And we begin

[HailMan06]

Didn't someone do pixel counting in past years?

I think it was Friv (TheGlobalWarmer) who did that.

[michsnowfreak]

Well, it will be near impossible to top 2013-14 here, not only was it Detroits snowiest winter on record but snow depth records were also smashed....but the excitement of a coming winter will never grow old..plenty of snow for all

 

[Quixotic1]

Didn't someone do pixel counting in past years? 

Not sure who it was.  Think it was Frivolous123 or however the number went. 

[sebastiaan1973]

From the Belgium weather forum weerwoord.be

[The_Global_Warmer]

I think it was Friv (TheGlobalWarmer) who did that.

Wasn't me.

 

 

It looks like October will be off to a slow start as of now.  But IIRC that is what we want then it to ramp up thru October.

 

[Eskimo Joe]

Some nice gains on the other side of the globe the past few days.

 

[radarman]

Some nice gains on the other side of the globe the past few days.

 

Day268-2014.png

 

Winter cancel?

 

Seriously, this is why I think the SAI is strange though it continues to prove me wrong for doubting it.  Snows over there the next few days will hurt the index by bringing up the start point.  I can understand that it might cause the eventual EP standing waves to be out of phase with some other forcing that seems more driven by the calendar,but it still feels unnatural to be bummed out by this, even if we should be.  One day I'll say 'no mas' and call the SAI my daddy.

[Sunny and Warm]

Winter cancel?

 

Seriously, this is why I think the SAI is strange though it continues to prove me wrong for doubting it.  Snows over there the next few days will hurt the index by bringing up the start point.  I can understand that it might cause the eventual EP standing waves to be out of phase with some other forcing that seems more driven by the calendar,but it still feels unnatural to be bummed out by this, even if we should be.  One day I'll say 'no mas' and call the SAI my daddy.

Wasn't the SAI predictive quality out to lunch last year?

[radarman]

Wasn't the SAI predictive quality out to lunch last year?

 

No, I don't believe so.  In fact the very slow advance, particularly in the latter half of October, predicted a +AO which verified. 

[NEchestercountydude]

Isn't it fine if we build snow pack around the poles now, and starting in October we want it to build south of 60N?

[ORH_wxman]

No, I don't believe so.  In fact the very slow advance, particularly in the latter half of October, predicted a +AO which verified. 

 

 

It got the sign right, however, it overshot the magnitude of the AO though by quite a lot...we ended up with a slightly positive AO and it was predicting a very robust +AO.

 

It is almost as if a compromise between the Snow Advance Index (SAI) and the old snow cover extent metric (SCE) was the best predictor last winter.

 

 

There was actually a good article by capital weather gang where they interviewed Cohen before the winter and he talked about the two conflicting signals:

 

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

[radarman]

Isn't it fine if we build snow pack around the poles now, and starting in October we want it to build south of 60N?

 

That is correct, it's 25-60N, 0-180E  Some of the recent snow is south of 60N, but I am being a bit dramatic of course, I don't really think it spells doom, just find it interesting that in theory the snows may be a negative sign for winter.

[NEchestercountydude]

It got the sign right, however, it overshot the magnitude of the AO though by quite a lot...we ended up with a slightly positive AO and it was predicting a very robust +AO.

 

It is almost as if a compromise between the Snow Advance Index (SAI) and the old snow cover extent metric (SCE) was the best predictor last winter.

 

 

There was actually a good article by capital weather gang where they interviewed Cohen before the winter and he talked about the two conflicting signals:

 

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

 

After reading that it sounds like a compromise of maybe 60/40 in favor of SAI is the best way to predict the dominate phase of the AO in DJFM

[radarman]

It got the sign right, however, it overshot the magnitude of the AO though by quite a lot...we ended up with a slightly positive AO and it was predicting a very robust +AO.

 

It is almost as if a compromise between the Snow Advance Index (SAI) and the old snow cover extent metric (SCE) was the best predictor last winter.

 

 

There was actually a good article by capital weather gang where they interviewed Cohen before the winter and he talked about the two conflicting signals:

 

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

 

I'm glad he has his own doubts... makes me feel better. 

 

In that article he mentioned that he was expecting early winter to have the more robust snowcover response, and that the AO would increase later on... again he was half right... November was sky high, December lower, but still very high, and Jan pretty negative before it started going up again at a more modest pace. 

 

To me this indicates that the early snow does not simply push up the EP flux-> SSW -> downward propagation schedule... More like earlier EP flux response is out of phase with climo driven waves, causing a weaker response in the months of late Dec/Jan/Feb/early Mar.  I'm not entirely convinced that November AO especially, even the majority of Dec AO have anything to do with the snowcover signal. 

 

Let's say the background signal for last winter favored the strong vortex, per the solar max and the QBO, Judah may have actually been right if the signal started taking affect near the end of the year, and not late Nov, early Dec.  You get a steep drop early from the SCE signal followed by a rise from the SAI signal.

[ORH_wxman]

After reading that it sounds like a compromise of maybe 60/40 in favor of SAI is the best way to predict the dominate phase of the AO in DJFM

 

 

Yeah perhaps...though it is probably the type of thing where you don't normally need to compromise so much. He mentioned how rare it was to get such strong conflicting signals from the two metrics (he said he had never seen it that opposing in the 40 years of data). Usually they tend to point somewhat in the same direction. SAI is just better fine-tuned apparently.

 

It is still a work in progress though. You have to remember that the SAI only has daily snow cover data since 1997, so the sample of years isn't huge before you have to lower the resolution to weekly data in the first 25 years of measurements.

[NEchestercountydude]

Yeah perhaps...though it is probably the type of thing where you don't normally need to compromise so much. He mentioned how rare it was to get such strong conflicting signals from the two metrics (he said he had never seen it that opposing in the 40 years of data). Usually they tend to point somewhat in the same direction. SAI is just better fine-tuned apparently.

 

It is still a work in progress though. You have to remember that the SAI only has daily snow cover data since 1997, so the sample of years isn't huge before you have to lower the resolution to weekly data in the first 25 years of measurements.

 

The thing I cant rap my head around is why SAI is more important then than SCE. Is the rate of increase is important because its showing us the upcoming pattern of winter or something else? Now like you said the two signals rarely conflict each other, but when they do (last year) why exactly is SAI more accurate?

[griteater]

The thing I cant rap my head around is why SAI is more important then than SCE. Is the rate of increase is important because its showing us the upcoming pattern of winter or something else? Now like you said the two signals rarely conflict each other, but when they do (last year) why exactly is SAI more accurate?

 

From the Cohen and Jones SAI paper

 

"An important question that we have not answered is why the October SAI (Snow Advance Index) is more highly correlated with the DJF AO than the October SCE (Snow Cover Extent) index. One likely reason is that the SAI is limited to latitudes equatorward of 60°N while the SCE index includes all of Eurasia, which has a significant amount of snow cover north of 60°N. Assuming that the high albedo of snow cover is one if not the most important snow characteristic that influences the overlying atmosphere, this would favor the SAI, which is limited to regions that are exposed to a higher sun angle more so than the SCE, in predicting the atmospheric response to snow cover variability.  Another possibility is that the SAI is sensitive to the timing of snowfall, where snowfall at the end of the month contributes to higher values of the SAI and snowfall at the beginning of the month contributes to lower SAI values while the monthly‐mean SCE is insensitive to the timing of snowfall. A rigorous answer to the question is beyond the scope of this concise Letter and will require more in‐depth analysis."

[NEchestercountydude]

from millwx in the Mid Atlantic sub forum, Incredible write up on the topic.

 

 

" If you believe some of the previous discussions on this board, this is irrelevant.  And, in fact, it may actually be a bad thing.  Why?  Because it's TOO soon.  It is only the gains in October that matter.  I've even heard some (attempt to) attribute some physical processes as to why late September snowfall doesn't matter.

 

Frankly, it's bull, and you are correct to point it out.  This snow cover is a good thing.  I've run the numbers myself.  I admit, I'm kind of a data nerd, lol.  But I just get skeptical trusting other's analyses when we have all the data we need on hand.  Snow data is readily available online, as is the AO data.  I've run it all myself...

 

It is true, as the SAI promoters suggest, that the highest correlation between Eurasian snow cover and the AO is on the advance of the snow cover, NOT on an absolute value.  However, that correlation (-0.5472) occurs on the delta between Week 43 and Week 39.  Week 39 is 9/23 - 9/30.  So, the argument that last September snowfall is "too soon" or irrelevant is just total b.s.  Mother Nature doesn't look at a calendar and flip a switch on October 1st.  And the numbers hold to that.  Mind you, the correlation using the delta between Week 43 and 40 (Week 40 starts at 10/1) isn't much lower (-0.5217), but it IS lower.  This late September snow advance is relevant and it is good.  Arguably, since we're IN week 39 and looking at the delta, maybe we don't want Week 39 to be high yet (don't want to subtract off a large number), but just AFTER... which could imply that what I'm saying is wrong and, in fact, we don't want to start watching until Week 40.  BUT... I've done the correlations on all of the week pairings, and I can tell you that even Week 43 minus Week 38 (9/16 - 9/22) has a higher correlation (-0.5373) than Week 43-40.  Bottom line... late September matters... unquestionably.

 

Moreover, while I said that the correlations support the SAI proponents, I'm not totally sold on the increase being most important anyway.  The correlation between AO and just the total snow cover in a given week (not the increase) barely comes in under the peak increase correlation.  The top correlation to raw snow cover values is -0.5259 (in Week 43 ...not coincidentally, I'm sure, the end-week on the peak increase correlation... indicating that the first week that you're subtracting off is of almost negligible importance).

 

Furthermore, the AO is not a perfect harbinger of U.S. winter temps anyway.  The correlation between Eurasian snow cover and population-weighted U.S. temperatures (which is what we use in my industry... but the population in the U.S. leans heavily to the East... so, this is VERY relevant to this board) actually does NOT peak on a snow increase value.  It peaks on an absolute snow cover value.  And, in fact, it's not even remotely close.  The highest correlation you get between U.S. winter temps and Eurasian snow increase is a putrid +0.1673 (correlation is positive because it's on U.S. heating degree days, not temperature)... that's on Week 43 minus Week 37.  But whatever... that's a pretty poor correlation.  The BEST correlation is on the absolute snow cover value... NOT on the increase!  The top correlation (still modest, lower than the AO correlation itself, but MUCH better than the HDD correlation with the snow increase), is +0.3024.  That is the correlation of the U.S. winter temps with Week 42 (10/15 - 10/21).

 

So, for U.S. (mainly Eastern U.S.) temperatures what matters most in Eurasian snow cover is the absolute amount by mid or mid-to-late October.  It's really as simple as that.  Sometimes I think we over-complicate matters.  If Eurasian snow cover is high in mid-October, that's a positive sign for a cold winter (not a guarantee... a 0.3 correlation is far, far from a home run... just a good indicator).  Simple as that.  (And, as an aside, if the GFS ends up even in the rough ballpark of correctness, snow cover by mid-Oct should be above normal... but that's pretty far out there... and it's the GFS... so, we'll see.) "

[ORH_wxman]

from millwx in the Mid Atlantic sub forum, Incredible write up on the topic.

 

 

 

 

Yeah I just read that...good stuff. Though it is important to point out that two slightly different metrics are being discussed...he is discussing the population-weighted temperatures of the U.S. (biased to the east) while the SAI/SCE predicts the AO for the winter...as we know, the AO and US temps (esp eastern US) are highly correlated, but not the same thing.  

[griteater]

With respect to the SAI, there was discussion last year regarding changing the start and end date, and it was shown to not have an effect on the SAI slope

 

http://www.americanwx.com/bb/index.php/topic/41113-and-we-begin/?p=2456684

 

Except the SAI is calculated using a linear regression of the snow cover each day of the month of October. Changing a few of the values at the end and beginning of the period is likely not going to have much of an impact on the regression coefficient. It makes sense to use a regression coefficient for this reason because it doesn't weight the snow cover values at the beginning or the end of the month more than any of the intermediate values.

[NEchestercountydude]

Yeah I just read that...good stuff. Though it is important to point out that two slightly different metrics are being discussed...he is discussing the population-weighted temperatures of the U.S. (biased to the east) while the SAI/SCE predicts the AO for the winter...as we know, the AO and US temps (esp eastern US) are highly correlated, but not the same thing.  

 

Yeah you're right that his finding are more about the affects in our area. Americanwx is probably 80-90% people east of the Mississippi, so it pertains to a large portion of the board. Now I wouldn't want our fellow cold and snow lovers across the pond to read it and assume the same for them, but he nicely states that his findings are more for the eastern part of the US.

[millwx]

Yeah I just read that...good stuff. Though it is important to point out that two slightly different metrics are being discussed...he is discussing the population-weighted temperatures of the U.S. (biased to the east) while the SAI/SCE predicts the AO for the winter...as we know, the AO and US temps (esp eastern US) are highly correlated, but not the same thing.  

 

What I found most interesting (I didn't know what the results would be until I ran them... I came at this with no prejudice or expectations... well, except to find SOME link between snow cover and AO... and temps), is that while the snow increase (SAI) *IS* the most predictive for the AO (albeit barely... the raw extent numbers are nearly as good), it is much WORSE (than the raw extent numbers) when correlating to what ultimately matters... temperatures (Eastern U.S. temps in particularly).

 

I'll be blunt and say I have no idea why.  AO isn't a perfect correlator to temps (nothing is), but it's pretty good.  I'd have anticipated a similar signal in the AO correlations as in the temperature correlation.  That doesn't exist.  At all.  If you want to know what the AO is gonna be... you probably won't go too far wrong using snow extent, but you'd be better off using the SAI.

 

...but if you want to know what the temperature anomalies are gonna be, ditch the SAI.  Just go with the mid-October snow extent.  Much simpler and MUCH more accurate (correlation is twice as high... though still not tremendously high... everyone should remember, this is just one piece in a big puzzle; Eurasian snow isn't the magic key to the winter forecast... 2006 had very high Eurasian snow cover and the U.S. winter torched).

 

That was my biggest surprise... the change in snow cover (I'm not looking precisely at the SAI, just the Eurasian weekly snow cover from Rutgers) links strongly with the AO, but NOT with Eastern U.S. temperatures.  The raw snow cover itself links much better with the Eastern U.S. temperatures.  I have some theories on why that is.  I don't want to get into it too much here, because I may well be wrong, as I'm just scratching the surface with it.  But, my suspicion is that the Eurasian snow cover is either related to OR impacts things other than just the AO (my suspicion is the Pacific, based on some anecdotal observations).  So... because it is a complex interaction, this is not simply a "snow => AO => temperature" chain of events.  There are other snow-induced or snow-related influences that are complicating this and not making for a clean relationship.  Again, that's my theory on why the odd correlation disconnect occurs.

 

But whatever the reason, if we just follow the numbers, and if we're just interested in the bottom line ("what are the temps gonna be?"), then I'd just stick to the simple snow extent.

[radarman]

What I found most interesting (I didn't know what the results would be until I ran them... I came at this with no prejudice or expectations... well, except to find SOME link between snow cover and AO... and temps), is that the snow increase (SAI) *IS* the most predictive for the AO (albeit barely... the raw extent numbers are nearly as good), it is much WORSE (than the raw extent numbers) when correlating to what ultimately matters... temperatures (Eastern U.S. temps in particularly).

 

I'll be blunt and say I have no idea why.  AO isn't a perfect correlator to temps (nothing is), but it's pretty good.  I'd have anticipated a similar signal in the AO correlations as in the temperature correlation.  That doesn't exist.  At all.  If you want to know what the AO is gonna be... you probably won't go too far wrong using snow extent, but you'd be better off using the SAI.

 

...but if you want to know what the temperature anomalies are gonna be, ditch the SAI.  Just go with the mid-October snow extent.  Much simpler and MUCH more accurate (correlation is twice as high... though still not tremendously high... everyone should remember, this is just one piece in a big puzzle; Eurasian snow isn't the magic key to the winter forecast... 2006 had very high Eurasian snow cover and the U.S. winter torched).

 

That was my biggest surprise... the change in snow cover (I'm not looking precisely at the SAI, just the Eurasian weekly snow cover from Rutgers) links strongly with the AO, but NOT with Eastern U.S. temperatures.  The raw snow cover itself links much better with the Eastern U.S. temperatures.  I have some theories on why that is.  I don't want to get into it too much here, because I may well be wrong, as I'm just scratching the surface with it.  But, my suspicion is that the Eurasian snow cover is either related to OR impacts things other than just the AO (my suspicion is the Pacific, based on some anecdotal observations).  So... because it is a complex interaction, this is not simply a "snow => AO => temperature" chain of events.  There are other snow-induced or snow-related influences that are complicating this and not making for a clean relationship.  Again, that's my theory on why the odd correlation disconnect occurs.

 

But whatever the reason, if we just follow the numbers, and if we're just interested in the bottom line ("what are the temps gonna be?"), then I'd just stick to the simple snow extent.

 

Excellent posts today...

 

I actually consider 06/07 to be a classic success of the methodology.  Granted it didn't correspond well to the DJF index as a whole, but the massive SSW that occurred, the downward propagation from the stratosphere the troposphere, and the 40+ day stretch from the latter half of Jan through Feb is, in my opinion, the prototypical example of the snow cover response.  It was delayed a bit... One might have expected the SSW to have occurred a bit sooner, but as I mentioned earlier I think you can probably toss the correlation for snowcover and the AO in the first 20 days of Dec.  I have not put the time into calculating it, you may be able to confirm or deny based on your data, just a gut feeling in combination with how I believe the mechanism works.

 

Look at the EP flux (30 day av) from 11/30 and 12/30 in 2006.

 

 

By far and away the vertical component was the strongest that year that we've had since, and IMO made the pattern ripe for the flip.   08/09 and 09/10 were both strong as well, and both also featured February as their most negative AO month of winter.  Feb 09/10 was obviously off the charts negative, and what's notable about those months was not the magnitude of the max vals of vertical EP flux in Nov. and Dec.,  but their extent through the vertical column.  And I think it's fair to say, there were some other factors at work here, presumably a boost from the strongly negative QBO and the El Nino.

 

 

[ORH_wxman]

What I found most interesting (I didn't know what the results would be until I ran them... I came at this with no prejudice or expectations... well, except to find SOME link between snow cover and AO... and temps), is that while the snow increase (SAI) *IS* the most predictive for the AO (albeit barely... the raw extent numbers are nearly as good), it is much WORSE (than the raw extent numbers) when correlating to what ultimately matters... temperatures (Eastern U.S. temps in particularly).

 

I'll be blunt and say I have no idea why.  AO isn't a perfect correlator to temps (nothing is), but it's pretty good.  I'd have anticipated a similar signal in the AO correlations as in the temperature correlation.  That doesn't exist.  At all.  If you want to know what the AO is gonna be... you probably won't go too far wrong using snow extent, but you'd be better off using the SAI.

 

...but if you want to know what the temperature anomalies are gonna be, ditch the SAI.  Just go with the mid-October snow extent.  Much simpler and MUCH more accurate (correlation is twice as high... though still not tremendously high... everyone should remember, this is just one piece in a big puzzle; Eurasian snow isn't the magic key to the winter forecast... 2006 had very high Eurasian snow cover and the U.S. winter torched).

 

That was my biggest surprise... the change in snow cover (I'm not looking precisely at the SAI, just the Eurasian weekly snow cover from Rutgers) links strongly with the AO, but NOT with Eastern U.S. temperatures.  The raw snow cover itself links much better with the Eastern U.S. temperatures.  I have some theories on why that is.  I don't want to get into it too much here, because I may well be wrong, as I'm just scratching the surface with it.  But, my suspicion is that the Eurasian snow cover is either related to OR impacts things other than just the AO (my suspicion is the Pacific, based on some anecdotal observations).  So... because it is a complex interaction, this is not simply a "snow => AO => temperature" chain of events.  There are other snow-induced or snow-related influences that are complicating this and not making for a clean relationship.  Again, that's my theory on why the odd correlation disconnect occurs.

 

But whatever the reason, if we just follow the numbers, and if we're just interested in the bottom line ("what are the temps gonna be?"), then I'd just stick to the simple snow extent.

 

 

It might be interesting to see if there is some sort of response in the EPO WRT autumn snow cover...the EPO does share some domain with the AO, but not nearly as much as the NAO shares domain with the AO. Yet, the EPO is just as highly correlated to temperature in the populated areas of the midwest/lakes/Northeast as the AO...even a bit moreso than the AO in the northern belt of cities like a MKE-DET-BUF-BOS line.

 

Last year, we had a disconnect between the two where a slightly positive AO verified versus a highly negative EPO. (I think now NOAA uses a new metric called EP/NP which is basically the same thing, but the sign is reversed...makes it a bit confusing for those of us to still talk about the original EPO)

[radarman]

To add on, the two most strongly positive Feb's we've had in the last decade were 2011 and 2008, both preceded by weaker than average EP flux in Nov and Dec.  

 

It should be noted however, that with respect to snowcover, EP flux doesn't always match up well either... As I mentioned Nov. 2008 and Dec 2008 featured strong flux, but pretty lousy Oct. snowcover.

[hosj III]

Some serious gains south of 60 degrees, especially in Mongolia and Quebec

 

[NJwinter23]

Wasn't the SAI predictive quality out to lunch last year?

 

 

It got the sign right, however, it overshot the magnitude of the AO though by quite a lot...we ended up with a slightly positive AO and it was predicting a very robust +AO.

 

It is almost as if a compromise between the Snow Advance Index (SAI) and the old snow cover extent metric (SCE) was the best predictor last winter.

 

 

There was actually a good article by capital weather gang where they interviewed Cohen before the winter and he talked about the two conflicting signals:

 

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

 

Yeah Dec-Feb combined it ended up being pretty out to lunch, but if we break it down, it was really the jan-feb period that the AO went negative enough to throw off the index.

 

Following the SAI reading that argued for a ++AO last year, I'd argue that we saw a very good response up through the trop-strat in Novy and Dec (which combined had one of the highest +AO reading since 1950) The Pacific pattern had other ideas for sensible temps in the East obviously. 

 

Anomalously cold stratosphere/strong Polar Vortex was coupled well with the troposphere for the same period and argued for the ++AO signal

[Absolute Humidity]

Happy October everyone. May the Siberian snowfall express begin.

This doesn't look half bad,almost right where we want it, above 60N

[weathafella]

Happy October everyone. May the Siberian snowfall express begin.

This doesn't look half bad,almost right where we want it, above 60N

Today

[Sunny and Warm]

plenty cold enough.  This is for north of 80 degrees, so more often than not when you have temps on this graph above normal, it means the cold has been dispersed to lower latitudes temporarily.