GIS-based statistical model for the prediction of flood hazard susceptibility

At present, flood is the most significant environmental problem in the entire world. In this work, flood susceptibility (FS) analysis has been done in the Dwarkeswar River basin of Bengal basin, India. Fourteen flood causative factors extracted from different datasets like DEM, satellite images, geology, soil and rainfall data have been considered to predict FS. Three heuristic models and one statistical model fuzzy Logic (FL), frequency ratio (FR), multi-criteria decision analysis (MCDA) and logistic regression (LR) have been used. The validating datasets are used to validate these models. The result shows that 68.71%, 68.7%, 60.56% and 48.51% area of the basin is under the moderate to very high FS by the MCDA, FR, FL and LR, respectively. The ROC curve with AUC analysis has shown that the accuracy level of the LR model (AUC = 0.916) is very much successful to predict the flood. The rest of the models like FL, MCDA and FR (AUC = 0.893, 0.857 and 0.835, respectively) have lesser accuracy than the LR model. The elevation was the most dominating factor with coefficient value of 19.078 in preparation of the FS according to the LR model. The outcome of this study can be implemented by local and state authority to minimize the flood hazard.