Abstract
The large-scale variability of atmospheric and ocean circulation patterns cause seasonal climate changes in the Earth. In other words, climate elements are affected by phenomena like El Niño Southern Oscillation (ENSO), El Niño (NINO), and Northern Atlantic Oscillation (NAO). In this study the characteristics of the frost season over a 20-year period (1996–2015) from seven synoptic stations in western Iran were evaluated using support vector machine and random forest regression. Comparing determination coefficients obtained by these models between atmospheric and ocean circulation indices and the characteristics of the frost season showed a positive effect. Thus, the onset and the end of the frost season in this region were highly correlated with the Southern Oscillation Index (SOI) and NAO, respectively. In regions with lower correlations (central areas and some regions of Alvand Mountain), the role of the geographical factors, altitude and topography becomes more pronounced and the impact of the global indices is reduced. Cluster analysis was also conducted to detect patterns and to identify regions according to the effect of the atmospheric and oceanic indices on frost season, and three regions were identified. The largest correlations with global indices (in both models) belonged to the first and third classes, respectively. The results of this study could be applied for planning environmental and agricultural activities.