As I understand things, ERA5 uses HadISST2 sea ice cover prior to September 2007 and then uses OSI SAF data after August 2007. While HadISST2 is largely based on OSI SAF data for the satellite era, HadISST2 adjusts the OSI SAF data to better match information from sea ice charts; this can result in substantial differences between OSI SAF and HadISST2 estimates of sea ice concentration. Is any correction performed to ensure consistency between the HadISST2 and OSI SAF datasets prior to use in ERA5? I did not see any correction mentioned in this 2016 publication regarding the sea ice data used in ERA5.
For context, I ask this in part because I recently published a paper that discusses available sea ice concentration datasets for use in statistical inference of surface temperatures. For my study, one of the things I tried to do was append HadISST2 data with OSI SAF data, but I found that led to very large discontinuities, so I avoided that approach for my main results. I also know that Berkeley Earth appends HadISST2 data with satellite data from NSIDC, but they adjust the NSIDC data to try to make it consistent with the HadISST2 data.
As an aside, the 2016 publication mentions the following for sea ice uncertainty prior to September 2007: “For SIC, we use the differences between RA03 and TR14.” I suspect that the difference between the HadISST1 and HadISST2 sea ice datasets does not well represent uncertainty in sea ice concentrations, especially for the Antarctic, but I suppose that was the best information that was available in 2016. In my paper, I discussed some other sea ice concentration datasets that go back until 1850 (COBE-SST2 and G10010v2), but unfortunately, there is not good quantification in uncertainty of sea ice concentration estimates prior to the satellite era.
Also, out of curiosity, for ERA6, are there any plans to improve either the treatment of the discontinuity in 2007 or to better represent the uncertainty of sea ice concentration estimates prior to the satellite era? I did not see this mentioned in this recent ECMWF publication regarding sea ice for ERA6.
Dear Bruce,
Many thanks for your query. You can see from the Hersbach et al. 2020 paper ‘The ERA5 global reanalysis’, that the sea ice concentration input changes from HadISST2.0.0.0 to OSI-SAF in 1979, not 2007 (Table 7). In September 2007, we change the SST product from HadISST2.1.1.0 to OSTIA, and although we continue to use OSI-SAF for sea ice concentration, we start to use the operational version.
You can see from the newsletter article that you shared, that we will use ORAS6 as the lower boundary conditions for ERA6, and we aim to use OSI-SAF sea ice concentration data in ORAS6 from around 1979 once again.
Many thanks,
Alison
Hi Alison,
Thank you for clarifying that the change from HadISST2 sea ice to OSI SAF sea ice occurs in 1979 for ERA5 and will also occur in 1979 for ORAS6 and ERA6, and for clarifying there is no adjustment to the HadISST2 sea ice estimates to reduce its discontinuity relative to OSI SAF sea ice data.
You response is greatly appreciated!
Best regards,
Bruce