Predictive Flood Susceptibility Modelling with Machine Learning: Insights from The Baitarani River Basin, Odisha

Floods are among the most destructive natural hazards, causing widespread damage to ecosystems, communities, and critical infrastructure. Accurately understanding the drivers of flooding and identifying areas prone to inundation are essential for informed disaster management and planning. This study analyzes historical flood records from 2003 to 2023 to delineate flood-susceptible (FS) zones within the Baitarani River Basin (BRB). To predict these vulnerable areas, two machine learning (ML) models were applied: the Fuzzy Support Vector Machine (FSVM) and an enhanced version optimized using Simulated Annealing (SA-FSVM). These models were developed to capture the complex relationships between flood events and multiple contributing environmental factors. Model performance was evaluated using the Area Under the Receiver Operating Characteristic (AUROC) curve, a robust metric for assessing classification accuracy and reliability. Among the tested models, SA-FSVM achieved the highest predictive performance, with an AUROC value of 0.91. The findings reveal that low-lying coastal zones, particularly in the southeastern portions of the basin, are highly vulnerable to flooding. Overall, the study demonstrates the capability of advanced ML techniques for effective flood susceptibility mapping and highlights their importance in guiding flood mitigation strategies and enhancing disaster preparedness across the BRB.