The accordance of synoptic-scale circulation in atmospheric reanalyses and its methodological sensitivity

Several global-scale atmospheric reanalyses (ARs) have been developed since the mid-1990s providing outputs that reach back into the pre-satellite era. These datasets, despite their deficiencies, are routinely used in climate research, prompting the need for validation and inter-comparison studies. Here, we compare synoptic-scale atmospheric circulation in an ensemble of ARs, namely, in the NCEP/NCAR Reanalysis 1 (NCEP1), ECMWF ERA-40 reanalysis, the Japanese 55-year reanalysis (JRA-55), ECMWF Reanalysis v5 (ERA5), and two ARs that assimilate only a limited number of surface variables—NOAA-CIRES-DOE Twentieth Century Reanalysis version 3 (20CRv3) and ECMWF twentieth century reanalysis (ERA-20C).

A survey on the accordance of daily circulation types (CTs) identified in daily mean sea level pressure, 700 hPa and 500 hPa geopotential height fields is carried out for boreal (DJF) and austral (JJA) winters. The CTs are independently calculated in many 30° × 20° (longitude × latitude) overlapping regions by a k-means algorithm, each region centred on one of the total of 9,360 grid points that cover the Earth from 80°S to 80°N at the step of 2.5° × 2.5°.

Not surprisingly, marked differences can be found between ARs in the pre-satellite era, particularly at the surface and over the Southern Hemisphere. Although the accordance of ARs tends to increase over time, there are several regions in which significant differences persist into the 21st century (central South America, sub-equatorial Africa, northern parts of the Indian subcontinent, western parts of China, and Antarctica).

In the presentation, methodological issues (number of CTs, size of classification domain, choice of circulation variable, choice of classification method, and type of normalization of fields prior to their classification) will be discussed, and a physical interpretation of inter-reanalysis differences will be provided.