Combining regional flood frequency estimates with an at site flood frequency analysis using a Bayesian framework : practical considerations

In flood frequency analysis the estimation of rare flood quantiles is subject to considerable uncertainty, largely because of the short length of the gauged record. This uncertainty can be reduced, sometimes considerably, by augmenting the information in the gauged record with regional flood frequency information and, where available, information about large floods outside of the gauged record. The Bayesian approach is particularly well suited to making full use of these diverse sources of information. Case studies are presented, drawn from both synthetically generated and recorded data, illustrating the application of regional and historical information to at-site flood frequency analysis using the Flike software. Regional information was derived from Australian Rainfall and Runoff Project 5 and this paper illustrates how this information can be used as a prior in the at-site flood frequency analysis based on the Bayesian approach. Emphasis is given to providing an overview of using Flike and its key features. These features include the multiple Grubbs-Beck test to remove low outliers. Practical applications of analyses compliant with the latest Australian Rainfall and Runoff guidelines as well as illustrations of the assessment of Flike outputs are presented.