Modelling Rainfall Patterns in Meru and Embu Regions Using ARIMA Models

Abstract

Rainfall is the natural source of water, it has greater impact on agricultural activities and domestic consumptions. Since Meru and Embu regions are agricultural zones relying heavily on rainfed agriculture, it is important for farmers to know rainfall patterns prevailing in their regions.In this study we model rainfall patterns in Meru and Embu regions of Kenya using Monthly and yearly rainfall data. ARIMA model was developed using Box-Jenkins (BJ) Methodology and fit to monthly and yearly average amount of rainfall. The data were examined to check for the most appropriate class of ARIMA processes. This was done by selecting the order of the consecutive and seasonal differencing. The auto-correlation function (ACF) and the partial autocorrelation function (PACF) are the most important elements of time series analysis. Using the AIC criterion ARIMA (1,1,1)(0,1,1)₁₂ was identified as the best model_. This model was used to forecast monthly rainfall patterns for five years and found that future rainfall patterns will change with time as contributed by many factors. It was recommended that, future researchers should consider zoning regions, identify other factors contributing to change in rainfall patterns and apply developed Arima model.

Keywords: ARIMA model, Box-Jenkins methodology, forecasting, seasonal differencing.