Assessment of soil loss using RUSLE around Mongolian mining sites: a case study on soil erosion at the Baganuur lignite and Erdenet copper–molybdenum mines

Mining constitutes an integral part of Mongolia’s national economy and dominates the country’s export revenue. At the same time, a wide range of mining impacts on soil, water resources, the atmosphere and the biosphere have been documented across the country. This case study addresses the long-term soil degradation around two mining sites located in the semi-arid steppe zone of Mongolia: the open-cast lignite mine of Baganuur about 140 km east of Ulaanbaatar, and the open-pit copper–molybdenum mine of Erdenet about 240 km northwest of Ulaanbaatar, both of which started commercial extraction in the late 1970s. For the assessment of soil erosion, the RUSLE model was applied in different seasons for the period from 1989 to 2018 at 3-year intervals, considering both climatic variation and the expansion of the mines based on maps and satellite imagery. Rainfall erosivity was identified as the most dominant factor driving soil erosion in the study regions, with mining leading to local increases in soil erodibility. The highest soil erosion rates were found in both areas in July 2018, reaching 7.88 t ha–1 month–1 in the Erdenet area and 9.46 t ha–1 month–1 in the Baganuur area. The spatial patterns of soil erosion showed higher soil loss rates were in the vicinity of the mines and adjoining industrial sites. Particularly high soil losses were identified in July 1998, July and August in 2013 and July 2018 in both mining areas. The combination of the RUSLE model, remote sensing and ground truth data as and their processing by GIS was found to be a time-saving and cost-effective technique for continuous monitoring of soil erosion and planning of preventive measures in and around mining areas.


Introduction
Over the past few decades, Mongolia's soils have been affected by a wide range of degradation phenomena including soil erosion, desertification, nutrient depletion and various forms of soil pollution (Batkhishig 2013;Hofmann et al. 2016;Han et al. 2021). Soil degradation in Mongolia is driven by the combined effects of climate change and anthropogenic activities including mining, (over-)grazing, agriculture, urbanization and offroad transportation (Lehmkuhl and Batkhishig 2003;Batkhishig 2013;Chonokhuu et al. 2019). About 72% of the country's land is considered to be degraded (Eckert et al. 2015;Darbalaeva et al. 2020;Liang et al. 2021) and the country has been facing severe desertification (Liang et al. 2021) which according to concerned scientists affects up to 90% of the total pastureland (Darbalaeva et al. 2020). Previously constrained to the regions bordering the Gobi desert, land degradation has in recent years increasingly affected the central part of Mongolia and to a lesser degree also extended to its north (Wang et al. 2020). The combined annual cost of land degradation in the country is estimated at around 2.1 billion USD or 43% of the country's GDP (UNCCD 2018). Batkhishig (2013) therefore noted, that an accurate estimation of soil erosion is crucial in Mongolia, but that there is limited knowledge regarding the spatiotemporal pattern of erosion and the distinction between predominant driving factors.
Over the past few decades, several scientific studies have been conducted on desertification and soil erosion in Mongolia. Climate change and anthropogenic land cover changes resulting from overgrazing, mining activities and deforestration have been documented as reasons for severe soil erosion and land degradation (Batkhishig 2013;Han et al. 2021;Liang et al. 2021). Studies based on remote sensing have recently provided an overview about soil erosion and desertification in Mongolia and identified mining activities as one of the major causes (Eckert et al. 2015). Large areas of land in Mongolia are adversely affected by mining activities, and around 4.6% of the territory of Mongolia currently covered by more than 2700 valid mining licenses (MRPAM 2020). There is well documented knowledge on mining related water, soil and air pollution in Mongolia (Knippertz 2005;Karthe et al. 2014Karthe et al. , 2017Timofeev et al. 2016;Chonokhuu et al. 2019;Kosheleva et al. 2018), but rather limited and case-specific documentation of mining impacts on soil erosion. Most previous studies have examined sediment transportation and soil erosion in agricultural areas, often combining field measurements and modelling approaches (e.g., Priess et al. 2015). Moreover, several studies have focused on the sources and impacts of sediments in Mongolian river system and identified high livestock densities along the rivers as the core challenge (e.g., Hartwig et al. 2016;Theuring et al. 2015). Onda et al. (2007) and Kato et al. (2010) measured soil erosion and sediment accumulation with radionuclide methods in the two small catchments of the Baganuur stream (0.076 km 2 ) and the Kherlen Bayan Ulaan stream (0.069 km 2 ). During the 2003 and 2004 monitoring period, annual soil erosion was estimated to be 0.37 t ha -1 year -1 in Baganuur and 0.02 t ha -1 year -1 in the Kherlen Bayan Ulaan, with precipitation playing the major role for sediment mobilization and suspended transport. Batkhishig et al. (2019) showed by the 137-Cesium radionuclide fallout method that the average soil erosion rate on a mean escarpment of the Ikh Bogd Mountains around Lake Orog was 12.57 ± 1.08 t ha -1 year -1 and found a maximum soil erosion rate is 40.87 t ha -1 year -1 . This site is not impacted by mining, but its environment is comparable to the Erdenet and Baganuur mining sites but with steeper natural mountain slopes. Research into the causes of soil erosion has identified livestock and overgrazing as a main cause (Lehmkuhl and Batkhishig 2003) but also identified changing rainfall patterns and a trend toward strong convective rainfall events as a problem (Vandandorj et al. 2017).
Mining likely plays a significant but not sufficiently studied role for soil erosion in Mongolia. The country has more than 600 deposits of more than 80 types of minerals and ranks seventh in the world in terms of mineral resources (Davaasambuu 2018). The mining sector has been developing rapidly in recent decades and it is one of the country's most important sources of foreign direct investment (FDI), revenues and exports (Baatarzorig et al. 2018). For instance, in 2014-2016, the country's gross domestic product (GDP) increased by 1.2% due to decline in commodity prices and FDI in the mining sector. In 2017-2018, the GDP growth reached 5.3% due to a recovery of raw material prices and an FDI increase. This growth trend is expected to remain relatively stable in the future (Turmunkh 2020), averaging about 9% annually over the past decade (NSO 2018).
The economy in Mongolia is heavily dependent on the mining sector and revenues from mineral exports including particularly coal, gold and copper (Dagys et al. 2020). Although the rapid development of the mining sector is important for the country's economic development, it has been identified as an important driver of land degradation (Farrington et al. 2005). The effects of mining on water resources have been studied in some detail, including the sector's significant water in water-scarce regions, and the impacts of mining on ground and surface water pollution (Timofeev et al. 2016;Batbayar et al. 2017;Jarsjö et al. 2017;Nottebaum et al. 2020), and the challenges of aligning mining and water policies have recently been addressed (Schoderer et al. 2021). A few studies have investigated land degradation in major mining areas. The total damaged land area in the mining regions of the country was estimated at around 8.1 thousand hectare (ha) (Batmunkh 2021). Mining-related pollution has been linked to bioaccumulation by plants (Kosheleva et al. 2018) and several species of freshwater fish (Kaus et al. 2017), but also to a public health hazard (Nottebaum et al. 2020), thus pointing to a nexus of mineral extraction and the degradation of the abiotic and biotic environment.
In Mongolian mining areas, most previous work on soil degradation has focused on soil pollution, but little work has been done to assess mining-related erosion or to investigate the interlinkages between soil erosion and soil pollution. It has been reported that the generation of barren lands, mine tailings and stored overburdens has led to increased erosion particularly during the rainy season (Kayet et al. 2018). Some authors have described mining as growing concern regarding its role for land degradation (Khishigjargal et al. 2015) and observed a massive expansion of soil erosion due to mining (Batkhishig 2013). For the Tuul River Basin in Northern Mongolia, Jarsjö et al. (2017) applied the WATEM/ SEDEM model to quantify the input of metal-contaminated sediments from Zaamar Goldfield mining area into the river system, and found the mining activities to strongly exacerbate natural soil erosion. The rate of soil loss around the Zaamar Goldfield mining area was reported to be 10 t km −2 month −1 , and the study found that mining-related soil losses contributed to pollutant transport in the mining area. In order to harmonize the development of mining sector in Mongolia with environmental sustainability, it is crucial to study and monitor environmental degradation and develop science-based environmental rehabilitation measures. This particularly includes soil losses due to erosion, which have not been investigated sufficiently in the past.
In this study, we aimed at the quantification and assessment of spatiotemporal pattern of soil erosion in two mining cities of Erdenet and Baganuur, with particular focus on the vicinity of the open-pit mining areas. For the time period from 1989 until 2018, we used a combination of terrestrial data and satellite-based information. The main objectives of this study were (1) to estimate soil erosion rate using the RUSLE method supported by Geographic Information Systems (GIS); (2) to classify spatiotemporal erosion risks and (3) to evaluate the spatial distribution in the two cities of Erdenet and Baganuur with particular focus on identifying the role of open-pit mining areas. The final results of the study can be used by decision-makers at the local, regional and national level to develop science-based soil erosion prevention and rehabilitation plans. This information is relevant both for mining officials and for government agencies responsible for the environmental oversight of mining operations.

Study area
The objective of this study is to provide an integrated assessment of soil degradation at Mongolia's two largest open-pit mining sites: the Baganuur lignite mine and the Erdenet copper-molybdenum mining complex, both of which began commercial operation in 1978 ( Fig. 1 and Table 1). Baganuur lignite mine is located in Baganuur district of Ulaanbaatar city (N 47°47′10″ E108°21′40″) and is 140 km east of the city center. The total district covers an area of 622 km 2 , of which 416 km 2 is pastureland, 193 km 2 is settlement area, 0.047 km 2 belongs to the road and infrastructure network and 0.14 km 2 is governmental and other land. The city of Baganuur is home to a total population of 29,342 (NSO 2018). Baganuur district is located in a semi-arid steppe region at the Southern fringe of the Khentii Mountains, and the western bank of the Kherlen River which flows from north to south in the city's vicinity (Kim 2017). The district including mining sites is situated in a wide valley bounded by small hilly mountains in the west and east (Park et al. 2020). The mean annual temperature of Baganuur district is around-2.2 °C and the coldest month is January with an average temperature of-20.7 °C and the warmest month is July with an average temperature of + 18.2 °C (Climatedata 2020). The annual mean precipitation is around 250-280 mm and about 70% of the precipitation falls during the summer season from June to August (Kim 2017;Park et al. 2020). The territory of Baganuur area is filled with Cretaceous and Jurassic sediments with Quaternary alluvial, diluvial and aeolian sediments (Park et al. 2020). The Baganuur open lignite mine deposit was estimated at 812 million t in 2014 (Otgochuluu et al. 2015). Baganuur mining site has open pits, mining equipment and offices and the mine uses water through a municipal water supply system. Open-pit mining removes naturally vegetated areas and produces a high amount of waste that leads to an increase in soil pollution and soil erosion surrounding the area. Since the operation of Baganuur mining, 337 Mbcm overburden and 1100m 3 /h groundwater usage were estimated (Park et al. 2020). The Erdenet Copper and Molybdenum complex is located in Erdenet city (N 49°03′38″ E104°03′31″) in Orkhon province, Northern Mongolia. Erdenet city was established in 1976 as a center of mining and industry, and it is the country's second-largest city. Orkhon province, which covers the two districts of Bayan-Undur (where the city of Erdenet is located) and Jargalant, encompasses a total area of 844 km 2 , 464 km 2 of this are pasture land, 210 km 2 by settlement area, 9 km 2 by road and infrastructure, 2 km 2 by water bodies, and 140 km 2 by forest. The total population of Bayan-Undur district including Erdenet city is 101,909 (NSO 2018). The city is located in the Orkhon-Selenge volcanic sediment area, which consists of alkali-rich trachyandesite, granite intrusions and Carboniferous deposits (Battogtokh et al. 2013). The climate is continental with an average temperature of 0.7 °C and annual mean precipitation of 350 mm, the coldest month is January with an average temperature − 22.2 °C, and the warmest month is July with + 16.5 °C (Climatedata 2020) and 90% of total precipitation falls from May to September (Battogtokh et al. 2013). The city of Erdenet is located between two small streams, the Erdenet and the Gavil river, which merge into the Khangal river about 7 km downstream of the city in close proximity to the Cu Mo mine in the South and its large tailing pond in the North (Solongo et al. 2018). The Erdenet mine was established in 1978 and it is one of the largest copper and molybdenum mines in the world with metal deposits estimated as 7.6 Mt of copper (Cu) and 216 600 t of molybdenum (Mo) in 1991 (Munkhtsengel et al. 2006). The Erdenet open-pit mining complex consists of one tailings pond, dams, open-pit  mines, a mine waste dump, and an ore processing mill complex. A locally well-known environmental impact of the mine is "white-dust", which is mobilized and transported by wind when parts of the tailings pond dessicate. White-dust affects soil and water quality around the tailings pond, which is problematic as this area is used by animals, herders, and residents to grow crops, and as it affects the Khangal River, which is a source of drinking and irrigation water (Battogtokh et al. 2013).

Data sources
In this study, the following data related to soil characteristics, land cover, topography and precipitation were collected to determine the spatio-temporal pattern of soil erosion.

Soil erosion estimation
In this study, the monthly soil erosion for the Baganuur and Erdenet mining areas was estimated using the Revised Universal Soil Loss Equation (RUSLE) model. It is based on the Universal Soil Loss Equation (USLE), which is considered a very practical approach that has been applied in more than 100 countries . The USLE was developed in the late 1950s for assisting widespread application including soil conservation and soil management in a wide range of settings such as cropland, urban construction areas, recreational areas, and mine sites (Renard et al. 1991;USDA 2017). Due to some limitations of the original USLE method, the revised and updated USLE (RUSLE) was developed by Renard (1991). More recently, the application of remote sensing and GIS tools has become a standard practice for mapping soil erosion risks (Bahrawi et al. 2016). The RUSLE can easily combined with GIS and remote sensing approaches which can provide some of the input data (Kulimushi et al. 2021; see Table 2). Although the RUSLE model is widely used to estimate soil erosion worldwide, the model has been used in only a very few studies in Mongolia. The equation of the RUSLE model (Fig. 2) is based on (Schmidt et al. 2016(Schmidt et al. , 2019: where A is the quantification of soil loss in a month (t ha -1 year -1 ), R is the rainfall erosivity factor (MJ mm ha -1 h -1 year -1 ), K is the soil erodibility factor (t h MJ -1 mm -1 ), LS is the topographic steepness factor (unitless), C is the cover management factor (unitless), P is the support and conservation practice factor (unitless) (Schmidt et al. 2016(Schmidt et al. , 2019Kebede et al. 2021). The annual soil loss equation was modified for calculating monthly soil loss using monthly based rainfall and vegetation data (Schmidt et al. 2019).
where A is the monthly soil loss in t ha −1 month. −1 These input parameters are described in the following.

Rainfall erosivity factor (R)
The rainfall erosivity factor is one of the RUSLE input parameters that calculate precipitation and runoff impact on soil loss (Kayet et al. 2018;Lee et al. 2021). The amount Precipitation data Soil erosion: Rainfall intensity factor-R factor of rainfall and its intensity is considered by this factor (Maqsoom et al. 2020). For calculation of the R factor, satellite-based Worldclim precipitation data (WorldClim) with a spatial resolution of 21 km 2 was used. For regions poorly covered by meteorological data, this database provides moderate spatial resolution data including long-term global monthly precipitation data (Abatzoglou et al. 2018). Average monthly precipitation around the two cities for 1986-2018 is shown (Fig. 3).
In this study, the monthly precipitation data of April to October from 1989 to 2018 were used in the calculation. The rainfall erosivity factor was calculated based on the following equation (Renard and Freimund 1994;Duulatov et al. 2021) based on monthly precipitation data. Rainfall point data were converted into raster data using the Inverse Distance Weighted (IDW) method in Arcmap software. The IDW method has been used for many studies to estimate R factor and it predicts the values in cells with unknown points (Ouadja et al. 2021).
where P is the monthly rainfall (mm). The rainfall data is satellite-based Worldclim precipitation data for 1989-2018.
(3) R = 0.0483 + P 1.61 , Fig. 2 The methodological framework of RUSLE model The soil erodibility factor (K) factor quantifies the relative susceptibility and transportability of soil from the surface due to rainplash and runoff (Kayet et al. 2018;Hateffard et al. 2021). The K factor is commonly classified based on soil texture and and soil organic matter content (Parveen and Kumar 2012), and ranges from 0.02 for the least erodible soil to 0.64 for the highest erodible soil (USDA 2017). A high amount of clay in soils has a high K value due to its easy erodibility, whereas loamy soil has moderate erodibility (Maqsoom et al. 2020). K factor values are negatively correlated with coarse textured soils and positively correlated with fine fractioned soils (Huang et al. 2022). In this study, the K factor was determined according to the soil texture type and soil organic matter content. The point data were converted into raster data using the Inverse distance weighted (IDW) method in Arcmap software. Soil samples were collected from the top layer of the surface (0-10 cm) inside and outside of the fences of the mining sites (Fig. 1). A total of 48 soil samples were collected from Erdenet mining sites in September 2020 and a total of 50 soil samples were collected from Baganuur mining sites in June 2020. Soil sampling points were georeferenced with a global positioning system (GPS Garmin 64) and are shown in Fig. 1. A total of 19 soil sampling locations were covered in a previous study by Knippertz (2005). Other sampling locations in Baganuur and Erdenet were selected randomly but represented different land covers including open-pit mining areas, urban and ger settlement areas, and the industrial zone. The simple random method was applied as it provides unbiased estimates of the mean and variance (Worsham et al 2012). In addition, it time saving and thus allows for a greater number of sampling points to be covered during the same time. All soil samples were air-dried under a normal air condition for 48 h in a laboratory of German-Mongolian Institute for Resources and Technology (GMIT) and sieved to less than 2 mm. The dry sieve method was used to determine the particle size fractions of the soil. In the process, 100 g of soil of each soil sample was subsequently placed in a sieves of 3 different diameters in order to classify it to fractions from 0.05 to 2.0 mm, 0.002 to 0.05 mm and < 0.002 mm according to USDA 2017. A shaking machine was used for 8 min with each sieve. Then the dry sieved results were used to determine soil textures based on a triangular diagram according to USDA standards.
The soil organic matter content of the samples was determined using the Loss on ignition (LOI) method. This method is one of the most commonly used methods for estimating soil organic matter content (Heiri et al. 2001). The soil samples were air dried at 22-25 °C for 48 h and particle larger than 2 mm were removed. Then, 5.00 ± 0.01 g of soil from each sample was oven dried for 24 h at 105 °C. After that, the dried samples were cooled in a desiccator, then the samples were heated at 360 °C for 2 h and then cooled below 105 °C (Schulte and Hopkins 1996).
where S w1 is the soil weight after drying at 105 °C, S w2 is the soil weight after heating 360 °C. The soil erodibility factor was then estimated based on texture and soil organic matter content (see Table 3).

Topographic steepness factor (LS)
Topographic steepness factor (LS) is used to estimate the impact of topography on soil erosion. Soil loss amount per area is directly correlated with the length of the slope (Tessema et al. 2020). The higher and lower slope has high risk of soil erosion (Kayet et al. 2018). It can be estimated by field measurement or a Digital Elevation Model (DEM) with GIS (Maqsoom et al. 2020;Kayet et al. 2018;Hateffard et al. 2021). In this study, satellite-based DEM data (Fig. 1) from the Shuttle Radar Topography Mission (SRTM) was obtained from the U.S Geological Survey's web site (USGS). The spatial resolution of the SRTM DEM data is 30 m, and its vertical resolution is 16 m or better (Mukul et al. 2017). In this study, the following equation was used to calculate the topographic steepness factor (Parveen and Kumar. 2012;Hateffard et al. 2021).  where χ is the flow accumulation, λ is the raster cell dimension, β is the slope inclination in degrees (•).

Land cover and management factor (C)
The land cover and management factor (C) is used to estimate soil loss from an area of a particular vegetation cover. Besides soil properties and slope steepness, the C factor is one of the most important drivers of soil erodibility (Panagos et al. 2015). Vegetation cover protects the soil surface from the impact of raindrops, and reduces runoff velocity, and even partial vegetation cover significantly reduces soil erosion (Boussaadi and Mauzai 2021). In this study, the C factor was calculated based on the Normalized Difference Vegetation Index (NDVI) using the equation (Durigon et al. 2014) below.
The NDVI is one of the vegetation indices, that estimates the amount of green vegetation cover in a particular landscape. It ranges from − 1.0 for to 1.0, where higher values represent higher vegetation cover and lower values represent low vegetation cover or barren land surface. In this study, NDVIs for the years 1989 to 2018 were presented as seasonal information in the following way: spring (April to May), summer (June to August) and autumn (September to October). This corresponds with the typical definition of these seasons in Mongolia (where the period from November to March can be considered as winter). Only, cloud-free or nearly cloud-free Landsat images were selected (Land cloud cover (LCC) ≤ 10%) and downloaded (via USGS) calculating the NDVI.

Conservation factor (P)
The conservation factor (P) represents the effect of precautionary measures for reducing soil erosion rate such as land treatments of countering, compaction, and constructive measures to control runoff and sediment transport (Maqsoom et al. 2020). It ranges from 0 for high-quality preservation practice to 1 for poor protection practices (Kayet et al. 2018). In this study, the P factor for the whole area of both cities was set as 1 (not influential) to calculate the RUSLE model as no prevention practices were applied in the two regions, except for limited drainage systems in the urban area around both mining regions.

R factor estimation
The average monthly value of the R factor from April to October for 1989-2018 ranges from 43.7 to 63.36 MJ mm h -1 month -1 in Erdenet area and from 34.46 MJ mm h −1 month −1 to 42.86 MJ mm h −1 month −1 in the Baganuur area. Generally, the average monthly rainfall erosivity factor was higher in Erdenet area than in the Baganuur area due to higher rainfall in Erdenet. Moreover, the highest average monthly rainfall erosivity factor was found during the vegetation growing season (June to August) in both study areas, ranging from 112.91 to 156.84 MJ mm h −1 month −1 in Erdenet and from 95.43 to 112.41 MJ mm h −1 month −1 in Baganuur (Table 4).

C factor estimation
The values of the C factors in the two mining areas differ according to the density of vegetation cover. Average C values range from 0.02 to 0.073 around Erdenet city and from 0.03 to 0.068 around Baganuur city during the period from 1989 to 2018 (Table 5). The value of the C factor were higher in the vicinity of the tailings pond and openpit in both mining areas (Fig. 4) indicating higher vulnerability to erosion. The C factor decreased around both mining area in the growing season, which corresponds to the season of highest precipitation across Mongolia.
When rainfall was identified the dominant cause of temporal variations in soil erosion, the C factor based on the vegetation index NDVI was identified as a good predictor of spatial erosion pattern. On the one hand, the C factor reflects the seasonal pattern of vegetation growth. On the other hand, it clearly identifies barren lands that are a consequence of mining operations. The map of spatial distribution of NDVI is shown in Fig. 5 and Fig. 6 to show the difference between the highest and lowest precipitation during the peak season of vegetation cover (June to August) during the study period 1989-2018 in both areas.
The spatial patterns of NDVI in wet and dry month and years clearly show that vegetation cover increases when precipitation is higher. However, the results of soil erosion showed that soil erosion rate during the study period was higher in the peak vegetation season from June to August and lower in autumn and spring. Rainfall amount is the most important factor in increasing soil erosion rate. This result is consistent with previous studies (Vandandorj et al. 2017;Schimdt et al. 2019;Shojaei et al. 2020;Duulatov et al. 2021).

K factor estimation
The soil erodibility factor was defined based on SOM content and soil texture. The estimated organic matter content in Erdenet area was 0.16-0.41% and in Baganuur area  0.05-0.36%. The soil textures of the sampled soils were generally sandy soil (92% of the total samples in Erdenet and 94% in Baganuur), with some loamy and sandy loam textures in the Erdenet area, and loamy sand and sandy loam textures found in Baganuur. These findings agree with literature which describes the soils of Baganuur as sands consisting of quartz, albite and microline (Park et al. 2020), and fine sand and some silt as the predominant soil textures around Erdenet (Battogtokh et al. 2013).
Regarding to the SOM content and soil texture, the K values were selected from Table 3 as 0.12-loamy sand; 0.25-sandy loam; 0.27-sandy soil. The spatial patterns of the K factor were created in ArcMap software using the IDW technique.

LS factor estimation
The LS factor describes the combined effects of slope gradient and slope length (Maqsoom et al. 2020). In this study, the LS factor was calculated using Digital Elevation Model (DEM) data. The results of the LS factor ranged from 0 to 9.6 around the Erdenet mining area and 0-5.7 around the Baganuur mining area. In both areas, higher values of the LS factor were observed at higher altitudes due to longer and steeper slopes (Duulatov et al. 2021).

Assessing soil erosion for 1989 to 2018
The average monthly soil erosion rate was calculated in three different seasons: for spring (April to May), for summer (June to August), and for autumn (September to October) ( Table 6). The average soil erosion rate was higher in Erdenet area than in Baganuur area due to higher rainfall in Erdenet.
The results of soil erosion modelling from 1989 to 2018 show that both study areas are susceptible to soil erosion, especially during periods of higher precipitation. Previous studies have found that the average soil erosion due to water erosion in summer is higher than soil loss in winter and autumn (Schmidt et al. 2019) and precipitation erosivity is the most important factor influencing soil loss (Schmidt et al. 2016). The highest soil erosion rates in both study region occurred when precipitation peaked between June to August (Table 7), which in Mongolia typically for 85-90% of the annual precipitation (Batkhishig 2013).

Soil erosion rate in Erdenet area
Large parts of the total area (90.71%) experienced only few levels of erosion (0-0.1 t ha -1 month -1 ) when considering the entire 30 year observation period. In general, erosion levels increased during the summer months due to higher precipitation. Erosion rates above 2 t ha -1 month -1 were calculated only for the summer season, with limited areas even experiencing above 5 t ha -1 month -1 . The highest soil erosion rate (8.31 t ha -1 month -1 ) was observed in July 2018  and the lowest soil erosion rate (0.05 t ha -1 month -1 ) was calculated for April 2010. These events correlated with the highest rainfall (120.24 mm) in July 2018 and the lowest rainfall (8 mm) in April 2010 of the observation period. The comparison of soil erosion rate in the years with the highest and lowest rainfall in Erdenet is shown in Fig. 7 to illustrate the spatial distribution. Erosion rates were calculated in July 2018 for the approximate area in the vicinity of different land uses such as open-pit mining area, ger area, urban area, tailing pond and industrial area (Table 8). For instance, the calculation showed that a 20 ha area near the open-pit area experienced erosion rates of 1-2 t ha -1 month -1 . The highest soil erosion rates were found near tailings pond and industrial areas in July 2018.

Soil erosion rate in Baganuur area
The results of the average monthly soil erosion rate were quite similar of those of the Erdenet area, with slightly lower values calculated for Baganuur. This can be explained by the higher rainfall in the Erdenet area.
In the Baganuur area, higher soil erosion rates were observed during June to August, with 0.21% of the total area experiencing erosion of more than 2 t ha -1 month -1 . The average monthly soil erosion rate was lower in April to May and September to October (Table 7) when a threshold of 2 t ha -1 month -1 was not exceeded. In July 2018, i.e. the wettest month of the observation period, 0.8% of the total area experienced intensive erosion of more  than 2 t ha -1 month -1 . The highest soil loss rate (9.46 t ha -1 month -1 ) was found in July 2018 and the lowest soil erosion rate (0.056 t ha -1 month -1 ) was found in October 2018, corresponding to the highest amount of rainfall (134.65 mm) in July 2018 and the lowest amount of rainfall (4.46 mm) in October 2018. The comparison of the map with the highest and the lowest soil erosion rates is shown in Fig. 8. In general, a higher erosion rate was found near the open-pit mine and at higher elevations. Erosion rates were calculated in July 2018 for the approximate area considering different land uses (Table 8).
The estimated highest soil loss rates (> 5 t ha -1 month -1 ) as well as their area and the amount of eroded soil loss during 1989-2018 in both areas are shown in Table 9. In general, higher soil erosion rates were observed in the Erdenet region than in the Baganuur region. During the study period, the highest erosion in both areas was observed in July 1998, 2013 and July 2018, which is directly related to rainfall.

Discussion and conclusions
This study revealed that in the semi-arid steppe environment of two Mongolia's largest mining sites, the open-cast lignite mine of Baganuur and the open-pit copper-molybdenum mining complex of Erdenet, the spatio-temporal pattern of soil erosion are driven by precipitation pattern and anthropogenic impacts from mining. Some of the highest soil loss rates were found at the extraction sites, but also near mining-related infrastructures such as the tailing pond of Erdenet. The lower soil erosion rates in the vicinity of the open-pit in some respects could be explained by differences in vegetation cover, for instance, a higher vegetation cover found in surrounding areas of the open-pit is explained by lower anthropogenic influences in that area, such as fences. Slightly higher soil loss rates were found in urban and ger settlement areas in Baganuur than in the same settlement areas in Erdenet. This can be attributed to topographic differences between the settlement areas. For example, the urban and ger areas in Baganuur are at a lower elevation and have more opportunities to receive water through precipitation, flooding and snowmelt from the higher elevated areas, and there are several stream channels that run through the settlement (Fig. 1). Soil erosion rates were higher in the Erdenet area as compared to Baganuur, which is mostly due to higher precipitation. The highest monthly erosion rates were 8.31 t ha -1 month -1 for Erdenet in July 2018, and 9.46 t ha -1 month -1 for Baganuur in July 2018, which corresponded to the 30-year monthly rainfall maxima in both areas. The exceptionally high rainfall in 2018 can be attributed to El Niño conditions (UNESCAP 2019). In Mongolia, intensive flood events were recorded in July 2018 in several parts of the country, particularly central and northern areas (IFRC 2018). In the light of a trend towards a higher frequency and intensity of extreme precipitation events in the Mongolian steppe zones (Vandandorj et al. 2017), it is likely that such erosion rates will become more frequent in the future. From a soil conservation perspective, it is important to note that apart from steep natural mountain slopes, the highest erosion rates were calculated for the mining sites. Precautionary measures for limiting soil erosion should be considered during mining operation. In the long run, the rehabilitation of abandoned mining sites, and particularly the restoration of surface vegetation, can significantly help in reducing levels. Apart from these specific findings, the results of this study show that a combination of the RUSLE model and a GIS-based approach that integrates ground-based data (in our case, soil texture and soil organic matter to calculate the soil erodibility factor K) and remote sensing-based information (in our case, NDVI as a proxy for the land cover management factor C, the SRTM-DEM for calculating the topographic steepness factor LS, and a global precipitation dataset for deriving the rainfall intensity factor R) can be useful in operationalizing a soil erosion assessment over relatively large areas and long time periods. It is obvious that results should be interpreted with some caution, as all of the above-mentioned ways for quantifying the input parameters of the RUSLE model have their limitations; in our case, the soil erodibility factor was based on limited number of samples, and each of the RS-based datasets has some limitations (e.g. in terms of spatial or horizontal resolution). Nevertheless, the described approach can be a time-saving and cost-effective technique for the operative monitoring and prediction of soil erosion, including the specific situation of mining areas.

Appendix 1
Monthly soil erosion from 1989 to 2018 rate in Erdenet and Baganuur cities. tion to the German Academic Exchange Service (DAAD) for awarding a scholarship for the PhD sandwich program of RWTH Aachen University, Germany and the German-Mongolian Institute for Resources and Technology (GMIT), Mongolia. Also, great appreciation to Mr. Malte Giesenow who organized administrative and other related work of the project.
Funding Open Access funding enabled and organized by Projekt DEAL. This work was funded by Deutscher Akademischer Austauschdienst.

Data availability
The authors confirm that the data supporting the findings of this study are available within the article.

Conflict of interest
The authors have not disclosed any competing interests.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.