Comparison of conventional and differential evolution-based parameter estimation methods on the flood frequency analysis

Abstract

Accurate estimation of flood frequency is an important task for water resources management. This starts with appropriate selection of probability distribution to flood samples (annual maximum flows) that is of great importance for flood frequency analysis (FFA). In order to reach the most precise estimation, the probability distribution of the considered time series should be well defined and its parameters should be more accurately estimated. First time in the FFA literature, a differential evolution-based parameter estimation method is applied to obtain the parameters of probability distribution functions and is compared with the traditional maximum likelihood method (MLM) in the present study. For this purpose, eleven distributions have been used to describe the annual maximum flood series of nine gauging sites, with the performance of each distribution being investigated based on six criteria. The results revealed that a single distribution cannot be specified as the best-fit distribution for the study area. Moreover, it has been found that the applied approach improves the probability prediction of floods better than MLM method for efficient design of hydraulic structures, risk analysis and floodplain management.