cfgrib does not take the proper variable name

Question

Era5 data for wind

grib_ls -p shortName,name,paramId,table2Version,indicatorOfParameter era5_wdir-ws.grib
ws     Wind speed       10   128   10
ws     Wind speed       10   128   10
ws     Wind speed       10   128   10
...
wdir   Wind direction   3031   3   31
wdir   Wind direction   3031   3   31
wdir   Wind direction   3031   3   31

So, grib_ls gives the proper variable name. Under python, using xarray with cfgrib backend:

ws_dataset = xr.open_dataset(
    grib_file_path_wind_str,
    engine='cfgrib',
    indexpath='',    # Evita que cfgrib cree un fichero índice
    backend_kwargs={'filter_by_keys': {'table2Version': 128,
                                       'indicatorOfParameter': 10}}
)
wdir_dataset = xr.open_dataset(
        grib_file_path_wind_str,
        engine='cfgrib',
        indexpath='',
        backend_kwargs={'filter_by_keys': {'table2Version': 3,
                                           'indicatorOfParameter': 31}}
    )

And ws_dataset is:

<xarray.Dataset>
Dimensions:     (time: 8760, latitude: 13, longitude: 23)
Coordinates:
    number      int64 0
  * time        (time) datetime64[ns] 2010-01-01 ... 2010-12-31T23:00:00
    step        timedelta64[ns] 00:00:00
    surface     float64 0.0
  * latitude    (latitude) float64 30.0 29.75 29.5 29.25 ... 27.5 27.25 27.0
  * longitude   (longitude) float64 -18.5 -18.25 -18.0 ... -13.5 -13.25 -13.0
    valid_time  (time) datetime64[ns] ...
Data variables:
    ws          (time, latitude, longitude) float32 ...
Attributes:
    GRIB_edition:            1
    GRIB_centre:             ecmf
    GRIB_centreDescription:  European Centre for Medium-Range Weather Forecasts
    GRIB_subCentre:          0
    Conventions:             CF-1.7
    institution:             European Centre for Medium-Range Weather Forecasts
    history:                 2024-01-16T10:25 GRIB to CDM+CF via cfgrib-0.9.1...

So, the variable name has been properly set, but for wdir_dataset:

<xarray.Dataset>
Dimensions:     (time: 8760, latitude: 13, longitude: 23)
Coordinates:
    number      int64 ...
  * time        (time) datetime64[ns] 2010-01-01 ... 2010-12-31T23:00:00
    step        timedelta64[ns] ...
    surface     float64 ...
  * latitude    (latitude) float64 30.0 29.75 29.5 29.25 ... 27.5 27.25 27.0
  * longitude   (longitude) float64 -18.5 -18.25 -18.0 ... -13.5 -13.25 -13.0
    valid_time  (time) datetime64[ns] ...
Data variables:
    p3031       (time, latitude, longitude) float32 ...
Attributes:
    GRIB_edition:            1
    GRIB_centre:             ecmf
    GRIB_centreDescription:  European Centre for Medium-Range Weather Forecasts
    GRIB_subCentre:          0
    Conventions:             CF-1.7
    institution:             European Centre for Medium-Range Weather Forecasts
    history:                 2024-01-16T10:25 GRIB to CDM+CF via cfgrib-0.9.1...

Why does the grib tool grib_ls get the proper variable name and cfgrib not? I have installed both grib tools and python/xarray/cfgrib under a conda environment, and use them from that conda environment.

Answer 1

cfgrib uses eccodes under the hood, and not anymore the deprecated grib_api you're using, so that may be the source of the difference you're seeing. However, regardless of eccodes or grib_api the name of the variables only depends on the GRIB tables installed and shouldn't depend on the executable used. Understanding which grib tables are used is a pain in the ...., especially because the way cfgrib is installed (conda, pip...?) would mean it also installed other dependencies that may come with different GRIB tables. Plus, cfrgib is doing some additional preprocessing when interpreting the data as hypercubes.

If you're using only cfgrib I'd suggest you to forget about grib_ls and use the variable name that cfgrib is finding, which shouldn't change if you don't make any modification in the environment where it is installed.