An exhaustive global climate model performance assessment based on Lamb weather types

Sixty global climate model configurations contributing historical experiments to the Coupled Model Intercomparison Project Phase 5 and 6 are evaluated in terms of their capacity to reproduce the observed climatological frequencies of the Lamb weather types in the northern and southern hemisphere, as represented by reanalysis data. Large performance differences are obtained from one model family to another, a few of them yielding favourable results in almost the entire study area. The finer atmospheric resolution of the model versions used in CMIP6 is associated with better model performance and the obtained spatial error patterns are remarkably similar, even for some nominally different model families, which points to unexpected model dependencies. By comparison of the results for the two hemispheres, possible model tuning issues are addressed as well. Since more complete representations of the climate system are in principle preferable to simpler ones and also produce unique scenario features, a score describing GCM complexity in terms of prescribed and interactive climate system components is introduced as additional model selection criterion. In spite of the increasing uncertainty sources, the more complex models do generally not perform worse than the less complex ones. The study comes with an extensive model metadata archive based on a survey involving the model development teams themselves. This archive includes the names and versions of all considered climate system component models, integer codes describing the complexity of the coupled model configurations and other relevant metadata. It helps to avoid “black box” use of the GCMs and can be retrieved from https://github.com/SwenBrands/gcm-metadata-for-cmip/blob/main/get_historical_metadata.py