A weighted ensemble of regional climate projections for exploring the spatiotemporal evolution of multidimensional drought risks in a changing climate

Abstract

Understanding future drought risks plays a crucial role in developing climate change adaptation strategies and in enhancing disaster resilience. However, previous studies may lead to biased conclusions due to the neglect of two factors, including the relative performance of climate simulations and the uncertainty in drought characterization. In this study, Bayesian model averaging is used to merge five regional climate model simulations and to project future changes in hydroclimatic regimes over China under two representative emission scenarios (RCP4.5 and RCP8.5). Drought characteristics, including drought severity and duration, are extracted using the Standardized Precipitation Evapotranspiration Index (SPEI). A Bayesian copula approach is used to uncover underlying interactions of drought characteristics and associated uncertainties across 10 climate divisions of China. The regional return periods of drought characteristics are used to assess future changes in multidimensional drought risks and the probability of extreme droughts. Our findings reveal that the variations in drought characteristics are generally underestimated by the ensemble mean (AEM) simulation. The Bayesian framework improves the reliability and accuracy of hydroclimate simulations and better reproduces the drought regimes compared to the AEM simulation. The drought duration and severity are projected to substantially increase for most areas of China based on the Bayesian framework, but the AEM simulation may lead to multiple opposite behaviors, especially under RCP4.5. The estimated joint risk from drought duration and drought severity is expected to increase under both emission scenarios. The likelihood of extreme droughts is also projected to increase as the radiative forcing increases.