GFS: A global model with ~13 km resolution, providing a broad overview of weather patterns over large areas and longer forecast periods.
HRRR: Covers the continental US and Alaska with very high-resolution forecasts (~3 km). Updated hourly out to 18 hours, with extended runs out to 48 hours four times a day. Particularly useful for detailed, near-term weather predictions.
The Canadian GEM model offers both global (~15 km) and regional forecasts (~2.5 km HRDPS), with the regional version providing HRRR-comparable detail for Canada and surrounding areas. GEM updates twice a day vs. four times for GFS, so the two often disagree on near-term timing.
All three models incorporate a wide range of data, including satellite and radar observations, and employ sophisticated data assimilation techniques. HRRR, in particular, is updated hourly, reflecting the latest atmospheric conditions.
Each model employs different strategies to simulate physical processes such as cloud formation, precipitation, and land-surface interactions. These differences can lead to variations in forecast details, especially in complex terrains and for specific weather events like storms or heatwaves.