China's Loess Plateau is one of the regions in the world that is most seriously affected by soil erosion (Feng et al., 2010; Yuan et al., 2019*Soil erosion is affected by various factors, including rainfall, land use, and variations in the catchment surface, such as terrain, topography, and soil type (Clark, 2000; Ochoa et al., 2016; Wang et al., 2016; Rajbanshi and Bhattacharyaa, 2020; Wang et al., 2020*The vast expanse of continuous and widely spread loess on the Loess Plateau, the geomorphology of thousands of gullies, and the serious soil erosion have attracted the attention of domestic and foreign scholars (Li et al., 2020*Geomorphology is the manifestation of internal process, and it is the main basis for the research of its formation causes (Liu, 2016*The gullies on the Loess Plateau are shaped by the unique geographical environment and long-term severe soil erosion. Severe soil erosion has brought great disasters to the natural environment, economy, and society of the region (Zhang, 1993; Li et al., 2008; Gao et al., 2016; Li et al., 2019*Therefore, in-depth research on the landform characteristics and macroscopic spatial differences of the Loess Plateau will reveal its formation mechanism and development trend (Xiong et al., 2021), thereby further clarifying the spatial variance pattern of soil erosion, predicting the development process, and guiding the ecological restoration (Liu et al., 2017) and development of the Loess Plateau. The regional comprehensive management and development (Gao et al., 2016), and the implementation of the strategic policy in harmonious integration of the development of the National Western Development and the environment (Magdoff et al., 2011; Zhang et al., 2012) have important theoretical significance and application prospects.

Unique loess surface morphology must contain clues of the evolution of the loess morphology for a long time (Zhou et al., 2010*The shoulder-line is a topographic structure line that reflects the morphological characteristics of the loess landform best. The inter-gully land (positive terrain) and the gully land (negative terrain) (Luo 1956; Song et al., 1990) divided by the shoulder-line are staggered, thereby forming a unique loess landscape. In the research of Zhou et al. on the 1:1 million digital landform classification system of China's terrestrial landforms, they pointed out that “positive and negative terrains, as one of the five most basic interactive relationships in the classification of landform types, is a unity of the interdependence and role in the formation and evolution of landforms, the frequency of their replacement, the degree of contrast and structural characteristics are very crucial” (Zhou et al., 2009*The binary combination of PNTs has always been an important starting point for studying complex geoscience systems. However, in many geoscience studies, the qualitative description of PNTs is redundant with quantitative research. The reason is that PNTs' boundary is not obvious. In the severe gully erosion areas of the Loess Plateau (LPGE), the problem of the boundary between positive and negative topography can be solved effectively. The shoulder-lines, which are the natural boundary between the PNTs of the loess are more developed because of the vertical cleavage and collapsibility characteristics of loess (Zhou et al., 2010*Therefore, the shoulder-line is an excellent starting point to study the spatial differences of loess morphology and the evolution mechanism of geomorphology. Regarding the study of the loess landform along the shoulder-line, from the perspective of the whole watershed, some predecessors mainly used the inter-gully and gully divided by the shoulder-line to study the topographic and landform characteristics of the two (Jing, 1986; Liu, 1993), the classification system (Luo, 1956), the evolution of loess landforms, and the formation of gullies (Xiong et al., 2016; Lv et al., 2017) and to provide scientific basis for soil and water conservation in the Loess Plateau (Jing, 2004*From the perspective of the basin, most of the existing studies have focused on the areas below the shoulder-line, such as gully density (GD) (Tian et al., 2013), gully point characteristics (Jiang et al., 2013; Zhu et al., 2014) and gully retreat (Vanmaercke et al., 2016), gully network (Wu et al., 2017; Yang et al., 2017), gully profile group (Cao et al., 2019), gully cross section characteristics (Zhou et al., 2020), and the relationship between gully length and area (Li et al., 2017*The research on the positive topography and spatial combination of the PNTs is slightly insufficient. The research on positive topography mainly includes watershed profile (Tang et al., 2015), shallow gully erosion (Zheng et al., 2016), and soil erosion (Wang et al., 2013*The latter research is mainly in line with on the study of the spatial variations and evolution of the loess landforms based on the spatial change characteristics along the gully (Xiao et al., 2007; Zhou et al., 2013; Zhou et al., 2019) and the discussion of the relationship between water and sand (Zheng et al., 1998; Chen et al., 1999).

The above studies have laid the foundation for the development of loess landforms. However, shortcomings, which are mainly reflected in the fact that most studies have segmented the binary combination of the PNTs, are explored in a small area, and lacked the distribution and difference of erosion in large-scale areas, are still encountered. Academician Liu Dongsheng mentioned in the article “The history, current status and future of loess research in China” (Liu et al., 2001) that one of the problems that need to be solved urgently in the study of the Loess Plateau is: “The law of erosion of the loess landform and the time law of its formation we have not studied enough.”

Field surveys and remote sensing interpretation work have found in the LPGE that the distribution of PNTs and their combined morphological characteristics have become the most intuitive and external manifestation of the difference regularity of the loess morphology, which is an important clue for studying the spatial difference of loess morphological types and is also an effective carrier to express the macro-differentiation pattern of loess landform. The evolution of the loess landform is precisely the positive terrain being constantly eroded and the negative terrain constantly expanding. This research begins from the true characteristics of the natural surface. The most intuitive and unique shoulder-line of the loess landform landscape was used as the starting point. Along the line, the ratio of the gully area to the inter-gully area is used as a macro statistical indicator to measure the extent of the development of gully erosion in the region. Its spatial difference characteristics have always been the topic of landform landscape and soil erosion research, and the influences of precipitation, vegetation, land use types (LUTs), and loess thickness (LTs) under the different geographical zoning types are also investigated. This research is a useful exploration of the digital terrain analysis of the Loess Plateau from appearance to mechanism and has great significance to the soil and water conservation work of the plateau.

The LPGE is located in northwestern China and belongs to semi-arid and semi-humid climate with serious soil erosion and fragile ecological environment (Li et al., 2017) (Figure 1*It is also the main source of sediment in the Yellow River and the key region regarding soil and water conservation in China (Chen et al., 2004; Fu et al., 2017*The study area mainly refers to the core region of the obvious gully erosion in the Loess Plateau, which is characterized by gully development, typical loess cover, various types of landforms, no mountain bedrock exposure, and no plains (Zhou, 2011*The basic geographic data in Figure 1 were obtained from NGCC (http://www.ngcc.cn/ngcc/).

The region starts from the east of Liupan Mountain in the west to the west of Lvliang Mountain in the east, to the northern boundary of the Guanzhong Plain in the south and to the Mu Us Desert in the north (34°19°‒39°39°N, 106°09°‒112°32°E*It covers an area of 144,220 km² with an altitude ranging between 332 m and 2366 m above sea level. The types of regional geomorphology from south to north follow the loess tableland (LT) and rugged tableland areas in the south, the loess ridge (LR) in the central part, the loess hill (LH) in the middle and north, and the northern sandstorm transitional area (Zhou et al., 2010*At the same time, this area is the main depositional area of loess, and the thickness of the loess covered ranges from 50 to 300 m, which has obvious regional differences.

The high spatial resolution image from Google Earth was used as the data source. On the basis of ensuring data scientificness, completeness, and representativeness of the experimental areas, the data still need to be based on an area that is appropriate (Tian et al., 2013), has stable and typical geomorphological features (Tang et al., 2015; Zhou et al., 2009), and is reasonably distributed in the region to extract 172 sample areas' shoulder-line through a combination method of automatic and visual interpretation. The LT, LR, and LH areas should also be covered. In addition, due to limited data, this study used the 5-m resolution DEM data that were provided by Shaanxi Administration of Surveying, Mapping, and Geo-information. Moreover, the topographic factors that GD and hypsometric integral (HI) of some sample areas, are calculated. The sediment modulus (SM) data came from hydrological stations, and the data of normalized difference vegetation index (NDVI), precipitation, and LUTs were provided by the Resources and Environment Data Center of the Chinese Academy of Sciences. The loess thickness (LTs) data were compiled by the Second Hydrogeological Party of Bureau of Geology and Mineral Resources of Shaanxi Province.

In the LPGE, the shoulder-line has obvious boundaries and divides the complete watershed into the gully area and the inter-gully area. Huge differences can be noted in defining method and morphological characteristics along the shoulder-line in LT and loess hilly-gully area, which are the main landform types in the study (Figure 2*According to a large number of field investigations, gully development in the LT and broken tableland areas is relatively late, and a gentle slope transitional area between the PNTs of the LPGE is observed, the slope of which is roughly in the range of 5°-15°. In the hilly-gully area, the gullies for the hilly and hilly-ridge areas are relatively mature, and the slope between the tops of the elongated ridges (liang) and the gullies have larger slopes in the range of 15°-25° (Xiao et al., 2007), and the slope change characteristics are more evident.

At present, two main methods can be used to obtain the shoulder-line. One is based on high-resolution remote sensing images by using visual interpretation to extract manually. This method has high accuracy but low efficiency. The second is the automatic or semi-automatic extraction of the shoulder-line based on DEM (Xiao et al., 2007; Zhou et al., 2013; Yan et al., 2011) and image segmentation methods (Wang et al., 2015; Yang et al., 2019*Zhou and Xiao identify the candidate points of the shoulder-line based on slope turning (Figure 3*This method is derived from the definition of the shoulder-line itself, which has high accuracy and efficiency but poor continuity. Therefore, our study selected this method to obtain the candidate points along the shoulder-line based on DEM. Subsequently, they were superimposed with high-resolution remote sensing images, and visual interpretation was used to obtain a continuous and accurate shoulder-line of each sample. The area of samples selected in this experiment tends to be approximately 10 km².

In the LPGE, the positive and negative terrains represent the original accumulation of loess surface morphology and the surface morphology shaped by modern erosion, respectively. The shoulder-line is the boundary that divides the loess landform into the PNTs dominated by inter-gully and gully areas. Its dynamic change reveals that the gully expands continuously, the gully area grows gradually, and the landform is being eroded continuously (Figure 4).

The SND is used to characterize gully developmental morphology in a whole watershed.

The gully expansion is that the positive terrain is constantly eroded by the negative terrain in horizontal dimension. In other words, SND reflects the area ratio of the water erosion gully to the slope surface area without obvious water erosion (Zhou et al., 2011; Yang et al., 2019) and is expressed as:

where SN is the projected area of the negative terrain area on the horizontal dimension, and SP is the projected area of the positive terrain area on the horizontal dimension. When SND is greater than 1, the negative terrain is dominant. When SND is equal to 1, the area between PNTs are equal, and the development of watershed is at a medium level. The larger the value is, the closer the shoulder-line along the watershed line, the stronger the degree of traceable erosion and lateral erosion, and the more severe the historical sediment yield will be.

Geostatistical analysis means to perform the best unbiased interpolation estimation for the sample data based on a large number of sample points by considering the spatial position of the sample points and the distance between the samples. The basis for geostatistical analysis is Tobler's first law of geography, which states that the closer the distance is, the more similar the sample will be. Numerous interpolation methods exist for spatial interpolation through geostatistical analysis (Li et al., 2020*The Kriging interpolation method can reflect various terrain changes. It is widely used in the field of geoscience. Kriging interpolation requires a certain number of sample points that meet the normal distribution, and a spatial correlation is observed between the sample points. In this study, the large number of sample points and the spatial correlation between them met the requirements of the Kriging interpolation method. Therefore, it was used to study the spatial differences of SND. Before conducting spatial interpolation, we performed exploratory analysis for the variables. For the variables that are unsatisfied normal distribution, we conducted the function transformation to satisfy the variables' normal distribution in the histogram. Meanwhile, in the exploratory analysis stage, we found that the variables obey the second-order stationary that is needed for Kriging interpolation through the spatial autocorrelation detection of the semi-variable function. Generally speaking, ordinary Kriging needs to obey the second-order stationary hypothesis, whereas the universal Kriging is fit for the data that has a clear trend. We adopt universal Kriging interpolation because of the obvious spatial trend for the SND. Then, a SND spatial interpolation map was obtained.

The landform types of the eight key test areas basically represent the typical loess landform of different development stages, and along with it, the degree of landform development in these test areas is also quite different. Therefore, the comparison of the SND in these regions provides an insight into the recognition of the regularity of landform development in the LGPE. In this study, eight key test areas, such as Shenmu, Jiaxian, Suide, Yanchuan, Yan'an, Ganquan, Yijun, and Chunhua (Figure 5), were selected from north to south to calculate the SND. According to Table 1, as the degree of landform development continues to mature from south to north, the SND presents a continuous distribution pattern (Zhou et al., 2010*In the Yan'an area, which represents the loess hilly-ridge landform, the SND has reached 0.97 and continues to increase after a low value in Suide area. The supervision area of the Loess Plateau soil and water loss and the ecological construction policies, such as “returning farmland to forests,” and the comprehensive river basin management have been effectively practiced and have contributed to significant regional ecological environment restoration (Fu, 2014*In Jiaxian and Shenmu, the value has exceeded 1, indicating that the negative terrain area exceeds the positive terrain area, the surface erosion has been very serious, and it has become a key concern for water and soil conservation region. In the southern LT and broken tableland areas, the surface erosion is relatively weak, a large area of positive topography is observed. Meanwhile, in the loess hilly-ridge areas where the loess landform is more mature, the surface is more fragmented, and the erosion is more serious.

Existing studies have shown that the systematic exploration of the morphological characteristics, structural types and spatial distribution of the eroded landform is the key to a comprehensive understanding of the loess landform erosion morphology and its evolution (Arabameri et al., 2019; Liu, 2017*This research will explore the spatial differences of SND from two aspects.

The landform evolution of the LPGE has a certain regularity in the spatial distribution. From south to north, the characteristics of landform development are shown as the LT and rugged tableland areas in the south, the LR in the central part, the LH in the middle and north, and the northern sandstorm transitional area (Zhou, 2011*Based on eight key test areas (Figure 5 and Table1) and 164 general watershed samples (Figure 6 and Table 2) in the LPGE, the universal Kriging interpolation method in the ArcGIS geostatistical analysis is used to simulate the surface spatial trend of the SND, and the trend map is obtained. In Figure 7, SND is gradually increasing from southwest to northeast. The high values of SND appear in the northeast of the study area, especially in Shenmu of Shaanxi, Baode to Xingxian in Shanxi. In the northern area of Yanhe River basin, the value is above 0.87, and the regional geomorphology is very mature, whereas the low value areas of SND are mainly distributed in the LT and rugged tableland areas in the south of the study area. In the inclined elongated ridge area in Guyuan of Ningxia and Pingliang of Gansu, the SND is less than 0.43, the gully development is not obvious, and these areas are a large area of plains. The trend of SND is gradually transitional, and it coincides with previous studies on loess gully erosion landform classification (Zhou, 2011; Liu, 2017; Zhang, 2013; Li et al., 2020), indicating that the spatial pattern of SND in the loess gully area shows good consistency with the distribution of loess landform division.

The middle reaches of the Yellow River flow through the core area of LPGE. In this study, the Wuding River, Yanhe River, North-Luohe River, and Jinghe River from north to south are selected as the study areas. The flow length (FL) of every small watershed with its water outlet as the direction axis is used to explore the different characteristics of SND in each watershed.

The FL is a distance of the starting point of each small watershed to the outlet point along the river network, which makes the centroid of each watershed as the starting point, and its unit is kilometers.

The Wuding River belongs to the first level basin of the Yellow River. The landform in the basin is dominated by LR, which is seriously eroded and mature. As shown in Figure 8a, the value of SND in the basin continues to decrease with the increase in FL, which also indicates that the erosion in the lower reaches of this basin is more serious than that in the upper reaches. In addition, the northern part, which is far from the water outlet, belongs to the sandstorm transitional area, which is an alternate zone of water and wind erosion, and the erosion is relatively strong.

The Yanhe River, which flows from Ansai to Yanchuan in Shaanxi, also belongs to the first level of the Yellow River Basin and is a transitional area of loess hilly-ridge landform. The SND value of each basin with the increase in FL decreases more smoothly in Figure 8b, that is, the longer the FL is, the weaker the erosion will be.

Three landform types can be found in the North-Luohe River basin and Jinghe River basin, and they all belong to the second level basins of the Yellow River Basin. Existing studies found that most of the landform types at the outlet of the basins are LT and rugged tableland areas. In comparison with loess hilly-ridge areas in the upper and middle reaches of the basins, the LT areas are covered with large patches of positive terrain, the development of this landform is very young, and the erosion is relatively weak (Li et al., 2017*Thus, the value of the SND is relatively small and is the same as those in Figures 8c and 8d. From LT to LR and then to LH, the value of the SND increases with the FL. Such findings also explain the coupling between the developmental stage of the landform types and the SND to some extent.

In existing studies, some indicators, such as GD, HI and SM, have been proven to be able express the evolution stage of the landform better. GD represents the ratio of the gully line length to the watershed area, HI represents the relationship between the elevation and the area in a watershed, and SND represents the ratio of the area of positive terrain to the area of negative terrain in a watershed. From the perspective of geomorphology, these indicators all reflect the erosion and development of loess landforms. Therefore, they are correlated to some extent. Theoretically, for the younger landform area, the bigger the HI is, the smaller the GD will be, and the smaller the SND is. However, the notion is not completely consistent with our conclusion. The SND has a strong correlation with GD (R²=0.6674, P<0.01) (Figure 9a), and SM (R²=0.720, P<0.01) (Figure 9c), whereas the correlation with HI (R²=0.4656, P<0.01) (Figure 9b) is not so strong. Considering the accuracy and limitation of 5-m DEM, 90 samples were selected to calculate HI based on Zhu et al*2013), and GD was calculated on the basis of Tian et al*2013), SM data were obtained from 80 hydrological stations.

The HI is a three-dimensional index constructed on the basis of watershed elevation and area (Zhou et al., 2019*It is a macroscopic index and can be used to divide the development stage of regional geomorphology, revealing that the eroded material in the watershed accounts for the total. The SND is a two-dimensional plane index that represents that the degree of the positive terrain is eroded by the negative terrain and has nothing to do with elevation.

The correlation between SND and GD is higher. GD is a topographical feature factor that can reflect the development of gully in the LPGE, compared with SND based on a two- dimensional plane. The latter can not only indicate the direction in which the gully extends, but also the degree of lateral widening in multi-direction. In addition, GD is affected by terrain and DEM resolution, the extraction of gully line often needs to adopt appropriate thresholds, which are subjectively affected by human activities and are uncertain. Moreover, a large number of mixed gullies with inherited gullies and water erosion gullies in the LPGE were observed to have a greater influence on the GD.

SM is a comprehensive index. There is a good correlation between SM and SND (Figure 9c), which is also an internal performance of the development of landform erosion. However, the SM data are often due to the considerable area of the hydrological stations, the large difference in geographical conditions in the basin, and the complex erosion types in fact cannot reflect the erosion on different types of landform units or land use units. Furthermore, the observation values of the hydrological stations correspond to the amount of suspended sediments, and the load is not counted. Therefore, SND can be used as a reliable indicator to measure the degree of regional erosion.

Soil erosion on the Loess Plateau is restricted by many factors. Previous research has shown that the major natural factors affecting soil erosion include geology and topography, climate, soil, and vegetation. In this research, we used LTs, annual precipitation (AP), NDVI, and LUTs as influencing factors in different geographical divisions to study their effects. Geographical division map is provided by National Earth System Science Data Center, National Science & Technology Infrastructure of China (http://www.geodata.cn), including sandstorm transition (SST), loess wide gully (LWG), sand-covered loess hill (SCLH), loess inter-montane basin (LIMB), LH, LR, LT, loess rocky hill (LRH), bed-rocky mountains (BRM), and river alluvial plain (RAP*Considering that obvious gullies in SST, LIMB, BRM, and RAP are few, our study does not consider these types.

A huge area of continuously and widely spread loess can be found on the Loess Plateau, especially in Dongzhiyuan, Luochuanyuan, and the inclined elongated ridges of Guyuan in Ningxia. The LTs can reach more than 150 m. Figure 10a shows that the influence of LTs on SND is not obvious, and existing studies believe that the LTs have a greater impact on the vertical erosion index (Zhou et al., 2019), while SND is a sign of erosion in horizontal dimension. Field research shows that the sedimentary loess of the north of Qingyang in Gansu and south of Wuqi in Shaanxi has been undergoing a flowing water erosion controlled by Ordos ancient topography. The degree of erosion is greater, thereby forming a deep and slender dendritic gully extension. The surface erosion is not obvious, but the slope is large, which is attributed to the prime period of gully development.

Precipitation is an important factor that affects soil erosion and landform development (Zhou et al., 2019*By studying the relationship between AP and SND (Figure 10b), we found that the SND first increase and then decrease with the increment in AP. Rainfall and slope runoff are used as the main motivation of soil separation and sediment transport (Feng et al., 2010*However, high-intensity vegetation can be found in the LRH and LT areas, the total amount of rainfall is large but not concentrated, and this vegetation can not only directly intercept rainfall and reduce the amount of rainfall input into the forest land but can also weaken the energy of rainfall and convert the energy into penetration energy, and reduce soil erosion (Liu et al., 1994*Thus, the SND decreases with rainfall in these regions. Apart from of this, we believe that precipitation intensity and previous soil water content also have more important impact on erosion. Due to the limited time of the data in this paper, we hope to study these influences on the SND in the future.

The role of vegetation in preventing and controlling soil erosion has been widely discussed by scholars at home and abroad. Since people's research on soil erosion began, they have realized important role vegetation plays in preventing and controlling soil erosion (Chen et al., 1996; Yang et al., 2014*Vegetation affects soil erosion mainly by reducing the energy of rainfall so that the surface is not directly hit by rainfalls, increasing ground runoff infiltration time, improving soil physical and chemical properties, and enhancing soil erosion resistance (Carroll et al., 2000*NDVI is widely used as a vegetation index for monitoring and expressing the conditions of vegetation because it is closely linked with vegetation cover density, biomass, and leaf area index (Tucker et al., 1979*In our study, we use the NDVI data to study the influence of vegetation on SND. As shown in Figure 10b, with the gradual increase in NDVI from north to south, the SND first increases and then decreases when the vegetation coverage reaches a certain threshold, which is in good agreement with previous studies (Li et al., 2015; Zhou et al., 2019*However, their special values in our study need to be verified further.

Regarding the impact of LUTs on SND, we used the zonal statistical analysis method to calculate the land use categories under different geographical divisions. This study only shows that the LUTs in each zone accounts for more than 10% of the total area of the class, and then the average SND of each zone is calculated. As shown in Figure 11, except for the LRH area, dry land accounts for about 40% in other zones (Table 3), and low coverage grassland is only distributed in loess hilly-ridge and wide hilly-gully areas, the SCLH in the study area is relatively small. The more grassland with low coverage, the more serious the erosion. The possible reason is that this grassland type lacks water and is sparse, and the surface is dry; hence, the viscosity of the loess particles is low with the occurrence of runoff due to the lack of intercepting precipitation. Erosion occurs easily, and the surface loess is affected by runoff, thereby intensifying the erosion. The research is consistent with Lei X (2020), where in high-coverage grassland and woodland areas have weak erosion, while low coverage grassland has strong erosion.

This research uses a combination of semi-automatic and visual interpretation method to extract the shoulder-line, constructs SND, and uses geostatistical analysis method to explore the spatial differences of SND and reveal the spatial pattern of loess landform deeply and systematically. It is also of great significance to the planning of soil and water conservation on the Loess Plateau.

(1) The SND can be calculated formally.

(2) SND shows a pattern of gradual increase from southwest to northeast, and high values appear in Shenmu of Shaanxi, Shilou of Shanxi, and northern Yanhe River, whereas the low values are mainly distributed in the southern LT and the inclined elongated ridge area of Pingliang in Gansu and Guyuan in Ningxia. High value areas are the areas that are most seriously affected by soil erosion and sediment yield in the LPGE. The erosion modulus is above 9000 (t·km^‒2·a^‒1), and it is as high as 13,000 (t·km^‒2·a^‒1) in Hequ and Shilou. Once again, this result confirmed the importance of soil and water conservation to soil erosion.

(3) The first level basins of the middle reaches of the Yellow River, such as the Yanhe River and Wuding River, have the worst erosion. With increase in FL, the SND decreases in the first level basins and increases in the second level basins. The SND is an indicator of landform development types. In the watershed with single landform type or similar development stage, the influence of FL is more significant. In the large watershed with complex geomorphic types, the influence of FL on the SND is insignificant because of the difference of landform development stages.

(4) The SND, highly coupled with SM, has an obvious correlation with GD and a middle correlation with HI. Therefore, the SND of the LPGE can be used as an important parameter to measure the intensity of soil erosion and the stage of landform development. As for the mechanism factor analysis, the relationship between LTs and SND is not obvious, but SND increased first and then decreased with the increase of AP and NDVI in each geographical division, and we found that the TUTs of low coverage grassland have greater erosion potential. Soil erosion is restricted by many factors, it is expected that detailed studies will be carried out from the perspectives of rainfall intensity, aspect and so on in the future for providing scientific advice and references for soil and water conservation on the Loess Plateau.