Markov Chain Probability Models to Describe Bi-modal Rainfall Pattern in Sri Lanka

Abstract

Purpose: Sri Lanka is a tropical island with prominent four monsoonal seasons. The rainfall pattern of the country follows a bi-modal distribution with two peaks centred on two growing seasons. Model the bimodel rainfall distribution of Sri Lanka using the Markov probability chain model was the objective of this study. Research Method: Both first and second-order Markov probability models were developed for Anuradhapura, Badulla, Hambanthota and Katunayake using daily rainfall data for 1981–2011 period and the locations were selected to cover the major parts of the country. The model fit was done using Instat Statistical Programme. Findings: Both first and second-order Markov models successfully described bi-modal distribution of rainfall. In general, both transitional probabilities in the first order (p_rd and p_rr) and three transitional probabilities of second-order except rainfall after dry day and rainy day (p_rrd) followed a bi-modal pattern with two peaks. The sum of the logs of the rainfall amount (lr) and the amount of rainfall on rainy days (r_mean) also showed two peaks for two growing seasons. In both models, stations in the dry zone showed higher agreement in the simulated rainfall. Research Limitations: Lack of continuous long-term rainfall data is one of the major limitations. Originality/ Value: It is evident that both first and second-order Markov chain probability models are very much capable to explain the bi-modal rainfall distribution in Sri Lanka.