An overview of preparation of streamflow database for ARR Project 5 Regional Flood Method

For any regional flood frequency estimation (RFFE) study, compilation of streamflow data of the gauged catchments is an essential but tedious step. As a part of Australian Rainfall and Runoff (ARR) Project 5 (Stage 3), a database of 943 unregulated/rural catchments from all over Australia has been compiled. This has been possible due to active cooperation of various state water agencies and in-kind contribution by a large number of professionals. This paper presents important information on the streamflow database which has been used to develop Australian Rainfall and Runoff (ARR) Project 5 Regional Flood Frequency Estimation (RFFE) Model, referred to as ARR RFFE 2014 Model. A total of 877 catchments have been selected from the data-rich region and 66 catchments from the arid/semi-arid regions of Australia. The selected catchments of the data-rich region are generally smaller than 1,000 km² with a median size of 178 km². The length of the streamflow data from the data rich region ranges from 19 to 102 years with a median value of 37 years. In preparing the annual maximum (AM) flood series, it was found that about 7% of the data points were needed to be infilled. A Multiple Grubbs-Beck Test (MGBT), an option available in the current ARR FLIKE software, was used to censor the potentially influential low flows from the AM series" the MGBT often finds a greater number of influential lows than the original GB test. It was found that the MGBT is superior original GB test. In at-site flood frequency analysis, the LP3 distribution with Bayesian parameter estimation method was adopted. With regard to rating curve error, it was decided to ignore rating curve error in the ARR RFFE 2014 Model development with the understanding that further work is required.