Estimation of ARIMA model parameters for drought prediction using the genetic algorithm

Drought is a phenomenon that occurs slowly, affecting surface water and groundwater resources, which can reduce the water supply, worsen water quality and deteriorate agricultural products, and affect economic and political activities. Urmia Lake basin has been exposed to severe water stress in recent years. Therefore, in this study, drought monitoring in Tabriz synoptic station as one of the most important stations in Urmia Lake basin in the short-term, mid-term, and long-term steps of 53 years period was investigated using the Standard Precipitation and Evapotranspiration Index (SPEI). Then, the prediction of the drought was investigated using a hybrid model of the genetic algorithm (GA) and autoregressive integrated moving average model (ARIMA) approaches. The results showed that there were three long periods of drought related to 1961–1963, 1986–1992, and 1997–2009 during the statistical period. In the prediction section, the results showed that based on the Brock-Dechert-Scheinkman (BDS) test, in all SPEI time scales, the time series has predictability. Also, the prediction accuracy of the GA-ARIMA model has a direct correlation with the SPEI time scale. So that in the test section, the determination coefficient in the 1-month time scale (SPEI1) has increased from 0.34 to 0.93 in the 48-month time scale (SPEI48).