The potential value of early (1939–1967) upper‐air data in atmospheric climate reanalysis

Abstract

In recent years a number of reanalysis datasets have been published that cover the past century or more, including the ‘Twentieth Century Reanalysis’ 20CRv2 and the European Reanalysis of the twentieth century ERA‐20C. These datasets are widely used, showing the need for, and possible benefit of, reanalysis data products designed for climate applications. The twentieth‐century reanalyses so far have assimilated only surface observations, and rely on independent estimates of monthly averaged sea‐surface temperatures and sea ice concentrations as boundary conditions. While 20CRv2 uses only observations of surface and sea‐level pressure, ERA‐20C additionally assimilates marine winds. Here we describe an experimental reanalysis, referred to as ERA‐PreSAT, which covers the period 1939–1967 and also assimilates historical upper‐air data. Assessments of this dataset including comparisons with independent data show that (i) temperature biases in the Northern Hemisphere are largely reduced compared to reanalyses that assimilate surface data only, (ii) concentration of 1940s upper‐air data in the northern extratropics created a strong interhemispheric asymmetry which is likely not realistic, (iii) the forecast skill in the Northern Hemisphere has increased substantially compared to reanalyses that assimilate surface data only, (iv) day‐to‐day and (in the northern extratropics) month‐to‐month correlations with independent observations (of total column ozone, upper‐air data) increase over time, (v) interannual variability is well captured in the reanalysis, (vi) a signature of the stratospheric Quasi‐Biennial Oscillation is present as far back as the 1940s, and (vii) tropical cyclones are not well represented. The generally encouraging results from the experimental ERA‐PreSAT reanalysis underline that early upper‐air data greatly contribute to our knowledge on the troposphere and lower stratosphere over the twentieth century.

State of the Climate in 2016

“…Radiosonde and pilot balloon are the best sources for station-based upper air wind climatologies, with data back to the early 1940s in the northern extratropics. Reanalyses need upper air information to capture the QBO, as can be seen from the increasingly better depiction with the ERA-preSAT (Hersbach et al 2017) experimental assimilation run. Surface data-only reanalyses, which extend further back, currently do not capture the QBO.…”

CERA‐20C: A Coupled Reanalysis of the Twentieth Century

“…The quality of such historical reanalyses depends on the density of observations and cannot outperform a comparable system that uses all sources of observations, including satellite and upper‐air measurements. In fact, a research reanalysis for the 1939–1967 period which assimilates radiosonde data (ERA‐preSAT; Hersbach et al, ) correlates better with independent observations than 20CRV2c. The whitelisting approach to data selection, where observations are used only if they are known to be suitable for climate applications, reduces the artificial variability and spurious trends generated by the introduction of new instruments.…”

Can reanalysis products with only surface variables assimilated capture Madden–Julian oscillation characteristics?

“…In future assimilation experiment related to the MJO the assimilation of atmospheric low‐level signals may be given priority. Meanwhile, it is admitted that adding the upper‐air information such as the one from radiosondes into the assimilation system together with surface observations could further improve the quality of the data and reduce the variability biases (Hersbach et al ., ).…”