Author(s): W. C. D. K. Fernando; S. S. Wickramasuriya
Linked Author(s):
Keywords: Extreme rainfall; Extreme value analysis; L-moment method; GEV distribution; Return period
Abstract: Sri Lanka which has a tropical Asian monsoon climate has experienced several major flood disasters due to extreme rainfalls in the recent past. In September 2015, massive flooding occurred in the wet zone due to strong monsoonal rainfall and thousands of people were badly affected. In this context, it is essential to understand extreme rainfall patterns, which is needed in hydrologic studies. Thus, the impact of disasters can be minimized by reducing the uncertainty in computing the design rainfall magnitudes for different return periods. The main objective of this paper is to determine the most appropriate distribution for modelling the annual maximum daily rainfall series of five stations within the wet zone of Sri Lanka. Accordingly, five distributions, the Generalized Extreme Value (GEV), Three parameter Lognormal (LN3), Pearson Type 3 (P3), Generalized Pareto (GP) and Gumbel distribution were considered. The parameters of the distributions were estimated using L-moments and the goodness-of-fit (GOF) was tested using the L-moment ratio diagram, Kolmogorov–Smirnov (K-S) test and Chi-Square test. While the best fitting distribution varied from one station to another, generally the GEV distribution was found to be the most appropriate for all stations. It was evident that the Gumbel distribution was not appropriate for analyzing extreme rainfall series. The rainfall magnitudes for different return periods have also been calculated based on the GEV distribution. If the Gumbel distribution is used for calculating rainfall magnitudes at return periods of 100 or 200 years, one would be underestimating the real values by 20%-25%, thereby making a significant impact on design flood estimation.
Year: 2016