Evaluation of surface coal mines reclaimed to different vegetation types and their stability in semi-arid areas

Since it is difficult to monitor the recovery of vegetation after disturbance in mining areas in semi-arid zones, if the problem of monitoring this type of area can be solved, it can be applied to the vast majority of mining areas, and for this reason we chose a typical mining area, the Pingshuo mining area. Here the monitoring of vegetation recovery will be based on the identification of mining disturbances. Disturbance will be defined as the complete loss of vegetation due to mining-related activities. As shown in Fig. 1, the Pingshuo mining district is located in the northern part of the Shanxi Province, China, and was established under the jurisdiction of Shuozhou City, Shanxi Province. It is one of the largest coal concentration development areas in China(Xu et al., 2021). The geographical coordinates are 112°16′20″- 112°30′32 “E and 39°26′10″- 39°35′30 ”N. From east to west, the area measures 67.9 km by 69.5 km, with a total area of 517.48 km². This study area encompasses three districts in Shuozhou, namely Pinglu, Shuocheng and Shanyin, and is situated in the border area between the Loess Plateau and the northern Tushan Mountains. A climate with average annual precipitation of 428.2–449.0 mm and evaporation of 1786.6–2598.0 mm, of which 75 % of the precipitation occurs during July, August, and September(K. Yang et al., 2022). This type of natural environment makes the ecosystem in the study area weak in terms of resilience, severe in terms of soil erosion, and possesses only a limited amount of vegetation cover. Furthermore, long-term mining operations and town construction have significantly changed the land's cover and use, leading to severe ecological deterioration on the already delicate Loess Plateau. Land use management in the area has long been concerned about environmental degradation in mining sites as well as the ongoing reclamation of mining zones..

This study included all surface reflectance images (30 m x 30 m) covering the study area. The surface reflectance data consists of 34 time-series Landsat TM/OLI images spanning the period 1987 to 2021, including Landsat 5 TM, Landsat 7 ETM+, and Landsat 8 OLI satellites. Data were collected from the United States Geological Survey (USGS; https://glovis.usgs.gov/) and processed on GEE platform. In view of the vegetation characteristics as well as the interference of clouds and other atmospheric conditions, all imageries were taken from April to September with less than 10 % cloud cover(Liao et al., 2019). For the calibration of the sensor radiance of Landsat 5, 7 and 8 images for surface reflectance, the Landsat Ecosystem Disturbance Adaptive Processing System (LEDAPS) (Park et al., 2020) and Landsat Surface Reflectance Code (LaSRC) (Myroniuk et al., 2020) were used. The C implementation of Function of Mask (CFMASK) determines important pixel data quality flag information suggestive of clouds, shadows, water, and snow(Stillinger et al., 2019). Section 2.3 corresponds to step 2 of the flow diagram of the whole study (Fig. 2), and sections 2.3, 2.4, and 2.5 correspond to steps 2, 3, and 4, respectively.

Restoration effects are highly correlated with the type of vegetation used for reclamation. The vegetation types where reclamation occurred following disturbance were identified based on Google Earth observations of historical images and field surveys. A classification of the vegetation was then made based on the differences in ecosystem service supply capacity between forest, grassland, and cropland vegetation. CART classification in NDVI and machine learning has proven to be an effective tool for distinguishing different types of vegetation(Liu et al., 2020). Once the areas where mining disturbances occurred and reclaimed were identified, we defined the physical differences between different vegetation types, labeled the remote sensing image elements, and processed the data using GEE (Fig. 3). It should be noted that the whole classification process only collected NDVI time series data for the last year of the study interval, and only determined the final characteristics of spatial distributions of different vegetation types. It should be further noted that as part of this study, physical characteristics were extracted to improve the accuracy of land cover classifications, including the beginning and ending dates of the growing season, the length of the growing season, the seasonal magnitude and the maximum fitted NDVI values. A growing season begins and ends when the fitted NDVI curve reaches 50 % of its seasonal amplitude measured from left and right minima, respectively(Jia et al., 2014). As an additional measure of validation of the classification results obtained using the phenological features, four statistical temporal characteristics of the fused NDVI data series, including maximum, minimum, mean, and standard deviation values, were extracted to fully demonstrate the reliability of the study.

Vegetation growth in mining areas is a dynamic process that changes over time, and it exhibits different states at different times after planting, but it is not possible to visually identify what growth cycle it is in. A general trend of growth of the restored vegetation indicates a S-shaped exponential growth from slow to rapid and then from rapid to slow, reaching a steady state(Zeng et al., 2020). This basic law served as the basis for the design of experiments for data mining using the S-logistic fitting function for cumulative NDVI at the pixel level, as shown in Fig. 4, which restored the core logic of the entire process using S-shaped curves. In accordance with the growth characteristics of vegetation, the cumNDVI will gradually converge to a stable value, but not reach it. We therefore utilize this trend information in conjunction with the concept of limit in calculus to uncover the period of vegetation reclamation using the maximum value of the second-order derivative. While the maximum value of the first-order derivative of the curve represents the most rapid stage in the reclamation of vegetation in the mining area, its vicinity represents the most rapid stage in vegetation growth during this period. In this stage, the vegetation will undergo a process of slow recovery followed by rapid recovery, followed by gradual stabilization.

For the purpose of continuously monitoring the growth dynamics of the vegetation, the area and the type of vegetation that will be utilized for reclamation were initially determined (Fig. 5). All data used for this process was open source, and no dedicated database was required from the mine site. Reclamation areas were determined by using LandTrendr's algorithm to monitor 95 % of the NDVI for the full period of 2021, which is non-complex and easy to implement, and the final areas were determined by pixel aggregation, thus reducing errors(Yang et al., 2018). The majority of reclamation activities have been concentrated in northeastern and southwest corners of the mine area, resulting in a large and small vegetation distribution pattern. In terms of the final reclamation area structure, forest land, grassland, and cropland contributed 15.79 %, 42.61 %, and 41.59 % respectively. In comparison with the northeast corner, more vegetation area was reclaimed in the southwest, where forest, grasslands, and crops were spatially and massively regularly distributed. Forests are primarily concentrated near mine boundaries, forming three large patches and a scattered distribution. A belt of cultivated land surrounds the forest land and is distributed to the northeast of the gathering center. From a quantitative perspective, arable land occupies the majority of the mine area. Any reclaimed area will contain grasslands in between forest and crops as a supplement to the reclaimed vegetation. A reclaimed area of grassland and cropland was found in the northeastern gathering center, and the ratio of both was primarily dominated by cropland.

In accordance with Fig. 6, reclamation activities in the mine area vary greatly before and after 2004, with the majority of the activities occurring between 2010 and 2015. Reclamation activities began in the southwest corner of the mine area, and the reclamation area in the southwest peaked after 2010 and gradually shifted to the northeast within the next five years. As reclamation efforts continued in the southwest after 2016, the northeast has become the focus of these efforts. Different vegetation types represent important reclamation options, and all three types of vegetation have been identified in terms of spatial and temporal processes. It has been found that, forested lands were reclaimed very early, and from 1992 to 2003, forest lands dominated the vegetation reclamation process. From 2004 to 2009, however, the mine began to utilize cropland and grassland for reclamation, and both reclamation areas were comparable to forest lands, indicating that the reclamation policy had changed. The reclamation rate of cropland was much faster between 2010 and 2015 than the reclamation rate of forest and grassland, as well as the area of cropland reclamation being greater than the sum of the two, despite the fact that forest and grassland also increased during this period of time. Between 2016 and 2021, virtually no new forest was established, while cropland and grassland dominated the reclaimed area.

The reclamation areas for each vegetation type were counted year after year in order to demonstrate clearly the changes in area over time (Fig. 7). From 1986 to 2021, forest land, pasture land, and cropland were reclaimed in quantities of 9.10 km², 24.54 km², and 23.95 km², respectively, although reclamation rates varied from one vegetation type to another. Reclamation of grassland and cropland initiated later than that of forest land, but the extremely rapid pace soon exceeded the area of forest land reclamation, especially the rapid growth of cropland in 2010. Also, vegetation growth in the total reclaimed area did not follow the rhythm of mining activities, but after 2001, widespread planting of vegetation began. In the period 1998–2012, 2001–2017, and 2004–2017, respectively, forest, grasslands, and cropland increased significantly; however, their peak rates varied significantly. Forest lands reclaimed during 2009 reached 1.91 km², while grassland reclaimed during 2009 reached 1.96 km² and cropland reclaimed during 2013. The new areas of the three vegetation types have not changed in recent years and are basically zero, indicating that reclamation activities at the Pingshuo mine have been discontinued. However, the effect of different vegetation varies widely in practice, and it is unclear whether the reclaimed vegetation will remain stable until 2021.

For the purpose of analyzing a forest, grassland, and cropland's growth dynamics and monitoring their contribution to the final reclamation effect, CumNDVI was used to identify the stabilization process of vegetation. According to the method described in Section 2 for categorizing vegetation reclamation stages, vegetation belonging to reclamation stages in the reclamation area was extracted, which basically covered all vegetation identified in the reclamation area. According to our research, forest land, grassland, and cropland in the mine area stabilized on average after 6.50 years, 5.42 years, and 4.64 years, respectively (Fig. 8 (b)). Moreover, most of the reclaimed vegetation is still in the rapid development stage of reclamation and is primarily cultivated, with small areas of forests and grasslands reclaimed. There were two types of vegetation identified in the continuous reclamation site, vegetation aged 1–4 years and vegetation aged 4–7 years. These vegetation types were primarily cropland and grassland. It has been determined that the stability time in the forest area typically falls within the range of 7–10 years, occupying about 15.0 % of the total area (Fig. 8 (a)). Reclamation stability was determined by examining the time required by each vegetation type. Forest are generally stable vegetation; however, grasslands and croplands still require a greater period of time in order to become stable, especially those located in the northeast corner of the mine area.

A better understanding of the growth characteristics of reclaimed vegetation after mining is crucial for the development of a sustainable mining industry. One of the most critical factors in the reclamation process is whether the vegetation can be stabilized after years of growth. By doing so, it is possible to discern the contribution of planted vegetation to the ecological development of the mining area, which can prove very useful in subsequent replanting and replanting selection. In this study, we analyzed cumNDVI curves over time to determine reclamation stages, identified key nodes, and reconstructed the whole growth dynamics of vegetation. As feasible and highly efficient as current studies aimed at identifying vegetation recovery in mining areas based on time series data(Yang et al., 2018). Indicators of vegetation obtained from remote sensing can be obtained in a short time period, at high resolution, and with a long-term availability, which is related to the capability of Landsat satellites to mine and analyze mesoscale long-term time series remote sensing indicators(Myroniuk et al., 2020). With the advancement of remote sensing big data, these studies are now capable of providing a wide range of information regarding vegetation growth. Moreover, when compared with traditional NDVI metrics, cumNDVI incorporates the different distribution characteristics of a large number of pixels, thus eliminating the possibility of outliers. In this manner, some progress has been made towards restoring the dynamics of vegetation growth(W. Yang et al., 2022). Also, by utilizing the LandTrend algorithm on the GEE remote sensing cloud computing platform, it is possible to extract boundary and temporal information spatially by using pixel aggregation(Liu et al., 2020). By utilizing these new techniques, it has been possible to ensure accuracy of data results and the reliability of identification information when monitoring vegetation details at the site level. As an additional demonstration of the reliability of the reclamation stages delineated, the mean CumNDVI values for the three vegetation types within the study area were counted and plotted as line graphs in different years (Fig. 9). Our findings in Section 3.3 indicate that the mean stabilization time for forest land, grassland and cropland within the mine area is 6.50, 5.42 and 4.64 years, respectively. This statistic also indicates that it took eight, six, and five years for forest, grassland, and cropland to reach 70 % of maximum CumNDVI, respectively. As a result of the use of vegetation growth curves to simulate the growth of vegetation processes and evaluate their performance, the results of the modeling are slightly conservative, but they are reliable.

A monitoring of the vegetation maturation cycle in mining areas is crucial to determining the effectiveness of reclamation following mining, particularly when it comes to determining the growth cycle of different vegetation types. The results of our study indicate that 49.46 % of the vegetation within the entire mine area has not reached its peak of development, and this estimate is relatively conservative. Fig. 10 illustrates the spatial and temporal distribution of growth and development of the three types of vegetation from 1986 in order to underscore the significance of the study. The results of Fig. 10 (a-c) reveal that both woodland and cropland are still primarily at an early stage of development, with most development occurring between 1 and 7 years, whereas more than half of the grasslands are within the 7–10 year range. Therefore, Fig. 10 (d) was generated using thresholds for the growth of different vegetation to a stable stage. According to our findings, the vegetation in the mine area will gradually stabilize in the near future. In specific terms, the forest and grassland in the southwest part of the mine will gradually stabilize within two years, while the cropland in this part is on the verge of stabilizing within two years, and the northeast part will mature in 2–4 years. Grassland and cropland cover the northeast portion of the mine, where vegetation will primarily stabilize and grow over the next 2–4 years. Clearly, this is relevant to the sequence and timeframe involved in mining reclamation. Further demonstrating the accuracy of the reclamation results identification, the Pingshuo mine site has very accurate reclamation results and guidance regarding the rehabilitation of the site. Our method can also be used to provide reasonable assistance in mining or other areas where similar natural conditions and vegetation types are occurring, which allows reclamation activities to be evaluated accurately based on spatial and temporal information, rather than limited to the initial planting period.