GloFAS - Known Issues

GloFAS data downloaded from the CDS in NetCDF format are incorrect between Tuesday, March 19th and Thursday, March 21st, 2024 only. See this forum post for details.

What to do: Download the data again.

When data download is requested for a sub-region contains negative longitude values, the returned longitude value are shifted by 360 compared with the expected values. This occurs both when using the sub-region extraction tool on the CDS and when the data is downloaded directly from MARS using the 'area' keyword.

For instance, requesting an area (N,W,S,E) of: 89.975/-179.975/-59.975/-159.975

Returns a dataset with the following latitudes and longitudes:

latitude    (latitude) float64 89.97 89.92 89.88 ... -59.87 -59.92 -59.98

longitude   (longitude) float64 180.0 180.1 180.1 ... 199.9 200.0 200.0

Instead of 

latitude    (latitude) float64 89.97 89.92 89.88 ... -59.87 -59.92 -59.98

longitude   (longitude) float64 -180.0 -180.1 -180.1 ... -160.1 -160.0 160.0

To handle the data correctly, users must apply a longitude shift to their data. This is easily possible for example using xarray assign:

ds["longitude"] = ds["longitude"].where(ds["longitude"] < 180, ds["longitude"] - 360)

The longitude issue is due to the way ecCodes handles longitude since version v2.27.0. 

This will be resolved in the next release of ecCodes (v2.32.0). Please note that unlike the ongoing issue below, this issue cannot be resolved by downgrading your version of ecCodes, as it is caused by the version used at ECMWF internally.

This error causes the final longitude value of v3.1 GloFAS NetCDF data from the CDS to be incorrect. When examining the data (for instancing using xarray), it will look like this:

(longitude) float64 -179.9 -179.8 -179.8 ... 179.7 179.8 540.0   (final value should be 179.9)

This occurs because the conversion from GRIB to NetCDF via the CDS currently makes use of a version of ecCodes that contains a bug. ecCodes versions v2.27.0 and later have the issue.

If using the cfgrib library to work with v3.1 GRIB data, the GRIB data may appear to have the same issue. This is because the cfgrib library depends on ecCodes.

Until this is resolved in the next release of ecCodes (v2.32.0), we have removed NetCDF as a format option from the GloFAS CDS entries, and advise users to check the version of ecCodes they have installed if they are working with v3.1 data.

V4.0 GloFAS data is not affected by this issue.

This error caused the CDS to return incorrect GloFAS v4.0 data when an area was selected using the 'sub-region extraction' widget.

The following GloFAS datasets were affected:

• GloFAS medium-range forecasts
• GloFAS seasonal forecasts
• GloFAS medium-range reforecasts
• GloFAS seasonal reforecasts

Please discard data retrieved between 2023-07-26 to 2023-08-09 (13:00 UTC) if you used the sub-region extraction tool. Users are advised to redownload the data which will now be correct.

GloFAS historical v4.0 data was not affected.

It has come to our attention that there was an error in the CDS api downloads of GloFAS forecast and seasonal forecast data, where data requested in NetCDF format was being downloaded as a GRIB file. This error has now been corrected (as of 24/05/2023).

It has come to our attention that there was an error with the static file for upstream area in our documentation for GloFAS. This error has now been corrected (as of 28/03/2022) and the new file uploaded to the GloFAS wiki pages here. Users are advised to redownload the upstream area file, if previously downloaded before 28 March 2022.