Reliability and importance of structural diversity of climate model ensembles

摘要: We investigate the performance of newest generation multi-model ensemble (MME) from Coupled Model Intercomparison Project (CMIP5). compare to previous models (CMIP3) as well several single model ensembles (SMEs), which are constructed by varying components models. These SMEs range where parameter uncertainties sampled (perturbed physics ensembles) through an a number physical schemes switched (multi-physics ensemble). focus on assessing reliability against present-day climatology with rank histograms, but also effective degrees freedom (EDoF) fields variables makes statistical test more rigorous, and consider distances between observation members. find that features CMIP5 general broad scales, consistent those CMIP3, suggesting similar level for climatology. The spread MMEs tends towards being “over-dispersed” rather than “under-dispersed”. In general, examined tend insufficient dispersion histogram analysis identifies them statistically distinguishable many observations. EDoFs generally greater SMEs, structural changes lead characteristically richer behaviours is obtained parametric/physical-scheme-switching ensembles. For distance measures, observations members similarly spaced each other MMEs, whereas outside ensemble. suggest should represent important component uncertainty analysis.