Regional flood modelling in arid and semi-arid regions in Australia

Regional flood frequency analysis (RFFA) is often used in hydrological studies to estimate design floods for ungauged catchments. The RFFA techniques in arid and semi-arid regions are more difficult to develop, largely because of paucity of the recorded streamflow data, extreme temporal and spatial variability of floods and long periods in the streams without any significant flow. The current national design guideline, Australian Rainfall and Runoff (ARR), contains little information and guidance on RFFA in arid and semi-arid regions of Australia. This study develops a RFFA method based on the best available data set for the arid and semi-arid regions of Australia. Here, an index type RFFA procedure is adopted. Dimensionless regional flood frequency curves are developed in the form of regional growth factors. The study uses streamflow and catchment characteristics data from 45 catchments in the arid and semi-arid regions of New South Wales (NSW), Victoria (VIC), Queensland (QLD), South Australia (SA) and Northern Territory (NT). The mean annual rainfall values for these selected catchments are smaller than 450 mm. The partial duration series flood data are used in this study to estimate the flood quantiles for average recurrence intervals (ARI) of 2 to 100 years using the generalised Pareto distribution fitted with the method of L-moments. The areas and streamflow record lengths of the selected catchments range from 3.8 to 5,975 km(2) and 10 to 46 years, respectively. The index-flood, which is taken as the mean annual flood in this study, is related to catchment area and design rainfall intensity by ordinary least squares regression to provide a means of estimating the mean annual flood for an ungauged catchment. This estimated mean annual flood and the regional growth factors permit flood quantile estimation for different ARIs in the study regions. Separate sets of prediction equations for mean annual flood are developed for VIC & NSW (combined), QLD, SA and NT. For the testing of the developed regional flood estimation methods, a one-at-a-time validation procedure is adopted. A number of evaluation statistics have been employed to assess the performances of the developed RFFA models. It has been found that the developed prediction equations provide relatively accurate design flood quantile estimates in the study regions with median relative error in the range of 20% to 110%. The steepness of the developed growth curves reflects the variability of the flood discharge with ARIs and it has been found that SA has the steepest growth curve and QLD the least one. This implies that SA design floods in arid and semi-arid regions show greater variability across different ARIs as compared to those of other Australian states. The developed RFFA methods in the arid and semi-arid regions are based on relatively shorter streamflow records and a smaller number of stations as compared to the coastal regions of Australia and hence have relatively lower degree of accuracy.