So what the hell happened with the GFS?

[Mr. Windcredible!]

Clearly the Euro led the way on this one. The NAM caught on late but was pretty consistent once it caught did, highlighting the heavy deformation band. NAM had it's usual QPF bias but it wasn't as extreme as most of us suspected. But what the hell happened with the GFS? Even once snow had started for CT, the 12z run yesterday, it barely had the 1.5" QPF line back to HVN-IJD-ORH. Was it just its inability to properly handle the convective nature of the storm...initialization errors...or something more? Discuss..

[radarman]

I think this is stating the obvious but when the Euro is showing run after run after run of the same solution, particularly inside day 5 there is really almost no point even looking at the GFS.  Of course we all still do it, but Kevin does have a point that it often just causes unnecessary confusion.  It's only when the Euro is waffling and doesn't have a good handle that it's useful at this point.  

[Damage In Tolland]

I think this is stating the obvious but when the Euro is showing run after run after run of the same solution, particularly inside day 5 there is really almost no point even looking at the GFS.  Of course we all still do it, but Kevin does have a point that it often just causes unnecessary confusion.  It's only when the Euro is waffling and doesn't have a good handle that it's useful at this point.  

You had me at hello

[Mr. Windcredible!]

I think this is stating the obvious but when the Euro is showing run after run after run of the same solution, particularly inside day 5 there is really almost no point even looking at the GFS.  Of course we all still do it, but Kevin does have a point that it often just causes unnecessary confusion.  It's only when the Euro is waffling and doesn't have a good handle that it's useful at this point.  

 

The Euro has had a few fails in the past few years too IIRC. Although most of the time, I think when it's finally lost a storm it's still been 3-4 days out. So I guess when it's consistent and remains consistent within 72 hours...look out. And while we bad mouth the NAM lots as well..the Euro/NAM combo is hard to beat!But would still be curious to hear thoughts on why the GFS was so far off this time. Even when it's been lousy in the past it usually catches on by 24 hours out.

[hurricaneman]

I would have to assume either convective feedback or it had trouble resolving the phasing

[WintersGrasp]

Which model runs of the models had that insane band over LI? Looking at so many the past several days I forget which had it. I think the NAM did actually at some point....I know at one point on the NAM it WAS over NYC actually

[radarman]

The Euro has had a few fails in the past few years too IIRC. Although most of the time, I think when it's finally lost a storm it's still been 3-4 days out. So I guess when it's consistent and remains consistent within 72 hours...look out. And while we bad mouth the NAM lots as well..the Euro/NAM combo is hard to beat!But would still be curious to hear thoughts on why the GFS was so far off this time. Even when it's been lousy in the past it usually catches on by 24 hours out.

 

Yeah not saying the Euro is infallible, but it's sort of like why you shouldn't generally forecast historic snow totals until the last second... most of the time you'll be wrong.  Similarly with paying attention to the GFS when the EC is rock steady.  

[CardinalWinds]

Sort of related to this topic...CNN meteorologist Chad Myers had Dr. Uccellini on-air during the storm and after congratulating him on his new post as director of the NWS and heaping praises on his career, Myers took him to task on the general disparity between the superior Euro and the inferior GFS. I think Uccellini was surprised at this line of inquiry, but responded that the GFS is soon to be moved to a more powerful supercomputer to increase resolution and that plans to increase its data collection grid is also in the works. 

[ChrisM]

^^ he basically admitted that it sucked dude.

[CardinalWinds]

^^ he basically admitted that it sucked dude.

Well, yeah, but I wanted to try and say a bit more than that.

[ChrisM]

Well, yeah, but I wanted to try and say a bit more than that.

 

No I know, I just thought it was funny that he didn't even try to defend it really haha

[Mr. Windcredible!]

Which model runs of the models had that insane band over LI? Looking at so many the past several days I forget which had it. I think the NAM did actually at some point....I know at one point on the NAM it WAS over NYC actually

 

The Euro and NAM really started targeting CT/LI and eastern MA 48-60 hours out...and yeah...there was at least one run of the NAM (maybe 12 or 18z Thursday) that took the band as far west as NYC...but I think it was just a hiccup run and trended back east the few runs that followed.

 

Sort of related to this topic...CNN meteorologist Chad Myers had Dr. Uccellini on-air during the storm and after congratulating him on his new post as director of the NWS and heaping praises on his career, Myers took him to task on the general disparity between the superior Euro and the inferior GFS. I think Uccellini was surprised at this line of inquiry, but responded that the GFS is soon to be moved to a more powerful supercomputer to increase resolution and that plans to increase its data collection grid is also in the works. 

 

I figured it's issues had more to do with just resolution, but I've really got no idea. 

[N. OF PIKE]

i'd like more MET input, bc the euro really caused unnecessary confusion, and it mess'd with ryan hanrahan lol JK ryan .  it had mid level lows weaker and SE and was just crap with the track, i assume it had difficult with the southern stream convectively driven vort and i guess perhaps that effected the way it tracked and thus phased. 

[CardinalWinds]

No I know, I just thought it was funny that he didn't even try to defend it really haha

I think Myers blindsided him with the model talk. He wasn't ready for it, you could tell. Uccellini was clearly unaware of the constant harping on the subject Myers was doing on-air for the last 3 days...lol.

[Mr. Windcredible!]

i'd like more MET input, bc the euro really caused unnecessary confusion, and it mess'd with ryan hanrahan lol JK ryan .  it had mid level lows weaker and SE and was just crap with the track, i assume it had difficult with the southern stream convectively driven vort and i guess perhaps that effected the way it tracked and thus phased. 

 

You mean the GFS. But yeah, Ryan and his crew lowered their amounts yesterday after the 12z NAM thinking it was continuing to trend to the GFS. I had the same fear though, can't blame them on that one. I expected the 12z GFS to finally catch on but it was still only a very slight tick west. It wasn't until the 12z Euro and things were already getting underway that it was finally obvious the GFS was out to lunch.

[skierinvermont]

You mean the GFS. But yeah, Ryan and his crew lowered their amounts yesterday after the 12z NAM thinking it was continuing to trend to the GFS. I had the same fear though, can't blame them on that one. I expected the 12z GFS to finally catch on but it was still only a very slight tick west. It wasn't until the 12z Euro and things were already getting underway that it was finally obvious the GFS was out to lunch.

That's where 4 runs of the Euro would be handy... when the Euro and NAM started ticking east (Euro did so at 00z thursday night) it seemed like they might move to a 50/50 compromise with the GFS. But if anything the Euro ticked back west at 12z Friday. If we had a 6z run in between we could have had confirmation earlier.

 

Looking at the EC ensembles could have been informative. Did they tick east too 00z Friday? Or was that just on the OP?

[ORH_wxman]

That's where 4 runs of the Euro would be handy... when the Euro and NAM started ticking east (Euro did so at 00z thursday night) it seemed like they might move to a 50/50 compromise with the GFS. But if anything the Euro ticked back west at 12z Friday. If we had a 6z run in between we could have had confirmation earlier.

 

Looking at the EC ensembles could have been informative. Did they tick east too 00z Friday? Or was that just on the OP?

 

 

Yes they did, but not by a lot. I just pulled it up on WSI (they have a 10 day archive)...but 00z Friday became a lot more intense. That's partly due to being closer so the SLP isn't spread out over a larger area, but it seemed more than you usually see which suggested this storm was going to deepen very rapidly. The Euro was always one of the deepest models with this storm getting it into the mid 970s at least, or even lower. A lot of other guidance did not want to get it below the 980s for pressure.

 

 

The biggest flag for me on remaining quite bullish despite some ugly GGEM/GFS runs (and even a Ukie run) late in the period was the Euro's qpf output was very uncharacteristic on being so bullish. There's a reason it outscores the other models by a lot on QPF. It doesn't forecast extreme QPF blobs very often. But in thiscase, seeing it do it for about 5-6 consecutive runs by the time we got to 00z Friday made me pretty skeptical of other guidance that disagreed with it being much lighter.

[Ed Lizard]

 It doesn't forecast extreme QPF blobs very often.

 

 

 

 

 

Did that have anything to do with the GFS being too far offshore?  The somewhat coarser resolution and greater tendency to generate QPF blobs because of convective feedback issues?

 

 

I forgot which red tagger (probably in the Central forum) told me the SREF models from one of the particular model packages that generated what appeared to be supercells were actually generating QPF blobs due t lower resolution, and linked to something that was from HPC or NWS about looking for 500 mb reflections from the latent heat release in a QPF blob to diagnose what was a spurious convective feedback blob.

 

http://www.hpc.ncep.noaa.gov/mdd/part3c.html

 

 

The other thing I was thinking was initialization, but IIRC the US models initialize in similar fashion, so the Euro wasn't beating the GFS w/ 4D VAR. 

 

ETA  Dur.  Forgot about 5/22/12 ensKF change.  Other than a message about NAM initializing too warm on one of the runs, I don't think I saw initialization error messages...

[Clinch Leatherwood]

Which model runs of the models had that insane band over LI? Looking at so many the past several days I forget which had it. I think the NAM did actually at some point....I know at one point on the NAM it WAS over NYC actually

The euro didnt pick up on the band in li and ct even at 0z last night! It was pretty good but botched the stall point and caved towards the other guidance. People see the snow in their yard and figure the nam nailed it. Not so on the cape or in western areas. The model always overforecasts qpf so when things break that way it'll be more right. Euro was best but far from perfect particularly se and NW

Most models were good with the surface track gfs included. Where they had issues was with the mid level tracks as oceanwx noted out of gyx.

[CT Rain]

You mean the GFS. But yeah, Ryan and his crew lowered their amounts yesterday after the 12z NAM thinking it was continuing to trend to the GFS. I had the same fear though, can't blame them on that one. I expected the 12z GFS to finally catch on but it was still only a very slight tick west. It wasn't until the 12z Euro and things were already getting underway that it was finally obvious the GFS was out to lunch.

 

Yeah what concerned me was the SREFs going east, the GFS going east, and the NAM going east after the 00z Euro went east. The jump east wasn't substantial for places like PVD or ORH but the jump east was pretty concerning for us here in CT.

Then of course the Euro came in at 12z with another monster run and the concern was alleviated. It was just a sort of odd NWP evolution.

 

The Euro certainly isn't always right... it has had some notable struggles... but the GFS has really had some horrific, horrific performances. 

[Damage In Tolland]

The RPM actually nailed the CT totals at 6z Fri then backed off . It was right

[ineedsnow]

This storm ended up being historic for sure!!!     The concerns you guys had were warranted though... never would have thought the valley and some areas near it would have been kind of a jackpot zone though!

Yeah what concerned me was the SREFs going east, the GFS going east, and the NAM going east after the 00z Euro went east. The jump east wasn't substantial for places like PVD or ORH but the jump east was pretty concerning for us here in CT.

Then of course the Euro came in at 12z with another monster run and the concern was alleviated. It was just a sort of odd NWP evolution.

 

The Euro certainly isn't always right... it has had some notable struggles... but the GFS has really had some horrific, horrific performances. 

[Ginx snewx]

Save a horse ride the Euro in all STJ systems

[Baroclinic Zone]

Save a horse ride the Euro in all STJ systems

There really isn't anything special wrong with the GFS.  It's just knowing what patterns support some models vs others.  I'll be honest, I never believed the GFS with this system once the Euro and Ensm latched onto the amplified scenario.  Even the Euro was not exact with the placement of the heaviest precip.  It ended up 75-100mi W of where it was modeled.  But that's the nature of the beast.  These systems can never be 100% accurately depicted by the models.  That's where forecasters are needed.

[Ginx snewx]

There really isn't anything special wrong with the GFS. It's just knowing what patterns support some models vs others. I'll be honest, I never believed the GFS with this system once the Euro and Ensm latched onto the amplified scenario. Even the Euro was not exact with the placement of the heaviest precip. It ended up 75-100mi W of where it was modeled. But that's the nature of the beast. These systems can never be 100% accurately depicted by the models. That's where forecasters are needed.

yep the Euro nailed my QPF but was a little undergone to the west. Dry slots and meso processes are now cast issues. I did not buy the GFS QPF distribution but it's upper level features were not that bad. Anytime you see -6sd 850 inflow its huge.

[CTWeatherFreak]

Sort of related to this topic...CNN meteorologist Chad Myers had Dr. Uccellini on-air during the storm and after congratulating him on his new post as director of the NWS and heaping praises on his career, Myers took him to task on the general disparity between the superior Euro and the inferior GFS. I think Uccellini was surprised at this line of inquiry, but responded that the GFS is soon to be moved to a more powerful supercomputer to increase resolution and that plans to increase its data collection grid is also in the works. 

 

After which he admitted somewhat sheepishly that  it still wont be able to match the ECMWF in resolution!

[Clinch Leatherwood]

That's good news on the GFS being upgraded.

 

Couple of sort of related notes.  MM5 pinned down the stall 18 hours out pretty well.  It accurately depicted the coastal band for most of the day Sunday although it missed the intensity, also correctly stalled the mega band east of the Cape.  That's the one that if it'd been back 125 miles would have resulted in #1 snows in Boston. 

 

Some NAM and GFS maps in addition to the MM5.  The Euro missed the death band in CT too.  They all did really aside of the NAM which as this map shows had a death band for several runs in many places that received nothing of the sort.

 

No model was perfect, the track over the BM or near it was pretty good.  Look at the satellite Roger Smith posted showing it going right over it.  The problems for the models is the system was not stacked at that point, the mid levels were skewed NW from there which pushed the bands further west and probably delayed the stall (euro had it SW for 2-3 days of where it occurred and didn't back off until the night before, but really the day of the storm) until the centers stacked.  Ultimately I don't know what the ratios were in places like ORH. PVD and Boston, but the GEFS QPF wasn't bad, and the GFS not "terrible" aside of the death band.

 

JMHO

 

 

 

 

 

[Clinch Leatherwood]

Couple of other quick obs.  Here's the satellite shot someone else posted, and 2 boston forecasters a day in advance.  They did pretty well.

 

 

[Mr. Windcredible!]

Yes they did, but not by a lot. I just pulled it up on WSI (they have a 10 day archive)...but 00z Friday became a lot more intense. That's partly due to being closer so the SLP isn't spread out over a larger area, but it seemed more than you usually see which suggested this storm was going to deepen very rapidly. The Euro was always one of the deepest models with this storm getting it into the mid 970s at least, or even lower. A lot of other guidance did not want to get it below the 980s for pressure.

The biggest flag for me on remaining quite bullish despite some ugly GGEM/GFS runs (and even a Ukie run) late in the period was the Euro's qpf output was very uncharacteristic on being so bullish. There's a reason it outscores the other models by a lot on QPF. It doesn't forecast extreme QPF blobs very often. But in thiscase, seeing it do it for about 5-6 consecutive runs by the time we got to 00z Friday made me pretty skeptical of other guidance that disagreed with it being much lighter.

I agree the Euro QPF was impressive and gave some credence to the NAM. There was no doubt in my mind if Euro/NAM they had the track/timing of the phase right we were easily gonna get 24" amounts in CT...it was more a question of we're they right with the track and timing.

[CoastalWx]

GFS did have that nice QPF blob over LI extending north so it saw something. It also hinted at the more progressive pattern..otherwise the Cape snow they had on Saturday would ahve been ours in ern MA. So it wasn't it's greatest performance by any means..but it did hint at some thing that occurred.

 

It's all how you use guidance. Even the lousier performing model like the GFS hinted at a more progressive situation which blended in with the aggressive models...gave you a great forecast.