Weather Models

The GFS (Global Forecast System) is run by NOAA/NCEP four times daily out to 16 days at roughly 13 km resolution, and its output is fully free and public. The ECMWF model, commonly called the Euro, is run by the European Centre for Medium-Range Weather Forecasts twice daily at about 9 km resolution and has historically verified as the more accurate global model, particularly in the 3-to-7-day range and for tropical cyclone tracks.

By long-run verification scores the ECMWF (Euro) is generally the most accurate global model, with the GFS close behind and the UKMET and CMC also competitive. Accuracy depends heavily on the situation and time range, though: for thunderstorms and snow bands in the next 18 hours the high-resolution HRRR usually beats any global model, and no single model wins everywhere.

The HRRR (High-Resolution Rapid Refresh) is a 3 km convection-allowing model run hourly by NOAA out to 18 hours, with extended runs to 48 hours. Because it resolves individual thunderstorms rather than parameterizing them, it is the go-to model for short-term severe weather, heavy rain, and snowfall banding over the United States.

Spaghetti models are plots overlaying the forecast tracks from many individual model runs or ensemble members on one map, most often for a tropical cyclone. Tightly clustered lines indicate high confidence in the forecast track, while a wide spread indicates genuine uncertainty — the spread itself is the useful signal, not any single line.