Maximum entropy method for flood frequency analysis: A case study of the Grand River in Ontario, Canada

Abstract

Flood frequency analysis provides the fundamental information for the design of hydraulic and civil infrastructure facilities as well as water resources management. The available procedures to determine flood distributions and their parameters from historical data suffer from restriction on the family of presumed standard classical distributions, such as gamma distribution and extreme value distributions. This paper presents a distribution-free approach for flood frequency analysis by using the combination of maximum entropy method and Akaike’s information criterion. The determination procedure is given in a flowchart. It is distribution-free because no classical distributions were presumed in advance and the inference result gives a universal form of probability curves. Comparative studies from historical flood data show that the proposed method can, accurately and reasonably, characterize the probabilistic information of observed or transformed flood data.