OCTOBER PATTERN INDEX (OPI) MONITORING WINTER SEASON 2014-2015

[chionomaniac]

This index and its calculations fascinate me with such huge swings at this stage of the month, Time for another side by side comparison to have a look at what is different.

 

They look pretty similar to me GF, lol.

[Gael_Force]

They look pretty similar to me GF, lol.

 

......... fixed

[Interitus]

Lol what a waste if time.(interitus)

Assumption flaw-

That the AO = OPI

The best correlation you could run is as follows-

1) all pos OPI years correlate V the following winter AO & all neg OPI years correlate to the following winter AO

That gives you the lowest correlation score as it includes months with very low neutral scores.

 

 

Since when has it been assumed that the AO = OPI?

The correlations I've shown are reproducible by anyone with the freely available data and with regards to the Taymyr peninsula the research has been subject of a published paper linked to two days ago in this thread - https://forum.netweather.tv/topic/81637-eddy-heat-flux-over-eurasian-snow-cover/

The point is that it supports a possible link between October circulation and the following winter AO which may be measured by the OPI.

But as it is unclear exactly how the OPI is calculated it cannot be independently tested. We already know that the values appear to fit the historical data well so it is correlations of the OPI and winter AO which are the waste of time.

[Fingers]

Back to where we were this time yesterday at -2.75. Let's hope it doesn't rise on the 12z.

[karyo]

-2.75! We are doing well. :-)

Edited October 23, 2014 by karyo

[radiohead]

-2.75! We are doing well. :-)

Very little change from the 0Z. Starting to stabilize.

Edited October 23, 2014 by radiohead

[karyo]

Very little change from the 0Z. Starting to stabilize.

It is indeed and I am starting to get excited!

[Continental Arctic Front]

It is indeed and I am starting to get excited!

 

i'm holding back for a little month because it's still not evaluated by the prof. team. But when the blocking gets in the forecast around the end of november-dec and the OPI team aproves this OPI score then i will start to get excited (and more then usual! )

[Fingers]

It is indeed and I am starting to get excited!

Only 8 days of forecasting left, get that down to 5 and we are at the business end!

[karyo]

Only 8 days of forecasting left, get that down to 5 and we are at the business end!

I would normally be more cautious, especially if a cold spell was forecasted for the UK, but as southwesterlies are forecasted then nothing can go wrong!

[shaky]

My god -2.75 is excellent!!especially considering that the month is coming to a close!!!are we actually gone end up around -2.75 -2.50 :0!!!

[Mucka]

I have now seen Chino and Steve refer to the OPI as a hindcast. Maybe I am missing something in the meaning or something is getting lost in translation but doesn't every scientific theory uses observation and past data to formulate a theory that can hopefully make successful forecasts/predictions which will determine if it has any merit as a scientific principle?

Every theory is reverse engineered to explain the observed world and then tested for predictive accuracy isn't it?

 

Hindcast to me means that you modify your theory to fit new data when it was wrong.

Edited October 23, 2014 by Mucka

[chionomaniac]

I have now seen Chino and Steve refer to the OPI as a hindcast. Maybe I am missing something in the meaning or something is getting lost in translation but doesn't every scientific theory uses observation and past data to formulate a theory that can hopefully make successful forecasts/predictions which will determine if it has any merit as a scientific principle?

Every theory is reverse engineered to explain the observed world and then tested for predictive accuracy isn't it?

 

Hindcast to me means that you modify your theory to fit new data when it was wrong.

No, hindcast is using past data to test your mathematical theory. However, you have the benefit of the statistical data to build your theory - in fact you may have played around a number of times to modify your theory so that the backtest is improved. I very much doubt that the OPI team came out with their formula at the first time of asking. However, the true test comes in the forecast element.

 

http://en.wikipedia.org/wiki/Hindcast

[radiohead]

-2.39 after the 12Z. Certainly starting to look like we are getting better consistency today compared to previous days.

Edited October 23, 2014 by radiohead

[Isolated Frost]

-2.39, a third consecutive drop if we compare the 12z day by day (which we should really). Just 8 more now...

[SMU]

Since when has it been assumed that the AO = OPI?

The correlations I've shown are reproducible by anyone with the freely available data and with regards to the Taymyr peninsula the research has been subject of a published paper linked to two days ago in this thread - https://forum.netweather.tv/topic/81637-eddy-heat-flux-over-eurasian-snow-cover/

The point is that it supports a possible link between October circulation and the following winter AO which may be measured by the OPI.

But as it is unclear exactly how the OPI is calculated it cannot be independently tested. We already know that the values appear to fit the historical data well so it is correlations of the OPI and winter AO which are the waste of time.

 

Hi

 

heres your post

Over the 38 years of the OPI data, the October AO correlation with winter AO is a weak positive 0.279 and 65.8% of winters have the same AO sign as October (compared to ~50% by chance).

But looking at the October 500mb geopotential anomaly over only the north pole from NCAR/NCEP plots though shows that a positive anomaly followed by a negative winter AO and vice versa is correct 73.7% of the time with a correlation of 0.387.

 

 

I don't understand why you are trying to correlate the AO mode in October, with the AO mode in the winter- the statistics - even the AO table on the website shows the data where the relationship isn't SIGNIFICANT.

Also the problem with doing 'all' months & segregating them into a mere pos or negative is you have no neutral point, with this what you actually doing is reducing your correlation.

 

I believe to do it correctly you would group +0.5 Octobers ( & above ) & group -0.5 Octobers & below & see where the correlation is. ( or +1 / -1)

but to be honest that not part of the thread.

 

This is the OPI thread, & we have 3 ways of dealing with the lack of recorded visible data.

 

* dismiss it all - but without clear evidence to the contrary you cant

* do nothing at all but go with the general flow

* embrace it & work with it but understand that because of the hindcast timeline V the actual forecast timelines there is a caveat that should the AO index not follow the OPI for 2014/2015 15/16 etc then perhaps the data was initially skewed to display a high correlation.

 

We cannot argue the negative figures ive posted them, The best figure starts to emerge around -1.5. & would probably improve if you segregate W QBO years V E QBO years- however the sample size would be limited.

 

 

In terms of the abstract link, which is a little difficult to understand - its saying ( or should be ) that there is a positive relation with the height anomalies over the Taymyr region in October to the AO index in Winter.

The relationship should be +VE Positive heights in the region inbetween Kara sea & eastern Siberian Sea in October feedbacks to a positive height anomaly in that region in Winter.

 

So negative AO October ( STRONG) correlates to the strong Negative AO in winter. ( see data point in image below)

 

Taymyr anomaly October.

 

Subsequent Winter

 

 

Theres nothing new in all this An index called the POL ( polar Eurasia ) has been measured for a long time by the NOAA-

If you look at the positive phase of the POL for October it has the taymyr region as low heights-

http://www.cpc.ncep.noaa.gov/data/teledoc/poleur_map.shtml

 

So a NEGATIVE POL is what we want with huge heights in that region.

ftp://ftp.cpc.ncep.noaa.gov/wd52dg/data/indices/poleur_index.tim

OCT 2009 very negative.....

 

 

 

Going back to the OPI - The data presented suggests a 0.8 to 0.9 correlation with the Winter AO which becomes very strong after the index gets below -1.5 ( also above +1.5)

 

throw in E QBO years & I would hazard a guess that the correlation is near on 1 for the few sample size....

 

S

PS quote you.

 

But as it is unclear exactly how the OPI is calculated it cannot be independently tested. We already know that the values appear to fit the historical data well so it is correlations of the OPI and winter AO which are the waste of time

 

Whats actually pointless is you posting in a thread where you don't believe the data because YOU physically cant prove or disprove the original data supplied.. so vis a vee all correlations claimed by the OP are a waste of time because you don't have the information.

 

I can see a one correlation here infact. -

 

the input of your posts detracting from the purpose of the thread & what its trying to achieve has a high correlation with the recycle bin.

Edited October 23, 2014 by Steve Murr

[Mucka]

No, hindcast is using past data to test your mathematical theory. However, you have the benefit of the statistical data to build your theory - in fact you may have played around a number of times to modify your theory so that the backtest is improved. I very much doubt that the OPI team came out with their formula at the first time of asking. However, the true test comes in the forecast element.

 

http://en.wikipedia.org/wiki/Hindcast

 

Hi Chino,

 

Maybe this is just semantics but what is a mathematical theory built on if not past data whether that be in statistical form or observational form?

And surely the test of the theory of the October Pattern Index is whether it can forecast future patterns not explain previous ones which would be part of the development and refinement of the theory?

Edited October 23, 2014 by Mucka

[SMU]

-2.39, a third consecutive drop if we compare the 12z day by day (which we should really). Just 8 more now...

 

That's not a bad finishing zone...

 

S

[lorenzo]

Has anyone managed to get a PDF of the Taymyr paper yet, everywhere I looked today it was paywalled?

[Nouska]

Steve mentions the Eurasian index above - there is a recent paper on this that I posted in the last strat thread. Some interesting theories that relate to the solar cycle fluctuations.

 

http://www.atmos-chem-phys.net/13/6275/2013/acp-13-6275-2013.pdf

[nick sussex]

Hi Chino,

 

Maybe this is just semantics but what is a mathematical theory built on if not past data whether that be in statistical form or observational form?

And surely the test of the theory of the October Pattern Index is whether it can forecast future patterns not explain previous ones which would be part of the development and refinement of the theory?

It's a complicated problem for meteorology compared with lets say a situation where you have a simple hypothesis and test it in other scientific areas.

 

If you developed a theory that was testing the OPI for future forecast patterns the researcher would be six feet under before the theory could be properly tested, however if you build a theory around certain historic patterns and see how that's repeated then you can still build up a good picture.

 

Lets take the NAO for example in western Europe, its commonly accepted that a negative NAO is linked to colder winters but that is based on our reviewing past patterns, if we don't look back then trying to develop a theory would be almost impossible. Where exactly would we start?

 

Now of course it maybe possible that years down the line the OPI will be viewed just as the NAO is or  the AO.

[bluearmy]

judging from tonights NOAA cpc update, we may have to wait a few more days before we see a landing zone < 0.5 spread

'SIGNIFICANT UNCERTAINTY IS EVIDENT IN THE MODEL FORECASTS OVER ALASKA AND THE  NORTH PACIFIC EVEN AT RELATIVELY EARLY LEAD TIMES.' 

Edited October 23, 2014 by bluearmy

[karyo]

 

judging from tonights NOAA cpc update, we may have to wait a few more days before we see a landing zone < 0.5 spread

'SIGNIFICANT UNCERTAINTY IS EVIDENT IN THE MODEL FORECASTS OVER ALASKA AND THE  NORTH PACIFIC EVEN AT RELATIVELY EARLY LEAD TIMES.' 

Could this mean that there can be more blocking there than the models currently show?

[Crocodile23]

No, hindcast is using past data to test your mathematical theory. However, you have the benefit of the statistical data to build your theory - in fact you may have played around a number of times to modify your theory so that the backtest is improved.

 

This is not necessarily a mistake. If the procedure is done correctly and you succeed in a high correlation of your prediciton then nothing is wrong of course and you have done a good job. But of course the question is always if you haven't ovedone it with your model and your number of inputs is larger compared to the observation and the phenomenon you want to explain. I think you(English dudes) call it overfitting or something right?

 

And the ultimate and perhaps only test is of course the FUTURE. If your model can explain the next new "values",- in our case the next AO-, with similar correlation as the hindcasted data, then you are ok and your model really works.

Edited October 23, 2014 by Crocodile23

[bluearmy]

Could this mean that there can be more blocking there than the models currently show?

 

of course - but then it could mean the opposite !