|Series/Collection||ERA Report Series|
|Authors||Poli, P, Hersbach, H, Berrisford, P, authors, other|
|Event Series/Collection||ERA Report|
|Place of publication||Shinfield Park, Reading|
The ERA-20C deterministic (single-member) reanalysis uses spatio-temporally varying background errors derived from the earlier ERA-20C ensemble, and background error correlations from ECMWF operations in 2013. The resulting climate trend estimations are more realistic than the earlier ensemble production. The observation quality controls are also generally improved, except for tropical cyclone bogus observations, largely rejected in both productions. In ERA-20C deterministic, the inter-annual variability of total column water vapor over tropical oceans (total column ozone over northern high latitudes) agrees more with observational products HOAPS and RSS (respectively, Multi-Sensor Reanalysis-2) than ERA-Interim and JRA-55 (respectively, than the model integration ERA-20CM). The realism of the general circulation of ERA-20C deterministic appears adequate in the Northern hemisphere mostly after World War II. In the Southern hemisphere, the quality of the circulation is much more doubtful until the late 1980s, which coincides with a wider deployment of drifting buoys. This is supported by several metrics and suggests that the barriers to improving our knowledge of the 20th century weather lie with the availability of observations. As the ERA-20C deterministic reanalysis only provides a single realisation, the use of the spread between the various members of the earlier ERA-20C ensemble production is explored as a proxy for estimating uncertainties. For two-metre temperature, we find that the ensemble spread may be used to assess uncertainties in synoptic weather charts, but not uncertainties related to climate time-scales; the spread between monthly means is probably too small to be realistic. This can be explained by all sources of uncertainties prescribed with a short memory and no time-correlation, except for the SST forcing. Finally, the report proposes several leads for improvement for a future repeat of a 20th century reanalysis at ECMWF.