January 2026 Medium/Long Range Discussion

[mitchnick]

I'll be surprised if the Eps aren't close to the Gefs for the threat.

[DMAC98]

5 minutes ago, stormtracker said:

Dang it  runs at light speed on StormVista. Still on TAU 114 on Pivotal

[SnowenOutThere]

Euro was actually extremely close to a complete big dog but unfortunately got squashed by a fast moving lobe in the NS. If that was even 12 hours delayed we would be in business. Good to see though.

 

[winter_warlock]

Ok question for u all  how accurate is the AI GFS and AI euro compared to the others?

[bncho]

Just now, winter_warlock said:

Ok question for u all  how accurate is the AI GFS and AI euro compared to the others?

The AI Euro is a deadly fuckin model, especially inside D4. It was the only model that reliably sniffed out that the Feb 2025 storm was going out to sea, without really wavering.

[Scarlet Pimpernel]

Question for anyone more knowledgeable about the internals of AI models, both the Euro and GFS versions.  Do they tend to "hold onto" or continue to reflect previous runs in general, for a couple of cycles?  That is, if they show a great hit, do they tend to still sort of show that in the next run and perhaps gradually diminish over time?  I ask that in terms of the standard dynamical models which rely upon initial conditions based upon a short forecast from the previous run, so some of the "information" gets propagated so to speak.  This doesn't happen all the time, but there are notable events (ahem, some storm in Dec. 2010!) that held on to "good" runs for us before finally later giving up on that idea.  Vice-versa has also happened I'm sure (but we tend to recall the bad busts!).  I don't know if AI models do the same or if they just take an initial field, and based on the same long training sample they were derived with, come up with whatever forecast regardless of what the immediate preceding run(s) showed.  Not sure if I asked this correctly and hope it makes sense, but I am curious.

[Terpeast]

1 minute ago, winter_warlock said:

Ok question for u all  how accurate is the AI GFS and AI euro compared to the others?

Not sure about the ops, but AIFS ensembles have been doing an excellent job at 500mb. 

[bncho]

1 minute ago, Scarlet Pimpernel said:

Question for anyone more knowledgeable about the internals of AI models, both the Euro and GFS versions.  Do they tend to "hold onto" or continue to reflect previous runs in general, for a couple of cycles?  That is, if they show a great hit, do they tend to still sort of show that in the next run and perhaps gradually diminish over time?  I ask that in terms of the standard dynamical models which rely upon initial conditions based upon a short forecast from the previous run, so some of the "information" gets propagated so to speak.  This doesn't happen all the time, but there are notable events (ahem, some storm in Dec. 2010!) that held on to "good" runs for us before finally later giving up on that idea.  Vice-versa has also happened I'm sure (but we tend to recall the bad busts!).  I don't know if AI models do the same or if they just take an initial field and based on the same long training sample come up with whatever forecast regardless of what the immediate preceeding run(s) showed.  Not sure if I asked this correctly and hope it makes sense, but I am curious.

@high risk my MVP wil likely be able to provide the answer

[winter_warlock]

1 minute ago, Terpeast said:

Not sure about the ops, but AIFS ensembles have been doing an excellent job at 500mb. 

 Oh ok ty bro!!!

[winter_warlock]

2 minutes ago, bncho said:

The AI Euro is a deadly fuckin model, especially inside D4. It was the only model that reliably sniffed out that the Feb 2025 storm was going out to sea, without really wavering.

 Oh wow nice!!!

[bncho]

The GEFS has like 6+" of snow in DC at hr 288 holy shit

[winter_warlock]

28 minutes ago, TSSN+ said:

People cancel winter here in November. What do you expect? 

Funny but very true!!

[WxUSAF]

I wouldn’t pay much attention to the AI models for mesoscale features. But synoptic scale? Euro AI is a cheat code inside D4-5. GFSAI outperforming physics-based GFS as well

[JenkinsJinkies]

5 minutes ago, bncho said:

The AI Euro is a deadly fuckin model, especially inside D4. It was the only model that reliably sniffed out that the Feb 2025 storm was going out to sea, without really wavering.

Then it would throw us a bone by shifting very slightly north and west for a run only for the very next run to shift even further south and east that it was before.

[winter_warlock]

Just now, bncho said:

The GEFS has like 8" of snow in DC at the end of its run holy shit

 Nicee... So far this seems like our best shot of a nice hit all winter!!

[bncho]

Just now, winter_warlock said:

 Nicee... So far this seems like our best shot of a nice hit all winter!!

i was looking at the wrong ensemble run lol, post was edited

[winter_warlock]

1 minute ago, WxUSAF said:

I wouldn’t pay much attention to the AI models for mesoscale features. But synoptic scale? Euro AI is a cheat code inside D4-5. GFSAI outperforming physics-based GFS as well

 Wow nice!!! 

[Scarlet Pimpernel]

I know from looking at various types of events (not just for snow) that the AIFS (Euro) has done a stunningly good job out to 5 or even 7 days for situations that are rather tricky.  And it tended to lock in, not go all over the place.  Granted, those cases were just specific individual ones that maybe were more favorable for it, but you get the idea.  I'm pretty impressed with how the AI models have done.

[winter_warlock]

1 minute ago, bncho said:

i was looking at the wrong ensemble run lol, post was edited

 Well still would be a decent hit

[Scarlet Pimpernel]

2 minutes ago, WxUSAF said:

I wouldn’t pay much attention to the AI models for mesoscale features. But synoptic scale? Euro AI is a cheat code inside D4-5. GFSAI outperforming physics-based GFS as well

Totally agree...at this point (until they improve or there's more mesoscale AI type models) they won't capture small features like banding, convective elements, and the like.  But for overall synoptic setup and general low and precip locations as you say, it's rather bang-on in a lot of cases.  

[bncho]

pretty cool signal for 8 days out

[Scarlet Pimpernel]

5 minutes ago, bncho said:

pretty cool signal for 8 days out

Oy, Gevalt, yes please!  But seriously...ignoring the specific amounts, the ensembles (mean at least) clearly has the main swath of snow (and precip) through the area here.  Going with just that, it's clear the deterministic GFS is currently outside the main envelope here with giving GA/SC/NC the heaviest snow and/or ice.

(ETA:  Gotta put out the usual disclaimer that I'm not saying the GFS will be wrong, just that with the current indications it's looking like an outlier solution or at least too much suppression.)

[stormy]

The Euro AI deterministic 24/25 event can be given a 20% chance of verification 7 days out.

That's better than 10% on deterministic Op.

AI ens just downloading.

[Eskimo Joe]

12 GEFS-AI is money for this Frid-Sat potential. Cold powder, 850s, never get above -6, and some gul moisture to boot.

[Terpeast]

5 minutes ago, Scarlet Pimpernel said:

Oy, Gevalt, yes please!  But seriously...ignoring the specific amounts, the ensembles (mean at least) clearly has the main swath of snow (and precip) through the area here.  Going with just that, it's clear the deterministic GFS is currently outside the main envelope here with giving GA/SC/NC the heaviest snow and/or ice.

(ETA:  Gotta put out the usual disclaimer that I'm not saying the GFS will be wrong, just that with the current indications it's looking like an outlier solution or at least too much suppression.)

You can see a secondary swath across the deep south, so its clear that there are a few suppressed members. But still are in the minority. The op just happens to be one of them this run. Would not be surprised to see a very different 18z op run. 

[Scarlet Pimpernel]

Just now, Terpeast said:

You can see a secondary swath across the deep south, so its clear that there are a few suppressed members. But still are in the minority. 

Yeah, good and fair point.  I don't have subscriptions to all that information so was just going by what I could make out on the zoomed-in image above.  All the same, what you mention does indicate some suppressed members but in the minority.

[bncho]

By the way, the mean temperature for next Sunday afternoon is 18 degrees in DC on the GEFS. Which is just absolutely LOL worthy.

[bncho]

JFC the Euro AI Ensemble has a 3 day snow mean for Jan 25-27 is like 6+" I think, have to wait for slow ass wxbell to confirm

[Heisy]

The big thing is getting that STJ energy to eject E

12z ai made a huge progression change in doing that.

The prior 6z run buried it under the ridge

If it really does head east this is the best chance for a MECS in a decade.




.

[Kevin Reilly]

33 minutes ago, bncho said:

pretty cool signal for 8 days out

Talk about region wide.  Has a Wes-southwest to east-northeast flow to it.