Explainable Machine Learning Framework for Flood Risk Assessment in the Himalayan Upper Ganga Basin
The Research group of Dr Somil Swarnkar [Department of Earth and Environmental Science] developed an integrated framework for flood susceptibility, vulnerability, and risk assessment in the Upper Ganga Basin of the Indian Himalaya using explainable machine learning and multivariate probabilistic analysis. The study combines advanced machine-learning models, including Random Forest, XGBoost, Extra Trees, and LightGBM, with Bayesian hyperparameter optimization and SHAP (SHapley Additive exPlanations) to accurately identify flood-prone regions and reveal the key factors controlling flood occurrence. The framework further integrates flood hazard, population exposure, settlement distribution, and road density through a joint cumulative distribution function to quantify flood vulnerability and generate spatially explicit flood risk maps. Results indicate that elevation, land use/land cover, proximity to rivers, and vegetation are the dominant drivers of flood susceptibility, while densely populated valley corridors exhibit the highest vulnerability. This transparent and data-driven approach provides valuable scientific support for flood risk reduction, early warning prioritization, disaster preparedness, and climate-resilient land-use planning in the Himalayan region. For more details, kindly visit: https://doi.org/10.1016/j.indic.2026.101350.
