Monte Carlo simulation of flood frequency curves from rainfall

Abstract

The currently adopted rainfall-based design flood estimation techniques, for example the Design Event Approach, do not account for the probabilistic nature of the key variables except for the rainfall depth. This arbitrary treatment of key inputs and parameters can lead to inconsistencies and significant bias in flood estimates for a given average recurrence interval. This paper presents a Monte Carlo simulation technique that makes explicit allowance for the probability-distributed nature of the key flood producing variables and the dependencies between them to determine derived flood frequency curves. The proposed approach employs joint probability principles to develop a design flood estimation technique that can incorporate commonly applied rainfall–runoff models and design data. The application of the proposed technique to three catchments in Victoria has shown that the new method provides a relatively precise reproduction of the observed frequency curves. The new technique is relatively easy to apply for catchments with good rainfall data and a limited streamflow record. The technique thus shows a strong potential to become a practical design tool; further work is needed to allow its routine application in a wider range of design situations.