Effects of Vegetation Restoration on Rill Erosion on the Chinese Loess Plateau

Abstract

Accurately describing rill erosion is important for building physical process models of soil erosion. Vegetation restoration may lead to differences in soil properties and root characteristics, and thus likely affects the rill erosion process to flowing water erosion, reflected by rill erodibility (Kr) and critical shear stress (τc). However, few studies have been conducted to evaluate this effect of vegetation restoration on the Loess Plateau of China. The present work chose eight typical herbaceous plants, commonly appearing at diverse vegetation succession periods on the Chinese Loess Plateau, for discovering how soil properties and root characteristics affect rill erosion. In total, we collected 240 undisturbed soil samples for overland flow scouring within the hydraulic flume in the presence of six shear stresses (range, 5.94–18.58 Pa). According to our findings, Kr and τc were 0.015–0.411 s m−1 and 0.050–6.059 Pa among eight typical grasslands. Cultivated grasslands have high Kr and low τc; the Kr was 5–25 times greater than that of other grasslands, and τc was 89%–98% less than that of other grasslands. Plants with tap root systems exhibited a 5‐fold increase in Kr value and a 50% reduction in τc compared to those with fibrous root systems. The Kr and τc were affected by the root‐soil complex, with interaction effects of root and soil being 54% and 50%, respectively. The Kr decreased with increasing bulk density (BD), soil cohesion, soil organic matter, and soil aggregate as power functions and decreased with increasing root surface area density (RSAD) and root length density as exponential functions. The τc increased with specific root length (SRL) and BD as power functions. Those above‐mentioned soil properties and root characteristics indicated that Kr was dominantly affected by cohesion and RSAD, and τc was affected by BD and SRL. Kr was simulated through soil cohesion (Coh) and RSAD, and τc was simulated by soil BD and SRL as power functions. Our constructed model achieved satisfactory performance.