OceanStWx

All that confluence over the Maritimes would definitely support a nice high in a favorable CAD position. It will be interesting that it all starts with a backdoor more or less.

The GFS with 32 shoved back into Mass is a huge red flag. That model loves to drive a warm front to YUL.

Not uncommon for winter events to all be under 10,000 ft clouds. And you're like 7,000 ft on our beam.

Beautiful band of snow, it's just not sticking around.

ASOS is really good with visibility, but vis is only a proxy to snow rates. I've definitely seen 1 1/2SM -SN that was 1"/hr because it was perfect aggregates. Those events are exceptions not the rule though.

Google maps also allows distance measurements now too. It's funny that I remember one of the first things I learned in Ithaca was that it was a half mile from Bradfield to the Clock Tower. And if you lost sight of the American flag on Barton it was 1/4SM +SN.

Trading in my ICAO Ks for Ts. I'll just pack 7 banana hammocks and be done with it.

Well timed high will eff up your record highs every time. Only ice I hope to see this weekend will be in my rum drinks.

Right, one run of the Euro with a snow map isn't really a forecast to follow. The EPS did have higher probabilities going back a few runs, but taking > 6" probs from 30% to 0 shouldn't be considered that wild a swing. Even a 70% chance of > 3" is a 1/3 chance you get less.

It's not a black and white. There is a chance of 1-3" but it's not a good chance. Straight EPS is like 30% for 1+ and 10% for 3+. I think my forecast at the moment would be sub-advisory, and if we have to work up from there we can.

Not a lot of hope there either. Not a lot of hits there for SNE, and really not a lot of QPF for NNE.

I know we joke, but that's obviously the perception. Always verify a cutter never a coastal. But I think it proves the point too. The high res guidance can be variable when there is a small window to thread, but you don't notice the shifts when 500 miles can't even save the event for you. But ask someone outside ORD about that same cutter and they will be talking about how much snow changes run to run. I mean MSLP isn't great either, but it's also not exactly sensible weather. It does work both ways at times, but generally MSLP is really a response to forcing from aloft. If that's the case you can see how there are just more places for a model to fail before getting to that level.

What is a sensible weather score? Inches of snow in your backyard? Because modeled snow is going to be a terrible score. QPF will be bad as well.

I can't speak to the specifics but I'm pretty sure the differences are statically significant, despite the AC difference being a small number. And overall you're right, that 500 heights don't always translate to sensible weather. QPF is not a good metric for models. It's better than it used to be, but still one of the worst variables to predict. Typically models are better at finding a window QPF will occur, but bad at finding the right amount of QPF. Think of it this way. Whether you get a wound up coastal with 1" QPF or a fropa with 0.01" in snow showers the model still had the trough for that period. High AC score, low QPF score.

I think it's a impossibility to expect a relatively high resolution model (13 km) to lock in a solution for days on end or even run to run. Increased resolution can lead to increased variability in models. It's why we preach focusing on the larger scale features (i.e. jet, heights, etc), they are more likely to be stable than smaller scale features (i.e. QPF -> snow).

Demand. If you can offer higher resolution, in my backyard type forecasts that's what people are going to want. And we don't have the computing power to run two versions. And running a coarse model and downscaling it to 13 km isn't really helping improve things either. There is an argument that we have too much data at our fingertips, or maybe that we have too much data that we don't understand fully at our fingertips. A 20 km GFS run is not the same as a 90 km GFS run, and forecasters need to change their thinking about how to use it to improve the forecast. Rip and reading may have worked for a broad brushed forecast at 90 km, but rip and reading at 20 km can make you look pretty bad at times.

I mean when you make a image out of 13 km f-gen it looks like a 2000 mile long series of mountain waves. I can glean no information out of that. But a 90 km GFS field showed clear regions of f-gen that could be used to diagnose the best area for potential banding. That's why I'm kind of okay with the Euro data we get in AWIPS being 80 km still, despite the model being 13 km res.

Ooo, that's going to be close. Downtown Harwich is like 70.06W but extends east to about 70.00W, and the ACK ferry landing is like 70.09W but Sankaty Head is like 69.97W.

There are a couple reasons why this isn't the case. One is that you have to find a way to manage convection. You either explicitly resolve it (convection-allowing) or parameterize it. Either way you are making assumptions that you are either parameterizing it correctly or correctly modeling its location and strength. Two is that as you improve your resolution you also sharpen gradients and increase the max/min values of features. This can dramatically affect the forecasts farther and farther out in time. You can imagine that an 80 km Euro on day 4 having broad QPF amounts would show a potential event for everyone, but a 13 km Euro 4 days out may show a sharp northern edge and convince NNE that they are going to get nothing. What if the model trends north then? It will look like a bust, whereas years ago it wouldn't have seemed that far off.

Still has pretty decent probabilities for most of the state for > 3". But if you're looking for a stripe of warning criteria snow I wouldn't be painting it through SNE or the NNE mountains yet.

Still quite a bit of variability in the EPS members. Some skunk SNE, others skunk NNE. I like the location of EPS probabilities of > 6" for the highest potential corridor right now. I would say because there is a bit of a bi-modal distribution of north or south members, the actual probability of > 6" is likely higher where it snows, it's just getting smoothed out here.

Meh, it's one model cycle of many left to go. Ensembles still look like there's plenty of potential. I think it's increasingly clear that forecasters need to stop relying so much on deterministic run to run variability. With resolution down to 13 km in most instance you are just going to get far too much variability given the detail they show. 10-15 years ago deterministic runs were 80 km and features were much more broad and could be applied in the same way ensemble features are now.

Is he also our farthest east poster? Narrowly beats James and ACKwaves

What an amazing idea honestly. "Let's take the NAM, but run it with GFS boundary conditions. And let's start it at 84 hours." What could possibly go wrong.