Assessment of Nonstationary Drought Frequency under Climate Change Using Copula and Bayesian Hierarchical Models

Abstract

Characterization of nonstationarity in drought metrics due to the effects of combined forcings of natural climate variability and anthropogenic climate change are critical to effective adaptive management of future droughts. In this study, we propose a nonstationary copula–Bayesian hierarchical model framework to perform drought severity–duration–frequency (S-D-F) analysis. The methodology is demonstrated for a study region in Oregon, where significant temporal trends in meteorological drought have been observed. Based on the deviance information criterion (DIC), the nonstationary Bayesian hierarchical model with prior and hyperprior distributions is the best choice for nonstationary frequency analysis. The S-D-F curves developed for historical and future climate are compared to understand the different impacts of climate change on meteorological drought patterns in the study region. The average drought severity is projected to increase by up to 25% under representative concentration pathway (RCP) 4.5 scenario in the 2021–2040 period at a few locations in the study region. Similarly, under the RCP 8.5 scenario, changes in projected drought characteristics are indicative that drought conditions may exacerbate by the end of the 21st century. Severe drought events are also projected to have lower return periods by the nonstationary models. The study highlights the importance of applying the nonstationary S-D-F curves in water resource systems design and analysis.