Statistical modeling of the Ghanaian weather using markov chains: A case study of Accra, Kumasi and Tamale

Abstract

In this dissertation, Markov chain was used to model the rainfall pattern and sunshine duration for three regional capitals in Ghana (Tamale, Kumasi and Accra). Data on mean monthly averages and daily figures of measured parameters (sunshine duration and rainfall) for five years were used. To achieve these goals, a statistical basis was presented before adopting the three main approaches for modeling. Transition Matrix shows the short-run behavior of Markov chain; the steady-state probability, indicating the long-run behavior of Markov chain; and finally, the mean first passage times. The chain for the transition matrix for each region was found to be ergodic. The steady-state probability gave the possibility of weather change in a particular day on the condition that it was rainy or sunny for the last two days or the previous day. The mean first passage times of the weather gave the average number of daily rainfall before sunshine and an average number of daily sunshine before rainfall. The study showed that among the three towns, Kumasi has the highest probability of rainfall followed by Tamale and then Accra. Secondly, Tamale has the highest probability of duration of sunshine followed by Accra and then Kumasi. Kumasi has the highest average number of occurrence of daily rainfall compared to Tamale and Accra. Finally, Tamale has the highest average number of occurrence of daily sunshine compared to Accra and Kumasi.