January Medium/Long Range: A snowy January ahead?

[VA Mad Man]

2 hours ago, adelphi_sky said:

That would be an awesome snow chase. Sipping Sazerac hot chocolate eating beignets watch the snow fall on Bourbon Street. 

THIS!

[Terpeast]

25 minutes ago, Heisy said:

Icon trending towards that Monday/Tuesday overrunning potential. Mitch it’s really amazing how the models generally just follow the Ai when it comes to H5. I wouldn’t use it for temps or precip total, but man H5 is on point





.

For now, this is the best use case for AI in weather. We are not there yet for mesoscale. 

[JenkinsJinkies]

DCA needs 5.4 inches to match climo, is that possible before the month is out?

[psuhoffman]

33 minutes ago, osfan24 said:

Just give us our long-awaited HECS and we can all relax and just enjoy whatever else winter has in store for us.

We’ve had one non Nino HECS in the last 40 years and 2 in the last 75 years!  We’ve had 7 in a Nino even though those years only make up like 30% of the winters!   

I think it’s safe to say we shouldn’t expect an HECS in any non Nino year. We’re missing one of the two key ingredients needed to get those level events, a strong STJ.  The other being a really strong well placed 50/50 to keep the cold in place as the STJ moisture attacks, typically from blocking but we’ve had a couple where there was a 50/50 without blocking. 

[Maestrobjwa]

27 minutes ago, psuhoffman said:

We’ve had one non Nino HECS in the last 40 years and 2 in the last 75 years!  We’ve had 7 in a Nino even though those years only make up like 30% of the winters!   

I think it’s safe to say we shouldn’t expect an HECS in any non Nino year. We’re missing one of the two key ingredients needed to get those level events, a strong STJ.  The other being a really strong well placed 50/50 to keep the cold in place as the STJ moisture attacks, typically from blocking but we’ve had a couple where there was a 50/50 without blocking. 

I'm still a little unclear about the STJ. So, we've had seveal southern waves this year...but those are NOT stj but something else? If am active gulf not always an "stj" per se?

[stormtracker]

Nice overrunning event on the GFS for next week

[87storms]

That would be an awesome snow chase. Sipping Sazerac hot chocolate eating beignets watch the snow fall on Bourbon Street. 

You ain’t lying. Was just down there last week. The food was legit. Went to bourbon st one night…it’s like an adult boardwalk lol. I enjoyed Sazerac, though honestly thought beignets were a bit overrated. Pretty impressive turnout for AMS…heard they broke an attendance record.

[mitchnick]

2 minutes ago, stormtracker said:

Nice overrunning event on the GFS for next week

Same idea as the AI, but a little wetter and 12hrs later. Close enough. 

[stormtracker]

Just now, mitchnick said:

Same idea as the AI, but a little wetter and 12hrs later. Close enough. 

Wish it was a lil more colder.   timing is good on that first slug of moisture.  Looks like a 2-4/3-5 type deal

[mitchnick]

234hrs Gfs is getting ready for an arctic dump into the Conus.

https://www.pivotalweather.com/model.php?rh=2025012012&fh=228&dpdt=&mc=&r=na&p=850th&m=gfs

[87storms]

234hrs Gfs is getting ready for an arctic dump into the Conus.
https://www.pivotalweather.com/model.php?rh=2025012012&fh=228&dpdt=&mc=&r=na&p=850th&m=gfs

We might be getting too good at snow to avoid another week of tracking.

[clskinsfan]

27 minutes ago, Terpeast said:

For now, this is the best use case for AI in weather. We are not there yet for mesoscale. 

The future of AI in weather is going to be really interesting to watch. AI's increase in power over just the past year has been eye opening to me. I thought it would be some fringe thing. But no. It is now fully capable of programming at levels above the best programmers in the world. It's growth has been insane. 

[psuhoffman]

20 minutes ago, Maestrobjwa said:

I'm still a little unclear about the STJ. So, we've had seveal southern waves this year...but those are NOT stj but something else? If am active gulf not always an "stj" per se?

We need a STRONG stj wave to get an HECS.  That’s the difference in many events between a 6-12” event and those 20”+ snowstorms.   We rarely get those juiced up southern stream bowling bowls throwing some 2”+ precipitatable water moisture feed at us in non Nino years. 
 

Further north they can get a hecs without that if the northern streak phases and bombs a storm and it develops a mature CCB with a crazy deform.  But we’re too far south for that.  We can sometimes get the southern edge of that and in many of our HECS storms we do get an additional 6-12” from that to put us over the top but we had 12+ already just from the STJ overrunning precip.  We are too far sorry typically to get 20” from a NS phasing coastal bomb unless it has a strong STJ component initially.  

 

[psuhoffman]

23 minutes ago, 87storms said:

You ain’t lying. Was just down there last week. The food was legit. Went to bourbon st one night…it’s like an adult boardwalk lol. I enjoyed Sazerac, though honestly thought beignets were a bit overrated. Pretty impressive turnout for AMS…heard they broke an attendance record.

Ya it would be great, until they all freak out over 3” and the town closes like it’s Armageddon. 

[Darkstorm]

8 minutes ago, clskinsfan said:

The future of AI in weather is going to be really interesting to watch. AI's increase in power over just the past year has been eye opening to me. I thought it would be some fringe thing. But no. It is now fully capable of programming at levels above the best programmers in the world. It's growth has been insane. 

Speaking from experience. It's junior/mid-level at best with programming. It's not reliable and can produce some junky and unsecure code. And man oh man can it hallucinate. I'd argue its a good complement and a productivity booster versus a replacement of what humans are capable of.

[JenkinsJinkies]

49 minutes ago, psuhoffman said:

We’ve had one non Nino HECS in the last 40 years and 2 in the last 75 years!  We’ve had 7 in a Nino even though those years only make up like 30% of the winters!   

I think it’s safe to say we shouldn’t expect an HECS in any non Nino year. We’re missing one of the two key ingredients needed to get those level events, a strong STJ.  The other being a really strong well placed 50/50 to keep the cold in place as the STJ moisture attacks, typically from blocking but we’ve had a couple where there was a 50/50 without blocking. 

So if this PDO change sticks could we be in for one the next time we have a Niño?

[87storms]

Ya it would be great, until they all freak out over 3” and the town closes like it’s Armageddon. 

Can’t blame them lol. There’s gonna be some cool pics coming out of there, assuming the snow delivers.

[Terpeast]

16 minutes ago, clskinsfan said:

The future of AI in weather is going to be really interesting to watch. AI's increase in power over just the past year has been eye opening to me. I thought it would be some fringe thing. But no. It is now fully capable of programming at levels above the best programmers in the world. It's growth has been insane. 

It really has. I do some of this stuff for a living and there’s always some new thing every week. 

[87storms]

Speaking from experience. It's junior/mid-level at best with programming. It's not reliable and can produce some junky and unsecure code. And man oh man can it hallucinate. I'd argue its a good complement and a productivity booster versus a replacement of what humans are capable of.

The euro discussion I attended at ams mentioned ai as more of a hybrid solution, if memory holds (last week was a brain dump). I’d be curious to see how much overfitting would impact the models. Might result in more missed fluke events (just brainstorming).

[adelphi_sky]

1 minute ago, Terpeast said:

It really has. I do some of this stuff for a living and there’s always some new thing every week. 

I am in IT and the amount of DevOps tools that are released every year makes your head spin. Before it was the regular Microsoft, Apache, Oracle software. Now there's hundreds of DevOps software. When I start a new project someone mentions that they need this software. Something I never heard of before. It's always something new.

[Terpeast]

Just now, adelphi_sky said:

I am in IT and the amount of DevOps tools that are released every year makes your head spin. Before it was the regular Microsoft, Apache, Oracle software. Now there's hundreds of DevOps software. When I start a new project someone mentions that they need this software. Something I never heard of before. It's always something new.

I suspect that we’re in the bubble phase of the AI hype though. Two years from now, all that froth will dissipate and the so and so things we’d never heard of will fall away back into obscurity while we return to the big boys like Microsoft, Apache, etc. 

[87storms]

It really has. I do some of this stuff for a living and there’s always some new thing every week. 

I think the “ai” in this case is more about utilizing ml (data driven vs physics), not so much using ChatGPT to write code.

[Terpeast]

1 minute ago, 87storms said:

I think the “ai” in this case is more about utilizing ml (data driven vs physics), not so much using ChatGPT to write code.

Yep that’s what I mean. ChatGPT’s code is only basic/educational at best, but nowhere near production-level quality. ML is more in my wheelhouse

[clskinsfan]

4 minutes ago, Terpeast said:

Yep that’s what I mean. ChatGPT’s code is only basic/educational at best, but nowhere near production-level quality. ML is more in my wheelhouse

Yes. Open source AI is not where the future is. We have no idea what someone is working on behind closed doors right now. But at the pace of AI's improvement the future will be amazing. If not a little scary imo. 

[87storms]

9 minutes ago, Terpeast said:

Yep that’s what I mean. ChatGPT’s code is only basic/educational at best, but nowhere near production-level quality. ML is more in my wheelhouse

Yea, it’s pretty interesting. Been trying to better understand the math behind svn’s and neural networks. Understanding the loss function and gradient descent are key pieces, but it’s a thick topic, that’s for sure.

Anyhow, heard it might snow in New Orleans lol

[DDweatherman]

11 minutes ago, 87storms said:

I think the “ai” in this case is more about utilizing ml (data driven vs physics), not so much using ChatGPT to write code.

I was going to say, these are ML more than they are AI to me. Training models. 

[anotherman]

People are going to excitedly open this thread hoping that a big storm is on the models only to find it’s a discussion on AI.

[Darkstorm]

14 minutes ago, Terpeast said:

I suspect that we’re in the bubble phase of the AI hype though. Two years from now, all that froth will dissipate and the so and so things we’d never heard of will fall away back into obscurity while we return to the big boys like Microsoft, Apache, etc. 

Legitimately there is an entire cycle that actually follows that from Gartner. We are definitely in the hype cycle where everyone pegs it as the solution in about 18-24 months is where we actually start to see the legitimate use cases pop up. Same thing happen with crypto/block chain.

[Ralph Wiggum]

33 minutes ago, JenkinsJinkies said:

So if this PDO change sticks could we be in for one the next time we have a Niño?

Absolutely would give a solid chance.

[87storms]

I was going to say, these are ML more than they are AI to me. Training models. 

Yea, the term AI is loose…even ChatGPT is based off ml, but for forecast models the output is a forecast, not a snippet of Python.