Modeling and Predicting the Monthly Average Temperature of Isfahan using SARIMA Model

Increasing the earth temperature causes anomalies in the planet climate, which affects all aspects of human life. In this study, the determination of temperature changes and the most appropriate model for estimating temperature changes was carried out using the SARIMA time series model in Isfahan. For this purpose, the long-term monthly average temperature of Isfahan during the years 1951-2017 were used in MINITAB software medium. Then, using the time series, an initial guessing pattern was extracted as follows: SARIMA (0, 0, 4) (0, 1, 1) 12 and SARIMA (4, 0, 0) (5, 1, 0) 12 trail and error. The method of the goodness of fit these two patterns resulted in a final pattern SARIMA (1, 0, 1) (0, 1, 1). In the next stage, accuracy and preciseness of this model were confirmed by AIC statistics and analysis of self-correlation charts, the histogram of residual patterns, and other parameters. Finally, based on fitted models, the forecast was made for the next 10 years. The results of this study indicate that these models had almost good accuracy for predicting temperature changes over the coming years. In addition, the assumption of the independency of the residuals was confirmed by the correlation between the model and the remainder of the model due to the inclusion of all self-correlations in their standard limits, and then the histogram of the residual parts of the M1 pattern showed normalilty of the data.