Journal of Applied & Computational Mathematics

Modeling Rainfall Data in Kenya Using Bayesian Vector Autoregressive

Abstract

Gitonga Harun Mwangi*, Joseph Koske and Mathew Kosgei

Time series modeling and forecasting has ultimate importance in various practical domains in the world. Many significant models have been proposed to improve the accuracy of their prediction. Global warming has been a big challenge to the world in affecting the normality of the day to day economic and non-economic activities. It causes far-reaching weather changes, which are characterized by precipitation or temperature fluctuations. Rainfall prediction is one of the most important and challenging tasks in the recent today’s world. In Kenya unstable weather patterns which are associated with global warming have been experienced to a greater extent. The objective of this study was to modeled rainfall patterns in Kenya by use of Bayesian Vector Autoregressive (BVAR). To achieve this objective, the data was first statistically diagnosed using Augmented Dicker Fuller and Granger Causality test. The BVAR model was developed using multiple regression analysis in a system of equations. The model sensitivity was performed using confusion matrix and the F-test was used to compare the variances of the actual and forecasted rainfall values. After the first differencing the data was found to be stationary where Augmented Dicker Fuller (ADF) test was statistically significant with P-values <0.05. The Granger Causality test found that; temperature, atmospheric pressure, wind speed and relative humidity influenced the rainfall time series models in all the regions. The model sensitivity was performed using confusion matrix. The BVAR model developed was statistically significant (R2=0.9896). The sensitivity of the model was 82.22%, making it appropriate for forecasting. In conclusion the Bayesian Vector Autoregressive model developed is suitable and sensitive for forecasting rainfall patterns.