Modelling the elements of flash flood hydrograph using genetic programming

Abstract

A novel approach is proposed in this work on constructing the flash flood hydrograph by modelling the elements of the hydrograph namely the time to start of the initial flood (t_i), the time to peak discharge (t_p), the peak discharge (Q_p) and the base time (t_b) using Genetic Programming (GP). The proposed method is applied to the Kickapoo River catchment in Wisconsin, USA. It is demonstrated that even under limited data scenario, for a poorly gauged station, GP is able to model the elements of hydrograph with reasonably high accuracy thereby offering considerable lead time to predict the flash flood. The mathematical models developed by GP also offer some understanding of the influence of rainfall events and the stream discharge in producing the flash floods.