Markov chain modeling of daily rainfall in Lay Gaint Woreda, South Gonder Zone, Ethiopia

Abstract

An understanding and knowledge of seasonal rainfall distribution, e.g. length of the growing season and rainfall extremes, is very important in agro-based economies like Ethiopia, where 95% of the farmers depend on rainfed agricultural production. The distribution pattern of rainfall rather than the total amount of rainfall within the entire period of time is more important for studying the pattern of rainfall occurrence. Zero, first and second order Markov chain was used to describe the characteristics of rainfall occurrences in this woreda. The states considered were; dry (d) and rainy (r).The overall chance of rain and the fitted curve tells us that the chance of getting rain in the main rainy season is quadruple as compared to the small rainy season. The first order Markov chain model indicates that the probability of getting rain in the small rainy season is significantly dependent on whether the earlier date was dry or wet while the second order Marko chain indicates that during the main rainy season the dependence of the probability of rain on the previous two dates' conditions is less as compared with the small rainy season. Rainfall amounts are very variable and are usually modeled by a gamma distribution.Therefore, the pattern of rainfall is somewhat unimodial having only one extreme value in August. Onset, cessation and length of growing season of rainfall for the main rainy season show medium variation compared to the small rainy season.