Classifications of winter atmospheric circulation patterns: validation of CMIP5 GCMs over Europe and the North Atlantic

Abstract

Winter atmospheric circulation over the Euro-Atlantic domain and three subdomains (British Isles, Central Europe, and Eastern Mediterranean) is validated in outputs of historical runs of 32 global climate models (GCMs) from phase 5 of the Coupled Model Intercomparison Project (CMIP5). Eight automated classifications of daily SLP patterns from five reanalysis datasets are produced for each domain in order to analyse the effect of the choices of methods and reference data on results. The results show that the ranking of GCMs fundamentally depends on which classification is used; therefore, only parallel usage of multiple classifications can provide robust rankings of models. Considering all eight classifications, three models (HadGEM2-CC, MIROC4h, and CNRM-CM5) are among the best in simulating the frequency of circulation types (CTs) over all four domains. Regardless the domain, the bias in CT frequency of the worst GCMs is larger than 50% of the frequency in the reference reanalysis dataset. Conversely, the best GCM for each domain differs from the reference reanalysis by about 10–20%, which is nearly the same result as found for the NOAA-CIRES Twentieth Century Reanalysis (version 2). The persistence of circulation is simulated better than the frequency with errors rarely exceeding 15%. The GCMs overestimate the frequency of westerly circulation over all domains (by about 7% over the British Isles, 21% over Central Europe, and almost 70% over the Eastern Mediterranean) and also cyclonic CTs, while easterly and anticyclonic CTs are typically underestimated by 30–40%.