The best probability models for dry and wet spells in Peninsular Malaysia during monsoon seasons

Abstract

AbstractThe identification of the most appropriate models to represent the rainfall distribution is very important as it can be utilised in various applications such as those in water resource management, agricultural planning department and hydrological sectors. Various types of probability models have been introduced and applied to the distribution of dry (wet) spells by previous researchers in the field. With the application of these probability models, this study will investigate the best probability models to describe the distribution of dry and wet spells at each of the 38 rainfall stations in Peninsular Malaysia for the period of 1975–2004. The analysis will be conducted on annual basis and also through monsoon seasons. Thirteen types of probability models with one up to three parameters including the family of geometric and log series distributions will be considered. In selecting the best models to represent the distribution of the dry (wet) spells, the Akaike's information criteria (AIC) and the Kolmogorov‐Smirnov goodness‐of‐fit (GOF) test are used. The results revealed that the modified log series (MLD) and the compound geometric (CGD) distributions were found to be the best models for the dry and wet spells respectively during the monsoon seasons. Meanwhile, the success of the mixed log series with the truncated Poisson distribution (MLTPD) was also proven in the study since this model was the most frequent best model selected in representing the observed distribution of both the annual dry and wet spells in most of the stations over the Peninsula. Copyright © 2009 Royal Meteorological Society