资源环境科技发展态势分析平台

Atmospheric reanalyses have been widely used to study large-scale atmospheric circulation and its links to local weather and to validate climate models. Only little effort has so far been made to compare reanalyses over the Euro-Atlantic domain, with the exception of a few studies analyzing North Atlantic cyclones. In particular, studies utilizing automated classifications of circulation patterns-one of the most popular methods in synoptic climatology-have paid little or no attention to the issue of reanalysis evaluation. Here, five reanalyses [ERA-40; NCEP-1; JRA-55; Twentieth Century Reanalysis, version 2 (20CRv2); and ECMWF twentieth-century reanalysis (ERA-20C)] are compared as to the frequency of occurrence of circulation types (CTs) over eight European domains in winters 1961-2000. Eight different classifications are used in parallel with the intention to eliminate possible artifacts of individual classification methods. This also helps document how substantial effect a choice of method can have if one quantifies differences between reanalyses. In general, ERA-40, NCEP-1, and JRA-55 exhibit a fairly small portion of days (under 8%) classified to different CTs if pairs of reanalyses are compared, with two exceptions: over Iceland, NCEP-1 shows disproportionately high frequencies of CTs with cyclones shifted south-and eastward; over the eastern Mediterranean region, ERA-40 and NCEP-1 disagree on classification of about 22% of days. The 20CRv2 is significantly different from other reanalyses over all domains and has a clearly suppressed frequency of zonal CTs. Finally, validation of 32 CMIP5 models over the eastern Mediterranean region reveals that using different reanalyses can considerably alter errors in the CT frequency of models and their rank.