Searching for an appropriate probability distribution for at-site flood frequency analysis in south-east Australia

Design flood estimation is often required in hydrologic design, flood risk assessment, and various environmental and ecological studies. Flood frequency analysis is commonly used to estimate design floods where recorded streamflow data of reasonable length are available and where there has been little land use changes. In Australian Rainfall and Runoff, Log Pearson Type 3 (LP3) distribution has been recommended for at-site flood frequency analysis. More recently, there has been an increased focus on the use of Generalised Extreme Value (GEV) distribution in Australia similar to many other countries. This paper uses data from 104 gauged sites across Victoria, New South Wales and Australian Capital Territory to assess the relative performances of five probability distributions for flood frequency analysis: LP3, GEV, Log Normal, Extreme Value 1 and Normal. The flood frequency estimates obtained from these distributions are compared with non-parametric estimates based on the site's recorded streamflow data. The study considers six average recurrence intervals (ARIs): 2, 5, 10, 20, 50 and 100 years. It has been found that LP3, GEV and extreme value 1 distributions perform very simarly, in that the median estimation error is about 10%.