All you have to do to get my attention is tag me (use @ before my user name).
For example I see on the COD site there is now a NAMNEST and a HRRR15, which I don't think were there last year. How/when should the HRRR15 be used compared to the "regular" HRRR?
I believe HRRR15 is simply the regular HRRR run with output every 15 minutes.
This is correct. And actually, this has been going on for about as long as the HRRR has been around. It's just that, in past years, the 15-minute output was only shown on the ESRL site.
The NAM NEST refers to the convection-allowing nest within the NAM (parent is still 12 km grid spacing). This has been the 4-km NAM you've probably used many times in recent years. Earlier this spring it was upgraded. The data assimilation scheme and some other timing parameters were improved (including the microphysics), and the grid spacing was decreased to 3 km.
Other model changes you may find useful are generally on the ESRL site: rapidrefresh.noaa.gov/hrrr. Two particularly fascinating and helpful HRRR products include the time-lagged ensemble HRRR (HRRTLE:
https://rapidrefresh.noaa.gov/hrrr/hrrrtle/), and the HRRR ensemble (HRRRE:
https://rapidrefresh.noaa.gov/hrrr/HRRRE/). The HRRTLE simply takes successively initialized runs of the operational HRRR and combines them as if they were an ensemble by creating probabilistic fields for separate runs all valid at the same time. For example, you could get a probability of composite reflectiivty exceeding 40 dBZ in the HRRRTLE valid at 00Z by using a...
12-hour forecast from the 12Z HRRR, an
11-hour forecast from the 13Z HRRR, a
10-hour forecast from the 14Z HRRR, a
...,
...,
..., and a
1-hour forecast from the 23Z HRRR
since each field would be valid at 00Z. It's a poor man's way of creating an ensemble from deterministic forecasts, and in general TLEs don't verify as well as their full blown ensemble bretheren, but they can still be useful, especially in smaller numbers (when the temporal distance between TLE members is not so great), and they essentially add zero cost computationally.
The HRRRE is going to become an operational entity eventually. I studied convection-allowing ensembles during my PhD, and I can tell you they are revolutionary. It's probably mostly due to technological constraints and bureaucracy that CAM ensembles have not already become operational. And no one else in the world is really doing this...this is a unique venture by the US weather enterprise. I would guess most folks will need some time to get used to looking at probabilistic products from convection-allowing ensembles, but the NCAR ensemble is such an ensemble that has already been available for a season or two now. The HRRRE is very similar to the NCAR ensemble. I do not know where the ensemble diversity is coming from, as ESRL has not released that information to me, but I suspect initial condition uncertainty from an ensemble Kalman filter is the method, just as in the NCAR ensemble.
Don't forget that the MPAS is also running for the 2017 Hazardous Weather Testbed spring experiment, so from now through early June. You can find graphics here:
http://www2.mmm.ucar.edu/imagearchive/mpas/images.php?p=hwt2017&r=central&i=1&f=rain24h&d=2017043000
And there are other experimental convection allowing ensembles out there. NSSL still runs theirs, and my current group at OU is also running an experimental storm scale ensemble this year. There are graphics, but they are pretty crude and I'll see if they improve before releasing the URL to that.
