A hemispheric-wide performance assessment of the CMIP models based on recurrent atmospheric circulation patterns for use in regional climate studies

The present study provides an exhaustive performance assessment of the the global climate models participating in the Coupled Model Intercomparison Project Phase 5 and 6 for the northern hemisphere extratropics. Meant as a direct answer to the request for model selection criteria tailored to the requirements of the downscaling and bias correction communities, the evaluation is based on the climatological frequencies of recurrent regional atmospheric circulation patterns as defined by Lamb (1972). In total, 126 historical integrations from 51 different models are assessed, taking into account the effects of internal model variability and reanalysis uncertainty. Three problematic regions (the southwestern United States, the Gobi desert, Greenland and the surrounding seas) are detected where reanalysis uncertainty considerably alters any attempt to rank the models. In the remaining regions, this kind of uncertainty is generally negligible. It is shown that the best performing models are not necessarily the most complex ones, where complexity is here defined as the number of realms considered by the coupled model configuration. Internal model variability only has a small effect on the overall results and, even at the grid-box scale, shifts the ranks only by a few integers. On average, the model versions used in CMIP6 perform better than their CMIP5 predecessors and so do classical atmosphere-ocean global circulation models (AOGCMs) if compared with fully comprehensive Earth System Models (ESMs). Both findings can be explained by a moderately strong relationship between the models’ performance and the horizontal resolution of the atmospheric model component (r = 0.65), which tends to be finer in CMIP6 if compared with CMIP5 and in the AOGCMs if compared with the ESMs. The study provides an objective model ranking based on performance for any region of then northern hemisphere extratropics that can be used for model selection in regional climate studies. Since model complexity should be also taken into account, the overall ranking is complemented by separate rankings for the AOGCM and ESM families. The role of model complexity as a selection criteria should be further discussed in the future. Reference: Lamb, H.: British Isles Weather types and a register of daily sequence of circulation patterns, 1861-1971, Geophysical Memoir, 116, 85pp., HMSO, 1972.