RRFS to Become Operational August 31, 2026

The Rapid Refresh Forecast System (RRFS) and RRFS Ensemble Forecast System (REFS) is scheduled to become operational with the 1200 UTC run on August 31, 2026.  The RRFS ensemble has been intermittently available on weather.utah.edu now for a couple of years (yes, I know it hasn't been available in a few weeks).  It's been a long slog for the National Weather Service to get the RRFS operational, so let's hope this proves to be a productive upgrade.

The RRFS and REFS will replace the NAM, HREF, SREF, and HiresW operational modling systems, if you happen to use those.  These will all be retired on the same day the RRFS and REFS become operational, along with their derived products.  I won't be sad to see any of these go.  

As is often the case, they have made the acronyms and ensemble model configurations as complicated as possible.  The so-called deterministic RRFS will run out to 18 hours hourly and out to 84 hours for the 0000, 0600, 1200, and 1800 UTC cycles.  It will cover the North America, including Alaska, at 3 km grid spacing.  There will also be a relocatable 1.5 km fire weather run. 

The RRFS will also produce forecasts from five ensemble members out to 60 hours for the 0000, 0600, 1200, and 1800 UTC cycles.  These members will involve a mix of differing initial conditions, lateral boundary conditions, and model physics.  The REFS is based on a combination of the deterministic and five ensemble members from the most recent and prior RRFS runs.  It is thus a time-lagged ensemble.  

I'm not sure to what degree I'll use the time-lagged REFS or its products.  I may stick with the six-member "RRFS ensemble" as we have been doing just because it is easier and I lack the time and mind to deal with complex things.  

The full announcent is available at https://www.weather.gov/media/notification/pdf_2026/scn26-48_RRFS_and_REFS_Implementation.pdf, if you are interested in some of the details.  

As I mentioned, I hope this upgrade proves productive.  The RRFS development has had many problems and is greatly delayed.  My time in Austria, with access to many modeling systems and ensembles that are run at higher resolution and with more members than the RRFS and the RRFS ensemble, has shown me that the US is now well behind Europe in operational numerical weather prediction (see I Have Seen the Future).  The RRFS won't fully close the gap to where we could be, but we will all hopefuly benefit from an advance in operational modeling capabilities.