Climate-induced soil erosion sensitivity in a humid tropical watershed under multi-model CMIP6 projections

Soil erosion in humid tropical watersheds is strongly influenced by rainfall variability, while uncertainty in future climate projections complicates the assessment of climate-related erosion risk. This study evaluates climate-driven changes in rainfall-derived soil erosion sensitivity in the Hulu Selangor Basin, Malaysia, using a GIS-based Revised Universal Soil Loss Equation (RUSLE) framework coupled with three CMIP6 climate models under SSP2-4.5 and SSP5-8.5. Historical rainfall observations from 2001 to 2020 were used to establish baseline conditions, while future precipitation projections covered 2021–2060. The historical basin exhibited predominantly very low erosion sensitivity, with 89.5% of the area classified as very low, 7.8% as low, 1.2% as moderate, 0.6% as high and 0.9% as very high, with a mean annual soil loss of 15.89 ton/ha/yr. The historical mean rainfall-derived R factor was 17,183.47 MJ·mm/h/ha/yr. Validation results showed that BCC-CSM2-MR provided the lowest rainfall bias and error among the three models. Future projections exhibited substantial inter-model uncertainty: the ensemble mean R increased by 6.0% under SSP2-4.5 during 2021–2040 and by 5.9% during 2041–2060, while SSP5-8.5 produced changes of 0.1% and 6.7%, respectively, relative to the historical mean. MIROC-ES2L generally produced the upper end of the projected erosivity range. These results demonstrate that climate-model uncertainty is an important consideration in assessing future rainfall-driven erosion sensitivity in humid tropical watersheds.