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
{
"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 ID | Coverage | Notes |
|---|---|---|
icon_d2 | Germany, Switzerland, Austria | High resolution (~2 km), 48h horizon |
icon_eu | Europe | ~7 km, 120h horizon |
icon_global | Global | ~13 km, 180h horizon |
ecmwf_ifs025 | Global | ECMWF high-res, 240h horizon |
gfs_seamless | Global | US GFS blended, 384h horizon |
ukmo_seamless | Global | UK Met Office, 168h horizon |
arpege_europe | Europe | Meteo-France, 96h horizon |
knmi_harmonie_arome_europe | Europe | KNMI, 60h horizon |
knmi_harmonie_arome_netherlands | Netherlands, Belgium | KNMI high-res, 48h horizon |
dmi_harmonie_arome_europe | Europe | Danish Met, 60h horizon |
meteoswiss_icon_ch1 | Switzerland | ~1 km, 33h horizon |
meteoswiss_icon_ch2 | Switzerland | ~2 km, 45h horizon |
metno_nordic | Scandinavia | Norwegian 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.