GeoAI-based soil erosion risk assessment in the Brahmaputra River Basin: a synergistic approach using RUSLE and advanced machine learning.

Soil erosion is a critical environmental issue in the Brahmaputra River Basin, threatening agricultural productivity, water resources, and ecological balance. This study employs the revised universal soil loss equation (RUSLE) alongside remote sensing, geographic information systems (GIS), and advanced machine learning models like random forest (RF) and gradient boosting (GB) to analyze soil erosion patterns from 2005 to 2024. The analysis revealed that average annual soil loss increased from 15.8 tons/ha/year in 2005 to 25.4 tons/ha/year in 2024, marking a 60.76% rise over two decades. Peak erosion rates were observed in 2020, with localized hotspots recording up to 32,130 tons/ha/year. Spatial analysis from 2005 to 2024 indicated substantial variability, with soil loss values ranging from - 7.024 to 9034 tons/ha in 2005. Topographic influence, quantified using the LS factor, revealed that 47.2% of the basin area has slopes steeper than 16°, significantly contributing to elevated erosion risk. The rainfall erosivity (R-factor) fluctuated throughout the period, peaking at 2305.73 MJ mm/ha h year in 2015 but declining to 799.21 MJ mm/ha h year by 2024, indicating a temporal shift in rainfall patterns. Vegetation cover improvements during this time reduced the mean C-factor from 0.52 to 0.34, though 13.8% of the basin (approximately 3.05 million ha) still falls under high to very high erosion risk zones. RF model predictions achieved an R² of 0.915 and RMSE of 4.82, while GB attained an R² of 0.952 with RMSE of 3.97, indicating superior predictive performance. These findings underscore the urgent need for targeted soil conservation measures, afforestation programs, and sustainable watershed management. The integration of AI-driven modeling with remote sensing and GIS provides a robust framework for long-term soil erosion monitoring, enabling informed decision-making for climate adaptation in the region.