theweatheroutlook

Remove ads from site

Brian Gaze
30 November 2023 13:32:58



Which makes me wonder....are the models programmed with climate change assumptions these days....so that instead of running with scenarios purely based on historical data, they are delivering outcomes with a big dollop of a "climate change will make our weather more extreme" assumption added in, with a greater emphasis for the runs excess of 8 days ?

Originally Posted by: Timmytour 



No.
Brian Gaze
Berkhamsted
TWO Buzz - get the latest news and views 
"I'm not socialist, I know that. I don't believe in sharing my money." - Gary Numan
Nick Gilly
30 November 2023 14:08:56

No.

Originally Posted by: Brian Gaze 



If the models aren't factoring in climate change, does that not make them less reliable for longer range forecasts?
warrenb
30 November 2023 14:12:20
ECM short ensemble for 06z run is hilarious, literally take your pick from +4 to -8, with a fairly equal scatter between.

​​​​​​https://www.wetterzentrale.de/en/show_diagrams.php?geoid=49069&model=ecm&var=2&run=6&lid=ENS&bw=1 
Tonbridge 45m 
Brian Gaze
30 November 2023 14:28:04

If the models aren't factoring in climate change, does that not make them less reliable for longer range forecasts?

Originally Posted by: Nick Gilly 



The models are fed with data which represents the current state of the atmosphere and oceans. They then apply mathematical equations based on the laws of physics to generate the solution. The laws of physics are not altered by climate change.
Brian Gaze
Berkhamsted
TWO Buzz - get the latest news and views 
"I'm not socialist, I know that. I don't believe in sharing my money." - Gary Numan
+2
Saint Snow
30 November 2023 14:42:13

The models are fed with data which represents the current state of the atmosphere and oceans. They then apply mathematical equations based on the laws of physics to generate the solution. The laws of physics are not altered by climate change.

Originally Posted by: Brian Gaze 




😄

 

Martin
Home: St Helens (26m asl) Work: Manchester (75m asl)
A TWO addict since 14/12/01
"How can wealth persuade poverty to use its political freedom to keep wealth in power? Here lies the whole art of Conservative politics."
Aneurin Bevan
squish
30 November 2023 15:22:12
12z ICON undercuts the cold air mass out to +144
D.E.W on Dartmoor. 300m asl
Timmytour
30 November 2023 15:26:25
Then I guess somewhere in the algorithms used by the likes of GFS there has just got to be the some  sort of instruction to perform duff mathematical equations that end up translating into modelling a storm that virtually wipes London of the map   (see 6z 15DEC23) instead of getting anywhere close to what really happens.....


 
Broxbourne, Herts 133ft ASL
squish
30 November 2023 15:29:38
ICON tantalising out to +180.

good start to the 12zs
D.E.W on Dartmoor. 300m asl
Gandalf The White
30 November 2023 15:35:55

Then I guess somewhere in the algorithms used by the likes of GFS there has just got to be the some  sort of instruction to perform duff mathematical equations that end up translating into modelling a storm that virtually wipes London of the map   (see 6z 15DEC23) instead of getting anywhere close to what really happens.....


 

Originally Posted by: Timmytour 



But if what you say was true then the model would be producing faulty output all the time, which clearly isn’t the case.

At the limits of the model output range, 16 days away, you can guarantee that virtually no chart will verify as it appears.  But it doesn’t mean the formulae are faulty, it means the opening data isn’t accurate and the behaviour of the atmosphere is too complex to predict so far ahead.
Location: South Cambridgeshire
130 metres ASL
52.0N 0.1E


DEW
  • DEW
  • Advanced Member
30 November 2023 15:56:17

In 14 days time the GFS 6z puts the south east into a "storm of the century"

I'm struck by how often there are such modelled runs of late.  And of course, some never materialise at all while those that do to some extent, do so with nowhere near the same degree of intensity as initially envisaged by the models.

Which makes me wonder....are the models programmed with climate change assumptions these days....so that instead of running with scenarios purely based on historical data, they are delivering outcomes with a big dollop of a "climate change will make our weather more extreme" assumption added in, with a greater emphasis for the runs excess of 8 days ?

Originally Posted by: Timmytour 



I'm struck by how unstable the GFS model output is of late.

The 6z output referred to has LP 940mb N England Sat 16th at 6am
The 12z output has HP 1040mb completely covering England Fri 15th at 12 noon (Sat chart presumably still downloading)

A fascinating shift of fortune. 

So is the butterfly wing effect causing a dramatic effect in outcomes i.e. the weather is more chaotic than ever, or are the GFS algorithms over-sensitive to starting conditions?
War does not determine who is right, only who is left - Bertrand Russell

Chichester 12m asl
Timmytour
30 November 2023 16:17:23

But if what you say was true then the model would be producing faulty output all the time, which clearly isn’t the case.

At the limits of the model output range, 16 days away, you can guarantee that virtually no chart will verify as it appears.  But it doesn’t mean the formulae are faulty, it means the opening data isn’t accurate and the behaviour of the atmosphere is too complex to predict so far ahead.

Originally Posted by: Gandalf The White 



It's not the accuracy of the modelling I am querying, it's the types of outcomes it occasionally throws up at a certain distance out.

If it shows a sunny day and it turns out to be raining, that's one thing.

But to model the types of conditions today's GFS 6z showed for the south-east for at T342 is a one in three hundred year scenario in real life.  But for the amount of times such scenarios get modelled you'd be forgiven for thinking it's a one in three year event!! 

 
Broxbourne, Herts 133ft ASL
Heavy Weather 2013
30 November 2023 16:23:43
Got to say 12z GFS at 198hrs looks ok, it’s a mess, but polar vortex non existent also trough disruption over the UK. 
Mark
Beckton, E London
Less than 500m from the end of London City Airport runway.
squish
30 November 2023 16:24:54
12z GFS not a million miles from ICON 

just a bit further north with the action …but substantially further south from the 06z.

again not far from something pretty memorable somewhere ! 

 
D.E.W on Dartmoor. 300m asl
nsrobins
30 November 2023 16:28:28

12z ICON undercuts the cold air mass out to +144

Originally Posted by: squish 


ICON like a dog with a bone at the moment.
Neil
Fareham, Hampshire 28m ASL (near estuary)
Stormchaser, Member TORRO
doctormog
30 November 2023 16:30:31
And the GFS 12z op  run to +200 hr remains cold here. There are fine margins and lots of uncertainty in the now medium term. 
Aberdeen
Current conditions (personal WS) 
Ally Pally Snowman
30 November 2023 16:33:17

Got to say 12z GFS at 198hrs looks ok, it’s a mess, but polar vortex non existent also trough disruption over the UK. 

Originally Posted by: Heavy Weather 2013 



Yep easterly by day 9. Interesting that the models keep refusing to reset to zonal. Something different this year maybe.
 
Bishop's Stortford 85m ASL.
doctormog
30 November 2023 16:40:33

Yep easterly by day 9. Interesting that the models keep refusing to reset to zonal. Something different this year maybe.
 

Originally Posted by: Ally Pally Snowman 



Perhaps. Or it could still be a long fetch SWly by day 7!
Aberdeen
Current conditions (personal WS) 
overland
30 November 2023 16:42:59
It is weird how the 0z runs have been more progressive recently, but ones later in the day go the other way. It's been mentioned a few times, but I've not paid any attention as it's not logical with the way the models work.  However it keeps happening and it's getting to the stage where it might be more than a statistical anomaly.
Mumbles, Swansea. 80m asl
doctormog
30 November 2023 16:47:14

It is weird how the 0z runs have been more progressive recently, but ones later in the day go the other way. It's been mentioned a few times, but I've not paid any attention as it's not logical with the way the models work.  However it keeps happening and it's getting to the stage where it might be more than a statistical anomaly.

Originally Posted by: overland 



I have noticed that for a number of years and I wonder if it is to do with the range of input data sources at the initiation times of the various different runs. Perhaps it’s my mind playing tricks on me.
Aberdeen
Current conditions (personal WS) 
Retron
30 November 2023 16:58:40

I have noticed that for a number of years and I wonder if it is to do with the range of input data sources at the initiation times of the various different runs. Perhaps it’s my mind playing tricks on me.

Originally Posted by: doctormog 


It used to be that the 0z and 12z runs had a fuller set of data. They used to verify more highly than the 6z and 18z runs too.

It also used to be the case that a lack of aeroplane data over Christmas led to all sorts of wacky blocking popping up, only to vanish as flights got back to data a couple of days later - this happened pretty much every year, the exception being when the outlook was already looking blocked (in our region).

These days there's much more data being fed in, including vast amounts of satellite data which simply wasn't available even 10 years ago. It's enough to have reduced the impact of lack of data over the past couple of Christmases.

I still wouldn't be surprised, mind you, to see the 0z and 12z runs continuing to verify more highly than the others!

EDIT: I also wonder if the ever-increasing resolution of the GFS in particular is causing problems. The "dartboard phenomenon" seems to have become worse in the past few years, and I wonder if it's because the model is now able to resolve very small-scale phantom features which end up throwing things awry - whereas with fewer grid points, they wouldn't have appeared at all, thus wouldn't have gone on to cause issues.

It should also be mentioned that ECM has also increased its resolution massively over the years, but doesn't suffer as much from the "dartboards". They've always had better results overall than GFS, though, so presumably their algorithms are better able to get rid of those phantom features...
 
Leysdown, north Kent
Rob K
30 November 2023 16:59:48



Which makes me wonder....are the models programmed with climate change assumptions these days....so that instead of running with scenarios purely based on historical data, they are delivering outcomes with a big dollop of a "climate change will make our weather more extreme" assumption added in, with a greater emphasis for the runs excess of 8 days ?

Originally Posted by: Timmytour 



As far as I understand the models, they are not programmed with either climate change assumptions OR historical data.

All they do is model the atmosphere as a fluid - you take measurements of the temperature, pressure, wind speed etc etc at various points, and the model just calculates what happens to the fluid flow. They are not programmed with "if the pressure rises here then there will be a Scandinavian high over here, because that's what happened in 19xx".
Yateley, NE Hampshire, 73m asl
"But who wants to be foretold the weather? It is bad enough when it comes, without our having the misery of knowing about it beforehand." — Jerome K. Jerome
overland
30 November 2023 17:10:02

I have noticed that for a number of years and I wonder if it is to do with the range of input data sources at the initiation times of the various different runs. Perhaps it’s my mind playing tricks on me.

Originally Posted by: doctormog 



It's an interesting one as we (I!) only really pay attention when there is cold on offer so haven't noticed it at other times.
Mumbles, Swansea. 80m asl
Retron
30 November 2023 17:11:54

They are not programmed with "if the pressure rises here then there will be a Scandinavian high over here, because that's what happened in 19xx".

Originally Posted by: Rob K 


True enough, although it's worth mentioning that's how Graphcast works, the new ECM/Google ML model.

The holy grail would be merging the two codebases, but that'd be a hell of a job. In other words, "the physics over our grid points would say x will happen, but checking with ML shows that 9 out of 10 times something pops up and messes it up, so we'll go with that instead".

You could call it the "No white Christmases ever again for Kent" model! 😂

(There were also "bias corrected" versions of some models available in the past, but the difference between those and the normal models was minimal at best in terms of overall accuracy. It could well be that there are certain tweaks applied to the traditional models to reduce known errors...)

And doing more digging, how's this for interesting? It's from the GFS changelog back in 2019.
https://www.emc.ncep.noaa.gov/emc/docs/FV3GFS_OD_Briefs_10-01-18_4-1-2019.pdf 

Initial discovery of odd convection in FV3GFS (Mark Klein, WPC)
Fixed an issue with climatological tendencies applied to the model (4x/day!)

Fit-to-radiosondes
GFS and FV3GFS winds too weak, but FV3GFS is closer to obs.

FV3GFS winds better in troposphere but worse in stratosphere
Fit-to-analyses FV3GFS reduced GFS cold bias in middle to upper stratosphere.

FV3GFS wind RMS worse in stratosphere but comparable to EC
 
Leysdown, north Kent
+1
Heavy Weather 2013
30 November 2023 17:22:40
This isn’t a dig at anyone, but you know winter is here when start talking about input data and verification. Winter is here. Yay.
Mark
Beckton, E London
Less than 500m from the end of London City Airport runway.
The Beast from the East
30 November 2023 17:24:07
Is ICON the new Nogaps? At least we might get a snowy breakdown out of this spell. Just hope its not the only one of the season, like last year
"We have some alternative facts for you"
Kelly-Ann Conway - special adviser to the President
Users browsing this topic
    Ads