Investigation on at-site flood frequency analysis in South-East Australia

Abstract

Flood frequency analysis is frequently adopted in flood risk assessment for catchments where recorded streamflow data of reasonable length are available. In the Australian Rainfall and Runoff (ARR) 1987, the national guideline for design flow estimation, Log Pearson Type 3 (LP3) distribution fitted with method of product moments (MPM) is recommended for at-site flood frequency analysis. This paper uses annual maximum flood data from 18 gauged sites across South-east Australia to assess the relative performances of the currently recommended method of at-site flood frequency analysis (LP3-MPM) and some of the techniques developed in recent years e.g. Generalised Extreme Value (GEV)-LH moments, LP3-Bayesian maximum likelihood (BML) and Generalised Pareto (GPA)–L moments. Several measures are adopted to compare the performances of various flood frequency analysis methods in the range of average recurrence intervals of 2 years to 100 years including statistical hypothesis testing and comparison of the quantile estimates obtained from the fitted distributions with graphical estimates. It is found that the GPA-L moments and GEV-LH moments methods provide the best fit to the observed flood data of the selected stations followed by the LP3-BML method. The currently recommended ARR method (LP3-MPM) does not perform as well as the above three methods.