Page tree
Skip to end of metadata
Go to start of metadata

GFAS archived data (2003 to yesterday) are available for download through:

(warning)Before proceeding with the steps below, users must be ECMWF registered and have their computer set up as described on How to download data via the ECMWF WebAPI (steps 1-4).

1. Learn about CAMS GFAS data

  1. Read the CAMS GFAS data documentation and get familiar with the dataset.
  2. Browse the CAMS GFAS catalogue and make sure the data you are interested in is available.

2. License agreement

Read the CAMS data license and accept it at the bottom of the page.

3. Run a CAMS GFAS test retrieval script

This is to verify that your computer is set up correctly.

  1. Copy this GFAS Python script to a text file and save it, for example as ""

  2. Run the script. On most computers you would do this by opening a command prompt and typing


The test request will download data from ECMWF and save as file "" in the directory you issued the command from. The file should be around 384 MB.

(warning) If you get an error message, most likely your computer setup is wrong, or you did not accept the data license, please go back to the previous set-up steps.

Note that the data retrieval is not designed to be instant. The test request above usually takes a few minutes to complete, but depending on demand it can take longer.

If it all worked fine, you are now ready to retrieve the data you need by following the instructions below.

4. Create your custom CAMS GFAS data retrieval script

  1. As a starting point, use the Python test script used above as a text file and save it, for example as "".
  2. Adapt the template Python script to your requirements, taking the following remarks into account:
    1. Change "param" to add the parameters of interest (separated by "/"). List of parameters is available in the GFAS documentation.
    2. For an output in NetCDF format, add "format": "netcdf" in your script. By default, output will be in GRIB format.
    3. Change "target" to the desired output path and file name, e.g.  "" or "./data/data1.grib". The default path is the current working directory.

    4. Because the current ECMWF online tools available for post-processing the data (area subsetting and regridding) do not use a mass conserving interpolation algorithm we advise against using "area" and "grid" keywords in GFAS requests. The recommended solution is to download the data at the provided resolution and perform interpolation separately.
    5. To download gridded observations from the satellite instruments used in GFAS, browse through the GFAS satellite data catalogue to check for data availability. Then change "type" to "type": "gsd" in your script. Also add "ident": "see values in table below" and "instrument": "see values in table below"in your script. Please check GFAS documentation for further information.

      SatelliteInstrumentSatellite ID ("ident")
      Instrument ID ("instrument")

Please read!

(warning) Limits: There are limits of the number and size of downloads.

(lightbulb) Data tapes and retrieval efficiency:

To retrieve data efficiently (and get your data quicker!) you should retrieve all the data you need from one tape, then from the next tape, and so on. For GFAS, this means retrieving all the data you need for one month, then for the next month, and so on.

5. Run your data retrieval script

On most computers you would do this by opening a command prompt and typing


This will retrieve the data as specified and download it as a single file to your computer into the current directory (or whichever "target" directory you specified).

Note that the data retrieval is not designed to be instant. A larger request can take hours and even days to complete.

For long running processes you can check the progress of your request in your job list.

To report an issue or bug please contact Copernicus Support at ECMWF.