Dataset for Application of explainable artificial intelligence to decode water-induced soil erosion in Lidder watershed of the Greater Himalayas

The data pertains to the research paper "Application of explainable artificial intelligence to decode water-induced soil erosion in Lidder watershed of the Greater Himalayas" which undertakes the behavioural assessment of soil erosion by water using the advanced machine and deep learning techniques and explainable AI techniques. Soil erosion susceptibility in the Lidder watershed of the Greater Himalayas is evaluated by integrating the Revised Universal Soil Loss Equation (RUSLE) with six machine and deep learning models using 24 environmental conditioning factors.

Average annual soil loss is estimated at 58.81 t ha⁻¹ yr⁻¹, with more than 60% of the watershed exhibiting high to extreme erosion risk. Random Forest provides the highest predictive accuracy among the tested models. SHAP-based explainable AI identifies slope, elevation, surface temperature, stream density, and bare ground as the primary drivers of erosion, while vegetation cover, built-up areas, and agricultural terracing mitigate erosion susceptibility in mountainous environments.