December 19-20 Talking Points - Part 3

[A-L-E-X]

Come to think of it-- my memory might have been right.  Take a look at the scores through 2006-- 12z GFS is significantly ahead of the other runs.

[Sundog]

Looks like convective feedback issues on the GFS this run? There's a mysterious blob of convection SE of the low and a vort that might be robbing our storm on this run.

In any case, I haven't really lost any enthusiasm since seeing the Euro today. 0z tonight and 12z tomorrow will be very crucial. If the models hang on for 12z tomorrow we should be pretty much a lock for a substantial storm. If they begin to fall east again tonight and tomorrow, it's almost certainly over as all the relevant data will be sampled by then. The Euro must have noticed something to make it and its ensembles into that crazy solution today (after starting to trend last night), since it was flat as a pancake for numerous runs prior. It's probably unlikely it becomes that deep, but a storm like this has to be watched very carefully because of the chance for a stall due to the blocked up pattern. It also won't take much at all in terms of energy to spin a storm up here. The Polar vortex is in a favorable position and backing away as the storm approaches. It's really all on the southern stream.

I also remember distinctly the NAM and GFS playing these convective feedback games before the 12/19 storm last year.

The GFS and NAM gave us cirrus about 78 hours before the storm. And it wasn't even close. They were pushing the storm off North Carolina Virginia still.

[A-L-E-X]

Looks like convective feedback issues on the GFS this run? There's a mysterious blob of convection SE of the low and a vort that might be robbing our storm on this run.

In any case, I haven't really lost any enthusiasm since seeing the Euro today. 0z tonight and 12z tomorrow will be very crucial. If the models hang on for 12z tomorrow we should be pretty much a lock for a substantial storm. If they begin to fall east again tonight and tomorrow, it's almost certainly over as all the relevant data will be sampled by then. The Euro must have noticed something to make it and its ensembles into that crazy solution today (after starting to trend last night), since it was flat as a pancake for numerous runs prior. It's probably unlikely it becomes that deep, but a storm like this has to be watched very carefully because of the chance for a stall due to the blocked up pattern. It also won't take much at all in terms of energy to spin a storm up here. The Polar vortex is in a favorable position and backing away as the storm approaches. It's really all on the southern stream.

I also remember distinctly the NAM and GFS playing these convective feedback games before the 12/19 storm last year.

We have like 3 noreasters on this forum and its getting confusing trying to tell them apart lol.  The one with no numbers in their name said this:

18z GFS is having convective feedback issues...take a look at how it develops 2 lows, each centered around a ball of convection... According to the 500mb it was on the right path in the begining by developing a broad surface low near FLA, then out of nowhere the low jumps/ revdevelops due east....

-Nor

[nzucker]

Lol I cant believe youre serious about going by 18z modeling information. The fact is the GFS and NAM are both inferior to the products put out in Europe. I just wish we used the 4DVAR data assimilation scheme.

I'm not just saying that because of the recently released 18z runs....I think the majority of guidance for the last four days has had the storm being mostly offshore, with only a few runs hitting NYC Metro and slightly more delivering for SNE. Although the trends were very good on the ECM lately, only one out of the last six ECM runs has had any significant snow getting to NYC, and the GFS has also been waffling about how late the northern stream gets injected into the trough so not that many of the GFS/NAM runs have been a hit here either. The 12z UKMET is also a miss and that's the 2nd best model behind the ECM. The fact is we don't have a potent southern stream shortwave so we have to rely on the northern stream to get there in time to amplify the trough before it's already shifted too far off the coast and washed out the baroclinic zone with a much weaker system. It's a high-risk/high-reward situation: if the storm phases properly, we'd be looking at a MECS/HECS almost certainly, but the whole phase is just very difficult to time and very fragile. People were acting like the storm was a sure thing when they saw the 12z ECM and I am just urging caution.

[dtk]

18z model runs in the past have been very unreliable..I have seen them lose storms and have it come back at 00z...I have also seen them make bombs and lose it 6 hours later..really the next big cycle is the 0z runs tonight

I want to reply to the "18z suite is crap since they use old data" nonsense (mostly from the previous thread). Think of it like this, if an 18z model initialization used NO data/observations at all, it would look identical to the previous cycle's run (i.e. no observations were used, a 120h forecast from 18z would look identical to a 126h forecast from 12z). This is explicitly true for global models with are cycled, and only somewhat true for the NAM and other regional models that use partial cycling. Additionally, and 18z regional model run will have the advantage of using a new global forecast for boundary conditions.

In a time mean (average) sense, the 18z run will always be better than a 12z run since it DOES use updated observations, improves the analysis, and therefor improves the forecast. For an individual case (especially when looking at something like downstream energy transfers, phasing, coastal cyclogenesis), very small changes to the initial conditions can have extreme impacts on the model simulation (resulting from large error growth rates for the storm of interest, nonlinearity, etc.)....hence all of the model flip-flopping the last few days. So even though the updated forecast will be better in a global/verification sense, it could potentially make some component of that forecast actually worse (i.e. having a storm develop too far off shore, too quickly, what have you).

Lastly, even though we don't have much in terms of raobs (which are awesome obs, so we intentionally draw closely to them) at 18z, we have a ton of other observations (satellite radiances and derived products, aircraft data, wind profilers, surface observations, satellite cloud track winds, GPS radio occultation, and so on).

[A-L-E-X]

The GFS and NAM gave us cirrus about 78 hours before the storm. And it wasn't even close. They were pushing the storm off North Carolina Virginia still.

Euro locked in first with that storm, but no model had a good handle on the final track of that storm until 48 hrs before the event started.  That usually seems to be the case lol.

[A-L-E-X]

I want to reply to the "18z suite is crap since they use old data" nonsense (mostly from the previous thread).  Think of it like this, if an 18z model initialization used NO data/observations at all, it would look identical to the previous cycle's run (i.e. no observations were used, a 120h forecast from 18z would look identical to a 126h forecast from 12z).  This is explicitly true for global models with are cycled, and only somewhat true for the NAM and other regional models that use partial cycling.  Additionally, and 18z regional model run will have the advantage of using a new global forecast for boundary conditions.

In a time mean (average) sense, the 18z run will always be better than a 12z run since it DOES use updated observations, improves the analysis, and therefor improves the forecast.  For an individual case (especially when looking at something like downstream energy transfers, phasing, coastal cyclogenesis), very small changes to the initial conditions can have extreme impacts on the model simulation (resulting from large error growth rates for the storm of interest, nonlinearity, etc.)....hence all of the model flip-flopping the last few days.  So even though the updated forecast will be better in a global/verification sense, it could potentially make some component of that forecast actually worse (i.e. having a storm develop too far off shore, too quickly, what have you).

Lastly, even though we don't have much in terms of raobs (which are awesome obs, so we intentionally draw closely to them) at 18z, we have a ton of other observations (satellite radiances and derived products, aircraft data, wind profilers, surface observations, satellite cloud track winds, GPS radio occultation, and so on).

This still doesnt change the fact that the GFS is an inferior product because it doesnt use the Euro's data assimilation scheme.  It's high time that was changed.

[jrodd321]

I still think this storm's going east and going to miss us. Very poor trends on the 18z GFS and NAM; I feel that the southern stream shortwave is going to run out too far ahead of the more potent northern stream in the zonal flow off the Pacific, and that the system won't get pulled back towards the coast in time.

Lmao just hilarious. 18z doesn't establish "trends" dude. Hahahhah

[Rainshadow]

Looks like convective feedback issues on the GFS this run? There's a mysterious blob of convection SE of the low and a vort that might be robbing our storm on this run.

In any case, I haven't really lost any enthusiasm since seeing the Euro today. 0z tonight and 12z tomorrow will be very crucial. If the models hang on for 12z tomorrow we should be pretty much a lock for a substantial storm. If they begin to fall east again tonight and tomorrow, it's almost certainly over as all the relevant data will be sampled by then. The Euro must have noticed something to make it and its ensembles into that crazy solution today (after starting to trend last night), since it was flat as a pancake for numerous runs prior. It's probably unlikely it becomes that deep, but a storm like this has to be watched very carefully because of the chance for a stall due to the blocked up pattern. It also won't take much at all in terms of energy to spin a storm up here. The Polar vortex is in a favorable position and backing away as the storm approaches. It's really all on the southern stream.

I also remember distinctly the NAM and GFS playing these convective feedback games before the 12/19 storm last year.

I'm not saying the solution is right, actually philosophically I know its wrong, but the trof isn't as sharp or goes negative tilt as quickly as it did on the 12z run. Any convective feedback issues is just adding extra salt.

[A-L-E-X]

I'm not just saying that because of the recently released 18z runs....I think the majority of guidance for the last four days has had the storm being mostly offshore, with only a few runs hitting NYC Metro and slightly more delivering for SNE. Although the trends were very good on the ECM lately, only one out of the last six ECM runs has had any significant snow getting to NYC, and the GFS has also been waffling about how late the northern stream gets injected into the trough so not that many of the GFS/NAM runs have been a hit here either. The 12z UKMET is also a miss and that's the 2nd best model behind the ECM. The fact is we don't have a potent southern stream shortwave so we have to rely on the northern stream to get there in time to amplify the trough before it's already shifted too far off the coast and washed out the baroclinic zone with a much weaker system. It's a high-risk/high-reward situation: if the storm phases properly, we'd be looking at a MECS/HECS almost certainly, but the whole phase is just very difficult to time and very fragile. People were acting like the storm was a sure thing when they saw the 12z ECM and I am just urging caution.

Yeah, this forecast has a high margin for error, that's why all the options are still equally likely at this point.  You cant start locking anything in until 48 hours before its supposed to start.  That was even the case at this time last year.

[Rainshadow]

This still doesnt change the fact that the GFS is an inferior product because it doesnt use the Euro's data assimilation scheme. It's high time that was changed.

Send money.

[tmagan]

Kind of reminds me of last December's storm (was it?) where thirty-six hours in advance with nearly every model indicating a significant snowstorm, the 06Z NAM came in totally dry for the area, and 5 minutes later KOKX issues a blizzard watch. KOKX turned out to be right.

[Zir0b]

FWIW, the 18z UKMET looks pretty at 72 hours.

[A-L-E-X]

Send money.  

Lol I feel like we live in a third world country sometimes.  Not to go on a political rant here, but we arent as great as we once used to be-- that much is really obvious (very similar to what happened to the Roman Empire actually.)  There's little signs of it everywhere.  I would gladly pay more in taxes to get this stuff fixed.  

[A-L-E-X]

Kind of reminds me of last December's storm (was it?) where thirty-six hours in advance with nearly every model indicating a significant snowstorm, the 06Z NAM came in totally dry for the area, and 5 minutes later KOKX issues a blizzard watch. KOKX turned out to be right.

Lol yea-- that storm was supposed to miss us to the south.  2/6 took care of that though.

[Rainshadow]

Lol I feel like we live in a third world country sometimes. Not to go on a political rant here, but we arent as great as we once used to be-- that much is really obvious (very similar to what happened to the Roman Empire actually.) There's little signs of it everywhere. I would gladly pay more in taxes to get this stuff fixed.

I think there are plans for the GFS to go 4DVAR in 2012 or 2013.

[jm1220]

The GFS and NAM gave us cirrus about 78 hours before the storm. And it wasn't even close. They were pushing the storm off North Carolina Virginia still.

I remember giving up on it less than 48 hours out, then I think the Euro was the first to suggest it making it all the way past Boston, and soon the others caught on. But I also remember a renegade GFS/NAM run in the 36 hours prior to the storm developing a weird strung out low south of the main storm and causing both to flounder around and go out to sea. They finally caught on afterwards and brought the storm in.

Without that weird feature over the Atlantic on the GFS, I think it would look a lot like 12z this run, with one consolidated storm. Overall I would say the favorability for a storm is about equal, with there being a little less energy but a little more ridging on the coast due to the PV retrograding further. The NAM does this all the time as well and has been doing strange things with the storm for the last couple of runs now.

But there's still considerable time for the models to start waffling again with all the new incoming data. 72 hours is a lot of time still, I agree there, especially for a fragile setup such as this. It's not as clear cut as a massive blob of moisture emanating from the Gulf like we often see in the Nino winters.

[blazess556]

FWIW, the 18z UKMET looks pretty at 72 hours.

Thanks to 4DVAR

[A-L-E-X]

I think there are plans for the GFS to go 4DVAR in 2012 or 2013.

Thank God.  Im thinking the Yankees should donate the money theyre not spending on Lee also

[dtk]

I think there are plans for the GFS to go 4DVAR in 2012 or 2013.

Not true, we don't have it ready yet. We're hoping to implement an hybrid ensemble-var system within the next 18 months (I'm currently the one working on the testing the prototype). If I had to venture a guess, some flavor of 4DVAR or hybrid ens-4DVAR would follow that, probably on our next supercomputer, in 2014.

[A-L-E-X]

You have no knowledge of the inner workings of NCEP, what it takes to implement and run 4DVAR, computational feasibility, or probably even how/why it's better....  If it was so easy to "just do 4DVAR", don't you think we'd already be doing it.

Also, whomever posted those cycle by cycle stats is correct....there is no statistically significant difference between any of the 4 runs for a given lead time.  There was a time (when we didn't use satellite data as well as we do now, didn't have GPSRO, etc.) where this wasn't the case....but those times are long gone.

But we do know that the Euro has somewhat higher verification scores than the other models, with the UKMET coming in second.  I know why its not being done-- its because of a lack of funds.  Dont talk down to me, your met degree doesnt give you extra knowledge of data assimilation schemes that other science fields use also-- namely oceanography (especially since its pretty apparent that meteorologists lack in scientific knowledge behind other branches of science; sorry had to be honest).  And Im shocked that you would defend a system that lags behind-- but maybe that's because you're a part of it.

Enough said.

[Rainshadow]

Not true, we don't have it ready yet. We're hoping to implement an hybrid ensemble-var system within the next 18 months (I'm currently the one working on the testing the prototype). If I had to venture a guess, some flavor of 4DVAR or hybrid ens-4DVAR would follow that, probably on our next supercomputer, in 2014.

Thank-you for the information.

[tombo82685]

does some one have a link to the 18z ukie, or can post the map at hr 72?

[Sundog]

You have no knowledge of the inner workings of NCEP, what it takes to implement and run 4DVAR, computational feasibility, or probably even how/why it's better.... If it was so easy to "just do 4DVAR", don't you think we'd already be doing it.

Also, whomever posted those cycle by cycle stats is correct....there is no statistically significant difference between any of the 4 runs for a given lead time. There was a time (when we didn't use satellite data as well as we do now, didn't have GPSRO, etc.) where this wasn't the case....but those times are long gone.

Serious question, what exactly is holding back the NWS from implementing the higher quality data assimilation scheme?

[A-L-E-X]

Not true, we don't have it ready yet.  We're hoping to implement an hybrid ensemble-var system within the next 18 months (I'm currently the one working on the testing the prototype).  If I had to venture a guess, some flavor of 4DVAR or hybrid ens-4DVAR would follow that, probably on our next supercomputer, in 2014.

It's about time-- literally.

[A-L-E-X]

Nice to see the 18z UKMET (with the better data assimilation scheme) coming in west.

[famartin]

And ya know, I know people are saying that the 6z and 18z of the GFS are just as accurate as the 0z and 12z but I distinctly remember seeing a verification score graphic a couple of years ago showing otherwise.

I do as well. I think I recall seeing that either the 6Z or 18Z was the least accurate... but I no longer recall which.

[ag3]

Back to the storm.

As Tombo asked,

Can someone post link or image of 18z UKIE?

Thanks in advance.

[Sundog]

I do as well. I think I recall seeing that either the 6Z or 18Z was the least accurate... but I no longer recall which.

I could be wrong, but I think I remember the 6z taking that honorable title.

[supermeh]

does some one have a link to the 18z ukie, or can post the map at hr 72?

It was on the general weather discussion section. Came in west of 12z. Also, RGEm, fwiw, looked good