Modeling monthly rainfall in Katunayake region using seasonal arima model

Abstract

The accurate forecast of rainfall is much important as the rainfall is one of the factors which is bound to human beings in routine life. The prediction of rainfall on a seasonal time scale has been attempted by various research groups using different techniques. In the present study, a univariate time series seasonal autoregressive integrated moving average (SARIMA) model has been developed for monthly rainfall data from a period of January, 2001 to January, 2016 (181 observations) in Katunayake region, Sri Lanka. For the model selection it was used 157 observations while the rest 24 observations were used to validate the developed model. The Johnson transformation was used to transform observations in order to correct the non-normality of the residuals. Based on the results, the SARIMA (2,0,2) (2,0,1)12 model was found to be most suitable for forecasting the mean rainfall. The Akaike Information Criterion (AIC), Schwarz Information Criterion (SIC) and Durbin Watson statistics were used to test the validity of the developed model in different stages. This model is appropriate to forecast the monthly rainfall for the future months to assist decision and policy makers to establish priorities for water demand, storage and disaster management.