Probabilistic prediction of SPI categories in Iran using sea surface temperature climate indices

This study examines probabilistic prediction of the standardized precipitation index (SPI) categories (i.e., dry, normal and wet conditions) in Iran regressed onto the combination of the North Atlantic Oscillation (NAO) index and several sea surface temperature (SST) indices including Niño4, Niño3.4, Niño3, Niño1 + 2, western Pacific (WP; 0º–15ºN, 130º–160ºE), the eastern Mediterranean Sea (EM; 30º–38ºN, 20º–35ºE) and the Indian Ocean Dipole (IOD). The ordinal regression models (ORM) based on the logistic function are applied to determine the best predictor variables. Seasonal precipitation during the two wet seasons of October-December (OND) and January-March (JFM) for 50 synoptic stations across Iran for the period 1967–2017 are used in this research. 3 month SPI at the end of December and March, which provides SPI values over OND and JFM, is constructed based on the Gamma probability distribution. The SPI categories for OND and JFM precipitation averaged over Iran are considered as the predictand variables in the ORM. The linear trend analysis of JFM SPI values indicates that the risk of drought has been enhanced in this season. Among all individual predictors, the SST anomalies over the central Pacific Ocean has the strongest teleconnection with OND SPI categories. Based on the minimum Akaike information criterion (AIC), the combination of Niño3.4 and WP gives the best model for probabilistic prediction of wet and dry events in OND. Unlike the OND, the SST anomalies over different parts of the Pacific Ocean are not strongly related to the SPI values of the JFM season in Iran. Among all indices, only the SST anomaly variations over the eastern Mediterranean Sea are statistically teleconnected to JFM SPI categories and can be used to predict dry and wet events probability in Iran.