Met Office Winter forecast - Updated Feb 6th

[TEITS]

I agree entirely with your last sentence Steve and much of what you say and I'm glad my points have raised the discussion. The more research, the better. There is a key in there, it just hasn't yet been found.

Your 90% success rate, quoted there and just so nobody comes back with this, as figures like that tend to stick, is not a prediction success rate. I do appreciate it could bode well for the future.

Again, I mentioned that no-one, to my knowledge has achieved a 70% success rate at predicting seasonal weather accurately. That actually means that the success rates are less than 70%, maybe 2 out of 3. The Met Office and others, have a number of paying customers and some large utilities either increase, or run down their stocks, based upon the LRF they pay for, that is true; last year's leak of the "colder than average winter" came from an ill-advised article by a UNISON member, following a Met Office presentation to their commercial customers, on exactly that topic, but only 2/3 is one hell of a gamble and I'm sure they know it. It would not be hard to imagine the furore when a cold winter does come along (and it will) and the 1 in 3 chance of the forecast being wrong drops on that year (which, statistically, it will). It is that which will be remembered.

The number of commercial customers has grown in recent years and their expectation of getting their money's worth, with a correct forecast has grown with it. I only hope that the advances in forecasting techniques improve before the combination of a cold winter and a less than 70% success rate collide.

Paul

PS If I'm wrong about the <70% and an organisatiuon has demonstrated a >70% accuracy, whose dataset is statistically significant at the 95% level (ie is long enough to demonstrate reliability), tell us, anyone, please. WF? You seem to have some knowledge here? Now that would be a forecast worth investigating. I know Piers Corbyn claims a success rate greater than that, but he has never made his research available to the scientific community, claiming (perhaps rightly) commercial confidentiality.

Im curious Dawlish at what timeframe do you class a forecast as being a LRF week,fornightly,monthly,seasonal?. The reason I say this is because due to our varied climate I believe any forecast beyond a week in the UK can be classed as a LRF. Because of this I would rate the Met O weekly,fortnightly forecast's as being very accurate.

Another point I wish to address is your comment in your sig "Event Spot 90% success". Now Im not having a go but I think im right in saying that you use the GFS for these spot's?. Do you wait to see a pattern to emerge at +240 for example and if so then claim you have spotted something?, if this is the case then you cannot really claim any credit for this because the supercomputer is doing the forecast for you. What would be more impressive is to spot something when the GFS is all over the place at this timeframe.

In my experience, those who shout loudest about their supposed wealth of knowledge are often the very same people who lack a true understanding in that subject.

I can relate to this. When I first joined this forum I admit I was rather cocky, arrogant, loud, brash and thought I knew more than I did. In time I have realised my knowledge is nowhere near some on here hence my lower profile these days.

[Guest Mr_S]

Yes accuracy would be needed (or prefered) in a LRF but fact is as i mentioned in my other post within this topic, No person - or computer, given the length to which the METO are INDICATING the winter *could* go.. Aka December-ish - could be 100% spot-on, Dead on balls accurate.

I Refuse to belive that anyone or anything could actually sit there and forecast for one season, the day by day events for things such as pressure change, cloud cover, mm of rain, mph of wind, conditions in general, air temp, EVERYTHING... And to forecast is to such accuracy.. you can nail it down to the second, minute, hour of "said day" which from some of the posts im reading are what they are expecting of the METO.

This isnt a post to cause problems and im sorry if it has come across strongly worded and will openly apologise on the board in public if anyone takes this post as a direct "shot" at them - its not. It's just i feel some are expecting way to much from the metoffice that when in reality 1) i doubt their own ability, and 2) its MONTHS away.

Look as TEITS Said,

The reason I say this is because due to our varied climate I believe any forecast beyond a week in the UK can be classed as a LRF

A week in our weather is difficult to forecast, which is pretty much true.

Look at Gordon, he had the majority of named Forecasters HERE hanging fire on casting where he was going for definate up to around 24hrs away - and thats a storm we can SEE on satalite and can actually track.

I'll get off my soap box now

ps, TEITS, Steve, WF... Good to see you all around for the coming seasons once again welcome back

[coldfingers]

Long range forecasts are interesting and long may people continue making them 70% is better than nothing......

Well certainly for down here last winter's LRF by the MetO was 100% accurate. We had exactly what they said, less precipitation than average and colder than average.

Hope they can achieve such success this year but if not I wont be moaning about it.

[Dawlish]

I have read many of your posts and these are what lead me to my conclusions. In my experience, those who shout loudest about their supposed wealth of knowledge are often the very same people who lack a true understanding in that subject.

I am sorry, but it was you who claimed to have not seen a forecasting organisation with a verifiable success rate of long range forecasts above 70%. If you now want to go back on that figure it is yet another example of you having far less knowledge than you like to claim. I don't expect that there is an organisation who can claim with any substantiation that their long range (3 months or more) forecasts are continually at or around 70% accurate.

This does not mean though that there is no value in long range forecasts, you're attempting to speak for commercial organisations with whom you have no link or understanding. The advances in long range forecasting have helped to save and make significant sums of money for those who have used them. Forward planning is essential in many, if not every industry and any tool which can aid that, even if it is relatively experimental is of use.

You also must remember that the forecasts and press releases you see publically are not the products that commercial organisations receive, so in essence you are correct, the press release regarding this winter from the met-office does not have a great deal of commercial use. That is irrelevant though as it wasn't produced with commercial clients in mind.

Methinks you go a little too far in questioning knowledge when you know little of the person, but I'll leave others to judge on that one and try not to leave insults in return.

If, as you say, WF, "I don't expect that there is an organisation who can claim with any substantiation that their long range (3 months or more) forecasts are continually at or around 70% accurate", doesn't that lend, at least some, credence to my point that long-range forecasting is not a terribly accurate process? Indeed, can you lead us to an organisation that claims a verifiable accuracy greater than the hindcast accuracy of the Met Office's NAO forecast (between 64 and 66%) which led me to suggest that no organisation has an accuracy of over 70%, in the first place?

For me to suggest that the actual accuracy may be even less than the 70% I first quoted is hardly going back on my assertion - think about it, it actually firms up the opinion, not diminishes it.

Regards, Paul

[BrickFielder]

The quoted hindcast accuracy of the Met Office's NAO forecast (between 64 and 66%) is based on the performance of the HADCM3 and GLOSEA models. 66% is a good representation of the performance of the MetOffice's seasonal forecast accuracy over the last 20 years. The metOfficeare now using HADCM4, glosea2 and HADGEM and are currently testing HADGEM2.

The demeter project documentation suggests the best accuracy for a single model is currently about 81% but this was an equatorial seasonal forcast. In fact most long range forecasts covering equatorial areas are usually reasonably accurate.

Demeter Project

The UCL Climate Extremes group have developed skillful probabilistic forecasts for the winter seasonal and monthly AO. The physical basis for these models is knowledge of prior zonal winds in the Northern Hemisphere stratosphere and upper troposphere. Robust hind casts for the period 1958/9 to 2005/6 show that the model anticipates correctly the anomaly sign of the winter AO in 68% of years, and the anomaly sign of the month-ahead December, January and February AO in 71%-79% of years.

All in all I would say 70% accuracy is a fair reflection on the performance of long range model forecasts although I have a feeling the MetOffice forecasts may now be sneaking above that figure.

[Dawlish]

The quoted hindcast accuracy of the Met Office's NAO forecast (between 64 and 66%) is based on the performance of the HADCM3 and GLOSEA models. 66% is a good representation of the performance of the MetOffice's seasonal forecast accuracy over the last 20 years. The metOfficeare now using HADCM4, glosea2 and HADGEM and are currently testing HADGEM2.

The demeter project documentation suggests the best accuracy for a single model is currently about 81% but this was an equatorial seasonal forcast. In fact most long range forecasts covering equatorial areas are usually reasonably accurate.

Demeter Project

All in all I would say 70% accuracy is a fair reflection on the performance of long range model forecasts although I have a feeling the MetOffice forecasts may now be sneaking above that figure.

Thank; sounds about right BF. I'm sure the forecast accuracy will improve, last winter will have helped a little, as they were certainly the best of all the agencies who issued a forecast. Dunno if the Met Office models are under, or even slightly above that figure of 70%. It would be helpful to all if they published their forecast accuracy on their site. It's a shame we are speculating on it. I might be completely wrong and the Met Office may be hiding their LRF accuracy and its statistical confidence under a 90%+ accuracy and 95% confidence barrel, but I'm very sure that's not true.

In the end, thank goodness there is so much inaccuracy! I've said many times that I'd lose interest in forecasting, if anyone got it right all the time. It woukld just be so BORING to know what the coming winter was actually going to be like.

Paul

PS I haven't forgotton your enquiries Eye, I'll get there. Lots of other things on today!

[Dawlish]

Im curious Dawlish at what timeframe do you class a forecast as being a LRF week,fornightly,monthly,seasonal?. The reason I say this is because due to our varied climate I believe any forecast beyond a week in the UK can be classed as a LRF. Because of this I would rate the Met O weekly,fortnightly forecast's as being very accurate.

I'm close to agreeing with you, Eye and 12 months ago, I would probably have agreed with your 1 week idea, very closely, though after my experiments with forecasting events, this year, I'd extend that to a max of around T+300.

Paul

Another point I wish to address is your comment in your sig "Event Spot 90% success". Now Im not having a go but I think im right in saying that you use the GFS for these spot's?. Do you wait to see a pattern to emerge at +240 for example and if so then claim you have spotted something?, if this is the case then you cannot really claim any credit for this because the supercomputer is doing the forecast for you. What would be more impressive is to spot something when the GFS is all over the place at this timeframe.

I can relate to this. When I first joined this forum I admit I was rather cocky, arrogant, loud, brash and thought I knew more than I did. In time I have realised my knowledge is nowhere near some on here hence my lower profile these days.

This might explain. It from a post of mine, some time back:

"I feel that; if an "event" shows itself, T=300, or more, sticks for a few runs and disappears, it is worth remembering. If it re-appears, several days later and sticks again, my confidence in it actually happening increases. At some point, perhaps about T+240+ to T+200, I'll be confident enough to call it as a spot, if it is still there. At that time, judge whether I'm right, or wrong (or not!) when the "event" finally happens, or disappears into randomness.

My reasoning for this is coincidence: reappearing events, after the parameters have changed (the numerical inputs) over several days, suggest something which is coincidental but any sticking on the reappearance suggests something greater than coincidence. I mean; why would a gfs event at long range, plotted from numerical inputs, reappear? Surely at that distance in the future it should be random?

My null hypothesis is that events cannot be forecast at >T+200+. If I achieve anything but the null hypothesis, with a 75% success, I'll believe there is something in this. 66-75%, it is worth further testing. <66% and it is too random for me to believe and I'll accept my null hypothesis. You may have to put up with me for sometime, or I may get bored, or whatever (!), but for the time being, I'm enjoying testing the theory!

I'm prepared to do this in public and I'm prepared to fall flat on my bum. There's the interest. It is an unknown, it is a theory I've developed over the last year of using the gfs charts and I'm interested enough to test it. My changes in confidence will be shown in any odds that I offer on the event actually happening. It'll be fun for some to get involved, but there is absolutely no need(!); these boards are all about one person's interpretation vs another's - that's the basis of other netweather competitions too. I view the odds as a learning tool about the gfs, for myself and anyone else who may be interested. I might get sick of doing that sometime and it might be a fad, but for the time being, I'm finding it fun and it concentrates my own mind on following the gfs closely, even if no-one takes me up on the odds I offer!"

Regards, Paul

[smich]

Paul, I'm an infrequent poster here these days (and an even more inexperienced forecaster for that matter) but I'm stung into action by this thread! Your enthusiasm for the subject is to your credit and I don't think you deserve some of the posts aimed at you here.

Personally, I think anything approaching 70% accuracy would persuade me to reach for my last £100 million if it could be proven accurate. But the odds can get so long in this game. What if I offered you 100/1 that 2 of the last 3 months would exceed the CET average by THREE DEGREES!!!!It would have been a no-brainer, and it could happen!

IMHO, LRForecasting will improve enormously in the next few years, but whether the definitive measurements of GW over the last few years will contribute to that I'm not so sure. It's new territory. Teleconnections and SSTs will be the benchmark data for the timebeing, I think, for better or worse.

In any case big up yerself for (albeit inadvertantly?) provoking this debate!

Smich

[bigboblee]

I think I would trust the METO forecast as much as any other 'reputable' organisation - and probably more so. I would say last years winter prediction from them was a complete success in the context of what could reliably be expected to be accurate - if you are to use the 70% accuracy rule as a baseline expectation. The only really inaccurate part of the prediction was the -NAO which bumbled along around neutral, but in most other respects it was fairly spot on IMO.

A seasonal forecast should have personal expectations managed along the baseline parameters/precentages of accuracy it is set up with - otherwise disappointment and criticism is bound to follow. Such disappointment and criticism can be misplaced though if personal expectation and interpretation of results exceed the baseline parameters and percentages.

On the basis of current predictions for this winter those who want cold weather and snow have reasonable cause for some optimism - we will see.

Tamara

I am sorry if i am being a right old sceptic here but i JUST do not get what all the hype is about?

I read the MetO's revision and there claim that winter will turn colder later on and upon reading the weather page in the Times today I am EVEN more sceptical.

They repeat the comments on the UKMO's website but go further and say that it starts from the theory of two Swiss scientists who say that the extreme cold of 1941 to 1942 was POSSIBLY caused by an El Nino!!!!!!!!

Is that why the MetO changed its tune? The Work of just two scientists?

It clarifies the Late Winter Cold spell as coming Feb to Mar 2007,NOW really we are surely entering silly territory here as when was the last time you saw ANY forecaster foretell the weather THAT far ahead?

We are entering the first week of October and they are telling us to expect an event 5 - 6 months ahead - COME ON!

Remember back in the spring the MetO issued a Summer 06 forecast that said the summer would be wet with below average temperatures,then after the heat in July the prediction was August would be EVEN hotter and look how that turned out - mild yes but no - one could call it record breaking.

Before anyone calls me a mild - ramper,i would still pour a lot of scorn on the LRF if it predicted record breaking heat for 5 to 6 months ahead into the Winter - SORRY I just do'nt get it!!!!!!!!!!!!!!!!!!!!!!!!

[kold weather]

Not much I can really add to this deabte but I'd personally take anything that the Met-office says over most peoples...unles I can see evidence that would lead me to believe that actually it may not turn out right, such as unforecasted strengthening of El ninos that models may not pick up quite so well!

One good sign is that the jet stream overall still seems to be quite southerly tracking so far and this hasn't changed since summer either, though the jet's reent northern swing weas helped solely by the ex-hurricanes. This is also agreeing with Steve said about unusual storm tracks.

[The Eagle]

Im concerned about the models this year because there seems to be alot more inconsistency. ATM we cant look at 168hrs and see it as even credible and no models agree even at 120 - 144hrs. I think it may be more tricky for us this season, though hopefully that will change as we head toward December. As regard the MO forecast, its far to vague really.

[bigboblee]

Not much I can really add to this deabte but I'd personally take anything that the Met-office says over most peoples...unles I can see evidence that would lead me to believe that actually it may not turn out right, such as unforecasted strengthening of El ninos that models may not pick up quite so well!

The last El Nino was what 1997 - 1998(?) and if memory serves me right 1998 was relatively mild - NO severe winter there,BUT according to my diary (JUST checked for that winter) here in the south east of the Uk and my town Herne Bay in particular there was a period of heavy snow/blizzards in early January followed by wet and windy weather from late January interspersed with calm foggy days which lasted well into March and the spring of that year and it got progressively mild too - NOt proof to poo - pooh the whole MetO thing BUT food for thought???????????

[Big Bear]

Remember back in the spring the MetO issued a Summer 06 forecast that said the summer would be wet with below average temperatures,then after the heat in July the prediction was August would be EVEN hotter and look how that turned out - mild yes but no - one could call it record breaking.

Actually, August was average.

0.1 degrees above average to be precise.

The Met office may have got last years LRF quite right but remember the success rate is based on wild fluctuations between reasonably accurate - to wildy innacurate and we just don't know what is around the corner besides another probable mild & westerly dominated Dec/Jan/Feb

[SMU]

I am sorry if i am being a right old sceptic here but i JUST do not get what all the hype is about?

I read the MetO's revision and there claim that winter will turn colder later on and upon reading the weather page in the Times today I am EVEN more sceptical.

They repeat the comments on the UKMO's website but go further and say that it starts from the theory of two Swiss scientists who say that the extreme cold of 1941 to 1942 was POSSIBLY caused by an El Nino!!!!!!!!

Is that why the MetO changed its tune? The Work of just two scientists?

It clarifies the Late Winter Cold spell as coming Feb to Mar 2007,NOW really we are surely entering silly territory here as when was the last time you saw ANY forecaster foretell the weather THAT far ahead?

We are entering the first week of October and they are telling us to expect an event 5 - 6 months ahead - COME ON!

Remember back in the spring the MetO issued a Summer 06 forecast that said the summer would be wet with below average temperatures,then after the heat in July the prediction was August would be EVEN hotter and look how that turned out - mild yes but no - one could call it record breaking.

Before anyone calls me a mild - ramper,i would still pour a lot of scorn on the LRF if it predicted record breaking heat for 5 to 6 months ahead into the Winter - SORRY I just do'nt get it!!!!!!!!!!!!!!!!!!!!!!!!

Ignore the guy in the times-

S

[kold weather]

1997-1998 El nino was immense, only the 1983 El nino can match it in terms of strength, the El nino this year isn't going to be that severe, or at least its not going to have time to get that severe before our winter is over.

Certainly though the southern part of the jet is looking stronger then in recent years which is also a good sign, though that did allow for July to be so very southerly as well when those LP's couldn't get no further eastwards then the Azores region. This has also shown on the hurricane season and shear has been quite disruptive dow nthere year, a good sign of a stronger then normal southerly jet.

(I presume you are talking about 1997 winter, started out pretty cold for Dec and Jan, then turned very mild in Feb)

[bigboblee]

Im concerned about the models this year because there seems to be alot more inconsistency. ATM we cant look at 168hrs and see it as even credible and no models agree even at 120 - 144hrs. I think it may be more tricky for us this season, though hopefully that will change as we head toward December. As regard the MO forecast, its far to vague really.

Vague it is my friend,WELL it might be cold this winter or it might not and there COULD be an exceptional cold spell later OR NOT!!!!!

Then i ask why mention those 29 years 1971 to 2000? seems like they are saying take your pick here as Winter COULD follow the same trend of one of these!

Why did they simplly not say HELP we do'nt know as the forecast models are all over the place and telling us different things? :

Actually, August was average.

0.1 degrees above average to be precise.

The Met office may have got last years LRF quite right but remember the success rate is based on wild fluctuations between reasonably accurate - to wildy innacurate and we just don't know what is around the corner besides another probable mild & westerly dominated Dec/Jan/Feb

That will do me!!!!!!!!!!!!!!!!

[shuggee]

It's getting a bit confusing (for me anyway - so please bear with me as I'm simple) having two threads about the forthcoming winter. So I'm popping this thread into the Media/Internet Forecast Discussion area to help separate the two discussions - which are intrinsically linked but distinct.

So please continue to discuss the Meto Forecast here

[Big Bear]

Hi Steve

When are you releasing your first winter LRF installment?

I really respect your knowledge & take your forecasts with no added salt.

[bigboblee]

1997-1998 El nino was immense, only the 1983 El nino can match it in terms of strength, the El nino this year isn't going to be that severe, or at least its not going to have time to get that severe before our winter is over.

Certainly though the southern part of the jet is looking stronger then in recent years which is also a good sign, though that did allow for July to be so very southerly as well when those LP's couldn't get no further eastwards then the Azores region. This has also shown on the hurricane season and shear has been quite disruptive dow nthere year, a good sign of a stronger then normal southerly jet.

(I presume you are talking about 1997 winter, started out pretty cold for Dec and Jan, then turned very mild in Feb)

Thats the one matey.

Ignore the guy in the times-

S

AAAAAAAAARRRRRRRRRRRRRGGGGGGGHHHHHHHHHHHH!!!!!!!!!!!!!!!!! So just who do i believe? God i hate this time of year as here we go again with the mild vs cold dbate and im getting a headache!

[SMU]

Hi Steve

When are you releasing your first winter LRF installment?

I really respect your knowledge & take your forecasts with no added salt.

It was/ is going to be Nov sometime- I did however want to get the forecast out before the next Meto update- however that may not be possible-

Anyway to clear up confusion in this thread my BEST guess around the METO offices change of forecast are for 3 reasons-

1) Atlantic SSt's- The Base for the UK met office forecast- at the time of the original thoughts the Massive +VE anomaly looked like pointing towards a mild +VE/Neutral NAO Winter- However we have undergone changes in the last 5/6 weeks-

Atlantic SST's

Maps courtesy of NineInchNails Eastern US

The formation of colder Surface temps & The migration North of the Large +VE anomaly has thrown enough of a spanner in the works to preclude the additional warning from the Meto- If that colder surface continues to grow then the atlantic takes on a more familiar -NAO tripole-

Heres what were aiming for- NEG NAO SSTA

2) El- nino to come in WEAKER than expected - The inital Long range Elnino forecasts did point to a stronger event, however we are now looking at a weaker affair- probably weak to moderate which is 'good' in terms of analogues for -NAO & Colder UK Winters-

3) The AO- wildcard- In the milder winters of the 90's the AO was very Positive, however a succession of warming events in the stratosphere last winter assisted in Large +VE anomlies over the pole for most of the Winter, these drove Heigher heights South PACIFIC side at the start of the Winter & Heigher heights south Towards Europe at the end of the Winter-

Clearly September ISNT the time to be issuing a Winter forecast & parameters are Still changing- there is STILL enough scope to go either way, the Met office release is sensible for this early stage going on the preliminary indicators- The update in November will hold a lot more weight.........

Steve

Thats the one matey.

AAAAAAAAARRRRRRRRRRRRRGGGGGGGHHHHHHHHHHHH!!!!!!!!!!!!!!!!! So just who do i believe? God i hate this time of year as here we go again with the mild vs cold dbate and im getting a headache!

Paul S is a Journalist NOT a Meteorologist.........

S

[lesta_snow]

what was the winter average in 1941/1942 eg. dec,jan,feb

[SMU]

what was the winter average in 1941/1942 eg. dec,jan,feb

6.6C- Ave 2.2C-

Dec 5.6C, jan 0.9C Feb 0.1C-

I would take another 1941/42 Winter, Im sure some would differ though-

El NINO - especially a weak to moderate one CANNOT be the ONLY factor in contributing to a severe Winter.....

S

[lesta_snow]

thx mate

[Dawlish]

Paul S is a Journalist NOT a Meteorologist.........

S

OOOOO steady on Steve. Does that mean that anyone without qualifications in Meteorology can't comment on weather and climate and be believed?

That certainly counts me out, for my fellowship of the Royal Meteorological Society was gained in the days before it was an indication of professional competence and it certainly discounts Prof. Gordon Manley. Geography degrees apparently don't count and are looked down upon in meteorological academia, for sure, as I have found out to my cost! Lewis Fry Richardson had a degree in Physics, not meteorlogy! Piers Corbyn has a degree in Astrophysics, not Meteorology! I could go on and on!

Paul Simons in the Tombs has more than a good understanding of meteorology and, though he may not have a meteorology degree (I don't know that for a fact, I'm inferring it from your post), he is certainly a meteorologist, in my eyes.

Paul

[TEITS]

what was the winter average in 1941/1942 eg. dec,jan,feb

Some great chart's for you to look at.

http://www.wetterzentrale.de/archive/slp/1...slp19420217.gif

http://www.wetterzentrale.de/archive/slp/1...slp19420224.gif