November 21-23, 2023 | NE Winter Storm Potential

[PA road DAWG]

2 hours ago, Penn State said:

I was just looking at the CPC 6-10 Day Analogs.. December of 2002 is rising on the charts. There was a pretty good snowstorm around December 5th of ‘02. Here’s a link to some information, maps, and data. Below is the analog. Might set up well for just after this period TBH. 

I’ll never ever forget the 12/25/02 nor’easter!   That was a good one.  

[MaineJay]

7 hours ago, 1816 said:

This is the biggest key in my opinion: "Matthew Chantry, who works on machine learning forecasting at the ECMWF, says that Google DeepMind’s GraphCast has emerged as the strongest of the AI contenders. “Over time it will consistently be just a little bit better,” he says."

It mentions in this article how even ai missed the bus on Otis. As well as some other shortcomings. But when you have a system that can dive into every facet and angle and literally learn from its mistakes, the sky is the limit for improvement. 

I think a part that's missing here, these machine learning systems are trained on the data from the supercomputers.  I don't believe it can be an either/or situation.  I think if it more of an extension of the current models.  I don't think the AI can do it in it's own. There's also human hands guiding it along the way.

 I worry that s lot of this is designed to take the "responsibility" away from people.  If machine learning makes a mistake, who's at fault?  The programmers can say the data it was trained on is "bad", so it's effectively someone else's fault.

   I remember when Panasonic was touting it's model was beating all the other models by incorporating more airplane data.  If memory serves, IBM acquired the tech, and developed deep thunder. I'm not sure it's even run anymore.

   Do I think ML and AI can improve forecasting? Yes. But there are still the same limitations of grid points and knowing the state of the atmosphere. Initial conditions at precisely 0z and 12z will always be impossible to know. 

 I'm happy that we can't accurately predict the weather 5 days out. Life would be boring if we knew everything in advance. Knowing everything ahead of time isn't living, it's going through the motions.  

 I fear we are driving head first.

 

  

[Penn State]

On the 12z GFS.. I'm almost getting a CAD feeling for central PA. There may be enough shallow cold to start things as frozen before a changeover to rain, assuming this moves to the West of the Appalachians. 

[clm]

[TLChip]

15 hours ago, 1816 said:

It mentions in this article how even ai missed the bus on Otis.

To be fair no model saw Otis. I think one sniffed it out for a run. Will be one for the books, location, strength, how quickly it developed, all uncommon in that area. 

[LUCC]

3 hours ago, TLChip said:

To be fair no model saw Otis. I think one sniffed it out for a run. Will be one for the books, location, strength, how quickly it developed, all uncommon in that area. 

Otis? I've been out of the loop apparently. 

[StretchCT]

A lot of wind with this one.  That 81mph gust is over Montauk.  Gusts to 40 can shut down airports.  It's the busiest travel day and possibly DC-Portland will be interrupted.  CLE and PIT too. 

 

Edited November 15, 2023 by StretchCT

[StretchCT]

Here's the Euro.  It doesn't bomb out the storm like the GFS does. 

 

CMC is pretty high too, but not as alarming. 

[StretchCT]

Graphcast's take on this threat 😴

[Penn State]

I would love for the primary low to give way to the coastal sooner.. but there’s no evidence of that occurring to my eye. 

[Penn State]

18z GFS.. The snow map shows that front end snow in parts of the Shenandoah Valley and SCPA. That’s what I was thinking with the CAD scenario earlier. New England does well! 

[MaineJay]

2 hours ago, LUCC said:

Otis? I've been out of the loop apparently. 

Hit Acapulco a few weeks ago. Went from a tropical storm to a cat 5 in 24 hours, and one of the strongest in history. No model had it more than a cat 1 at best.   

[Penn State]

ICON is also showing some snow for the interior. 

[Penn State]

15 hours ago, MaineJay said:

I think a part that's missing here, these machine learning systems are trained on the data from the supercomputers.  I don't believe it can be an either/or situation.  I think if it more of an extension of the current models.  I don't think the AI can do it in it's own. There's also human hands guiding it along the way.

 I worry that s lot of this is designed to take the "responsibility" away from people.  If machine learning makes a mistake, who's at fault?  The programmers can say the data it was trained on is "bad", so it's effectively someone else's fault.

   I remember when Panasonic was touting it's model was beating all the other models by incorporating more airplane data.  If memory serves, IBM acquired the tech, and developed deep thunder. I'm not sure it's even run anymore.

   Do I think ML and AI can improve forecasting? Yes. But there are still the same limitations of grid points and knowing the state of the atmosphere. Initial conditions at precisely 0z and 12z will always be impossible to know. 

 I'm happy that we can't accurately predict the weather 5 days out. Life would be boring if we knew everything in advance. Knowing everything ahead of time isn't living, it's going through the motions.  

 I fear we are driving head first.

 

  

I was thinking about this.. In education, we are starting to tackle AI. Instead of rejecting it.. we are being told to embrace it. Students are using it to write papers (which is cheating.. obviously). However, we are able to use it to develop lesson plans, rubrics, presentations, etc. At any rate.. we have been told to ask students questions that involve personal experience because it’s something AI can’t generate. This is where I think good meteorology wins against AI. Experience an intuition.. similar to a Bernie Rayno.. can’t be replaced, even by AI. At least.. I don’t think so.. Thankfully lol 

[Mainiac]

Currently engaged in AI work, I can confidently assert that these Language Model advancements aren't as far along as many believe. Comparing them to driverless cars seems apt: we once envisioned leisurely watching TV while our timeshare smart cars chauffeured us to work a decade ago. However, the reality is starkly different. Smart cars navigate only preplanned routes in warm climates with minimal rain, overseen by a team in a central hub. They're far from being capable of picking us up in a snowstorm on poorly maintained roads without markings in remote areas—it might take decades for them to reach that level.

The initial 99% progress is manageable, yet a functional, widespread smart car hinges on solving the final 1%. This truth extends to AI applications involving human lives.

In domains where satisfactory performance suffices, AI is making transformative strides. However, in contexts where precision and accuracy are paramount, the wait might be extensive.

 

^^ On a completely related side note, I asked ChatGPT to write that for me using a bit of prompt engineering, which will be the only white collar job left in America in 2050, lol.  😉 

 

[Mainiac]

3 hours ago, Penn State said:

 

This is a snow map I can get behind. 

[StretchCT]

3 hours ago, Penn State said:

18z GFS.. The snow map shows that front end snow in parts of the Shenandoah Valley and SCPA. That’s what I was thinking with the CAD scenario earlier. New England does well! 

oh that's close

 

[bigben89]

That was fun while it lasted. Definitely getting cold after Thanksgiving next week. Does look promising for some mood flakes 11/25.

Edited November 16, 2023 by bigben89

[MaineJay]

On 11/14/2023 at 10:57 PM, PA road DAWG said:

I’ll never ever forget the 12/25/02 nor’easter!   That was a good one.  

Noting that winter officially started at 1457z November 14 with this post.

[Penn State]

2 hours ago, bigben89 said:

That was fun while it lasted. Definitely getting cold after Thanksgiving next week. Does look promising for some mood flakes 11/25.

No need to prepare the eulogy just yet! 06Z GFS was still significant for parts of central PA through NE. I think there'll be enough cold air around for some folks to cash in. 

[LUCC]

Definitely going to get cold and a couple small disturbances around after Turkey Day, hopefully we can get some flakes to fly in the S. MidAtl area before Dec. 

Edited November 16, 2023 by LUCC

[StretchCT]

GFS snow trend in SNE. GFS separates into two systems, one through the lakes, another up the coast.  Euro/CMC do not and focus on the GLC only.  

ICON just pushes it over the Apps and reforms over the Ches bay.  Ukie looks the same, maybe a bit more south but warmer

Edited November 16, 2023 by StretchCT

[NWOhioChaser]

I have to drive from Toledo to Bloomingburg, NY on Tuesday so I hope this system will behave until I get there. 

Edited November 16, 2023 by NWOhioChaser

[StretchCT]

12z Ukie still on board with the low just plowing through to the coast. 

 

12z ICON develops the GLC more, the coastal later

 GFS ENS even looking more like the Euro with only a couple of members going coastal.

Will do the 5 day challenge on this tomorrow, but it looks like others are moving to the Euro.

[StretchCT]

No one posted this?