Quantifying the uncertainty of internal variability in future projections of seasonal soil moisture droughts over...

Understanding and quantifying drought projection uncertainty at regional scales is critical for climate adaptations and mitigations. The model uncertainty has been well represented by multi-model ensemble through the implementation of Coupled Model Intercomparison Projects (CMIPs). However, the uncertainty from internal variability is usually quantified by statistical fitting due to insufficient initial-condition ensembles for each global climate model (GCM), resulted in an underestimation of the uncertainty. In this study, Single Model Initial-condition Large Ensembles (SMILEs) that represent internal variability based on GCMs with different initial conditions, are combined with CMIP5 and CMIP6 GCMs to separate the uncertainty of seasonal soil drought projection over China. All three datasets show that internal variability dominates uncertainty for the near-term drought projection, and the internal variability uncertainty is exceeded by model uncertainty for the long-term projection. By using SMILEs as a benchmark, we revisit the method from Hawkins and Sutton (2009; hereafter, HS09) and find that this method performs well for drought projection at national scale over China. For drought projections at regional scale, however, HS09 method underestimates the uncertainty of internal variability for drought frequency, duration and intensity by 27%-54%, 15%-47% and 16%-31%, respectively. Our study highlights the importance of the selected approach for addressing the internal variability in the near-term projection of regional extremes and related adaptations.