December Medium/Long Range Discussion

[NorthArlington101]

3 minutes ago, mitchnick said:

That's odd. Operational shows this on Pivotal. 

I’m showing the ens. I keep hearing random chatter that they’ve got a broken output but nobody has offered any proof, so I’ll choose to ignore it. Looks plausible to me… 

[TSSN+]

Only model that far south so it’s an outlier. 
 

 

[Eskimo Joe]

9 hours ago, winter_warlock said:

Gfs hr 126

High is in a bad spot. Southeast winds just eroding the cad.

[SomeguyfromTakomaPark]

Wow ai ensembles lol.  Absolute eye candy.  Do I fall for it? Hook line and sinker!!! U know the drill. 

[mitchnick]

41 minutes ago, Heisy said:

The one posted is the Ai ensemble mean


.

I know, it's odd that the operational would have to clearly be a huge outlier.  Hasn't it learned anything? 

[Ji]

i saw the individual panels for the Euro AI ensembles. Just wild lol

[TSSN+]

4 minutes ago, mitchnick said:

I know, it's odd that the operational would have to clearly be a huge outlier.  Hasn't it learned anything? 

Nope, it’s just like us. Ain’t learnt nothin. Haha 

[MillvilleWx]

Caution when assessing the AI model at range, especially for setups that lean more delicate and have greater emphasis on energy ejections and thermal gradients involved. AI is a great model and actually performs very well inside 72hrs when it comes to 5H progression and MSLP characteristics. It can still be semi-erratic and off beyond the 72hr mark, although I will add that if the model doesn't waver much and it maintains credence inside 72hrs, it probably will be right more often than not. It's something to keep in mind when looking at the AIFS/AIFS-ENS package. It's a tool like any other piece of guidance, and it's verification is pretty damn good right now because it doesn't utilize the typical thermodynamic and dynamic functions that allow for greater swings in output compared to dynamical models. You'll know it might be on to something when it doesn't waver for a while and we begin reaching that critical point inside 48hrs. 

[NorthArlington101]

CMC ens last night were also good… for the little it’s worth. 

[poolz1]

Give me the set up at the end of the Can run last night.  Endless overrunning into a stout HP.  I think these set ups only occur on model runs though.  

[Chris78]

2 minutes ago, poolz1 said:

Give me the set up at the end of the Can run last night.  Endless overrunning into a stout HP.  I think these set ups only occur on model runs though.  

That was a perfect set up. Cold High anchored NW of us in Canada and a huge slug of moisture running Sw to NE.

As the run ended we had 6"+ on the ground and several panels of snow left to go past 240 hours.

North of DC would of been easily 12+ amounts.

[Eskimo Joe]

5 minutes ago, Chris78 said:

That was a perfect set up. Cold High anchored NW of us in Canada and a huge slug of moisture running Sw to NE.

As the run ended we had 6"+ on the ground and several panels of snow left to go past 240 hours.

North of DC would of been easily 12+ amounts.

 

10 minutes ago, poolz1 said:

Give me the set up at the end of the Can run last night.  Endless overrunning into a stout HP.  I think these set ups only occur on model runs though.  

Yes, THAT is how we win and its inside an established pattern. Not a wave breaking into something as we're setting up. Really would like to see at least a slightly -NAO to help lock in our cold air though.

[poolz1]

7 minutes ago, Chris78 said:

That was a perfect set up. Cold High anchored NW of us in Canada and a huge slug of moisture running Sw to NE.

As the run ended we had 6"+ on the ground and several panels of snow left to go past 240 hours.

North of DC would of been easily 12+ amounts.

I dug into the details too! lol  I'll live in fantasy model land as much as I want.

[mitchnick]

You guys made me look at the Gem. Dang, that's a dream. 

[mitchnick]

12z Icon went bad on us. 

Nothing. Big shift north.

[TSSN+]

1 minute ago, mitchnick said:

12z Icon went bad on us. 

Nothing. Big shift north.

Good thing it’s the icon and worth about 1 cent 

[87storms]

1 hour ago, MillvilleWx said:

Caution when assessing the AI model at range, especially for setups that lean more delicate and have greater emphasis on energy ejections and thermal gradients involved. AI is a great model and actually performs very well inside 72hrs when it comes to 5H progression and MSLP characteristics. It can still be semi-erratic and off beyond the 72hr mark, although I will add that if the model doesn't waver much and it maintains credence inside 72hrs, it probably will be right more often than not. It's something to keep in mind when looking at the AIFS/AIFS-ENS package. It's a tool like any other piece of guidance, and it's verification is pretty damn good right now because it doesn't utilize the typical thermodynamic and dynamic functions that allow for greater swings in output compared to dynamical models. You'll know it might be on to something when it doesn't waver for a while and we begin reaching that critical point inside 48hrs. 

Weather model evaluation seems similar to machine learning on just about any dataset.  Want to reduce variance and have a more generalized prediction (it'll probably snow, but not sure how much)?  Random forest classifiers can help.  Want to reduce bias and get more precise with potential overfitting (don't worry about downsloping, it's gonna snow 9.8")?  Gradient boosting.  Want a mixed bag, but high cost and a potential black box (it's gonna snow somewhere between 1 and 15"), go with ensembles that have a variety of algorithms.

Judging by the median (which I'm assuming is the wise thing to look at lol), looks like the AIFS ensembles are leaning towards snow tv, which is still a win in early December.

[TSSN+]

Gfs still on car topper alert for northern Maryland crew sunday morning.

 

 

[Ji]

21 minutes ago, TSSN+ said:

Good thing it’s the icon and worth about 1 cent 

you still dont want any model being that warm/north

[SnowenOutThere]

GFS looks to have a bit more confluence so may be a bit cooler

[Ji]

1 minute ago, SnowenOutThere said:

GFS looks to have a bit more confluence so may be a bit cooler

it should at least hold from 6z

[NorthArlington101]

Nice thump at 117. Looks better