Abstract:
This work provides a review and general consideration of models for rain fall data. From the time series plots and test for seasonal effects on the original data, it is evident that the series have seasonal effect and that prompted the need for seasonal differences. With an additive model, the series were decomposed, and the trend and seasonal indices with respect to months were estimated. This informed the use of SARIMA model to mimic the data generating process. It was also discovered that SARIMA (0, 0, 2)*(1, 1, 1)12fits appropriately to the series. In order to capture the behavior of the series in frequency domain, Fourier series model was fitted to the data. The empirical relationships of the Fourier residual modification (FSARIMA model) on the fitted SARIMA model was incorporated, Using some model performance measures (SSE, RMSE, MAE and MBE) on the above models it was discovered that the Fourier series model out perform other models as it yielded minimum error values in all the measures of goodness of fits.
