Matty may not talk sense about many things (  ) but he's bang on the money when it comes to LRFs. File under fiction.

---

Why would anyone recommend any seasonal forecast, let alone strongly recommend? That gives the impression there's an ounce of worth to any of them, which there isn't. All the technical presentation in the world changes nothing.

When I first visited TWO in the early noughties, the only thing I really read was the long range forecasts. I don't agree that they have no credibility; they just arn't going to be as accurate as short term ones.

Well you're wrong I'm afraid.

As I've said before - show me the link to the agency that produces regularly accurate seasonal forecasts and I'll hold my hands up. There isn't one. Even the professionals will tell you seasonal forecasting is currently impossible.

I would go as far as to say they are a complete load of bull and a waste of time if it weren't that nothing has ever been achieved by giving up. Of course it has to be continued to be explored, but to suggest there any one out there is better than another, let alone from an amateur site, is total and utter nonsense, and misleading nonsense at that. I have a real bugbear about this. It annoys the hell out of me when people confidently claim something that is totally wrong.

I propose an experiment then. You can probably find a web archive of Brians TWO seasonal forecats. He predicts precipatation and temperature CETs each quater in five bands. Get say a few years worth of his forecasts and compare to the actual CETs. Of course all he has to do is show his forecasts are significantly better than what you would expect by chance. I would then assume temperature/precip is distributed normally as an approximation and then take his predictions and compare the standard deviation (although you may have to transform from a binomial which is discrete) with the mean of brians error. If the later is significantly smaller than the former, then that would be evidence in his favour.

My plan was much easier. Show me a link to the agency that consistently produces accurate seasonal forecasts.

 

How can you say "Accurate" when conclusions on LRF tend to state "colder/Warmer than average" etc  Thats all anyone is looking for in accuracy when it comes to LRF.  What will the weather be like over the season?.  Therefor I think Quantum has a point.  All you need to do is proove that the averages panned out correctly.  should be a simple task really.

---

Why would anyone recommend any seasonal forecast, let alone strongly recommend? That gives the impression there's an ounce of worth to any of them, which there isn't. All the technical presentation in the world changes nothing.

When I first visited TWO in the early noughties, the only thing I really read was the long range forecasts. I don't agree that they have no credibility; they just arn't going to be as accurate as short term ones.

Well you're wrong I'm afraid.

As I've said before - show me the link to the agency that produces regularly accurate seasonal forecasts and I'll hold my hands up. There isn't one. Even the professionals will tell you seasonal forecasting is currently impossible.

I would go as far as to say they are a complete load of bull and a waste of time if it weren't that nothing has ever been achieved by giving up. Of course it has to be continued to be explored, but to suggest there any one out there is better than another, let alone from an amateur site, is total and utter nonsense, and misleading nonsense at that. I have a real bugbear about this. It annoys the hell out of me when people confidently claim something that is totally wrong.

I propose an experiment then. You can probably find a web archive of Brians TWO seasonal forecats. He predicts precipatation and temperature CETs each quater in five bands. Get say a few years worth of his forecasts and compare to the actual CETs. Of course all he has to do is show his forecasts are significantly better than what you would expect by chance. I would then assume temperature/precip is distributed normally as an approximation and then take his predictions and compare the standard deviation (although you may have to transform from a binomial which is discrete) with the mean of brians error. If the later is significantly smaller than the former, then that would be evidence in his favour.

My plan was much easier. Show me a link to the agency that consistently produces accurate seasonal forecasts.

 

How can you say "Accurate" when conclusions on LRF tend to state "colder/Warmer than average" etc  Thats all anyone is looking for in accuracy when it comes to LRF.  What will the weather be like over the season?.  Therefor I think Quantum has a point.  All you need to do is proove that the averages panned out correctly.  should be a simple task really.

Proving that it is statistically significant might be more problematic, however, if you are looking at such broad descriptions. As an example, when a LRF says there is a 60% chance of a colder than average winter, and then there is, you need to make sure the "60% chance" element of the forecast isn't lost in your calculations of its accuracy. Working out whether there is a proper correlation would take many, many years of such figures.

---

This particular forecast should be easier to verify than most. The patterns and weather expected are all shown through NCEP/NCAR reanalysis - come the end of February then do the same exercise to compare the actual to the projected. Tells you nowt about the conditions in x,y,z location - cold can be through a variety of pathways in the winter season.

Irrespective of that, anybody who has spent the time to compile such a detailed analysis of the atmosphere is deserving of a thanks rather than scorn.

Proving that it is statistically significant might be more problematic, however, if you are looking at such broad descriptions. As an example, when a LRF says there is a 60% chance of a colder than average winter, and then there is, you need to make sure the "60% chance" element of the forecast isn't lost in your calculations of its accuracy. Working out whether there is a proper correlation would take many, many years of such figures.

IMO you don't need figures to prove that the science behind it is bunk. Even allowing for the idea that it might be possible to predict the probabilities of large-scale circulation patterns, trying to relate those to the temperature of the area we are interested in (the UK, presumably) is pretty futile.

Say you have blocking in the north Atlantic area. A shift of the block by 500 miles or so east or west - utterly insignificant on a global scale switches the UK weather from cold northerlies to mild southwesterlies. That can happen with forecasts 10 days out, so hoping for accuracy three months out is laughable.

---

Matty may not talk sense about many things (  ) but he's bang on the money when it comes to LRFs. File under fiction.

Won't stop me writing mine every year though Rob.

I do it for fun, but I do put some work into it (mainly researching past patterns).

That's all.

---

This particular forecast should be easier to verify than most. The patterns and weather expected are all shown through NCEP/NCAR reanalysis - come the end of February then do the same exercise to compare the actual to the projected. Tells you nowt about the conditions in x,y,z location - cold can be through a variety of pathways in the winter season.

Irrespective of that, anybody who has spent the time to compile such a detailed analysis of the atmosphere is deserving of a thanks rather than scorn.

 

agreed

 

The science for LRF has to start somewhere so we can discover what works... unfortunatley its a slow process, and probably would need a period of around 20 yrs to bed in some ideas/processes and to see the return in information.

 

We shouldnt knock anyone willing to do this level of research.

---

In order to verify wheather quantatative forecats are accurate, the statistics could get very nasty. I think I suggested in my above post using a binomial distribution as a normal approximation or something, but its a very crude calculation even so. Plus  you have to factor in the decreasing accuracy over the range of the forecast, the calculations could turn out extremely nasty. But you would have to do calculations of some description, otherwise you are basing accuracy on oppinion, which is subject to all kind of confounding factors like confirmation bias. To make it simplier, perhaps just take Brians 3rd month in his forecast. Divide his predictions into two variables that can take integar values from -3 to 3 ; each representing the error in his CET prediction in temp and precip according to his catogaries (much below, below, slightly below, avg, slightly above, above, much above). Then I suppose you would want to assume these variables are distributed normally, unfortunately being discrete they won't be - so you should approximate the binomial to a normal using a continuity correction. Then work out the standard deviation of the distribution. The standard deviation of the actual CET would be the same as the standard deviation of the error you would expect to obtain if you had simply predicted average every time (or the same as chance). If the standard deviation of the error is smaller, than that of the standard deviation of the CET then that would imply Brian is doing better than would be expected by chance. But then you would also have to test for statistical significance, which I think would be a standard deviation of a standard deviation of the sample.

My statistics are a little rusty, so I might have some of this wrong. But its the best method I can think of at the moment, the idea of actually performing this calculation gives me a headache though.   

---

Forget all this warmist rubbish, it will be a damn cold winter as the very quiet sun confirms .

Severe winters do appear to be more frequent during low activity cycles, but even during the Maunder Minimum [approx 1645-1715] not every winter was cold. Winter 1685/1686 had a CET of +6.33c which is the 6th warmest in the entire CET record.

I'm in Matty's corner too, I respect those that give long range forecasts but they really aint worth the time and effort imo especially when the same forecasters cannot forecast the weather for the end of next week let alone end of Winter.

Look how much chopping and changing and opinions there have been over the last few days in the MO thread. Those that give a view after every run more often than not change there view on a daily basis because of whats been thrown infront of them.

Let's get past t120 and a couple days after nailed on 90% time then look past that, till then it's really not worth it unless you wantt he 50/50 gamble of getting it wrong.

 

---

He's talking about North America, UK looks like cold December milder Jan/Feb. above average overall for us.

850 mb Korean Met A winter forecast. December upper right, Jan lower right, Feb upper left, total lower left. BRRR pic.twitter.com/HrcmcO75Qy

— Joe ******i (@BigJoe******i) November 28, 2013

---

Matty may not talk sense about many things (  ) but he's bang on the money when it comes to LRFs. File under fiction.

Won't stop me writing mine every year though Rob.

I do it for fun, but I do put some work into it (mainly researching past patterns).

That's all.

And a fine job it was , even though I binned it

---

I would like to thank Quantum for linking to my first ever winter forecast and for your comments.

The NW winter forecast is the first I have ever attempted and I now know how difficult it is to collect ones thoughts and research and put it all down on paper and realise how much hard work must go into a forecast and commend Brian on producing one every year.

 

However, I do take issue with those comments that consider attempting a long range forecast is a waste of time. I do not believe this to be the case. We may not be at a position just yet when the verification rates of longer ranges forecasts is particularly high, however, this shouldn't stop us trying to improve the success in this challenging field.

It may have not gone unnoticed to some of the more astute on here that when certain factors are present there is a correlation to the atmospheric conditions and the positioning of long wave troughs and ridges. This is no coincidence and even though the weather patterns may seem chaotic in fact there is indeed some order behind this chaos. A prime example of this can be demonstrated during the active phase of a strong La Nina in the NH winter. The La Nina conditions can be well forecast in advance and so with that can the atmospheric response wrt long wave trough and ridge positioning. These can be demonstrated through reanalysis of H500 anomalies. And this isn't limited to the ENSO state - the same can be shown on other factors such as the QBO, solar flux, SST's October snow cover rate and gain and so on and so forth. We are only at the beginning of the complex job of fitting all these jigsaw pieces together, but by writing this off without even trying we will never progress.

I suspect that those who believe that it is a waste of time to attempt to put the atmospheric jigsaw pieces into some kind of order are being a little short sighted - perhaps they would have been saying that a five day forecast would have been a waste of time 40  years ago, or still believe that the stratosphere has no influence on the underlying tropospheric pattern. Certainly the energy company who now pay my predecessor for his long range forecasts do not believe it to be waste of time....

I welcome those who try and increase the understanding of the world we live in and will not try to belittle their achievements whether they are successful or not.

And on that note I would like to wish Brian the best of luck on his forthcoming winter forecast and would like to acknowledge and appreciate the time and effort that he has put into, even before it is released. I hope that others here will have the good grace to do the same and if they can't then perhaps it is better not to comment at all.

c

I would like to thank Quantum for linking to my first ever winter forecast and for your comments.

The NW winter forecast is the first I have ever attempted and I now know how difficult it is to collect ones thoughts and research and put it all down on paper and realise how much hard work must go into a forecast and commend Brian on producing one every year.

 

However, I do take issue with those comments that consider attempting a long range forecast is a waste of time. I do not believe this to be the case. We may not be at a position just yet when the verification rates of longer ranges forecasts is particularly high, however, this shouldn't stop us trying to improve the success in this challenging field.

It may have not gone unnoticed to some of the more astute on here that when certain factors are present there is a correlation to the atmospheric conditions and the positioning of long wave troughs and ridges. This is no coincidence and even though the weather patterns may seem chaotic in fact there is indeed some order behind this chaos. A prime example of this can be demonstrated during the active phase of a strong La Nina in the NH winter. The La Nina conditions can be well forecast in advance and so with that can the atmospheric response wrt long wave trough and ridge positioning. These can be demonstrated through reanalysis of H500 anomalies. And this isn't limited to the ENSO state - the same can be shown on other factors such as the QBO, solar flux, SST's October snow cover rate and gain and so on and so forth. We are only at the beginning of the complex job of fitting all these jigsaw pieces together, but by writing this off without even trying we will never progress.

I suspect that those who believe that it is a waste of time to attempt to put the atmospheric jigsaw pieces into some kind of order are being a little short sighted - perhaps they would have been saying that a five day forecast would have been a waste of time 40  years ago, or still believe that the stratosphere has no influence on the underlying tropospheric pattern. Certainly the energy company who now pay my predecessor for his long range forecasts do not believe it to be waste of time....

I welcome those who try and increase the understanding of the world we live in and will not try to belittle their achievements whether they are successful or not.

And on that note I would like to wish Brian the best of luck on his forthcoming winter forecast and would like to acknowledge and appreciate the time and effort that he has put into, even before it is released. I hope that others here will have the good grace to do the same and if they can't then perhaps it is better not to comment at all.

c

 

Exactly.

 

just because we are only just beginning to understand LRF...doesnt mean they are pointless.  These things take time to write, asses, evaluate, reflect and fine tune.

---

 

 

However, I do take issue with those comments that consider attempting a long range forecast is a waste of time. I do not believe this to be the case. We may not be at a position just yet when the verification rates of longer ranges forecasts is particularly high, however, this shouldn't stop us trying to improve the success in this challenging field.

That's all very well but can you tell us what will happen at the end of next week regarding this cold spell on the horizon

Everyone has there own view, don't take it to heart

---

My problem is with long-range forecasts that describe conditions on the ground week-by-week or even month-by-month. To me that's just science fiction, i.e. there's indeed some background research, but after that it's purely the product of the writer's imagination.

But I do like research which investigates whether certain long-term and identifiable conditions, such as sea-surface temps or solar cycles, make out-of-ordinary weather more or less likely over the course of a season, without necessarily pinning down when precisely it might occur. After all, that's what we'd like to know in practical terms as individuals, planners or businesses.

---

Oh dear. A characteristically defensive response from the forecast creator, similar to anyone who dare question the science of stratospheric warming.

Anyway, I'm sure everyone on here understands long range forecast attempts aren't a waste of time per se; as mentioned on page one, nothing has ever been successfully achieved by giving up. The issue people have - some people - is that it's currently a million miles away from sustained accuracy that proves the theory. Therefore, any claims of one LRF being better than another are extremely speculative at best as things stand. It's nothing to do with being "astute" as you put it. Trying to pour scorn on people's perceived lack of intellegence is belittling and adds nothing to this weather site. It's a weapon you have used in the past re Strat warming. Anyone who doesn't jump on the bandwagon of praise is lambasted as thick, trolling, or otherwise.

On the basis of that I thank you for your effort and wish you luck with your forecast. Don't take criticism so personally and you'll do fine.


m.

---

I think long term forecasting is fraught with disadvantages if people are going to say "it will happen"

However the Met Office seem to put some confidence in their GloSea4 long range model, but they usually only give a hint of what "may" happen but things usually pan out worse or better when that time comes - but even they the professionals get things wrong, no one is perfect even if people give off a smug personality that seems otherwise, we all make mistakes and so do computers.

---

In addition to our esteemed leader Brian Gaze, Gav from gavs weather vids, and the metoffice. The other winter forecast that I would strongly recommend is from the other place . Being serious, though, I thought I would post the link here because the technical forecast is extremely detailed and the quality looks excellent. 

http://www.netweather.tv/index.cgi?action=winter-forecast-in-depth;sess=

They suggest a slightly colder/drier than average winter, with some blocking but 'not a repeat of 1947'!

 

One of the bester presented LRFs going IMO - i'm not suggesting it is any more accurate than anybody elses LRF, but I like the presentation of it and use of data to back it up in a nice digestible format.

 

 

---

Oh dear. A characteristically defensive response from the forecast creator, similar to anyone who dare question the science of stratospheric warming. Anyway, I'm sure everyone on here understands long range forecast attempts aren't a waste of time per se; as mentioned on page one, nothing has ever been successfully achieved by giving up. The issue people have - some people - is that it's currently a million miles away from sustained accuracy that proves the theory. Therefore, any claims of one LRF being better than another are extremely speculative at best as things stand. It's nothing to do with being "astute" as you put it. Trying to pour scorn on people's perceived lack of intellegence is belittling and adds nothing to this weather site. It's a weapon you have used in the past re Strat warming. Anyone who doesn't jump on the bandwagon of praise is lambasted as thick, trolling, or otherwise. On the basis of that I thank you for your effort and wish you luck with your forecast. Don't take criticism so personally and you'll do fine. m.

With all due respect MattyH I think you are being a little unnecessarily confrontational yourself. In your first reply to me you effectivly said long range forecasts were all but useless (which is odd considering that you are an admin of a forum that was most originally famous for its long range forecasts). As for your comment about comparision of long range forecasts, then again I disagree. There is evidence to suggest looking at certain factors like SSTs does influence the weather, so forecasters that base their forecasts on multiple quantifiable phenomena are likely to be more accurate than those that don't. To label high quality forecasts like Brians, Gavin P, netweather and the metoffice with certain others just doesn't seem right. 

---