How root and soil properties affect soil detachment capacity in different grass–shrub plots: A flume experiment

Grasses and shrubs play important roles in preventing soil erosion. However, the mechanism of their impact on soil detachment is still unclear. Therefore, this study examined the following combinations of typical grass–shrub sample plots on the Loess Plateau: Caragana korshinskii (JJ), C. korshinskii and Agropyron cristatum (JC), C. korshinskii and Bothriochloa ischaemum (JB), and C. korshinskii and Artemisia gmelinii (JH). The variation characteristics of the soil detachment rate with depth (0–30 cm) and at different locations (0 cm, 50 cm) from the base of shrubs under a fixed flow rate and slope were explored. The spatial heterogeneity of the detachment rate in the grass–shrub plots was strong. In different plots, the soil detachment rate at 50 cm from the base of the shrub was relatively low. The ability of root length to reduce the detachment rate varied. In the JJ sample plot, the root length at the shrub base had a relatively strong ability to reduce the detachment rate, while it showed the opposite trend in the other plots. The ability to reduce the detachment rate was higher in the JC and JB plots at 0–5 cm than in the JH plots. However, at 20–30 cm, the opposite trend was observed. Overall, with the exception of the JC plots, the soil detachment rate in the other mixed plots was lower than that in single-species plots. In this study, soil aggregates and root length density were the main factors affecting the soil detachment rate. The combination of the two could also better calculate the soil detachment rate (R2 = 0.80). Compared with traditional regression methods, the back propagation neural network model had a relatively high R2 (0.86) when predicting the soil detachment rate. This study revealed the mechanisms underlying differences in soil detachment capacity in different grass–shrub sample plots.