Bivariate peaks-over-threshold modeling of compound heat-drought extremes and their spatial dependence under climate change

Climate-related risks occurring worldwide are becoming increasingly diverse, and extreme events of greater intensity are occurring more frequently than in the past. Consequently, there is growing interest in compound risks arising from multiple extreme phenomena occurring simultaneously. This study aims to provide a unified statistical framework for analyzing climate-change risks exhibiting nonstationarity by employing the Peaks-Over-Threshold approach to the bivariate extreme value distribution, specifically through the generalized Pareto distribution. The primary focus is to detect concurrent extreme events—abnormal heat and drought during the Korean summer—and to quantify their dependence structure. The results show that exceedances of both maximum temperature and drought above predefined thresholds are well characterized by a bivariate GPD model. The spatial distribution of joint exceedances is concentrated primarily in parts of Jeollanam-do and Gyeongsangbuk-do, while union exceedances are mainly observed in the Gyeongsangbuk-do region. Within the joint-exceedance domain, maximum temperature and drought exhibit a weak positive dependence.