Author(s): H. M. S. M. Herath; P. R. Sarukkalige; V. T. V. Nguyen
Linked Author(s): Van-Thanh-Van Nguyen
Keywords: GEV; Extreme rainfall; Scaling concept; Non Central Moment; IDF relations
Abstract: Scaling behavior of extreme rainfall events is considered as an important phenomenon in modelling of annual maximum rainfall. Generalized Extreme Value (GEV) distribution is one of the most commonly used probability distribution to evaluate the statistical properties of extreme rainfall events. This study, evaluates the suitability of GEV distribution to assess the scaling property of extreme rainfall series observed in Western Australia. Subjected annual extreme rainfall series are 24 hours, 12 hours, 6 hours, 2 hours, 1 hour, 30 min. ., 12 min. and 6 min. annual maximum rainfalls of eight rain gauge stations and record lengths are varied between 15 years to 30 years. In the first part of the study, importance of parameter estimation methods are evaluated. Three parameter estimation methods; Non Central Moment, L Moment and Maximum Likelihood Estimation are assessed and subsequently estimated GEV parameters are modelled as a function of duration to evaluate the uncertainty of parameter estimation. Root Mean Square Error (RMSE), Percent Bias (PBIAS) and Nash–Sutcliffe Model Efficiency Coefficient (NSE) taking as indicators, modelled rainfall depths are compared with observed rainfall depths. Comparison shows that L moment method and Non Central Moment method are highly capable in GEV parameter estimation for all stations. As the second part of the study, scaling behavior of annual maximum rainfall series are analyzed. Scaling behavior of moments and Non central Moments of annual maximum rainfall events are assessed and it displays simple scaling behavior. Further, estimated GEV parameter for daily and sub daily annual extreme rainfall series are analyzed and they confirm that the Non Central Moment method is the most suitable GEV parameters estimation method in rainfall scaling studies. Finally, IDF relations are developed to sub daily rainfall events by means of a simple scaling model.
Year: 2014