high risk

Goddamnit, y'all. I have literally worked as a RAP/HRRR developer, and I would still use the JMA or NAVGEM before I used the RAP/HRRR system for snow amounts.

minor addendum: it's still very much experimental and is not scheduled for implementation now until 2025

It's not the kiss of death, but major red flags should always be raised with regards to accepting a 12 km NAM solution when its 3 km nest shows something different.

It always warms my heart to see the snow depth maps posted here, but this is not the type of event for which they do well. They limit accumulation when surface temps and soil are warm, and they'll never capture the ability of heavy rates to overcome marginal thermodynamics. I'd probably either average the 10:1 and snow depth products or mentally adjust the 10:1 maps downward. Kuchera maps might be good too - I hate how generous they are with colder temps, but they seem to properly limit accumulations when the temperature is marginal.

bright banding??

I'll take a few of those convective bursts of snow at the end of the event, please.

I don't think that this is a storm for which the super generous Kuchera ratios will work well. It's not super cold, and lift does not appear to be maximized in the DGZ. With cold ground and temps a few degrees below freezing, this might be one of those cases in which the 10:1 maps actually work pretty well.

One other thing, and perhaps I missed discussion of this: a few CAMs and the GFS show a period of light snow here Thursday evening with the lead wave.

Surface temperatures are super marginal, so it might have some difficulty sticking on roads once we get past 8am or so. That said, the road surfaces should be plenty cold going into the event, so rates might still get the job done.

Nice! I don't have that sort of precision on my station, but it is showing 6 this morning.

The 12 km parent and 3 km nest have different diffusion, so differences in 500 evolution can and do happen.

The northward shift in that run also makes us a bit warmer at the surface, which reduces the ability of the snow to stick during the daytime hours, although surfaces at least will overall be cold.

correction: The actual model run will run on time as it does 99% of the time. The TT graphics processing, run on a 1984 Commodore 64, will be ready by bedtime.

Depends whether they can clear all of the side streets and sidewalks today and give them time to dry out this afternoon before the temps plummet later. Anything that isn't dry will freeze hard tonight, so I would say that there could be delays or closings again tomorrow.

4" here in southern Howard County

Traffic maps across the region look far worse than I expected to see during the daylight hours on a holiday. Likely to look much worse when the sun sets and rates increase.

I agree 100%, and the NAM Nest is that model. It's sometimes a bit too slow in eroding the cold air, but that's better than the GFS and HRRR which wipe it out way too fast.

The NAM struggles with synoptics for sure, but once it figures those out, it's pretty damn good with winter event details. And to be clear, I'm focused on the 3 km NAM Nest, as the 12 km parent is meh. There is no model better than the NAM Nest for resolving cold air damming, and while it sometimes runs too cold in the low levels when sorting out precip details, it's also really good at capturing warm layers that screw up a snow profile. The RRFS has yet to prove competence in these areas, and the HRRR isn't great either, so we will miss the NAM on some days when it's gone.

Correct. It's really tough to generate meaningful spread from a single model core in the short range. And the struggles of the FV3 core aren't helping things.

18Z NAM (both 3 and 12 km) looks better because the snow gets more impressive earlier, but it also slows things down much faster Monday night, and the highest totals are northwest of the I-95 corridor

Agreed. While this could still potentially move further in the wrong direction (and to be clear, I think we'd be good for a while regardless), the NAM Nest is fine for snow for most of the area through the prime portion of the event. In fact, the "Ferrier" accumulation, which is completely driven by the model microphysics, has a nice accumulation for many, indicating that the model has the hydrometeors falling as snow for plenty of the key hours, even if it turns to some light freezing rain Tuesday morning:

The RRFS (Rapid Refresh Forecast System, a 3 km ensemble, is being developed. It hasn't been easy, because it has to be at least as good as the HREF to replace it, and it's far easier to get meaningful spread from a a system with multiple different models (HREF) than a system with just a single model core.

That's the key point. It's badly out of date and wasn't a great system when it was "in date".

YES

The main focus is on us getting NAM'ed, but I'll just make a quick point that this isn't really true. The SREF is a mix of 13 ARW and 13 NMMB runs using a mix of RAP, NAM, and GFS initial conditions (and multiple physics). Yes, half of the members have the same model core as the NAM, and roughly 1/3 use NAM initial conditions, but it's not a NAM ensemble.