Spatially adaptive modeling of soil erosion susceptibility using geographically weighted regression integrated with remote sensing and GIS techniques

Abstract

Soil erosion poses a major threat to sustainable land use in Ethiopia's highland regions, where steep terrain and expanding agriculture accelerate environmental degradation. This study assesses erosion susceptibility in the Gubalafto district using a spatially adaptive modeling approach-Geographically Weighted Regression (GWR)-integrated with remote sensing and GIS techniques. High-resolution terrain data (12.5 × 12.5 m) were used to derive key topographic and hydrological indicators, including slope, LS factor, curvature, valley depth, channel network base level, channel network distance, and wetness index. The GWR model revealed significant spatial variation, with erosion hotspots concentrated in steep, sparsely vegetated areas near drainage networks. Validation metrics confirmed the model's reliability and spatial sensitivity. The resulting erosion susceptibility map offers a decision-support tool for targeted watershed interventions. It identifies priority zones for terracing, reforestation, and land-use regulation, enabling more efficient allocation of conservation resources. The key finding is that slope, LS factor, and proximity to drainage channels are the dominant predictors of erosion, providing a spatially precise framework for sustainable land management in vulnerable highland ecosystems.