摘要：Partitioning uncertainty in projections of future climate changeinto contributions from internal variability, model response uncertaintyand emissions scenarios has historically relied on making assumptions aboutforced changes in the mean and variability. With the advent of multiplesingle-model initial-condition large ensembles (SMILEs), these assumptionscan be scrutinized, as they allow a more robust separation between sourcesof uncertainty. Here, the framework from Hawkins and Sutton (2009) foruncertainty partitioning is revisited for temperature and precipitationprojections using seven SMILEs and the Coupled Model Intercomparison Project CMIP5 and CMIP6 archives. The original approach is shown to workwell at global scales (potential method bias < 20 %), while atlocal to regional scales such as British Isles temperature or Sahelprecipitation, there is a notable potential method bias (up to 50 %), andmore accurate partitioning of uncertainty is achieved through the use ofSMILEs. Whenever internal variability and forced changes therein areimportant, the need to evaluate and improve the representation ofvariability in models is evident. The available SMILEs are shown to be agood representation of the CMIP5 model diversity in many situations, makingthem a useful tool for interpreting CMIP5. CMIP6 often shows larger absoluteand relative model uncertainty than CMIP5, although part of this differencecan be reconciled with the higher average transient climate response inCMIP6. This study demonstrates the added value of a collection of SMILEs forquantifying and diagnosing uncertainty in climate projections.
