Lol so Exacta weather don't rely on super computers, weather balloons  and on the ground surveillance then.

none whatsoever. No one can forecast what the weather is going to be like in two months time. It really is always that simple. 

---

Not true, the nino can be foretasted several months in advanced to a high degree of accuracy. For long range forecasts the accuracy is a function of latitude, Arctic regions are nigh impossible, whereas the equator is doable. 

For the UK its much less reliable ofc but LRFs do exist that are better than chance, in particular the metoffice which is a good example of how LRFs should be done, probabilistically. Specifics is ofc nonsense, but it is true that low solar activity is likely to make a decade with colder winters for example, and that one can say with reasonable confidence this year will be colder than last year simply because of regression to the mean.

 

---

Not true. There is no consistently accurate method for forecasting even trends for the UK at that range. 

---

Well yes in a sense no LRF is going to have anywhere near the accuracy of a short term forecast, that being said they are better than nothing which is what is being implied here.

 

---

 

Im im not convinced they are better than anyone's guess over all. Our climate here is so variable that even short range forecasts are often inaccurate. You've only to look at the short range models of the last 24 hours. They've gone from Atlantic storms, to southerly plumes and back again. There is no reliable method for this part of the world two months out. There just isn't. Any accuracy is tenuous at very best. 

Continental trends at that range may well have an improved accuracy down the sheer repetitiveness of seasons on a continent such as North America. 

---

Perhaps this might convince you, this is an example of a 3 month verification for the ECMWF seasonal model

This definitely looks better than a guess, notice the pattern that I mentioned, that the accuracy decreases with lattiude. Its virtually 100% accurate in the equatorial west pacific.

---

Steady on Q!

---

The CFS shows similar verification patterns, it seems to me that a professional long range forecaster from the met office would do better than the raw model output otherwise there would be little point in employing them. The evidence seems to infer that they are better than a guess, but are strictly probabilistic (e.g 40% cold, 30% avg, 30% mild). In the tropics they do much better.

 

---

It does convince me that what I wrote in the last post was more accurate than a season LRF for the UK. 

---

Matty, a guess would have an average correlation of exactly 0, now there will be some statistical uncertainty to consider as well so 0.1 is still 0 if the uncertainty is larger than 0.1, but statistically significant positive average correlations, no matter how small are better than guesses.

---

we can split hairs all day, but you can't argue that the accuracy reliability remains so slim it's negligible. Point is, someone asked how much credence we should give other agencies offering a forecast for two months time, and the answer remains none really. That's not writing them off as agencies - fair play to them for having a go, but they are still going to be inaccurate on a very regular basis

---

If I made a LRF saying there was a 30% chance of a cold winter 10 times and 3 out of those 10 times there was a cold winter, that's still a good forecast it just doesn't have anywhere near as much utility as a SRF.

 

---

Does anyone else only see 21 pages to this thread until they actually click on it?

---

. 

Percentages aren't that simple. All 10 could still be completely wrong and at some stage in the future you could get a couple right in a row. All I see there is a 70% failure rate. 

---

Well yes, ideally you'd have an infinite sample size, but that's a problem of reflecting on the quality of the forecast, not the quality of the forecast itself. LRFs have very low utility for the UK, I agree with that, but it doesn't mean that LRFs are just guesses either. The issue is the people that treat them like SRFs saying "the UK is about to see the coldest winter in a century" e.c.t. No! The UK is slightly more likely to see the coldest winter in a century than the average year would be a reasonable LRF. 

---

Agree with that, but seeing as we are talking statistically, any amateur forecaster with a bit of sense will always forecast an average UK winter of mainly mild with the odd cold snap. He or she will be accurate - even taking into account the vagueness of such a forecast - more often than not. This is still a guess though, albeit an educated one based on probability, so therefore his or her guess is more likely to be accurate on any given year than any global model which attempts to use current weather patterns to forecast a season. 

I have a headache now. Beer required. Well it is Friday. 

---

As far as LRFs go, predicting an average winter every-year is definitely a good way to go (although obviously completely useless), but I submit that we can do better than that (perhaps not enough to satisfy many) and a professional LRFer would be able to not only beat the roll of a continuous random variable dice but also beat the guy that always bets on the most favoured horse. Obviously no one is ever going to assign a high probability to a very unlikely scenario, but the metoffice is confident enough to go away from 50% cold 50% mild to perhaps 70% 30% and has evidence to back this up.   

---

Jeeze - that was an intense few pages wasnt it? Problem with statistics and odds etc is its purely down to ones interpretation. So two people can read the same data and see two different things. Plus - the forecast is never wrong. ever. Personally speaking I work in IT - From an IT perspective when you test software - the aim is to test out a system and find as many defects as possible. Perhaps this would be a better way to develop forecasting technology. As in - make it fail to understand why it fails. Rather than trying to find 1000 ways to make it work. Expectations are lower and its safer to review retrospectively rather than trying to find the future holy grail. Perhaps thats what OPI is moving towards.

What i believe is that quite often - upon reflection of a season (especially in winter) once its over. Its common to look back and think - ahh the pattern of the past winter had the signs there during the previous season. IE blocking, or storm, or lack of precip etc. As a consequence i am always looking at the current weather pattern and theme to try and pick out any common themes that might persist through the winter.

Currently the Theme i am seeing is a High pressure dominated pattern. rather than one of Low pressure dominated one. Last season we had a few early storms which were a bit odd compared to recent autums. The previous winter had a weak but stubborn HP to our east north east which led to slider lows, and marginal but heavy snow events later in winter. 2010 had a major snow event in november - we all know what happened next.

With this in mind given the current pattern I see we could either get very mild drizzly weather, or very sharp northerly /easterly blasts. But dry overall. With probably majority of any heavy snow/rain hitting the continent. Less chance of storms, as has been the case recently. Less chance of zonal weather, and days on end of strong winds from the west.

could be wrong i guess. But its alot more interesting that arguing % of this happening and % of that happening.

---

You turn up here aghast at the last few pages and then throw that into the mix?

Here, come and share my beer. It'll make everything feel better. 

---

already had beers - hence reply.

---

Update

Review of winter 2014/5 indicators.Updated.

There is little change to the indicators set out at the end of August.My conclusion reviewing the September data is that overall a warmer than average winter is indicated.

1.ENSO.

Latest diagnostic discussion from NOAA suggests a weak El Nino in range 0.5-0.9."Weak El Nino is expected to begin in 1 to 2 months time and last into the Spring".Overall El Nino's tend to be followed by warmer than average winters with the link getting stronger for the higher El Nino levels.Weak El Ninos return 50.50 for the following winter.

On balance I would say the ENSO signal points to a warmer than average winter.

2.QBO

Last year had a very long and very strong +ve QBO and was followed by a strong +veNAO and warm winter.However the record of the link  between QBO and winter CET is patchy.For what its worth the QBO is now in negative phase (-23) which points to below average winter temps.

3.SST's

I always find the pattern (May/June) difficult to interpret but  back in May there didn't seem to be the clear tripole which links to NAO-ve winters.Latest forecast for SST's  for DJF has negative anomalies in the central Atlantic to the west of the UK.What is this likely indicate in terms of pressure patterns? My own verdict is that the spring  SST's favour a warmer than average winter.

4.Sunspots

The solar cycle peaked in Feb 2014 at 102.6 and is now falling away slowly down to at 87.6 in September.The only real verifiable link to lower than average CET is around  the solar minimum(University of Reading studies).So little influence likely at this stage of the solar cycle.

5.Model predictions

Meto Glosea contingency probability indicators shows the following for the 3 month Dec,Jan,Feb.

Temperature.20% probability of below average temp 60-80% of above average.

Rainfall:Close to avaerage.

Pressure:Above normal 20% probability,below normal 40-60 % probability ie favouring a +ve NAO.

CFS currently shows average temperature and pressure over the UK

6.Seasonal predictors

Spring 2014 was a Very Warm wet season and summer is likely to finish overall very warm and dry Autumn is shaping up to be warmer than average.Both of these types tend to be followed by warmer than average winters.Autumn type to early to call.

OVERALL PREDICTION

A repeat  of last month ie winter 2014/15 is likely to be warmer than average.

Overall your logic is flawed as a weak nino is more conducive  of a cold blocked pattern when coupled with a negative QBO, the SST's are a waste of space hence why the MetO dumped them a few years ago having been embarrassed one time too many after utilising them for their LRF. As for the GLOSEA model this time last year it was showing HLB, so again this model should be discounted at this range. 

Overall prediction, haven't the foggiest as the weather will do just what it wants to do, as always.

An interesting few pages of posts here. In reality an amateur forecaster who goes for mild with a few odd ( northerly toppler)colder snaps is not making a forecast , merely describing the weather of most of the winters from 1988-2008. The crux of winter forecasting is being able to spot the trends which will predict the likelhood of extreme winters at either end of the scale ie mild and stormy like last year or the coldest/snowiest for 30years like 2009/10. It is only when we find a tool or tools that can do this that winter forecasting becomes really relevant.

I admire any science based efforts to try and help in this regard