Jump to content
Sign in to follow this  
Riccardo

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

Recommended Posts

Yes  MS ^^ go to the OPI link & he how the axis parameters are driving the negative results.

Edited by Steve Murr

Share this post


Link to post
Share on other sites

The loading pattern for negative OPI years that Steve has posted show a very positive anomaly over the Taymyr peninsula - this seems to tie in with the new research link that I put in the NHsnow thread.

Share this post


Link to post
Share on other sites

 

Ah right. Well, if anyone wants to use the one I made, go ahead.

 

Year: OPI

1976:  -1.75
1977:  -0.95
1978:  -1.8
1979:  -0.5
1980:  -0.05
1981:  -0.4
1982:  -1.1
1983:  -0.3
1984:  -1.8
1985:  -1.9
1986:  -1.3
1987:  -0.45
1988:  1.45
1989:  0.25
1990:  0.55
1991:  1.1
1992:  1.75
1993:  -0.15
1994:  0.7
1995:  -0.65
1996:  -0.75
1997:  -0.75
1998:  0.1
1999:  -0.2
2000:  -0.75
2001:  0.45
2002:  -0.9
2003:  -0.7
2004:  0.3
2005:  -0.7
2006:  0.85
2007:  0.75
2008:  0.25
2009:  -3.15
2010:  -0.85
2011:  0.65
2012:  -1.65
2013:  1.6

 

 

Lovely list, many thanks. :)  I might use this as the basis for a post on my blog tonight or tomorrow, if that's OK?

 

Don't suppose theres any way of finding out what the OPI was in 46 and 62 is there? :wink:

Edited by Gavin P

Share this post


Link to post
Share on other sites

 

 

I have asked some follow up questions which should clarify the process further and also answer such things as to whether strong trends are considered when the final forecast is compiled or if only the final OPI is used regardless.

good to get an update on what it actually means.

We can each of course suggest what might be the most constructive way to do it. As now part actual part 10 day prediction, or just on the average of the actual data through the month. One thing that might help the reduction in swings I read about on here is to use just one GFS run. Over the years that is probably a toss up between the 00 and 12z. But choose one. Comparing the single output each day for D+10 days or so has almost always proved less volatile when using just one run each day.

My views only and no doubt a number will disagree. But the daily up/down comments if you read the thread carefully do reinforce what most of us know if some don't accept fully, one run per day=less volatility in outputs and less of the same on the thread discussing it.

To me using actual data and allowing that to accumulate to an average would be the best way. It would also allow anyone to see how the OPI ACTUAL index changes over the 31 days. There may be some relevance in how that changes to the outcome for the winter assuming that once peer reviewed it can be used as another predictor.

Very helpful to have that post though so we now all know what is going on. Good luck with the work Guido, only by work such as this does any form of science move forward.

Share this post


Link to post
Share on other sites
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.

Now the figures aren't up to what is claimed for the OPI but shows they might have stumbled on something.

 

65.8% accuracy over 38 data points vs 50% by random chance. 0.279 correlation. A weak correlation indeed. Considering 38 data points - that's possibly within the margin of error if there was no correlation. There's probably something to the OPI but these numbers don't suggest a holy grail of long range forecasting. It's never going to be that easy. If something sounds too good to be true...

 

No to mention the fact that a -AO winter does not in any way guarantee cold snowy weather... I don't understand why some are putting so much weight and importance on what the OPI figure will be? Maybe I'm missing something.

Share this post


Link to post
Share on other sites

65.8% accuracy over 38 data points vs 50% by random chance. 0.279 correlation. A weak correlation indeed. Considering 38 data points - that's possibly within the margin of error if there was no correlation. There's probably something to the OPI but these numbers don't suggest a holy grail of long range forecasting. It's never going to be that easy. If something sounds too good to be true...

 

No to mention the fact that a -AO winter does not in any way guarantee cold snowy weather... I don't understand why some are putting so much weight and importance on what the OPI figure will be? Maybe I'm missing something.

 

made in, everything depends on the shape of the axis and the inclination of the polar vortex and all that, we know that early November

Share this post


Link to post
Share on other sites

65.8% accuracy over 38 data points vs 50% by random chance. 0.279 correlation. A weak correlation indeed. Considering 38 data points - that's possibly within the margin of error if there was no correlation. There's probably something to the OPI but these numbers don't suggest a holy grail of long range forecasting. It's never going to be that easy. If something sounds too good to be true...

 

No to mention the fact that a -AO winter does not in any way guarantee cold snowy weather... I don't understand why some are putting so much weight and importance on what the OPI figure will be? Maybe I'm missing something.

 

Someone posted this earlier, but if you just look at the relationship with DJF CET anomalies together with the OPI for the past 13 years then there is almost certainly some connection. Perhaps it's something not directly related to the AO? But this doesn't look like coincidence.

 

tGzkWwC.png

Share this post


Link to post
Share on other sites

65.8% accuracy over 38 data points vs 50% by random chance. 0.279 correlation. A weak correlation indeed. Considering 38 data points - that's possibly within the margin of error if there was no correlation. There's probably something to the OPI but these numbers don't suggest a holy grail of long range forecasting. It's never going to be that easy. If something sounds too good to be true...

 

No to mention the fact that a -AO winter does not in any way guarantee cold snowy weather... I don't understand why some are putting so much weight and importance on what the OPI figure will be? Maybe I'm missing something.

 

0.279 is the october AO - winter AO correlation. NOT the OPI winter AO correlation.

 

October AO is but one factor in the OPI, The OPI has a 90% correlation with winter AO - , albeit based on the data set it was presumably calibrated on.

 

So far we have had one test of the OPI last year. The AO predicted was the right sign but less positive than may have been expected from the the OPI. This year we will get a second test and if it is a strong negative (as it looks to be) it should be a good test of the model.

Share this post


Link to post
Share on other sites

65.8% accuracy over 38 data points vs 50% by random chance. 0.279 correlation. A weak correlation indeed. Considering 38 data points - that's possibly within the margin of error if there was no correlation. There's probably something to the OPI but these numbers don't suggest a holy grail of long range forecasting. It's never going to be that easy. If something sounds too good to be true...

 

No to mention the fact that a -AO winter does not in any way guarantee cold snowy weather... I don't understand why some are putting so much weight and importance on what the OPI figure will be? Maybe I'm missing something.

 

Sorry, 0.279 is between the October AO and winter AO, 0.387 is between October north pole geopotential and winter AO.

Taymyr is the new buzzword - just compared the October geopotential on the end of the Taymyr peninsular vs winter AO and managed to get a correlation of 0.62.

The OPI manages above 0.9 fitted to historical data, though to quote John von Neumann - "With four parameters I can fit an elephant, and with five I can make him wiggle his trunk"

Share this post


Link to post
Share on other sites

The OPI manages above 0.9 fitted to historical data, though to quote John von Neumann - "With four parameters I can fit an elephant, and with five I can make him wiggle his trunk"

Good Quote. OPI has 7 parameters doesn't it? Presumably with 7 he could make it dance.

 

Edit: Actually I think it has 5 the other two t are derived from the 5 inputs.

Edited by SomeLikeItHot

Share this post


Link to post
Share on other sites

I like simplicity whenever possible

So to see what happened I took all the OPI indicators that showed a -ve value of 1.5 or more and then put the CET figures for Dec/Jan?feb with the coldest ones.

This is what it showed

OPI data for possible ‘cold’ winter months

 

Year: OPI

1976:  -1.75

1977:  -0.95

1978:  -1.8

1979:  -0.5

1980:  -0.05

1981:  -0.4

1982:  -1.1

1983:  -0.3

1984:  -1.8

1985:  -1.9

1986:  -1.3

1987:  -0.45

1988:  1.45

1989:  0.25

1990:  0.55

1991:  1.1

1992:  1.75

1993:  -0.15

1994:  0.7

1995:  -0.65

1996:  -0.75

1997:  -0.75

1998:  0.1

1999:  -0.2

2000:  -0.75

2001:  0.45

2002:  -0.9

2003:  -0.7

2004:  0.3

2005:  -0.7

2006:  0.85

2007:  0.75

2008:  0.25

2009:  -3.15

2010:  -0.85

2011:  0.65

2012:  -1.65

2013:  1.6

The following winters after an OPI of -1.65 or lower

1976/7=2.0/2.8/5.2

19789=3.9/-0.4/1.2

1984/5=5.2/0.8/2.1

1985/6=6.3/3.5/-1.1

2009/10=3.1/1.4/2.8

2012/13=4.8/3.5/3.2

CET values for Dec/Jan/Feb 1971-2000=5.1/4.2/4.2

 

Winters not show up by OPI

1981/2=0.3/2.6/4.8      OPI=-0.4

1982/3=4.4/6.7/1.7                  -1.1

1990/91=4.3/3.3/1.5                0.55

1995/6=2.3/4.3/2.5                  -0.65

1996/7=2.9/2.5/6.7                  -0.75

2010/11=only Dec 

Now you make of those figures what you want. How much statistical correlation there is between the lowest OPI values and the winter months I do not know. My statistical recall from my inter Bsc days has gone like snow on a hot oven!

 

to add, someone might like to take each October CET preceding each CET 'cold' winter and see if that correlates with them. To be honest until you do a full statistical test on any work then it is possible that any apparent correlation is not what it seems. Lots of fun though especially as the cold lovers so obviously will almost die for a link to be true. Some way off that being proved I am afraid.

 

from Steve

Lol what a waste if time

 

not sure which of us he refers to but a touch unkind Steve. Can you prove that your theory, which is what it is, is any sounder or more scientifically based than anyone else so far?

To me, until this is peer reviewed it is interesting and allows us all a bit of fun?

Edited by johnholmes

Share this post


Link to post
Share on other sites

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.

Run the same exercise working on a the baseline of
+ or -0.5 months
Then
+ or -1 months up to months of 3.

The correlation will go up slowly until a big jump to high correlation

Thats the point where you would draw the line of calling the index a neutral year to a positive or negative one

I suggest the neutral line is between 1 & 1.5...... (& -1 & -1.5)

S

Edited by Steve Murr

Share this post


Link to post
Share on other sites

I like simplicity whenever possible

So to see what happened I took all the OPI indicators that showed a -ve value of 1.5 or more and then put the CET figures for Dec/Jan?feb with the coldest ones.

This is what it showed

OPI data for possible ‘cold’ winter months

 

Year: OPI

1976:  -1.75

1977:  -0.95

1978:  -1.8

1979:  -0.5

1980:  -0.05

1981:  -0.4

1982:  -1.1

1983:  -0.3

1984:  -1.8

1985:  -1.9

1986:  -1.3

1987:  -0.45

1988:  1.45

1989:  0.25

1990:  0.55

1991:  1.1

1992:  1.75

1993:  -0.15

1994:  0.7

1995:  -0.65

1996:  -0.75

1997:  -0.75

1998:  0.1

1999:  -0.2

2000:  -0.75

2001:  0.45

2002:  -0.9

2003:  -0.7

2004:  0.3

2005:  -0.7

2006:  0.85

2007:  0.75

2008:  0.25

2009:  -3.15

2010:  -0.85

2011:  0.65

2012:  -1.65

2013:  1.6

The following winters after an OPI of -1.65 or lower

1976/7=2.0/2.8/5.2

19789=3.9/-0.4/1.2

1984/5=5.2/0.8/2.1

1985/6=6.3/3.5/-1.1

2009/10=3.1/1.4/2.8

2012/13=4.8/5.7/6.2

CET values for Dec/Jan/Feb 1971-2000=5.1/4.2/4.2

 

Winters not show up by OPI

1981/2=0.3/2.6/4.8      OPI=-0.4

1982/3=4.4/6.7/1.7                  -1.1

1990/91=4.3/3.3/1.5                0.55

1995/6=2.3/4.3/2.5                  -0.65

1996/7=2.9/2.5/6.7                  -0.75

2010/11=only Dec 

Now you make of those figures what you want. How much statistical correlation there is between the lowest OPI values and the winter months I do not know. My statistical recall from my inter Bsc days has gone like snow on a hot oven!

 

to add, someone might like to take each October CET preceding each CET 'cold' winter and see if that correlates with them. To be honest until you do a full statistical test on any work then it is possible that any apparent correlation is not what it seems. Lots of fun though especially as the cold lovers so obviously will almost die for a link to be true. Some way off that being proved I am afraid.

 

from Steve

Lol what a waste if time

 

not sure which of us he refers to but a touch unkind Steve. Can you prove that your theory, which is what it is, is any sounder or more scientifically based than anyone else so far?

To me, until this is peer reviewed it is interesting and allows us all a bit of fun?

1976 to 1987 a run of neg OPI 12 years in a row!

 

i wonder if this has any kind of link with solar activity.

and i assume that not all these winters had a neg ao and nao?

and how many of them neg OPI years were east QBO ?

Share this post


Link to post
Share on other sites

John - I make the CET figures for Dec-Feb (2012/2013):

 

4.8, 3.5, 3.2

 

which is lower than you've quoted (looks like you quoted Jan/Feb 2014 instead of 2013). The most notable thing about that winter was the March value of course being 2.7

Edited by beng

Share this post


Link to post
Share on other sites

I like simplicity whenever possible

So to see what happened I took all the OPI indicators that showed a -ve value of 1.5 or more and then put the CET figures for Dec/Jan?feb with the coldest ones.

This is what it showed

OPI data for possible ‘cold’ winter months

 

Year: OPI

1976:  -1.75

1977:  -0.95

1978:  -1.8

1979:  -0.5

1980:  -0.05

1981:  -0.4

1982:  -1.1

1983:  -0.3

1984:  -1.8

1985:  -1.9

1986:  -1.3

1987:  -0.45

1988:  1.45

1989:  0.25

1990:  0.55

1991:  1.1

1992:  1.75

1993:  -0.15

1994:  0.7

1995:  -0.65

1996:  -0.75

1997:  -0.75

1998:  0.1

1999:  -0.2

2000:  -0.75

2001:  0.45

2002:  -0.9

2003:  -0.7

2004:  0.3

2005:  -0.7

2006:  0.85

2007:  0.75

2008:  0.25

2009:  -3.15

2010:  -0.85

2011:  0.65

2012:  -1.65

2013:  1.6

The following winters after an OPI of -1.65 or lower

1976/7=2.0/2.8/5.2

19789=3.9/-0.4/1.2

1984/5=5.2/0.8/2.1

1985/6=6.3/3.5/-1.1

2009/10=3.1/1.4/2.8

2012/13=4.8/5.7/6.2

CET values for Dec/Jan/Feb 1971-2000=5.1/4.2/4.2

 

Winters not show up by OPI

1981/2=0.3/2.6/4.8      OPI=-0.4

1982/3=4.4/6.7/1.7                  -1.1

1990/91=4.3/3.3/1.5                0.55

1995/6=2.3/4.3/2.5                  -0.65

1996/7=2.9/2.5/6.7                  -0.75

2010/11=only Dec 

Now you make of those figures what you want. How much statistical correlation there is between the lowest OPI values and the winter months I do not know. My statistical recall from my inter Bsc days has gone like snow on a hot oven!

 

to add, someone might like to take each October CET preceding each CET 'cold' winter and see if that correlates with them. To be honest until you do a full statistical test on any work then it is possible that any apparent correlation is not what it seems. Lots of fun though especially as the cold lovers so obviously will almost die for a link to be true. Some way off that being proved I am afraid.

 

from Steve

Lol what a waste if time

 

not sure which of us he refers to but a touch unkind Steve. Can you prove that your theory, which is what it is, is any sounder or more scientifically based than anyone else so far?

To me, until this is peer reviewed it is interesting and allows us all a bit of fun?

Thanks John, an interesting analysis and there does appear to be a correlation without a doubt but how strong a correlation remains to be seen with such limited data to fall back on. I look forward to this being peer reviewed at some point as I certainly feel it offers another layer to our limited forecasting skills for LRF.

Share this post


Link to post
Share on other sites

John - I make the CET figures for Dec-Feb (2012/2013):

 

4.8, 3.5, 3.2

 

which is lower than you've quoted (looks like you quoted Jan/Feb 2014 instead of 2013). The most notable thing about that winter was the March value of course being 2.7

 

 

ta-my mistake

Share this post


Link to post
Share on other sites

The comment around wasting time was aimed at all the work of calculating the wrong correlation!!!

Anyone fancy doing it correctly PM me

S

Share this post


Link to post
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Restore formatting

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

Sign in to follow this  

×
×
  • Create New...