Regionalization of storm duration for determining derived flood frequency curve : a case study for Victoria in Australia

A holistic approach of design flood estimation such as the Monte Carlo simulation technique involves the simulation of thousands of storm and runoff events to determine a derived flood frequency curve. The implementation of such a technique requires the specification of the distributions of various input variables to the rainfall runoff model such as storm duration, storm intensity and initial and continuing losses. This paper presents a case study which focuses on the regionalization of the distribution of storm duration in the state of Victoria, Australia. This in particular compares the one-parameter exponential and two-parameter Gamma distributions in approximating the distribution of storm duration from 91 pluviograph stations in Victoria. Based on the Kolmogorov-Smirnov and Anderson-Darling tests, it has been found that the two-parameter Gamma distribution provides a better fit to the storm duration data in Victoria than the one-parameter exponential distribution. The application of the fitted Gamma distribution in the Monte Carlo simulation technique for generating flood frequency curves shows that this approximates the observed flood frequency curves for the selected test catchments quite well. The methodology presented in this paper can be adapted to other states of Australia or other countries, in particular where a sufficient quantity of continuous rainfall and stream flow data are available. This would particularly be useful in hydro logical study of the important/large water infrastructure projects.