• Hello, please take a minute to check out our awesome content, contributed by the wonderful members of our community. We hope you'll add your own thoughts and opinions by making a free account!

Jan. 17-18, 2026 SE Winter Weather Threat

Not sure I recall this big of a disagreement by the gfs and its ensembles vs other modeling this close to game time. But I never can fully trust the gfs. But even its AI version is much more amped than euro models. This is not settled and I’m not fully discounting the gfs right now.


Sent from my iPhone using Tapatalk
 
iwantsouthernsnow123 said:
Generally its quite the opposite, you want those trends within this timeframe. NW tick often happens. FFC NWS mentions that this will not be really figured out possibly until 24-48 hours till.
Click to expand...
Like I mentioned, models have gotten better over the last 5 years. Much less run-to-run variation inside of 84 hours. Unfortunately, like Atlanta, Toccoa will likely not see any measurable snowfall from this event.

I want snow, but at some point we have to be realistic here. The ECMWF/AIFS are the gold standard in NWP. Doesn't mean that they're perfect, but expecting such drastic changes in such a short period is merely wishcasting.

When it is GFS vs the world, the world wins 99/100 times.
 
bouncycorn said:
I don't think Atlanta will receive measurable snowfall from this event, FWIW. Best chance for snowfall will be in the Carolinas.
Click to expand...
I don't think Atlanta will receive measurable snowfall from this event, FWIW. Best chance for snowfall will be in the Carolinas.
i 100% disagree
 
anc.py said:
Wouldn't that legitimately just shut down the entire Atlanta metro for at least a day before a warmup would melt some of the accum?

Footnote, if it doesn't melt all of it, then it's going to stay for 4 days due to a cold spell lol
Click to expand...

Pretty much the entire southeast


Sent from my iPhone using Tapatalk
 
bouncycorn said:
Like I mentioned, models have gotten better over the last 5 years. Much less run-to-run variation inside of 84 hours. Unfortunately, like Atlanta, Toccoa will likely not see any measurable snowfall from this event.

I want snow, but at some point we have to be realistic here. The ECMWF/AIFS are the gold standard in NWP. Doesn't mean that they're perfect, but expecting such drastic changes in such a short period is merely wishcasting.

When it is GFS vs the world, the world wins 99/100 times.
Click to expand...
Where is your model?
 
bouncycorn said:
Everyone wants this NW trend to continue, we've just run out of time for any further substantial NW trend IMO. Especially after seeing the 00z AIFS/IFS outputs.

If we in Atlanta were seeing these model outputs at 144hrs, I'd be excited. Unfortunately, we're about 84hrs out from the 'event'. And models have gotten better over the last 5 years.
Click to expand...
How have we ran out of time for the NW trend to continue? I’m very curious on your theory
 
iwantsouthernsnow123 said:
Does the phasing look off to you? Like almost like euro isnt handling things well?
Click to expand...
there's not really any phasing with this to speak of. i just thought the structure of the shortwave looked better and was surprised it didn't bear fruit. i wouldn't be shocked if it's noise and the eps improves
 
accu35 said:
As Webb mentioned earlier, we may have little more NW trend up until go time.
Click to expand...

There's nothing preventing it until we get better data which won't happen for at least 6-12 hours I assume, considering the area of interest.

However as TheBatman mentioned, something's got to give soon and it'll happen quick lol
 
Ross said:
there's not really any phasing with this to speak of. i just thought the structure of the shortwave looked better and was surprised it didn't bear fruit. i wouldn't be shocked if it's noise and the eps improves
Click to expand...
I have noticed that euro doesn't push as much southern stream energy up as the other models which leads me to believe maybe it could be heading that way. The positive tilt worries me but it may not be the worst thing ever.
 
Guys 12z models today gonna be huge!! Another piece of the puzzle will start to come in as the data is ingested to the model systems. Just every chill.
 
weather nerd said:
what's the reasoning
Click to expand...
I’m not a certified met but have been watching weather for almost a decade at this point. The NW trend has only failed once (last year) also in overrunning events the precip field has always over performed. Also almost every major model ensembles has at least a dusting for the Atlanta metro
 
NorthGaWinter4 said:
How have we ran out of time for the NW trend to continue? I’m very curious on your theory
Click to expand...
Conventional weather models tend to have a 'NW trend' component, especially between 84 and 144hrs. Usually that starts to slow from a strong NW trend to wobbles inside of 84 hours.

New generation models (AIFS/AIFS-ENS/WeatherNext2/Pangu, etc) have very little variation inside of 84 hours, especially in terms of strong trends in one direction.

I mean, look at the last 4 runs of the AIFS. Just small wobbles. This is what I generally expect over the next 3 days -- the new generation models will continue to have small wobbles while conventional models fall into line with what they are showing. This has been the case for the last 2-3 years.

I am a lead weather model engineer for a company that focuses on predicting snow around the world. I live and breathe this stuff every day.

As bad as I want snow, I can't deny what I am seeing here.
trend-ecmwf_aifs-2026011500-f084.qpf_006h-imp.conus.gif
 
bouncycorn said:
Conventional weather models tend to have a 'NW trend' component, especially between 84 and 144hrs. Usually that starts to slow from a strong NW trend to wobbles inside of 84 hours.

New generation models (AIFS/AIFS-ENS/WeatherNext2/Pangu, etc) have very little variation inside of 84 hours, especially in terms of strong trends in one direction.

I mean, look at the last 4 runs of the AIFS. Just small wobbles. This is what I generally expect over the next 3 days -- the new generation models will continue to have small wobbles while conventional models fall into line with what they are showing. This has been the case for the last 2-3 years.

I am a lead weather model engineer for a company that focuses on predicting snow around the world. I live and breathe this stuff every day.

As bad as I want snow, I can't deny what I am seeing here.
View attachment 183603
Click to expand...
Since we're going to bring up AIFS specifically. Thought it was necessary to add GFS AI as well rather than just using 1 AI model. I get you're a model developer but at the same time can't just look at 1 model and call it a day.
 

Attachments

  • gfs-ai-conus-t850_mslp_prcp6hr-1768435200-1768716000-1768716000-20.gif
    gfs-ai-conus-t850_mslp_prcp6hr-1768435200-1768716000-1768716000-20.gif
    2.1 MB · Views: 46
NorthGaWinter4 said:
I’m not a certified met but have been watching weather for almost a decade at this point. The NW trend has only failed once (last year) also in overrunning events the precip field has always over performed. Also almost every major model ensembles has at least a dusting for the Atlanta metro
Click to expand...
were the majority of these NW trends from actual low pressure systems (which im well aware of) or just an expansive precipitation field instead?
 
bouncycorn said:
Conventional weather models tend to have a 'NW trend' component, especially between 84 and 144hrs. Usually that starts to slow from a strong NW trend to wobbles inside of 84 hours.

New generation models (AIFS/AIFS-ENS/WeatherNext2/Pangu, etc) have very little variation inside of 84 hours, especially in terms of strong trends in one direction.

I mean, look at the last 4 runs of the AIFS. Just small wobbles. This is what I generally expect over the next 3 days -- the new generation models will continue to have small wobbles while conventional models fall into line with what they are showing. This has been the case for the last 2-3 years.

I am a lead weather model engineer for a company that focuses on predicting snow around the world. I live and breathe this stuff every day.

As bad as I want snow, I can't deny what I am seeing here.
View attachment 183603
Click to expand...
appreciate your expertise. hard to argue against any of that. have you noticed any storm modes that are harder on the ai models?
 
Ross said:
appreciate your expertise. hard to argue against any of that. have you noticed any storm modes that are harder on the ai models?
Click to expand...
Nope, across the board, the AIFS/AIFS-ENS is the best publicly-available model suite available currently. AIGFS is quite a bit worse -- on par with the conventional ECMWF-IFS.

There will certainly be cases where the ECMWF data assimilation techniques probablistically get it 'wrong' which would impact all downstream models (ECMWF-IFS, AIFS, EPS, AIFS-ENS, GRAF).. but most of the time, ECMWF has superior data assimilation techniques. And, typically, if there are issues with the ECMWF DA, you'd see divergence between the ECMWF-based/initialized models and models that use different DA (GFS, ICON, CMC, UKM).

In this case, the GFS/AIGFS are on an island of their own.. so this makes me think that there could be data assimilation issues with GFS.
 
bouncycorn said:
Nope, across the board, the AIFS/AIFS-ENS is the best publicly-available model suite available currently. AIGFS is quite a bit worse -- on par with the conventional ECMWF-IFS.

There will certainly be cases where the ECMWF data assimilation techniques probablistically get it 'wrong' which would impact all downstream models (ECMWF-IFS, AIFS, EPS, AIFS-ENS, GRAF).. but most of the time, ECMWF has superior data assimilation techniques. And, typically, if there are issues with the ECMWF DA, you'd see divergence between the ECMWF-based/initialized models and models that use different DA (GFS, ICON, CMC, UKM).

In this case, the GFS/AIGFS are on an island of their own.. so this makes me think that there could be data assimilation issues with GFS.
Click to expand...
RRFS generally agrees with GFS. It seems like AIFS, especially euro and euro AI appears to be struggling making use of the southern stream while some other models are making rather practical use out of it. Obviously RRFS long range is questionable but the general phase is similar to GFS and Euro just makes a lot of use out of the southern jet.
 
models-2026011500-f054.500wh-mean.conus.gif
I unfortunately think the initial lead wave the GEFS is horribly on its own with (and @ F+054 it's doing this? Jesus).

It's what leads to that neutral/negative tilt quickly as opposed to the positive tilt through GA.
 
bouncycorn said:
Nope, across the board, the AIFS/AIFS-ENS is the best publicly-available model suite available currently. AIGFS is quite a bit worse -- on par with the conventional ECMWF-IFS.

There will certainly be cases where the ECMWF data assimilation techniques probablistically get it 'wrong' which would impact all downstream models (ECMWF-IFS, AIFS, EPS, AIFS-ENS, GRAF).. but most of the time, ECMWF has superior data assimilation techniques. And, typically, if there are issues with the ECMWF DA, you'd see divergence between the ECMWF-based/initialized models and models that use different DA (GFS, ICON, CMC, UKM).

In this case, the GFS/AIGFS are on an island of their own.. so this makes me think that there could be data assimilation issues with GFS.
Click to expand...
this is heavy stuff. reading your findings gives me a sense of, if the aifs is locked in, then why even bother with anything else
 
Ross said:
this is heavy stuff. reading your findings gives me a sense of, if the aifs is locked in, then why even bother with anything else
Click to expand...
I get AIFS is definitely advancing but I'm not convinced there isn't any fail modes. Can't forget what happened last year

Sampling not long ago appears to look generally better for the south
 
iwantsouthernsnow123 said:
I get AIFS is definitely advancing but I'm not convinced there isn't any fail modes. Can't forget what happened last year

Sampling not long ago appears to look generally better for the south
Click to expand...
the siren song of advocating for a nw trend is alluring but hard to argue against bouncycorn's thesis. just gotta live through it to find out
 
You must log in or register to reply here.
Back
Top