the models... winners and losers in the blizzard

[green tube]

i'm still in a snow coma, but my impression is....

biggest loser: ggem. never got it right.

2nd biggest loser: ukie... had it OTS day after day until the last minute.

biggest winner: GFS. it was the first to bring it back to the coast after all models had it OTS.

2nd place: euro. had a big hit. took it away. then played catch up to the GFS.

is that about right?

i know early in the week the euro was the only one showing a big hit.... but it lost it.

to me, once they all lost it, the scoreboard is reset to 00.

[SnowGoose69]

I posted in the main forum just a few minutes ago that I'm not totally sure what caused the error in the day 3-5 range....DT posted specifically that the 1/25/00 event was missed due to the models viewing the northern stream disturbance as a kicker as opposed to it being able to pull the trough negative....in this event it was either the same thing or simply bad model sampling of that disturbance til it reached far enough south near the border....I score the GFS higher than the Euro as well since the Euro was right for the wrong reason in the medium-long range...the reason the Euro had the event beyond day 5 was that it had the system too slow out of the SW....if not for that bias the Euro was probably a miss all along...so basically the GFS was probably correct even in the long range when it had a miss....the UKMET cannot be trusted the rest of this winter because its progressive bias is being fueled further by the La Nina...its been too east with every event so far in December...the GGEM I think was very good until the short range, it correctly had the miss before that northern disturbance was resolved....I think the GGEM is our most reliable model this winter as I've posted already since its west/warm bias is being negated by the La Nina a bit....the Euro and UKMET are going to continue having issues being too slow out of the SW and too progressive respectively.

[Amped]

What it came down to was not which was the best model not which model had the best initialization scheme, but which model had the best initialization of the one S/W. Or, which model initialized it too far west instead of too far east.

This was one of those setups that wasat least 10x as hard to predict 72hrs out than 2/5/2010 was. Nature doesn't give us the same challenge every time.

Luckily all models caught it 00z on the 25th, much better than 1/25/2000.

[SnowGoose69]

What it came down to was not which was the best model not which model had the best initialization scheme, but which model had the best initialization of the one S/W. Or, which model initialized it too far west instead of too far east.

This was one of those setups that wasat least 10x as hard to predict 72hrs out than 2/5/2010 was. Nature doesn't give us the same challenge every time.

Luckily all models caught it 00z on the 25th, much better than 1/25/2000.

This event was another great example along with 2/5 that would have been a disaster 10 years ago and especially 20....I said that in 2000 the 2/5 event would have been forecast as a hit at least as far northward as POU/BDL/AVP/MSV because the NGM consistently threw precipitation too far north and west on East Coast storms, the ETA had a severe NW bias in those days and the AVN was just bad...that event may very well have been a 12-18 inch forecast for NYC that would have verified as zero...I'm not sure but maybe in the 12-24 hour range the models would have caught on somewhat....this event would likely have been missed as well like 1/25 though it probably would have been caught as well just before.

[JBG]

2nd place: euro. had a big hit. took it away. then played catch up to the GFS.

Definite rehab for the Euro after the December 19 whiff.

[green tube]

thanks for the insight snowgoose. :thumbs_up:

i hate the actual smiley, so i typed it out

[baroclinic_instability]

GFS definitely won this event since it was the first to catch back on to the idea approximately 3 days before the storm developed. There was a verifiable trend from day 6-7 down to day 3-4, and that was the southern PV anomaly would be weaker than initially expected. We analyzed that southern PV anomaly probably more than any other feature of this storm, and for good reason. Early epic runs of the ECM that were crushing the Mid-Atlantic had a very deep phase with the southern PV into the GOM, and the southern PV itself was rather impressive. At the same time, the GFS was still playing catch up to the Euro as it had an odd northern solution with a very early phase that incited weak cyclogenesis and changed the orientation of the low level cold air more towards the SE which eventually led to a solution that was OTS. All models, once that southern PV anomaly finally came onshore and into the observing network eventually showed a much weaker southern stream that was barely strong enough to incite cyclogenesis as it phased with the northern stream. In other words, we were barely phasing anything, and this showed as the later models were essentially "deforming" the upper level height field of the weak southern PV anomaly and ejecting it rapidly northeast. At this point, all guidance went OTS and the threat looked much smaller. Then the GFS caught on to something late...first it was a perfectly timed phase with the southern anomaly. It had to be late enough or the solution would miss the GOM and it would eventually head OTS. Secondly, it had to be far enough south to develop a surface depression in the GOM. Third, the backside trough itself eventually ended up being more amplified than any model was suggesting early on owing to the strength of the Rockies ridge. Lastly, the southern stream PV was developing much more convection than early guidance suggested as the system headed near the GOM, and this convection aided in the development of low level convergence over the GOM. As discussed earlier, the GOM was key in developing a bomb that hooked into the coast since it essentially "priimed" the system for rapid positive feedback development over the Gulf Stream. GOM essentially worked to both increase the amount of low level warm air advection as well as to decrease the static stability through latent heat release through condensation in the low levels of the troposphere owing to convergence near the boundary layer. In the end, the numerical guidance struggled with a number of different factors in this event, and eventually the storm developed for different reasons than we initially believed it would require. As meteorologists/forecasters, we key in on the important "players" and we watch trends and determine what probability there is of a particular event. The weakening of that southern stream PV was something we all could see, and once it developed onshore and into the observing network, it was clear it was going to be much weaker than we all initially believed. ECM lost the storm OTS, GFS went farther east after it was beginning to trend W towards the ECM, CMC was OTS, UK was OTS, ensembles shifted eastward. It looked bad, and this was when DT talked about the pooch, when the weather weenies began posting hate threads for the ECM, and general anarchy developed. It was a tough storm for everyone. Nobody was eventually completely right or wrong, and the models, IMO, did fine even though not one of them was "perfect". We had plenty of lead time (as early as 6-7 days out initially), and a more definite 2-3 days out before the storm finally hit. 50 years ago, a storm like this likely would have killed hundreds, maybe thousands. We have come a long way in weather prediction.

[KEITH L.I]

My model scores

GFS B+

EURO B

SREF B-

NAM B-

GGEM D

UK F

NOGAPS F

[Sampson]

From my very simplistic perspective:

The Euro had the storm in the long range as well as the GFS. The GFS showed it first then abruptly lost it just as the Euro OP started to consistently show the coastal solution - although as snowgoose states, for the wrong reason.

Then the GFS started to slowly shift west as the Euro continued to show the coastal. It seemed that as soon as the GFS came closer to the Euro, the Euro jumped ship and had a huge shift east. it It had 2 blips and then returned to a compromised version of what it depicted earlier. It seems very similar to what happened with the 19th threat. I don't recall if the GFS showed a solution closer to the final track in the longer range, but it took the system too far east in mid-long range, while the Euro was consistently staying west. As the GFS started shifting west, the Euro gave us a heart attack with two runs along the coast, only to return to where it began. Not sure if that's telling wrt any bias that either may have.

Ultimately, I think the Euro did the best job of depicting the energy in the GOM, while all the models struggled with the PV and its interaction with the low. The GFS did start to correct itself first in that regard.

[Dsnowx53]

The RGEM was atrocious, too. It was generally way too progressive.

[wxlady]

The main problem I had with the models was the inability to deal with the sub-gridscale accumulation gradients. I saw way too many ridiculous QPF forecasts of 0.15-0.25" in Frederick County, MD on several runs that it gave me false hope for something more than trace to an inch of snow. Baltimore really didn't get much either. Just for giggles, here's one of the more impressive crazy gradients for snow totals in New Jersey...

Storm report from Mt Holly WFO

Somerset County, NJ:

..SOMERSET COUNTY

4 ENE FRANKLIN TWP 19.5 700 AM 12/27 COCORAHS

5 ESE HILLSBOROUGH T 16.5 700 AM 12/27 COCORAHS

1 NW BERNARDS TWP 12.0 700 AM 12/27 COCORAHS

ENE SOMERVILLE 10.5 800 AM 12/27 COCORAHS

2 NNE MONTGOMERY TWP 9.0 800 AM 12/27 COCORAHS

1 NNW BERNARDS TWP 8.4 630 AM 12/27 COCORAHS

3 NE MONTGOMERY TWP 8.0 745 AM 12/27 COCORAHS

2 W BRIDGEWATER TWP 7.1 700 AM 12/27 COCORAHS

1 ENE BERNARDS TWP 7.0 700 AM 12/27 COCORAHS

3 NW BRIDGEWATER TWP 7.0 700 AM 12/27 COCORAHS

1 N HILLSBOROUGH TWP 6.1 900 AM 12/27 COCORAHS

[Amped]

The main problem I had with the models was the inability to deal with the sub-gridscale accumulation gradients. I saw way too many ridiculous QPF forecasts of 0.15-0.25" in Frederick County, MD on several runs that it gave me false hope for something more than trace to an inch of snow. Baltimore really didn't get much either. Just for giggles, here's one of the more impressive crazy gradients for snow totals in New Jersey...

Storm report from Mt Holly WFO

Somerset County, NJ:

..SOMERSET COUNTY

4 ENE FRANKLIN TWP 19.5 700 AM 12/27 COCORAHS

5 ESE HILLSBOROUGH T 16.5 700 AM 12/27 COCORAHS

1 NW BERNARDS TWP 12.0 700 AM 12/27 COCORAHS

ENE SOMERVILLE 10.5 800 AM 12/27 COCORAHS

2 NNE MONTGOMERY TWP 9.0 800 AM 12/27 COCORAHS

1 NNW BERNARDS TWP 8.4 630 AM 12/27 COCORAHS

3 NE MONTGOMERY TWP 8.0 745 AM 12/27 COCORAHS

2 W BRIDGEWATER TWP 7.1 700 AM 12/27 COCORAHS

1 ENE BERNARDS TWP 7.0 700 AM 12/27 COCORAHS

3 NW BRIDGEWATER TWP 7.0 700 AM 12/27 COCORAHS

1 N HILLSBOROUGH TWP 6.1 900 AM 12/27 COCORAHS

That kind of stuff is the luck of the draw.

If you move a sharp gradient 20 miles east or west, the result is a huge error. The same concept makes forecasting small hurricanes a nightmare and large ones are still no picnic. In fact "Gradient" and "sensitivity to displacement." almost mean the same thing. Fell free to use them interchangeably, one implies the other.

[forkyfork]

the xmas eve runs of the gfs nailed the NJ snow max

[eduggs]

I posted in the main forum just a few minutes ago that I'm not totally sure what caused the error in the day 3-5 range....DT posted specifically that the 1/25/00 event was missed due to the models viewing the northern stream disturbance as a kicker as opposed to it being able to pull the trough negative....in this event it was either the same thing or simply bad model sampling of that disturbance til it reached far enough south near the border....I score the GFS higher than the Euro as well since the Euro was right for the wrong reason in the medium-long range...the reason the Euro had the event beyond day 5 was that it had the system too slow out of the SW....if not for that bias the Euro was probably a miss all along...so basically the GFS was probably correct even in the long range when it had a miss....the UKMET cannot be trusted the rest of this winter because its progressive bias is being fueled further by the La Nina...its been too east with every event so far in December...the GGEM I think was very good until the short range, it correctly had the miss before that northern disturbance was resolved....I think the GGEM is our most reliable model this winter as I've posted already since its west/warm bias is being negated by the La Nina a bit....the Euro and UKMET are going to continue having issues being too slow out of the SW and too progressive respectively.

I don't agree with this. The Euro was the first model to correctly nail the delayed timing of the event and to correctly target the digging of the northern stream all the way to the gulf coast. This was a northern stream dominant system. The GFS (and most other models) were too fast with the southern s/w (and therefore too flat and progressive) until it finally caught on at the end. The Euro blizzard depictions days 5-7 were really pretty close to the end result. I see no evidence of bias affecting the solution here.

[Jet-Phase]

I don't agree with this. The Euro was the first model to correctly nail the delayed timing of the event and to correctly target the digging of the northern stream all the way to the gulf coast. This was a northern stream dominant system. The GFS (and most other models) were too fast with the southern s/w (and therefore too flat and progressive) until it finally caught on at the end. The Euro blizzard depictions days 5-7 were really pretty close to the end result. I see no evidence of bias affecting the solution here.

I have to give it to the GFS. It showed the storm as early as 12-18-2010 on several runs even though the event happened a day later. It recognize the potential signal of something significant happening earlier than any other model 8/9 days in advance. It lost it for a few days and then was first to bring it back. Check out the GFS run on 12-18-2010:

[green tube]

GFS definitely won this event since it was the first to catch back on to the idea approximately 3 days before the storm developed. There was a verifiable trend from day 6-7 down to day 3-4, and that was the southern PV anomaly would be weaker than initially expected. We analyzed that southern PV anomaly probably more than any other feature of this storm, and for good reason. Early epic runs of the ECM that were crushing the Mid-Atlantic had a very deep phase with the southern PV into the GOM, and the southern PV itself was rather impressive. At the same time, the GFS was still playing catch up to the Euro as it had an odd northern solution with a very early phase that incited weak cyclogenesis and changed the orientation of the low level cold air more towards the SE which eventually led to a solution that was OTS. All models, once that southern PV anomaly finally came onshore and into the observing network eventually showed a much weaker southern stream that was barely strong enough to incite cyclogenesis as it phased with the northern stream. In other words, we were barely phasing anything, and this showed as the later models were essentially "deforming" the upper level height field of the weak southern PV anomaly and ejecting it rapidly northeast. At this point, all guidance went OTS and the threat looked much smaller. Then the GFS caught on to something late...first it was a perfectly timed phase with the southern anomaly. It had to be late enough or the solution would miss the GOM and it would eventually head OTS. Secondly, it had to be far enough south to develop a surface depression in the GOM. Third, the backside trough itself eventually ended up being more amplified than any model was suggesting early on owing to the strength of the Rockies ridge. Lastly, the southern stream PV was developing much more convection than early guidance suggested as the system headed near the GOM, and this convection aided in the development of low level convergence over the GOM. As discussed earlier, the GOM was key in developing a bomb that hooked into the coast since it essentially "priimed" the system for rapid positive feedback development over the Gulf Stream. GOM essentially worked to both increase the amount of low level warm air advection as well as to decrease the static stability through latent heat release through condensation in the low levels of the troposphere owing to convergence near the boundary layer. In the end, the numerical guidance struggled with a number of different factors in this event, and eventually the storm developed for different reasons than we initially believed it would require. As meteorologists/forecasters, we key in on the important "players" and we watch trends and determine what probability there is of a particular event. The weakening of that southern stream PV was something we all could see, and once it developed onshore and into the observing network, it was clear it was going to be much weaker than we all initially believed. ECM lost the storm OTS, GFS went farther east after it was beginning to trend W towards the ECM, CMC was OTS, UK was OTS, ensembles shifted eastward. It looked bad, and this was when DT talked about the pooch, when the weather weenies began posting hate threads for the ECM, and general anarchy developed. It was a tough storm for everyone. Nobody was eventually completely right or wrong, and the models, IMO, did fine even though not one of them was "perfect". We had plenty of lead time (as early as 6-7 days out initially), and a more definite 2-3 days out before the storm finally hit. 50 years ago, a storm like this likely would have killed hundreds, maybe thousands. We have come a long way in weather prediction.

a virtual thesis on the evolution of the event! very much appreciated.

[baroclinic_instability]

My model scores

GFS B+

EURO B

SREF B-

NAM B-

GGEM D

UK F

NOGAPS F

Mostly agree with this assessment except the NAM is a definite D. It would be an F, but its final runs before the event were not all that awful. Global Gem would prolly be a C- since it was holding onto a storm through much of the day 3-5 range.

[jm1220]

I was one of those completely blindsided by this event and said to "move on" multiple times before it came back from the dead. I really thought it was over once the Euro slid it out and the entire evolution looked to become yet another sheared out and broad trough that could never get itself together before it was out past the Maritimes. That increased digging of the northern stream and perfect timing of the southern wave was key in inciting the phase in the correct place, and forming a storm in time to nail us. That southern wave was quite intense indeed, and when combined with the northern stream created probably one of NYC's all time great blizzards. I definitely agree that a few decades ago a storm like this would have been quite deadly, for the fact it likely wouldn't have been forecast until it was here and just the intensity of it. The winds were just incredible and nothing like I've witnessed in a winter storm since 1996. And in terms of overall impact to the NYC metro area, it at least seems to rival 1996 and perhaps even the Lindsay Storm.

[eduggs]

I have to give it to the GFS. It showed the storm as early as 12-18-2010 on several runs even though the event happened a day later. It recognize the potential signal of something significant happening earlier than any other model 8/9 days in advance. It lost it for a few days and then was first to bring it back. Check out the GFS run on 12-18-2010:

I agree that the GFS really latched onto this threat in the long range, which is kind of amazing. However the GFS solution you posted was more than a day too early because at that time it was moving the southern s/w west to east across the country instead of dampening it toward the GOM as the northern stream dug and carved out a deep trof. The Euro, for example, had the s/w timing and amplitude more correct. The mid/long range was kind of a wash - which reiterates that one model usually doesn't completely dominate... they each tend to pick up different aspects of a patterns' eventual evolution.

The GFS being the first to bring back the threat after the big shift east was the most impressive thing to me.