high risk

also worth noting that the GFS is considerably warmer at the start of Sunday than most other guidance.

The GFS usually underdoes the amount of sleet and freezing rain in the transition zone. It mostly shows rain or snow here in the evening hours, and I don't buy that at all.

This is true, but when the GFS and NAM have large synoptic differences at longer ranges, it's not often that the NAM will be correct. It's simply a lot to ask of *any* regional model to nail synoptic details beyond day 2. One the snyoptics align, one should absolutely use the NAM (and preferably the NAM Nest) for temperatures and important mesoscale details.

Sort of. I said that the GFS was upgraded early in 2021, while the GEFS was not. So we can therefore not treat the ops GFS as a true control run for the GEFS. As for whether to buy the inland GFS track or the more coastal track of the GEFS mean, I have no idea.

NAM has low teens here Saturday night with dew points around 0. GFS keeps a slight northerly component to the sfc winds ahead of the sfc low on Sunday and has a known bias for scouring out low-level cold air too quickly. I don't disagree at all that we'll be able to torch above the ground, taking the current track verbatim, but the surface cold air is not going to give up easily at all.

I can't argue with this. It's certainly possible that higher resolution is the key to a further west solution. I'm just trying to point out that resolution isn't the only significant difference between the GFS and GEFS. Really wish that the GFS and GEFS could be upgraded simultaneously.....

This is how it *should* be set up, but it's unfortunately not the case right now with the GEFS. The operational GFS underwent a significant upgrade this past March, including changes to the PBL scheme, radiation, and data assimilation. Those changes were not made to the GEFS. So, even the GEFS control run cannot be considered only a lower-resolution version of the operational GFS. Does that fact that the GEFS members and GFS have larger differences than just resolution explain why no GEFS member looks like the GFS solution? Impossible to say....

The ensemble members are run at lower resolution than the operational GFS. And likely more importantly, the GFS had an upgrade early in 2021 that has not been applied to the GEFS.

To be clear, this isn't a big deal in any way, but it's not a TT issue. They're using the p-type information directly out of the GFS (and do that for all of the models) and not are computing it on their own. In lighter precip, the GFS can sometimes overmix the lower levels and warm them up a bit too much, and that appears to be what is happening in that example.

I've been bullish on this event, but I can't argue with what everyone else has observed in the guidance this afternoon and evening: while I think it's unwise to completely dismiss the threat further south, the highest chances of seeing precip tomorrow morning are going to be well north of DC.

The big takeaway from the morning model runs is that a lot of the guidance delays the arrival of the precip and allows an opportunity to warm up before arrival. That said: 1) The HRRR still has a lead band of precip (out ahead of the main frontal band) moving through the northern half of the area around 12-13Z. As said previously, even a few hundredths would create huge problems with temperatures likely still in the upper 20s then 2) Even if you consider a solution like the NAM Nest which delays precip until 16Z or so, it's about 32 or 33 degrees when it arrives. Given how cold surfaces will be after a very cold day and night, you can absolutely still get icing on roads and sidewalks in that scenario. The biggest threat is certainly for areas a good distance north and northwest of DC, but I do believe that there is a still a decent threat for everyone along and north of I-66 in VA and Rt 50 in MD, and a slight threat likely extends a little bit south of there if the HRRR solution is correct.

Spot on. It's pretty clearly going to be well down into the 20s Saturday night before the temperature rises later at late. All road surfaces will be super cold, so just getting to 32 or 33 Sunday morning won't end the icing threat. It's also worth noting that the NAM Nest appears to be doing its usual thing of being slow with precip arrival, and the earlier start shown by some of the other guidance certainly suggests a fairly disruptive (and likely a high-impact) icing event. The only thing keeping it from being an extreme high-impact event will be that it's on a Sunday morning.

Correct. It's initialized with the previous GFS cycle and has a fairly similar overall configuration (with higher resolution).

One quick note: maps like these for any models except RAP/HRRR are completely derived by the site that displays them, and they involve a lot of estimating and interpolating. Only the RAP/HRRR have explicit totals by type. That said, most of the guidance now shows at most a very brief period of rain at the start.

one hell of a band in the NAM3. Looks like a freakin' squall line.... edit: Ninja'ed by @Deck Pic

While I don't get too worked up over HRRR snow details 36 hours out, the 18z forecast is consistent with the idea that has been discussed about good amounts to the west with the shortwave dynamics and then a secondary max to the southeast as the coastal low spins up. Unpleasant screw zone in between, although its totals there would be acceptable, at least to me. And the band of heavier totals with the coastal low isn't far southeast of I-95 at all....

True, although temps should rise quickly later tonight as the southerly winds and clouds increase. But you make a good point that temps here in the early evening have plummeted, so all surfaces will be really cold in the early morning, and any very light rain will freeze on the cold surfaces even if the air temp warms to near 32.

I think that what is happening is that the MAG site where you got this image displays "snow depth change" and not "positive snow depth change" like Tropical Tidbits does. The distinction is that the snow depth change is going to tally snow on the ground today that melts before Thursday night as a negative. So, if an area has 4" melt and then gets 4" of new snow, that map will show a 0. (Even negative values are possible, but that map starts at 0.) The positive snow depth map on Tropical Tidbits looks way different and shows several inches of new snow across VA/MD/DE.

Sneaky freezing drizzle potential early Wednesday? Several CAMs show some spotty light precip early Wednesday generally on the east side of the Potomac. Some of it is shown as rain, but some of it is freezing rain. Air temperatures are progged to be right around freezing, but the surface temperatures will likely be colder than that.

I-95 in central Virginia still a complete mess. People stuck in their cars for many hours and running out of gas now.

Superstorm '22 is a lock. Got it!

5.8" in southern Howard County. Feeling fortunate to have stayed just south of the northwestern edge of the snow for several hours.

This traffic map does an incredible job showing which areas have been crushed and where the northern edge of the snow has resided.

Will be retired. NAM (parent and nest), RAP/HRRR, and Hi-Res Windows will be turned off when the RRFS is implemented (or very soon after).

Sort of. Yes, the NAM (and RAP/HRRR and Hi-Res Windows) will be subsumed by a hi-res, hourly FV3 ensemble (called the RRFS), but the configuration of that hi-res FV3 is not what the configuration of the RRFS members will be (different physics, different initialization....). Think of this run that you showed as sort of a hi-res GFS for now.