Dissecting The Bust

[BxEngine]

What went wrong? Is it as simple as the Euro being off by a bit, but it just so happened to be off over the most populated metro areas in the US?

Is the new GFS now a model to be taken seriously?

No politics or upton/mt holly bashing (understand the difference between bashing and legitimate criticism please!), keep it civil, and try to back up your thoughts/arguments with data. Ready, go:

[Clinch Leatherwood]

Over reliance on a once infallible model coupled with incorporated "improvements" to the model that have introduced a demonstraraable slow bias with east coast lows developing under negatively tilting toughs

[mranger48]

U know what's the most dangerous

Aspect of this whole thing imo? is we get another storm threat of a

Dangerous magnitude sometime soon and people don't take it seriously because of what happened with this storm. This could be a huge problem....

Sent from my iPhone using Tapatalk

[weatherpruf]

Over reliance on a once infallible model coupled with incorporated "improvements" to the model that have introduced a demonstraraable slow bias with east coast lows developing under negatively tilting toughs

 

 

Over reliance on a once infallible model coupled with incorporated "improvements" to the model that have introduced a demonstraraable slow bias with east coast lows developing under negatively tilting toughs

Sounds good to me.

[LongBeachSurfFreak]

Well for people such as my self stuck at work in manhattan still up from yesterday this is a nightmare. All I want to do Is drive my truck home and go to sleep but I can't. So it's definitely screwing my life up pretty bad. Yes I'm super bitter. I understand that 50 miles would have been huge. And as much as I said upton and I are no longer friends I understand. This easily could have been NYCs biggest storm and it will be out in the forks. I just want the travel ban lifted it's not snowing at all in manhattan. Not at all. Man new jersey is even more ridiculous.

[Ridingtime]

The Euro needs to be dethroned from its place of superiority and placed on a level equal to that of the GFS. The NAM needs to be chucked entirely. And short range models, such as the RGEM should always be taken seriously.

People, mets included it seems, had their incredible love of snow bow down too strongly to the Euro and neglect models that went against their hopes and wishes. All while giving an unreasonable amount of credence to a bipolar model known as the NAM. Consider this storm a wake up call.

[Derecho!]

parachuting in as a tropical weenie to note that I don't quite get how models are viewed and used for winter storms.

it's abundantly clear that consensus models are greatly superior to any of the individual models that make them up for tropical track forecasts, even very broad consensus models.

Yet for winter wx it's always viewed as some sort of Mano-a-mano death match where one model is "great" and the rest are deemed "garbage" (though this may change every 12 hours).

Granted, TCs have a discrete trackable center so it's easier, but isn't it pretty clear that what actually happened with this storm was basically the average of the major model ouputs?

I know it seems boring and wimpy to simply assert a given winter storm will do something that is a blend of all the models but you'll end up more accurate than the current view of models as some sort of steel cage death match.

[tmagan]

I have followed the RGEM ensembles and never did any member produce any excessive QPF for NYC for this storm.

 

Here is the 12Z Monday RGEM ensemble for KNYC:

 

[Brian5671]

The Euro/NAM rule did not work this go around. Forecasters bought into their solutions lock stock and barrel.  Usually that combo is deadly within 24-48 hours.  Seeing the RGEM not on board was a red flag...it's one thing to have the GFS out in lala land, but when it had other support, that was the red flag that should have been noticed a bit earlier.

[real]

More people should have seen the red flag yesterday midday when Euro scaled back to like 18 inches in NYC. And GFS was even lower, at about a foot.  And just ditched the NAM completely as an unreliable model for major East Coast storms. And whatever happened to always scaling back some on model precip amounts. Saw many Mets saying yesterday "the time for model watching was over."  But if they would have taken the Euro and GFS yesterday at midday, and scaled back amounts some, they would have had a forecast of about 8 to 16 inches for NYC, which would have been a reasonable call.

 

Also, Upton did seem to jump in way too early with the Blizzard Warnings on Sunday. I get the need to warn people -- especially  considering it was a weekend and storm that was sneaking up on everyone. But throwing out 2 to 3 feet as a first call likely influenced other meteorologists as well.  The forecasts on Sunday should have stayed Blizzard Watch, and maybe toss out language that up to 20 inches was possible but forecasts would need refined on Monday

[Drz1111]

parachuting in as a tropical weenie to note that I don't quite get how models are viewed and used for winter storms.

it's abundantly clear that consensus models are greatly superior to any of the individual models that make them up for tropical track forecasts, even very broad consensus models.

Yet for winter wx it's always viewed as some sort of Mano-a-mano death match where one model is "great" and the rest are deemed "garbage" (though this may change every 12 hours).

Granted, TCs have a discrete trackable center so it's easier, but isn't it pretty clear that what actually happened with this storm was basically the average of the major model ouputs?

I know it seems boring and wimpy to simply assert a given winter storm will do something that is a blend of all the models but you'll end up more accurate than the current view of models as some sort of steel cage death match.

I think NYC/NJ ended up significantly below model consensus from 24-36 hrs out. At that point even the "low" model cluster was spitting out 12-15". There's a lesson about models here, but this was also just a whiff.

(the big difference is in tropical, no one fliches if a forecast is off by 100 miles, or a storm comes in at a cat lower than forecast, because that level of uncertainty is assumed and expressed by the media/NHC.)

[winterwarlock]

The Nam should not even be looked at yet people are always riding it as legit

[BxEngine]

More people should have seen the red flag yesterday midday when Euro scaled back to like 18 inches in NYC. And GFS was even lower, at about a foot. And just ditched the NAM completely as an unreliable model for major East Coast storms. And whatever happened to always scaling back some on model precip amounts. Saw many Mets saying yesterday "the time for model watching was over." But if they would have taken the Euro and GFS yesterday at midday, and scaled back amounts some, they would have had a forecast of about 8 to 16 inches for NYC, which would have been a reasonable call.

Also, Upton did seem to jump in way too early with the Blizzard Warnings on Sunday. I get the need to warn people -- especially considering it was a weekend and storm that was sneaking up on everyone. But throwing out 2 to 3 feet as a first call likely influenced other meteorologists as well. The forecasts on Sunday should have stayed Blizzard Watch, and maybe toss out language that up to 20 inches was possible but forecasts would need refined on Monday

This is a 100% hindsight is 20/20 post. Ive never seen a pro met discuss cutting back Euro and gfs totals on a consistent basis...not to mention you needed to cut the Euro totals by 80-90% in the areas just northwest of the city.

[The_Global_Warmer]

parachuting in as a tropical weenie to note that I don't quite get how models are viewed and used for winter storms.

it's abundantly clear that consensus models are greatly superior to any of the individual models that make them up for tropical track forecasts, even very broad consensus models.

Yet for winter wx it's always viewed as some sort of Mano-a-mano death match where one model is "great" and the rest are deemed "garbage" (though this may change every 12 hours).

Granted, TCs have a discrete trackable center so it's easier, but isn't it pretty clear that what actually happened with this storm was basically the average of the major model ouputs?

I know it seems boring and wimpy to simply assert a given winter storm will do something that is a blend of all the models but you'll end up more accurate than the current view of models as some sort of steel cage death match.

Fantastic post.

I think social media and these forums are as also a bit to blame.

We all do it.

The anticipatory high for snow lovers is powerful.

Powerful enough to destroy rational judgment.

[BxEngine]

Fantastic post.

I think social media and these forums are as also a bit to blame.

We all do it.

The anticipatory high for snow lovers is powerful.

Powerful enough to destroy rational judgment.

Again...is cutting euro output 90% to get your true predicted total "rational"?

[tmagan]

The Blizzard of 1996 produced an extreme gradient on its northwestern side.

[chietanen]

I wonder how much of a factor the ice storm last week was in going worst case scenario for a forecast. That was somewhat of a bust by the NWS (among others) and there were thousands of accidents and several deaths.

[The_Global_Warmer]

Again...is cutting euro output 90% to get your true predicted total "rational"?

Why was the gfs,rgem, rap, hrrr,ukmet, and ggem all ignored?

The euro isn't that much better.

By mid-day yesterday there were very clear indications things were not going to be like the Nam

By 18z the gfs and rgem both came in way East.

Then the rap and hrrr from 18z or so onwards were even further East.

There were countless runs leading up to this as well.

[eyewall]

It was clear there was too much "modelology" going on and there was a slow response to radar/sat/ob trends last night. Not enough "looking out the window" so to speak.

[wxman]

I think it's simple, ANY snow forecast for NJ that relies on any type of a capture scenario needs to be discounted by 80%.  No matter what any model shows, they don't work out for us, March 2001 and now this (problem is there is always '78 that shows it's possible).

[Clinch Leatherwood]

It was clear there was too much "modelology" going on and there was a slow response to radar/sat/ob trends last night. Not enough "looking out the window" so to speak.

 

Very true. 

 

The bottom line is there have been at least three negatively tilted troughs with a stalled low since later February 2014.  I believe 1 or 2 produced blizzard warnings for the Cape based on the Euro while others particularly the RGEM were not biting on the severity and closed off later etc.   Each time the euro was too quick to develop and capture the low resulting in a displacement SW.  Each instance involved the model playing catchup right to the end 0h.   This instance a season later -   I really believe in the upgrade they tweaked something for the worse.

 

We got away from biases as models were upgraded and the Euro had few that were identifiable aside of holding back energy in the 4 corners.  I'm about convinced we can add this now as a demonstrable bias, and one that needs to be considered particularly when it's the west outlier in a severe storm while the RGEM is east.

[Brian5671]

I think it's simple, ANY snow forecast for NJ that relies on any type of a capture scenario needs to be discounted by 80%.  No matter what any model shows, they don't work out for us, March 2001 and now this (problem is there is always '78 that shows it's possible).

that's a good point. The capture often occurs a bit later than modeled and that makes a big difference for areas S and W of NYC especially

[ILoveWinter]

At least some lessons were learned, I mean its easy to monday quarterback in a situation like this.  But obviously one lesson would be that you cant ignore the consensus of a majority of global models (despite their reputation) especially when they have short term model support.  I did have a bit of uneasiness that there was so much disparity between the Euro/NAM vs the rest even as the storm was at zero hour, but like many others I brushed those doubts aside and fully expected a blizzard.  On to the next one!

[WintersGrasp]

In my opinion, there were many things that could and should have been handled differently. Here is my take :

So the Euro suddenly started showing a huge snowy solution for all of our area. It was actually said then to supposedly have convective feedback issues. As it continued to show that similar solution for another run and then again, it suddenly did not have convective feedback issues ? I said from the beginning as the hype began, I am not comfortable being in that heavy snow zone with such a cutoff without much other model support 3 days before the event......as this continued to play out, other models that did support a similar but less extreme scenario gradually moved away from it to put our area on the western fringe of heavy snow and in many model runs, the extreme western fringe.....

GFS, a model that although has proven to not be the most reliable, is still the model that many of our forecasts are based on.....was tossed.

The GGEM showed some snowy solutions, but then moved away from it quite strikingly, and was seemingly tossed as well.

.....The "kicker" here for me is that the RGEM (which goes out to only 48 hours and in my experience has proven to be awesome many times as we get closer and closer to an event) continued to paint a less-snowy solution for western portions of the area over and over and over again.....at times showing a foot or so but then backing on that call to approximately the totals that are coming out now. So yes, the RGEM was also tossed it seems.

and then we have the NAM.....unanimously known to be the most unreliable model in the world. To the NAM, we average 250-500+ inches of snow per year in and around the NYC area, with 750 Inches NW of the city.....so the NAM wasn't biting on the storm. Then.....it had a "NAMish" run and gave our areas 2-4 feet of snow. Suddenly, it was welcomed with open arms and put alongside the euro as the "models of choice".....I guess because it supported the original model of choice?

When Upton pulled the trigger overnight Saturday/Sunday morning, the forecast immediately started calling for 20-30 inches of snow. Snow maps showed 24-36 inches of snow. As all other models continued to move even further away from such a solution, the predictions were scaled back to 18-24". Then (which i really, really cannot understand) as the euro at its 12z run on Monday (the day of the storm's beginning) showed a significant jump east with the precip, the NAM decided to do one of it's crazy runs and suddenly even though the model of choice (euro) was moving AWAY from it's solution, the unreliable NAM was suddenly taken from the "tossed" pile and placed at the top of the preferred models list. RGEM stood its ground, becoming less and less favorable for our areas (further west the least snow yet) and still no attention was given to the formerly revered, good old RGEM (even though it had support from basically EVERY SINGLE OTHER MODEL, with even the Euro moving away from huge snows in NJ/NYC/LHV/parts of LI, etc.....so what to do at this point when NO model (other than the NAM) seem to be trending or moving toward a snowier solution?.....bump totals back up to 20-30" of course ! Short term models CONSISTENTLY did not like the snowy solutions. Run after run after run of the RAP/HRRR showed the bands not being able to push too far east past LI. I heard something about them not being able to handle this type of storm ? I can't help but feel like so many theories were made up and assumptions made just to support the snowy scenario. The rest is history. About 5" later, here is my storm.

I'm usually pretty "calm" on here, but as I expressed my views in the main thread as to why I just did not see this panning out, I was met with a ton of criticism by hobbyists and some forecasters. That is fine, although the reasoning wasn't too sensible and became stale after a while. The euro being best, all of the other models supposedly not being able to handle the dynamics, snow ratios will be so great that a lower amount of liquid will turn into 20" of snow still, etc etc.....while all the while saying "you don't just make forecasts based on weather models" (even though ironically those who said it were basing their thinking on a single weather model and ignoring all others). This is a most frustrating situation as I feel the writing was on the wall well before any of this had to happen. Though I was cautious and had that "this just does not make sense and probably will not work out" in my mind the entire time, eventually I just kind of said to myself "I know so little compared to the actual mets.....there must be really good reason to be ignoring every single other weather model in favor of the euro, and when it started caving too.....they must have some really good reason to suddenly favor the NAM".

I really never thought we'd see another bust like March 2001, but as this one was developing, it was clear that it could be unfolding again, and did

[pazzo83]

Let's call this what it is: reckless forecasting and apparently a wholesale inability to communicate probability and risk (in terms of the forecast) to the general public.  This probably cost the NY metro area upwards of $1bn in terms of lost revenues/wages plus the costs of shutting down a metro area of 20+ million.

[PB GFI]

Let's call this what it is: reckless forecasting and apparently a wholesale inability to communicate probability and risk (in terms of the forecast) to the general public.  This probably cost the NY metro area upwards of $1bn in terms of lost revenues/wages plus the costs of shutting down a metro area of 20+ million.

Chris I could not disagree more .

A few things the GFS had just gotten beat badly on the east coast last Sunday where 2 inches of rain fell and the GFS was wide right until 48 hours out and the Euro saw it from  5 days out and never moved .

So the def SE bias had to be in BOX OKX and DIX minds .

 

The best skill score model printed out 2.5 inches of liquid for 10 straight runs  and a lot of that coming in a 12 to 18 period . So that meant incredible rates and a repeat of 13 needed to be avoided . Forget the NAM for a moment , during NEMO so much snow fell in Nassau and Suffolk county many were stranded on the LIE and it became a matter of life and death for some .

 

When you see that much snow forecasted their first responsibilities is to the general public and the obligation is to make sure first responders can get where they need to .

24 to 36 was what the EURO was forecasting for 2.5 days . It was not a 1 off run .  The forecast busted from Nassau county west .

There is still 20 in Nassau and 30 will show up in Suffolk . 

We are talking about 75 miles and if 20 -30 fell at KNYC and it was that close I think they made the right call .

 

We weather nuts hated the outcome if you are in the city and points west but it was that close and they has no choice .

You cant play catch up if 30 inches  and rates of 2 to 4 inches an hour are on the table

[gravitylover]

Modelology bit everybody in the behind.  Sure a few models picked up on the storm potential a week or so ago and guess what?  We got a storm.  Right "on time" and for all intents a fairly significant one.  This is what a weather model is supposed to do, alert mets (and us) to the potential but then meteorology NEEDS to take over.  There were a few warning signs early yesterday that the storm was going to move just a bit further offshore than the hype was spewing but once that machine gets rolling it can't be stopped.  

 

So was it a bust?  No.  Was it hyped to the point that it made things difficult and maybe unpleasant for a lot of people?  Yes.  Would we rather have had people stuck out in it and then our first responders having no choice but to endanger themselves to help others?  I think not.  After all is said and done I bet a lot of municipalities will choose to do the same thing again.  Not having full town crews working double and triple time probably saved a s**t ton of money, not running a fleet of trucks all night saved a bunch of fuel and road salts and the extra taxes collected on all of the alcohol sold can't be a bad thing 

[pazzo83]

Chris I could not disagree more .

A few things the GFS had just gotten beat badly on the east coast last Sunday where 2 inches of rain fell and the GFS was wide right until 48 hours out and the Euro saw it from  5 days out and never moved .

So the def SE bias had to be in BOX OKX and DIX minds .

 

The best skill score model printed out 2.5 inches of liquid for 10 straight runs  and a lot of that coming in a 12 to 18 period . So that meant incredible rates and a repeat of 13 needed to be avoided . Forget the NAM for a moment , during NEMO so much snow fell in Nassau and Suffolk county many were stranded on the LIE and it became a matter of life and death for some .

 

When you see that much snow forecasted their first responsibilities is to the general public and the obligation is to make sure first responders can get where they need to .

24 to 36 was what the EURO was forecasting for 2.5 days . It was not a 1 off run .  The forecast busted from Nassau county west .

There is still 20 in Nassau and 30 will show up in Suffolk . 

We are talking about 75 miles and if 20 -30 fell at KNYC and it was that close I think they made the right call .

 

We weather nuts hated the outcome if you are in the city and points west but it was that close and they has no choice .

You cant play catch up if 30 inches  and rates of 2 to 4 inches an hour are on the table

 

I understand, but if you are going to call for a near-certain historic storm, you better have near-unanimous model agreement.

[87storms]

from the outside looking in, i thought the western edge was too sharp even on the euro to discredit the other models.  it certainly didn't appear to be a lock.  even going by the euro, i think 12-18"+ would have been a more appropriate call to start things off with adjustments as needed.

[NYCSNOWMAN2020]

At least some lessons were learned, I mean its easy to monday quarterback in a situation like this. But obviously one lesson would be that you cant ignore the consensus of a majority of global models (despite their reputation) especially when they have short term model support. I did have a bit of uneasiness that there was so much disparity between the Euro/NAM vs the rest even as the storm was at zero hour, but like many others I brushed those doubts aside and fully expected a blizzard. On to the next one!

well said