Flood Risk Assement Using AHP, Frequency Ratio, Logistic Regression, and Random Forest in Naraipur Municipality

Abstract

Floods, a natural hazard characterized by the overflow of water from its normal riverbed onto dry land, are a recurrent problem. The study area we choose is Narainapur Municipality. This study aims to develop a detailed flood risk assessment using the Analytic Hierarchy Process, Frequency Ratio, Logistic Regression,and Random Forest models. A flood inventory map was created, with 102 flooded and 95 non-flooded points Flood conditioning factors include slope, NDVI, soil type, distance from streams and roads, rainfall, TWI, aspect, and curvature. The Logistic Regression model (AUC = 0.944) achieved the highest AUC value, indicating superior predictive accuracy compared to other models. The Random Forest model (AUC = 0.936) also performed well, followed by the Frequency Ratio model (AUC = 0.855) and the AHP model (AUC = 0.822). These results highlight the effectiveness of machine learning-based models in our study area.The findings of this study provide valuable insights for flood risk management, offering a scientific basis for better planning and decision-making in vulnerable regions.