Spatio-temporal variations of black soil erosion under the scenario of soil organic carbon change based on RUSLE and random forest

Introduction The Northeast black soil area is an important marketable grain base in China. However, due to soil erosion, the black soil layer has been gradually thinning and its quality deteriorating. Therefore, accurately assessing the extent of soil erosion in this region is essential for the protection and sustainable utilization of black soil resources. Methods In this study, linear and nonlinear models were compared combined with remote sensing images to invert soil organic carbon (SOC). In the scenario of SOC change, temporally variable soil erodibility factor were obtained. Then based on the RUSLE model and GIS technology, land use, rainfall, soil texture and digital elevation model (DEM) were used to evaluate the temporal and spatial variation characteristics of soil erosion in black soil region from 1995 to 2020 in Hulan River Basin. The main influencing factors were explored by random forest model and analyzes in combination with eco-geological. Results The findings are as follows: The random forest (RF) model was optimal for SOC inversion (2020: R 2 = 0.64, RMSE = 0.70, 2010: R 2 = 0.66, RMSE = 0.35). The erosion intensity was mainly slight or mild while mean annual soil loss firstly decreased then increased from 1995 to 2020 reaching a rate of 1020.16 t km −2 y −1 by 2020. Rainfall and topography were the main driving factors of soil conservation changes, and soil erosion was more likely to occur in the eco-geological environment of the neutral rock hilly woodland area. Discussion The results provide insights into spatial distribution characteristics of black soils erosion which are crucial for preventing further degradation and ensuring national food security.