ABSTRACT

This Research work the seasonality of monthly rainfall of Umudike was Considered. Also the time series node I suitable for the data was fitted and this was used to forecast the future values of the data. In order to achieve these objectives the data required for analysis were collected from the Agromet unit of National Root Crop Research Institute Umudike, Abia State. The set of data was the monthly rainfall of Umudike in Abia State from 2000 through 2004. The preliminary analysis revealed that monthly rainfall was stationary, jeasonal and without significant trend. The computer software MINITAB Package was used to carry out series of analysis which included seasonal differencing of the data, Autocorrelation and partial autocorrelation functions of the differenced data, so as to fit the stochastic model to the monthly rainfall series. Also the Akaike Information Criteria (AIC) method was applied in order to selectthe best model. The ;~IC of about six (6) possible first order seasonal models were calculated. In addition the Autocorrelation functions (ACF) of the residual of models fit were calculated. The mean square of the forecasted from the actual values of three adequate models were examined. All the analysis considered revealed that the monthly rainfall was stationary with SARIMA(I, O, O) 1 0, l, l),, model. Again the autocorrelation coefficients and partial autocorrelation coefficients of the residual of the model fit was computed with the MINITAB Package, it was discover that the residual values did not show any cut off at any lag, which implied that the residual values are uncorrelated. This jrther explained the fact that the model fit was the best for the data set with minimum error. Hence <n (o, 6°) secondly the et is independently and identically distributed (i i d). The model and estimate of the parameters were obtained and this showed that the model of the monthly rainfall of Umudike in Abia State ,} non-seasonal autoregressive of orde 1, seasonal differencing 1 and seasonal moving average of order ~ by acronym SARIMA(1, 0, 0) (0. 1, 1), The model revealed that monthly rainfall data in one year has a aasonal influence on proceeding year's value. In addition the values were being influenced by seasonalast year's random shocks. Finally, the future forecast of monthly rainfall indicated a seasonal fluctuating future time series .