Comparative Study Among Different Time Series Models for Monthly Rainfall Forecasting in Shiraz Synoptic Station, Iran

In this research, monthly rainfall of Shiraz synoptic station from March 1971 to February 2016 was studied using different time series models by ITSM Software. Results showed that the ARMA (1,12) model based on Hannan-Rissanen method was the best model which fitted to the data. Then, to assess the verification and accuracy of the model, the monthly rainfall for 60 months (from March 2011 to February 2016) was forecasted and compared with the observed rainfall values in this period. The determination coefficient of 99.86 percent (R2=0.9986) and positive correlation (P˂0.05) between the observed data and the predicted values by the ARMA (1,12) model illustrates the goodness of this model in prediction. Finally, based on this model, monthly rainfall values were predicted for the next 60 months that the model had not been trained. Results showed the forecasting ability of the chosen model. So, it can conclude that the ARMA (1,12) model is the best-fitted model overall.