Reducing biases and representation errors in the global historical insitu upper-air network

In preparation for the next generation Copernicus reanalysis ERA6, we aim at augmenting the global insitu upper-air dataset with additional data and metadata, in order to reduce observation and representation errors in those data. 

For all available data, the balloon drift is calculated from the available wind data information. Also the actual launch time is supplied for as many ascents as possible, to reduce representation errors. Results indicate that background departures (from ERA5) are almost uniformly reduced for temperature, humidity and wind, with strongest reductions near the extratropical jet streams, if balloon drift is taken into account. 

The data set also contains background departures, calculated offline, for data not assimilated in ERA5, which helps spotting spurious data episodes. Background departures were calculated with respect to ERA5, before 1940 a bias-adjusted version of the NOAA/CIRES/DOE reanalysis v3 was used as reference. For the first time, we also include ascents launched from ships. 

The background departures form the basis for comprehensive statistics-based adjustment of biases in temperature, wind direction and also humidity, using the RAOBCORE/RICH method. The corresponding software has been completely rewritten in python and has an improved handling of data gaps. 

Results from bias-adjusted records indicate realistic spatial trend heterogeneity and very good fit to reprocessed satellite data products. Background departures from ERA5 increase substantially when going back to the early 1950s and 1940s, even after bias adjustments,  but are still better than departures found when comparing to the 20th century reanalysis.