GFS Forecast 

Last update: April 2020

    IntroductionInputs - Outputs - Correction of temperature with the altitude


The Global Forecast System (GFS) is a weather forecast model produced by the National Centers for Environmental Prediction (NCEP). Dozens of atmospheric and land-soil variables are available worldwide, among which we selected a few parameters as listed in "output variables". This service delivers the most recent spatial resolution of 0.25°, corresponding approximately to 28 km. The data are available from about a month ago up to day+3.

The manual access via the Web service interface is for free and unlimited.

We also provide an automatic access to the data, but this service is for pay. Please contact us for a quotation. All the information to build GFS forecasts automatic commands.

How is built the 3 hour dataset available on SoDa?

The GFS model Cycle Runtime (CR) is 0 h, 6 h, 12 h, and 18 h. At each CR, a 3 hour forecast is provided by GFS up to 7 days ahead, but we only keep 3 days ahead. The 3 hour time step GFS dataset available on SoDa is built as follow:

  • From a month ago up to the last CR available included, each CR + the first 3 hour forecast are conserved and won't change anymore,
  • from the last CR available, only 3 hour forecast up to 3 days ahead are provided. CR+6 will be updated and thus modified as soon as a new CR is computed and available from GFS.

NB: There is a large time lag of 6 to 7 hours between the "last CR available" and the actual time due to processing constraints. For instance, the last CR available at 07:00 AM is 00:00 (7 hours = 5 hours of GFS processing + 2 hours of SoDa processing). We contemplate to add a column of "code" in the retrieved CSV file that would indicate which GFS values are stored and won't change in the GFS database, to distinguish from those who will be updated during the next run. We still need to discuss this point with the team as this option would modify the output format and consequently break the automatic processing GFS chains of customers.

How are generated data for a sampling rate below 3 hours?

A linear interpolation is applied. Please note that to generate Global Horizontal Irradiation below 3 hours, we didn't use an intelligent interpolation with Kc.



  • Latitude and longitude: enter the values in decimal degrees (we advise 3 or 4 digits after comma), or directly click on the map to select your point, or you can also use the "Search Address" tool on the top of the map.
  • Start date and End date: format yyyy-mm-dd. MERRA-2 data are available from Jan. 1980 and are regularly updated but available with approx. a month of delay.
    Please pay attention to ask for complete months when asking for monthly values. If incomplete, the given monthly value results from an interpolation based on the selected days.
  • Time step: "1 min", "5 min", "10 min", "15 min", "30 min", "hour", "day", "week", or "month". NB: "1 min" and "5 min" time steps are limited to one month of data per delivery. Original time step of the data is 3 hours.
  • Output format: CSV only


The output file format is CSV. When pushing the "process" button, wait a while and right-click on the appeared link to retrieve your file. Here is the output column content:

  • Temperature (K): temperature at 2 m above ground in Kelvins (K)
  • Relative humidity (%): Relative humidity at 2 m above ground in percent (%)
  • Pressure (hPa): Pressure at ground level in hPa
  • Wind speed (m/s): Wind speed at 10 m above ground in m/s
  • Wind direction (deg): Wind direction at 10 m above ground in degrees (0 means from North, 90 from East...)
  • Rainfall (kg/m2): Rainfall (= rain depth in mm)
  • Snowfall (kg/m2): Snowfall
  • Snow depth (m): Snow depth in meters
  • Short-wave irradiation (wh/m2): Surface downward short-wave irradiation (broadband=GHI)

Correction of the temperature with the altitude

By default, the temperature is given for the altitude of GFS (approx. 28 km of spatial resolution), available in the 6th headerline entitled: "# Site altitude" in the retrieved output CSV file. It is possible to modulate the temperature with the actual altitude of the selected point with for instance the altitude contained in the Digital Elevation Model SRTM (approx. 100 m of spatial resolution) to take into account the decrease of temperature with height. Here is the equation (Wikipedia):

Temperature_corrected (C°) = Temperature (K) - 272.15 - 0.65*(Altitude_GFS - Altitude_SRTM)/100;

Top of page