A hybrid method for efficiently estimating the annual flood distribution under a changing climate

Abstract

In case of catchments with insufficient streamflow data or ungauged catchments, derived flood frequency approaches are often required in order to estimate design flows for flood protection or water resources management purposes. It is illustrated in this paper using a simple synthetic case study, that existing derive flood frequency approaches (design storm, design soil moisture, and event-based methods) will require long continuous simulation (CS) runs of rainfall and rainfall-runoff models to estimate ARI of extreme runoff events under climate change. Long CS runs are impractical for anything but the simplest rainfall-runoff models. To address this issue, a novel approach combining the concepts of both continuous and event-based simulation is developed to provide an efficient and reliable procedure to estimate the annual flood distribution. The methodology uses the total probability theory and the peak over threshold theory. A simple synthetic case study is used to illustrate the efficacy of the procedure. The advantages of this approach are that it drastically reduces the computational demand of a long continuous simulation in case of prediction under a changing climate, overcomes the problematic ARI neutrality assumption made in the design storm approach, and overcomes the need to "calibrate" the design soil moisture as required in many event-based methods. Future work will look at the development of the procedure using more realistic case studies.