A probabilistic model for estimation of large floods in ungauged catchments : application to South-east Australia

As part of the on-going revision of Australian Rainfall and Runoff, various regional flood estimation methods are being tested to identify appropriate method(s) for general application. This paper focuses on a Probabilistic Model, which assumes that standardised maximum observed flood data over the available period of records of each site in a region can be pooled together and described by a single probability distribution. The Probabilistic Model can be used for estimation of 'large' to 'rare' floods. The Probabilistic Model is developed and tested in this paper using data from 227 catchments from Victoria and NSW. This involves development of prediction equations using Generalised Least Squares regression for the mean and coefficient of variation of the annual maximum flood series as a function of catchment characteristics variables. An independent test shows that the Probabilistic Model provides quite reasonable design flood estimation in the study area for average recurrence intervals in the range of 20 to 200 years, with median relative error values in the range of 10 to 35%.