Version 2.1 of the AWI CryoSat-2 sea ice product was introduced with the start of 2018/2019 Arctic winter season in October 2018. A number of changes where introduced, mainly in the form of updates and improvements of auxiliary data and an improved computation of uncertainties of gridded data. All changes are detailed here:
Updated Snow Depth Climatology
Previous versions of the AWI CryoSat-2 product relied on a modifed Warren 99 climatology (50% over sea ice) as input data set for the entire northern hemisphere and thus using the monthly snow depth and density far beyond its intended scope. Since the Warren climatology is presented as a 2D quadratic fit it will provide information for all locations, but the result in regions outside the central Arctic Ocean might not be in a sensible range or physically possible. A filter for an assumed valid snow depth range between 0 an 60 cm therefore led to significant data gaps in some month.
In version 2.1 we introduce an incremental update of a snow climatology by merging Warren data in the central Arctic with monthly mean snow depth from AMSR2 provided by the Institute Of Environmental Physics (IUP) at the University of Bremen outside the Arctic Ocean. Both data sets are blended by a regional weight function that ensures smooth transition between the Warren and AMSR2 domains. The weight function is also used to apply the 50% snow depth reduction over first-year sea ice only in the Warren region.
As for snow density, we use a monthly mean sea ice density for the Warren region for the entire Arctic. This simplified approach also reduces issues with the 2D quadratic fit of snow density in some regions.
The monthly grid files of the updated snow depth climatology can be download from the AWI ftp.
Updated Level-3 (Grid) Uncertainties of Freeboard & Thickness
In the previous version, the all uncertainties of geophysical parameters in Level-3 (gridded) products where computed from averages of the corresponding Level-2 (trajectory) points inside the grid cell. This approach neglected that some random error components are reduced by averaging. The resulting uncertainty was therefore the maximum uncertainty of the algorithm and very likely overestimated the actual uncertainty of 25 km grid points.
In version 2.1 the uncertainty of radar freeboard, freeboard and sea ice thickness is computed with the list of geophysical variables and the usual error propagation function. However, for random components such as range noise, we use the error of a weighted mean and for systematic errors such as mean Level-2 error as before. The result is a Level-3 uncertainty which consistently lower and in a more plausible range. The version 2.0 uncertainty is also available in version 2.1 products under the name
Added Region Code (Modified NSIDC Regional Mask)
From version 2.1 the data products will contain a region code field that can be used to automatically extract sea ice thickness or any other parameters for a given region. The region code comes in form of a number and is available for both daily trajectory data (l2p) and gridded data (l3c).
The region code was extracted from the NSIDC regional mask provided by the ARCUS Sea Ice Prediction Network and the Baltic Sea was added with code 16.The names of the flags are provided by the
flag_meaning variable attribute in the netCDF attribute and also documented .
Updated Mean Sea Surface (DTU18)
The mean sea surface data source for the freeboard estimation has been updated from DTU15 to DTU18 (1 minute grid) of the Danish National Space Center (DTU).
The changes of DTU18 in the Arctic Ocean compared to the earlier version are summarized in the presentation:
Version 2.1 also introduced a number of minor changes and additions:
- Modified filenaming and global attributes of the netcdf files to be more compliant with the Data Standards Requirements for ESA CCI Data Producers
- Set the sea ice concentration value automatically to 100% close to the north pole if the operational OSI-SAF sea ice concentration product (OSI-401) yields a NaN in the pole hole of the passive microwave sensor. This resulted to a loss of sea ice freeboard/thickness in version 2.0 due to a failure of the surface type classification algorithm. The assumption of sea ice coverage near the pole (> 87N) can be safely made for the October through April period.
- Renamed the variable
sea_ice_thickness_uncertainty. (Please also see the change of Level-3 uncertainties outlined above).
- No labels