Forecast Fresh
NOAA's new Rapid Refresh Forecast System: one 3 km model for North America, a new run every hour, and an ensemble at every point. We collect all six published members the minute they land and show you where they agree — and where they don't.
6 OCT 2026 · 12Z—
On the board
Every six hours NOAA runs the RRFS ensemble to 60 hours. Its members start landing about an hour and a quarter after the run begins and finish about three hours in; plumes appear here as soon as the first 21 hours of all six are in. The full board.
Tomorrow, twelve ways
six members + the previous run's sixOpen now
Every member, one place
Plumes for temperature, dew point, gusts, precipitation, snow and pressure; lanes that show each member's rain, snow, ice, sky and radar hour by hour; and plain sentences about where the members split.
The RRFS on a live map
Forecast Eagle's explorer, drawn from Fresh's own server: every domain from Puerto Rico to Alaska, every hourly run, 15-minute frames, city values, wind, isobars, 3-D terrain, and a swipe against HRRR, RTMA and URMA.
Every member's storms
Rotating storms tracked through all six members, with threat areas and probability swaths for each day; surface lows followed hour by hour, with Hart cyclone phase space for every member — warm core, cold core, and the transition between.
94 products, 103 regions
The grads3 map suite (radar, severe, winter, upper air, smoke, flight rules) for every 00/06/12/18Z run to 84 hours, and a run-to-run loop that shows how the forecast for one moment has changed.
The arrivals board
Which member-hours NOAA has published and which are already in our store, for every run we hold. Watch a cycle arrive.
What RRFS is, and how Fresh works
The model, the ensemble, what we store and how fast, and what is coming next.
Next exits
being builtAny member, any threshold
Member maps, stamps and probability-of-anything maps drawn on demand from the six members already in the store.
How RRFS verifies
RRFS against HRRR at 2,200 ASOS airports, hour by hour since 12 August: error by lead and time of day, station by station, rain and flight-category skill, and how honest the ensemble is.
Bias-corrected RRFS
Once the archive is deep enough, each point's recent errors correct its next forecast; then RRFS blends with the rest of the fleet.