GribStream Blog
RRFS on GribStream: pressure levels refreshed and 2D fields added
GribStream re-onboarded NOAA's experimental RRFS CONUS feeds, keeping rrfsprslev for pressure-level guidance and adding rrfs2dfld for surface, near-surface, cloud, radiation, precipitation, fire-weather, and storm-diagnostic fields.
We re-onboarded NOAA's experimental Rapid Refresh Forecast System (RRFS) feed on GribStream after NOAA changed how some RRFS output is organized in the AWS bucket.
The short version: upper-air pressure-level RRFS data remains available as rrfsprslev, and the 2D fields now have their own dataset, rrfs2dfld. If you use 2 m temperature, 10 m wind, surface weather, precipitation, cloud, radiation, soil, fire-weather, or storm-diagnostic fields from RRFS, use rrfs2dfld.
RRFS is still experimental. NOAA's AWS registry describes this as a real-time pre-implementation prototype, not an operational 24/7 monitored feed. That matters: files can be delayed, missing, renamed, split, or moved as NOAA continues the transition toward operations. We are treating this re-onboarding as the best current match to the shape that NOAA says should approximate the future operational distribution, but users should expect some instability until RRFS is fully operational.
What changed
Our original RRFS support focused on pressure-level GRIB2 files. That was useful for upper-air workflows, but the RRFS feed has been moving toward a clearer split between pressure-level products and 2D products.
GribStream now exposes that split directly:
- RRFS 2D fields:
rrfs2dfld - RRFS pressure levels:
rrfsprslev - RRFS Ensemble pressure levels:
refsprslev
We kept the existing rrfsprslev and refsprslev dataset codes so existing URLs and upper-air integrations keep their identity. The new rrfs2dfld code is the companion dataset for 2D RRFS fields.
The current GribStream scope is the 3 km CONUS grid with a short rolling archive. We are intentionally not onboarding every RRFS subset yet. Alaska, Hawaii, Puerto Rico, broader North America, and focused experimental fire-weather domains may be added later if the feed layout stabilizes around them.
Which dataset should you use?
Use rrfsprslev for isobaric upper-air fields such as:
- temperature, humidity, and wind on pressure levels
- geopotential height
- vertical velocity
- vorticity
- cloud and hydrometeor variables on pressure levels where available
Use rrfs2dfld for 2D and near-surface fields such as:
- 2 m temperature and dew point
- 10 m wind
- accumulated precipitation
- mean sea-level pressure and surface pressure
- ceiling, visibility, cloud cover, radiation fluxes
- soil, canopy, land-surface, fire-weather, and convective diagnostics
For example, a 2 m temperature selector belongs in rrfs2dfld:
{ "name": "TMP", "level": "2 m above ground", "info": "" }
A 500 mb temperature selector belongs in rrfsprslev:
{ "name": "TMP", "level": "500 mb", "info": "" }
The safest way to build requests is to check the live parameter catalog for the exact selector you need:
- RRFS 2D field inventory
- RRFS pressure-level inventory
/api/v2/catalog/datasets/rrfs2dfld/parameters/api/v2/catalog/datasets/rrfsprslev/parameters
Why RRFS matters
RRFS is NOAA's next-generation, rapidly updated, convection-allowing forecast system. The official AWS registry describes the planned operational configuration as a 3 km system that supports short-range guidance for aviation, severe convection, renewable energy, heavy precipitation, and winter weather. NCEP/EMC's RRFS and REFS evaluation page describes RRFS as NOAA's next-generation convection-allowing prediction system, built on the Unified Forecast System with both deterministic and ensemble forecasts.
For GribStream users, the practical value is storm-scale forecast data that is closer to the workflows currently served by HRRR, RAP, NAM Nest, HREF, and other regional systems. RRFS is not simply another dataset with a new name. It is part of NOAA's broader move toward consolidating regional guidance around UFS-based systems.
What to expect next
Until RRFS becomes operational, treat the GribStream RRFS datasets as experimental. They are useful today, but they are not yet the same kind of stable operational feed as GFS, HRRR, RAP, or NBM.
We will keep the current dataset codes stable where possible. If NOAA changes the feed again, we will update GribStream again and document the correction. Once RRFS becomes operational, we expect to revisit the archive depth, domains, ensemble products, and any operational NOMADS-backed distribution that becomes available.