December 19-20 Talking Points - Part 3

[earthlight]

Havent seen the Ukie H5 yet, but that would be a good sign if it was west.

And jeez...didn't mean to spark such a heated debate regarding NCEP models...just saying the Euro has a greater resolution and more thorough initialization, thus it's probably a good idea to give it's solution some extra weight, especially with it's ensemble mean agreeing.

[Sn0waddict]

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

http://vortex.plymouth.edu/grbcalc.html

[Chris L]

All right, I am back, after several hours of being away and having issues. And from now on, I won't be making silly or dumb posts like I used to....

And since I can only make a few limited post at a time. Here it is.

The 12z ECMWF is a very lovely sight for us in the New York Metro Area. Since it has trended westward and has classic features surrounding its rapid cyclogenesis evolution, as the PV, weak Southern Stream, and the super ampified Northern stream, merges as it develops a fairly significant cutoff low that makes everyone quite happy and quite flustered in every area of the nervous system, myself included, once I saw the evolution, it is absolutely ripe for a potential historic event for all of us that reside here. What I like most about the ECMWF is that it has the right AMOUNT of SPEED to merge and give us the solution, the right amount of everything, which makes it ideal for a heck of a cyclogenesis and at the benchmark at circa. 985 millbars surface low. Of course, the GGEM has trended west, and GFS (Sans 18z, but it is clearly having issues with the southern stream and the speed, which is why the sfc low is going eastward. I won't comment about the 18z NAM because, as of now, its useful for trends. Yes it did go the wrong way,but don't forget here, we are dealing with a fairly complex setup that offers us a series of hope and despair. Kind of like a Seinfeld episode where George and Kramer plan a scheme, a complex scheme of whatever it might be; and when it goes right....Win for George and Kramer.. Okay enough about that analog, now what the point is here, if we DO get the ECMWF's plan, we will likely be matching or exceeding the totals from the 2010 February storms, in a strong Nina mind you; impressive indeed. BUT... One has to keep their guard up in case of sudden shifts like Chris Farley and Richard Hayden in Tommy Boy movie. Agian probably a bad analog, but so be it. With the block being there, develops a cutoff low, potential is THERE for a slow moving storm before it leaves the East Coast. Oh yes, I know a side note, but ECMWF for next weekend looks very pretty developing a cutoff low JUST off shore...

Tonight: ECMWF is the tiebreaker...

Be well everyone.

[dtk]

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. 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.

First, apologies for the tone (I don't intend to talk down to anyone). I haven't been at NCEP all that long but I am in fact part of the global data assimilation team (and actually, I have worked with DA in oceanography as well....).

Just a side note, I'll never run away from the fact that the EC has better verification scores....you can't argue with the truth, but there is no way to prove it's because they run 4DVAR and simply don't have a superior forecast model (likely a combination of both).

Do I wish we were running 4DVAR? Of course I do.....but given what's happened I'm proud to say that we're getting way more out of our "crappy 3DVAR" system than any of us ever would have imagined. I'm amazed sometimes that we remain competitive with the EC and UK and are typically better than some centers that DO run 4DVAR (Canada, Navy, JMA).

Having said all that, we are in fact working (with extremely limited resources) on developing some sort of 4DVAR....and the difficulty is not just about funds, though that plays a role.

Anyway, I only intended to reply/comment on the 18z/06z misconceptions. This has gotten a bit OT, and I think there is recent thread that touched on some of this already. I'm happy to take the conversation there and answer any/all questions.

[A-L-E-X]

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.

What I remember seeing was (in this order, from most accurate to least): 12z, 0z, 18z, 6z

[gkrangers]

Thank-you for the information.

Ditto. Glad someone on the inside was able to chime in on he model stuff. We need a model facts and myths wiki or something...

[A-L-E-X]

First, apologies for the tone (I don't intend to talk down to anyone).  I haven't been at NCEP all that long but I am in fact part of the global data assimilation team (and actually, I have worked with DA in oceanography as well....).  

Just a side note, I'll never run away from the fact that the EC has better verification scores....you can't argue with the truth, but there is no way to prove it's because they run 4DVAR and simply don't have a superior forecast model (likely a combination of both).  

Do I wish we were running 4DVAR?  Of course I do.....but given what's happened I'm proud to say that we're getting way more out of our "crappy 3DVAR" system than any of us ever would have imagined.  I'm amazed sometimes that we remain competitive with the EC and UK and are typically better than some centers that DO run 4DVAR (Canada, Navy, JMA).

Having said all that, we are in fact working (with extremely limited resources) on developing some sort of 4DVAR....and the difficulty is not just about funds, though that plays a role.  

Anyway, I only intended to reply/comment on the 18z/06z misconceptions.  This has gotten a bit OT, and I think there is recent thread that touched on some of this already.  I'm happy to take the conversation there and answer any/all questions.

I would glady spend more money on this.  I know you guys do a great job with the resources you have, and it could be even better with more resources.

As Americans we spend a crapload on stuff we dont need and we let pass things that are really important-- like advancement in science.  Oceanography is a fascinating field; I dont know which part of the country you're from but our program out here at SUNY Stony Brook is really good. 

[earthlight]

Just a simple question here, wondering if you have an answer since you work at NCEP. I was told the GFS, when run with 4DVAR initialization, outperforms the Euro. I assumed it was a rumor, but figure now would be a good time to put it to bed.

[gkrangers]

Havent seen the Ukie H5 yet, but that would be a good sign if it was west.

And jeez...didn't mean to spark such a heated debate regarding NCEP models...just saying the Euro has a greater resolution and more thorough initialization, thus it's probably a good idea to give it's solution some extra weight, especially with it's ensemble mean agreeing.

And because it fooking buries us.

[dtk]

Just a simple question here, wondering if you have an answer since you work at NCEP. I was told the GFS, when run with 4DVAR initialization, outperforms the Euro. I assumed it was a rumor, but figure now would be a good time to put it to bed.

It's false. No such run has ever been done since there is no tangent-linear (TLM) nor adjoint (AD) versions of the GFS (the TLM and AD versions of the model are a necessary component to carry out the 4D part of 4DVAR).

Now, what some people at NCEP have done are to create ECMWF analysis look-alikes to initialize the GFS, and the results generally improve the skill of the forecasts, but this isn't really a test then of 4DVAR or 3DVAR (since the EC model goes into creating the ECMWF analysis, etc).

[dtk]

I would glady spend more money on this. I know you guys do a great job with the resources you have, and it could be even better with more resources.

As Americans we spend a crapload on stuff we dont need and we let pass things that are really important-- like advancement in science. Oceanography is a fascinating field; I dont know which part of the country you're from but our program out here at SUNY Stony Brook is really good.

I'm definitely more of an atmosphere guy, but I've dabbled some with oceanography....since SSTs are such a huge part in model forcing. BTW, any / all things I share here are my opinion only, as I'm not entitled to speak for NCEP or fellow employees. I just think it's important for us to be much more open in terms of how things work and where we're headed (again, MY opinion).

Lastly, I'm probably the last person in the world to actually defend the GFS....in fact, being part of the analysis team, I end up being the first one to complain about the model .

[Chris L]

The thing to look for tonight on the ECMWF is to keep an eye on the southern stream because that is what is giving us the loaded QPF amounts, I wouldn't be surprised if it trended stronger a little bit, (or weaker given the Nina) and give us the same amount of liquid for the storm and snow that we so desire for. Speed is very important also.

[A-L-E-X]

I'm definitely more of an atmosphere guy, but I've dabbled some with oceanography....since SSTs are such a huge part in model forcing.  BTW, any / all things I share here are my opinion only, as I'm not entitled to speak for NCEP or fellow employees.  I just think it's important for us to be much more open in terms of how things work and where we're headed (again, MY opinion).  

Lastly, I'm probably the last person in the world to actually defend the GFS....in fact, being part of the analysis team, I end up being the first one to complain about the model .

Ive also heard that another reason its hard to upgrade to 4DVAR is because of the processing time it takes.  Being a computer guy, I know that there have been significant advances made in the last few years in parallel processing using GPGPU and CUDA (basically using graphics processors to do the work, because theyre more powerful than CPUs, even those of supercomputers-- with hundreds or thousands of cores)-- maybe this can be incorporated into modeling as well, to improve the resolution and the data assimilation scheme.  I know quantum computing is a ways down the line (maybe 10 years or so), but I would think the future looks bright as far as weather modeling is concerned-- or at least until we run into chaos and Heisenberg's Uncertainty Principle lol 

[dtk]

Ive also heard that another reason its hard to upgrade to 4DVAR is because of the processing time it takes. Being a computer guy, I know that there have been significant advances made in the last few years in parallel processing using GPGPU and CUDA (basically using graphics processors to do the work, because theyre more powerful than CPUs, even those of supercomputers-- with hundreds or thousands of cores)-- maybe this can be incorporated into modeling as well, to improve the resolution and the data assimilation scheme. I know quantum computing is a ways down the line (maybe 10 years or so), but I would think the future looks bright as far as weather modeling is concerned-- or at least until we run into chaos and Heisenberg's Uncertainty Principle lol

Actually, because of the advances in computing, I think people are starting to really think hard about transitioning toward ensemble-based things (which are much easier to take advantage of such architectures)....including at the European Center ;-).

[A-L-E-X]

Actually, because of the advances in computing, I think people are starting to really think hard about transitioning toward ensemble-based things (which are much easier to take advantage of such architectures)....including at the European Center ;-).

Nice idea-- you could increase the number of ensembles and parallel process them more quickly and have a larger sample size to work with

[earthlight]

Just saw the 18z UKMET, seems to fall somewhere between the GFS and Euro, definitely more towards the Euro. It has a 1003mb low probably 50 miles off OBX at hr 72.

[bluewave]

OKX thinks that 12z Euro was suffering from convective feedback problems.

LONG TERM /SATURDAY NIGHT THROUGH THURSDAY/...MODELS HAVE COME INTO MUCH BETTER AGREEMENT ON THE MEDIUM TO LONGRANGE. FOR NOW HAVE GONE WITH A BLEND OF THE 00Z ECMWF/12Z GFS/12ZCMC GLOBAL THROUGH MONDAY...THEN REPLACED THE 00Z ECMWF WITH THE12Z ECMWF FROM MONDAY NIGHT ON. THE 00Z ECMWF DEVELOPS A WARM COREAT 850 HPA WITH ITS LOW BY SUNDAY NIGHT...APPARENTLY DUE TOCONVECTIVE FEEDBACK PROBLEMS...SO ITS LOW IS PROBABLY A TAD ON THESTRONG SIDE...AND A TAD TOO FAR WEST...IN ADDITION TO ITSQUESTIONABLE THERMAL FIELDS...SO PREFERRED NOT TO USE IT DURINGTHE CORE OF THE POTENTIAL STORM FOR EARLY NEXT WEEK.

http://forecast.weather.gov/product.php?site=NWS&issuedby=OKX&product=AFD&format=CI&version=1&glossary=1&hig

[earthlight]

OKX thinks that 12z Euro was suffering from convective feedback problems.

I find it so weird that HPC model diagnostics is riding the ECMWF and a west track while Upton is throwing it out.

[earthlight]

...MID ATLANTIC/NORTHEAST...

THE ECMWF/CANADIAN/GFS/NAM MODELS HAVE TRENDED WEST THIS CYCLE

WITH THE CYCLONE TRACK OF THE SYSTEM FORECAST TO DEVELOP ALONG THE

SOUTHEAST COAST AND MOVE OVER THE COASTAL

WATERS OF THE MID ATLANTIC/NEW ENGLAND. THE SYSTEM REMAINS A

TOUGH CHALLENGE DUE TO RUN TO RUN FLUCTUATIONS IN TRACK/INTENSITY

AS THERE ARE CONFLICTING SIGNALS CONCERNING SNOWSTORM POTENTIAL

RESULTING FROM TIMING AND INTENSITY CHANGES. THE STRONGER MODEL

SOLUTIONS AND THOSE WITH A FURTHER WEST TRACK LIKE THE 12Z ECMWF

RESULT IN GREATER SNOWFALL POTENTIAL. GIVEN THE ECMWF 300 MB JET

STRUCTURE AND 700 MB WAVE COMING ON TO THE WEST COAST IN CA LOOK

SIMILAR TO THE PRIOR 00Z RUN...I DO NOT SEE PACIFIC SAMPLING OF

THE DATA AS THE CAUSE OF THE RUN TO RUN CHANGE. IT APPEARS TO BE

MORE TIED TO TYPICAL PHASING DIFFERENCES BETWEEN CONFLUENT FLOW OF

TWO JET STREAMS IN THE EAST...AND SUBJECT TO FURTHER CHANGE. LOW

RISKS OF FOUR AND EIGHT INCHES ARE SHOWN IN PARTS OF THE MID

ATLANTIC TO LONG ISLAND/SOUTHERN NEW ENGLAND UNTIL THE MODELS CAN

SETTLE ON THE EVOLUTION OF THE CYCLONE AND RESULTANT ADJUSTMENTS

IN WHERE THE SNOW/RAIN TRANSITION ZONE SETS UP AND HOW FAR WEST

HEAVY PRECIPITATION EXTENDS. CONFIDENCE REMAINS LOW...AS IS

TYPICAL FOR A POTENTIAL EAST COAST SNOW EVENT. CONSEQUENTLY...A

BLEND OF MODEL/SREF MEAN SOLUTIONS WAS USED FOR THE FORECAST.

THE PROBABILITY OF SIGNIFICANT ICING IS LESS THAN 10 PERCENT.

PETERSEN

[ptb127]

I find it so weird that HPC model diagnostics is riding the ECMWF and a west track while Upton is throwing it out.

That is somewhat odd

[tombo82685]

OKX thinks that 12z Euro was suffering from convective feedback problems.

LONG TERM /SATURDAY NIGHT THROUGH THURSDAY/...MODELS HAVE COME INTO MUCH BETTER AGREEMENT ON THE MEDIUM TO LONGRANGE. FOR NOW HAVE GONE WITH A BLEND OF THE 00Z ECMWF/12Z GFS/12ZCMC GLOBAL THROUGH MONDAY...THEN REPLACED THE 00Z ECMWF WITH THE12Z ECMWF FROM MONDAY NIGHT ON. THE 00Z ECMWF DEVELOPS A WARM COREAT 850 HPA WITH ITS LOW BY SUNDAY NIGHT...APPARENTLY DUE TOCONVECTIVE FEEDBACK PROBLEMS...SO ITS LOW IS PROBABLY A TAD ON THESTRONG SIDE...AND A TAD TOO FAR WEST...IN ADDITION TO ITSQUESTIONABLE THERMAL FIELDS...SO PREFERRED NOT TO USE IT DURINGTHE CORE OF THE POTENTIAL STORM FOR EARLY NEXT WEEK.

http://forecast.weat...&glossary=1

what i don't get about these things, is convective feedback would throw the placement of the low off, thats it im assuming. The euro from the get go was more amplified and had a stronger s/w, which i would assume isn't thrown off by convective feedback

[earthlight]

what i don't get about these things, is convective feedback would throw the placement of the low off, thats it im assuming. The euro from the get go was more amplified and had a stronger s/w, which i would assume isn't thrown off by convective feedback

Convective feedback issues can seep through every aspect of a model run. If Upton is correct, the ECMWF has convective feedback which is causing a warm core system and a ridiculously deep surface low, which then leads to a domino effect of more QPF, stronger CCB, etc.

[A-L-E-X]

I find it so weird that HPC model diagnostics is riding the ECMWF and a west track while Upton is throwing it out.

The margin of error for this thing is so huge that it seems pointless to ride on anything lol.  This thing has us glued from run to run

[A-L-E-X]

Convective feedback issues can seep through every aspect of a model run. If Upton is correct, the ECMWF has convective feedback which is causing a warm core system and a ridiculously deep surface low, which then leads to a domino effect of more QPF, stronger CCB, etc.

The funny thing is convective feedback was also being implicated in the 18z GFS flatline look lol.

[bluewave]

what i don't get about these things, is convective feedback would throw the placement of the low off, thats it im assuming. The euro from the get go was more amplified and had a stronger s/w, which i would assume isn't thrown off by convective feedback

Yeah,we'll need to see in the next few runs how soon the low closes off to get the exact track.

[IsentropicLift]

OKX thinks that 12z Euro was suffering from convective feedback problems.

LONG TERM /SATURDAY NIGHT THROUGH THURSDAY/...MODELS HAVE COME INTO MUCH BETTER AGREEMENT ON THE MEDIUM TO LONGRANGE. FOR NOW HAVE GONE WITH A BLEND OF THE 00Z ECMWF/12Z GFS/12ZCMC GLOBAL THROUGH MONDAY...THEN REPLACED THE 00Z ECMWF WITH THE12Z ECMWF FROM MONDAY NIGHT ON. THE 00Z ECMWF DEVELOPS A WARM COREAT 850 HPA WITH ITS LOW BY SUNDAY NIGHT...APPARENTLY DUE TOCONVECTIVE FEEDBACK PROBLEMS...SO ITS LOW IS PROBABLY A TAD ON THESTRONG SIDE...AND A TAD TOO FAR WEST...IN ADDITION TO ITSQUESTIONABLE THERMAL FIELDS...SO PREFERRED NOT TO USE IT DURINGTHE CORE OF THE POTENTIAL STORM FOR EARLY NEXT WEEK.

http://forecast.weat...&glossary=1

Actually its the 00z in which he thought was

[jm1220]

Convective feedback issues can seep through every aspect of a model run. If Upton is correct, the ECMWF has convective feedback which is causing a warm core system and a ridiculously deep surface low, which then leads to a domino effect of more QPF, stronger CCB, etc.

I guess you call the Euro solution a REAL snowicane then.

[tombo82685]

Convective feedback issues can seep through every aspect of a model run. If Upton is correct, the ECMWF has convective feedback which is causing a warm core system and a ridiculously deep surface low, which then leads to a domino effect of more QPF, stronger CCB, etc.

yea i get that, but from the get go the trof and the southern s/w was stronger than 0z, that wouldn't be effected by it, since the storm hasn't even formed yet.

[joey]

The funny thing is convective feedback was also being implicated in the 18z GFS flatline look lol.

LOL. That's when you know everyone needs to stop overanalyzing and take a few hours break.

[tombo82685]

Actually its the 00z in which he thought was

that explains it then, it was the 0z with feedback issues?