A Reliable U-trough Runoff Collection Method for Quantifying the Migration Loads of Nutrients at Different Soil Layers under Natural Rainfall

: Long-term quantiﬁcation of the migration loads of subsurface runoff (SSR) and its collateral soil nutrients among different soil layers are still restricted by the runoff collection method. This study tested the reliability of the U-trough collection methods (UCM), compared with the seepage plate collection method (SPM), in monitoring the runoff, sediment and nutrient migration loads from different soil layers (L 1 : 0–20 cm depth; L 2 : 20–40 cm depth; L 3 : 40–60 cm depth) for two calendar years under natural rainfall events. The results suggested that the U-trough could collect nearly 10 times the SSR sample volume of the seepage plate and keep the sampling probability more than 95% at each soil layer. The annual SSR ﬂux from L 1 to L 3 was 403.4 mm, 271.9 mm, and 237.4 mm under the UCM, 14.35%, 10.56%, and 8.41% lower than those under the SPM, respectively. The annual net migration loads of sediment, TN, and TP from the L 1 layer under the UCM were 49.562 t/km 2 , 19.113 t/km 2 and 0.291 t/km 2 , and 86.62%, 41.21% and 81.78% of them were intercepted by the subsoil layers (L 2 and L 3 ), respectively. While their migration loads under the SPM were 48.708 t/km 2 , 22.342 t/km 2 and 0.291 t/km 2 , and 88.24%, 53.06% and 80.42% of them were intercepted, respectively. Under both methods, the average leached total n (TN), total p (TP) concentrations per rainfall event and their annual migrated loads at each soil layer showed no signiﬁcant difference. In conclusion, the UCM was a reliable quantitative method for subsurface runoff, sediment, and soil nutrient migration loads from diverse soil layers of purple soil sloping cultivated lands. Further studies are needed to testify the availability in other lands. of fertilizer applications for summer maize and one-time fertilizer application for winter mustard, was set up in each year in this study. The fertilization was applied at a depth of 5–10 cm to avoid surface loss. This study had been conducted over two consecutive years’ ﬁeld monitoring.


Introduction
The surface-subsurface water pollution is generally owing to the agricultural nonpoint source pollution [1]. Surface runoff (SR) and underground leaching are the main concurring sources for agricultural pollutants to enter the water environment [2,3]. The subsurface runoff (SSR), as an essential part of runoff, presents important influences on soil nutrients leaching loss in the watershed [4]. The factors influencing SSR include soil physical and chemical properties [5,6], geobiont [7,8], plant roots [9,10], ground slope [11,12] and roughness [13,14], climate including rainfall [15,16] and freeze-thaw [17], surface coverage [18,19], and farming practices [20,21]. However, such complex and diverse factors make soil nutrients' underground loss difficult to measure directly [22,23]. Therefore, it is vital to develop a reliable method of monitoring soil nutrient migration at different soil layers to control groundwater pollution or non-point source agricultural pollution.
Nowadays, several methods have been explored to study soil leakage water. The tile drainage system was widely used to monitor the SSR and its collateral soil nutrients migration in the United States and Canada. Bosch et al. [24,25] and Sadhukhan et al. [26] installed loop-porous-corrugated-tile drains equidistantly at a depth of 0.8-1.5m. The length of each tile drainage was consistent with the land plot length. Otherwise, according to the topography, the tile drains would be spaced irregularly in the depression areas [27][28][29]. The fluid collection buckets were then allocated at the end of the down-slope tile-drain to collect runoff samples [30,31]. This method can reduce the groundwater level and pressure in the soil and improve the horizontal hydraulic conductivity. The soil water can be effectively discharged out to increase the total amount of SSR [32]. That is why the drainage system is widely used. However, it has not been used for the quantitative study in the vertical direction. Bjerkholt et al. [33] adopted a porous corrugated polyethylene (PE) pipeline drainage system to study the soil particle and nutrient loss characteristic. However, due to underground soil's small hydraulic conductivity, the flow collected in this system was limited [34].
Conflux troughs, fixed outside of the down-slope soil body, were also available for collecting SSR flowing out from the soil section. Huang et al. [35] used these troughs at the runoff flume's down-end to collect SSR in the laboratory. Zhu [3], Deng et al. [36][37][38], and Wang et al. [39][40][41] adopted the conflux troughs at the lower end of the sloping land to collect SR and SSR in field runoff plots. The collected runoff was then introduced into the runoff storage bucket by water pipes [42]. Mazur [43] adopted this method to study the runoff's quantity and quality of different soil layer cross-sections (0 m, 0.00-0.25 m, 0.25-0.50 m; 0.50-0.75 m) at an eroded loess slope land. Additionally, Zheng et al. [44] studied the nitrogen loss characteristics by collecting the SSR permeated out of the A h (soil depth: 0-30 cm), B s (soil depth: 30-60 cm), and B sv (soil depth: 60-105 cm) sections of red soil slopes in southeast China. Nevertheless, these conflux troughs can only get the groundwater oozing laterally from the cross-section, but not be suitable for quantifying the migration characteristics in the vertical soil layers.
Diverse lysimeters also are standard tools for soil water collection. Liao et al. [45] collected and measured SSR's nutrient concentrations through zero-lysimeters to analyze the tea garden's nitrogen leaching loss characteristic. Li et al. [46,47] studied soil nitrogen loss characteristics by collecting the water flux and soil solution in the vertical direction through a self-made lysimeter made from a PVC pipe. Biernat et al. [48] explored the nitrate leaching characteristic by taking ceramic suction cups as lysimeters. The ceramic suction cups had a length of 54 mm, a diameter of 20 mm, and numerous 1 µm pores. Izydorczyk et al. [49] analyzed the change of nitrate and phosphate content in groundwater by the piezometers. These lysimeter methods can only collect a small amount of soil water sample, while they are heavily limited by the rainfall amount. Additionally, Comino et al. [50] caved a vertical profile (50 cm depth and 150 cm width) under a simulator ring to observe the infiltration dynamic and observe the subsurface flow with a mental collector under simulated rainfall. However, it was impossible to quantify the subsurface flow because the soil boundary was open. Therefore, they are not available for quantitative analysis of runoff and soil nutrient migration load.
The accuracy of quantitative research on soil nutrient loss depends on the feasibility of its collection method. The environment in labs was in control, but the complex environmental factors in the field were out of human's control so that the nutrient loss in the lab might be higher than [51], or similar to [29], or lower [52] than the underground losses. Therefore, the quantitative study on soil nutrient migration, especially on the field's stratified migration of soil nutrients, is still subjected to its collection method. Zhang, et al. [53] quantified the nitrogen loss in the Three Gorges Reservoir area's sloping land by the seepage-plate collection method (SPM). Hangen et al. [54] quantified the forest soil's preferential flow by the SPM. A U-trough collection method (UCM) was explored in this study. This study com-pared the SPM with UCM by its collected amounts of runoff, sediment, and their nutrient (n and p) concentrations and their annual migration loads to testify the reliability of UCM on qualifying soil nutrients migration among different soil layers under natural rainfall events. It concluded that the UCM was a reliable method of quantifying the stratified soil's runoff, sediment and nutrient migration loads. Notably, the UCM was applied as an invention patent (application No. 202010204850.9) in China on March 22, 2020. Figure 1 showed the necessary information of experimental plots in this study. The experimental plots are located in the National Monitoring Base for Purple Soil Fertility and Fertilizer Efficiency (29 • 48 38" N, 106 • 24 40.6" E, Figure 1a), Beibei, Chongqing, China. The climate is a typical subtropical humid climate according to the Climatological Classification of China. The National Meteorological Station of Beibei (Station No. 57511) showed that the annual average temperature is 18.3 • C, and the mean annual precipitation is 1105.4 mm. The test plots were built on February 28, 2018. As shown in Figure 1, the size of each plot is 9 m (length) × 3 m (width) × 0.8 m (depth). The flank walls and floors of these grids were made from reinforced concretes. The soil layers of each plot from top to bottom were 20 cm depth of surface-cultivated mellow soil (L 1 ), 20 cm depth of plow pan soil (L 2 ), and 20 cm depth of clastic purple shale parent materials (L 3 ), respectively, as shown in Figure 1b,c. The experimental soil, developed from the Jurassic Shaximiao Formation, is purple soil, classified as Regosols in Food and Agriculture Organization of the United Nations (FAO) Taxonomy or Entisols in United States Department of Agriculture (USDA) Taxonomy. It is a unique soil type in China. Purple soil has the characteristics of a dualistic geotechnical structure of overlying soil and underlying rock [55]. Plentiful soil pores, strong permeability, and weak water holding capacity of purple soil may cause severe soil erosion [56] under the climate environment of concentrated rainfall and heat [11], especially in slope-cultivated land area [57]. The basic physicochemical properties of the top layer soil were pH (6.29), total nitrogen (1.14 g/kg), total phosphorus (0.72 g/kg), total potassium (21.65 g/kg), soil organic matter (14.96 g/kg), soil bulk density (1.22 g/cm 3 ), and soil porosity (53.96%). The surface slope of all grids is 15 • [58-60].

Experimental Design
This research aims to verify the reliability of UCM on soil water, sediment and nutrient migration load at different soil layers (L 1 : 0-20 cm, L 2 : 20-40 cm, L 3 : 40-60 cm), compared to the SPM [53]. The UCM was designed to collect the SSR of a consecutive belt-region from the top slope to the bottom slope, while the SPM was used to collect the SSR of a dot area at the top slope and the bottom slope. As shown in Figure 1b,c, two U-troughs or two seepage plates were installed and fixed at each soil layer's bottom. Each U-trough was a polyvinyl chloride (PVC) flume with 10 cm width, 5 cm depth, and its length was consistent with the length of the experimental plot. In each plot, the vertical projection spacing of all U-troughs was 40 cm. The vertical projection spacing of all the seepage plates was 70 cm in the transversal slope direction and was 4.5 m in the longitudinal slope direction. All collection devices of the UCM and SPM methods were kept parallel to the slope surface. Both kinds of collection devices were filled with 2-4 mm diameter clean quartz sands and covered by two layers of 200-mesh gauzes to improve the hydraulic conductivity. Moreover, all devices of the UCM were connected to the buckets outside of the downslope by discharge tubes, while all SPM devices were connected to buckets by PE pipes. Total SSR in each test plot, except this part collected by devices, was collected by a subdrain trench stuffed with the quartz sand (2-4 mm) and was drained into a bucket by a water discharge tube. Total SR in each test plot was collected by an open trench and then discharged into a bucket through a discharge tube. Each bucket had nine shunt holes, and one of the shunt holes connected to a sub-bucket through a water pipeline. This research had been conducted with three duplicates of each collection system. Briefly, the corn-mustard rotation, two times of fertilizer applications for summer maize and one-time fertilizer application for winter mustard, was set up in each year in this study. The fertilization was applied at a depth of 5-10 cm to avoid surface loss. This study had been conducted over two consecutive years' field monitoring.
pan soil (L2), and 20 cm depth of clastic purple shale parent materials (L3), respectively, as shown in Figure 1b,c. The experimental soil, developed from the Jurassic Shaximiao Formation, is purple soil, classified as Regosols in Food and Agriculture Organization of the United Nations (FAO) Taxonomy or Entisols in United States Department of Agriculture (USDA) Taxonomy. It is a unique soil type in China. Purple soil has the characteristics of a dualistic geotechnical structure of overlying soil and underlying rock [55]. Plentiful soil pores, strong permeability, and weak water holding capacity of purple soil may cause severe soil erosion [56] under the climate environment of concentrated rainfall and heat [11], especially in slope-cultivated land area [57]. The basic physicochemical properties of the top layer soil were pH (6.29), total nitrogen (1.14 g/kg), total phosphorus (0.72 g/kg), total potassium (21.65 g/kg), soil organic matter (14.96 g/kg), soil bulk density (1.22 g/cm 3 ), and soil porosity (53.96%). The surface slope of all grids is 15° [58][59][60].

Experimental Design
This research aims to verify the reliability of UCM on soil water, sediment and nutrient migration load at different soil layers (L1: 0-20 cm, L2: 20-40 cm, L3: 40-60 cm), compared to the SPM [53]. The UCM was designed to collect the SSR of a consecutive beltregion from the top slope to the bottom slope, while the SPM was used to collect the SSR of a dot area at the top slope and the bottom slope. As shown in Figure 1b,c, two U-troughs or two seepage plates were installed and fixed at each soil layer's bottom. Each U-trough was a polyvinyl chloride (PVC) flume with 10 cm width, 5 cm depth, and its length was consistent with the length of the experimental plot. In each plot, the vertical projection spacing of all U-troughs was 40 cm. The vertical projection spacing of all the seepage

Experiment Data Collections and Processing
The volume of each fluid sample was the volume of water in the corresponding bucket of each device. When the collected SSR by the device was more than 1 L, we sampled 1 L and divided it into two 0.5 L for the nutrient and sediment concentration analysis; When the collected SSR by the device was 0.5-1 L, we only sampled 0.5 L for the nutrient concentration analysis; When the collected SSR by the device was less than 0.5 L, we did Sustainability 2021, 13, 2050 5 of 15 not sample. When the collected SSR by the device was more than 1 L, we sampled 1 L and divided it into two 0.5 L for the nutrient and sediment concentration analysis; when the collected SSR by the device was 0.5-1 L, we only sampled 0.5 L for the nutrient concentration analysis; when the collected SSR by the device was less than 0.5 L, we did not sample. Each soil layer's SSR volume per rainfall event and the sampling probability during the test period were calculated as follows: where R is the SSR volume (L) per rainfall event flowing out the specified layer; r is the collected volume (L) by the device; A is the test plot area (m 2 ); a is the collecting area (m 2 ) of the device; SP is the sampling probability (%), i is the times of rainfall-runoff events that met the minimum sampling requirements (collected SSR volume was more than 0.5L); n is the total rainfall-runoff events during the test period.
In the i th rainfall event, the flux of SSR permeating from each soil layer and its total flux in a specified period could be obtained by these equations as follows: where R ni is the SSR flux (mm) of the n th soil layer in the i th rainfall event; r ni is the average collected volume (L) of SSR through both devices at the same layer; R n is the total SSR flux (mm) of the n th soil layer; k is the number of rainfall events in a specified period; n is the number of soil layers. The amount of sediment was from each SSR sample by the filtration and drying method [61]. The amount of collected sediment that met the minimum sampling requirement was 1.0 g for sediment nutrient analysis and 10.0 g for the particle size distribution analysis under a single rainfall-runoff event. Otherwise, the total mixed sediment amount of several consecutive rainfall events would be obtained. The following equations could calculate the sediment migration intensities of each soil layer.
where S ni is the sediment migration intensity (t/km 2 ) of the n th soil layer in the i th rainfall event; s ni is the dried weight (g) of sediment in each SSR sample of the n th soil layer; S n is the total sediment migration intensity (t/km 2 ) of the n th soil layer in a specified period; other parameters have the same meanings as above. The runoff sample from the inside of the bucket was thoroughly stirred before sampling. The soil nutrient migration intensities were obtained from the following equations. where Q nij is the migration intensity (t/km 2 ) of the j th nutrient at the n th soil layer in the i th rainfall event; C nij is the concentration (mg/L) of the j th nutrient at the n th soil layer; Q nj is the total migration intensity (t/km 2 ) of the test grid's j th nutrient at the n th soil layer in a specified period; other parameters have the same meanings as above.
Hourly meteorological data were from a small-sized automatic collection weather station near the research plots. The average annual precipitation was 1009.2 mm in this study. Notably, during the whole experiment period, 312 rainfall events occurred, but only 51 rainfall-runoff events had been observed and collected. During the first cycle year of the test period, at the L 1 , L 2 , and L 3 layers, 20, 18, 18 valid rainfall-runoff events in UCM and 16, 11 and 11 valid events in SPM were observed and collected, respectively. While during the second cycle year, at the L 1 , L 2 , and L 3 layers, 31, 31, 31 valid rainfall-runoff events in UCM and 27, 23, 22 valid events in SPM were observed and collected, respectively. This study only reported results from the concentration and migration intensities of total nitrogen (TN) and total phosphorus (TP). The TN and TP concentrations were determined by the Kjeldahl method and the ammonium-molybdate spectrophotometry method, respectively [62]. The sediment samples were dried in an oven at 105 • C to a constant weight. The sediment particle size distribution (PSD) was determined by the conventional pipette method [61]. The basic physicochemical properties of soil were determined as Yang's methods [63].

Statistical Analysis
The IBM SPSS Statistics 25 software was the data analysis tool, while the Graph Pad Prism 8 was the mapping software in this paper. The Pearson Correlation was used for the correlation analysis. The Least Significant Difference (LSD) method at p = 0.05 was performed to elucidate any significant differences. The average data were the results of normalizing to the corresponding numbers of their valid events. Figure 2 presented that the UCM presented significant differences in the collected SSR compared to the SPM. As shown in Figure 2a, in the whole test period, the average collected SSR volumes per rainfall event from the L 1 , L 2 , and L 3 layers under the UCM were 9.69, 9.74, and 9.97 times of those under the SPM, respectively. Additionally, both methods showed that the deeper the soil layer, the lower volumes were collected from the SSR. Noteworthily, the collected SSR volumes under both methods at the L 1 layer were significantly higher than those at the other two layers. Figure 2b shows that the SSR flux from each soil layer presented a significantly positively linear correlation with natural rainfall under both methods. However, the flux from the SPM was more seriously affected by the rainfall. With the increase of rainfall, the SSR leaching from each soil layer increased obviously. The topsoil layer was substantially more affected by the rainfall than the two lower layers. Table 1 presents that the UCM had significantly higher sampling probabilities at each soil layer than the SPM. The annual sampling probability of collected SSR volume met the minimum sampling requirement reached up to 95.00-100% under the UCM, compared to 63.4-83.00% under the SPM. The annual SSR flux from L 1 to L 3 was 403.4 mm, 271.9 mm, and 237.4 mm under the UCM, 14.35%, 10.56%, and 8.41% lower than those under the SPM, respectively. The difference between both methods in the annual SSR flux was significant between the L 1 and L 2 layers. However, it was not significant at the eventual loss flux from the L 3 layer. The difference between the infiltrated flux and each soil layer's outflow flux was the intercepted amount at this layer. The upper soil layer's outflow flux was the infiltrated flux of the next soil layer. The total permeating water flux through the soil surface (L 0 ) was the difference between the precipitation and SR flux; 57.31%, 16.03%, and 1.53% of annual total permeating water flux (944.9 mm) was retained by the L 1 , L 2 , L 3 layers under the UCM, respectively. In comparison, the proportions were 50.81%, 17.44%, and 4.68% under the SPM. Briefly, the sharp decrease of SSR flux from L 0 to L 1 indicated   10.56%, and 8.41% lower than those under the SPM, respectively. The difference between both methods in the annual SSR flux was significant between the L1 and L2 layers. However, it was not significant at the eventual loss flux from the L3 layer. The difference between the infiltrated flux and each soil layer's outflow flux was the intercepted amount at this layer. The upper soil layer's outflow flux was the infiltrated flux of the next soil layer. The total permeating water flux through the soil surface (L0) was the difference between the precipitation and SR flux; 57.31%, 16.03%, and 1.53% of annual total permeating water flux (944.9 mm) was retained by the L1, L2, L3 layers under the UCM, respectively. In comparison, the proportions were 50.81%, 17.44%, and 4.68% under the SPM. Briefly, the sharp decrease of SSR flux from L0 to L1 indicated that most of the infiltration flux was intercepted by the L1 layer. The deeper the soil layer, the lower the interception capability of this soil layer.   Note: The capital letters indicate significant differences between different methods, and the lowercase letters indicate significant differences between different soil layers at p = 0.05; L 0 is the annual average rainwater flux permeated into the soil, namely the difference between the annual average rainfall and SR; L 1 , L 2 , and L 3 represent the annual migration loads out of the corresponding soil layer, respectively. SD represents the standard deviation. SP presents the sampling probability.

Sediment Migration Analysis
Figure 3 existed that the difference in sediment migrations under both methods was significant. In Figure 3a, during the test period, the average collected sediment amounts per rainfall event at the L 1 , L 2 , and L 3 layers under UCM were 6.22, 3.61, 4.97 times those under SPM. The deeper the soil layer, the lower the sediment amount collected. The sediment collected by the SPM was basically less than the minimum sampling requirement of 1.0 g under a single rainfall-runoff event. Figure 3b indicates that each soil layer's migrated sediment load presented a significantly positively linear correlation with rainfall under both methods. However, when the rainfall was higher than 20 mm, the migrated sediment loads from the SPM at each soil layer were higher and more substantially affected by rainfall. With increased rainfall, the sediment migration amounts at each soil layer increased obviously. The topsoil layer was more obviously affected by rainfall than the two lower layers. Figure 3c indicates that the sediment particle size distribution (PSD) was significantly different in each soil layer. The fine particle with a size less than 0.05 mm dominated the sediment at each soil layer. The cumulative weight percentage of sediment particles below 0.002 mm at L 1 was 55.09%, while 86.40% at L 2 , and 76.48% at L 3 . Remarkably, the particles' size at L 1 was more extensive than that at other layers, and the L 2 layer had the smallest sediment particle.    The difference of sediment migrating load between the two adjacent layers was the net intercepted amount. As shown in Table 1, the deeper the soil layer, the lower the net intercepted amount. The annual sediment migrating out of the L 1 layer was 49.562 t/km 2 under UCM, while it was 48.708 t/km 2 under SPM. The annual sediment migrating out of the L 2 layer under the UCM was 12.475 t/km 2 , which meant that the L 2 layer intercepted 74.83% of the L 1 layer's net leached loss. The annual sediment migrating out of the L 3 layer was 6.63 t/km 2 under UCM. The L 3 layer intercepted 5.845 t/km 2 of sediment, accounted for 11.79%, 46.85% of the L 1 and L 2 layers' net leached loss, respectively. Though the sediment migrating loads under a single rainfall event were significantly different between the UCM and SPM (shown in Figure 3b), the annual net sediment migrating loads of both methods at each layer had no significant differences. Briefly, 86.62% and 88.24% of the topsoil layer's leached sediment under the UCM and SPM were withheld by the subsoil layers (L 2 and L 3 ), respectively. According to Pearson correlation analysis (shown in Table 2), the different soil layers' annual sediment migration loads under UCM showed no significant relationship with the annual SSR flux. In contrast, under SPM, they showed a significant relationship.  Cases  3  3  3  3  Cases  3  3  3  3 ** The correlation is significant at p = 0.01(1-tailed); * The correlation is significant at p = 0.05(1-tailed); SSR is the subsurface runoff; S is the sediment; TN is the total nitrogen; TP is the total phosphorus; Pearson CC. is the Pearson correlation coefficient.  Figure 4b indicated that the TN migration amount at each soil layer presented a significantly positively linear correlation with rainfall under the UCM and SPM. However, the TN concentrations from the SPM under the same rainfall event at each soil layer were higher than those from the UCM. In other words, the SPM was more substantially affected by rainfall than the UCM. Figure  4c shows no significant differences in TP concentrations between the L 2 and L 3 layers, except in the topsoil layer (L 1 ). The average TP concentration under the UCM was 0.71 mg/L at L 1 , 0.30 mg/L at L 2 , and 0.22 mg/L at L 3 , while 0.59 mg/L, 0.29 mg/L, and 0.22 mg/L under the SPM, respectively. Figure 4d indicated that the TP migration amount at each soil layer presented a significantly positively linear correlation with natural rainfall under both ways. However, the TP concentrations from the SPM under the same rainfall event at each soil layer were higher, but not significantly, than those from the UCM.

Nutrient Migration Analysis
in the topsoil layer (L1). The average TP concentration under the UCM was 0.71 mg/L at L1, 0.30 mg/L at L2, and 0.22 mg/L at L3, while 0.59 mg/L, 0.29 mg/L, and 0.22 mg/L under the SPM, respectively. Figure 4d indicated that the TP migration amount at each soil layer presented a significantly positively linear correlation with natural rainfall under both ways. However, the TP concentrations from the SPM under the same rainfall event at each soil layer were higher, but not significantly, than those from the UCM.  As shown in Table 1, the annual leached TN out of the L 1 layer was 19.113 t/km 2 under UCM, while it was 22.342 t/km 2 under SPM. Under UCM, the annual migrated sediment out of the L 2 layer was 13.795 t/km 2 , namely that the L 2 layer intercepted 27.83% of the L 1 layer's leached TN. The L 3 layer intercepted 2.558 t/km 2 of TN, and accounted for 13.39%, 18.15% of the L 1 and L 2 layers' leached TN, respectively. The annual migrated TN loads of both collection methods at each soil layer had no significant difference. Briefly, 41.21% and 53.06% of the topsoil layer's leached TN under UCM and SPM were withheld by the subsoil layers (L 2 and L 3 ), respectively. Note that 58.79%, 46.94% of the L 1 layer's annual leached TN was finally lost from the bottom of the test plot through the SSR. The annual net leached TP out of the L 1 layer under the UCM was 0.291 t/km 2 , while it was 0.286 t/km 2 under the SPM. The annual migrated TP out of the L 2 layer under the UCM was 0.081 t/km 2 , which meant a net of 0.210 t/km 2 (72.31% of the L 1 layer's annual leached TP) was intercepted by the L 2 layer. The L 3 layer intercepted 0.028 t/km 2 of TP, and accounted for 9.61%, 34.71% of the L 1 and L 2 layers' net leached TP loss, respectively. The annual migrated TP loads of both collection systems at each layer presented no significant difference. Also, 81.78% and 80.42% of the topsoil layer's leached TP under UCM and SPM were withheld by the subsoil layers (L 2 and L 3 ), respectively. Only 18.21%, 19.58% of the L 1 layer's annual net leached TP finally lost from the bottom of the test plot through the SSR.
According to Pearson correlation analysis (shown in Table 2), under UCM, the different soil layer's annual migrated TN loads had no significant relationship with the annual SSR flux but showed significant positive relationships with the annual sediment loads; the different soil layer's annual migrated TP loads had no significant relationships with the annual sediment loads but showed extremely significant positive relationships with the annual SSR flux. While under SPM, the annual migrated TN loads presented significant relationships with annual SSR flux and extremely significant relationships with the annual sediment loads; the annual migrated TP loads presented significant relationships with annual SSR flux but showed no significant relationship with the annual sediment loads. Additionally, the annual migrated TN loads had no significant relationship with the annual migrated TP loads under UCM but showed significant positive relationships under SPM.

Discussion
The difference between the two methods was due to the difference in their collection devices. It was easy to find that the average collected SSR volume and collected sediment amount from the collected devices, SSR and sediment migrating loads per rainfall event were significantly different under UCM and SPM. These differences were rooted in the difference of various devices' efficient collection areas. In this study, the collecting area of the U-trough was ten times that of the seepage plate. It can be proved in Figure 2a that the average collected SSR volume per rainfall event at each soil layer under UCM was nearly 10 times that under SPM. Undoubtedly, these differences eventually led to the difference in sampling probabilities of both methods. It can be testified in Table 1 that 95.00% of the annual rainfall-runoff events were observed and collected at all soil layers under the UCM, while only 62.99% under the SPM. That is because the SSR flux during some light rainfall events was hard to observe and collect under the SPM. Therefore, the UCM could collect more leakage water from the upper soil layer and be more sensitive to rainfall, especially to the minor rainfall events, than the SPM. The runoff flux, sediment, and nutrient per rainfall event under the SPM were higher than those under the UCM. There might be two reasons as following: firstly, the PE pipe connected to the seepage plate in the soil was empty, which can improve the soil water transmitting ability indirectly; secondly, the lower outlet of the PE pipe was open, then the water in the pipe flowed downward under the gravity might produce negative pressure on the seepage plate and cause a siphonic effect. Therefore, it can accelerate the SSR to flow through the SPM and improve the capability to transport the sediment and nutrient, which can be deduced in Figures 2b, 3b and 4b.
Notably, though the differences in SSR flux per rainfall at each soil layer between the UCM and SPM were significant, the differences in the annual sediment, TN and TP migration loads between the UCM and SPM were not significant. That is because more valid rainfall events were observed and collected under the UCM than those under the SPM, as mentioned above. Thus, the more the total SSR flux could be obtained under UCM to make up the difference. It can be proved in Table 1 that the sampling probabilities under UCM were higher than those under SPM, but the differences in annual sediment, TN and TP migration loads between both methods were not significant. Additionally, compared with the SPM's point-like layout, the UCM's collecting area was a continuous area from the top to the slope's bottom. It might effectively reduce soil heterogeneity of slope, but it still needs more further studies to be proved.
The difference in sediment, SSR, TN and TP migration loads between different soil layers was due to the difference in soil structure characteristics of different soil layers. The topsoil layer was abundant in soil pores [64] and always presented lower moisture than the subsoil layer under the comprehensive affection of gravity and evapotranspiration during the no-rain period. [6,65]. Thus, in the rainy time, the topsoil layer showed high water absorption and storage capacity and absorbed rainfall-runoff firstly to make up soil moisture loss. Then, the SSR in the topsoil layer diffused downward gradually. There might not be enough SSR flux to spread down to the lower layers after the topsoil layer's interception, especially during some minor rainfall events. The soil pores in the soil can provide preferential flow pathways [66], which might accelerate the nutrient transmitting, especially for phosphorus migrating. That is why the annual leached TN and TP loads from the topsoil layer were quite high, as shown in Table 1. Subsoil with strong adsorption capacity could be considered an appropriate filter to leached nutrients, especially leached phosphorus [66]. The soil subsidence under gravity, leaching and less disturbance made the plow pan soil in a compact structure with fewer soil pores [61,67], which might stop sediment particles of large size from being transmitted downward. It can be proved in Figure 3c. The plow pan soil could intercept most sediment and phosphorus leaching loss of the topsoil [68], which can be proved in Table 1 and Figure 4c. Only a small portion of leached phosphorus was eventually lost, which can be tested in Figure 4c. The detrital shale parent material layer with rich cracks, strong permeability, and weak water-holding capacity [56] would contribute to the sediment, nutrient transmitting. Hence, this layer presented no significant interception ability of leaching sediment and nutrient, which can be seen in Table 1. Additionally, according to Pearson correlation analysis (shown in Table 2), the different soil layers' annual sediment migration loads showed no significant relationship with the annual SSR flux under UCM but showed significant relationships under SPM. Under UCM, the different soil layers' annual migrated TN loads showed significant positive relationships with the annual sediment loads; the different soil layer's annual migrated TP loads showed extremely significant positive relationships with the annual SSR flux. While under SPM, the annual migrated TN loads presented significant relationships with annual SSR flux and extremely significant relationships with the annual sediment loads; the annual migrated TP loads presented significant relationships with annual SSR flux. Additionally, the annual migrated TN loads had no significant relationship with the annual migrated TP loads under UCM but showed significant positive relationships under SPM.

Conclusions
The average leached TN, TP concentrations per rainfall event and their annual migrated loads at each soil layer between the SPM and UCM showed no significant difference. Under both methods, the different soil layer's annual TP migration loads showed significant relationships with the annual SSR. In contrast, the different soil layers' annual TP migration loads showed significant positive relationships with the annual sediment loads. Both methods also indicated that the topsoil layer was the main layer to retain rainfall-runoff and showed the source of leached sediment and soil nutrient. The subsoil layers, especially the plow pan soil layer, presented a significant interception role of leached sediment and soil nutrients. However, the collecting area of devices significantly affected each soil layer's sediment, TN and TP migration loads per rainfall event. The U-trough with a larger collecting area could collect higher SSR volume than the seepage plate and keep a higher sampling probability at each soil layer. In conclusion, the UCM was proved to be a reliable quantitative method for subsurface runoff, sediment, and soil nutrient migration loads from diverse soil layers of purple soil sloping cultivated lands. Additionally, we just did the comparison test of both methods in the purple soil sloping cropland in this study, so it needs further studies to testify their availability in other land-use.

Patents
The UCM has been applied for an invention patent (application No.: 202010204850.9) in China on 22 March 2020.

Abbreviations
The following abbreviations are used in this manuscript: UCM U-trough collection method SPM Seepage plate collection method SR Surface runoff SSR Subsurface runoff SP Sampling probability TP Total phosphorus TN Total nitrogen