Hello Horst,
meteoblue uses a combination of its own weather models and selected models from other providers, then blends them together for each forecast.
The two in-house model families are NMM and NEMS. NMM was the first model run by meteoblue, in operation since 2007, and is optimised for complex topography. NEMS is its successor, running since 2013, and is a multiscale model (from global down to local domains) with notably improved cloud and precipitation forecasting. Since 2013, meteoblue has also computed its own global forecast, so it no longer depends on outside models for the large-scale picture.
Alongside these, meteoblue integrates well-known external models, mainly from national weather services, such as GFS (NCEP), IFS (ECMWF) and ICON (DWD), plus reanalysis and observational datasets like ERA5, IMERG and METEOSAT for historical and supporting data.
These are then combined through the meteoblue Learning MultiModel (mLM), which uses machine learning to weigh the different models against real observations and pick whichever is performing best for a given location and situation. That combined result is what you see as the standard forecast.
If you would like to see the individual models side by side for any location, the MultiModel diagram on the website shows each one separately, along with its resolution.
Best regards,
The meteoblue team
