October Pattern Index! Predicting Winter AO from October, with 90+% accuracy?

[DTWXRISK]

I think your issue is with the forecasting firm not the research. But it is also clear that you unfairly criticized the research, too, as others have pointed out.

 

If your problem is that people are suggesting a warm winter based solely on the SAI...fine (I think any sensible forecaster would agree that it's not too bright to base a temperature forecast on 1 variable)... but if you actually think the SAI suggests a -AO this year, you are seriously misreading the variable.

 

 

 DR COHEN says  I might  not be   mis reading anything ...   

 

Here's the facts on the actual, physical processes right now: there aren't too many autumns as suggestive as 2013 for a +AO. Even if you take away the OPI and SAI, the wave/jet behavior is classic leading to a +NAM. The anomalous PV-cyclonic streamfunction over Siberia with poleward anticyclones are ugly for -AO predictions. The SAI / OPI just elucidate these observations. While there have been wave disturbances, none of these can be traced back to the actual snow cover, especially that of the late September period since the vortex is still forming!

 

The vortex really comes on in October and peaks in January, strength-wise. Upwelling waves in early autumn are basically useless in affecting the NAM state.

[DTWXRISK]

I agree with this  

 

 

 

The split came to fruition and we had a -AO winter. 

 

In 2012 this method correctly predicted the sign of the DJ AO by November 24th (possibly earlier if you believed stratosphere forecasts from Nov 19th onward).

 

Checking my posts from 2011, In 2011 this method correctly predicted the sign of the DJ AO by November 17th. 

 

So as a predictive method the past two years, the method is 2/2. If the PV is intact by the end of November you will have a +AO DJ (2011-2012). If it is disrupted by the the end of November you will have a -AO DJ (2012-2013). 

 

 

For 2013, I am not ready to declare a +AO based on asian stratosphere temperatures (although it is looking that way). Time remains for an SSW to save us. But SAI and OPI are both indicating a +AO. While their lead time is longer, the correlation is not as great. If the PV remains intact 2 weeks from now, we can be more certain of a +AO winter.

[DTWXRISK]

 YES   I should of used in SCE.. and in my   winter forecast  I  DID

 

If you want to latch on to something that supports your argument, it can't be the SAI. Instead, you could use Cohen's earlier work, the average Eurasian snow cover extent (SCE) during October. That was a little above normal this October, which would weakly support a -AO winter. Just be careful in doing so, since the correlation between that index and winter AO is much less convincing than the correlation between the SAI and the winter AO. And with the fact that things so far in November are supporting the SAI over the SCE, I'd be even more cautious about latching onto the SCE.

 

 

 

EDIT: Also, re: your last paragraph, statistically speaking, snow falling on October 2nd (or even October 8th) is detrimental to the SAI. There is no great change on October 1st where snowfall goes from being detrimental to the SAI to it being a boon. Since a regression line is fit through the days, the "best case scenario" is for generally low snow early on (say in the first half of the timeseries) and then high snow later (in the second half). That's why shifting the start date this year doesn't actually improve the SAI number.

[HM]

Well then what are we talking about here? You criticized the SAI, not the SCE. You also brought up the "process"...so I showed you the process and how it isn't all that strong right now (100mb heat flux). So if you originally meant SCE, then we have nothing to debate.

[Harry]

Well then what are we talking about here? You criticized the SAI, not the SCE. You also brought up the "process"...so I showed you the process and how it isn't all that strong right now (100mb heat flux). So if you originally meant SCE, then we have nothing to debate.

 

I believe DT is debating the idea that the AO will be so positive as a few have eluded to which that Judah Cohen appears to agree with him on?

[HM]

I believe DT is debating the idea that the AO will be so positive as a few have eluded to which that Judah Cohen appears to agree with him on?

 

If he wants to disagree with other forecasts, fine. But he said the following:

 

 

"What Judah Cohen found was a PROCESS  which exists between the rate of snow cover  buildup in Siberia   and how it impacts the phase of the AO.   It is the process  that is important here ...  NOT the fact that  the snow  fell on September 26 vs. October 2.

 

To suggest otherwise   is arguing that the process   has a  "magical start date "   of October 1  and ends of OCT 31.

 I do not see WHY   the snow  falling Siberia on one side of October 1   -- say  SEPT 26  is detrimental  to the  SAI  but if the same snow   falls   4 or  5  days later -Say OCT 2...   has a totally different and significant  implications for the SAI."

 

The "magical start date" has been debunked countless times. Also, I showed him the process on a pretty graph (poleward eddy heat flux has been avg to below). He then said he meant SCE, so naturally one has to ask...what the heck were we talking about because it definitely was the SAI!

 

I only care about the merits of the research and the SAI. What people have been alluding to is not my concern.

 

[The_Global_Warmer]

If the OPI and SAI were predicting a nuetral AO for DJF and DJFM like -.2 to .2.  Or anything more negative than that.  None of this conversation would be taking place.

 

DT would be writing posts about how the SAI and OPI are calling for a -AO and would surely incorporate that into his forecast.  There is no way DT or anyone else would be having the skeptism that is being seen if that was the case.

 

 

It looks like 1976 was year one.  So we have had what 37 or so winters.  Which is quite a few the entire time series correlates to over 90 percent accuracy.

 

 

 

For some reason.  Which could be ENSO, solar, who knows.  Since 2000. 13 winters. Thirteen 90 day periods.  About 1170 days worth of winter and the correlation on the OPI foreacst comes to 97%.  Thats unreal. 

 

 

 

 

 

So now all of a sudden in a year where the OPI is showing a 1.64 AO+.  Folks expect the correlation to break down at record levels?

 

 

Yeah maybe we will have a .5 AO+ for the winter.  We have seen the OPI break in the past. In fact most of the uncertainty is jsut from a few winters where it was off.

 

But 1982-83 when the OPI forecast was off by slightly over 1.0 on the AO scale. Also featured a massively powerful NINO.  With ONI monthlies busting into the 2.0+ range through the winter.

 

 

1988-89 had a very strong NINA with ONI of -1.6, -1.9. -1.9, -1.7, -1.5 from Oct-Feb.

 

 

Then we have 1997-98 and 1998-99.

 

The winter of 1997-98 had the Super NINO.

 

The winter of 1998-99 had a strong NINA.

 

It's obvously no coincidence that the most innacurate years also featured wild ENSO swings on the warm or cool side.

 

 

 

This year the ONI is -0.3 and has been for four months. The ONI will probably go up closer to 0.0 as the winter goes on.  But it's steady and nuetral. 

 

 

So it's really hard to believe we will see the same breakdown in the OPI that we have almost exclusively seen in major NINA or NINO winters.

 

 

 

 

 

 

[tacoman25]

I think some of the problem lies with people seeing correlating factors strongly pointing to a +AO, and thinking that means it will be a strong +AO for DJF. A strong signal can lead to a weak, but still correct, result.

[N. OF PIKE]

 

 

It looks like 1976 was year one.  So we have had what 37 or so winters.  Which is quite a few the entire time series correlates to over 90 percent accuracy.

 

 

 

 

 

 

Do we know if there was a strong SSW event that caused any of the year(s) where the OPI predicted a +AO and the correlation failed

[The_Global_Warmer]

Do we know if there was a strong SSW event that caused any of the year(s) where the OPI predicted a +AO and the correlation failed

 

That is a great question that I can not answer becuase I do not have direct access to historical lists of SSW events.  HM, ORH, or Skier could probably answer that.

 

It's a solid question because it's obvious other events outside the PV formation and placement play a large role in our global or NH regime. 

 

 

My defense of the OPI is based off the math.  Plus the obvious tie in to extreme ENSO regimes.  If we figure out how to correct for ENSO we could probably push the OPI's accuracy over 95%. 

 

 

In the extreme ENSO regime winters the OPI did much worse.  A few of them upwards of 1.0.  So at the max bust on record the 2013-14 OPI prediction of a 1.64AO+ could seemingly be around a .6 to .7AO+. 

 

So yeah like Tacoman said above it's definitely possible for the AO to be weakly positive in the .5 to .8 range vs the predicted 1.64.  But in terms of going negative then the SAI and OPI would have to have their correlations completely broken.  And if a raging NINO or raging NINA can't break it past a 1.0.  What could possibly do that?  I would think prior SSW would show up.  I am sure we have had SSWs since 2000 and the OPI is almost perfect since then.

 

Which I wonder if it has to do with the next generation of weather satelittes going up around that time. 

 

My dog in this fight is to defend the Science.  Whether the OPI is showing a 1.64+ or -1.64 I don't care.  I have no control over it.  I want the OPI to be right because it predicts percentage wise about as good as the 5 day euro.  And when you really think about it the 5 day Euro forecasts are pretty damn accurate on a large scale and the OPI is a large scale predictor. 

 

 

This is the beggening of the last walls of our ignorance of our atmospheric processes starting to break down.  The AO is the beggening.  We will soon or maybe in progress attm do this for all of our teleconnectors then we can start piecing them together to enhance everything in the medium to long range.  And I believe that will be our ticket to extremely precise 20, 30, 40, 50 day forecasting.  Pending how much infuence things like the sun have to disrupt our ability to forecast accuractly the global weather in a much longer range than now.

 

We may never be able to get 100 percent but when are breaking 90 percent accuracy with a forecast made on Nov 1st for 30 days later to 120 days later for a major atmospheric pattern index we are definitely making massive progress towards a much better understanding of it all. 

 

Which will be a great human achievement for many reasons, but it may kill the snow chasing passion we all share if we know whats coming so far in advance. 

 

[The_Global_Warmer]

I dont mean 40 days forecasts nailing precip and high temps.  I mean 40 day forecasts of the NAO, AO, PNA, EPO and so on the same level as the OPI for the AO or even better. 

 

I would think knowing that would allow us to better program our shorter range weather models to have more accurate 10-15 or even 15-25 day patterns.

 

 

If we have a 95% or higher confidence interval that all of our teleconnector will be in a certain alignment in 20 days I would assume we can help program the computer model to not veer from that global pattern regime. So even if the small stuff like QPF and temps are still off.  At least the models would be predicting within a range of certainly of the pressure fields.

 

that should help the model do better with the smaller things as well I would assume.  It could reach a point where the computer models are built to work within that 95 perent confidence range of the main pressure fields so it can not veer to solutions that break that 95 percent mark within the 20 day period.  This could I think help steer the model to realistic long range weather forecasting.

[Harry]

If he wants to disagree with other forecasts, fine. But he said the following:

 

 

"What Judah Cohen found was a PROCESS  which exists between the rate of snow cover  buildup in Siberia   and how it impacts the phase of the AO.   It is the process  that is important here ...  NOT the fact that  the snow  fell on September 26 vs. October 2.

 

To suggest otherwise   is arguing that the process   has a  "magical start date "   of October 1  and ends of OCT 31.

 I do not see WHY   the snow  falling Siberia on one side of October 1   -- say  SEPT 26  is detrimental  to the  SAI  but if the same snow   falls   4 or  5  days later -Say OCT 2...   has a totally different and significant  implications for the SAI."

 

The "magical start date" has been debunked countless times. Also, I showed him the process on a pretty graph (poleward eddy heat flux has been avg to below). He then said he meant SCE, so naturally one has to ask...what the heck were we talking about because it definitely was the SAI!

 

I only care about the merits of the research and the SAI. What people have been alluding to is not my concern.

 

 

See below.. But yeah he may wanna respond to clear things up a bit just to be sure.

 

 

 SORRY  SKI   and HM  but  JUDAH  COHEN says I am right ...sort of  

you know the   guy who developed and  found the  SAI / AO connection 

 

http://www.washingtonpost.com/blogs/capital-weather-gang/wp/2013/11/20/judah-cohens-winter-outlook-a-downer-for-east-coast-winter-weather-lovers/

 

But there’s one wrinkle this year, Cohen says. Even though Siberian snow advanced much more slowly than normal this October, its overall extent was substantially above normal. Such a mixed signal makes this year’s AO forecast more uncertain, Cohen stresses.

 

QUOTE FROM  COHEN

 

 This October was highly anomalous in that the snow cover extent [sCE] and the snow advance index [sAI] strongly diverged. The SCE was well above normal, which is a robust prediction of a negative winter AO while the SAI was well below normal, which is a robust prediction of a positive winter AO. This has never happened over the forty years that we have calculated both indices. I view the SAI as the more reliable predictor but I also think it would be a mistake to dismiss the high Siberian SCE from this past October. How the two predictions evolve this winter should be interesting.

 

 

This implied to me that Dave is debating the AO because of what COHEN said. Thus a warmer winter nor a +AO is a given from what Cohen is suggesting. Per OPI the winter AO will be very positive from what i have gathered as well as DT. Per COHEN there is some doubts vs this OPI which says it will be a +AO and thus where the debate is at.

 

Me? I'll take a wait and see approach.

[Harry]

A negative AO will help those NYC south,look at some recent winters where they got colder more snowier years than us further north , then look at 2007-2008 where  those N of NYC got big snows while south of 40 had single digit snowfall totals!

 

 

Yep.. Personally for my backyard i'll take my chances with a neutral AO and or between -1.000 and +1.000! Best winters ( snowiest ) in these parts tend to happen when the AO is in that range for the month.

[skierinvermont]

Is Cohen actually correct in saying SCE was anomalously high?

 

He does point out that SAI is the more reliable indicator - indicating a +AO. Saying the year will be 'interesting' is not the same thing as saying a +AO is not the most probable outcome. A +AO is the most probable outcome and I believe Cohen would agree with that statement. His work certainly supports it.

[ORH_wxman]

Is Cohen actually correct in saying SCE was anomalously high?

 

He does point out that SAI is the more reliable indicator - indicating a +AO. Saying the year will be 'interesting' is not the same thing as saying a +AO is not the most probable outcome. A +AO is the most probable outcome and I believe Cohen would agree with that statement. His work certainly supports it.

 

 

Yes, this October was ranked 4th for Eurasian snow cover out of 46 years on the Rutgers site.

[skierinvermont]

Yes, this October was ranked 4th for Eurasian snow cover out of 46 years on the Rutgers site.

 

Ah yes I remember that discussion. 

[The_Global_Warmer]

I thought it came in 7th. That might extent and not area.

[ORH_wxman]

I thought it came in 7th. That might extent and not area.

 

Northern Hemisphere came in 7th, the Cohen studies are only concerned with Eurasia...which came in 4th.

[The_Global_Warmer]

Oh I gotcha.

My guess is the snow fell to earlier to have a large enough impact.

[Mallow]

He does point out that SAI is the more reliable indicator - indicating a +AO. Saying the year will be 'interesting' is not the same thing as saying a +AO is not the most probable outcome. A +AO is the most probable outcome and I believe Cohen would agree with that statement. His work certainly supports it.

 

Correct.

[donsutherland1]

FYI, in his winter forecast, Glenn Schwartz referenced the OPI.

 

http://www.nbcphiladelphia.com/weather/stories/Hurricanes-Long-Range-Winter-Forecast-232835491.html

 

He also makes a statement that I, for one, find refreshing and consistent with the best of scientific pursuit of knowledge that is, IMO, worth repeating here:

 

But my interest is always to explore new research and test it myself, published or not.

[Absolute Humidity]

FYI, in his winter forecast, Glenn Schwartz referenced the OPI.

 

http://www.nbcphiladelphia.com/weather/stories/Hurricanes-Long-Range-Winter-Forecast-232835491.html

 

He also makes a statement that I, for one, find refreshing and consistent with the best of scientific pursuit of knowledge that is, IMO, worth repeating here:

 

But my interest is always to explore new research and test it myself, published or not.

Some nice kudos from Glenn to Ricardo and OPI. Good stuff, thanks for posting this.

[Crocodile23]

So far, and in the near future as the NOAA prediction says, December will be with big positive AO so the OPI prediction so far is right on track.

A disgusting situation in the first 20 days of December for European weather, with a colossal ridge and anticyclone covering everything for weeks.

[The_Global_Warmer]

Time for an update on how the OPI, Al Marino proprietary NAO forecasting method, and Cohen's method are doing.  I can't post them all at one time and I don't want to lose the post.  I am starting with the NAO and will get to the rest as soon as I can.  I can say all of them appear to be a bit off.  But one of them is horribly wrong.

 

 

First the NAO prediction called the TRIPOLE METHOD by Al Marino:

 

Prediction for DJF: -0.77.

 

December: +0.95

January: +0.29

NAO index through January: +0.64

 

 

So far this method is actually on pace to bust harder than any other year in it's history and it's not even close.

 

Through January the prediction was off by 1.41.

 

On top of that the NAO is averaging almost right at +1.0 for the month of February through the 18th.  Forecasts call for it to stay positive the rest of the month.  Mostly between +.5 and +1.0.

 

I am not sure what to make of it.  I have said before that my gut told me this was a big scam.  However I do not know Al Marino and I do not want to call him a fraud without an explanation from someone or Al himself about how this is possible after 33 years of amazing accuracy a bust at least twice as bad as the nearest "wrong" forecast.

 

 

Math guys what are the odds or probabilities of this given what the data set showed over the last 33 years?

 

 

 

 

 

[donsutherland1]

Time for an update on how the OPI, Al Marino proprietary NAO forecasting method, and Cohen's method are doing.  I can't post them all at one time and I don't want to lose the post.  I am starting with the NAO and will get to the rest as soon as I can.  I can say all of them appear to be a bit off.  But one of them is horribly wrong.

 

 

First the NAO prediction called the TRIPOLE METHOD by Al Marino:

 

Prediction for DJF: -0.77.

 

December: +0.95

January: +0.29

NAO index through January: +0.64

 

 

So far this method is actually on pace to bust harder than any other year in it's history and it's not even close.

 

Through January the prediction was off by 1.41...

 

 

 

 

 

Some thoughts:

 

1. The 1950-79 data should be included to broaden the sample. If the relationship is not normally distributed, the existing sample size might be too small.

 

2. As I don't know the indicators that were used, it's difficult to ascertain whether, in fact, there is a curvilinear relationship between the independent variables and dependent variable, much less one that is reasonably reliable. If not, then a polynomic equation won't really provide a useful prediction over the long-term.

 

Looking closely at the two points above, one will be in a better position to determine whether the apparent large error this year is a relatively infrequent error or if there is a systemic problem with the model that was developed. By itself, the magnitude of the error merits a close look at what happened.

 

Also, beyond those two points, given that one indicator was selected from September, one must wonder whether the ongoing general but sometimes irregular decline in summer  Arctic sea ice may also be skewing the results. Notice, for example, in 2008 when the summer Arctic sea ice extent staged a partial recovery from the 2007 minimum, the tool was also wrong. In 2009, when the minimum was still higher, even as the tool got the right "sign," the difference was greater than 1.

 

Finally, I'm not sure that the tool uses atmospheric variables. Hence, one may be oversimplifying if one assumes largely or solely an SSTA-NAO relationship. More than likely, there's a top-down and bottom-up relationship where the atmosphere impacts SSTAs and SSTAs influence the atmosphere. The coupled ocean-atmosphere relationship is complex, so one might need to include some atmospheric variables, if they are not already incorporated in the tool.

[ORH_wxman]

Through the end of January, the average AO was about +0.25. At the moment, February looks to finish near neutral or slightly positive....therefore, the average winter AO will likely be close to the +0.25 through the end of January. We'll know the final numbers in about 10 days.

[The_Global_Warmer]

Adding up the daily AO numbers on the CPC site so far Feb is roughly 0.31+.

 

If it finishes the month around that.  The DJF period will be roughly 0.29+.

 

Which would also be the largest bust in OPI history with it coming in predicting a 1.64 AO+.

 

Cohen wouldn't be as bad and he would still have March.

 

Not sure about the NAO.  But the AO is really close to being in the right spot for that major +AO to verify. 

 

 

 

 

[ORH_wxman]

 

Through the end of January, the average AO was about +0.25. At the moment, February looks to finish near neutral or slightly positive....therefore, the average winter AO will likely be close to the +0.25 through the end of January. We'll know the final numbers in about 10 days.

 

The CPC has updated its figures and the final February AO was +0.044.

Weighting the months properly since February only has 28 days, the final DJF AO was +0.1885.

[Ginx snewx]

there is no IF. It's obviously solid. Wow, that really sucks. The pacific will have to drop the hammer on us for there not to be a pretty warm winter if we see 1.5 AO+ or so. Damn.

Last time the OPI predicted the AO to be this high. We saw this:

2

this was very off,predicted 1.6 final .2

[ORH_wxman]

this was very off,predicted 1.6 final .4

 

 

Final winter AO was actually 0.188 or near +0.20. Very close to neutral.

 

 

It will be interesting to investigate why this method failed this year.