Gridded numerical forecast model data

There are numerical forecast model data sets available through the Download Service of the Finnish Meteorological Institute. Some of them are mentioned below:

  • HARMONIE (MEPS) weather model results,

  • Hydrodynamic sea current model results for the Baltic Sea, 

  • OAAS sea level forecast, and

  • WAM wave model results for the Baltic Sea.

All these forecast model results are made available as time series for specific points (see the Timeseries data for examples). Additionally the HARMONIE (MEPS), Hydrodynamic and the WAM results are made available in gridded format, where the values of predicted properties are given in a single files for each of the grid points covering the geographic forecast area. Currently only the data near the ground or the sea surface (surface level) is provided.

The gridded numerical model data is available through the following WFS stored queries:

  • fmi::forecast::harmonie::surface::grid

  • fmi::forecast::hydrodyn::grid

  • fmi::forecast::wam::grid

The following example WFS query returns all the currently available HARMONIE (MEPS) weather model results:

The returned response contains information about each available model run result, like

  • the start and end times of the forecast period (phenomenonTime),

  • available geophysical properties (observedProperties),

  • the geometry of the forecast area (featureOfInterest),

  • nominal model run time (om:parameter named "analysisTime"), and

  • the result of the forecast Observation event

As with the radar data request, a collection of GridSeriesObservation features is returned, one for each available forecast model run (see Data models for more information about the returned data model). Within each returned GridSeriesObservation feature, there is the om:result -element, which contains the result of the Observation event. However, for gridded forecast model datasets, the om:result element does not contain the entire data grid, because the data files are too big to be encoded inside the returned XML document. Instead they contain a link to download the binary data file using a separate request:

<om:result> ... <gml:rangeSet> <gml:File> <gml:rangeParameters/> <gml:fileReference>,GeopHeight,Temperature,DewPoint,Humidity,WindDirection,WindSpeedMS,WindUMS,WindVMS,PrecipitationAmount,CAPE,TotalCloudCover,LowCloudCover,MediumCloudCover,HighCloudCover,RadiationGlobal,Visibility,WindGust,RadiationGlobalAccumulation,RadiationNetSurfaceLWAccumulation,RadiationNetSurfaceSWAccumulation,RadiationSWAccumulation&\ bbox=-18.1181798511537,49.7657866393714,54.237,75.2270233434913&origintime=2022-11-09T06:00:00Z&starttime=2022-11-09T06:00:00Z&endtime=2022-11-12T00:00:00Z&format=grib2&projection=EPSG:4326&levels=0&timestep=60</gml:fileReference> <gml:fileStructure codeSpace=""/> <gml:mimeType>application/octet-stream</gml:mimeType> </gml:File> </gml:rangeSet> ... </om:result>

For forecast model data, the binary download URL inside the WFS response points to the Binary Download Service at with GRIB2 as the response format in the above example. You need to make another HTTP request to download the binary data file for each om:result.

Downloading binary forecast data directly

As with the radar data, the binary forecast model files can also be directly downloaded without making the WFS request first. Unlike in the radar case, the binary files are not made available through WMS layers, but using a proprietary Binary Download Service under

Choose your parameters carefully

The Binary Download Service supports multiple parameters that may be used with forecast model data queries.

Notice, forecast model data files can be extremely large. Therefore, it is important to use query parameters to limit queries as strictly as possibly and to the exact area of interest.

Query parameters:

  • starttime=t (ISO time format, for example: 2013-03-05T12:00:00Z)

  • endtime=t (ISO time format, for example: 2013-03-05T12:00:00Z)

  • origintime=t (ISO time format, for example: 2013-03-05T12:00:00Z) (this is the nominal model run time)

  • timestep=n (in minutes, for example: 60,180,360,720,1440)

  • timesteps=n (number of timesteps included in response)

  • param=x (meteorological parameter, see list below)

  • projection=x (for example: projection=epsg:4326)

  • format=x (grib1 | grib2 | netcdf)

  • gridstep=x,y (chooses every N:th point in x- and y-direction)

  • gridsize=x,y (x- and y-direction)

  • gridresolution=x,y (distance between points in x- and y-direction)

  • bbox=left,bottom,right,top (for example: bbox=22,64,24,68)

  • gridcenter=lon,lat,xkm,ykm (for example: gridcenter=23.7,61.5,100,100)

Supported meteorological parameters:

  • Temperature

  • DewPoint

  • Humidity

  • WindUMS

  • WindVMS

  • Pressure

  • Precipitation1h

  • TotalCloudCover

  • GeopHeight

Supported projections

Only some projections are supported because of the GRIB-format characteristics. You are of course free to test which one work. But, some of the projections that are known to work are listed below:

  • EPSG:4326

  • EPSG:3995

  • stereographic (Polar Stereographic, latitude of origin 60, central meridian 0)

If projection is not included in query, the original projection of data is used.

Examples for downloading HARMONIE (MEPS) data in GRIB2 and NetCDF format

The GRIB2 data files can be decoded using the open source GRIB API software package provided by the ECMWF.