Identification of the best-fit probability distributions in at-site flood frequency analysis : a case study for Australia using 127 stations

Selection of an appropriate probability distribution and associated parameter estimation procedure are of prime importance in at-site flood frequency analysis (FFA). The choice of the probability distribution for a given application is generally made arbitrarily as there is no sound physical basis to justify the selection. In this study, an attempt is made to investigate the suitability of as many as fifteen different probability distributions and three parameter estimation methods based on a large Australian annual maximum flood data set. A total of four goodness-of-fit tests are adopted i.e. the Akaike information criterion (AIC), the Bayesian information criterion (BIC), Anderson-Darling test (AD) and Kolmogorov-Smirnov test (KS) to identify the best-fit probability distributions. Furthermore, the L-moments ratio diagram is used to make a visual assessment. It has been found that a single distribution cannot be specified as the best-fit distribution for all the Australian states as it was done in the Australian Rainfall and Runoff 1987. Although the log-Pearson 3 (LP3) and generalised extreme value (GEV) distributions have been identified as the best-fit distributions for the majority of the cases, they represent the best-fit distributions for 28% and 27% cases, respectively.