General Model Output Discussion

[Magadan]

no model has any bias in spite of a number on here maintaining they do.

All base their output on the laws of thermodynamics-all have slightly different soft ware packages (not a really apt name for the complexity of these) which is likely to give differing results even using the same basic data input. But none have preprogamming in to make for a default set up at any stage.

They have differing numbers of levels through the atmosphere, differing methods of doing the mathematics, differing time scales, differing length grid points and some differing length grid points at different time intervals.

Some times I am quite surprised with all these differences at just how close the results are for them. Just goes to show that the basic laws they all follow do work IF we had an absolutely accurate set of data which can never ever happen. Nor can the human made rules for the computers ever be as perfect as they would have to be to get perfect answers to the weather, be it 1 hour ahead or 1 month or 1 year-fortunately in my view.

this is the link into UK Met web site about their models so may be of interest to you and others; I would imagine that you can log on to both GFS (NOAA) and ECMWF, and get a similar amount of information about how they programmes their models.

hope it helps anyway

http://www.metoffice.gov.uk/science/creating/first_steps/

So am I right in this: they take a snapshot, as it were, of the current conditions and then try to mathematically model what it will do in the next few moments and repeat this until they get a prediction? I can see why weather forecasting this way is fraught with so much uncertainty. The equations must be nightmarishly complex (I'm speaking from the viewpoint of being an aeronautical engineer in the past and I know how troublesome it is to model fluid dynamics).

This is going to sound naive / stupid, but has anyone ever tried to document and electronically store snapshots of atmospheric conditions going back decades so that the computers can use any current snapshot today to search for a "best fit" from the past and then see what happened next the last time things were setup as closely? It would probably be even less accurate than the current computational modeling, given the lack of in-depth data from the past and the impossibility of two identical weather thumbprints within what is a "chaotic" system.

[SMU]

Indeed, a small spread

Cheers-

Remember our little banter around the 1-7 CET fot the UK- I said 0-1c was very conservative!!!!!!

Steve

[Thundery wintry showers]

Those ensembles have a very solid look and the general evolution looks pretty nailed out to the back end of next week.

Signs that the Northern arm of the jet is awakening from it's slumber and I would think that HP closer to the UK would be the evolution thereafter, but that's not to say we couldn't see another reload if the AO remains favourable and heights stay low over Europe.

I agree.

For southern areas it could be potentially snowy even towards mid-January if the southern arm of the jet stays strong enough to keep the high pressure centred over the north allowing easterly winds to affect central and southern districts.

[TEITS]

Here is a chart that will please some of those who we're complaining how dry it is.

http://www.meteociel.fr/ukmo/fax/fax48s.gif

This trough lined from E Anglia to Belfast will move SE during Saturday. At the moment the heaviest precip will be towards the E side of the trough but the UKMO even suggests snowfall into Wales.

A good example of never looking too far ahead when it comes to precipitation because these troughs often appear nearer the time.

[johnholmes]

hi M

yes remind me another day and I'll try and copy then post in just one of the basic formula-I have no knowledge of your side of fluid dynamics but yep its complex on the weather side.

before computers forecasters used to use empirical methods-something I've talked about on the forum on several occasions.

When computers came along I do know some of this was changed into whatever pc gobbledygook was required for a whole host of things-it is called MOS now Model Output Statistics, so some, I'm not sure, how much, of that is part of each model algorithms, if that is the correct terminology?

I'm sure if you write to the Met o, e mail, and ask those questions that they will direct it to the proper department/person who would answer you.

If you have a problem then pm me and I'll see if my diminishing number of contacts within the Office might help.

hope that helps a little?

[Optimus Prime]

According to the GFS maximum temperatures from Monday through next week WIDELY the maximum temperatures will be similar to January 1997 - below freezing.

[TEITS]

A clear example of an E.ly on the way when Kent has the lowest mean!

http://charts.netweather.tv/ensimages/ens.20091231/12/t850Kent.png

Even a couple of runs touch -15C!

[snowking]

Here is a chart that will please some of those who we're complaining how dry it is.

http://www.meteociel.fr/ukmo/fax/fax48s.gif

This trough lined from E Anglia to Belfast will move SE during Saturday. At the moment the heaviest precip will be towards the E side of the trough but the UKMO even suggests snowfall into Wales.

A good example of never looking too far ahead when it comes to precipitation because these troughs often appear nearer the time.

Yep a perfect example as even their own NAE model doesnt really suggest this

http://expert.weatheronline.co.uk/daten/proficharts/en/nae/2009/12/31/basis12/ukuk/prty/10010212_2_3112.gif

SK

[wysiwyg]

So am I right in this: they take a snapshot, as it were, of the current conditions and then try to mathematically model what it will do in the next few moments and repeat this until they get a prediction? I can see why weather forecasting this way is fraught with so much uncertainty. The equations must be nightmarishly complex (I'm speaking from the viewpoint of being an aeronautical engineer in the past and I know how troublesome it is to model fluid dynamics).

This is going to sound naive / stupid, but has anyone ever tried to document and electronically store snapshots of atmospheric conditions going back decades so that the computers can use any current snapshot today to search for a "best fit" from the past and then see what happened next the last time things were setup as closely? It would probably be even less accurate than the current computational modeling, given the lack of in-depth data from the past and the impossibility of two identical weather thumbprints within what is a "chaotic" system.

Just to add to what JH said, my understanding of these models is that they are indeed "trained" using existing known conditions. In other words the coefficients in the algorithms are "tuned" on known data. I am only basing this on knowledge of other computational models that are used to model fluid and thermal dynamics. It is possible to use biased data to cause a training bias within these models, however I suspect that given the maturity of some of them that any bias will have been ironed out as more and more data has been used in each training run of the next "beta" version. As JH mentioned, the real achilles heel for these models is the lack of solid input data. The pre-screening process interpolates many of the points between each of the real data points, data has to be time shifted as it is not all collected at exactly T0, etc etc - bottom line for me is they are a valuable tool but frankly rather useless in isolation.

[winter watcher]

90%- very high

UKMO looks a little progressive & deep with the shortwave at 144, this has unanamous model agreement- wait to you see the London ensembles- they will be tightly clustered-

S

Thanks Steve very much really do appreciate a response back. Happy new year to you mate and what a new years present it is to see these MEGA DREAM charts

[camperstron]

I might have this all wrong, but I think that the various models are really useful for getting an overall picture of how conditions might be. Right now it's all pointing to a cracking cold spell - but other events can evolve more rapidly and the overall trends can then dictate how they might pan out.

So even though the models are showing us a dry, cold period - over that timescale anything could happen.

Apologies if this is off topic for this thread, but I'm a very novice chart reader!!

[Nick L]

A clear example of an E.ly on the way when Kent has the lowest mean!

http://charts.netwea...12/t850Kent.png

Even a couple of runs touch -15C!

I think the most remarkable thing about that chart is the incredible agreement for such a long way out for very low temperatures.

[❄️❄️❄️❄️❄️❄️❄️❄️❄️]

A clear example of an E.ly on the way when Kent has the lowest mean!

http://charts.netweather.tv/ensimages/ens.20091231/12/t850Kent.png

Even a couple of runs touch -15C!

Pretty amazing to see ensembles for the somewhere in south of the UK where no member even touches -5C for 10 days!

[kold weather]

All I'll say about the ensembles is for the UK and cold, they are probably just about the best set we've ever had...many are still bitterly cold on the 14th which would make this a 2 week cold spell if that happened....and on our way to the coldest month since Feb 86 as well if some of the more extreme runs came off....

[cold snap]

A good example of never looking too far ahead when it comes to precipitation because these troughs often appear nearer the time.

[Frosty.]

What an amazing gfs 12z cold or very cold for the next 384 hours with N'lies, NE'lys, E'lys and no sign of SW'lys. My only concern is how the gfs flip flops from run to run, sometimes quite alarmingly, for example, we had a very good 12z yesterday followed by an 18z which was only half as good. There is no doubt we are in for another freeze and this one will probably be more severe than the last.

[Snowly does it]

All I'll say about the ensembles is for the UK and cold, they are probably just about the best set we've ever had...many are still bitterly cold on the 14th which would make this a 2 week cold spell if that happened....and on our way to the coldest month since Feb 86 as well if some of the more extreme runs came off....

Just get a nice surface HP over the UK and pulling a continental feed in and Feb 86 will look positively tropical!

First 27 days of Jan 1814 were sub zero - let's aim for the stars and beat that record

OK, fantasy cold aside, the outlook is very cold with snow showers possible throughout the next 8-10 days (anything falling from tonight on will be of snow I would say, apart from right on coasts where the sea ameliorates maybe) - where and when more organised snowfall occurs (it will) is down to more of a now and nearcast as the detail is painted in on the exact movement and flow.

The GH needs watching as does the trend for it's longevity - in terms of it collapsing, little danger for cold lovers in that as for the foreseeable there is a hugh area of LP propping up any SE movement and this will tend to increase the chances of a rebuild over Scandi and a renewed/continued push from the East.

Just like the December spell, this won't shift easily - and by the time it does (if it does in Jan) we will be at the ppoint of winter the cold stuff can really kick in.

Yum yum

[Snowyowl9]

I can`t see an end to this cold spell from T24......

Happy new year.

http://91.121.93.17/pics/Rtavn241.png

http://91.121.93.17/pics/Rukm1441.gif

http://91.121.93.17/pics/Rtavn3841.png

[BristolBaggie]

I can't see why everyone is getting excited to be honest. There is no point having a cold mainly dry 2 weeks.

It would be such a waste to have such an unusual cold spell to end up with no snow.

No snow = Rubbish winter

[keithlucky]

I can't see why everyone is getting excited to be honest. There is no point having a cold mainly dry 2 weeks.

It would be such a waste to have such an unusual cold spell to end up with no snow.

No snow = Rubbish winter

Patience good things come to those who wait.

[Duncan McAlister]

Patience good things come to those who wait.

@BristolBaggie

I think with the cold entrenched, the breakdown when it comes will be one mother of a battleground

Edited December 31, 2009 by Duncan McAlister

[Optimus Prime]

Patience good things come to those who wait.

Yeah a ban by the looks of things.

Snow will only be picked up nearer to the time, at the moment models are indicating very cold. The precipitation will come much closer to time.

[PB22]

Here comes the ECM 12Z: http://www.wzkarten.de/pics/Recm721.html

For those worrying about dry and cold, all I will say is that in order to get snow, you have to have the cold in place first. Get the cold established, and then who knows what may suddenly crop up.

[kold weather]

A clear example of an E.ly on the way when Kent has the lowest mean!

http://charts.netwea...12/t850Kent.png

Even a couple of runs touch -15C!

Yep some amazing runs from the 12z ensembles, with most of the milder runs at 850hpa having a cold surface feed coming in...

[Paul White.]

I can't see why everyone is getting excited to be honest. There is no point having a cold mainly dry 2 weeks.

It would be such a waste to have such an unusual cold spell to end up with no snow.

No snow = Rubbish winter

This is not a model output post. Model mood thread would be better. The 12z GFS looks very good for some sustained and prolonged cold plenty of convenction opportunities.