-2.08 starting the last day of October, not bad at all and should be good enough to test some theories.

Currently off the page being finalised, but last figure I saw was I think -2.18 based on the 12z yesterday.

---

Yeah, say's they are making a final revision to OPI and a forecast will be issued soon;

http://app.til.it/opi/

Will be very exciting waiting for this!

---

I expect that an aberrant minus sign had crept into the calculations and they are just taking that off!

---

With negative values we've seen this October then I tend to agree with that, the correlation with  values under  -1.5 are startling and this is by far the best long range forecasting tool for the winter ahead. Beats the pants off  GLOSEA and all the other junk models, even with the limited datasets.

  

Will it be available via the normal channels such as Twatter and Tundra or shall we expect a forced download to our iphones?

---

The padlock symbol and Pin number space on that page look less than promising for those of us hoping to see a final figure or a forecast.

Never mind. We know the figure is below -2 and we know that such a low value has been achieved just once in 38 years (2009).

Most importantly imo Steve Murr at The Other Forum has calculated that when the OPI is lower than -0.5 and the QBO is neutral or easterly (as it will be this winter), then 90% of subsequent winter months show a negative Arctic Oscillation. Which means that regardless of the weather on the ground we're almost guaranteed plenty of fun on the Model Output thread from a month's time. And let's face it, when you consider how many people seemed to spend Dec. 2010 on here looking for the breakdown, that's arguably more important than any actual forecast or actual weather.

 

---

I think they will be making the forecast public. That's certainly the impression I've gotten anyway.

---

Yes. That's the impression I got, but then I saw that sinister padlock...

---

Last year's results were in "American Weather" on 7th November so it doesn't sound like it will be long to wait

---

Most of us are interested in improving the accuracy of  forecasting winter temperature ,looking particularly  for cold winters and the increased snow risk this brings.

There are however two hurdles to jump.The first is being able to forecast the AO or NAO. The METO NAO forecasts we believe achieve ,using a number of variables ,66% accuracy this being measured in terms of getting the correct sign for the NAO index.The advantage of the OPI  to AO route is that it uses only one variable and appears to have a remarkably high correlation to the  AO. However at present we do not know what forecasts were made for the AO only what actually happened.

The second hurdle is this-even if we could forecast the AO or NAO with total accuracy how does this translate  into CET? On the NAO front the correlation between NAO and CET is  0.66.If we look at the 50/51 to 13/14 data, taking the lower values <1.5 there are 15 winters and 13 of them finished up cold( ie below mean.However given the sample size is small and the correlation between NAO and CET values is only 0.35

If we now look at AO v CET the overall correlation is very similar to that of the NAO 0.67.Taking the lower values and using  <1.5 gives 8 winters 60,63,66,69,70,77,86,09.Of these 6 finished up cold but correlaltion is weak at 0.25.So statistically the  the links are weak.

However we now have fascinating live experiment.The OPI points towards a AO- winter.It would be good to have the METO  NAOforecast when it is made usually some time in November if anyone can get hold of it!

 

I've spoken with the lovely Riccardo Valente today. Forecast should be made public within the coming week.

They've very kindly said I can publish it on my Blog page. 

---

I've been thinking about this AO forecasting tool and weighing my inner sceptic against the fact that, unless we are all being subjected to some massive hoax, there is a remarkable correlation with winter AO.

Then pondering, as have others, how far that gets us with forecasting actual UK winter weather. Clearly the correlation between CET and the NAO is pretty strong, for obvious reasons: the  NAO is a statistical expression of the pressure pattern in our immediate vicinity. The AO-CET correlation seems reasonably strong - certainly statistically significant but not strong enough to bet your financial future on.

But I say all this as someone who generally dislikes cold weather and would love nothing more than a continuation of this Autumn's balmy conditions for the next 5 months. Yet still enjoys the occasional cold snowy outbreak, because of the excitement it sparks in my son and the short lived prettiness of thick snow. So what does a negative AO do for the likelihood of at least one sustained cold outbreak in Southern Britain during the course of an otherwise bog standard winter? I'd be willing to bet the relationship is much stronger than with mean winter temperatures. Does anyone have the stats at hand?

---

I was thinking about the OPI last night and have a few questions:

1) Why is it hosted on a top level domain which appears to be related to car / van hire?

http://www.til.it/

2) Why were the updates to the OPI during October irregular? This suggested a manual process which would be quite odd and possibly prone to large errors.

3) Why are there no links or abstracts to research on the OPI apart from on weather forums? The universities running the project surely should give an overview on their web site/s? Even 20 years ago when I worked as a research assistant with universities in Europe this was the case.

4) Why has the peer review process taken so long? Surely a paper should have been made available by now because this was being discussed in October 2013 if not earlier

I very much appreciate Riccardo Valente taking the time to register and post about the OPI on TWO, but I think rather than blindly accepting this we do need to know more. 

 

---

I can only answer the latter one but I don't think it's been submitted for peer review as of yet, as they are still ironing out a few details. I maybe wrong and I'm sure someone from the OPI team will rectify me if that's the case.

My concern is that the correlation of 0.91 between OPI(actuals) and AO (actual DJF) looks too good to be true.Normally things that look too good to be true....

 

I'm not sure what the difficulty is as far as we amateurs are concerned.

We know the parameters being investigated and can see them for ourselves 4 times a day on GFS during October, i.e. the 500 hPa geopotential anomaly across the northern hemisphere.

We know that for a negative OPI we need red across the Arctic aligning as close as possible to an axis running from Canada across the pole to Siberia. The precise formula for calculating these anomalies and angles in numerical form is the subject of the eventual paper, but the authors have revealed enough for all of us to check that the numbers they publish are in line with what we are seeing on the charts.

They have answered the question of irregular updates by pointing out that they ignore GFS runs that look suspicious or unreliable, exactly as the NOAA does in its forecast discussion on a daily basis. These updates don't actually matter in themselves, as the whole index is calculated at the end of the month on the basis of what actually transpired, with the model runs being irrelevant. The updates are just a means of keeping us interested/excited/despairing as we approach winter.

Beyond that I'm speculating regarding the questions Brian has raised.

Is a year necessarily a long time for peer review, especially on a subject which can only be tested every 12 months? If the authors wish to keep track of the daily GFS updates during October via the online equivalent of the back of a cigarette packet, well, that's up to them. If I were employed academically and were engaged in some slightly oddball research, I'd be inclined to keep it away from my employer's official media too. That way your employer will not get needlessly embarrassed if it all goes wrong.

---

The data series for OPI and AO are available via Google queries (or Gavin's vid in the case of OPI !)

If you have the energy you can enter them into Excel and do the correlation yourself

I was a bit suspicious so I did that and got 0.91 as well

but as the saying goes "a correlation does not mean a causation

---

Also there is an issue that you can devise a model that fits perfectly provided you reduce the DOF enough. So if you have 2 data points a linear model always works, 3 data points and a quadratic model always works, 4 data points and a cubic model always works and so on. So just because a model can produce a very good correlation doesn't mean it is necessarily a good fit, although of course it might be. The real test comes in the prediction. However despite my skeptism I was reassured that the author citations are real scientists; I would have been more concerned if the names yielded no results or results in an unrelated field. 

---

Something that perhaps should be taken note of regarding the OPI this year, is that if you look at the six years with an OPI of -1.5 or less, four of those years were cold or very cold, and the other two were only fractionally warmer than average.  This year will be the first "very warm" year that coincides with a very negative OPI.  So, you could argue we are in unknown territory, and we have no way of knowing how much that excessive warmth will override the signal that we're getting from the OPI.   In the same way that the excessive warmth of September and October 2011 may well have scuppered the chances of a cold Winter in 2012 perhaps?   

 

 

 

 

 

---

Just in case anyone else other than me hadn't noticed, the data we have been looking at through October is still there if you click on the the OPI final data. The last value shown is 2.12 based on the 12z from the 31st.

---

 

Creating a paper can take a year even once you have a lot of the data and Peer review can take a couple of years on top of that. These are at the longer ends of the spectrum.

Once you have some data and a hypothesis you spend a few months writing a paper and proving results. You will then circulate the paper with peers and incorporate their feedback, this can take another 3-6 months. You then might need to collect more data, or re-write large sections.

You then submit to the journal- typically journals want to give you a review back in 90 days but often it can be more like 6-9 months! Your paper might be rejected then based on the reviews. If there is constructive feedback you will spend some months incorporating that and then sending the paper to a new journal where you will await the same 3-9 month period. These steps can be repeated many times until you ae lucky or give up.

 

Once you have a paper initially accepted with revision you then typically have 3 months to make the changes and improvements. You submit the payer and will wait up to another 3 months to hear back on whether the changes were acceptable. If they are then there is still some back and forth to sort out minor issues, this can take a month. If you are lucky they will publish the paper online at that point - but some journals will only put the paper online once it is in print, which might not be for a year or longer while the journal organizes its papers. However, even when hearing back from the revised paper there can be pitfalls: quite possibly they will have deemed that you have resolved the reviewer concerns so the paper is rejected - you then start back at square one and find a new journal. Rarer still but it some fields it is very common that you will have to make more significant changes and submit it back to the journal, each time inuring 3- 6 months delay. There can be 2-3 rounds of this. Even after all of that you might get the paper rejected at the 3rd round of reviewing a couple of years down the road. This is when you quit academia and get a real salary in industry.

 

 

Of course it is possible to go from an idea to a published paper under 12 months but there are many, may factors. 

Having a correlation of 0.9 might not be enough to get a publication, the reviewers may want a sound theory about why the correlation exists, the causation, and a detailed model. Developing the model can take months and months. Even if everything is scientifically sound there are many reasons for rejection.  I reject papers if there is a certain number of English mistake reached by the first page for example. A paper might also be rejected because the editor is not interested in that kind of work in their journal.  You might get reviewed by competitors who don't want you to get published before them. Many editors are just incompetent fools who dont read the paper or the reviewer comments and will just reject if one of the reviewers says reject.  And then their is the whole political side of it - there are academic niches based on culture/country where people only accept papers from their in-network colleagues.

As the OPI forecast comes out today thought I would bring this back up

---

I'm ready - I've got the beers and popcorn in and Deal or No Deal is on record

---

I have to say I can't wait to see the forecast because if you look at all the long range models and what the met are saying you would feel a little isolated if you were to make a forecast for below average temps for the UK when they are going for a wet and mild winter. I hope the team stay true to themselves and put it out there as they see it without any sitting on the fence like some do.

As I have said in the Model thread there is so much in favour of a colder than average winter but the computers say no 😕