NOAA starts a new hourly forecast model for all of North America, replacing older systems

The National Weather Service on November 3, 2026, began running a new short-range weather model, the Rapid Refresh Forecast System (RRFS), NOAA says, in place of a group of older regional systems. NOAA claims it gives the most detailed hourly short-term forecasts for the whole of North America and looks further ahead than other models. Its grid spans the contiguous United States, Alaska, Hawaii, Puerto Rico, Canada, Mexico, Central America and the Caribbean, along with coastal and offshore waters where forecasters used to move between models.

The change reaches beyond forecasters. A National Weather Service notice dated October 2 said the North American Mesoscale system (NAM), the Short-Range Ensemble Forecast (SREF), the High-Resolution Ensemble Forecast (HREF), the High Resolution Window (HiresW) and the NAM Model Output Statistics would stop at 1200 UTC on November 3, and it linked lists of variables that RRFS will not generate. By our reading, anyone pulling those files from NOAA's servers has to move to the new products. HiresW for Guam stays on, the notice says. NOAA's lab page now says RRFS replaced the NAM, HREF, SREF and HiresW in 2026, while the agency's release, dated November 5, says NOAA “will discontinue” older systems as RRFS goes operational.

How it compares

NOAA's release gives RRFS 84 hours of lead time against 48 for the High-Resolution Rapid Refresh (HRRR), figures the agency corrected on November 5 after first publishing the wrong ones. The lab page adds a detail the release leaves out: the longer runs come every six hours, while the hourly runs go out 18 hours, the same pattern the HRRR follows. By our reading, the gain over the HRRR lies in reach and coverage, since both use a 3-kilometer grid.

RRFS was also built to run as an ensemble, the Rapid Refresh Ensemble Forecast System, or REFS. It produces a set of slightly different simulations so forecasters can see how sure a forecast is and read the odds of local events such as severe hail or a flash flood. NOAA's lab says testbed and field evaluations put it level with or ahead of HREF, which it replaces. The page gives no numbers.

Smoke, dust and fire

The system's smoke and dust component, RRFS-SD, succeeds the experimental HRRR-Smoke model. NOAA calls it the agency's first high-resolution model of its kind to enter Weather Service operations. Fed by satellite and weather model data, it projects the path of fire and dust-storm particles over horizons of 18 to 84 hours. It covers both high-altitude smoke, which matters for aviation and air temperature, and particles near the ground, which affect health and visibility, and NOAA says the output supports air quality alerts.

A new Hourly Wildfire Potential forecast rates the risk of rapid, intense spread. For fire weather, winter storms along travel routes or severe thunderstorms, meteorologists can also point a finer 1.5-kilometer grid at a chosen area.

RRFS is the biggest improvement to our regional weather models in years

What the release does not show

NOAA promises earlier and more accurate forecasts. Neither its release nor its lab's RRFS page publishes scores against the models RRFS replaces, so the size of any gain cannot be judged from those documents. The lab states an ambition rather than a result: it wants RRFS to become what it calls “the single best regional model in the world.”

What comes next

NOAA's lab says a second version, built on the Model for Prediction Across Scales (MPAS), could arrive as early as 2028. In September, NOAA picked the company Tomorrow.io for contract support on its Earth Prediction Innovation Center, including work tied to bringing MPAS into the Unified Forecast System, the single modeling framework NOAA is moving toward.