Comparison of the calculated frost event return period based on copula models under climate change: a case study of Chadegan region in Isfahan province- Iran

Abstract

Spring frost, is one of the important phenomena that damage agricultural production in cold areas. Predicting the occurrence of frost events can be valuable for managing and mitigating frost risk in orchards. In this study, copula models were applied to calculate the joint bivariate return period of frost event in both historical (1984–2014) and future (2023–2053) periods in Chadegan’s almond orchard. For the future period, a combination of 10 general circulation models (GCMs) from Coupled Model Intercomparison Project phase 6 (CMIP6) under three Shared Socioeconomic Pathway scenarios (SSPs) SSP1-2.6, SSP2-4.5 and SSP5-8.8 was employed using a weighting approach. The results indicated that the Generalized Pareto (GP) and Inverse Gaussian were the best marginal distribution functions of the severity (S) and duration (D), respectively. The Frank copula best explained the relationship between severity and duration of frost event. According to the joint bivariate return period of frost event, the extreme frost occurred more frequently in the future period under three SSPs compared to the historical period. In both historical and future periods, in “AND” mode, the frost event with S ≥ 6 ̊C and D ≥ 4 ̊C days, would be more likely to return in 64.71 years and about 14 years, respectively. In "OR" mode, the joint bivariate return period of mentioned frost event increase slightly in future period (3 years for SSPs) compared to the historical (1.54 years). This probabilistic assessment was pointed as a strong toll for predicting the return period of frost event in Chadegan.