NOAA’s new forecast model updates hourly and sees storms 84 hours out

National Weather Service Director Ken Graham described the Rapid Refresh Forecast System as the biggest improvement to the agency’s regional weather models in years, and the model went into operation on Nov. 3. It refreshes every hour on a 3-kilometer grid, and across the contiguous United States it carries forecasts out to 84 hours, up from the 48 hours of the High-Resolution Rapid Refresh model. NOAA announced the milestone on Nov. 5.

Save more, waste less: The Cool Down’s newsletter sends smart, simple changes that help you and the planet in your everyday life.

Subscribe to The Cool Down →

We may earn a commission if you sign up.

The model, abbreviated RRFS, merges the hourly-update design of the High-Resolution Rapid Refresh with the continent-wide domain of the North American Mesoscale model. Forecasters get one system where they previously switched among several, and coastal and offshore areas no longer require a hand-off between models.

Hourly updates, a 3-kilometer grid and the 84-hour reach

According to NOAA’s release on the debut, RRFS predicts conditions for each 3 km by 3 km square of land surface and covers the contiguous United States, Alaska, Hawaii, Puerto Rico, Canada, Mexico, Central America and the Caribbean. NOAA also says a finer 1.5 km mode can be switched on for demand-driven work such as fire weather, complex winter weather along travel corridors and severe thunderstorms.

The 84-hour figure needs one qualifier from the operational paperwork. NWS Service Change Notice 26-48 states that RRFS runs to 84 hours on the 00, 06, 12 and 18 UTC cycles and to 18 hours on the other hourly cycles. The model therefore refreshes every hour, but the longest look-ahead arrives four times a day. The companion ensemble, REFS, runs to 60 hours with five members and is built from RRFS output.

NAM, SREF, HREF and HiresW leave service

Graham’s agency retired five older systems on the same day. A separate notice says the National Centers for Environmental Prediction discontinued the NAM, SREF, HREF, HiresW and NAM Model Output Statistics at 1200 UTC on Nov. 3, with the date allowed to slip if a critical weather day intervened. NWS had proposed the change more than a year earlier, in a 2025 public notice that asked for comments through July 26, 2025 and argued the move would unify and simplify the convective-scale models in the operational suite.

Running old and new together was not on the table. According to NWS answers published after a user webinar, simultaneous operation of the legacy models and RRFS was not feasible. The same document says the High-Resolution Rapid Refresh itself is not retired with this first RRFS version, and that its eventual retirement is tied to a later upgrade with timing still to be decided.

Smoke, dust and wildfire potential in the new package

NOAA’s release names a first for the agency. RRFS-SD, the smoke and dust version, is the first high-resolution smoke and dust model to enter NWS operations, succeeding the experimental HRRR-Smoke. It forecasts aerosol transport from fires and dust storms over 18 to 84 hours, covering both high-altitude smoke and ground-level particulates. It also brings NOAA’s first Hourly Wildfire Potential forecast, which assesses the risk of rapid, intense fire spread.

The ensemble side targets local probabilities. NOAA says the Rapid Refresh Ensemble Forecast System runs multiple simulations on the 3 km grid and estimates the chance of events such as severe hail or flash flooding, a probability framing that single deterministic runs cannot supply.

Access to the output is constrained by bandwidth. Service Change Notice 26-48 says full output goes out on the NOMADS server while only a subset travels over the NOAAPORT satellite broadcast, and that the North American grid also comes as a coarser 13 km product. Hawaii and Puerto Rico run at 2.5 km, and the notice says a real-time parallel feed had been available since Aug. 12, 2026, so outside users could test before the switch. The webinar notes add that no model output statistics product is planned for RRFS: station BUFR files replace the old FOUS bulletins, and the National Blend of Models is the preferred post-processed guidance. They also state that there is no public RRFS archive or backfill, which limits researchers who want to study past events.

Verification, known weaknesses and who built it

The NWS webinar notes are candid about trade-offs. They say RRFS generally performs well for two-meter temperature, dew point and quantitative precipitation forecasts, though it may carry a high or heavy precipitation bias, and that on average it beats the NAM on precipitation verification. The NAM handles cold-air damming well, and RRFS likely does not match it there; improving that is listed as a future priority. For three derechos in June 2026, RRFS and REFS performed on par with the HRRR and HREF.

NOAA credits its Global Systems Laboratory and the NWS Office of Modeling and Development with leading the work, alongside other NWS entities, universities, federal agencies and international partners. NOAA Administrator Neil Jacobs said RRFS delivers more accurate, earlier forecasts to support the safety and security of Americans.

Neil Jacobs and Ken Graham framed the launch around public safety, while Steve Thur, the NOAA Research assistant administrator, said the public will be better informed because of NOAA research. The practical test comes with the first outbreak of severe storms or a fast-moving winter system, when forecasters can compare the hourly cycles against what actually happens.

The unresolved piece is the follow-on. NWS has not set a date for the model upgrade that would let the High-Resolution Rapid Refresh retire, so forecasters will run RRFS beside that older model for some time to come.

This article was produced with the assistance of AI and reviewed by Morning Overview editors prior to publication.

Save more, waste less: Subscribe to The Cool Down →

We may earn a commission if you sign up.