Probabilistic analysis of extreme droughts in Southern Maharashtra using bivariate copulas

Drought is a multivariate natural calamity, often characterized by severity and duration. In this paper, the bivariate joint dependence between drought properties severity and duration are modeled using upper tail-dependent copulas such as Gumbel–Hougaard, Galambos and Student’s t copula families. In order to model extreme drought condition, univariate return period of drought properties are obtained using Peak over Threshold (POT) approach. Then, return period obtained using POT approach is compared with copula-based bivariate return period. The methodology is employed to investigate meteorological droughts in Marathwada region, Maharashtra state. The relative performance of different copulas showed that the joint behavior of drought properties are satisfactorily modeled using Archimedean class of Gumbel–Hougaard copula. On comparing univariate return period obtained using POT approach with copula-based bivariate return periods, an increase in drought risk in the region is observed while considering joint probability of occurrence of any one of the drought variable, whereas opposite is observed to be true when joint probability of both drought variables occur simultaneously.