Uncertainty of the future changes in interannual precipitation variability under global warming based on SMILEs and CMIP6

The change in interannual precipitation variability (PIAV), especially the El Niño–Southern Oscillation-driven part over the Pacific, has sparked worldwide concern. However, it is plagued by substantial uncertainty, such as model uncertainty, internal variability, and scenario uncertainty. Single-model initial-condition large ensembles (SMILEs) and a polynomial fitting method were suggested to separate these uncertainty sources. However, the applicability of a widely used polynomial fitting method in the uncertainty separation of PIAV projection remains unknown. This study compares three sources of uncertainty estimated from five SMILEs and 28 models with one ensemble member in phase 6 of the Coupled Model Intercomparison Project (CMIP6). Results show that the internal uncertainty based on models with one ensemble member calculated using the polynomial fitting method is significantly underestimated compared to SMILEs. However, internal variability in CMIP6 as represented in the pre-industrial control run, aligns closely with SMILEs. At 1.5°C warming above preindustrial level, internal variability dominates globally, masking the externally forced PIAV signal. At 2.0°C warming, both internal and model uncertainties are significant over regions like Central Africa, the equatorial Indian Ocean, the Maritime Continent, and the Arctic, while internal variability still dominates elsewhere. In some regions, the forced signal becomes distinguishable from internal variability. This study reveals the limitations of the polynomial fitting method in separating PIAV projection uncertainties and emphasizes the importance of SMILEs for accurately quantifying uncertainty sources. It also suggests that improving the intermodel agreement at warming levels of 1.5°C and 2.0°C will not substantially reduce uncertainty in most regions.