Jump to content
Snow?
Local
Radar
Cold?
IGNORED

Long Range Forecasting - Fact Or Fiction


Paul

Recommended Posts

Posted
  • Location: Aviemore
  • Location: Aviemore

Welcome to the first weekly talking point, this new feature will be a regular Sunday night addition with a new weather or climate related topic to discuss every week. Some week's these will be regarding something topical, so the week's events may dictate the subject of the discussion, but on other occasions the subjects will be wide ranging and hopefully suggested by members of the community. So if you have a subject you think would make a good weekly talking point, please drop me or any of the other team members a pm.

Weekly Talking Point #1: Long Range Forecasting - Fact or Fiction?

With Winter closing in and a lot of people's thoughts turning to the long range outlook for it, we thought this would be a good discussion to get the ball rolling..

With so many differing techniques and forecasts available from various forecasters, do you think there's science behind long range forecasts, or is it purely guesswork?

Do the often sensationalist and inaccurate stories in the press regarding lrf's damage the reputation of long range forecasting as a whole?

Does the level of detail in a long range forecast affect your view on it, for instance - do you think it's possible for a forecast at a range of 1-2 months or more to be specific about particular days or weeks?

If you think there is some science behind long range forecasting, do you think there has been progression in recent years, and if so do you think that in the coming years there's a possibility that lrf's will become increasingly more accurate?

Everyone is welcome to get involved in this discussion, so please do get stuck in - the items above are purely to get the ball rolling so please don't feel you have to comment on or answer them. The only requirement is that your posts are based on the original question/topic.

Link to comment
Share on other sites

Posted
  • Location: Crowborough, East Sussex 180mASL
  • Location: Crowborough, East Sussex 180mASL

Now there's a good question.

NWP range is limited by so many sources of error: currently available computational grid resolution, interpolation, extrapolation (and interpretation) of data from monitoring stations, mesoscale topology, ocean dynamics, insolation coupling, cloud modelling.

Then there is chaos theory and the unpredictablility of atomic decay.

Teleconnections, oscillation indices, torques etc. all offer a bigger picture but only to the extent of balancing probabilities.

Because these techniques are rooted in statistics, there will never be an accurate and repeatable way of coupling the outputs of these emerging lr techniques with that of short range NWP. We're left with interpretation limited by experience and a good deal of random chance.

Don't get me wrong, lrp in the mould of Galacier Point, Chiono' and a few others will statistically give a better chance of accuracy tipped on the positive side of 50:50 (say 70:30 if I'm being generous). It will never be entirely accurate, hence will always be accompanied by a healthy dose of caveat emptor.

ffO.

Link to comment
Share on other sites

Posted
  • Location: .Hackenthorpe south east Sheffield
  • Weather Preferences: winter.
  • Location: .Hackenthorpe south east Sheffield

long range has improved over the years look at the last two trom net weather can only get better

Link to comment
Share on other sites

Posted
  • Location: Rossland BC Canada
  • Location: Rossland BC Canada

Long range forecasting in general is improving. While the Met Office may have had a bad experience with it, on a more global scale many people are finding some success with a variety of methods. I think that on the U.S. weather forum that resembles Net-weather (Eastern U.S.) you wouldn't even find a thread like this because people have some broad general belief in the consensus of long range forecasts. Now one might debate whether that's because North American weather is for some reason easier to predict in monthly or seasonal terms, or whether there are better techniques being used, or some combination of the two.

Despite this difference in perspective, I think long-range forecasting has improved for the U.K. and even right here on Net-weather, and I would include the official Net-weather forecasts in that assessment but not limited to that one source. Fred (BFTP) and I have collaborated on the past two winters and at the very least we feel that our efforts were better than random chance. Some people had much better comments than that, but I would just leave it at "some progress has been made" as our own assessment. We'll have another one in due course and see what the reception to that is. Some people are never going to accept progress in this field, especially if it comes from anywhere but the dead centre of the met establishment. Decades of conditioning have rendered some peoples' minds incapable of accepting that innovation can come from anywhere but that central source using the big and expensive computers (because after all, if it did, that would just make the big and expensive computers even more of an issue).

Other people at the other end of the spectrum make claims that their methods are infallible, near 100% accurate and go out with astoundingly dramatic forecasts of events that almost lie outside known climatology. These people speak of monthly averages like -7 C (I seem to recall this from Jan 2008) which didn't even happen in 1740, then say nothing when the average is actually +6 C. Of course this does nobody in the field any good. Fred and I have been trying to contain our enthusiasm for cold in the past two winters and we gave out some numbers that were not very sensational and which generally verified but were a bit conservative for last January. But I feel that we were definitely improving on purely random chance.

I think that the question will go on being unresolved and recycled every year in the same way with the same players taking on the same roles, until such time as a more senior authority than anyone present here takes it upon themselves to make a Pronouncement followed by the usual gradual shift from mavericks vs establishment to new science vs a few holdouts. This is rather similar to what has transpired in the AGW debate since 2005 although my assessment there is that we are about halfway through the paradigm shift and not necessarily locked into a final outcome. But it's clear that AGW has lost ground to the skeptics in general terms. Of course, the fact that most establishment scientists are in the AGW lobby, and most long range forecasters tend to be AGW skeptics even if they are meteorologists (this is especially true in the USA) just blends the two debates into one bigger debate where the common theme is a different appreciation of the scale and causes of natural variability.

There would not likely be one "smoking gun" event that would suddenly make long-range forecasting respectable, but in North America, the media have access to all sorts of credible sources and companies are engaged in selling such forecasts commercially from well known and credible sources (not entirely outside the mainstream or what you might call one-track theorists). So here, the public mind-set is different, it's not so divided into people who will buy a tabloid, get a forecast and live or die with it, it's more like a mainstream media coverage of a consensus position that usually has at least some resemblance to the reality to follow. For example, the hot summer just ended in many parts of the U.S. was fairly widely predicted. Last winter, although it's fair to say the heavy snow exceeded any forecasts, the trend of all long range forecasts collected in one study was very similar to what happened, only not as extreme (the Mid-Atlantic states were forecast to have the largest anomalies for snow). Now different people used different techniques to get these forecasts but with a few exceptions this is what dozens of people in the field were saying. This winter they are generally saying mild and dry becoming cold and near normal for snowfall as the (eastern) winter goes on. There are many forecasts out for colder and snowier weather than normal in the west, at least away from the coast. This is largely based on "la Nina" climatology rather than any spectacular theories or in-house black box models. Oddly enough, this is exactly the scenario that my own research method is showing although the la Nina has no input into that (I would guess it's a related effect of some overall cause).

As to Europe, it's fair to say that the state of the art is either behind, or perceived to be behind. I'm not really sure which it is, but feel it's more likely to be perception over actual fact. If you factor out the usual suspects who one could predict with 100% certainty will express doubt about long-range forecasting, and polled the members at large here, you'd get a picture of what people are thinking about this subject. I would hazard the guess that there has been a slight shift towards acceptance just as there has been a shift towards skepticism about AGW. But this is certainly a science in its early stages. Nobody with possibly one or two exceptions would claim that we are anywhere closer than halfway to the goal at this point in time. I would expect there might be a more reliable and widely used long-range forecasting available and accepted by about 2018 to 2025. It won't suddenly appear next year or in 2012. But I doubt that confidence in long-range forecasting will do anything but grow gradually from now on. And the official Net-weather forecast might be the most significant element in that gradual change, if the company (as opposed to the forum) continues to have faith in the right people and keeps an eye open to the consensus of the right and the wrong people at the same time.

We'll see -- obviously I can't prove these conjectures. This is my honest opinion about it. But we never know for sure what's around the corner, there could be a breakthrough that moves us from the kind of statistical or tele-connection approaches that are used today, to equations or deterministic methods that exceed anything we know about, so that wild card exists too, and given the number of people studying the problem with high speed computers available to them, it would seem almost curmudgeonly to suggest that nobody among all of them would figure out exactly what's going on, and how to forecast it. Bear in mind, we don't really know why the weather does what it does even on the short time scale, but we know what it's going to do within very small error margins thanks to the ability of computer models. It wasn't that long ago that the 72 hour forecast was received with the same skepticism as today's monthly range forecast might be. But unless certain theories are accepted, I don't think anyone in conventional meteorology can tell you why there's a deep low south of Iceland today and why there wasn't one there on this date in some other year.

Edited by Roger J Smith
Link to comment
Share on other sites

Posted
  • Location: Stanwell(south side of Heathrow Ap)
  • Weather Preferences: Thunderstorms, squally fronts, snow, frost, very mild if no snow or frost
  • Location: Stanwell(south side of Heathrow Ap)

Good post (above post)rjsmith.

I think that being in between the Atlantic and North sea, makes lrf much more difficult than places like spain and canada(as example), Our weather changes so frequently. I feel that forecasting 3months out is possible to get a general idea, for example..wetter milder..cold and dry, but complications can pop up and change the forecast, i would say that lrf-ing 2/3months ahead is best doing first and second half of the month. I think computers would soon be more advanced to foresee the complications that can arise and change the forecast, when a seasonal forecast goes out and the news/media gets it they print it, then as can happen within weeks something changes and the forecast goes wrong, i think our weather is getting more complicated but the forecasters are keeping up with it.

Edited by nimbilus
Link to comment
Share on other sites

Posted
  • Location: Isle of Lewis
  • Weather Preferences: Sun in summer, snow in winter, wind in Autumn and rainbows in the spring!
  • Location: Isle of Lewis

Id say it is improving all the time, gien that it has to contend with quite complex (ok understatement) very complex feedback mechanisms, such as oceanic temperatures, albedo, salt water concentrations, cloud cover and so forth. I feel it does better when Europe is under the influce of igh Pressure and the NAO /AO is negative. Maybe higher pressure is easier to predict?

I also think the problem isnt with the models but with the jo public who either ridicule them for being out because they will only remember what they want to remember! Or they interpret what they want to see. Or how one big event, though not significant in terms of the overall long range forecast figures can raise heighten peoples perception that this scenario was the norm the whole winter. Or for not understanding the purpose of such models and what they actually show. Net weather predicted a colder than average winter last year, I think also drier as well.... it was spot on. The model predicted a cold end to winter the year before.. it was spot on there too.

This winter the CFS, so far, has gone for a cooler than average winter over the UK, and to be frank, given the atlantic has failer to deliver the killer punch I thought that it might and with the scandi high being quite a significant area of high pressure... I think it has it right again.

Link to comment
Share on other sites

Posted
  • Location: Napton on the Hill Warwickshire 500ft
  • Weather Preferences: Snow and heatwave
  • Location: Napton on the Hill Warwickshire 500ft

With so many differing techniques and forecasts available from various forecasters, do you think there's science behind long range forecasts, or is it purely guesswork?

There is of course a science behind long range forecasting and it would be disrespectful to suggest those have abundant knowledge in their relevant fields are just putting forward ‘guess work’.

I have no empirical evidence but I would suggest the weekly forecast seems to have improved in the last 30 years.

I’m sure there has been some improvement in monthly forecast but beyond that?.

It’s not ‘guess work’ but clearly it beyond anyone at the moment to predict how December will turn out.

Link to comment
Share on other sites

Posted
  • Location: Wexford, Ireland. 80 metres asl
  • Location: Wexford, Ireland. 80 metres asl

I think the only way a long range forecast or any weather forecast for that matter is a computer model that takes into account every single variable possible, no matter how small, of course it may be thousands of years before we know all of these variables.

LRF's will get better gradually with time. if you look where the internet was 20 years ago and where it is now, theres no reason long range models can't go through the same transformation.

Link to comment
Share on other sites

Posted
  • Location: Bangor, Northern Ireland (20m asl, near coast)
  • Weather Preferences: Any weather will do.
  • Location: Bangor, Northern Ireland (20m asl, near coast)

Long range forecasting has improved over the years, particularly for the UK as our climate is very changeable.

It seems we can now determine certain spells of weather 2 months in advance, ie Unsettled, Settled, Warm, Cold.

However the devil is in the detail so to speak as that is what really is lacking and for a very good reason.

Predicting a single days weather even over a week away will not work unless the scenario is fitting (ie, HP overhead), because systems over that period of time can slow, speed up, move in a different direction, etc, so when the actual day comes that frontal system we were predicting could still be 100 miles off the coast of Ireland, or it could be 200 miles further East. Little things like that make the difference to the detail.

If there was a method of feeding our computers the data to suggest a slow/speedup of fronts etc....maybe that could all change?:pardon:

For example, I gave my uncle a forecast for his bike ride down to the SW of Ireland on Friday and I gave him the forecast Friday - Sunday.

I used meto fax charts to determine any frontal rain event, netweather surface wind maps, etc. I gave him this forecast on Thursday.

Now, Friday I got spot on, rain in the morning in Belfast, followed by showers as he headed SW but some good sunny spells, a little breezy and temperatures around average. Saturday I also got spot on. But Sunday, when I looked at meto faxes on the Wednesday night Sunday had a very wet and windy theme about it, but when Sunday finally came, it was dry as a bone all day apart from a little rain am.

I looked at radar and lone behold, the front had been shifted further East quicker than expected.....

Link to comment
Share on other sites

Posted
  • Location: Redhill, Surrey
  • Weather Preferences: Southerly tracking LPs, heavy snow. Also 25c and calm
  • Location: Redhill, Surrey

It is still a process in learning but I would suggest that there is much more than luck involved. Looking at the various methods used and the interesting similar outcomes forecast suggests that there is a method to the madness.

Take last 2 winters for instance, no signal pointed to the prospect of a massive PV over Iceland and constant 2 thousand mile SW'lies for the UK so IMO there was no point in forecasting that. I think we are capable now of reading such general patterns some distance ahead.

Take GP for example, the composites he posts are anomalous responses to teleconnection phases and it has to be said the general atmospheric pattern Stew forcasts is followed and he gets many more hits than misses [the misses are usually only detail too not overall set up]. That is not guesswork and my hats off to him.

I bought into Roger's theory of planetary magnetic field influences on climatic pressure patterns a few years ago because I believed and believe it is real. At the time I used lunar cycles only but collaboration with Roger and adding in Solar phasing I think we have progressed with our LRFs. Like GPs teleconnections there isn't just one driver, some are stronger but others 'kink the pattern' shall I say and I think this is why we are at the stage of calling the overall pattern and getting some very good hits when we pick up on the 'kinking' signals, but we tend to be needing to be much nearer the time to read/pick up on them hence my view that we are still a process in learning but an improving one.

Why have the MetO been more unsuccessful of late?...another time maybe

Many know that some of us at this early stage are for sure touting a cold start to winter with HP more likely to our NW. I hope and anticipate that Roger and I will collaborate again this autumn for our witner LRF. I have a feeling that we are seeing the same start.

BFTP

Edited by BLAST FROM THE PAST
Link to comment
Share on other sites

Posted
  • Location: Bangor, Northern Ireland (20m asl, near coast)
  • Weather Preferences: Any weather will do.
  • Location: Bangor, Northern Ireland (20m asl, near coast)

I think it's safe to say unless the forecaster has sat twiddling their fingers then a forecast is NOT Guesswork.

When they put the effort in and actually backed up their forecast with data then wether or not the forecast happens it is definitely not guesswork. So I'd go as far to say that most eligible long range forecasts can't be guesswork.

Link to comment
Share on other sites

Posted
  • Location: Napton on the Hill Warwickshire 500ft
  • Weather Preferences: Snow and heatwave
  • Location: Napton on the Hill Warwickshire 500ft

I think it's safe to say unless the forecaster has sat twiddling their fingers then a forecast is NOT Guesswork.

When they put the effort in and actually backed up their forecast with data then wether or not the forecast happens it is definitely not guesswork. So I'd go as far to say that most eligible long range forecasts can't be guesswork.

I would agree the forecast isn’t ‘guesswork’ but the long term outcomes of that forecast often ‘suggest’ guess work.

Subtle difference.

Link to comment
Share on other sites

Posted
  • Location: Sheffield South Yorkshire 160M Powering the Sheffield Shield
  • Weather Preferences: Any Extreme
  • Location: Sheffield South Yorkshire 160M Powering the Sheffield Shield

Despite all the slagging off the Met Office got they didn't do a bad job overall. Much better than other companies.

The Barbie summer was the death-knell and showed a complete lack of understanding on how the media report stories and love to misquote.

The other thing I notice is that people tend to read what they want to read and manage to come up with a completely different forecast at times yet swear blind by it. Even some people doing their own forecast manage to change it by moving dates to fit the actually weather or get it wrong yet still say they're right.

I tend use CFS for monthly forecasts sometimes it goes wrong other times it's pretty good. Have a go at it make a note of what you've forecast and see how you're doing.

Link to comment
Share on other sites

Posted
  • Location: Beccles, Suffolk.
  • Weather Preferences: Thunder, snow, heat, sunshine...
  • Location: Beccles, Suffolk.

I tend to err on the side of fiction, myself. I do however have the belief that it'll be computers that'll one-day crack the thing. Although it may need quantum computers to handle the required huge amount of data???

Then there's Piers Corbyn and PWS et al who so love to allow themsleves retrospective 'windows' in which to shoe-horn whatever actually happened, afterwards...I guess I'm a tad cynical??

But none of the above can detract from the invaluable indications provided by GP, chiono and JH et al!

Link to comment
Share on other sites

Posted
  • Location: Hayward’s Heath - home, Brighton/East Grinstead - work.
  • Weather Preferences: Snow and storms
  • Location: Hayward’s Heath - home, Brighton/East Grinstead - work.

A good question Paul, to start the weekly debates.

I am in the camp where I think it is possible to give a long range forecast and they are not a waste of time.

I tend to break down forecasting into a series of time related periods, namely 0-3 days, 3 days to 2 weeks, 2-6 weeks, and 6 weeks plus. The area that I tend to concentrate on is around 2-6 weeks - more an intermediate range.

Stewart (GP) has consistently shown that whereas it may be impossible to tie down particular events to particular time, it is possible using the various teleconnections available, to point out the patterns that are likely to occur within this timeframe and the reasons why these patterns occur. A good example is currently on offer with the present modeling. Stewart has picked out long before any models, the likelihood of retrogression of the Scandi High, once the increase in relative AAM had subsided (around 60ºN) and the background La Nina conditions would prevail. No guesswork there!

I do not believe in forecasting methods that suggest a particular event to take place months in advance (eg a storm at a particular location for the 20th Jan 2011), however it would be quite reasonable to suggest that the jet stream may be enhanced/ reduced for a period of time due to well explained reasonable factors (eg stratospheric conditions).

All in all, long range forecasting and analogue pattern matching using teleconnective factors is a very useful way in being able to forecast certain scenarios for an intermediate period and patterns for those months further out. It is an area where a lot more research and understanding is needed and one that I think will improve significantly in the future.

c

Link to comment
Share on other sites

Posted
  • Location: Evesham, Worcs, Albion
  • Location: Evesham, Worcs, Albion

Quite a topical subject this - did Paul know the media were going to be running with PWS's winter froecast in the papers today? :doh:

Anyway, came across a very good piece by Abhijeet Ahluwalia today in which he comments that

The Met Office has chosen to give up making long-term predictions, not least because it is routinely hounded by tabloid newspapers that struggle to grasp the concept of statistical probability as opposed to iron-clad certainty

Also adding that

Michael Lawrence, a Met Office forecaster, questioned the veracity of the PWS claims, saying: "What these forecasters do is pit themselves in opposition to what we say and if they get it right they get a lot of publicity."

Met Office rival predicts another snowy winter, but can they be trusted?

Link to comment
Share on other sites

Posted
  • Location: Dorset
  • Location: Dorset

Long range forecasting is and always will be a game of luck.

The skill in forecasting is really to minimise the amount of luck you need and then play the percentage game.

Some are better than others at minimising this, such as GP and to be fair also the METO I think.

But at the end of the day all long range forecasts are a mixture of luck and knowledge, the more knowledge the better but none of them IMO can reduce the luck component below 50% for the UK.

Link to comment
Share on other sites

Posted
  • Location: Redhill, Surrey
  • Weather Preferences: Southerly tracking LPs, heavy snow. Also 25c and calm
  • Location: Redhill, Surrey

I do not believe in forecasting methods that suggest a particular event to take place months in advance (eg a storm at a particular location for the 20th Jan 2011), however it would be quite reasonable to suggest that the jet stream may be enhanced/ reduced for a period of time due to well explained reasonable factors (eg stratospheric conditions).

/quote]

I personally believe that it is possible. It is IMO the sheer fact on how the jetstream reacts to forcings that allows such 'events' occurring and being 'predicted'.

Just my view and as always I will chuck date events into my forecast if they are likely to be of significance, just like the snow events in Jan this year.

Its exciting to prepare one and also to read others. I know whose is my favourite [no not RJS and mine :doh: ] but I won't reveal that....not yet anyway.

BFTP

Link to comment
Share on other sites

Posted
  • Location: Napton on the Hill Warwickshire 500ft
  • Weather Preferences: Snow and heatwave
  • Location: Napton on the Hill Warwickshire 500ft

Long range forecasting is and always will be a game of luck.

The skill in forecasting is really to minimise the amount of luck you need and then play the percentage game.

Some are better than others at minimising this, such as GP and to be fair also the METO I think.

But at the end of the day all long range forecasts are a mixture of luck and knowledge, the more knowledge the better but none of them IMO can reduce the luck component below 50% for the UK.

Do you think in the next 20/50 years it will get any better ?.

I personally wouldn’t like a 95% certainty this winter will be mild/cold etc.

Well I could live with the cold bit.

In world dominate by predictability the weather gives a sense of our own limitations

I read somewhere about ‘fragmentation’ as the way forward, but cant see too find much on that subject on the net.

Link to comment
Share on other sites

Posted
  • Location: Aldborough, North Norfolk
  • Location: Aldborough, North Norfolk

Do you think in the next 20/50 years it will get any better ?.

I personally wouldn’t like a 95% certainty this winter will be mild/cold etc.

Well I could live with the cold bit.

In world dominate by predictability the weather gives a sense of our own limitations

I read somewhere about ‘fragmentation’ as the way forward, but cant see too find much on that subject on the net.

Me? I do think that the quality of prediction will increase, but I don't believe they will ever be able to forecast what will happen in 3 months, way too many variables that there is no control over. not forgetting things like teleconnections that complicate it even further.

So currently not much more than educated guesswork and remodelling old patterns, in the future, computers will help, but will still take an expert in teh field to make a judgement call

Link to comment
Share on other sites

Posted
  • Location: Rossland BC Canada
  • Location: Rossland BC Canada

Having devoted much of a lifetime to this general question, I have some rather different views on what is or is not possible. We need to be very clear on the paradigms involved here. I find that many, even experienced weather people, are not very clear on the paradigms (because we spend 99% of our time in the short-range paradigm where everyone is used to the playing field). Long range weather forecasting is not going to develop as a sort of extension of short range techniques out into an ever-growing future time frame, although this may work to improve accuracy from day four to about day fifteen. Beyond that, you need theories to develop long-range forecasting. The energy of current weather systems in almost all cases is completely dissipated within ten days. A few exceptions may be noted, but if the intellectual challenge is to predict the weather three months from today, then one would need a working theory of what caused weather events to happen in the first place. There could never be a computer program so complex that it would take today's input and maintain it accurately in 12-hour updates for 90 days. Don't even think about it. That would be a huge waste of time, money and resources to try to do that.

The whole point of our research (Fred and myself and some others out there too) is to find external energy sources that produce weather events. That would need to be studied from the two somewhat independent points of view of general circulation (the set up of ridges and troughs especially relative to normal values) and the timing of events. If my objective was to give a forecast for Christmas Day, for example, a large part of the guesswork implied would be reduced if I had an accurate idea what the pattern would be on that date. But to get the details right, I would need to know what energy regime was superimposed on that pattern. If I suspected easterly flow around strong blocking over Scandinavia, then would there be a Channel low, or a more distant but significant low, or just a dry flow devoid of energy, or perhaps a retrograde energy centre coming across the North Sea? If the research theory yielded that sort of accuracy in event and timing, and you had the right pattern overall, then it would be not much different from reading off the 48h GFS to make a long range forecast for that time and place.

Now I don't think that our method, or any other method, has reached a reliable stage in doing that, but I do think it is now within reach. Certainly it would need to be computer-generated to work optimally and yet you can see that it could be done without computers, just as Isaac Newton solved the riddle of gravitation without a computer. The difficulty of this challenge in purely scientific terms is not really much different from what he faced, the difference is that we are dealing with a much more complex system than one moon circling one planet and the need to find an equation of motion to describe that. The nature of the problem is such that any one worker will likely have to confine his research to one or two data sets for one or two locations, as you can see from my own case, enormous amounts of time and effort are required to make any progress at all, then you have the practical question of interacting with different groups of people because there is not much point in a person generating progress in this field somewhere off in a cave or isolated setting, then getting old, dying and nobody knowing about it. But that does seem to be the weather establishment's plan for my life and the lives of any other people who dare to shake the tree of complacency that says, just wait fifty years, and computers will figure this out.

Once theories of cause and effect are accepted and refined, then computers will be invaluable. But the idea that a computer model will somehow take the initial input (time zero) and extend out beyond the current FI period are hopelessly unconnected to atmospheric realities. The very slow progress of accurate 10-16 day results in the GFS should be the tip off there. It's nobody's fault, the atmosphere simply doesn't retain energy that long, the cycles of energy for storms in particular are typically 5-10 days. Now somebody (actually it may have been a former prof of mine, Dr Ken Hare, a U.K. emigrant to Canada who taught at the Univ. of Toronto back around 1970, if not him, then he was quoting the actual author) once famously said that a hurricane might be the outcome of a butterfly flapping its wings in Africa. Actually, it's more likely the result of Mercury moving through a planetary field sector of an outer planet and the earth getting in the way. Now that energy event is about on the same scale in absolute terms as received here, but it makes the point, if something that low-energy can initiate a hurricane, now how can a computer model simulate this 90-120 days in advance? How would a computer model know when a butterfly might decide to flap its wings in Africa? Or when a packet of energy might form over the northern Gulf of Mexico to create the genesis of a winter storm? Now if those butterfly wings are actually subtle energy inputs from predictable astronomical events or geomagnetic variations, you've got something. If not, then all the effort in the world will fail to solve this riddle, it just won't be possible to model such things accurately.

Yet, if the large scale features are based on the energy set-up of external drivers, then it stands to reason that the variations within the large scale features would become predictable as second-order variables.

Now a reality check is always good and we should understand that progress in this field is not on the same time scale as short-range forecasting. In some early stages, we should be pretty pleased if long-range forecasts can start to identify time scales within a few days (as with the onset of last winter's cold spell), and we should understand that this constant objection about "shifting days around" is really a misread of the different time scales from the fixed perspective of people used to short range forecasting. Of course, it's a big error if the storm expected on Sunday to arrive on a Tuesday night arrives on Thursday morning. But if one predicts a major event for 27 January and it arrives on 29 January, that may actually be a very good long-range hit. The verification would be complicated by how likely it is at random to occur on any date. If I were to say "a garden variety very average ordinary Atlantic front will cross Ireland on November 4th" and one crossed Ireland on Nov 5th that would prove very little given that such fronts cross Ireland perhaps every 2-3 days in certain weather patterns. But the same error for a major London snowstorm would be much less problematic, although still something that would qualify for further study as to the causes of the timing error. See what I mean here? The time scales for long-range forecasting are not those of short-range forecasting. But at some point if you allow too much leeway then you are allowing random chance to provide hits instead of theory validation. So that makes assessment of the more complex forms of long-range forecasting difficult. I loathe all anecdotal discussions of them because I know that some people will never accept any forecast success no matter how obvious to other workers in this field, whereas others will claim success on really dubious things like a snowstorm somewhere else on earth on their predicted date, or that sort of thing. But I feel that people can reach a consensus about what is a good long range forecast. Some of the politics will have to be swept away first (the kind that allow people to say or at least think, over my dead body will that guy ever get any credit, because that's actually more than half the problem at this point in time).

Final point, I think the Met Office made their own bed by constantly shifting their forecasts towards the warm end of the spectrum to conform with AGW orthodoxy, and when this began to backfire, they tried to offload blame to tabloid newspapers who had supposedly "misunderstood the range of probabilities." This is somewhat disingenous. There may be some truth to it, but what's more apparent is that they misunderstood the range of outcomes. They were unable to visualize a cold and snowy winter as being possible, and took some rather faint signs of warmth in 2009 to trump up to the barbeque summer. Now to be fair, the summer was close to being just that in the southeast. But I think it illustrated that they had a warm bias from 2007 on, which made earlier successes look like they may have been more to do with the bias than the methodology. They should have outsourced to Netweather in 2008, then they would be looking pretty good about now.

Link to comment
Share on other sites

Posted
  • Location: Redhill, Surrey
  • Weather Preferences: Southerly tracking LPs, heavy snow. Also 25c and calm
  • Location: Redhill, Surrey

Cooking my boy's tea and not able to prepare response as above like RJS so I'm mighty glad it has been done, but I really conform to that post 100%. Going to pour myself a cabernet sauvignon, 2003 France...a very good year!!

BFTP

Link to comment
Share on other sites

Posted
  • Location: Lincoln, Lincolnshire
  • Weather Preferences: Sunshine, convective precipitation, snow, thunderstorms, "episodic" months.
  • Location: Lincoln, Lincolnshire

I think there's a good deal of sense in what RJS posts above, I think the future of long-range forecasting definitely lies in the pinpointing of overall patterns and external drivers rather than, say, specific timings of specific events. This ties in also with Glacier Point and Chionomaniac using teleconnections and angular momentun frequently, which can give a general guide on the likelihood of different weather patterns recurring from weeks, even months, in advance.

My month-ahead forecasts were originally based largely on previous experiences of model outputs, typical synoptic progressions (in effect pattern matching on a "one week to the next" basis) and typical errors that the models make, where I then have to estimate what would be likely to happen instead. In essence it was all down to educated guesswork, some of it being knowledge but also a fair amount of "luck" (see Iceberg's prior post). In more recent times I've added the teleconnections analysis to the mix of variables used, e.g. the MJO strongly pointed towards an easterly type arising towards mid-October and it looks like we are, indeed, going to get one.

One major problem with long-range forecasting is that far too many sources either predict something that will make a lot of headlines, or predict what the forecaster wants to see happen (or both).

I don't think the Met Office actually did too badly overall; it was unfortunate that the "fails" occurred in a cluster in 2008-09, resulting in the previous successful forecasts (clustered around 2005-07) being forgotten by the media and the public. However, irrespective of this, I agree with their decision to focus on shorter-range LRFs and rely upon the "probability maps" for seasonal long-range stuff- for instance, the probability maps were a lot closer to getting last winter correct than the written forecast suggested, envisaging a colder-than-average winter in most of Eurasia but keeping the cold air locked too far east and leaving Britain in mild air.

Link to comment
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
×
×
  • Create New...