Model Output Discussion - Into 2024

[Ali1977]

Can we finally get some agreement on the 12zs and ECM - hope so

[Lukesluckybunch]

Just now, Ali1977 said:

Can we finally get some agreement on the 12zs and ECM - hope so

Let's hope so..all them colder runs in the gefs..absolutely staggering!

[IanT]

1 hour ago, Ice Day said:

So, at lunchtime on the 15th we have a long-fetch northerly with the -8 isotherm well clear of the south coast.

 

Brutal windchill by day 10 nationwide

Only 9 sleeps to @Scott Ingham day…

[Ben Sainsbury]

49 minutes ago, Battleground Snow said:

Do you think that because the EPS are run at the same high resolution as the OP, that it sometimes times causes them all to see the wrong solution early on in the run?

It always seems like the mean and the op are very close up to around 144, where as the gfs can be quite different to it's mean at that point.

It's quite amazing the difference in the cold pool distribution at 72 hours compared to the UKMO, but then again I guess we only notice it because it's over the UK, when it's a micro detail globally 

I am no expert in modelling, but I don't really see why the EPS having the same resolution as the OP should result in more wrong solutions early on. One of the main reasons to have the EPS the same resolution as the OP is so they have the same ability to resolve smaller features. On a localised scale I guess this could lead to wider range of solutions regarding winds etc, but for the broader pattern I don't see why a higher resolution to previous would result in more "wrong" solutions early on.

Sorry, not particularly helpful!

Edited January 6 by Ben Sainsbury

[IanT]

49 minutes ago, BleakMidwinter said:

I liked the *Ingham Singularity* but I can't remember who suggested it... 

 

38 minutes ago, Scott Ingham said:

I liked that as well! it was posted a few days ago yeah I can’t remember by who!

That was me…

 

[Mcconnor8]

21 minutes ago, Ali1977 said:

Can we finally get some agreement on the 12zs and ECM - hope so

Would be good to but wouldn't put any money on it, still far enough out for there to be a wide range of solutions and the models love to leave us guessing.

[Guest]

1 hour ago, nick sussex said:

The GFS 06 hrs run is a good illustration in that sometimes smaller is better ! 

 

Sometimes!...Mrs SM has been well and truly spoiled over the years!

[Rob 79812010]

So much to be positive about about in the models. Beeb monthly update is very encouraging. They still mention effects of ssw in feb

[Audaxian]

20 minutes ago, Ben Sainsbury said:

I am no expert in modelling, but I don't really see why the EPS having the same resolution as the OP should result in more wrong solutions early on. One of the main reasons to have the EPS the same resolution as the OP is so they have the same ability to resolve smaller features. On a localised scale I guess this could lead to wider range of solutions regarding winds etc, but for the broader pattern I don't see why a higher resolution to previous would result in more "wrong" solutions early on.

Sorry, not particularly helpful!

I think the hypothesis here is that same resolution = more likely for the models to "follow the same route" and generate less variation. So if the OP gets it wrong, it's possibly now more likely that the ensembles follow it than it was previously. But if the OP gets it right, same applies - ensembles more likely to follow along.

I suspect, however, that won't be too big an issue. I would think that while all the runs start with the same initial input data, they will all have small variations very quickly. Higher resolution could even result in more divergence as those small variations become more developed than they would at lower resolution, and thus potentially more influential on the evolution of the run.

Edited January 6 by Audaxian

[TSNWK]

36 minutes ago, Ali1977 said:

Can we finally get some agreement on the 12zs and ECM - hope so

So long as its the CORRECT! agreement

[jmp223]

1 hour ago, Mike Poole said:

And at the end of all of that we have this:

Hi Mike, is that a vortex split event rather than a displacement?  Just curious as to whether this has any bearing on our chance of cold.  I know 2/3 SSWs lead to colder weather, however I wonder if it is type dependent?  I would assume it is a good thing the warming is on our side-by that I mean to our NE, as opposed to occuring over the Canadian side.  I assume by that the effects would be more pronounced for the UK.

I'm only really starting to study SSWs and their "teleconnectors" seriously this winter. I found a great document that explains the role Rossby waves play in all of this, everyone can find it here.  I recommend everyone who isn't already an "expert" give it a read

Sudden Stratospheric Warming (SSW) | SKYbrary Aviation Safety

SKYBRARY.AERO

---

Description First discovered in 1952, a Sudden Stratospheric Warming (SSW) refers to a swift jump in temperatures in the stratosphere that is sometimes linked to the onset of cold weather in...

 

Edited January 6 by jmp223
Rossby, not Rosseby lol

[The Enforcer]

GEFS06z P16 is annoying in that it follows the overall pattern, but at the microscale, it never gets cold enough:

[Cold Winter Night]

1 hour ago, Battleground Snow said:

Do you think that because the EPS are run at the same high resolution as the OP, that it sometimes times causes them all to see the wrong solution early on in the run?

It always seems like the mean and the op are very close up to around 144, where as the gfs can be quite different to it's mean at that point.

It's quite amazing the difference in the cold pool distribution at 72 hours compared to the UKMO, but then again I guess we only notice it because it's over the UK, when it's a micro detail globally 

57 minutes ago, Mike Poole said:

It is an interesting idea.  The new EPS is the first time an ensemble set has been run at the same resolution as the op, so I don’t think we know the answer.  There is this idea that higher resolution can cause an error early on mean after that they just compute the wrong solution more accurately.  I am not sure I wholly buy it, maybe after the new EPS has been running for some time we may get an answer.

On the GFS side, I think there may be other differences between the GFS and the GEFS apart from resolution - I think they both now use the FV3 core, but the GFS did get an upgrade last year which most felt was a retrograde step, not sure if GEFS has this or not.

A small downside of a higher resolution, compared to lower resolution is indeed that a higher resolution run can pick up an unimportant detail and make a big deal out of it than a lower resolution run that would miss the detail in the first place.
It's not uncommon to see a couple of members in an ensemble go in wildly different directions, super mild or super cold, and sometimes the Operational is that member.
(My undocumented speculation: I suppose in the EPS the probability for that would be close to 1 in 51. Maybe it's a bit better because of the unperturbed starting data, so let's say once a month the Op goes crazy).

The upside on the other hand is that it can pick up details that do matter and rightfully make a big deal of it.

This is exactly one of the reason there is an ensemble in the first place. One or two members may pick up the wrong detail and blow it up unnecessarily, 51 members won't (or rarely).

This is also why it's best to follow an Op only to day 5 or 6, and to use the rest of it only for your entertainment. The ensemble mean and the clusters will be much more helpful after that.

[Scott Ingham]

4 minutes ago, Cold Winter Night said:

A small downside of a higher resolution, compared to lower resolution is indeed that a higher resolution run can pick up an unimportant detail and make a big deal out of it than a lower resolution run that would miss the detail in the first place.
It's not uncommon to see a couple of members in an ensemble go in wildly different directions, super mild or super cold, and sometimes the Operational is that member.
(My undocumented speculation: I suppose in the EPS the probability for that would be close to 1 in 51. Maybe it's a bit better because of the unperturbed starting data, so let's say once a month the Op goes crazy).

The upside on the other hand is that it can pick up details that do matter and rightfully make a big deal of it.

This is exactly one of the reason there is an ensemble in the first place. One or two members may pick up the wrong detail and blow it up unnecessarily, 51 members won't (or rarely).

This is also why it's best to follow an Op only to day 5 or 6, and to use the rest of it only for your entertainment. The ensemble mean and the clusters will be much more helpful after that.

Yea exactly this! I agree with all of this.

[The Enforcer]

Rather than having dozens of runs at low resolution, would it not be better to have fewer runs, but all at high resolution?

[Froze were the Days]

Always the GFS with the eye candy (which has come back again after dropping the decent runs) and ECM being the grinch...I know what model I'd bet on at the moment, haven't looked at UKMO 168 any good?

[WINTRY WALES]

Quick question. I know we have a graph which compare the performance of the various models (ECM, GFS, etc) but do they do that for each per run. Like which one performs best 06z, 12z etc…is the performance of those captured too?? 

[Mcconnor8]

2 minutes ago, WINTRY WALES said:

Quick question. I know we have a graph which compare the performance of the various models (ECM, GFS, etc) but do they do that for each per run. Like which one performs best 06z, 12z etc…is the performance of those captured too?? 

Verification stats for GFS between 0z, 6z, 12z and 18z are very similar

[Gypo]

13 minutes ago, jmp223 said:

Hi Mike, is that a vortex split event rather than a displacement?  Just curious as to whether this has any bearing on our chance of cold.  I know 2/3 SSWs lead to colder weather, however I wonder if it is type dependent?  I would assume it is a good thing the warming is on our side-by that I mean to our NE, as opposed to occuring over the Canadian side.  I assume by that the effects would be more pronounced for the UK.

I'm only really starting to study SSWs and their "teleconnectors" seriously this winter. I found a great document that explains the role Rossby waves play in all of this, everyone can find it here.  I recommend everyone who isn't already an "expert" give it a read

Sudden Stratospheric Warming (SSW) | SKYbrary Aviation Safety

SKYBRARY.AERO

---

Description First discovered in 1952, a Sudden Stratospheric Warming (SSW) refers to a swift jump in temperatures in the stratosphere that is sometimes linked to the onset of cold weather in...

 

I am probably wrong but I'm sure displacements lead to greater polar cap warming?

This looks like a displacement-split, with sister vortices ending up over western Europe/Canada and then Russia/Mongolia/China unless my map readings terrible?

[Met4Cast]

4 minutes ago, WINTRY WALES said:

Quick question. I know we have a graph which compare the performance of the various models (ECM, GFS, etc) but do they do that for each per run. Like which one performs best 06z, 12z etc…is the performance of those captured too?? 

Yes - There’s no real statistical difference between them when averaged out. 

[IanT]

15 minutes ago, Cold Winter Night said:

A small downside of a higher resolution, compared to lower resolution is indeed that a higher resolution run can pick up an unimportant detail and make a big deal out of it than a lower resolution run that would miss the detail in the first place.
It's not uncommon to see a couple of members in an ensemble go in wildly different directions, super mild or super cold, and sometimes the Operational is that member.
(My undocumented speculation: I suppose in the EPS the probability for that would be close to 1 in 51. Maybe it's a bit better because of the unperturbed starting data, so let's say once a month the Op goes crazy).

The upside on the other hand is that it can pick up details that do matter and rightfully make a big deal of it.

This is exactly one of the reason there is an ensemble in the first place. One or two members may pick up the wrong detail and blow it up unnecessarily, 51 members won't (or rarely).

This is also why it's best to follow an Op only to day 5 or 6, and to use the rest of it only for your entertainment. The ensemble mean and the clusters will be much more helpful after that.

It’s very simple. The model/resolution/run that shows the most wintry outcome for SE England is “correct”. The others are “struggling” and their output is “for the bin”

[TSNWK]

1 minute ago, Met4Cast said:

Yes - There’s no real statistical difference between them when averaged out. 

Is there one of those charts for day 6 please.  That is where ecm goes wrong

[sheikhy]

Ecm further north with easterly at 48 hours!!

[Audaxian]

11 minutes ago, Met4Cast said:

Yes - There’s no real statistical difference between them when averaged out. 

Higher number better or worse? Just looking at that specific chart, you can see that 17th sticks out (all the models either right on the button or way off it), and then if you look at the days before the 17th vs the days after the 17th there's a noticeable shift from being mostly clustered in the 0.3 to 0.6 range, vs mostly clustered 0.5 to 0.7.

I'm assuming these appear with a bit of lag? How long would that be - a few days or are we talking week or longer? If it's a short enough lag, we could be using these charts to "estimate" how reliable the model runs we're seeing currently might be; eg if they've been trending "bad" pile on some extra salt?

Edited January 6 by Audaxian

[MATTWOLVES 3]

Mogreps has seen the light and is saying..hey folks let's go back down

And again with the current warming of the strat...remember the lag time for the models to factor this in..Next 2 or 3 weeks could become very spicy.

Edited January 6 by MATTWOLVES 3