high risk

As you noted, the core was changed, but most of the actual physics (except for the microphysics) were retained. I think that the PBL scheme plays a role in these warm layer cases.

It's the "new" GFS, but it has the old physics, so I suspect that its inability to capture mid-level warm layers remains.

The regular members of this subforum are perfectly capable of making uninformed posts, without the need for any outside help, but thanks.

Agreed on all counts. The GFS typically shows way too small of a small transition zone between rain and snow. I do think that eastern Howard may get a decent sleet mess out of this.

It's worth noting that in the CAMs (especially the NAM nest) that turn central and north-central MD over to sleet for a long duration, the warm layer is up around the 750 or 700 mb level. The GFS is typically very poor at capturing those warm layers if they're well above, say, the 850 level. PSU made a reference earlier to the March 2017 storm in which the GFS missed forecasting most of the sleet because it kept things much colder aloft. It isn't the same GFS now, but I would not be surprised if this weakness holds.

the front end thump for those of us in Howard/Montgomery isn't nearly as good on the NAM nest

100% true for this area. Well, ok, 90% true.

exactly. no one should ever post a 10:1 map when the precip type images clearly show a ton of sleet. The NCEP models tally snow and sleet together and put out a liquid total; a 10:1 ratio applied to that has no chance to be even in the ballpark.

You sure about that? Both the 12 and 3 km versions seems to have snow in your area by 2 or 3pm. (I'm commenting on the precip timing, as it does look like you will go sleet.)

Right. Those maps don't account for actual ability to build on surfaces. They just look for rain falling with a sfc temperature at or below 32. You're spot on that without it being very cold today/tonight (i.e. mid to upper 20s), rain falling at 31 or 32 isn't likely to be a major deal. I do, however, think that parts of this area will get a big 'ole sleet bomb out of this.

Ok, but the HRRR has been running stupidly cold for winter precip events at longer ranges. For today's cold rain, a couple of the longer range runs had 4-6" through the DC area.

appears to be a respectable front-end thump for those of us north of the DC Beltway Wednesday afternoon. Temperature profiles have no margin for error, but verbatim, it would be cranking here:

Fair enough. It could be wrong; I'm just reacting to the folks who said that the NAM showing an inland low track proved that it should be immediately euthanized.

Since it seems to be the afternoon for apologies on this board, perhaps one needs to be made to the NAM......

FWIW, and I am not saying that it's right, the CMC evolution/track is more similar to the 00z NAM than to the 00z GFS....

Chill out. The NAM was frozen with its 2017 upgrade (the 3 km component has been running for way, way longer) and is slated for retirement in the 2023-2024 time frame. It's also well known that the 3 km nest is a far superior model to its 12 km parent, although it too certainly has limitations.

12z CMC definitely disappoints after seeing the 12z GFS

Absolutely. I'm a huge fan of the positive snow depth change as an alternative to the 10:1 maps, but they absolutely are driven by soil type and especially soil temperature; they tend to be a bit underdone early or late in the season when the ground temps have been warm for a few days.

not really. Yes, the NAM (and other regional models) will be replaced by a system that is run hourly (like the HRRR), but the new system will be FV3-based. regarding the Monday event, the HRRR is the craziest for central MD and DC, but several other hi-res models show a change to snow on the backside of the system, at least across central MD.

I'm not sure about the EPS, but the GEFS members are run at 25 km resolution, while the ops GFS is 13 km. So yes, that could be making a difference.

Definitely much more dispersive than the previous version of the GEFS.

The previous version of the GEFS was extremely underdispersive, but the new GEFSv12 has way more spread. That was one of the strengths of GEFSv12 noted in the summary presentations on the evaluation web page: https://www.emc.ncep.noaa.gov/users/meg/gefsv12/ (It also helps that 10 additional members were added.) If most members agree something at day 5, that's a much bigger deal in the new version.

This is what happened on 11/15/18. The NAM was one of the coldest models for that event too and was largely dismissed, but it had the right idea of big rates overcoming a very marginal air mass.

The parallel GFS had statistically significant synoptic range improvement over the ops GFS in the medium range over the full course of the retrospective runs. But that certainly doesn't preclude it doing worse on any single event. I wish we were able to run it more than once per day right now.....

Honestly, those details haven't been sorted out yet. It's not an output in the RAP/HRRR (which have actual snow accumulation as an output), although their SLR methodology is documented. The plan is to carry over the RAP/HRRR products into the RRFS (replacing all of the regional models in 2023) and build it into the 2024 GFS/GEFS upgrade.