonomous Banner, Yakeshi, and Oroqen Autonomous Banner. Vegetation coverage of 18.73 pixels showed an increasing trend, and they were located in Xin Barag Left Banner, Xin Barag Right Banner, Argun, and ZhaLanTun. In Xilingol, GNDVI slope values were in the range of −0.005 to 0.005. An insignificant change was seen in vegetation coverage of 79.96% pixels. There was a decreasing trend of vegetation coverage of 7.31% pixels distributed in Sonid Right Banner and West Ujimqin Banner. Vegetation coverage of 12.73% pixels tended to increase and they were located in Xilin Hot, Abag Banner, and Sonid Left Banner. In Chifeng, GNDVI slope values ranged from −0.006 to 0.007. The proportion of pixels with no change was 48.7%, suggesting that vegetation coverage of these pixels was unchanged over the past 34 years. Vegetation coverage of 26.43% pixels increased and they were distributed in Aohan Banner, Wengniute Banner, Ningcheng, and Kalaqin Banner. In Tongliao, GNDVI slope values were in the range of −0.006 to 0.008. Vegetation coverage of 51.85% pixels in Naiman Banner, Kulun Banner, Kailu, Horqin Left Wing Rear Banner, and Horqin Left Wing Middle Banner exhibited an increasing trend. In Hinggan League, GNDVI slope values ranged from 0.004 to 0.007. There were 57.55% pixels with no change. The proportion of increased pixels (including significant increased and extremely significant increased) decreased in the order of Tongliao > Hinggan League > Chifeng > Hulunbuir > Xilingol, suggesting that the change of vegetation in Tongliao is better than in other cities [22].
Banner. In Tongliao, GNDVI slope values were in the range of −0.006 to 0.008. Vegetation coverage of 51.85% pixels in Naiman Banner, Kulun Banner, Kailu, Horqin Left Wing Rear Banner, and Horqin Left Wing Middle Banner exhibited an increasing trend. In Hinggan League, GNDVI slope values ranged from 0.004 to 0.007. There were 57.55% pixels with no change. The proportion of increased pixels (including significant increased and extremely significant increased) decreased in the order of Tongliao ＞ Hinggan League ＞ Chifeng ＞ Hulunbuir ＞ Xilingol, suggesting that the change of vegetation in Tongliao is better than in other cities [22].

Vegetation Coverage Correlation between 25 Large-Scale Mining Areas and Its Buffer Zones
Correlation analysis of GNDVI between 25 large-scale mining areas and its buffer zones is exhibited in Figure 4. In general, before and after mining, there was a positive correlation of

Vegetation Coverage Correlation between 25 Large-Scale Mining Areas and Its Buffer Zones
Correlation analysis of GNDVI between 25 large-scale mining areas and its buffer zones is exhibited in Figure 4. In general, before and after mining, there was a positive correlation of vegetation coverage between mining area and its buffer zone for each mining area. For 16 mining areas, greater correlation coefficients were found in vegetation coverage between mining area and its 10 km buffer zones.
Before mining, for Lingdong, Tianshun and Duolun, the correlation coefficients of vegetation coverage between mining areas and their 50 km buffer zones were less than 0.5 (p < 0.05). For Zhahanao'er, except 10 buffer zone, the correlation coefficients of vegetation coverage were smaller than 0.5 (p < 0.05) between mining area and other buffer zones. For Liujia, smaller correlation coefficients were found in vegetation coverage among mining area and its 20, 30, and 50 km buffer zones. However, for the above five mining areas, a significant growth was seen in the correlation coefficients in vegetation coverage between mining areas and their buffer zones after mining. For Tiebei, except 10 km, there were poor correlation coefficients between mining area and its buffer zones after mining. vegetation coverage between mining area and its buffer zone for each mining area. For 16 mining areas, greater correlation coefficients were found in vegetation coverage between mining area and its 10 km buffer zones. Before mining, for Lingdong, Tianshun and Duolun, the correlation coefficients of vegetation coverage between mining areas and their 50 km buffer zones were less than 0.5 (p < 0.05). For Zhahanao'er, except 10 buffer zone, the correlation coefficients of vegetation coverage were smaller than 0.5 (p < 0.05) between mining area and other buffer zones. For Liujia, smaller correlation coefficients were found in vegetation coverage among mining area and its 20, 30, and 50 km buffer zones. However, for the above five mining areas, a significant growth was seen in the correlation coefficients in vegetation coverage between mining areas and their buffer zones after mining. For Tiebei, except 10 km, there were poor correlation coefficients between mining area and its buffer zones after mining.

Correlation between Vegetation Coverage and Climatic Factors of 25 Large-Scale Mining Areas
The correlation coefficients among GNDVI and temperature and precipitation of each mining area before and after mining are shown in Ⅱ , and Yuanbaoshan, precipitation played an important role in vegetation change after mining (p < 0.05 and p < 0.01). For, Baorixile, Linglu, and Tianshun, there was a significant relationship between GNDVI and precipitation before mining, but an insignificant relationship was seen after mining. In eight mining areas, temperature and precipitation had significant impacts on vegetation change.

Mines
Before Mining After Mining Temperature Precipitation Temperature Precipitation Figure 4. Correlation of GNDVI between mining area and its buffer zones before and after mining.

Correlation between Vegetation Coverage and Climatic Factors of 25 Large-Scale Mining Areas
The correlation coefficients among GNDVI and temperature and precipitation of each mining area before and after mining are shown in  < 0.05 and p < 0.01). For, Baorixile, Linglu, and Tianshun, there was a significant relationship between GNDVI and precipitation before mining, but an insignificant relationship was seen after mining. In eight mining areas, temperature and precipitation had significant impacts on vegetation change. Notes: * and ** represent p < 0.05 and p < 0.01, respectively. -represents that there is no value.

Correlation between Vegetation Coverage and Human Factors of 25 Large-scale Mining Areas
Residual reflects the impact produced by human activities. Figure 5 shows the residual slope values and t-test. For Baorixile and Laogongyingzi, human activities showed a significant influence on the vegetation change in the past 34 years (p < 0.05), especially promoting vegetation growth after mining. For Baiyinhua Electricity, Zhahanao'er, Zhanihe, and Jinyuanli, there was a significant increasing residual trend (slope >0, p < 0.05) after mining, indicating human activities have a positive effect on vegetation growth. For Liujia, we found a decreasing residual trend (slope <0, p < 0.05) after mining, which suggests that the degradation of vegetation was caused by human activities.

Vegetation Change Before and After Mining in Mining Mrea in Eastern Inner Mongolia
Over the past 34 years, vegetation change in more than half of eastern Inner Mongolia was not obvious. Similar findings were reported by Chen et al. and Mu et al. [53,54]. However, significant changes were found in vegetation coverage in mining areas. In general, from 1982 to 2015, vegetation change in most mining areas in Hulunbuir and Chifeng tended to increase, but in Xilingol and Tongliao tended to decrease ( Figure 6). In terms of vegetation change, seven mining areas presented a decreasing tendency before mining, but an increasing tendency after mining. Meanwhile, a more significant growth was found after mining in seven mining areas. Moreover, there were three mining areas with a less obvious decreasing tendency after mining. This suggests that vegetation change in most mining areas in eastern Inner Mongolia showed an increasing trend after mining. Changing degree of vegetation coverage can be expressed by the variance of GNDVI. As shown in Figure 7, except Laogongyingzi, vegetation coverage exhibited more obvious fluctuations after mining. This also reflected the effects of human activities on the vegetation growth in mining areas.

Vegetation Change Before and After Mining in Mining Mrea in Eastern Inner Mongolia
Over the past 34 years, vegetation change in more than half of eastern Inner Mongolia was not obvious. Similar findings were reported by Chen et al. and Mu et al. [53,54]. However, significant changes were found in vegetation coverage in mining areas. In general, from 1982 to 2015, vegetation change in most mining areas in Hulunbuir and Chifeng tended to increase, but in Xilingol and Tongliao tended to decrease ( Figure 6). In terms of vegetation change, seven mining areas presented a decreasing tendency before mining, but an increasing tendency after mining. Meanwhile, a more significant growth was found after mining in seven mining areas. Moreover, there were three mining areas with a less obvious decreasing tendency after mining. This suggests that vegetation change in most mining areas in eastern Inner Mongolia showed an increasing trend after mining. Changing degree of vegetation coverage can be expressed by the variance of GNDVI. As shown in Figure 7

Vegetation Change Before and After Mining in Mining Mrea in Eastern Inner Mongolia
Over the past 34 years, vegetation change in more than half of eastern Inner Mongolia was not obvious. Similar findings were reported by Chen et al. and Mu et al. [53,54]. However, significant changes were found in vegetation coverage in mining areas. In general, from 1982 to 2015, vegetation change in most mining areas in Hulunbuir and Chifeng tended to increase, but in Xilingol and Tongliao tended to decrease ( Figure 6). In terms of vegetation change, seven mining areas presented a decreasing tendency before mining, but an increasing tendency after mining. Meanwhile, a more significant growth was found after mining in seven mining areas. Moreover, there were three mining areas with a less obvious decreasing tendency after mining. This suggests that vegetation change in most mining areas in eastern Inner Mongolia showed an increasing trend after mining. Changing degree of vegetation coverage can be expressed by the variance of GNDVI. As shown in Figure 7, except Laogongyingzi, vegetation coverage exhibited more obvious fluctuations after mining. This also reflected the effects of human activities on the vegetation growth in mining areas.

Response of Vegetation Change to Climatic Factors in Mining Area
Existing research has shown that precipitation during the growing season plays a more important role than temperature in vegetation growth in semi-arid and arid regions [55]. As shown in Table 4, vegetation changes were affected significantly by temperature during the growing season in only two mining areas. For Yimin, vegetation remained basically unchanged from 1982 to 2015 ( Figure 6). The vegetation fluctuation after mining was the result of change in temperature and precipitation, with little influence from mining activities [56]. For Huolinhe No. 1, Li et al. [57] and Wu et al. [58] reported that temperature change had a significant impact on agricultural production in the past 50 years. There was a significant positive correlation between vegetation change and precipitation in 14 mining areas. These mining areas were located in Hulunbuir and Xilingol. Meng et al. [21] found that there was a significant change in precipitation during the growing season (spring, summer, and autumn) in Hulunbuir and Xilingol. In addition, Zhang et al. [59] and Hang et al. [60] reported similar findings. In addition, the weakening correlation between vegetation and temperature may be due to drought in the Northern Hemisphere [60].

Do Human Activities Affect Vegetation Growth in Mining Area?
Figure 5a shows that human activities had a positive or negative impact on vegetation growth. Vegetation degradation in a quarter of mining areas was caused by human activities after mining. Zhang et al. [13] and Bao et al. [22] reported that underground and open-pit mining results in vegetation degradation in mining areas. However, in this study, human activities promoted vegetation growth in approximately three quarters of mining areas after mining. This could be explained by the application of concurrent mining and reclamation. Remediation technologies mainly included slope treatment, soil fertilization, and bioremediation. For instance, for Baorixile, from 1998 to 2000, the GNDVI increasing tendency indicated that human activities had little effect on vegetation coverage in the early stage of mining [61]. The GNDVI in 2007 was lowest, which might have been caused by the reconstruction and expansion project in 2006 [62]. Meanwhile, a near-term governance planning (2010-2013) was formulated and implemented to protect the environment by applying slope treatment and soil maturation [63]. For Zhanihe, strip controlled mining technology was used for slope stabilizing in 2013, which prevented the vegetation degradation from 2013 to 2014 [64]. For Zhahanao'er, slope protection technology by spraying vegetation on carrying soil was applied in 2009, leading to an increase in vegetation coverage between 2010 and 2011 [65]. In addition,

Response of Vegetation Change to Climatic Factors in Mining Area
Existing research has shown that precipitation during the growing season plays a more important role than temperature in vegetation growth in semi-arid and arid regions [55]. As shown in Table 4, vegetation changes were affected significantly by temperature during the growing season in only two mining areas. For Yimin, vegetation remained basically unchanged from 1982 to 2015 ( Figure 6). The vegetation fluctuation after mining was the result of change in temperature and precipitation, with little influence from mining activities [56]. For Huolinhe No. 1, Li et al. [57] and Wu et al. [58] reported that temperature change had a significant impact on agricultural production in the past 50 years. There was a significant positive correlation between vegetation change and precipitation in 14 mining areas. These mining areas were located in Hulunbuir and Xilingol. Meng et al. [21] found that there was a significant change in precipitation during the growing season (spring, summer, and autumn) in Hulunbuir and Xilingol. In addition, Zhang et al. [59] and Hang et al. [60] reported similar findings. In addition, the weakening correlation between vegetation and temperature may be due to drought in the Northern Hemisphere [60].

Do Human Activities Affect Vegetation Growth in Mining Area?
Figure 5a shows that human activities had a positive or negative impact on vegetation growth. Vegetation degradation in a quarter of mining areas was caused by human activities after mining. Zhang et al. [13] and Bao et al. [22] reported that underground and open-pit mining results in vegetation degradation in mining areas. However, in this study, human activities promoted vegetation growth in approximately three quarters of mining areas after mining. This could be explained by the application of concurrent mining and reclamation. Remediation technologies mainly included slope treatment, soil fertilization, and bioremediation. For instance, for Baorixile, from 1998 to 2000, the GNDVI increasing tendency indicated that human activities had little effect on vegetation coverage in the early stage of mining [61]. The GNDVI in 2007 was lowest, which might have been caused by the reconstruction and expansion project in 2006 [62]. Meanwhile, a near-term governance planning (2010-2013) was formulated and implemented to protect the environment by applying slope treatment and soil maturation [63]. For Zhanihe, strip controlled mining technology was used for slope stabilizing in 2013, which prevented the vegetation degradation from 2013 to 2014 [64]. For Zhahanao'er, slope protection technology by spraying vegetation on carrying soil was applied in 2009, leading to an increase in vegetation coverage between 2010 and 2011 [65]. In addition, selecting proper plants was crucial to vegetation restoration in mining area. Leymus chinensis and hippophae rhamnoides were always selected and planted in Baorixile with developed root and drought tolerance [63]. Management of replanting vegetation played an active role in the ecological remediation of mining area [66]. Hence, we should take similar and suitable measures to protect vegetation for Linglu, Lingdong, Duolun, Tianshun, and Shengli in Yakeshi.

Conclusions
From 1982 to 2015, vegetation change in more than half of eastern Inner Mongolia was insignificant, but significant changes were found in mining areas. Vegetation coverage of 60% of large-scale mining areas showed an increasing tendency after mining. During the growing season, the responses of vegetation change to precipitation were obvious in mining areas. Human activities in mining areas promoted vegetation growth. These findings help to better understand the effects of active human activities on vegetation growth in mining areas, and to pay more attention to vegetation degraded areas. We will explore other influencing factors (e.g., land use change, overgrazing, and soil condition) in the future.