Daily rainfall analysis, particularly on wet and dry spells provides important application in a number of fields including meteorology, agriculture, hydrology, tourism, transportation, industries, health, and water resource management. Information provided by this analysis is crucial in providing clues on the spatial and temporal structure of the daily monsoon season. This study was conducted, partly to identify the characteristics and the persistence of dry and wet spells, and also to determine best fitting models that describes the observed distribution of dry and wet spells for daily monsoon season in Peninsular Malaysia. Covering three different regions, daily rainfall data were collected from 59 stations over the Peninsular Malaysia for the period of 35 years (1975 to 2010). The first and second order of Markov chain models as well as the existing Geometric distribution will be used in fitting the distribution of wet and dry spells. In order to identify the most appropriate model that describes the distribution of wet and dry spells, the less complicated and parsimonious option is often preferred among researchers. Meanwhile, there are cases where the more complex models failed to fit the observed data, thus necessitating geometric distribution with the reciprocal of median as a potential alternative. The Geo_Med model is thus proposed in this study. The performance of the proposed models and already existing ones were assessed using the Chi-Square and Kolmogorov-Smirnov goodness-of-fit tests. Generally, findings from this study indicated that the existing second order of Markov chain model is found to be the most adequately describes the distribution of dry spells at most of the studied locations. The successfulness of the newly proposed model, Geo_Med is also proven in representing the observed distribution of wet spells at southern region of Peninsular Malaysia during both monsoon seasons.
Geometric median, markov chain models, wet and dry spells
