high risk

Not true: HRRR and the three HiResWindows are all 3km. That said, I don't trust any of those for thermal profiles.

The reason that the NAM and Hires Windows are being retired is because it's impossible to maintain a model suite with so many different cores. That said, I need to figure out why I'm spending time defending the NAM....

The fact that it's one of the driest solutions is absolutely odd, and I don't mind treating it as a massive outlier. But I'm not sure that missing how significant the initial thump is will matter with how quickly we changeover, and I also don't think that it makes sense to compare radar to model forecast 400 miles away to determine what will happen here. Anyhow, you'll change over way later than we will no matter what......

The blanket "the NAM sucks" stuff is wild to me when the strength of the NAM Nest is nailing thermal profiles inside of Day 2. Maybe it's going to be too fast with the changeover, but complete dismissing of it makes no sense.

Agree fully. The wet content of what's on the ground followed by the deep freeze means that it will take a long time to get sidewalks and side streets back in order

So, the problem is that the NAM Nest is part of the HREF....... That means, though, that the other members (HRRR and HiRes Windows) are much snowier. They're all less icy than the RRFS and ECMWF.

The FV3 RRFS is also going bonkers with the freezing rain. I *think* it's way, way overdone and favor the sleet bomb idea of the NAM, but this will have to watched closely tomorrow.

There are plenty of valid arguments against the NAM, but making that point by showing changes across cycles of an forecast of what the lowest level radar reflectivity will look like at one particular moment, and then comparing that to a 6h-averaged precip rate from a different model doesn't make your case.

God, I really hope you meant "rimed".

Love that you showed the para! Friendly reminder that the NBM winter fields update at 01, 07, 13, and 19Z, so instead of showing the 12Z, it's best to wait one hour and use a version that incorporates more recent guidance.

The arctic front doesn't come through until Friday evening, but when it does, the freeze is on. Multiple models show upper 30s at 4pm falling to low teens by 1am.

Yes, but we're still going to get 2 years of the FV3-based RRFS in operations.

Big warm layers can easily reside in between those two levels. Some of the forecast soundings I have seen for this event show that exact scenario. The NAM precip type code is rock solid, based on the forecasted temperature profile.

Are you looking at the actual sounding, or just the temperatures at those levels??

They differ in more than just resolution, so they effectively are two different models (even though they have obvious commonalities). As you said, though, the 3 km is a far superior model, except for when it deepens hurricanes down to 850 mb.

The Euro and its ensemble contribute heavily to the NBM, which is the starting point for these forecasts. And I disagree with you: several inches of snow followed by a ton of sleet would be a very big deal at these temperatures.

No, no, sorry. I clearly wasn't nearly as clever as I thought with that post. I am expecting that the happy hour NAM Nest cycle right before the storm will go bonkers.

Apologies for bursting your bubble, but @wxmeddler is correct. There is a 1.33 km relocatable NAM Nest used for fire weather purposes that can be used for other types of events during non-fire-weather season. There used to be graphics online, but I don't think that there are anymore. And I think they've stopped running it for events like these.

The AI ensembles tend to me under dispersive even in the medium range, so it's really no surprise that you're seeing very little spread as we enter the short range.

I'm sure that this has been hashed out, but I want to emphasize that we're in that time range at which the differences in NAM precip type could simply be driven by synoptic errors, common for a regional model at this range, but they can't be discounted completely due to the NAM being the absolute best at precip types *once it has nailed down the synoptics*. I also would never fully trust the GFS precip type maps *even when it has nailed down the synoptic*. It often underplays the coverage of IP/ZR. Oh, and a friendly reminder that the model cycles all start at the same time every day, and they don't ever "stop". (They can crash once or twice per year, but it's so rare.) There can easily be dissemination issues, usually driven by way too many hits to a server during exciting events, which delay the arrival of everyone's favorite online maps.

Curious 18z GFS evolution for the Thursday wave

I could tell you what its Critical Success Index value for 24h snowfall at Day 3 is, but what would you do with that number? For the most part, it’s as good as the inputs it uses, but the operational version right now is running high, because it’s bias correcting QPF upward based on a very limited sample of recent significant QPF events in the very dry Mid-Atlantic.

So, you don’t actually know that. The winter part of the NBM at this range only includes the NAM, GFS and its ensembles, and the Euro and its ensembles. The para includes some of the Canadian system as well.

Just some speculation that this cycle will NAM the crap out of us.

If you liked that GFS run, you're going to LOVE the 18Z NAM Nest on Saturday.