Water-Saving Soil Conservation Measures Should Be Used in Northern China: Evidence from Runoff Plot Data

Most of the current studies on soil conservation measures mainly focus on their soil control effect, neglecting their impact on water quantity. In the present study, the latest seven years (2014– 2020) of monitored data from 22 runoff plots in the upstream catchment of the Miyun Reservoir, Beijing were used to evaluate the effects of slope, rainfall, and soil conservation measures on soil and water loss, and some implications were given in this water-scarce region. Excluding the impact of soil conservation measures, soil loss increased with the slope gradient and slope length. Runoff and soil loss were greatly affected by the rainfall amount and maximum 30-min rainfall intensity on the bare and cultivated slopes, or by rainfall amount and rainfall duration on almost all of the plots with soil conservation measures. The results indicated that the bare soil suffered the most severe soil loss, with a mean annual soil loss rate (SLR) of 4325 t km−2 year−1, followed by the cultivated lands without any measure, with an annual SLR of above 3205 t km−2 year−1. Contour tillage cannot effectively control soil loss on steep slopes. The vegetation measures and terrace, level bench, and fish scale pits, as well as their combinations, can decrease runoff by above 86% and decrease soil loss by 95%, respectively. Water-saving measures should be implemented in the study region. The measures, such as vegetation coverage, terracing, contour tillage, etc., should be carefully implemented on slopes. Bare and cultivated lands should further be implemented with soil conservation measures in this and similar regions in the world.


Introduction
Water erosion is one of the biggest ecological and environmental problems worldwide [1]. It can induce both on-site fertile topsoil and nutrient losses and off-site sediment siltation in river systems, and cause water resource use problems [2][3][4][5]. Severe hazards induced by water erosion have been noticed a long time ago, both in developing and developed countries [6,7].
Soil conservation measures, such as contour tillage, terrace, and vegetation plantation, were widely used in the world to combat water erosion, and many studies have been published until very recently. For example, Zhao et al. [8] summarized the runoff plot data on the Chinese Loess Plateau, and pointed out that soil conservation measures had greatly reduced soil and water losses, but not down to background levels (i.e., 500 t km −2 year −1 ). Wolka et al. [9] reviewed the effects of soil conservation measures on runoff, soil loss, and crop yields in Sub-Saharan Africa that showed that the effect of crop slope barrier soil conservation techniques on crop yields varied with the rainfall slope. Maetens et al. [10] evaluated the effects of soil conservation measures in reducing plot runoff and soil loss in Europe and the Mediterranean. Xiong et al. [1] also gave a global review of the effects of soil conservation measures on soil loss control. These studies greatly improve our understanding of the effects of soil conservation measures on soil and water losses. However, most of the studies focus on the reduction effects of soil conservation measures, such as In the study region, the main vegetation is Robinia pseudoacacia, Platycladus orientalis, and economic forest, and the main crops are corn and wheat. Around 60.9% of the sandy loam soil contains particles larger than 0.05 mm in diameter, 18.22% contains particles 0.05-0.005 mm in diameter, 9.22% contains particles 0.005-0.001 mm in diameter, and 14.88% contains particles less than 0.001 mm in diameter [25].

Runoff Plot and Data Collection
In the Shixia catchment, there were 22 runoff plots in total ( Figure 1). The boundaries of each plot were made of bricks and cement. The height of the plot boundary is around 30 cm above ground, with a 40-cm deep insertion into the soils to prevent runoff from leaving or entering the plot. Among the 22 plots, one plot is 5 m long, two plots are 20 m long, and the other plots are 10 m long. The slope degrees of these plots vary from 3.5° to 27°. These plots include five cultivated plots, four bare plots, and thirteen plots with different soil conservation measures. The soil conservation measures implemented on these plots include contour tillage, fish scale pit, level bench, terrace, forest, grass, shrub, and their combinations. Corn was planted on the cultivated plots. Detailed information is provided in Table 1. In the study region, the main vegetation is Robinia pseudoacacia, Platycladus orientalis, and economic forest, and the main crops are corn and wheat. Around 60.9% of the sandy loam soil contains particles larger than 0.05 mm in diameter, 18.22% contains particles 0.05-0.005 mm in diameter, 9.22% contains particles 0.005-0.001 mm in diameter, and 14.88% contains particles less than 0.001 mm in diameter [25].

Runoff Plot and Data Collection
In the Shixia catchment, there were 22 runoff plots in total ( Figure 1). The boundaries of each plot were made of bricks and cement. The height of the plot boundary is around 30 cm above ground, with a 40-cm deep insertion into the soils to prevent runoff from leaving or entering the plot. Among the 22 plots, one plot is 5 m long, two plots are 20 m long, and the other plots are 10 m long. The slope degrees of these plots vary from 3.5 • to 27 • . These plots include five cultivated plots, four bare plots, and thirteen plots with different soil conservation measures. The soil conservation measures implemented on these plots include contour tillage, fish scale pit, level bench, terrace, forest, grass, shrub, and their combinations. Corn was planted on the cultivated plots. Detailed information is provided in Table 1.
In 2014-2020, runoff and sediment discharges were collected with a nine-hole diversion bucket and a tank at the end of each plot. After each rainfall event, the runoff amount was measured. Evenly mixed water and sediment samples were also collected with 1000 mL flasks and transported to a laboratory where sediment concentrations were determined with a dry method [15].
Surface runoff depth (H; mm) for each rainfall event was then calculated with the total rainfall amount and the area of the runoff plot. The event soil loss rate (SLR) was also calculated using the total runoff amount, sediment concentration, and runoff plot area. Annual H and annual SLR were then obtained through summing the events for each plot.
A rain gauge and rain barrel near the plots were used to record rainfall information. In the present study, an erosive rainfall was defined as the one that induced erosion on any of the plots, as proposed by Zhu and Zhu (2014). For each erosive rainfall event, rainfall duration (RD), rainfall amount (P), mean rainfall intensity (I m ), and maximum rainfall intensities at 30 min (I 30 ) and at 60 min (I 60 ) were obtained from rainfall process data.

Data Treatment and Statistical Analysis
In the present study, Pearson correlation analysis and regression analysis were used to determine the relationships between H, SLR, and the five rainfall eigenvalues. Fisher's protected least significant difference test was also used to compare the means of Hs and SLRs from the plots. Treatments were considered significant if p < 0.05. All of the statistics were conducted using SPSS version 14.0 for Windows.

Rainfall Characteristics
The mean annual Ps in 2014-2020 ranged from 385 mm in 2020 to 532.2 mm in 2018, with an average of 475.1 mm, a coefficient of variation of 0.11, and a skewness of −0.86. The annual erosive Ps varied from 239.2 mm in 2020 to 434.7 mm in 2018, with an average of 297.1 mm, or 62.3% of the annual mean P ( Figure 2 and Table 2). In 2014-2020, over 50% of the annual Ps produced runoff on the plots. These percentages were higher than those on the Chinese Loess Plateau reported by Liang et al. (2020). The mean annual Ps in 2014-2020 ranged from 385 mm in 2020 to 532.2 mm in 2018, with an average of 475.1 mm, a coefficient of variation of 0.11, and a skewness of −0.86. The annual erosive Ps varied from 239.2 mm in 2020 to 434.7 mm in 2018, with an average of 297.1 mm, or 62.3% of the annual mean P ( Figure 2 and Table 2). In 2014-2020, over 50% of the annual Ps produced runoff on the plots. These percentages were higher than those on the Chinese Loess Plateau reported by Liang et al. (2020).  In 2014-2020, there were 66 erosive rainfall events. The event Ps ranged from 4.8 to 108.1mm, with an average of 31.5 mm. The Ps in the range of 10-20 mm occupied over one third of the total number of rainfall events. All of these erosive rainfall events ranged from 0.3 to 32.3 h in RD, with an average of 6.9 h, from 1.0 to 54.0 mm h −1 in Im with an average of 9.5 mm h −1 , and from 3.6 to 61.7 mm h −1 and 6.0 to 64.2 mm h −1 in I60 and I30, respectively ( Figure 3). Under future climate change, more extreme rainfall events could occur [13], inducing more severe soil erosion in the study region.  In 2014-2020, there were 66 erosive rainfall events. The event Ps ranged from 4.8 to 108.1mm, with an average of 31.5 mm. The Ps in the range of 10-20 mm occupied over one third of the total number of rainfall events. All of these erosive rainfall events ranged from 0.3 to 32.3 h in RD, with an average of 6.9 h, from 1.0 to 54.0 mm h −1 in I m with an average of 9.5 mm h −1 , and from 3.6 to 61.7 mm h −1 and 6.0 to 64.2 mm h −1 in I 60 and I 30 , respectively ( Figure 3). Under future climate change, more extreme rainfall events could occur [13], inducing more severe soil erosion in the study region.

Runoff and Soil Loss
Annual and event Hs varied widely among the 22 runoff plots ( Figure 4a and Table 3). The mean annual Hs in 2014-2020 ranged from 0 on plots 8 and 10 to 100.61 mm on plot 22. The mean annual Hs were less than 18.03 mm on most of the plots, which were significantly less than those on the bare and cultivated plots 1, 3, 4, 6, 21, and 22. Event Hs also differed greatly. The maximum event H was 10.67 mm on plot 22, and the minimum H was zero on plots 8 and 10 ( Table 3). Most of the plots had a mean event H less than 2.00 mm. The event Hs on plots 1, 3, 4, 6, 21, and 22 were significantly higher than those on other plots.

Runoff and Soil Loss
Annual and event Hs varied widely among the 22 runoff plots ( Figure 4a and Table  3). The mean annual Hs in 2014-2020 ranged from 0 on plots 8 and 10 to 100.61 mm on plot 22. The mean annual Hs were less than 18.03 mm on most of the plots, which were significantly less than those on the bare and cultivated plots 1, 3, 4, 6, 21, and 22. Event Hs also differed greatly. The maximum event H was 10.67 mm on plot 22, and the minimum H was zero on plots 8 and 10 ( Table 3). Most of the plots had a mean event H less than 2.00 mm. The event Hs on plots 1, 3, 4, 6, 21, and 22 were significantly higher than those on other plots.      Notes: * indicates the recorded minimum value to generate runoff in the 66 rainfall events. TRA, MAP, Accu. P, and RGT represent the threshold of the runoff amount, mean annual rainfall amount, accumulated rainfall amount, and runoff generation times, respectively. The mean values of runoff depth and soil loss rate on the columns with the same letter were not significantly at the 0.05 level.
Annual SLR also varied from 0 on plots 8 and 10 to 6006.84 t km −2 year −1 on plot 21 (Figure 4b), and annual SLRs on the plots 1, 3, 4, 6, 21, and 22 were higher than 2500 t km −2 year −1 , which is above twelve times the tolerate value (i.e., 200 t km −2 year −1 ) in the region [15]. In the Miyun Reservoir catchment, these lands are around 600 km 2 , occupying around 5% of the catchment area. Although this percentage is smaller, it should also be paid attention to, because it is still an important pollutant source [13,14], implying that these kinds of lands should have soil conservation measures implemented. In contrast, the annual SLRs from plots 2, 5, and 7-19 were much lower. In terms of event SLRs, the maximum SLR was up to 637 t km −2 event −1 from plot 21, and plots 1, 3, 4, 6, and 22 also had high event SLRs. Conversely, other plots had a much lower event SLR than 25 t km −2 event −1 , and plots 8 and 10 did not yield sediment (Table 3).

Impact of the Slope Gradient and Slope Length
The effect of the slope gradient on runoff and SLR is a controversial issue. As the slope gradients increase, surface runoff and soil loss can increase, decrease, and/or change little, resulting from the influence of complex factors, such as a forming or disruption of soil crust [27,28], rill and/or ephemeral gully development [29,30], differential soil cracking [31], and/or ponding depth [32,33], etc. Sometimes, a threshold maximum value of H or SLR exists with increasing slope gradients, resulting from the interaction of runoff generation and its sediment carrying capacity [28,34]. In the present study, both annual H and annual SLR increased first and then decreased, with a threshold annual H of 100.1 mm and a threshold annual SLR of 4748.29 t km −2 year −1 occurring at the 14.4 • slope ( Figure 5). A number of studies [35][36][37][38] found that the threshold slope gradients of H and SLR varied from 25 • to 50 • , which are much steeper than the one used in this study. The threshold slope gradient of 14.4 • of annual H and annual SLR can be explained by the impact of soil conservation measures. In the present study, the bare plots with a slope degree of 14.4 • had the highest annual H and annual SLR, whereas other plots with slope degrees lower and/or higher than 14.4 • were implemented with soil conservation measures, and yielded lower annual Hs and annual SLRs. For the plots with the same treatment, the changes of Hs and SLRs with the slope gradient varied depending on plot treatments. For example, higher annual Hs and SLRs occurred on contour tillage plot 6 than those of contour tillage plots 18 and 20. Similar phenomenon also occurred for the terraced orchard plots 2 and 7 ( Figure 4). This phenomenon could be attributed to the decrease in depressional storage and ponding depth on the plots with higher slope degrees [33,39]. However, for the shrub plots (i.e., plots 5, 9, 11, 13, 14, and 19), a threshold slope seemed to exist, whereas Hs and SLRs on plots 21 and 22 with a lower slope gradient were much higher than that of plot 4, resulting from the effect of the slope length.
Water 2021, 13, 853 9 of 17 plained by the impact of soil conservation measures. In the present study, the bare plots with a slope degree of 14.4° had the highest annual H and annual SLR, whereas other plots with slope degrees lower and/or higher than 14.4° were implemented with soil conservation measures, and yielded lower annual Hs and annual SLRs. For the plots with the same treatment, the changes of Hs and SLRs with the slope gradient varied depending on plot treatments. For example, higher annual Hs and SLRs occurred on contour tillage plot 6 than those of contour tillage plots 18 and 20. Similar phenomenon also occurred for the terraced orchard plots 2 and 7 ( Figure 4). This phenomenon could be attributed to the decrease in depressional storage and ponding depth on the plots with higher slope degrees [33,39]. However, for the shrub plots (i.e., plots 5, 9, 11, 13, 14, and 19), a threshold slope seemed to exist, whereas Hs and SLRs on plots 21 and 22 with a lower slope gradient were much higher than that of plot 4, resulting from the effect of the slope length.  Table 1.
Runoff and soil loss also present different changing patterns with increasing slope length [40][41][42]. In the present study, only three slope lengths (i.e., 5 m, 10 m, and 20 m) were included ( Table 1). The scattered dots in Figure 6 illustrate that annual Hs and SLRs increased or decreased when the slope lengths increased from 5 m to 20 m, resulting from the impact of soil conservation measures. Further analysis was done when their impacts were excluded. The annual H decreased when the slope length increased from 5 m to 10 m for plots 11 and 14, which had similar coverage rates and slope degrees (Table 1). This is because the time and volume of rainfall per unit area required before the runoff starts are higher in the longer plot than the smaller one [43]. However, for the bare plots 3, 4, 21, and 22, the annual H increased with increasing slope length. The difference could result  Table 1.
Runoff and soil loss also present different changing patterns with increasing slope length [40][41][42]. In the present study, only three slope lengths (i.e., 5 m, 10 m, and 20 m) were included ( Table 1). The scattered dots in Figure 6 illustrate that annual Hs and SLRs increased or decreased when the slope lengths increased from 5 m to 20 m, resulting from the impact of soil conservation measures. Further analysis was done when their impacts were excluded. The annual H decreased when the slope length increased from 5 m to 10 m for plots 11 and 14, which had similar coverage rates and slope degrees ( Table 1). This is because the time and volume of rainfall per unit area required before the runoff starts are higher in the longer plot than the smaller one [43]. However, for the bare plots 3, 4, 21, and 22, the annual H increased with increasing slope length. The difference could result from different earth surface conditions. According to study results by Mermut et al. [44] and Fang et al. [28], the soil texture in the study region made the soil crust easily form on the bare plot, which reduced the infiltration rate and increased the runoff generation capacity [33]. Although soil crust was not detected in the present study, it has been verified by previous studies through the use of natural rainfall [33] and rainfall simulation experiments [28]. Given the same land surface, annual SLRs increased with increasing slope length. For example, the mean annual SLRs of bare plots 21 and 22 were higher than those of bare plots 3 and 4, resulting from an increased surface runoff amount with increasing slope length. In contrast, a higher SLR on brush plot 14 than that on brush plot 11 could be explained by increased sediment concentration in runoff due to the increased concentrated flow on a longer plot [28]. from different earth surface conditions. According to study results by Mermut et al. [44] and Fang et al. [28], the soil texture in the study region made the soil crust easily form on the bare plot, which reduced the infiltration rate and increased the runoff generation capacity [33]. Although soil crust was not detected in the present study, it has been verified by previous studies through the use of natural rainfall [33] and rainfall simulation experiments [28]. Given the same land surface, annual SLRs increased with increasing slope length. For example, the mean annual SLRs of bare plots 21 and 22 were higher than those of bare plots 3 and 4, resulting from an increased surface runoff amount with increasing slope length. In contrast, a higher SLR on brush plot 14 than that on brush plot 11 could be explained by increased sediment concentration in runoff due to the increased concentrated flow on a longer plot [28].

Impact of Rainfall Characteristics
Rainfall is the external force inducing surface runoff and soil loss. In the present study, five rainfall eigenvalues were used to explore the impact of rainfall characteristics on event H and SLR. Pearson correlation coefficients indicated that event H was significantly correlated with event P and rainfall intensity (I30 and/or I60) at the 0.01 level, excluding the terraced plots 2 and 7, and the densely vegetated plot 12 (Table 4), the event H and SLR of which, however, were significantly correlated with P and RD. In respect to event SLR, event SLR was significantly correlated with P, I30, and/or both of them on most of the plots. Different responses of H and SLR from the plots to different rainfall eigen-

Impact of Rainfall Characteristics
Rainfall is the external force inducing surface runoff and soil loss. In the present study, five rainfall eigenvalues were used to explore the impact of rainfall characteristics on event H and SLR. Pearson correlation coefficients indicated that event H was significantly correlated with event P and rainfall intensity (I 30 and/or I 60 ) at the 0.01 level, excluding the terraced plots 2 and 7, and the densely vegetated plot 12 (Table 4), the event H and SLR of which, however, were significantly correlated with P and RD. In respect to event SLR, event SLR was significantly correlated with P, I 30 , and/or both of them on most of the plots. Different responses of H and SLR from the plots to different rainfall eigenvalues can result from the protection capacities of the measures [45][46][47]. For instance, the vegetated measures can effectively protect the splash impact of rainfall intensity, resulting in insignificant correlations of H and rainfall intensity for the vegetated plots. Note: * represents significance at the 0.05 level, and ** represents significance at the 0.01 level. P, RD, I m , I 30 , and I 60 represent rainfall amount, rainfall duration, mean rainfall intensity, and maximum 30-and 60-min rainfall intensity for each erosive rainfall event.
Similar to the studies by Fang et al. [33] and Zhu and Zhu [7], a large percent of annual runoff and soil loss occurred during several rainfall events. In the present study, we compared soil loss by the maximum annual erosive event to annual soil loss through four exampled plots, which represented different soil conditions (i.e., plots 1-3 and 14) in Figure 7. Soil loss from cultivated plot 1 by the maximum annual soil erosive event in 2014-2020 ranged from 1658.67 to 5104.98 t km -2 , which contributed to a mean value of 40.50% of the annual total soil loss, ranging from 30.72% in 2019 to 61.13% in 2018. The magnitude and contributions of the maximum annual erosive event on bare plot 3 were similar to those of plot 1. However, much higher contributions of the maximum annual erosive event occurred on plots 2 and 14, with terrace and shrub measures, respectively. The maximum contributions of the maximum annual erosive events were up to 85% on plot 2 and 89% on plot 14, although their maximum soil loss rates were much less than those on the bare and cultivated plots.
four exampled plots, which represented different soil conditions (i.e., plots 1-3 and 14) in Figure 7. Soil loss from cultivated plot 1 by the maximum annual soil erosive event in 2014-2020 ranged from 1658.67 to 5104.98 t km -2 , which contributed to a mean value of 40.50% of the annual total soil loss, ranging from 30.72% in 2019 to 61.13% in 2018. The magnitude and contributions of the maximum annual erosive event on bare plot 3 were similar to those of plot 1. However, much higher contributions of the maximum annual erosive event occurred on plots 2 and 14, with terrace and shrub measures, respectively. The maximum contributions of the maximum annual erosive events were up to 85% on plot 2 and 89% on plot 14, although their maximum soil loss rates were much less than those on the bare and cultivated plots.  Table 1.

Impact of Soil Conservation Measures
In order to study the impact of soil conservation measures, ten kinds of land use and soil conservation measures were obtained by using the 22 plots in Table 1 (Figure 8). Among the ten treatments, mean annual Hs differed greatly, and changed from 85.80 mm on bare plots, 72.83 mm on the cultivated plot without any measure, 67.74 mm on the cultivated plots with contour tillage, 11.49 mm on the grass plots, followed by the shrub plots, orchard plots with terrace, and other plots. Noticeably, no runoff occurred on plot 8 with forest plus grass measures and plot 10 with forest and fish scale pit measures. In comparison to the bare plots, the mean runoff reduction efficiencies were above 86% for the plots with soil conservation measures, excluding plots 1 and 6. Based on this result, the large-scale implementation of soil conservation measures since 2000 in the study region could, to some extent, explain the decreased discharge into the Miyun Reservoir,  Table 1.

Impact of Soil Conservation Measures
In order to study the impact of soil conservation measures, ten kinds of land use and soil conservation measures were obtained by using the 22 plots in Table 1 (Figure 8). Among the ten treatments, mean annual Hs differed greatly, and changed from 85.80 mm on bare plots, 72.83 mm on the cultivated plot without any measure, 67.74 mm on the cultivated plots with contour tillage, 11.49 mm on the grass plots, followed by the shrub plots, orchard plots with terrace, and other plots. Noticeably, no runoff occurred on plot 8 with forest plus grass measures and plot 10 with forest and fish scale pit measures. In comparison to the bare plots, the mean runoff reduction efficiencies were above 86% for the plots with soil conservation measures, excluding plots 1 and 6. Based on this result, the large-scale implementation of soil conservation measures since 2000 in the study region could, to some extent, explain the decreased discharge into the Miyun Reservoir, which agrees with the former studies at a catchment scale [13][14][15]. In order to ensure water quantity of the Miyun Reservoir, soil conservation measures should be water-saving, which allows more runoff downstream to the reservoir.
Annual SLRs presented the same changing patterns. The bare plots had the largest SLR, with an average of 4325.79 t km −2 year −1 , followed by the cultivated plot with no soil conservation measures, with an average of 3205.23 t km −2 year −1 , and the cultivated plot with contour tillage, with an average of 1067.62 t km −2 year −1 . In contrast, other kinds of plots had much lower mean annual SLRs. The mean annual SLR was 109.39 t km −2 year −1 for the grass plots, 74.16 t km −2 year −1 for the orchard plot with terrace, 55.97 t km −2 year −1 for the shrub plots, and 16.3 t km −2 year −1 for the forest plus shrub plots. No sediment ran out of the plots with forest plus fish scale pit and/or grass measures. Noticeably, almost all of the plots with soil conservation measures had higher soil loss reduction efficiencies of above 95%. The effective reductions of soil and water losses by vegetated measures were consistent with a number of studies [47][48][49][50][51]. Consistent with previous studies [26,47], terrace was effective in reducing runoff and soil loss. In the present study, soil loss from cultivated plot 17 with terrace was nearly 0 (i.e., 3.47 t km −2 year −1 ), and no runoff and soil loss occurred on plot 10, which had terrace and forest measures implemented, although its vegetation coverage was not very high. This implies that the combination of vegetation and terrace measures was more effective than a single measure in reducing soil and water losses [8]. In the present study, the mean annual SLR on plot 6 with a slope degree of 14.4 • was up to 2541 t km −2 year −1 , which is twelve times above the tolerable rate, although contour tillage was also implemented on this plot (Figure 3). In contrast, contour tillage can effectively reduce soil loss from plots 18 and 20, which have lower slope degrees (Table 1). This implies that contour tillage is not reasonable when it is implemented on steep slopes in the study region. which agrees with the former studies at a catchment scale [13][14][15]. In order to ensure water quantity of the Miyun Reservoir, soil conservation measures should be water-saving, which allows more runoff downstream to the reservoir. On the x-axis, bare represents plots 3, 4, 21, and 22, cultivated represents plot 1, cultivated + contour represents plots 6, 18, and 20, and cultivated + terrace represents plot 17. Orchard represents plot 2 and 7, shrub represents plots 5,9,11,13,14, and 19, forest + grass represents plot 8, forest + fish-scale pit represents plot 10, forest + shrub represents plot 12, and grass represents plots 15 and 16 in Table 1.
Annual SLRs presented the same changing patterns. The bare plots had the largest SLR, with an average of 4325.79 t km −2 year −1 , followed by the cultivated plot with no soil conservation measures, with an average of 3205.23 t km −2 year −1 , and the cultivated plot with contour tillage, with an average of 1067.62 t km −2 year −1 . In contrast, other kinds of plots had much lower mean annual SLRs. The mean annual SLR was 109.39 t km −2 year −1 for the grass plots, 74.16 t km −2 year −1 for the orchard plot with terrace, 55.97 t km −2 year −1 for the shrub plots, and 16.3 t km −2 year −1 for the forest plus shrub plots. No sediment ran On the x-axis, bare represents plots 3, 4, 21, and 22, cultivated represents plot 1, cultivated + contour represents plots 6, 18, and 20, and cultivated + terrace represents plot 17. Orchard represents plot 2 and 7, shrub represents plots 5, 9, 11, 13, 14, and 19, forest + grass represents plot 8, forest + fish-scale pit represents plot 10, forest + shrub represents plot 12, and grass represents plots 15 and 16 in Table 1.
The implemented soil conservation measures also significantly interfered with the relationships between event H and event SLR. Table 5 demonstrated that, except for the bare and cultivated plots, the SLR cannot be fitted at the 0.01 level. Note: the zero values for the 66 erosive events were assigned to 0.01 for power function regression. The relationships between the H and SLR were not given for plots 8 and 10 because no runoff occurred on these two plots.

Conclusions
The latest seven years (2014-2020) of monitored data from 22 runoff plots in the upstream catchment of the Miyun Reservoir in Beijing, Northern China were employed to study soil and water loss characteristics on slopes and their responses to slope, rainfall, and soil conservation measures, and some findings and implications were obtained. The bare land suffered the severest soil loss, with a mean annual SLR of 4325 t km −2 year −1 . Cultivated lands without any measures also experienced severe soil loss, with an annual SLR above 3205 t km −2 year −1 . Contour tillage cannot effectively control soil loss on the cultivated land with steep slopes. Soil conservation measures, such as forest, shrub, grass, terrace, and fish scale pits, as well as their combinations, can effectively reduce soil and water losses, with soil and water loss reduction efficiencies higher than 95% and 86%, respectively.
To protect drinking water safety, water quantity is a priority in the study region, and water-saving soil control measures should be considered to resolve the contradiction of soil loss control and water shortage in the downstream reservoir. The vegetation coverage of forest, shrub, and grass should be sparser, which allows more runoff to run downslope and simultaneously and effectively control soil loss. The size of the terrace on gentle slopes should be carefully designed. Bare and cultivated lands suffered the most severe soil erosion, with SLRs above 2732 t km −2 year −1 , and soil conservation measures should be implemented. Cultivated land with contour tillage still suffered a very high SLR (i.e., 2541 t km −2 year −1 ), and terrace or other vegetation measures (e.g., grass strips) should be implemented on this kind of land, because these types of measures can greatly reduce SLRs in the study region (Figure 4b).
Future studies can be conducted using physically based soil erosion models, such as the Water Erosion Prediction Plan (WEPP) and Soil and Water Assessment Tool (SWAT), to evaluate the effect of soil conservation measures on different slopes on downstream water resources.
Funding: This work was financially funded by the Beijing Natural Science Foundation (grant number 8202045) and the National Natural Science Foundation of China (grant number 41977066).

Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.

Data Availability Statement:
The data presented in the present study are available on request from the corresponding author.