Evaluating the effectiveness of time series models in determining the best model for predicting annual rainfalls in selected stations in northwest of Iran

Rainfall is considered as one of the most important climatic elements and main and determining components in the water balance of any region. In addition to being semi-arid, the north-western region of Iran, due to its mountainous nature, witnesses many temporal and spatial changes in terms of precipitation. In this research, a time series was used to predict the annual rainfall of selected stations in the northwest of Iran including: Tabriz, Orumiyeh, Saqez, Zanjan, Sanandaj and Khoy during a statistical period of 61 years (1961-2021). In order to evaluate the stationarity of the model, the autocorrelation function was used, and the unstable data were converted to static data using the differentiation method. Then, the random models were evaluated to determine the best model to fit the rainfall of the stations. The distribution of time series and the trend analysis of the equation of the regression line of precipitation in the north-west synoptic stations in terms of millimeters show that the slope of the line in all stations has a decreasing trend and the decrease in precipitation in the stations is between 1 and 3 mm per year. It is estimated. Among the moving average (MA), autoregression (AR), autocorrelated moving average (ARMA) and autocorrelated cumulative moving average (ARIMA) models, according to the absolute value of the T statistic, it is greater than 2, P-value is less than 0.5 0 and the lowest value of Akaike information criterion (AIC), the (0, 0, 1) AR model for Saqez, Zanjan, Orumiyeh, Khoi and Tabriz stations and the (1, 0, 1) ARIMA model for Sanandaj station were determined as the best models. The forecast results show an increase in rainfall for 2023 and a decrease in rainfall for 2024 and 2025.