Outta gas and Outta Time: Early March Winter Storm finale

[Ji]

5 minutes ago, bncho said:

I lost 3" of snow since 0z

thats the issue---its not the track....the wave is dying

[MillvilleWx]

4 hours ago, snowfan said:

WPC got this out before the 6Z GFS OP. LOL

 

Our products have deadlines, so we can’t adjust the overnight forecast with the 06z guidance. If things remain like they are, it’ll be reflected in the next update. 

[wasnow215]

1 minute ago, MillvilleWx said:

Our products have deadlines, so we can’t adjust the overnight forecast with the 06z guidance. If things remain like they are, it’ll be reflected in the next update. 

Do you agree that models are "overcooking" a storm and it won't be much for anyone?

[MillvilleWx]

Just now, wasnow215 said:

Do you agree that models are "overcooking" a storm and it won't be much for anyone?

Took some time away from models after a crazy 7 day stretch at the winter desk. Just from a glance, I haven’t been too enthused. Still a shot at some snow, but max potential is probably 2-4”. Need something with more umph this time of year to really get anything appreciable. Judging by the ensembles after this one, I’m gearing up for spring mode. Baseball is on the horizon and I’m ready to tackle some outdoor walks, hiking, and meals outside. I’ll always welcome some snow though, so if it happens, I’ll accept. 

[rjvanals]

Icon brings some snow north of I-70 

[Weather Will]

WB 12Z ICON

[mitchnick]

1 minute ago, Weather Will said:

WB 12Z ICON

Similar to 6z EuroAI 

[psuhoffman]

3 hours ago, mitchnick said:

See my post re EuroAI.  Lol

These models have done nothing but ruin this winter.

Back in the fall several people made the correct point that this winter would be heavily northern stream dominant, and given the state of the northern stream lately (fast and chaotic) it would mean whatever snow threats we did end up getting would be unlikely to resolve at any significant leads.  Several people referenced 2013-14 as an example, when we got numerous snow events but they were far from resolved until inside 24 hours!  Some, like early Dec 2013, was a positive bust in the nowcast!  There was a negative bust for all except the PA line people in early Feb also.  Both cases the going forecast as the storm began was WAY off...one turned snowier and one not.  

But I've found it entertaining and sometimes frustrating to see all the "why are the models sucking arse" posts all winter when this was a known thing coming in, it was predicted over and over...and yet people still expected 100 hour forecasts to end up accurate, knowing that this pattern was not one models would resolve details on at any lead let along 100 hours plus.  

2 hours ago, Terpeast said:

On a CWG post on facebook I saw a quote from Wes Junker saying that the atmosphere doesn't really support a storm here, the models are "overcooking" it rn. Upside is only a dusting and maybe a few lucky inches. He's one I respect and knows what he's talking about, so that dampened any remaining enthusiasm for tracking this (not that I had much to begin with).

 

big picture he is totally right.  But...sometimes late in the season this little boundary waves can be sneaky good given the increased baroclinicity.  This also seems to be increasing in recent years...maybe elephant related?  While it helps us less and less often, when we do get a flush hit from a weak little boundary wave they sometimes are way more than you would think just looking at the synoptics.  That isn't something you would forecast from range though, its just something to root for as a "sometimes this can be sneaky good if we get lucky" thing.  

[psuhoffman]

Icon has been better this winter, from my observations.  Not great, no model has been great, but its not been any further off than anything else and frankly has been more consistent than some of the other guidance for several events.  

[87storms]

10 minutes ago, MillvilleWx said:

Took some time away from models after a crazy 7 day stretch at the winter desk. Just from a glance, I haven’t been too enthused. Still a shot at some snow, but max potential is probably 2-4”. Need something with more umph this time of year to really get anything appreciable. Judging by the ensembles after this one, I’m gearing up for spring mode. Baseball is on the horizon and I’m ready to tackle some outdoor walks, hiking, and meals outside. I’ll always welcome some snow though, so if it happens, I’ll accept. 

After 4 months of a statistically cold winter, you’re ready for spring? Unacceptable. That next week system doesn’t look as wavy as I’d like, but nice to see a cutoff low of sorts out west. They still need precip.

 

[TSSN+]

16 minutes ago, MillvilleWx said:

Took some time away from models after a crazy 7 day stretch at the winter desk. Just from a glance, I haven’t been too enthused. Still a shot at some snow, but max potential is probably 2-4”. Need something with more umph this time of year to really get anything appreciable. Judging by the ensembles after this one, I’m gearing up for spring mode. Baseball is on the horizon and I’m ready to tackle some outdoor walks, hiking, and meals outside. I’ll always welcome some snow though, so if it happens, I’ll accept. 

 Bring on the shorts and tshirts! 

[baltosquid]

4 minutes ago, psuhoffman said:

Icon has been better this winter, from my observations.  Not great, no model has been great, but its not been any further off than anything else and frankly has been more consistent than some of the other guidance for several events.  

Feels like at times it is too dry/weak but it’s generally pretty consistent with that imo so makes it easier to work with than if it was erratic. Definitely worthwhile if not quite as much as its peers.

[LordBaltimore]

7 minutes ago, 87storms said:

After 4 months of a statistically cold winter, you’re ready for spring? Unacceptable. That next week system doesn’t look as wavy as I’d like, but nice to see a cutoff low of sorts out west. They still need precip.

 

My AC needs a new capacitor and I'm lazy so it'll probably take me a few weeks at least to fix. Would be perfect timing for a March/April full of backdoor cold-fronts

[psuhoffman]

3 minutes ago, baltosquid said:

Feels like at times it is too dry/weak but it’s generally pretty consistent with that imo so makes it easier to work with than if it was erratic. Definitely worthwhile if not quite as much as its peers.

Maybe this is perception bias, but when it came to my ability to predict what a significant storm would end up doing from the guidance, it was easier for me in the late 90s and early 2000s with the old school MRF/AVN/GGEM/ECMWF and short range ETA/NGM.  Those models were way way way less accurate, but they tended to be less accurate in a more consistent way.  They each had very very very universally consistent bias errors and if you knew how to correct for them they were useful.  Now...they are all more accurate in that they are more likely to be closer to the actual truth.  But they are much higher resolution and their errors tend to be less consistently in the same direction.  This makes it much harder to correct for them and determine what their errors are.  Not trying to be controversial here, and I could be wrong...but at times I felt it was easier to forecast using the models 20 years ago in the medium range than now.  

[mitchnick]

GfsAi miss to the south 

[TSSN+]

2 minutes ago, mitchnick said:

GfsAi miss to the south 

Gfs is coming north 

[stormtracker]

So far, GFS seems a little wetter/better than 6z

[stormtracker]

Much improved.  The see saw continues 

[TSSN+]

1 minute ago, stormtracker said:

So far, GFS seems a little wetter/better than 6z

Ya it comes north but needs a few more bumps north.  

[87storms]

Gfs was pretty atrocious for today’s system until the last minute. Whatever the euro says is still the gold standard imo. Many will disagree, but for mby at least, it also performed better than the gfs for Sunday’s rainstorm.

[stormtracker]

It's the Bob Chill show...starring Robert Chill! 

[87storms]

This was the gfs yesterday for today’s system: https://www.tropicaltidbits.com/analysis/models/?model=gfs&region=neus&pkg=mslp_pcpn_frzn&runtime=2026022512&fh=30

Talk to me when the euro shows me a snowstorm.

[TSSN+]

Cue up the Benny Hill music 

 

[Bob Chill]

7 minutes ago, stormtracker said:

It's the Bob Chill show...starring Robert Chill! 

Thank you for the intro good sir. With great honor I accept my 6" of digital snow and bare ground award on Monday. 

[psuhoffman]

RGEM did not look like it was going to be suppressed at 84 hours.  Guess we will see with the GGEM soon 

[mitchnick]

4 minutes ago, TSSN+ said:

Cue up the Benny Hill music 

 

It's headed back north with the last frame.

[TSSN+]

1 minute ago, psuhoffman said:

RGEM did not look like it was going to be suppressed at 84 hours.  Guess we will see with the GGEM soon 

I’ll tell ya right now. It’s trash

 

[mitchnick]

1 minute ago, Bob Chill said:

Thank you for the intro good sir. With great honor I accept my 6" of digital snow and bare ground award on Monday. 

You can have this one. I'll  take the one the first week of April. Lol

[psuhoffman]

1 minute ago, TSSN+ said:

I’ll tell ya right now. It’s trash

 

It really dampens the wave around 90-96 hours.  It doesn't eject enough energy, most of it hangs back, the wave dies with no mid and upper level support at all as it slides east.  

[87storms]

39 minutes ago, psuhoffman said:

Maybe this is perception bias, but when it came to my ability to predict what a significant storm would end up doing from the guidance, it was easier for me in the late 90s and early 2000s with the old school MRF/AVN/GGEM/ECMWF and short range ETA/NGM.  Those models were way way way less accurate, but they tended to be less accurate in a more consistent way.  They each had very very very universally consistent bias errors and if you knew how to correct for them they were useful.  Now...they are all more accurate in that they are more likely to be closer to the actual truth.  But they are much higher resolution and their errors tend to be less consistently in the same direction.  This makes it much harder to correct for them and determine what their errors are.  Not trying to be controversial here, and I could be wrong...but at times I felt it was easier to forecast using the models 20 years ago in the medium range than now.  

The bias/variance trade-off in ML is probably one of the most important things anyone can learn when building models. I imagine it’s even more of a factor as model resolution improves and training sets get larger. Would be an interesting topic to explore.