Skip to content

Models ​

The global Models list defines which Open-Meteo weather models njord polls for every location. Each model produces a separate Home Assistant device with its own set of sensors.

Configuration ​

json
{
  "Njord": {
    "Models": [
      "icon_eu",
      "ecmwf_ifs025",
      "gfs_seamless"
    ]
  }
}

Models can also be added per location to include regional models that only cover specific areas.

Verified model IDs ​

These models have been tested with njord and are known to work:

Model IDCoverageNotes
icon_d2Germany, Switzerland, AustriaHigh resolution (~2 km), 48h horizon
icon_euEurope~7 km, 120h horizon
icon_globalGlobal~13 km, 180h horizon
ecmwf_ifs025GlobalECMWF high-res, 240h horizon
gfs_seamlessGlobalUS GFS blended, 384h horizon
ukmo_seamlessGlobalUK Met Office, 168h horizon
arpege_europeEuropeMeteo-France, 96h horizon
knmi_harmonie_arome_europeEuropeKNMI, 60h horizon
knmi_harmonie_arome_netherlandsNetherlands, BelgiumKNMI high-res, 48h horizon
dmi_harmonie_arome_europeEuropeDanish Met, 60h horizon
meteoswiss_icon_ch1Switzerland~1 km, 33h horizon
meteoswiss_icon_ch2Switzerland~2 km, 45h horizon
metno_nordicScandinaviaNorwegian Met, 60h horizon

See the full model catalog for all available models with resolution and coverage details.

Coverage validation at startup ​

njord validates each model against a built-in coverage registry of geographic bounding boxes. If a model does not cover a configured location, startup logs a warning. The API itself returns HTTP 400 for out-of-coverage requests, which njord handles by skipping the model for that location.

WARNING

Adding a regional model (like icon_d2 or meteoswiss_icon_ch1) to the global model list will cause API errors for locations outside its coverage area. Use per-location models instead.

Choosing models ​

  • Start with 2 to 3 global models (e.g. ecmwf_ifs025, gfs_seamless, icon_global) for broad coverage.
  • Add regional models per location for higher resolution where available.
  • More models improve consensus accuracy but increase API usage. See budget.

Open-Meteo data is licensed under CC BY 4.0.