Estimating distribution parameters of extreme event using TL-moment method

Abstract

Good knowledge of flood magnitude and frequency is necessary in designing different types of flood protection projects. The use of flood frequency analysis (FFA) can help hydrologists in mitigating the problem of extreme flooding. The main problem in hydrologic design faced by the hydrologists is the estimation of high flow quantile. L-Moments, popular among hydrologist in FFA, is said to be over sensitive towards the lower part of the distribution and gives insufficient weight to large sample values. As an alternative, the trimmed L-Moments (TL-moments) method is proposed to be used in FFA since it has the ability to give zero weight on the extreme values. The aim of this study is to compare the performance of L-moments, TL-moments (1,0), TL-moments (2,0), TL-moments (3,0) and TL-moments (4,0) in FFA application. The four distributions, named generalized logistic (GLO), generalized extreme value (GEV), three parameter lognormal (LN3), and Pearson 3 (P3) distributions, were chosen and an estimation of the distributions using TL-moments (r,0), r = 1, 2, 3, 4 was formulated. The comparison is done using Monte Carlo simulation and annual maximum streamflow data over stations in Peninsular Malaysia. Simulation results show that TL-moments give comparable and better parameter estimates than those by L-moments, particularly when estimating the high flow quantiles. Furthermore, the generalized extreme value (GEV) and three parameters lognormal (LN3) distributions obtained from TL-moments method suit with the actual maximum streamflows of stations in Johor