Evaluation of soil erosion rates in the hilly-gully region of the Loess Plateau in China in the past 60 years using global fallout plutonium

land-use types. Widespread soil erosion at rates of 5.1 – 40.5 t/ha/yr. in this region in the past six decades was estimated. The influences of human activities in the past decades, type of land utilization, level of vegetation coverage, and terrain on the soil erosion in this region were discussed, and the accumulation of the eroded soil in the study sites was explored. Different types of land utilization showed diverse soil erosion rates (forest (slight) < grassland (light) < apple orchard ≈ cornfield (moderate)), indicating that natural vegetation rehabilitation, particularly restoring forest with high vegetation coverage is a practically effective conservation measure for soil erosion control; convex micro-topography on sloping fields is critical in alleviating soil loss by depositing eroded soil.


Introduction
Soil erosion is a severe ecological threat, detaching about 75 billion tons of soil per year from the world's terrestrial ecosystems (Pimentel and Kounang, 1998). The large erosion deprives the fertility of surface soil, and the deposition of the eroded soil on the riverbed and reservoirs further threaten the sustainability of water systems (Telles et al., 2011;Issaka and Ashraf, 2017). Uncontrolled human deforestation and land reclamation are the primary causes of soil erosion across the world, especially in the loess landscape. The Loess Plateau in northwestern China is one of the world's most susceptible regions to soil erosion, where 60 % of the area is eroded at rates of 50-100 t/ha/yr. (Cai, 2001;Zhao et al., 2013). The loess hilly-gully region located in the middle of the Loess Plateau is an important agricultural area owing to the large scale of loess platform terrain Huang and Gallichand, 2006). Meanwhile, this region have suffered serious soil erosion in the past few decades due to the damage of natural vegetation coverage by deforestation and land reclamation along with the expanding populations in the 1950s-1970s (Wei et al., 2006;Zhao et al., 2013). To prevent the ecological problems induced by soil erosion, various measures have been launched to mitigate soil erosion in some watersheds of the Loess Plautus since the 1970s, such as afforestation terracing, etc. (Zhao et al., 2013). From the 1990s, extensive measures were implemented in the Loess Plateau, such as the Forest Conservation Project and Grain-to-Green Project by returning the reclaimed sloping farmland to forest or grassland, to recover and improve the vegetation coverage (Lü et al., 2012). However, the extent of soil erosion in the hilly-gully region of the middle Loess Plateau during the past decades and the effectiveness of the current countermeasures on soil erosion have not been well evaluated.
Field survey, runoff plots, modeling and remote sensing are commonly used techniques to estimate soil erosion rate (Lal, 1994;Yue et al., 2005;Zheng et al., 2007;Mabit et al., 2008;Xu et al., 2015;Sepuru and Dube, 2018;Parsons, 2019;Zhang et al., 2019). Most of these methods are either time/labor/resource-consuming or qualitative evaluations for large areas (e.g., more than 10-30 km), which cannot provide precise estimations on soil erosion rates at the specific sites and insights into the soil erosion history in the past decades.
Radionuclides are effective tracers to monitor environmental processes in oceanography, pedology, atmospherics, etc. (Walling et al., 1999;Alvarado et al., 2014;Qiao et al., 2020). Based on the tight association of some naturally occurring and anthropogenic radionuclides with soil particles, this technique has been applied to trace the movement of soil particles, e.g. the soil erosion and redistribution processes (Zapata and Nguyen, 2009). The naturally occurring 7 Be and 210 Pb and anthropogenic 137 Cs, 239,240 Pu are the commonly used radionuclides for different erosion scenarios with the timescale stretching from months to decades (Walling et al., 1999;Zheng et al., 2007;Mabit et al., 2008;Xu et al., 2015;Zhang et al., 2019). The short-lived 7 Be (T 1/2 = 53.3 d) induced from cosmic radiations is suitable for short-term erosion events, such as individual rainfall events . 210 Pb (T 1/2 = 22.2 yr.), a decay product of 222 Rn escaped from soil to air, can be applied for a time-scale of decades, but its application is significantly limited by the difficulties on the accurate determination of low-level 210 Pb present in the soil. 137 Cs originated from the global fallout of nuclear weapon testing in the 1950s-1980s is the most commonly used radiotracer for the assessment of medium-term soil erosion (Zhang et al., 1990(Zhang et al., , 2003. As a gamma emitting radionuclide, 137 Cs can be readily measured by gamma spectrometry without chemical separation., However, due to its relatively short half-life (T 1/2 = 30.2 yr.), more than 75 % of the globalfallout-derived 137 Cs have decayed away since its maximum fallout in 1963, causing its measurement more and more challenging (Gering et al., 2002).
The long-lived 239 Pu (T 1/2 = 2.4 × 10 4 yr.) and 240 Pu (T 1/2 = 6.5 × 10 3 yr.) with the same origination and feature of strong association to soil particles as 137 Cs are ideal replacement of 137 Cs for this purpose (Alewell et al., 2017). The rapid development of mass spectrometry techniques, especially inductively coupled plasma mass spectrometry (ICP-MS), has enabled the determination of ultra-trace levels of 239 Pu and 240 Pu in environmental samples (Xing et al., 2018) and made it an alternative of 137 Cs to estimate soil erosion. Global-fallout derived 239,240 Pu (sum of 239 Pu and 240 Pu) has been used to evaluate soil erosion in different areas with diverse vegetation types or terrains, such as in the forested catchment of Germany (Calitri et al., 2020), the loess landscape of Poland (Loba et al., 2021), the wet-dry tropics of Australia (Lal et al., 2020), the Central Swiss Alps (Musso et al., 2020), and the bay region of northeastern China (Xu et al., 2013;Zhang et al., 2016). The application of plutonium isotopes in soil erosion estimation in the Loess Plateau is very limited. Zhang et al. (2019) estimated soil erosion rates in two slopes in a catchment (covered by artificial forest and natural grass, respectively) in the Loess Plateau using plutonium isotopes for the first time. The investigated sites in their study were specialized restoration areas with only two land utilization types.
In this study, we aim to investigate soil erosion rates in the hilly-gully region with different land utilization and vegetation types by determining 239 Pu and 240 Pu in soil profiles collected from the uncultivated sites (forest and grassland) and cultivated sites (apple orchard and cornfield) in the Heimugou (Luochuan County), a typical watershed in the hilly-gully region of the middle Loess Plateau of China, to understand the change of the erosion rate and the major influencing parameters in the past 60 years, as well as to evaluate the effectiveness of the soil erosion mitigation measures.  (Fig. 1). This 7-km long gully spreads from south to north with slopes of 20-60 • on both sides and flat land on the top. A stream flows at the bottom of the catchment to the Xianguhe River which joins to the Luohe River and finally flows to the Yellow River. The Heimugou Catchment is a typical watershed in the Beiluo River Basin in the middle Loess Plateau that has contributed about 7.1 × 10 7 t of soil particles to the Yellow River through the erosion processes during 1958-2012 (Zhu et al., 2004;He et al., 2016). The altitude difference of the catchment is about 250 m from the lowest point at the bottom entrance (elevation = 870 m) to the highest point at the top (elevation = 1120 m).

Study sites and sampling
The mean annual precipitation in this region is 533 mm in the past 60 years, mostly (60-70 %) occurring from July to September (Fig. S1). The mean annual evaporation (1629 mm) is about three times higher than the mean annual precipitation, resulting in a relatively dry condition of the soil for most of the year. The annual mean wind speed is quite constant throughout the whole year, with an annual mean value of 2.2 m/s (Fig. S2). The main soil types in the Heimugou Catchment are Quaternary loess, Neogene red clay, and Triassic sandstone-mud-stone. The soil type of all sampling sites in this study is Lishi loess, a kind of Quaternary loess, developed in the late Middle Pleistocene with a siltdominated texture that is easily eroded . The major natural vegetation species in the hilly-gully region of the Loess Plateau are Robinia pseudoacacia for trees and Agropyron cristatum, Imperata cylindrica (L.) Beauv, and Bothriochloa ischaemum for grasses, and the main crop types are maize, wheat, and apples. In the study area, most of the farming land is located on flat loess tableland, and the slopes on both sides of the gully are grassland and forest (part of them also on the flat Loess tableland). Eight soil cores with 40-65 cm depth were collected from four natural grassland sites, two forest sites, and two arable land sites in the Heimugou Catchment in July 2020 ( Fig. 1, Table 1). Four grassland sites are successively located on a grassland slope with 80-90 % grass coverage and no visible disturbance (Fig. 2). Among them, one site is located on the top flat area (G1); two sites are from the slope with gradients of 20 • (G2) and 40 • (G3), respectively; and one site is on a platform with a gradient of 2 • (G4) at the bottom of the slope and nearby a vertical scarp of 100-150 m height above the catchment bottom.
Two forest soil cores (F1 and F2) were collected from a top flat area covered with condensed locust trees and grass (vegetation coverage > 90 %) in the north of the Heimugou Catchment without disturbance by human activities since the 1950s. Two soil cores were collected from an apple orchard (A1) and a cornfield (A2) on the flat arable field area of the Heimugou Catchment, respectively. The apple orchard (total 2668 m 2 ) was used for cultivating winter wheat in 1950-1995 and has been planting apple trees since 1995. Tree spacing is 4 m between the rows and 3 m between the plants on a row. The cornfield was typical rain-fed farmland with the maize's growing period from April to October. This field was seeded with winter wheat from 1950 to 1985 and then converted to cultivate maize since 1985.
The soil cores from natural grassland (G1, G2, G3, and G4) and forest (F1 and F2) were manually collected using a stainless-steel spade (10 × 10 cm) with 2-cm intervals in the upper 30 cm and 5-cm intervals for the depth of 30-65 cm. Two 60-cm deep soil cores (A1 and A2) were collected with the stainless-steel soil tube auger (4.0 cm in diameter) and divided into 5-cm intervals. The collected soil samples were sealed in plastic bags and transported to the laboratory. After removal of plant roots and stones (>2 mm), the samples were weighed and dried in an oven at 105 • C until constant weight. An aliquot of soil sample was taken for measurement of grain size, and the remaining soil was ground and sieved through an 80-mesh sieve.

Radiochemical analysis of plutonium isotopes in soil samples
A modified method by Zhang et al. (2019) was applied for the determination of plutonium isotopes in the soil samples. In brief, 5-10 g of the homogenized soil powder was ashed at 450 • C for 12-15 h to remove organic matter. The difference of the sample weights before and after the ashing was measured as the loss-on-ignition (LOI) to estimate the organic matter content in the sample. 242 Pu tracer (~5 mBq) was spiked to the sample to monitor the chemical yield of plutonium during chemical separation. The sample was then digested with 80-100 mL of Aqua regia at 150 • C for 30 min and at 200 • C for 2 h. The leachate was filtrated through a glass fiber filter, and the residue was rinsed with diluted 0.5 M HNO 3 . 2.0 M NaOH solution was added to the leachate to adjust pH8-9, and the formed hydroxides co-precipitate was separated by centrifuging to remove the matrix elements. The precipitate was dissolved with 3-5 mL of conc. HCl, 200-300 mg of K 2 S 2 O 5 was added to the sample solution to reduce plutonium to Pu(III). 20 % ammonia solution was then added to the sample solution to adjust pH8-9 to coprecipitate Pu(OH) 3 with Fe(OH) 2 . After separation of the precipitate by centrifuge, 1-2 mL of conc. HCl was added to dissolve the precipitate, and then conc. HNO 3 was added to convert Pu(III) to Pu(IV) and prepare the sample in 3.0 M HNO 3 solution. The prepared sample solution was  * LOI (loss on ignition approximates soil organic matter) in percentage compared to the total mass of the soil, the average, and range.
# the percentages of different sizes of soil particles (20-200 μm, 2-20 μm, and < 2 μm). § bulk density and pH are the average values of soil in the whole cores. NA: not measured. loaded to a 2-mL TEVA resin column preconditioned with 40 mL of 3.0 M HNO 3 . After rinsing with 80 mL of 3.0 M HNO 3 and 40 mL of 6.0 M HCl to remove uranium, thorium and other remaining interfering elements, Pu on the column was eluted with 20 mL of 0.1 M NH 2 OH•HCl-2.0 M HCl solution by reducing Pu(IV) to Pu(III). The eluate was evaporated to dryness at 200 • C, and the residue was dissolved with concentrated HNO 3 . After heating at 200 • C to decompose NH 2 OH•HCl, the solution was evaporated to dryness. The residue was dissolved with 4 mL of 0.5 M HNO 3 . Plutonium isotopes ( 239 Pu, 240 Pu, and 242 Pu) in the sample solution were measured using a triple quadrupole ICP-MS (Agilent 8800). NH 3 -He was applied as the reaction gas to eliminate the interferences of 238 UH + and 238 UH 2 + . The measurement sensitivity for 239 Pu and 240 Pu was 710 cps/ppt. The chemical yields of plutonium in the chemical separation were measured to be 75-95 %. Based on using the procedure blank, the detection limits of the analytical methods for 239 Pu and 240 Pu were estimated to be 2.0 × 10 -3 and 3.0 × 10 -4 mBq/ g, respectively.

Estimation of soil erosion rate
The soil erosion rate was estimated by the comparison of the 239,240 Pu inventory in an investigated site with that in the reference sites without obvious soil erosion and accumulation. As various degrees of soil erosion occurred in all sample sites, we used a reference site from the Dongzhuanggou catchment of the Loess Plateau (as discussed in 3.2) nearby the studies area. The assumptions of this approach are: (1) the plutonium isotopes in the study area only originated from the global fallout; and (2) the sampling sites and reference sites have the same 239,240 Pu deposition and depth distribution (Alewell et al., 2017). The 239,240 Pu inventory in soil core (Bq/m 2 ) was calculated using the following equation: where A is the cross-sectional area (m 2 ) of the soil core, M i is the mass of the ith depth increment of a soil core (kg), and C i is the 239,240 Pu concentration (Bq/kg) of the ith depth increment. The Modelling Deposition and Erosion rates with Radio-Nuclides (MODERN) model was used to calculate soil erosion rates in this work (Arata et al., 2016a,b). Different from the previous models (e.g. the Proportional Model (Zapata, 2002), the Profile distribution Model Walling and Quine, 1990), the Inventory Method (Lal et al., 2013;Portes et al., 2018), the Mass Balance Model (Zapata, 2002) and the Diffusion and Migration Model (Meusburger et al., 2018), the MODERN conversion model could freely define the depth profile of plutonium at the reference site, which is more flexible to adapt to the natural scenarios. Besides, MODERN model allows simulation of different disturbing activities (e.g. tillage, erosion, and deposition) at the reference site, thus assuring the comparability of reference and sampling sites.
The detailed calculation method using the MODERN model in this work are presented in the Supporting Information. Two stepwise functions for the depth profile of 239,240 Pu inventory at the reference sites were established for two different scenarios of the studied sites (Fig. S3). For unploughed sites (forest and grassland sites), the stepwise function was based on the original depth profile of 239,240 Pu inventory determined at the reference site. Whereas for the ploughed sites (apple orchard and cornfield), considering the homogenized effect of ploughing on the vertical distribution of 239,240 Pu, the corresponding stepwise function was adapted from that for uncultivated sites by assuming an average inventory value at all the layers above the ploughing depth (20 cm in this study).
To estimate the thickness of eroded soil, the 239,240 Pu accumulative inventory at the sampling site was calculated for the whole depth profile. The MODERN model searched in one of the two established stepwise functions for reference site (unploughed or ploughed, depending on the scenario of the eroded site) to target a depth where the sum of all inventories in this increment and those below is equal to the 239,240 Pu accumulative inventory of the sampling site. With the calculated thickness of soil loss (x*, in cm) at the sampling site by MODERN, erosion rate Y in t/ha/yr. could be estimated by the following equation: where xm is the mass depth (kg/m 2 ) of the sampling site, d (cm) is the whole depth increment of soil core at the sampling site, t 1 and t 0 are the sampling year (2020) and the reference year (1963), respectively.

Level and distribution of plutonium in the soil profiles
The 239,240 Pu activity concentrations in all soil profiles collected in the Heimugou Catchment range from 0.002 to 0.34 mBq/g (Fig. 3). The peaks of 239,240 Pu concentration present at the depth of 10-20 cm in the uncultivated soil profiles (G1-G4, F1, and F2), followed by an exponential decline with depth in the deeper layers. The cultivated soil cores (A1 and A2) showed relatively homogeneous distributions of 239,240 Pu in the ploughing layer of 0-20 cm.
The 240 Pu/ 239 Pu atomic ratios (Fig. 4) in all soil profiles vary within 0.154-0.217 (excluding the results with high uncertainty in the deep layers) and have a mean value of 0.184 ± 0.021 (1σ), which agrees well with the typical ratio (0.18 ± 0.014) of global-fallout originated plutonium from atmospheric nuclear weapons tests (Kelley et al., 1999). The distributions of the 239,240 Pu inventories in each layer for the investigated sites with the total inventory in the entire soil column are shown in Fig. 5. A large variation was observed for the total inventories of 239,240 Pu from 33.87 ± 2.62 to 82.66 ± 4.98 Bq/m 2 in the investigated soil cores collected in the Heimugou Catchment region, which are significantly lower than the reported levels for the sites at similar latitudes in the United States, South Korea, Italy, and China (84.5-231 Bq/ m 2 , Table 2) (Lee et al., 1996;Ketterer et al., 2002;Xu et al., 2015;Portes et al., 2018;Raab et al., 2018;Ni et al., 2018;Cao et al., 2019;Zhang et al., 2019). These values are also much lower than the observed level at a reference site in Qingyang (35.7 • N, 107.5 • E; 111 ± 5 Bq/m 2 ) in the Loess Plateau near the sampling sites of this work (Zhang et al., 2019), indicating that remarkable soil erosion occurred in the study region.
Among the investigated soil profiles, the total 239,240 Pu inventories in the forest sites (61.00 ± 3.12 Bq/m 2 at site F1 and 79.04 ± 3.97 Bq/m 2 at site F2) are higher than those in the grassland sites (36-49 Bq/m 2 at sites G1, 2, and 4) except for site G3 with the maximum inventory (82.7 Bq/m 2 ). The sites G1, G2, G3, and G4 are located in a hilly grassland from top to the bottom (Fig. 2). The total 239,240 Pu inventories at sites G1, G2 and G4 are comparable (around 42 Bq/m 2 ), which is only about half of that at site G3. These three sites (G1, G2, and G4) show relatively low 239,240 Pu inventories and 239,240 Pu mainly presents in the uppermost 30 cm layer of soil, indicating an erosion of the upper layer of the soil no matter at the flat or the slope sites. Compared to the other sites in the grassland (G1, G2, G4) with gentle slopes (<20 • ), site G3 is at a steeply convex spot (40 • ) at the middle of the grassland with a deeper 239,240 Pu depth distribution (0-50 cm). Its peak value of 239,240 Pu inventory occurred at 20 cm depth in the soil core, which is different from the typical exponential declined trend or peaked value at the subsurface (3-5 cm depth) followed by an exponential declined trend of 239,240 Pu inventory with depth (Alewell et al., 2017). These features suggest that  there might be an accumulation of soil eroded from the upper part of the grassland slope at site G3. The low total 239,240 Pu inventories at two farmland sites (40.14 ± 2.71 at A1 and 33.87 ± 2.62 Bq/m 2 at A2) and the relatively homogeneous 239,240 Pu distribution in these two soil cores indicate a high soil erosion level and significant disturbance of the soil profile by the tillage activities in the 20 cm topsoil of these sites.

Soil erosion rates in the Heimugou Catchment
Based on the depth distribution of 239,240 Pu inventories, the soil erosion rates at each site were estimated using the MODERN model in comparison with the reference site. Because of the overall low 239,240 Pu inventories at all investigated sites in this work (including the undisturbed flat-top sites with forest or grass coverage), an alternative reference site (Re) in the Loess Plateau in Qingyang was used (Zhang et al. 2019). This site is located in another watershed (Dongzhuanggou), which is 175.2 km away from the studied area and has a similar climate (annual average precipitation of 556 mm) and soil type (Loess). This site has been well protected by vegetation conservation measures since 1954 without obvious soil erosion or deposition. Considering the similar environmental conditions and clear vegetation conservation history, the Re site is the best available reference soil profile for this research. The reference soil profile sample was collected without visible soil erosion and deposition and analyzed for 239 Pu and 240 Pu. An exponentially declining depth profile of 239,240 Pu concentrations in a soil core sampled from a flat natural grassland at this reference site has been reported (Zhang, et al. 2019), and the total inventory of 239,240 Pu in this soil core was estimated to be 111 ± 5 Bq/m 2 (Table 2 and Fig. S3), which is significantly higher than those observed in the investigated area and comparable with the values of other similar latitude regions in Table 2. However, the investigations of deposition and soil depth profiles of plutonium isotypes ( 239 Pu and 240 Pu) at reference sites in the hilly-gully region of the Loess Plateau are still limited, and more relevant data will be supplied to calibrate the soil erosion rates evaluated by the plutonium  isotopes in the study area. The estimated erosion rates at the forest, grassland and arable sites (Table 3) range from 5.1 t/ha/yr. at the natural forestland to 40.5 t/ha/ yr. at arable sites. These results are generally comparable with the reported average soil erosion rate of 34.1 t/ha/yr in Luochuan during 1997-2009, which was estimated by the Chinese Soil Loss Equation method based on remote sensing and field surveys (Li et al., 2013). Our results also agree with the estimated soil erosion rate of 2-57 t/ha/yr. in the Heimugou Catchment during 1963-1988 using the 137 Cs approach (Zhang et al., 1990). Because there is a potential soil deposition at the grassland's middle site G3 after an initial erosion (see detailed discussion in subsection 3.3), the MODERN model cannot be used to estimate the accurate soil erosion and accumulation results for site G3 based on the comparison of total 239,240 Pu inventories with the reference sites. Because the MODERN model, as well as other common conversion models (the Proportional Model, the Inventory Method, etc.), are more appropriate to quantify the single soil erosion rate or deposition rate of a site rather than the complicated scenarios like G3 site. According to the "Standard of Soil Erosion Classification and Gradation" (SL 190-2007) in China (Ministry of Water Resources of PR China, 2008;Zhang et al., 2015), the uncultivated forest sites (F1 and F2) could be categorized as "slight erosion" (<10 t/ha/yr.), grassland (G1, G2, and G4) as "light erosion" (10-25 t/ha/yr.) and the apple orchard (A1 and A2) as "moderate erosion" (25-50 t/ha/yr.).
The significant soil erosion rates observed at the investigated sites in this work should be attributed to the significantly reduced vegetation coverage caused by intensive reclamation and deforestation in the Loess Plateau in the 1950s-1970s (Rozelle et al., 1997;Zhao et al., 2013). Only 20-40 % of vegetation coverage was measured in the Heimugou Catchment in the late-1980s by remote sensing in the Beiluo River Basin where the Heimugou Catchment is located (Chen, 2013). The low vegetation coverage caused a bad reservation and fixation of soil and exacerbate soil erosion rates in the Loess Plateau (Zheng, 2006;Zhou et al., 2006). An average soil erosion rate of 20 t/ha/yr. was roughly estimated for a grassland slope in the studied region from 1963 (global fallout peak) to 1988 using the 137 Cs approach, corresponding to a total soil loss of 500 t/ha () in this period (Zhang et al., 1990). The average soil erosion rate of the Yangjuangou watershed (36.70 • N, 109.52 • E) in the hilly-gully region in the north of Heimugou Catchment was estimated to be about 89.8 t/ha/yr during 1991-1996 (Li et al., 1997;. A high average sedimentation rate of 52.7 t/ha/yr measured in check-dam systems in the Fangta catchment (36.78 • N, 109.24 • E) in the north of studied area during 1975-1987 was estimated (Wei et al., 2022). The abovementioned results confirmed that severe soil erosion happened in the hilly-gully region of the Loess Plateau in the past decades.
With the implementation of soil and water conservation practices, especially the Green to Grain program launched in 1999, the reclaimed farmland was returned to forest and grassland, and the vegetation coverage has been significantly improved in the Loess Plateau. A remarkably increased vegetation coverage up to 60-80 % in the Heimugou Catchment has been observed in the Beiluo River Basin by 2007 (Chen, 2013), which significantly mitigate the soil erosion in this region in the recent years.
The average soil erosion rate in the grassland slope (G1, G2, and G4) in the Heimugou Catchment was estimated to be 11.9 t/ha/yr with a total soil loss of 678 t/ha during 1963-2020. Compared with the reported total soil loss of 500 t/ha in 1963-1988 by Zhang et al., (1990), more than 70 % of soil loss occurred before the late 1980s, and soil erosion has significantly decreased in the recent decades. Similar erosion rates of 10.39 t/ha/yr. for shrub land and 11.86 t/ha/yr for grassland after 1960-2016 in the Wangmaogou Watershed (37.58 • N, 110.35 • E) of the Loess Plateau were also reported (Zhang et al., 2020). The soil erosion rates in the Yangjuangou Catchment (36.70 • N, 109.52 • E) estimated by 137 Cs tracer also revealed that the land covered with mature forest (>25 yr.) and grass have relatively lower soil erosion rates compared with young forest, orchard, and cropland (Fu et al., 2009). These results demonstrate that vegetation restoration practices have effectively suppressed soil erosion.

Key factors influencing soil erosion rate
A large variety of soil erosion rates at different sites was estimated in the investigated area using the 239,240 Pu tracer. Although the loose soil (loess) and concentrated precipitation (60 %-70 % in the summer) in the Loess Plateau could aggravate the soil erosion rates, the similar climate conditions in this relatively small watershed in the investigated area could not cause such a large variation in soil erosion rates.
The steep slope in the Heimugou Catchment (typically 20-60 • ) generally favors water erosion by increasing the velocity and flow rate of runoff (Fang et al., 2015;Wang et al., 2019), whereas the sampling sites in this study (except for the site G3) have relatively gentle and similar slopes (≤20 %), which has a limited impact on the variation of the soil erosion ratio in these sites.
The different land utilization and types of vegetation coverage in the Heimugou Catchment might be the key factors responsible for the different soil erosion rates. Increased soil erosion rates were observed for the sites in an order of forest < grassland < arable land in the Heimugou Catchment. The lowest soil erosion rates (5.1-7.6 t/ha/yr.) were observed at sites of the forest (F1 and F2), followed by grassland (G1, G2, and G4) with relatively low erosion rates of 10.1-13.0 t/ha/yr. This is probably owing to the fact that forests and grassland can effectively reduce the intensity of the kinetic energy of raindrops and water flows (Lacombe et al., 2018;Zhang et al., 2019) as well as the wind speed above the soil surface (Li et al., 2005;Wolfe and Nickling, 1993). The interlaced roots of vegetation with their exudates and microflora in the forest and grassland can also increase the cohesion and fixation of soil particles (Amézketa, 1999;Hudek et al., 2017;Musso et al., 2020;Pintaldi et al., 2018), inducing a suppression on soil erosion. Half soil erosion rates in the forests compared to that in the grassland should result from the different seasonal variabilities in vegetation coverage between these two types of land. The forest sites have relatively constant vegetation coverage throughout the whole year. The dominant species of tree in the forest in this region is the locust, a perennial deciduous tree species with a large forest canopy and extensive litter, which effectively protects the surface soil from erosion by raindrops and wind, even in the winter. Whereas natural grassland can only maintain high vegetation coverage in the growing season (i.e., from late spring to autumn), and the withering of grass in other seasons (i.e., winter) leads to the decrease of grass coverage, exposure of bare soil surface, and therefore increased Table 3 The estimated soil erosion depth, rate and intensity in the studied sites in the Heimugou Catchment. wind erosion occurring in the winter (Hou et al., 2020). The higher erosion rates at the grassland sites than those at the forest sites were also reported by other studies in the Loess Plateau and Sichuan Basin (Fu et al., 2009;Li et al., 2009). The highest soil erosion rate (32.5-40.5 t/ ha/y) was observed at the cultivated land sites (A1 and A2). The apple orchard (site A1) was mainly covered by low-density trees (about 12 m 2 / tree) and less other vegetation (e.g., grass). The cornfield (site A2) is only covered by cultivated plants in April to October. The less vegetation coverage and not well-developed root system in the surface soil of the cultivated land compared to the natural grassland and forest cause higher soil erosion by water and wind. The low content of organic matter in the cultivated land (Table 1 and Fig. S4) arising from the low vegetation coverage hampers the aggregation of soil particles and water infiltration, exacerbating water-induced erosion (Conforti et al., 2013). The studies in the upper Min River watershed and the Loess Plateau also showed that vegetation coverage has a considerable influence on the soil erosion rate (Jin et al., 2021;Zhou et al., 2008).

Re-location of the eroded soil in the grassland slope
The highest total inventory of 239,240 Pu (82.7 Bq/m 2 ) and the deepest distribution of 239,240 Pu (50 cm) in the soil core was observed at site G3 in the natural grassland over all investigated sites in the Heimugou Catchment (Fig. 5). The accumulated inventory and distribution of 239,240 Pu in the 20-50 cm section at site G3 are similar to the 0-30 cm sections at sites G1, G2, and G4 (Fig. 6). Based on the assumption that there were the same initial 239,240 Pu distributions in all soil profiles, there might be a similar soil erosion to the other grassland sites that occurred at site G3 in the early date but accompanied by deposition of 20 cm eroded soil at the top of soil core afterwards (Fig. 6). The significant accumulation of the eroded soil at site G3 might be attributed to the convex terrain where the fine particles eroded by the downwardflowing stream from the upper slope were deposited and accumulated ( Fig. 2a and 2b). This is supported by the significantly larger fractions of the average fine soil particle content (<2 µm: 9.3 %; and 2-20 µm: 59.2 %) in the soil profile at site G3 compared to those at sites F1, F2, G1, and G2 (<2µm: 6.2-8.1 %; and 2-20 µm: 37.1-46.3 %) (Fig. 7). Fine soil particles, compared to heavier coarse soil particles, could be easily removed by the inertial force of water flow (Wang and Shi, 2015), resulting in a deposition of the eroded fine particles at site G3. This is also supported by the observation that the eroded sediment contains 2.7-18.9 % more high-fine particles (<2 µm) than the on-site soil (Poesen and Savat, 1981;Tuo et al., 2014;Wang and Shi, 2015). This indicates that the convex slope could trap the eroded soil particles and reduce the overall loss of soil on the slope. This also agrees with the previous investigation in another catchment in the middle Loess Plateau (Zhang et al., 1990), where nearly twice 137 Cs inventory in the soil cores at the lower site with the convex terrain was observed compared with the upper side of the same slope. This support that the convex terrain is effective for deposition of the eroded soil from the upper part of the slope. An experimental study also indicated that an undulant microrelief consisting of mounds and troughs is more effective to mitigate soil loss than a smooth surface (Zhao et al., 2021). However, the total 239,240 Pu inventory at site G3 is still lower than that at the reference site, indicating that the deposition of eroded soil particles of site G3 from the upper slope may have only occurred in the later stage after the reclamation because the severe soil erosion eroded most of the 239,240 Pu inventories in the 1950s-1970s or even until the 1980s, while the deposition layers of G3 have relatively low 239,240 Pu inventories.
In the Heimugou Catchment, most of the eroded soil particles might be flushed to the bottom of the catchment along the steep slopes by the surface runoff after heavy rainstorms, leading to a large amount of soil deposition at the bottom of the catchment. Due to the typically interlaced structure of plateau, gully, and ridge in the Loess Plateau, the surface soil from the sloping areas is often moved by water flow and wind to the low-lying locations and deposited in the hilly-gully regions of the Loess Plateau (Hessel and Van Asch, 2003;Wei and Shao, 2007). The eroded soil particles, especially the fine soil particles might be also flowed to the rivers in each catchment and then to the Yellow River, and eventually be deposited in the estuarine area of the Yellow River in the Bohai Sea (Zhu et al., 2004;He et al., 2016).

Conclusion
The soil erosion in the Heimugou Catchment, a typical watershed located in the middle of the Loess Plateau, was investigated using a 239,240 Pu tracer. Land reclamation and deforestation in the 1950-1970s resulted in a wide and relatively severe soil erosion (with erosion rates of 5.1-40.7 t/ha/yr.) in this region. And the soil and water restoration practices (such as the Green to Grain project) implemented since the 1980s by recovering the vegetation coverage through returning the reclaimed sloping farmland to forest and grassland significantly suppressed soil erosion rates. Compared with the arable land, forest and grassland with higher vegetation coverage in most seasons of the whole year are prone to resist soil erosion, confirming the effective practice in  the Loess Plateau over the past 2-3 decades. The convex microtopography can play an important role in the resistance of soil loss in the slope area, which provides a practical strategy for the mitigation of soil loss in the Loess Plateau.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Data availability
Data will be made available on request.