Flood susceptibility mapping is crucial for risk assessment and disaster management, particularly in flood-prone regions such as the Kaljani River Basin in West Bengal, India. This study integrates Principal Component Analysis with Artificial Neural Network and Support Vector Machine to assess flood susceptibility. This study is the first to apply a PCA-ANN and SVM-based hybrid framework to flood susceptibility mapping in the Kaljani River Basin. A multicollinearity test identified high Variance Inflation Factor (VIF) values for elevation and slope, necessitating PCA to reduce redundancy and improve model performance. The PCA-ANN model enhanced accuracy from 56% (ANN) to 81.69%, while the SVM model outperformed both, achieving 91.55% accuracy, with high sensitivity (91.18%) and specificity (91.89%). The flood susceptibility maps revealed that Very High and high-risk zones (412.98 km2) are concentrated in low-lying areas near major rivers, while Moderate zones cover 375.13 km2. Very Low and Low susceptibility zones (431.87 km2) correspond to higher elevations with better drainage. Elevation, lineament density, and land use and land cover (LULC) were identified as the most significant flood susceptibility parameters by Geodetector analysis. These results provide important insights for the design of infrastructure in the Kaljani River Basin, targeted flood mitigation strategies, and sustainable land-use management.
Sarkar et al. (Tue,) studied this question.