Forecast models

Forecasts on Pluvicast are not one model's opinion. Different numerical weather prediction models cover different areas at different resolutions, and each is best at something else. These are the ones we use.

ICON-EU

The European nest of the German Weather Service (DWD) ICON model. Roughly 7 km grid over Europe, updated four times a day, reaching five days ahead. It drives the European temperature field on the globe and contributes to local forecasts. A good balance of resolution and range for continental-scale weather.

ALADIN

The high-resolution regional model of the Czech Hydrometeorological Institute, running on a grid of about 1 km over Czechia and its surroundings. Short range - up to about two days - but with detail that continental models cannot match: valley fog, local storms, mountain effects. It powers the fine Czech temperature layer.

GFS

The Global Forecast System of the US NOAA. Coarser (about 13 km) but truly global and free of charge, reaching sixteen days ahead. Used for places and ranges the regional models do not cover; the reliability of anything beyond a week is, as with all models, limited.

ECMWF

The European Centre for Medium-Range Weather Forecasts operates what is widely considered the most accurate global medium-range model. Its open data subset (about 25 km grid) contributes to our local forecasts as a second opinion next to GFS and the regional models.

Observations vs. forecasts

On the timeline, the difference matters: satellite, radar and lightning layers are observations - they exist only for the past. Model fields are forecasts - they reach into the future but are computed, not measured. Pluvicast deliberately shows both on one axis, so you can drag from what happened into what is expected and see where the data hand over.