Modelling soil erosion risk in rural sub-catchments of Zimbabwe using RUSLE, remote sensing and machine learning

Abstract

The study modelled soil erosion risk in the Shashe and Tugwi–Zibagwe rural sub-catchments in Zimbabwe. To derive land use and land cover (LULC) thematic maps for the years 2016, 2020 and 2023, analysis ready data (Sentinel 2) were applied using the Random Forest (RF) algorithm in the Google Earth Engine (GEE) platform. The Revised Universal Soil Loss Equation (RUSLE) model was applied to understand the drivers of soil loss in the sub-catchments. The rainfall erosivity (R), soil erodibility (K), length slope (LS), crop management (C) and conservation support practice factors (P) were derived in GEE and applied as input to determine soil erosion risk. The findings of the study show that, the Shashe sub-catchment had mean soil losses of 15.75, 45.25, and 23.51 t ha−1 year−1 for 2016, 2020, and 2023, respectively. In the Tugwi-Zibagwe sub-catchment, the mean soil losses were 11.62, 18.45, and 37.34 t ha−1 year−1 for the same years. The results also show that LULC changes were one of the major drivers to soil loss in the rural dominated sub-catchments. Results further show that, the area under cultivation was exposed to severe erosion which averaged 16–48 t ha−1 year−1 when compared to other land covers in the study areas. In conclusion, of all the two sub-catchments the Shashe experiences severe soil loss than Tugwi-Zibagwe due to variations in land use and covers. Soil loss also tends to be considerably high in areas along drainage networks and where vegetation clearance is evident. These findings highlight the pressing need for up-to-date soil management approaches to improve soil conservation in rural dominated sub-catchments of Zimbabwe.