Assessing land use/cover changes to soil erosion vulnerability using machine learning and RUSLE model

Abstract

Soil erosion is among the most pressing environmental problems worldwide, driven by agricultural intensification, climate change, and other uncontrolled human activities. Understanding how soil erosion threatens food productivity is important for preventing, reducing, and restoring land degradation, thereby contributing to attaining key sustainable development goals. This study employed the Revised Universal Soil Loss Equation (RUSLE) model and Support Vector Machine (SVM) algorithm, coupled with Geographical Information System (GIS), to estimate soil loss in the Federal University of Agriculture Abeokuta, Nigeria. Land use cover maps for 2002, 2012, and 2022 were generated through SVM. RUSLE parameters model was derived from remote sensing data, and erosion vulnerability zones were identified using GIS analysis. The results showed significant land use changes, including a substantial increase in built-up areas (42.7 %), and barelands (1.6 %), alongside a decrease in farmlands (30.9 %), and vegetation (23.7 %), respectively. Annual soil loss estimates of the studied area ranged in the order of 11 > 13 > 17 (t/ha/yr) with vulnerability soil loss classed as low (731 ha), moderate (421 ha), and high (491 ha). The findings from this study recommend adopting SVM and RUSLE for erosion assessment to inform sustainable land use and soil management practices. Future studies could integrate real-time erosion monitoring for improved conservation practices.