Regional flood estimation for NSW : comparison of Quantile Regression and Parameter Regression Techniques

Australian Rainfall Runoff (ARR) (1987) recommended a number of regional flood frequency estimation (RFFE) techniques to estimate design floods in ungauged catchments in Australia. These include Probabilistic Rational Method (PRM) for eastern New South Wales (NSW) and Victoria, and Index Flood Method (IFM) for western NSW. The PRM method uses a probabilistic runoff co-efficient map based on linear geographical interpolation, while the criterion of regional homogeneity required by the IFM method is rarely satisfied for Australian regional flood data. Recent studies on regression-based RFFE techniques have demonstrated that these can provide quite accurate flood quantile estimates in Australia using only a few predictor variables. Two regression-based methods, Quantile Regression Technique (QRT) and Parameter Regression Technique (PRT) have been tested recently in Australia. The QRT develops regression equation for individual quantile; however, PRT develops regression equations of the parameters of a probability distribution. For example, in the case log-Pearson Type 3 (LP3) distribution, the PRT develops regression equations for mean, standard deviation and skew of the logarithms of annual maximum flood series data. This study presents a comparison between the PRT and QRT methods for NSW State in Australia using data from 96 small to medium sized gauged catchments. Out of 96 catchments, 12 are selected randomly as test catchments to make an independent validation of the PRT and QRT methods. The results show that both the PRT and QRT perform very similarly, with relative error values in the range of 36% to 45%. Further studies using data from different geographical regions of Australia need to be conducted before recommending an appropriate RFFE method for application in Australia.