资源环境科技发展态势分析平台

The Southeastern USA has been repeatedly affected by severe droughts that have impacted the environment and economy of the region. In this study, the ability of 32 downscaled Coupled Model Intercomparison Project Phase 5 (CMIP5) models, downscaled using localized constructed analogs (LOCA), to simulate historical observations of dry spells from 1950 to 2005 are assessed using Perkins skill scores and significance tests. The analysis is split between cold and warm seasons based on timing of agricultural planting and harvesting dates of key crops. The models generally simulate the distribution of dry days well but there are significant differences between the ability of the best-and worst-performing models, particularly when it comes to the upper tail of the distribution. Only the top models provide a good estimate of extreme dry-spell lengths with simulations of 20 yr return values within +/- 5 d of observed values across the region. The findings provide guidance on selection of a suitable model ensemble for assessment of future drought risk in the Southeastern USA.