The Winter 2011/12 Forecast Thread

[A Winter's Tale]

Maybe because he used the phrase 'Negative Arctic Oscillation' so frequently, I got confused and forgot when he was referring to? However, you can't argue that he wasn't right about the snowfall pattern starting to become more widespread 15-20 December, as many Southern areas saw their first snow of the season on Friday. Plus, I didn't argue about his 'White Xmas' possibilities as I myself find that a little strange, so please don't try to make out that I was backing up that statement.

Listen here. The forecaster who has been spot on so far is Roger J Smith. This forecasters idea of widespread snow between 15-20 would be something like a severe cold snap, and I think some southerners on here would greatly disagree about how many southern areas so their first snow. I do have some respect Mark, but I think his big freeze forecast was just to get money and publicity. Look at Simon Keeling, that cold and dry Christmas has gone oot the window. But there's no point arguing or anything, just need to look at the models and stratosphere threads aswell as forecasters who actually use valid factors in their forecasts.

[damianslaw]

I've watched a few of Mark Vogan's presentations now, and I actually don't mind his style and I am not that fussed he doesn't back up his thoughts with scientific evidence etc.. he doesn't have a very bullish style unlike some I have to say.

However, I did find it quite perplexing how he suddenly kept mentioning the impact of the volcanic eruptions over Iceland and how this will be a key contributing factor for much colder winters in the decades to come. Not sure on that one Mark..

[TEITS]

Nicely done Dave.

Clear and straightforward so any person off the street can understand.

Give us a smile to the camera next time! :smiliz64: .

Cheers mate.

I did toy with the idea of making a longer more indepth forecast. However my main purpose was to make it short and simple. The forecast by James Madden is very irresponisble in my opinion especially as many people plan to travel during the xmas period. I just hope his forecast isn't published by the Daily Express!

I shall be interested to see what he has to say when he realises his forecast was woefully wrong. I can't see how he can use any spin to get out of this one!

[BLAST FROM THE PAST]

DECEMBER may be a roller coaster of mild and cold spells. The coldest part of the month is likely to come around the 15th to 20th. Before that, the first two weeks may include some strong winds from deep Atlantic lows, heavy rainfalls, brief intervals of hail or snow, but eventually, arctic air should develop and there could be snow on the ground at times just before Christmas. Then the Christmas holiday period is expected to turn milder with fog and rain especially around the 27th when it could once again become rather stormy. By New Years Eve it may be turning a lot colder.

Thus far this looks pretty close to the mark and looking at the models, still looking very good. All eyes now on the 27th and the now modelled storm for 27th. Will it won't it? Then period 7-10 Jan, could be the pivotal moment of winter.

BFTP

Edited December 22, 2011 by BLAST FROM THE PAST

[mountain shadow]

Thus far this looks pretty close to the mark and looking at the models, still looking very good. All eyes now on the 27th and the now modelled storm for 27th. Will it won't it?

BFTP

A bit disapointed with Nick in the Model Output thread rubbishing RJS's forecast because it names specific dates for events. What probability would a maths expert give on RJS accuracy being just down to guess work?

Still, the proof will be how accurate the rest of the winter is.

Edited December 22, 2011 by mountain shadow

[A Winter's Tale]

A bit disapointed with Nick in the Model Output thread rubbishing RJS's forecast because it names specific dates for events. What probability would a maths expert give on RJS accuracy being just down to guess work?

Still, the proof will be how accurate the rest of the winter is.

Hmmm, well the 27th is yet to come but the specific dates that have been passed such as the cold snap sometime between the 16th and 22nd was particulary excellent. His overall theme of the month was spot on and events such as the rollercoaster of mild and cold spells, the stormy week or two, the cold snap, the mild christmas have been brilliantly accurate.

[BLAST FROM THE PAST]

A bit disapointed with Nick in the Model Output thread rubbishing RJS's forecast because it names specific dates for events. What probability would a maths expert give on RJS accuracy being just down to guess work?

Still, the proof will be how accurate the rest of the winter is.

Indeed let's see how it continues and it will be judged on its total merits but nice to see a good start by some big hitters. Unsurprisingly I endorse it

BFTP

[phil nw.]

Thus far this looks pretty close to the mark and looking at the models, still looking very good. All eyes now on the 27th and the now modelled storm for 27th. Will it won't it?

BFTP

I can see what you mean about the modelled post Xmas low.

It has potential to deepen as shown on 12z ECM .

I have an open mind on alternative methods Fred such as yours and RFJ and will follow with interest.

I hope you understand that in the Mod.Disc Thread we have to be more objective and only refer to, shall i say, convential modelling output.

[BLAST FROM THE PAST]

Hmmm, well the 27th is yet to come but the specific dates that have been passed such as the cold snap sometime between the 16th and 22nd was particulary excellent. His overall theme of the month was spot on and events such as the rollercoaster of mild and cold spells, the stormy week or two, the cold snap, the mild christmas have been brilliantly accurate.

15-20 AWT

BFTP

I can see what you mean about the modelled post Xmas low.

It has potential to deepen as shown on 12z ECM .

I have an open mind on alternative methods Fred such as yours and RFJ and will follow with interest.

I hope you understand that in the Mod.Disc Thread we have to be more objective and only refer to, shall i say, convential modelling output.

Yes of course and agreed Phil hence why I've made a couple of entries here and PM'd Nick re furthering converstaion rather than off topic posts.

Fred

Edited December 22, 2011 by BLAST FROM THE PAST

[Bristawl Si]

Indeed let's see how it continues and it will be judged on its total merits but nice to see a good start by some big hitters. Unsurprisingly I endorse it

BFTP

Indeed let's see how it continues and it will be judged on its total merits but nice to see a good start by some big hitters. Unsurprisingly I endorse it

BFTP

Indeed, Roger's forecast has been pretty much spot on so far.

Looking forward to mid-Jan now....and it's interesting (and funny) to see some members in the Model Discussion Thread getting 'ahead of the curve' showing disappointment at little signs of change in the models over the next week.

GP's forecast, as well as Roger's, talks about change from mid-Jan onwards, so patience is required.

My own gut feeling is that this winter, for these parts (Imby terms), will see little cold and snow - a return to a more typical average winter for Bristol and surrounding areas.

Hope i'm wrong though.

Edited December 22, 2011 by Bristle boy

[Catacol]

Hmmm - i put a post in the model thread about the accuracy of RJS's forecast, come back to my puter and find my post mysteriously gone and people talking about some rubbishing of RJS. Did I provoke something?

Just to say again - staggering accuracy for me of RJS's forecast. Pattern correct, dates correct. Astonishing.

[nick sussex]

A bit disapointed with Nick in the Model Output thread rubbishing RJS's forecast because it names specific dates for events. What probability would a maths expert give on RJS accuracy being just down to guess work?

Still, the proof will be how accurate the rest of the winter is.

I don't think I was actually rubbishing it just questioning how a LRF can put dates on certain events.

As you said though the proof will be in its accuracy over the whole winter.

[BLAST FROM THE PAST]

Hmmm - i put a post in the model thread about the accuracy of RJS's forecast, come back to my puter and find my post mysteriously gone and people talking about some rubbishing of RJS. Did I provoke something?

Just to say again - staggering accuracy for me of RJS's forecast. Pattern correct, dates correct. Astonishing.

No but several of the posts were removed for being non model related hence why we're posting in here. I'm a ricidivist :smiliz34:

BFTP

[Catacol]

I don't think I was actually rubbishing it just questioning how a LRF can put dates on certain events.

As you said though the proof will be in its accuracy over the whole winter.

But surely that's the point? RJS put a 5 day bracket on the coldest spell which materialised, put 2 weeks on zonal stormy stuff which materialised beforehand, and now may have pinpointed the storm on 26/27 Dec. I am staggered - for all the talk of no SSW and strong vortex and amplitutde or non amplitude of the pattern etc etc Roger put a 4 week spell of UK weather together 4 weeks before it started and has got it right. Is that just luck? Could be I suppose, but he got much of last year right (the first year I read NetW) apart from Feb where his instincts were to go mild but bowed to BFTP's feelings that his data suggested possible cold. Had he stuck to his instincts then last year's forecast would also have been very close.

I am not trying to start an argument - just wondering aloud really whether the rather "odd" methods that he bases his forecasts on may actually have real substance.

Chicken and Egg? Which is which?

[leicsnow]

I can't find RJS's forecast from last year. I want to view it so I can compare to statistics? Anyone help providing me with a link?

And yes, his forecast does look very accurate indeed. Hope the rest of his forecast comes off!

[A Winter's Tale]

I can't find RJS's forecast from last year. I want to view it so I can compare to statistics? Anyone help providing me with a link?

And yes, his forecast does look very accurate indeed. Hope the rest of his forecast comes off!

http://forum.netweather.tv/topic/64878-rjs-and-bftp-winter-201011-lrf/

[Roger J Smith]

Last year is more or less irrelevant, the research program is extremely complex and requires me to reset mentally every year around April-May, something I have programmed into my work so that I don't lose track of what is already "working" from validation studies. People probably figure that these forecasts are the result of five minutes' peering into a crystal ball and a channelling session, but actually, they are based on extensive number crunching done at painfully low speeds on a home computer, research into possible new variables to add to known working factors, etc ... and then some in-depth pondering of what the output means in terms of a circulation pattern, because I don't have nearly the time or capacity here to design a global model, I am working from index values (like the ensemble data you see in shorter time frames). This is why specific dates or ranges of dates get mentioned, whenever sharp changes appear in the index values, these are ascribed to specific events that might be expected as forcings as the index values change.

Hope this does not sound too mysterious, it is meant to be very simple. An index value, in case not apparent on first reading, is a continuous measurement based on a given variable or weighted combination of variables, that corresponds to some meteorological parameter you are studying from a data base. It stands to reason that if you have a good data base and there are real signals, then a number crunch correctly designed and executed might expose those signals even for predictive purposes. The forecast is a measure of how well or poorly that work is being done at any given time.

But there is a subjective element, a sort of "how does this actually happen" stage to the forecast, that requires one to ponder analogues where similar trend curves might show up, and to go over the process, for example, is it progressive or retrograde, is it associated with a strong jet or a split flow, etc. My experience, and some of your feedback supports this, is as follows -- if the overall pattern is essentially well-predicted, then some surprisingly accurate details follow, because these details tend to be like passengers on a roller coaster, they show up at predictable times. But only if the roller coaster is heading through the UK and not say Iceland or Spain.

As my expectation from past forecasts is a rather modest 60 to 70 per cent range of gross validation (right side of normal) one might say that from that average level of performance, we might reasonably expect this current forecast to fall off in accuracy at stages in this next two months. It has been running well above 60 per cent in December so far and looks to stay on that pace for a while longer. So it's now a wait and see exercise, are we seeing a more accurate forecast level now that I have more data and in particular daily CET data in the mix, or will we end up with some periods of wrong circulation where the details are bound to be wrong as a consequence (you can't get a raging easterly blizzard in a retrograde Atlantic high, for example, nor can you get a mild southerly flow when the jet is pushing solidly south into the Med).

If over five years the track record was just 50-50 on gross validation and always hit or miss, then I would throw in the towel and stop trying to do this. But quite honestly, the results have suggested better than random and an upward trend, plus, I have a lot of experience with this system in North America and have seen enough better-than-random outcomes to encourage me to continue there also.

You would think the community would be all over this research, but I guess not. Should some research interest be shown? You tell me. Post 174 of the white Christmas thread could be used as a case in point. I mean, what do I have to do to demonstrate non-random significance? Maybe a reality check is needed here, there is no telling how many more years I have left to work and just because I can do this does not mean that I could walk into an office some place and five minutes later, everyone else could do it. This is the result of 30 years of effort.

Edited December 23, 2011 by Roger J Smith

[BLAST FROM THE PAST]

just because I can do this does not mean that I could walk into an office some place and five minutes later, everyone else could do it. This is the result of 30 years of effort.

Very true

BFTP

[Snowmad79]

Without disrespecting anyone concerning Rogers forecast last year, December had the most descriptive run-down. I'm only sceptical of that particular months forecast because it was released on November 15th 2010 when models were already starting to firm up a severe cold blast and around that very time BFTP was bigging it up as if he already new it was coming, when prior to the 15th 2010 the models weren't at all keen on any kind of cold snap... far from it.

In simple terms he went for Cold from MID December, a mild January and another arctic blast in February with periods of heavy snow... How it panned out was Severe cold at the end of November into December, warmed up for January and remained mildish and dry through Jan and February.

So when you think of that forecast as a whole, was it really that accurate ?

Again, no disrespect intended, but imo a bit too soon to be calling a winter forecast "very accurate"

[johnholmes]

.

I can understand your feelings Roger about so little interest from the areas you need to show that interest to get help proceeding with your research.

The figures you quote, of 60%+ and showing a rise I would have thought is pretty encouraging for lrf work.

Trying to mark ones own work is incredibly difficult trying not to have any bias. Again its not like issuing a forecast for a spot location or smallish area for 4-5 days ahead. So again trying to be objective in marking is hard. Does one call it a correct forecast if its within 2-3 days of a suggested event? Does one say an event will occur within x-y day timeframe and if its 1-2 days outside it is that still a hit? I'm not criticisning Roger just trying to explain to others how hard it is to be seen to be objective in forecasts such as you and GP prepare.

Keep doing it Roger I reckon its not just chance that so often you are as near as anyone else and you do it pretty regularly both short and longer range. I don't profess to be able to understand your methods in spite of your tutorials a year or two ago but there seems to be something there.

Edited December 23, 2011 by johnholmes

[VillagePlank]

What probability would a maths expert give on RJS accuracy being just down to guess work?

You'd need the forecast in a quantitative fashion (something like average CET, total precipitation, maximum pressure, minimum low for each week in question; not on a week to week basis, but on a seven day rolling period - to allow for serious events to be forecast, say, a day or so out - so think 1st Dec - 7th Dec, 2nd Dec - 8th Dec)

You'd also need to know all of the factors used to generate such a forecast (sun, moon, Jupiter, whatever) so we can work out the degrees of freedom

We can then compare observed with actual - we can then test out the hypothesis (that RJS forecast has skill) against the null hypothesis (that it's a random walk) and assign a statistical certainty as to how much down to chance the forecast was.

I'd be delighted to crunch the numbers.

Edited December 23, 2011 by Boar Wrinklestorm

[BLAST FROM THE PAST]

Without disrespecting anyone concerning Rogers forecast last year, December had the most descriptive run-down. I'm only sceptical of that particular months forecast because it was released on November 15th 2010 when models were already starting to firm up a severe cold blast and around that very time BFTP was bigging it up as if he already new it was coming, when prior to the 15th 2010 the models weren't at all keen on any kind of cold snap... far from it.

In simple terms he went for Cold from MID December, a mild January and another arctic blast in February with periods of heavy snow... How it panned out was Severe cold at the end of November into December, warmed up for January and remained mildish and dry through Jan and February.

So when you think of that forecast as a whole, was it really that accurate ?

Again, no disrespect intended, but imo a bit too soon to be calling a winter forecast "very accurate"

What sort of tone is that ST79? The forecasts are made not using GFS, ECM etc...so what's your point?

BFTP

[Catacol]

Without disrespecting anyone concerning Rogers forecast last year, December had the most descriptive run-down. I'm only sceptical of that particular months forecast because it was released on November 15th 2010 when models were already starting to firm up a severe cold blast and around that very time BFTP was bigging it up as if he already new it was coming, when prior to the 15th 2010 the models weren't at all keen on any kind of cold snap... far from it.

In simple terms he went for Cold from MID December, a mild January and another arctic blast in February with periods of heavy snow... How it panned out was Severe cold at the end of November into December, warmed up for January and remained mildish and dry through Jan and February.

So when you think of that forecast as a whole, was it really that accurate ?

Again, no disrespect intended, but imo a bit too soon to be calling a winter forecast "very accurate"

I dont think this gentle criticism stacks up. Are you suggesting that FI model runs in mid november that had started to show a northerly plunge was the basis for this forecast last year? The deep cold of 4 weeks later would have been a majorly risky call to make based on model runs that at FI range are almost always wrong.

No - it was done using Roger's methods that I dont understand but am increasingly wishing that I did.

Roger: as a long time lurker and reasonably able researcher and computer user, if you ever want someone to assist or work with you to help improve things then I would be very willing. I would need to start reading up on your methods from scratch, but I am impressed enough by my gut-feeling accuracy of your forecasts that I would like to know more.

[cooling climate]

Without disrespecting anyone concerning Rogers forecast last year, December had the most descriptive run-down. I'm only sceptical of that particular months forecast because it was released on November 15th 2010 when models were already starting to firm up a severe cold blast and around that very time BFTP was bigging it up as if he already new it was coming, when prior to the 15th 2010 the models weren't at all keen on any kind of cold snap... far from it.

In simple terms he went for Cold from MID December, a mild January and another arctic blast in February with periods of heavy snow... How it panned out was Severe cold at the end of November into December, warmed up for January and remained mildish and dry through Jan and February.

So when you think of that forecast as a whole, was it really that accurate ?

Again, no disrespect intended, but imo a bit too soon to be calling a winter forecast "very accurate"

Agree last years winter forcast was far from accurate apart from December which a lot of

forcasters predicted once the stratosphere models showed a warming end of October into

November. The rest of the forcast was way off.

As far as forcasting lows and storms on certain dates is almost certainly more luck than

judgement unless we are in a zonal flow which we are now. This is when low pressures come

along one after the other so to say we will have a storm in two or three days say will be a

fairly accurate guess. As far as forcasting the weather by phases of the Moon etc is imo

total poppy cock.

[Rollo]

Whilst I do not profess to understand all Roger writes he does have more than his share of successes and to describe his method as poppy cock does yourself as well as Roger a great disservice. There are no doubt many ways to forecast the weather other than the traditional ones and as time goes by forecasting will get better probably by a combination of many of todays unproven methods.