Analysis of Spatial and Temporal Characteristics and Spatial Flow Process of Soil Conservation Service in Jinghe Basin of China

The supply and demand of ecosystem services and their mutual feedback are important for the formulation of basin ecological environmental policies. Simulation of the spatial flow of ecosystem services can clarify the division of areas and can support policy development. This paper takes the Jinghe Basin in the Loess Plateau of China as the case-study area to simulate the spatial flow of soil conservation service on different scales. The results showed that (1) soil erosion situations in Jinghe Basin improved overall, with a trend of first aggravating and then recovering between 2005 and 2015; (2) the amount of annual soil conservation in the basin accounted for more than 75% of the potential soil erosion and showed a trend of first increasing and then decreasing; and (3) using digital elevation model (DEM) data and ArcGIS software, the experiment divided the basin into sub-basins (58 in total) and hydrological response units (HRUs) (e.g., 2181 HRUs in sub-basin #1), which were used to quantify the spatial flow direction and the corresponding amount of soil conservation service on the “HRU—river-sub-basin” scale. The divided supply and demand helped quantify the spatial flow pattern of soil conservation services from HRU to the sub-basin.


Introduction
Ecosystem services [1] are the environmental conditions and utilities formed by ecosystems and ecological processes critical for human survival and development. They are closely related to human life [2,3]. Ecosystem services [4] are generally divided into supply services, regulation services, culture services, and support services. Soil conservation service refers to the erosion control ability of the ecosystem to prevent soil loss and to store sediment and belongs to the scope of regulation services. The assessment of soil conservation services is often based on the revised universal soil loss equation (RUSLE) [5][6][7][8][9]. The RUSLE model is an empirical slope soil erosion prediction model formed by a large number of soil erosion experiments and quantitative observation data, and on this basis, many scholars have applied this model in a lot of research on soil erosion [10][11][12]. Its reasonable structure and strong representative parameters have been widely studied and applied in China [13].
An ecosystem service is used by humans, passed from one area to another so that an ecosystem service flow can be formed. Ecosystem service flow is a vital link connecting the service supply and demand of ecosystem services. Initial research on ecosystem service flow was research on spatial characteristics of the supply services [14]. Later, with an

Data Source and Processing
The research datasets include normalized difference vegetation index (NDVI) data, meteorological data, DEM data, soil data, land-use data, sediment data, and socioeconomic data. The data source and main processing processes are summarized as follows: NDVI data was downloaded from the National Aeronautics and Space Administration (NASA) website (https://www.nasa.gov/), using the MODIS Reprojection Tool (MRT) provided by NASA for batch processing of images. NDVI data were in grid format with a resolution of 250 m. In this paper, the rainfall rate data in the China meteorological forcing dataset [35] was selected as the basic rainfall data of the experiment. The data were downloaded from the National Tibetan Plateau Data Center (https://data.tpdc.ac.cn/zh-hans/). After Python language processing, the daily rainfall data were obtained. The data were in grid form with a resolution of 1.0 × 10 4 m. The DEM data were downloaded from the geospatial data cloud platform (http://www.gscloud.cn/). DEM data were in grid format with a resolution of 30 m. The soil data were downloaded from the Cold and Arid Regions Science Data Center (http://data.casnw.net/portal/). The soil data were in grid format with a resolution of 1.0 × 10 3 m. The land-use data were downloaded from the resource and environmental data sharing platform of the Chinese Academy of Sciences (http://www.resdc.cn/). The land-use data were in grid format with a resolution of 1.0 × 10 3 m. The sediment data were the measured annual sediment transport data of Zhang-

Data Source and Processing
The research datasets include normalized difference vegetation index (NDVI) data, meteorological data, DEM data, soil data, land-use data, sediment data, and socioeconomic data. The data source and main processing processes are summarized as follows: NDVI data was downloaded from the National Aeronautics and Space Administration (NASA) website (https://www.nasa.gov/), using the MODIS Reprojection Tool (MRT) provided by NASA for batch processing of images. NDVI data were in grid format with a resolution of 250 m. In this paper, the rainfall rate data in the China meteorological forcing dataset [35] was selected as the basic rainfall data of the experiment. The data were downloaded from the National Tibetan Plateau Data Center (https://data.tpdc.ac.cn/zh-hans/). After Python language processing, the daily rainfall data were obtained. The data were in grid form with a resolution of 1.0 × 10 4 m. The DEM data were downloaded from the geospatial data cloud platform (http://www.gscloud.cn/). DEM data were in grid format with a resolution of 30 m. The soil data were downloaded from the Cold and Arid Regions Science Data Center (http://data.casnw.net/portal/). The soil data were in grid format with a resolution of 1.0 × 10 3 m. The land-use data were downloaded from the resource and environmental data sharing platform of the Chinese Academy of Sciences (http://www.resdc.cn/). The land-use data were in grid format with a resolution of 1.0 × 10 3 m. The sediment data were the measured annual sediment transport data of Zhangjiashan Hydrological Station in the Jinghe Basin of China, which were collected from the Ministry of Water Resources of the People's Republic of China (http://www.mwr.gov.cn/sj/tjgb/zghlnsgb/). The socioeconomic data were collected from the relevant records of the National Bureau of Statistics (http://www.stats.gov.cn/). caused by spatial data, this paper transformed all spatial data into a unified coordinate system and resampled them to 30 m based on the resampling tool of ArcGIS 10.2.

Soil Erosion Model
Based on the RUSLE model [14], the amount of potential soil erosion (A p ), the amount of actual soil erosion (A r ), and the amount of soil conservation (A c ) in the Jinghe Basin were calculated from 2005 to 2015. The experiment selected observation data from the Zhangjiashan Hydrological Station at Jinghe Basin in the River Sediment Bulletin of China of the Ministry of Water Resources of the People's Republic of China to verify the actual soil erosion simulated by the model. A p refers to soil erosion in the ecosystem without vegetation cover and water and soil conservation measures, that is, C = 1, P = 1. A r refers to soil erosion in the ecosystem with vegetation cover, and water and soil conservation measures. A c is the difference between A p and A r . They can be calculated as follows: where A p is the amount of potential soil erosion (kg/m 2 ), A r is the amount of actual soil erosion (kg/m 2 ), A c is the amount of soil conservation (kg/m 2 ), R is the rainfall erosivity factor (J*mm/m 2 /s), K is the soil erodibility factor (kg*m 2 *s/m 2 /J/mm), LS is the slope length and steepness factor (dimensionless), C is the surface vegetation cover factor (dimensionless), and P is the soil conservation measure factor (dimensionless). The rainfall erosivity factor R refers to the potential erosivity caused by rainfall. It is the most important factor leading to soil loss. This experiment adopted the empirical formula proposed by Wischmeier et al. [36], which is more suitable for the Loess Plateau region [37], using monthly rainfall. The calculation formula is as follows: where P i is the monthly rainfall (mm) and P is the average annual rainfall (mm). The soil erodibility factor K refers to the calculation method in the erosion-productivity impact calculator (EPIC) model by Williams et al. [38]. This method was applied in small basins by Cai et al. [39]. The results calculated by this formula are very close to the measured values, and the method is simple. The calculation formula is as follows: where SAN is the sand content (%), SIL is silt content (%), CLA is clay content (%), and C is the organic carbon content (%). The slope length and steepness factor were calculated by the method proposed by Wischmeier et al. [11]. Since the Jinghe Basin is located in the Loess Plateau region and the slope is steep, the slope factor calculation adopted the steep slope calculation method proposed by Mccool et al. [40] and Liu et al. [41]. The formulas are as follows: where L is the slope length factor, S is the slope steepness factor, λ is the length of the horizontal projection of the slope, α is the slope length exponent, and θ is the slope (%) extracted from DEM. The vegetation coverage can be estimated more accurately by establishing the relationship between the vegetation index and the vegetation coverage factor [42]. This experiment adopted the calculation formula of the vegetation coverage factor of Cai et al. [39], which can be written as follows: where f c is the vegetation coverage (%), C is the vegetation coverage and management factor, NDV I is the normalized vegetation index value, and NDV I max and NDV I min are the maximum and minimum values of NDV I in the study area, respectively. The p value reflects the soil and water conservation measures, indicating the ratio of soil lost after special measures to the amount lost when planting along the slope. The experiment adopted the method of calculating the P factor proposed by Lufafa et al. [43] in 2003, and the calculation formula is as follows: where S is the slope (%).

Hydrological Information Extraction
Based on DEM data and ArcGIS 10.2, the experiment extracted the river network of Jinghe Basin and divided the sub-basins of Jinghe Basin. The specific steps are as follows: 1 use the fill tool of ArcGIS to fill the DEM, 2 use the flow direction tool of ArcGIS to extract the water flow direction, 3 use the flow accumulation tool of ArcGIS to calculate the basin accumulation, 4 use the raster calculator tool of ArcGIS to obtain the river network of the basin, 5 use the stream link tool of ArcGIS to capture the river pour point, and 6 use the watershed tool of ArcGIS to divide the basin.
In order to obtain accurate basin river network information, the experiment set 15 different water catchment thresholds between 1 × 10 4 m 2 -2.25 × 10 6 m 2 in the raster calculator tool of ArcGIS and obtained the river network under the corresponding water catchment threshold. At the same time, the experiment calculated the density of the river network under different water catchment thresholds, which was shown in Figure 2. The fitting relationship between the water catchment threshold and the river network density is y = 0.022 × x (−0.264) (R 2 = 0.9954), as shown in Equation (14), where x is the water catchment threshold and y is the river network density. By calculating the second derivative of Equation (14), we found no point in the definition domain where the second derivative of Equation (14) was 0 and where the derivative did not exist. With reference to previous experience [44], the power formula and the linear formula can be used to determine the appropriate catchment threshold for the river network extraction together. Therefore, this experiment calculated the combined solution for Equations (14) and (15) to obtain the catchment threshold suitable for the river basin.
Sustainability 2021, 13, 1794 6 of 17 In the following research, this paper refers to Equations (14) and (15) as a formula set. By solving the formula set, we found that, when b = 0.177 and x = 0.14, the equation has a unique solution; when b < 0.177, the formula set had no answer; and when b > 0.177, the formula set had two solutions. Determining the catchment threshold in this way avoided the subjectivity of artificial determination and simplified the complexity of the trial. Compared with the Jinghe Network in the 1:1 million topographic maps of China [45], the river network extracted based on ArcGIS was consistent. Therefore, this paper takes a catchment area of 1.4 × 10 5 m 2 (i.e., x = 1.4 × 10 5 m 2 ) for follow-up research work. In this case, the Jinghe Basin was divided into 58 sub-basins.

Dinf Algorithm
The Dinf algorithm takes the cell to be calculated as the center and diagonally divides the surrounding 8 cells (the diagonal direction of the central cell) in every 45°. The maximum slope and aspect in 8 intervals were calculated [46,47]. Then, the gap with the largest slope was selected and its direction r was regarded as the water flow direction of the current grid cell. The water flow was broken into 2 outflow directions: horizontal and vertical, corresponding to the surrounding 8 cells. An example is presented in Figure 3. The weight of flow distribution was determined by the angle between the slope direction and horizontal or vertical direction and the ratio of angle between the slope direction and diagonal direction. The water flow distribution plan can be seen in Figure 4. In the following research, this paper refers to Equations (14) and (15) as a formula set. By solving the formula set, we found that, when b = 0.177 and x = 0.14, the equation has a unique solution; when b < 0.177, the formula set had no answer; and when b > 0.177, the formula set had two solutions. Determining the catchment threshold in this way avoided the subjectivity of artificial determination and simplified the complexity of the trial. Compared with the Jinghe Network in the 1:1 million topographic maps of China [45], the river network extracted based on ArcGIS was consistent. Therefore, this paper takes a catchment area of 1.4 × 10 5 m 2 (i.e., x = 1.4 × 10 5 m 2 ) for follow-up research work. In this case, the Jinghe Basin was divided into 58 sub-basins.

Dinf Algorithm
The Dinf algorithm takes the cell to be calculated as the center and diagonally divides the surrounding 8 cells (the diagonal direction of the central cell) in every 45 • . The maximum slope and aspect in 8 intervals were calculated [46,47]. Then, the gap with the largest slope was selected and its direction r was regarded as the water flow direction of the current grid cell. The water flow was broken into 2 outflow directions: horizontal and vertical, corresponding to the surrounding 8 cells. An example is presented in Figure 3. The weight of flow distribution was determined by the angle between the slope direction and horizontal or vertical direction and the ratio of angle between the slope direction and diagonal direction. The water flow distribution plan can be seen in Figure 4.    Assuming that the maximum slope of the current grid unit (i, j) is in △2, the two downstream units involved are (i + 1, j) (i + 1, j + 1). If the current grid unit (i, j) is A, the flow calculation formula for the two downstream grids can be calculated using Equation (16):

Determination of Supply and Demand
 Determine the amount of supply and demand.
where c A (kg/m 2 ) is the amount of supply from soil conservation services, p A (kg/m 2 ) is the amount of potential soil erosion, r A (kg/m 2 ) is the amount of actual soil erosion, and r D (kg/m 2 ) is the amount of demand for soil conservation services.
 Determine the supply and demand Assuming that the maximum slope of the current grid unit (i, j) is in ∆2, the two downstream units involved are (i + 1, j) (i + 1, j + 1). If the current grid unit (i, j) is A, the flow calculation formula for the two downstream grids can be calculated using Equation (16):

Determination of Supply and Demand
1 Determine the amount of supply and demand.
where A c (kg/m 2 ) is the amount of supply from soil conservation services, A p (kg/m 2 ) is the amount of potential soil erosion, A r (kg/m 2 ) is the amount of actual soil erosion, and D r (kg/m 2 ) is the amount of demand for soil conservation services. In addition, this experiment normalized the supply and demand values of soil and water conservation services and divided them into five levels, namely low levels (0-0.2), low level (0.2-0.4), intermediate level (0.4-0.6), and high level (0.6-0.8) and a higher level (0.8-1) [48]. 2 Determine the supply and demand where A w (kg/m 2 ) is the difference between the supply and demand of soil conservation services. If the difference is greater than 0, it is a supply area. If it is less than 0, it is a demand area.

Results and Discussion
The experimental results showed that the average annual sediment transport modules in the Jinghe Basin from 2005 to 2015 was 1.597 kg/m 2 . With reference to previous research results [49], the sediment transport ratio of the Loess Plateau is approximately 1. This means that the soil erosion modulus of the basin is roughly equal to the sediment transport modules. The RUSLE model calculated the average annual soil erosion modulus of the Jinghe Basin from 2005 to 2015 as 1.441 kg/m 2 with a ratio of about 1.1 and a correlation coefficient (R 2 ) of 0.7392. The model accuracy verification is presented in Figure 5. ules in the Jinghe Basin from 2005 to 2015 was 1.597 kg/m 2 . With reference to previous research results [49], the sediment transport ratio of the Loess Plateau is approximately 1. This means that the soil erosion modulus of the basin is roughly equal to the sediment transport modules. The RUSLE model calculated the average annual soil erosion modulus of the Jinghe Basin from 2005 to 2015 as 1.441 kg/m 2 with a ratio of about 1.1 and a correlation coefficient (R 2 ) of 0.7392. The model accuracy verification is presented in Figure 5.

Temporal and Spatial Distribution Characteristics of Soil Erosion
The

Temporal and Spatial Distribution Characteristics of Soil Conservation Service
The spatial distribution of soil conservation in the Jinghe Basin in 2005, 2010, and 2015 is presented in Figure 7. The spatial distribution pattern of the amount of soil conservation service was relatively consistent with the spatial distribution pattern of soil erosion. R 2 between the basin soil conservation rate and NDVI from 2005 to 2015 was 0.60, and that between the soil conservation rate and rainfall for the same time was 0.87. In 2005, the amount of soil conservation in the southwest and southeast of the basin was low, while at other parts, it was relatively high. In 2010, the north-central and southeastern parts of the basin exhibited a high amount of soil conservation. In 2015, these areas with high amounts of soil conservation were located in the western and central-eastern regions of the basin. In 2005, 2010, and 2015, the amounts of soil conservation in the Jinghe Basin were 3.09 × 10 11 kg, 4.58 × 10 11 kg, and 2.26 × 10 11 kg, accounting for 78.86%, 79.54%, and 79.30% of the potential erosion in the corresponding years. During the study period, although the amount of soil conservation fluctuated under the influence of rainfall, the ratio of soil conservation to potential soil erosion increased year by year.

Temporal and Spatial Distribution Characteristics of Soil Conservation Service
The spatial distribution of soil conservation in the Jinghe Basin in 2005, 2010, and 2015 is presented in Figure 7. The spatial distribution pattern of the amount of soil conservation service was relatively consistent with the spatial distribution pattern of soil erosion. R 2 between the basin soil conservation rate and NDVI from 2005 to 2015 was 0.60, and that between the soil conservation rate and rainfall for the same time was 0.87. In 2005, the amount of soil conservation in the southwest and southeast of the basin was low, while at other parts, it was relatively high. In 2010, the north-central and southeastern parts of the basin exhibited a high amount of soil conservation. In 2015, these areas with high amounts of soil conservation were located in the western and central-eastern regions of the basin. In 2005, 2010, and 2015, the amounts of soil conservation in the Jinghe Basin were 3.09 × 10 11 kg, 4.58 × 10 11 kg, and 2.26 × 10 11 kg, accounting for 78.86%, 79.54%, and 79.30% of the potential erosion in the corresponding years. During the study period, although the amount of soil conservation fluctuated under the influence of rainfall, the ratio of soil conservation to potential soil erosion increased year by year.
In 2005, 2010, and 2015, the soil conservation amount for different land-use types was in the order of grassland > cultivated land > woodland > construction land > water area > unused land. During the study period, the proportion of soil conservation of grassland to the total soil conservation amount of the basin decreased from 47. Compared with the ratio of land-use types, it can be found that, during the study period, the area under grassland and cultivated lands decreased by 0.07% and 0.44% while the area of forest and construction land increased by 0.14% and 0.33%, respectively. However, in general, the basin soil conservation measures achieved results during the study period, and the economy of the area was improved.

Space Flow Path of Soil Conservation Service
This paper defines soil conservation service flow as reducing the amount of sediment carried by surface runoff due to restoration of the basin ecosystem, with improvements in ecological benefits. Soil conservation service flow is a directional ecological service flow [9,51]. The carrier of soil conservation service is runoff or the material or energy carried with runoff as a carrier. For example, the No.1 sub-basin was divided into 2181 hydrological response units (HRUs) based on DEM, land use, and soil. Figure 8 shows the spatial flow directions of soil conservation service at the HRU scale in the No.1 sub-basin of the Jinghe Basin and the corresponding amount of soil conservation service in 2015. Figure 9 shows the spatial flow directions of soil conservation service at the channel scale in the No.1 sub-basin of the Jinghe Basin and the corresponding amount of soil conservation service in 2015.

Space Flow Path of Soil Conservation Service
This paper defines soil conservation service flow as reducing the amount of sediment carried by surface runoff due to restoration of the basin ecosystem, with improvements in ecological benefits. Soil conservation service flow is a directional ecological service flow [9,51]. The carrier of soil conservation service is runoff or the material or energy carried with runoff as a carrier. For example, the No.1 sub-basin was divided into 2181 hydrological response units (HRUs) based on DEM, land use, and soil. Figure 8 shows the spatial flow directions of soil conservation service at the HRU scale in the No.1 sub-basin of the Jinghe Basin and the corresponding amount of soil conservation service in 2015. Figure 9 shows the spatial flow directions of soil conservation service at the channel scale in the No.1 sub-basin of the Jinghe Basin and the corresponding amount of soil conservation service in 2015.    This section takes the amount of soil conservation in 2015 as an example. The distribution of the supply levels and demand levels of soil conservation service in each subbasin of the Jinghe Basin in 2015 is presented in Figure 10. Sub-basin No.1 to the north of the basin; No.16 in the middle of the basin; and No. 26,No.38,No.45,and No.47 to the southwest of the basin had more quantity supplied than quantity demanded. In this subbasin, the slope was relatively gentle, the land-use was mainly low-coverage grassland  26,No.38,No.45,and No.47 to the southwest of the basin had more quantity supplied than quantity demanded. In this sub-basin, the slope was relatively gentle, the land-use was mainly low-coverage grassland and cultivated land, and the soil type was highly active leached soil. After applying fertilizer, the soil fertility was high and the soil erosion rate was low. The slope in the No. 16  supplied. The sub-basins No.3-8 and No.10-15 had steep slopes. The land-use types were mainly grassland and cultivated land with low vegetation coverage. The rainfall was high, and the erosion of soil particles by raindrops and runoff was strong, leading to severe soil erosion and nutrient loss. The sub-basins No. 20,No.25,No.30,and No.54 received heavy rainfall and had medium vegetation coverage. The main soil types were black with thin layers, which were loose and easy to lose. The sub-basins No.17-18 belonged to the Loess Plateau area, with large undulations and rugged slopes. According to Equation (19), the distribution of the supply area and demand area of the basin are presented in Figure 11. In 2015, there were five main supply areas for soil conservation service in the basin. The supply amount from these areas was greater than the demand. The remaining regions belonged to the demand area, i.e., the supply amount was less than the demand. There was a surplus of soil conservation service in these supply According to Equation (19), the distribution of the supply area and demand area of the basin are presented in Figure 11. In 2015, there were five main supply areas for soil conservation service in the basin. The supply amount from these areas was greater than the demand. The remaining regions belonged to the demand area, i.e., the supply amount was less than the demand. There was a surplus of soil conservation service in these supply areas, leading to a flow phenomenon. Following Wang et al. [52], we assumed that the soil conservation service from the supply areas would first flow to the most adjacent area (called connected flow) (represented by the solid arrow in Figure 11). In the demand area, it was a deficit status without the flow phenomenon (called terminating flow), represented by the dotted arrow in Figure 11. In addition, this paper assumed that a sub-basin could receive soil conservation service from multiple sub-basins through the soil conservation service flow. Based on the above assumptions and the consideration of the elevation of each sub-basin and according to the flow law of matter from high to low, we obtained the spatial flow path of the soil conservation service in the Jinghe Basin in 2015 at the sub-basin scale ( Figure 11) and the direction and quantity ( Table 2).
We finally mapped the spatial flow paths of the soil conservation service on different scales in the Jinghe Basin in 2015 ( Figure 12). Figure 11 shows the direction of the soil conservation service flow on the HRU-channel scale, and the direction and the amount of soil conservation service trend flow on the scale of the supply area to the demand area. These will provide support to improve the credibility of the soil conservation policy in the basin.
Combined with the socioeconomic data in 2015, we also calculated the value of ecological compensation in the Jinghe Basin. In 2015, the price of N fertilizer was 0.284 USD/kg, of P fertilizer was 0.08 USD/kg, and of K fertilizer was 0.67 USD/kg and the construction cost of the reservoir project was 0.87 USD/m 3 . The details of the calculation can be found in Reference [53]. According to the 2020 inflation rate (calculated according to the Chinese Consumer Price Index (CPI)), the total ecological compensation from the supply areas for soil conservation services in the Jinghe Basin (a total of 18 sub-basins) was 7.3 × 10 5 USD. The total ecological compensation from the demand areas (40 sub-basins in total) was 8.1 × 10 7 USD. The whole ecological compensation from the demand areas paid to the supply areas was 1.67 × 10 5 USD. This is expected to provide a reference for the formulation of ecological compensation policies in the basin.
conservation service from the supply areas would first flow to the most adjacent area (called connected flow) (represented by the solid arrow in Figure 11). In the demand area it was a deficit status without the flow phenomenon (called terminating flow), represented by the dotted arrow in Figure 11. In addition, this paper assumed that a sub-basin could receive soil conservation service from multiple sub-basins through the soil conservation service flow. Based on the above assumptions and the consideration of the elevation o each sub-basin and according to the flow law of matter from high to low, we obtained the spatial flow path of the soil conservation service in the Jinghe Basin in 2015 at the sub basin scale (Figure 11) and the direction and quantity (Table 2).    We finally mapped the spatial flow paths of the soil conservation service on different scales in the Jinghe Basin in 2015 ( Figure 12). Figure 11 shows the direction of the soil conservation service flow on the HRU-channel scale, and the direction and the amount of soil conservation service trend flow on the scale of the supply area to the demand area. These will provide support to improve the credibility of the soil conservation policy in the basin. Combined with the socioeconomic data in 2015, we also calculated the value of ecological compensation in the Jinghe Basin. In 2015, the price of N fertilizer was 0.284 USD/kg, of P fertilizer was 0.08 USD/kg, and of K fertilizer was 0.67 USD/kg and the construction cost of the reservoir project was 0.87 USD/m 3 . The details of the calculation can be found in Reference [53]. According to the 2020 inflation rate (calculated according to the Chinese Consumer Price Index (CPI)), the total ecological compensation from the supply areas for soil conservation services in the Jinghe Basin (a total of 18 sub-basins) was 7.3 × 10 5 USD. The total ecological compensation from the demand areas (40 sub-basins in total) was 8.1 × 10 7 USD. The whole ecological compensation from the demand areas paid to the supply areas was 1.67 × 10 5 USD. This is expected to provide a reference for the formulation of ecological compensation policies in the basin.

Discussion
Presently, different experts and scholars have different views on the flow of ecosystem services. Among them, there are two main understandings: (1) with emphasis on the process, the ecosystem service flow is a link between the ecosystem from the supply area to the demand area and the ecosystem service provided by the supply area is transferred to the beneficial area along a direction and path; (2) with emphasis on ultimate utility, the

Discussion
Presently, different experts and scholars have different views on the flow of ecosystem services. Among them, there are two main understandings: (1) with emphasis on the process, the ecosystem service flow is a link between the ecosystem from the supply area to the demand area and the ecosystem service provided by the supply area is transferred to the beneficial area along a direction and path; (2) with emphasis on ultimate utility, the ecosystem service flow is the ecosystem service finally obtained by a human being and the ultimate realization of ecosystem services. Based on the above opinions, this paper conducted a quantitative assessment of soil conservation services in the Jinghe Basin and separated the path and flow of ecosystem service flows to describe the path of ecosystem service flow and to calculate the flow amount of ecosystem service flow. This paper realized the multi-scale analysis of soil conservation service in "HRU-channel-sub-basin" and studied the specific paths and flow amount of soil conservation service flows on different scales. Soil erosion has always been one of the hot topics studied by scholars from various countries. China's Loess Plateau is a typical area for soil erosion research. This paper took soil conservation service as the research object and took the Jinghe Basin in the Loess Plateau region of China as the case area to evaluate the soil conservation service of the Jinghe Basin from 2005 to 2015 based on the RUSLE model. This paper took 2005, 2010, and 2015 as representatives to describe the temporal and spatial characteristics of soil erosion and soil conservation services of Jinghe Basin. On this basis, this paper took 2015 as representative to simulate the soil conservation service flow on three scales including "HRU-river-sub-basin", which is one of the innovations of this paper. Different from previous research, the division of HRUs, river channels, and sub-basins in this research was based on ArcGIS software, which is also one of the highlights of this article. Specifically, this paper used the azimuth to express the flow direction of the soil conservation service flow and gave specific values. This paper carried out a multi-scale expression of "HRUchannels-sub-basin" on the soil conservation services flow in 2015, which showed in more detail the spatial flow process of soil conservation services on different scales. This paper finally gave the general flow trend of soil conservation service throughout the basin, which can provide a scientific basis for formulating local comprehensive management policies and ecological safety policies in the future. Moreover, the rainfall erosivity factor (R) in the experiment was calculated based on Yangkun data. Compared with previous experiments, it avoided the influence of weather station data interpolation on the experiment. The soil conservation measure factor (P) in the experiment was based on the definite calculation formula proposed by LuFaFa [43] et al. Calculating the P factor through a certain formula effectively avoids the subjectivity of artificial assignment. However, the RUSLE model is an empirical model, which cannot fully demonstrate the process of soil erosion. In the future, we will try to simulate the service flow of soil conservation based on a mechanism model.

Conclusions
This paper proposed a new approach to the study of ecosystem service flow and realized multi-scale spatial flow simulation of soil conservation services in 2015. From 2005 to 2015, the correlation between the amount of soil erosion simulated by the RUSLE model and the measured value reached 0.74. During the study period, the amount of soil erosion in the Jinghe Basin showed a trend of first deepening and then reducing. Affected by rainfall, the amount of soil conservation showed a trend of first increasing and then decreasing, but the proportion of the amount of soil conservation to the amount of potential soil erosion during the study period showed an increasing trend. This paper took the soil conservation service in the Jinghe Basin in 2015 as an example and separately described the spatial flow path and flow volume of the soil conservation service at the scale of "HRU-channel-sub-basin". Among them, the division of HRUs, channels, and subbasins in the Jinghe Basin were all realized through the ArcGIS software. After comparing the supply and demand of soil conservation services, this paper divided the supply area and beneficiary area of soil conservation services in the basin in 2015 and calculated the ecological cost that the soil conservation service benefit area needs to provide to the supply area as 1.67 × 10 5 USD. The experimental results can provide a certain scientific reference for the formulation of comprehensive management policies for the river basin.