HREF/NAM/HRW Replacement with RRFS

[csteele2]

Upon seeing the announcement of RRFSv1 implementation in "early 2026", I looked into testing impacts of workflow. I noticed several with the retirement of NAM/HRW. I also noticed that parallel data is not available on nomads, but did find the s3 bucket seemed to go through a reorg recently, and I much prefer the faster/more efficient transfers from AWS without getting a ban hammer from nomads, so I started poking.

These files seem agreeable with cfgrib, so visualization of them isn't a problem, but they do not seem to agree with other grib utilities - especially ungrib (for use to drive WRF/MPAS). I seemed to get ungrib to accept them by reprojecting to regular lat/lon, but using the raphrrr Vtable in the wps repo doesn't work too well with the pressure-level files, and for nat levels, I find ungrib doesn't decode much of the surface and soil fields (ungrib log shows "-" for RH, PSFC, PSML, SOIL*, LANDSEA, etc.), which is another big problem. With some scouring I found a Vtable.RRFS, but it looks pretty close to raphrrr and doesn't really change the results..

My concern is that RAP/HRRR is unsupported with NCAR repo of MPAS, so I would be left with nothing to drive regional high-res simulations if the NCAR WPS repo does not work with RRFS files and/or RRFS inherits the same soil issue for MPAS (since it looks like it has the same LSM).

[EricJames-NOAA]

@csteele2 thanks for bringing up this issue. I've been looking into ungrib recently, and I believe ungrib is having trouble with the full RRFS domain GRIB2 files because of the file size. I was able to get ungrib to run on full North America RRFS domain files by splitting the GRIB2 file into smaller chunks.

For those specific surface / soil fields, I can try to look into what might be going on.