Extreme rainfall at Malaysia’s highlands / Aminuddin Mohd Baki, Ismail Atan and Fazzarieyyah Abd Halim

Many of the extreme rainfall events led to flash-flooding and consequent loss of life and substantial property damage. Therefore there is an urgent need to be able to forecast this event accurately Gumbel distribution is among the common probabilistic models used in hydrological extremes especially in the case of modelling rainfall extremes. This has been especially the case in modeling rainfall extremes. It is a well known that an estimation of extreme rainfall is very important for major hydraulic structure designs. In this study peak over threshold extreme rainfall series have been proposed as an alternative to annual maximum rainfall series in extreme rainfall analysis. A peak over threshold series consists of all the peaks above a certain threshold, whereas an annual maximum rainfall series contains only the maximum rainfall of each year. It has been argued that peak over threshold method uses more information about extreme rainfall because it works with more elements than the annual maxima series method. In this project, the data used were obtained from Department of Irrigation and Drainage Malaysia. The area used as a case study covered a selected area of Malaysia’s highland such as Bukit Peninjau at Pahang, Ladang Boh at Cameron Highland, Gunung Brinchang at Cameron Highland, Genting Sempah at Wilayah Persekutuan and Bukit larut at Perak. The data were extracted using annual maxima and peak over threshold methods and then fitted with Gumbel and Pareto distributions. The return periods were calculated and compared. The obvious advantage of such models over an annual maxima approach is that, by making more of the data in any given time and furthermore, the precision of estimates and return levels can be greatly improved. From the analysis most of the data shows that the return period becomes smaller as the level of threshold becomes larger, i.e. reducing the number of exceedances. Besides, the parameters estimated also shows decreasing with the increasing of number of threshold Furthermore, the quantile plot obtained for each station shows a good fit for both Gumbel distribution and Pareto distribution except for Genting Sempah and Gunung Brinchang which shows a poor fit for both distributions. It can be conclude that data series for both stations are not suitable to be fitted in Gumbel and Pareto distribution.