2. Materials and methods

2.1. Experiment materials

The experiment was carried out in the artificial rainfall hall of the State Key Laboratory of soil erosion and dryland agriculture, Institute of soil and water conservation, Ministry of water resources, Chinese Academy of Sciences. A movable and adjustable steel soil trough (5 m long × 1 m wide × 0.5 m high) was used. Experimental rainfall mode consisted of a downspray rainfall system. To ensure that all raindrops reached a final speed, rainfall height was set at 18 m (He et al., 2017).
According to the zonal division of soil in the Loess Plateau (Li et al., 1985), a sandy loam (SL; Suide), a light loam (LL; Ansai), a medium loam (ML; Changwu) and a heavy loam (HL; Yangling) were sampled from different soil zones from the north to the south across the plateau (Fig. 1). All samples were collected from the soil surface of local cultivated land. For the four soil types, soil bulk density was 1.34 g/cm3 (SL), 1.26 g/cm3 (LL), 1. 21 g/cm3 (ML) and 1. 13 g/cm3 (HL). Organic matter content was determined using the potassium dichromate heating method (Nelson and Sommer, 1982), mechanical composition was determined using the laser particle size analysis method (Vaezi et al., 2017), and a wet sieving method (Castro et al., 2002) was used to determine water stable aggregate content. For each analysis three replicas were conducted. Soil properties are shown in Table 1.

2.2. Experimental design and process

Before analysis, soil samples were air dried, passed through a 10 mm sieve, and had weeds and stones removed. In order to keep permeability of test soil as close as possible to that on the natural slope, the bottom of the test steel tank was filled with a 10 cm thick layer of fine sand upon which a permeable fine gauze was laid. The remainder of the steel tank was then filled with test soil using a layered filling method. During filling the soil was compacted at the same time. Each filling layer was 5 cm thick, and the total filling thickness in the tank was 30 cm. Bulk density of soil in the experimental tank was equivalent to that of the natural soil. In order to reduce the influence of the sidewall effect as much as possible, clapboards were compacted as much as possible when soil was loaded. After loading, the soil surface was leveled using a ruler and placed in the rainfall area for analysis. Four slope grades (10°, 15°, 20° and 25°) were used in the experiment and each slope was tested under two rainfall intensities (1.5 and 2.0 mm/min). In order to ensure that final rainfall was equal, corresponding rainfall durations were set at 60 min and 45 min, respectively.
During rainfall simulation, sediment and runoff samples were collected at the slope outlet, having a sampling interval of 1 min. Sediment volume was determined using the drying method. Slope runoff velocity was determined using the color tracer method (KMnO4solution). Velocity measurement included slope velocity, rill velocity after rill formation and rill velocity between rills. Starting from the slope section at the bottom of the slope, flow velocity at 4, 3, 2 and 1 m from the top of the slope was continuously measured. Before and after the experiment, DEM micro terrain surface data was collected using a 3D laser scanner (scanning 2), having an accuracy of 1 mm. At the same time, a high-definition camera was used to record slope morphology changes during the rainfall process.

2.3. Data analysis

DEM data form the soil trough was denoised using cyclone software. DEM for the whole slope was obtained after invalid points were removed. At the same time, ArcGIS software was used to extract rill length, width, depth and density data; other characteristic parameters of rill erosion were also calculated.