Comparison of hydrological modeling, Artificial Neural Networks and multi-criteria decision making approaches for determining Flood Source Areas

Abstract

Abstract Flood risk management is a critical task which necessitates flood forecasting and identifying flood source areas for implementation of prevention measures in a basin. Hydrological models, multi-criteria decision models (MCDM) and data-driven models such as Artificial Neural Networks (ANN) have been used for identifying flood source areas within a watershed. The aim of this study is comparing the results of hydrological modeling, MCDM and ANN approaches in order to identify and prioritize flood source areas. The study results show that the classification results of the hydrological model and the artificial neural network have a significant correlation; also the correlation between the TOPSIS method with the hydrological model (0.252) and the artificial neural network (0.233) indicates that none of the sub-basins in the Very high, High and Very low classes are similar in the above methods. Since the neural network model has simulated the HEC-HMS classifications very accurately, it can be concluded that this model has performed very well as compared to the TOPSIS multi-criteria decision-making method.