Evaluation of Critical Flood Conditioning Factors for the Generation of Flood Susceptibility Map for Egor Local Government Area, Edo State, Nigeria

The objective of this paper was to investigate nine (9) critical flood conditioning factors (elevation, slope, aspect, Topographic Wetness Index (TWI), flow direction, drainage density, flow accumulation, normalized difference vegetation index (NDVI), and rainfall to generate flood susceptibility map for Egor Local Government Area, Edo State, Nigeria using a GIS-based Artificial Neural Network (ANN) model by splitting into 70% training datasets and 30% testing datasets. The network was optimized through hyperparameter tuning and regularization to enhance accuracy and generalization. The resulting maps provide detailed spatial representations of the topographic, hydrological, and environmental characteristics influencing flood dynamics within the study area. The model's performance was validated using historical flood inventory data achieving the following: accuracy of 92%, a Kappa statistic of 0.84, sensitivity of 0.9580, and specificity of 0.8820. The AUC value of 0.948 indicates predictive capability. The flood susceptibility map categorizes flood risk into five distinct levels, from very low to very high, enabling targeted flood mitigation strategies and effective disaster response planning. This research shows the utility of machine learning techniques in flood risk assessment and highlights their potential for enhancing community resilience against flooding.