Integrated Machine Learning and Hydrodynamic Modeling for Flood Susceptibility Mapping in the Lower Narmada River Basin, India

Floods are among the most destructive natural disasters, causing significant economic, social, and environmental impacts, particularly in developing countries like India. Settlements in flood-prone areas and a lack of information and awareness exacerbate flood risks. This study proposes an integrated framework combining machine learning and a hydrodynamic model (HECRAS) to map flood susceptibility in the lower Narmada River basin, India. For this purpose, the study evaluates and applies Artificial Neural Network (ANN), Random Forest (RF), and Support Vector Machine (SVM) to develop flood susceptibility maps. The framework incorporates flood hazard factors such as elevation, topographical wetness index, slope, distance from the river network, drainage density, rainfall, and landuse landcover (LULC) characteristics, along with vulnerability factors like population density, agricultural production, and road–river intersections. The model will be trained using flood depth data from the hydrodynamic model. Moreover, the HECRAS model will be validated with historical flood events using Normalized Difference Water Index (NDWI) analysis from satellite imagery. The integrated approach is expected to achieve high predictive performance, with certain variables anticipated to be key contributors to flood risk. Results demonstrate the robustness of combining machine learning with hydrodynamic modeling for flood mapping, offering improved spatial and temporal accuracy. This study provides a reliable tool for policymakers and stakeholders to identify flood-prone areas, implement mitigation measures, and enhance flood disaster management strategies in the region.