From Static Erosion Maps to Transferable Risk Intelligence: A Process - Informed Machine Learning Framework for Morogoro, Tanzania

This study uses machine learning and the Revised Universal Soil Loss Equation, $A = R \times K \times LS \times C \times P$, to map soil erosion risk in Morogoro, Tanzania. Random Forest and XGBoost models use rainfall, soil, terrain and vegetation data to predict soil loss and classify erosion severity. The resulting maps can support targeted land conservation and restoration decisions.