Development of regional flood estimation methods using quantile regression technique : a case study for north-eastern part of Queensland

Several methods were recommended in Australian Rainfall and Runoff (ARR) 1987 for estimation of design floods on small to medium sized ungauged rural catchments in Queensland. These include Main Roads Department Rational Method, Department of Primary Industries Rational Method, Method of Eastgate, Swartz and Briggs, Runoff Routing and Synthetic Unit Hydrograph methods. Most of these techniques were based on limited streamflow data and of arbitrary nature and there is little information available on the relative accuracy of these methods. Since there has been accumulation of additional 20 years of streamflow data since the publication of ARR in 1987 and relatively more accurate method of at-site flood frequency analysis is available now, there is ample scope of achieving improvement in design flood estimation for ungauged catchments in Queensland similar to other states of Australia. This paper presents a pilot study which uses recorded streamflow and relevant climatic and catchment characteristics data from 76 catchments in North-eastern part of Queensland to develop and test a quantile regression technique for this region. A total of 10 independent test catchments were selected to assess the relative performances of the newly developed techniques. It has been found in this study that reasonable accuracy in design flood estimation can be achieved by using the newly developed techniques with the median error in flood quantile estimation ranges from 23 to 46% relative to the at-site flood frequency estimates. Work is continuing covering all of the state of Queensland as a separate research project; the findings of these studies will form the basis of the revision of regional flood estimations methods in the ARR for the state of Queensland.