Frequency Analysis And Evaluation Of Short Duration Storms For Peninsular Malaysia

The Department of Irrigation and Drainage (DID) published the Stormwater Management Manual (SWMM) in year 2000. The Manual requires Intensity-Duration- Frequency (IDF) values of low average recurrence interval (ARI) ranging from 1 month to 12 months for the design of water quantity and quality control facilities. However, these IDF values are currently not available because past analysis using annual maximum series (AMS) of rainfall depths could only derive the IDF values for ARls of 2 years and above. The SWMM has recommended as an interim solution to use simple coefficients to convert 2 year ARI rainfall intensity to obtain the 1,3,6 and 12 months ARls rainfall intensity for short durations. The coefficients were derived by fitting Gumbel distribution to the I hour duration rainfall depths obtained for the city of Ipoh and extrapolating the distribution to obtain the low ARIs . The coefficients need to be verified as they have been recommended for use in any location of Malaysia for any duration of rainfall. Twenty six rainfall stations distributed throughout Peninsular Malaysia were chosen and monthly maximum series (MMS) rainfall depth for 1 5 minutes, 30 minutes, 1 hour, 3 hours, 6 hours and 1 2 hours durations were extracted from rainfall records obtained from DID Hydrological Databank. For every station, frequency analysis was performed using the Gumbel distribution and method of moments with Gringorten Plotting Position formula to obtain rainfall intensity for 2, 3 , 6, 9, 1 2 , 1 5 and 1 8 months ARI. Concurrently, the same frequency analysis was performed using the same durations and length of data but by using AMS to obtain 2 years ARI rainfall intensities. The coefficients obtained from this analysis differed significantly from those recommended by the SWMM. As such, IDF curves were developed for the 26 rainfall stations, and are suggested for use in SWMM. Further analysis was performed on the MMS to determine the better method of estimate between the method of moment and the method of L-moment by computing the standard error of estimation based on Random Number Generation method. The findings were that the method of L-moment is a better method of estimation generally. Identification of appropriate families/parent distribution for various durations was determined from L-Moment Ratio Diagram. The study showed that Pearson Type 3 was best fit for 1 5 minutes, and 30 minutes durations, Generalised Normal curve for 1 hour duration and Generalised Extreme Value curve for 3 hours, 6 hours and 1 2 hours durations