Enhancing Urban Flood Loss Mapping by Integrating ANFIS Classifier With a Two-Dimensional Hydrodynamic Model

Flood loss mapping is one of the essential prerequisites for urban flood assessment studies to identify areas vulnerable to floods and to make cities safe and resilient. This study develops a neuro-fuzzy loss model to generate flood loss maps, classifying loss levels into several categories ranging from no loss to severe loss. Several key inputs, including the location of houses, as well as flood depth and velocity, were considered in the loss model. Two major flood events were simulated using HEC-RAS 2D: one (event 1) for model development and calibration using the roughness coefficient, and another (event 2) for validating and applying the proposed flood loss model. Subsequently, the outputs of the hydrodynamic model for event 2 were integrated with the data-driven loss model to create a flood loss map for the selected residential area. According to the results, the ANFIS-based method can classify flood losses with more than 80% accuracy, demonstrating its reliability as a tool for sustainable urban planning. The proposed model generates qualitative flood loss maps, which are vital prerequisites for urban planning aimed at enhancing city sustainability. This novel method can identify both vulnerable and sustainable areas for further urban development by assessing potential flood losses in the context of regional loss assessment studies.