Probabilistic floodplain mapping using HAND-based statistical approach

Abstract

Detection of 100-year floodplains is one of the major tasks in flood risk management. In recent years, a variety of DEM-based methods have been developed for preliminary estimations of 100-year floodplains over large regions. The higher efficiency of these methods for large-scale problems and data-scarce regions compared to the conventional hydrodynamic methods is a big advantage. However, unlike considerable advances in the field of probabilistic mapping by hydrodynamic models, these methods are mostly deterministic and cannot provide a probabilistic presentation of the floodplains. In this study, a new method is proposed to combine both advantages of probabilistic mapping compared to deterministic ones and DEM-based methods against conventional models. This method includes a probabilistic function, which uses a morphologic feature, Height Above Nearest Drainage (HAND), as the independent variable. HAND is defined as the difference in elevation between a given point and the nearest stream based on the flow direction and can be calculated from a Digital Elevation Model (DEM). The parameters of the probabilistic function are determined by using a heuristic optimization algorithm named Particle Swarm Optimization (PSO) by minimizing the error of a predicted 100-year floodplain map compared to a reference map. The results illustrate that a linear function with one parameter is an appropriate function for the study site. In addition, a comparison of the proposed method with its deterministic version demonstrates the higher effectiveness and reliability of the proposed probabilistic method for a flat watershed where the overpredictions and underpredictions generated by a deterministic threshold method are reduced.