Bivariate Time Series Analysis of Rainfall Pattern in Selected States in Nigeria

Abstract

The issues concerning global rainfall distribution and warming/climate change cannot be over emphasized since it affects virtually every part of live. The study used rainfall pattern in two states of Nigeria. Data on the monthly rainfall distribution in Imo and Rivers state, for a period of 37 years was examined. The result showed a continuous increase in the pattern of rainfall for a period of thirty seven years within the period under study. However the pattern was inconsistence for the remaining years with some kind of fluctuations. The monthly mean series plots showed a clear presence of trend with the peak period being September annually in both series with coefficient of determination- R square value of 83% and 86% for Imo and Rivers series respectively. The yearly mean series plots showed a clear irregular variation over the years in both series. The irregularity in the pattern of rainfall calls for serious commitment in joining the force for climate change abatement process. Pearson correlation coefficient between the two series is 0.80 (substantial correlation), which is the same result with cross correlation between the two rainfall series (0.79988 approx. 0.80). The series reveal the following characteristics: high correlations appear seasonality of order 12 in the monthly mean plots (every September), irregular variation and trend curve which is quadratic trend. ARIMAX model with independent variable was employed to identify the best bivariate time series model for prediction purpose. The result reveal an increase in Rivers series (Yt ) would tend to a linear combination of some increase of the preceding Imo series (It ) values which is vise visa. SARIMAX (1, 2) and SARIMAX (2, 1) models are identified as the best model with respect to Rivers on Imo series rainfall; via Imo on Rivers series rainfall.