A UAV- and eld-based investigation of the land degradation and soil erosion at opencast coal mine dumps after 5 years’ evolution of natural processes

11 Open-pit coal mining has a large impact on land surface, both at the mining pits themselves 12 and at waste sites. After artificial management is stopped, a reclaimed opencast coal mine dump is 13 affected by wind and water erosion from natural processes, resulting in land degradation and even 14 safety incidents. In this paper, the soil erosion and land degradation after 5 years of such natural 15 processes, at the Xilinhot open pit coal mine dump in Inner Mongolia, were investigated. A multi16 source data acquisition method was applied: the vegetation coverage index was extracted from GF17 1 satellite imagery, high-precision terrain characteristics and the location and degree of soil erosion 18 were obtained using an Unmanned Aerial Vehicle (UAV), and the physical properties of the topsoil 19 were obtained by field sampling. On this basis, the degree and spatial distribution of erosion cracks 20 were identified, and the causes of soil erosion and land degradation were analyzed using a 21 geographical detector. The results show that: 1) The multi-source data acquisition method can 22 provide effective basic data for the quantitative evaluation of the ecological environment at dumps; 23 2) slope aspect and vegetation fractional coverage are the main factors affecting the degree of 24 degradation and soil erosion. Based on this analysis, several countermeasures are proposed to 25 mitigate land degradation: 1) The windward slope be designed to imitate the natural landform; 2) 26 engineering measures should be applied at the slope to restrain soil erosion; 3) pioneer plants should 27 be widely planted on the platform at the early stage of reclamation. 28


32
Surface mining is the most widely used mining method in the world. In the United States, 33 Germany, Australia, Russia and other developed countries, the proportion of opencast mining 34 reached more than 60%. Although surface mining has many advantages, such as high safety, high 35 mining efficiency, high recovery rate, low cost and etc. (Kenndey, 1990), it drastically damages the 36 eco-environment and produces a large amount of coal waste. The dump formed by the resulting coal 37 waste stack takes up a large area of land, where the surface soil is poor and it is pressed by heavy 38 trucks, resulting in problems for plant roots and an increase in surface runoff. During the rainy 39 season, concentrated precipitation flows into the settlement cracks, causing disasters such as col-40 lapse, landslide and debris flow, and seriously threatening the lives of residents nearby (Neugirg et 41 al., 2016). Statistically, the land used for open-pit mine dumps accounts for 30%-50% of the total 42 land used for mining. As a large coal mining country, China's raw coal production accounts for more 43 than 70% of its total energy production (National Bureau of Statistics, 2017). There are a large 44 number of shallow coal seams in western China. The mining conditions are good and the coal quality 45 is excellent. The center of coal mining is predicted to move gradually westwards (Guo et al., 2018). 46 The proportion of open-pit production also increase gradually, from about 3.3% in 1998 to about 47 15% in 2015  and the annual output of coal is expected to reach 3.9×10 9 tons by 48 2020. The average annual additional land area occupied is up to 1×10 4 hm 2 (10 8 m 2 ), and the soil 49 erosion area increases annually by 3×10 3 hm 2 (3×10 7 m 2 ) (Qi, 2017).Therefore, it is important to 50 take account of the safety and stability of the dump when assessing the production safety and eco-51 nomic benefits of the open mining area. The monitoring and maintenance of the dump has become 52 an important aspect of the reclamation process of the open mining area. 53 The dump is mainly composed of waste material from open-pit mining. After the dump reaches 54 its full capacity, it is covered with a layer of topsoil to ensure a rapid restoration of vegetation. The 55 construction of the dump needs to be adapted to local conditions, and the appropriate disposal 56 method should be selected according to the geomorphologic characteristics of the mining area. Sev-57 eral types can be distinguished, such as conical waste dump, piles type of shape, board shape, shape 58 of the terrace, shape of slope, ridge shape, straight shape, or they may take the form of flat coverings. 59 (Petra et al., 2015). Because of the high cost of land acquisition in China, flat coverings are rarely 60 used in filling the dumps. The terraced landform of the Loess plateau can be imitated by designing 61 terraced dumps with a relative height difference of 100 m-150 m between the platform and the slope 62 intersect (Fig. 1a). Alternatively, in hill and gully regions of the Loess plateau, fly ash and gangue 63 can be used to carry out a mixed ecological filling of the open pit and gully (Fig. 1b), laying a 64 foundation for subsequent ecological restoration (Shanxi provincial bureau of quality and technical 65 supervision, 2016). However, in the eastern grassland area of northern China, the terrain is flat, there 66 is an insufficient supply of topsoil, and the precipitation can be highly intense. In order to reduce 67 the amount of land they occupy, most dumps adopt a ladder distribution (Liu et al., 2011). In view of a series of environmental problems and safety problems caused by surface mining, 71 the land reclamation in mining areas has been carried out in many countries. However, in order to 72 restore the mining eco-system to a stable top community as soon as possible, post-restoration mon-73 itoring is still needed. The monitoring of mine restoration mainly includes vegetation restoration, 74 soil quality and soil erosion. Vegetation restoration is the key to the restoration of dump sites. Veg-75 etation restoration can make full use of the function of the soil-plant composite system, improve the 76 local environment and promote a regional ecological balance (Srinivasan Madhusudan et al., 2015), 77 and also significantly improve the soil bulk density, soil moisture content and soil porosity (Wang 78 et al., 2016). Through the calculation of VFC, the vegetation growth status of the dump can be 79 effectively quantified, serving as an important reference for soil erosion. This should give priority 80 to local species with a high survival rate, strong resistance and an ability to improve the physical 81 and chemical properties of the soil. The ideal procedure is to gradually strengthen the composition, 82 structure and level of vegetation, improve the function of the soil and its water conservation and 83 self-renewal ability, and finally to achieve self-balance, establishing a complete and stable ecosys-84 tem . At present, in order to improve the success rate and greening rate of recla-85 mation over a short period of time, most of dump sites adopt artificial management. Silman, 2016; Siebert and Teizer, 2014) and changes in mine land use (Gui et al., 2008), and there 119 is little research carried out on mine dumps. In addition, there is a lack of systematic research on the 120 evolution of a dump, and the associated degradation mechanisms, under the influence of multiple

Study area 134
The experimental plot is located at the Western One site of the Shengli coal field, with geo-135 graphic coordinates of 115° 58′-116° 2′ 50″ E, 43° 59′-44° 2′ 15″N (Fig. 2), located in Xilinhot, 136 Inner Mongolia Autonomous Region, which is the largest prairie area in China. The mining area has 137 a typical temperate, semi-arid, continental monsoon climate, where the annual average temperature 138 is 0.3 ℃ and the average wind speed is above 8.5 m/s. The total annual precipitation averages 294.9 139 mm, with more than 70% occurring from June to August. The average annual effective evaporation 140 is 1811.3 mm, which is more than six times greater than the amount of precipitation. The coal mine 141 has a coal-bearing area of 342 km 2 , with proven reserves of 15.932 billion tons, most of which are 142 suitable for open-pit mining. There are three main external dumping sites, namely south dump, north 143 dump and side dump. All the three waste dumps had been reclaimed, with a green area of 8.64 km 2 . 144 145 Fig. 2 The study area location. 146

147
The specific study area is located in north dump of the Western One site of the Shengli coal 148 mine, with an area of 1.07 km 2 . The dump is divided into four platforms with an altitude range of 149 980 m-1040 m and a slope angle of 33°. North dump has been reclaimed since 2006, with a total 150 land area of 1.01 km 2 , in which the slope greening area is 0.355 km 2 , the platform greening area is 151 0.655 km 2 , and the green rate is 100%. Since the cessation of artificial conservation in 2013, the 152 vegetation at the dump site has been degraded, and concentrated precipitation in summer has led to 153 soil erosion, which can easily lead to slope erosion, collapse, debris flow and other disasters under 154 the action of gravity. 155

157
As a multi-stage dump, north dump is mainly composed of deep rock-soil, pulverized coal and 158 surface soil close to the coal seam, and siltstone, mudstone and gravel of different sizes away from 159 the stope. The soil is closely packed, with few internal gaps, the permeability coefficient is very 160 small, the seepage storage capacity is very low, and it contains essentially no humus. Without man-161 agement and protection, dust and soil erosion can easily occur. Therefore, topsoil with a thickness 162 of 30-50 cm has been added to cover the surface. This part of the soil has a high organic matter 163 content and is fully matured. The soil has favorable air permeability, in which the number of seed 164 banks and microorganisms is large, which is conducive to plant growth. After completion of the soil 165 covering and before the arrival of the rainy season, forage grass seeds and shrub seeds with salt-166 alkali resistance, drought resistance, strong adaptability and nitrogen fixation were planted. Artifi-167 cial maintenance and management was then carried out for 4 years, including watering, sprinkling 168 and irrigation, adding topdressing fertilizer, disease and insect pest control, soil cultivation and 169 planting. Artificial maintenance and management are quite important for vegetation construction 170 (shown in Fig. 3). However, after this was stopped, north dump deformed to different degrees, and 171 its vegetation degenerated, under the influence of wind and rain erosion (Fig. 4). The intense pre-172 cipitation which occurs during summer has led to soil and water loss, which can in turn easily lead 173 to erosion ditches, collapse, debris flow and other problems on the slope. poor, and usually consist of a mixture of coarse-grained particles and rock fragments. In addition, 211 the intensive traffic of heavy machinery used during reclamation can seriously compact soils, further 212 degrading its physical quality. This anthropogenic activity has a significant influence on the soil 213 stability to surface weathering, groundwater or infiltration, gully erosion and slope morphology 214 . 215 The physical properties of the topsoil of north dump were obtained by field sampling. Samples 216 from 117 points were collected in May 2017, uniformly spaced along the dump with a step length 217 of 100 m. The sampling points were supplemented at soil erosion locations (Fig. 6) and the sampling 218 depth was 15 cm. Samples at each sampling point were taken using a cutting ring, placed in a sealed 219 bag, fresh weight of soil was measured on site, and sent to the laboratory 4 days later to measure 220 their physical properties, including dry weight , porosity, water content and bulk density. The soil 221 moisture content was determined using a drying method. Fresh weight of soil was measured on 222 site, and dry weight was measured after drying in the laboratory. The soil bulk density and total 223 porosity were determined using a cutting ring method. We pressed the ring knife vertically into the 224 topsoil, and dig the ring knife out of the soil with a shovel and flatten the upper and lower ends. terized by the addition of a multispectral camera with high spatial and temporal resolution ( Table  235 1). It is widely used in the fields of geographical mapping, oceanic and atmospheric meteorological 236 observation, water conservation and forestry resources monitoring, fine management of urban areas 237 and transportation, epidemic situation assessment and public health emergency response, and sci-238 entific research on the Earth system. Based on the growth cycle of plants and the visibility of satellite 239 images, the multi-spectral image with a spatial resolution of 8 m obtained by GF-1 on August 11, 240 2016 was adopted. At this time, the plant growth condition was relatively good and there was rela-241 tively little cloud cover. To improve the quality and reduce the effects of terrain and atmospheric 242 noise, the images were preprocessed using geometric correction, radiometric calibration, and atmos-243 pheric correction from ENVI 5.2. 244 We take VFC max =1 and VFC min =0, so equation (2) where NDVI values with accumulative probability of 5% and 95% were taken as NDVI min and 266 NDVI max respectively. 267 VFC values range from 0 to 1, and the following results are found in this study, through field 268 investigation and comparison with remote sensing images. An area with VFC greater than 0.8 is a 269 high VFC area, and the plants are mainly pinus (Genus), ulmus pumila and armeniaca sibirica trees. 270 Between 0.4 and 0.8 represents a medium VFC area, and the vegetation mainly includes shrubs such 271 as C. korshinskii and hippophae rhamnoides. Lower than 0.4 corresponds to a low VFC area, which 272 is dominated by herbs such as medicago sativa and astragalus adsurgens and bare land. formats for other software. The first step is to input them to the Pix4D desktop to carry out image 297 alignment after adjusting for chromatic aberration, noise, and the white balance of the pictures. We 298 obtained the camera position corresponding to each picture, the internal camera orientation param-299 eters, and the sparse point cloud of the terrain, by using feature identification and feature matching. 300 The second step is to import and identify GCPs. The terrain profile, composed of a sparse point 301 cloud, can be seen where the aerotriangulation rays intersect, so that GCPs can be easily identified.

302
In the final step, the dense point cloud, 3D scene reconstruction, post-processing, the DSM and the 303 DOM are produced using a one-key procedure. The terrain of the whole dump can be obtained from 304 the DEM while particular locations on the dump can be extracted from the DOM. 305

Geographical detector 306
The geographical detector model was developed for exploring the relationships between spatial 307 patterns of landscapes and the factors which impact them, by Wang et al. (2010). Its principle is that 308 the spatial distributions of two attributes tend to be similar, if there are spatial relationships or inter-309 actions between them. The spatial data do not have to be stratified in geospatial terms, although the 310 attributes can be stratified. The similarity between two attributes can be represented by the so-called  employed in our study to objectively assign weights to different environment factors affecting soil 324 erosion hazards at the dump. As the dependent variable, soil erosion was described by the erosion 325 area. The area of gully erosion was described by the maximum length multiplied by the maximum 326 width, and the area of sheet erosion was identified manually. The independent variables were the 327 step slope gradient, elevation, VFC, degree of porosity, bulk density, soil moisture content and slope 328 aspect for each hazard location. In this way, the main driving factors that affect the amount of soil 329 erosion were assessed. The geographical detector consists of four parts, i.e., the factor detector, 330 interaction detector, risk detector, and ecological detector. We focused on the factor detector and the 331 interaction detector in this study. 332 334 The main soil components of the dump are siltstone, mudstone, carbonaceous mudstone and 335 chestnut soil. However, the proportion of chestnut soil is very small, with an average overburden 336 thickness of 30 cm, a sandy soil texture and a humus content between 1.5% and 3%. Table 2 shows 337 the physical properties of the soil samples. The average bulk density of the topsoil is as high as 1.51 338 g/cm 3 , and the average porosity of the soil is 35.89%. The soil tightness is much higher than that of 339 the soil under natural conditions. The average value of soil moisture content was 3.83%, indicating 340 the topsoil has a poor ability to conserve moisture, therefore it has a low fertility and is not conducive 341

Soil property analysis
to plant growth, making it susceptible to erosion. The sampling points were interpolated using the 342 Kriging method to obtain the physical properties (Mendes et al., 2019) of the entire dump (Fig. 7), 343 including soil moisture content, porosity and bulk density. The topmost platform is the most heavily 344 compacted, so that it has the smallest porosity, maximum bulk density and minimum water content. 345  The soil porosity of the dump site in the study area is low around and high in the middle, which 352 may be due to the fact that the top platform has been repeatedly rolled by heavy machinery, resulting 353 in soil hardening and topsoil hardening (Liu et al., 2016b). Although dust is reduced to a certain 354 extent, the soil with low porosity reduces the infiltration rate of water, resulting in a large amount 355 of surface water runoff, thus causing soil erosion. west of the dump. After long-term rolling, the platform in the west is sunken, which plays a certain 361 role in water collection. The low moisture content on the south side is due to sunny slopes (Pan et  362 al., 2017). On the north side, the dry and cold monsoon prevailing in spring and winter takes away 363 a large amount of water from the soil surface, so the soil moisture content is low. 364

Bulk density 365
The distribution of soil bulk density on slope surface is affected by vegetation and topography. 366 The moisture, nutrient and air content in soil also change with the slope position. As water flows 367 through the surface, it moves fine particles from the soil to the bottom of the slope. With the rain 368 erosion, some slope will also occur collapse. The soil surface structure from the upper slope to the 369 lower slope tends to change from compact to loose and porous (Lü et al., 2018).  pacted by heavy transport equipment and is more conducive to water infiltration and plant root 377 growth (Qi, 2017). In the high VFC area of the platform in the south of the dump, because of inten-378 sive planting during reclamation, the tree cluster distribution makes it more stable and not so vul-379 nerable to degradation. In different slope aspect, VFC of north slope and east slope was higher than 380 that of south slope and west slope. It indicates that the growth of vegetation is closely related to the 381 slope direction (Frankard et al., 2000). The north slope and the east slope are shady slopes with short 382 light time, weak plant transpiration and little water evaporation, while the south slope and the west 383 slope are sunny slopes with long light time, strong plant transpiration and much water evaporation, 384 which is difficult for vegetation restoration. 385 386 We utilized a SfM algorithm to process the aerial photos to acquire the geomorphology and 387

Geomorphology and gully erosion
gully erosion of the dump. The slope aspect (Fig. 8b) and gradient (Fig. 8c) were then obtained from 388 the DSM. The slope gradients of each platform are approximately equal, at about 33 degrees, how-389 ever, the step slope aspects are more variable (Fig. 9). In the north, northwest, northeast, east and 390 southeast directions, the step slope gradients are about 16 degrees, but in the west, southwest and 391 south directions, the step slope gradient is significantly shallower, even reaching as low as approx-392 imately 6 degrees. The greater the step slope gradient is, the more vulnerable the surface is to cause 393 Rock fragment movement (Nyssen et al., 2006a) and wind-water erosion (Zhang et al., 2019). The resolution of the DOM was 6 cm. Fig. 10 shows the distribution of erosion gullies and 400 collapsed areas. All the geological hazards are located in slope areas. The heavy mechanical equip-401 ment used in the dump compacts the dump platform, making it less susceptible to infiltration of 402 heavy rain. However, on the slopes of the dump, loose rocks and soils, poor stability, and low water 403 content make it vulnerable to water-wind erosion (Kainthola et al., 2011). The erosion gullies radar 404 chart (Fig. 11) shows the erosion extent on four platforms in different directions. It can be found in 405 combination with Table 3, the region with the most serious soil erosion was located west to north-  The factor detector reveals the impact of a single driving factor on the degree of soil erosion. 434 The power of determinant (P D,H ) of each factor is shown in Fig. 12. This shows that the most im-  Step slope Porosity Moisture content Bulk density Aspect VFC Elevation Power of determinant driving factors was bigger than that of each individual driving factor, there were significant differ-470 ences in the interaction strength of different factors (Table 4). The interactive effects of the step 471 slope gradient and elevation, soil moisture content and bulk density, and slope aspect and elevation 472 were greater than the maximum of their separate effects. However, the interactive effects of other 473 pairs were greater than the sum of the effects of the corresponding factors individually, indicating a 474 strong synergistic effect between the two factors. The interactive influence power of VFC with other 475 factors is higher than 0.36, indicating that VFC can significantly enhance the impact of various 476 environmental factors on abrasion degree. VFC can be used as an auxiliary indicator factor for the 477 monitoring of the aetiology of erosion and degradation (Tong et al., 2014). In addition, driving factor 478 pairs with strong synergistic effects also include: porosity and moisture content, porosity and bulk 479 density, slope aspect and moisture content, and slope aspect and bulk density. The physical proper-480 ties of the soil are intrinsically related, and the synergistic effect of slope aspect, moisture content 481 and bulk density is mainly reflected in the change of these physical properties caused by wind ero-482 sion. 483 Table 4 Results of the interaction detector 484 Step slope gradient Porosity Moisture Bulk density Slope aspect VFC Elevation Step  caused by wind erosion and water erosion, the following three suggestions are put forward accord-497

Discussions
ing to the characteristics of this area: 498 (1) Given the serious soil erosion on the windward slope in the west, we suggest reducing the 499 ladder slope gradient to the northwest. The slope would then better reflect the natural landform 500 before its disturbance by human influence, leading to a reduction in wind erosion. 501 (2) VFC is not only the main factor affecting soil erosion but also an auxiliary indicator for 502 monitoring soil erosion. Increasing VFC is important for the prevention and treatment of land deg-503 radation at dump sites. We can introduce dominant species and appropriately increase the density 504 and diversity of vegetation communities. At the same time, sand barriers and biological arpeggios 505 could be adopted to enhance the stability of vegetation community. 506 (3) Strong wind erosion leads to coarser soil surface particles and a great loss of nutrients. 507 Compaction by large machinery reduces soil porosity and increases bulk density, which is not con-508 ducive to plant growth. However, plant roots can reach deep into the soil, promote the agglomeration 509 of micro-aggregates with small particle size, reduce soil bulk density and increase soil porosity, 510 forming a positive feedback loop. Therefore, shrubs and herbaceous pioneer plants with strong 511 adaptability and good survival rates should be widely planted in the early stage of reclamation, so 512 as to provide wind protection, fix the soil and improve the soil structure. 513 The site-specific suggestions put forward in this paper are applicable to the whole grassland 514 area of Eastern Inner Mongolia. The grassland area of Eastern Inner Mongolia is more than 380,000 515 km 2 . More than 40 mining areas are scattered in the grasslands of eastern Inner Mongolia. The total 516 coal resources are 249 billion tons, and the annual mining volume is 150 -300 million tons. The 517 total area of the dump is more than 14,460 km 2 . 518 519 In this paper, a method of soil erosion monitoring based on multi-source data and combined 520

Limitation of the method and future work
with geographical detector can effectively identify the main factors causing soil erosion, but there 521 are still some scientific problems to be solved. On the one hand, the UAV platform can be equipped 522 with multi-spectrum or hyper-spectrum in the future to obtain more spectral information for con-523 structing vegetation parameters. Soil information can be obtained by inverting vegetation parame-524 ters, and soil erosion warning can be carried out. On the other hand, using UAV loaded RTK (Real 525 -time Kinematic) can greatly reduce the time of data acquisition, so as to construct a more rapid 526 and efficient monitoring method. 527 In addition, this thesis only proposes to alleviate wind erosion in the dump by reducing wind-528 ward slope, but the appropriate gradient is not discussed in depth. In fact, as far back as 1969 geo-529 morphic design approach has been used in landscape planning (McHarg, 1969). At present, the ge- nies will reduce the land acquisition area as much as possible. How to construct the terrain resistant 534 to wind and water erosion is an important direction of mine restoration in the future. 535 The combined methodology that integrating satellite remote sensing imagery for vegetation 536 monitoring, UAVS platform for accurate terrain and soil erosion investigation, and field sampling 537 for soil property analysis, could fill the gap of different method, that satisfied the demand of coal 538 mine dumps land reclamation monitoring, investigation, assessment. 539

540
This paper investigates an outer dump at which artificial management has been stopped, as an 541 example to explore the soil erosion after 5 years of evolution of natural processes, and to determine 542 the main factors influencing the degree of soil erosion. The main conclusions are as follows: 543 1) At a height of 100 m, UAV pictures combined with a SfM algorithm can be used to accu-544 rately and quickly obtain a 3D model of the dump with a total height of 60 m. The accuracy is 545 sufficient to position and identify the soil erosion area. 546 2) The step slope gradient of the coal mine dumps is significantly steeper than the natural land-547 form, which makes the windward slope vulnerable to wind erosion. It is suggested that the windward 548 slope be designed to imitate the natural landform, for example by reducing its gradient. 549 3) Vegetation coverage is the main factor affecting soil erosion. Measures such as sand barriers 550 and biological basketry should be applied at the slope to reduce and restrain gully erosion and planar 551 erosion. Pioneer plants that can improve soil structure should be widely planted on the platform at 552 the early stage of reclamation, so as to facilitate water infiltration and enhance the diversity and 553 stability of vegetation communities during the natural recovery period. 554 4) The combined methodology that integrating satellite remote sensing imagery for vegetation 555 monitoring, UAVS platform for accurate terrain and soil erosion investigation, and field sampling 556 for soil property analysis, could fill the gap of different method, that satisfied the demand of coal 557 mine dumps land reclamation monitoring, investigation, assessment.   The study area location. Note: The designations employed and the presentation of the material on this map do not imply the expression of any opinion whatsoever on the part of Research Square concerning the legal status of any country, territory, city or area o bbnhjr of its authorities, or concerning the delimitation of its frontiers or boundaries. This map has been provided by the authors.       Step slope. (a) The elevation pro le from the DSM is shown on the right, with its location shown on the map on the left. (b) Radar chart of the step slope gradient.