And we begin

[griteater]

The pattern turns terrible for snow advance as we go into November....but who needs it then?  perfect timing, lol

[WidreMann]

Anyone else concerned by the paucity of snow in NA? Usually we're filling up south of 60N, at least in Western Canada. It's great if we get a -AO and all, but if Canada is warm and snowless, it won't mean crap for the US.

[millwx]

The pattern turns terrible for snow advance as we go into November....but who needs it then?  perfect timing, lol

 

You actually DON'T want it then.  You want snow in Oct for the strat chain reaction.  The actual pattern that creates Eurasian snow is not one that's particularly favorable for a cold/snowy U.S.  In fact, the correlation actually REVERSES sign (albeit barely and negligibly) by as early as the end of November.

 

So, you're right, perfect timing.  We have no interest in seeing snow continue to inundate Eurasia in November.

[Bob Chill]

You actually DON'T want it then. You want snow in Oct for the strat chain reaction. The actual pattern that creates Eurasian snow is not one that's particularly favorable for a cold/snowy U.S. In fact, the correlation actually REVERSES sign (albeit barely and negligibly) by as early as the end of November.

So, you're right, perfect timing. We have no interest in seeing snow continue to inundate Eurasia in November.

I'm glad you posted this. I was thinking the same thing but just an inexperienced hunch. We've had several years in the last 10 or so where the coldest temps remained locked in the wrong hemisphere so even when we got blocking it really wasn't that cold here. Our source regions didn't have much to deliver.

Euro weeklies and even some hints on the global ens are starting to show ridging building in ak and the goa trough relaxing and or retrograding down the line.

If the pattern evolves that way through Nov and we get some blocking building then that's a fairly decent pattern to build some cold in nw and central Canada and also deliver it southward.

This is the wrong thread to discuss that but it would be nice to see things set up so our source regions can build up. Long ways to go before met winter. This thread has hit a home run in both analysis and outcome. It's about time to start looking at other things that can help get a good pattern going in the conus.

[millwx]

GaWx (and anyone else who may be interested),

 

Ok... I did the week re-alignments.  I still thinks it tricky/noisy/dirty, but I decided it was necessary as some years (like 1988) are a full week off (the week happens to fall at the end of the range, PLUS leap year).  As we can see in the daily plots, a few days can matter... never mind a full WEEK.  So, this re-alignment is really necessary.

 

It DOES, as we might expect - and is nice to see - add an increased focus.  All October weeks see an increase in the correlation, but the peak correlation got a greater increase than the surrounding ones.  Note, the improvement was better when correlating snow cover to U.S. HDDs rather than snow cover to winter AO... but the AO correlation did see a minor increase as well.

 

As for where it peaks, as mentioned, the focus gets even stronger.  The week beginning 10/18 (you can pick start weeks on any date you want, but for ease of data handling 9/20 was a good start date... and I just went from there, defining start weeks of 9/27, 10/4, 10/11, 10/18, etc)... anyway, as I was saying... the week beginning 10/18 has the peak correlation with U.S. winter HDDs (at +0.42).  The week beginning 10/11 does a bit better (0.31) than the week beginning 10/25 (0.26), implying that the optimal week-beginning data could actually be 10/17 or 10/16 or thereabout.  Also, it is not explained on the Rutgers site, but from my own analysis of their data, looking at past years, it LOOKS as though their weekly data is simply defined as a snapshot of the daily data at week's end.  My dates above give the date on the START of the week.  So, if you're looking for a singular peak date, it would be 10/24 (the last day on the week beginning 10/18) or perhaps a day or two earlier than that, as noted above.

 

It's also worth noting that the correlation utterly collapses in November.  The correlation for the week of 10/18-24 is +0.42.  For 10/25-10/31 it drops to +0.26.  For 11/1-11/7 it drops to +0.22.  For 11/8-11/14 it drops to +0.06.  For 11/15-11/21 it drops to -0.04.

 

Also note that, for reasons a little inexplicable to me, the AO correlation peaks a week later (week beginning 10/25 ...ending 10/31).  So, the rest of October isn't irrelevant.  However, the week before is ALMOST as strong (and the correlation TWO weeks prior is stronger than the correlation ONE week later)... indicating that the peak difference may really only be a few days... and could just be noise.

 

So, basically, for both the AO and temperature correlations the peak - for when Eurasian snow cover matters the most - can be approximated sometime around the 10/22 - 10/28 time frame.

[MN Transplant]

Anyone else concerned by the paucity of snow in NA? Usually we're filling up south of 60N, at least in Western Canada. It's great if we get a -AO and all, but if Canada is warm and snowless, it won't mean crap for the US.

 

Canada was warm and snowless in 09/10. 

[griteater]

nice work millwx...thanks for the analysis and post 

[das]

1) With this Nesdis update, there was a rise from 16.7 on 10/23 to 18.16 on 10/26, or +1.46. Per my earlier counting rectangle post, this jibes well. So, I believe this Nesdis update.

 

2) Based on my comparison of past Rutgers and Nesdis, it looks like Nesdis has a low bias vs. Rutgers in general. Six examples for late October suggesting this low bias:

 

2013 ~15.0 on Nesdis vs. ~16.3 on Rutgers

2012 ~16.4 on Nesdis vs. ~17.6 on Rutgers

2011 ~14.8 on Nesdis vs. ~15.9 on Rutgers

2010 ~13.0 on Nesdis vs. ~14.0 on Rutgers

2009 ~16.5 on Nesdis vs. ~18.3 on Rutgers

2006 ~13.1 on Nesdis vs. ~14.7 on Rutgers

 

 The ratios of Rutgers to Nesdis for these late Octobers average 1.09. Lowest was 1.07.

 

 However, for whatever reason, both 2008 and 2007 didn't follow this pattern as they were very close...i.e., ~1.00. Then again, those were two of the oldest years. Also, I was earlier thinking that the 19 msk barrier had likely been crossed on Rutgers. I'm guessing that there will be about a 7% increase this time to convert from Nesdis to Rutgers. So, 1.07 x 18.16 = 19.4. So, I'm educatedly guessing that Rutgers is currently ~19.4 msk for total Eurasian SCE (with ~19.0 on the low end). If that is a good estimate, that would mean that only 1976 was higher in late Oct. (~22 msk) as was just noted in this thread. Also, it would likely put the SAI among the elite top four or so.

 

Well, today was a crazy day. Let's see how these last few days of the month go.

 

******EDITED for careless errors in Nesdis/Rutgers comparisons as I had written them backward in most cases lol.

 

The reasons for the differences are largely due to resolution.  Rutgers uses 200km resolution (ICE uses 24km resolution) and NESDIS uses their native 4km resolution.  At 200km resolution, many features such as lakes, rivers, bays and peninsulas are not resolved and are simply coded as snow covered if 42% of the adjacent area is snow covered.  NESDIS also uses a 50% pixel coverage as their threshold, not 42%.  Rutgers and ICE have good reason for doing so, their mission is largely focused on comparative analysis so the data needs to be backwards compatible from before the remote sensing period.  The NESDIS data is far more accurate though.

[GaWx]

heavy_wx,
Would it be possible for you to temporarily take the 0-60N latitude constraint off so we can see what your GDAS map has for TOTAL Eurasian SCE? I see the great 9.19 sub-60 N, which I know is important (used for very important SAI) and jibes well with cfbaggett's latest 9.1 rough estimate. However, others including myself also follow total Eurasian because that's what the Rutgers data shows. TIA.

[GaWx]

27th image is out. Using the loop, it's easy to go back and forth between days to see the changes.

Steve,

Thanks for putting together this loop! It looks like today's update is going to end up as a very small net change from yesterday. I'm sticking to my idea of a max around 10/27-8 with only a rather modest net drop between now and 10/31 fwiw. We're looking golden!

[GaWx]

Cfbaggett,

Thanks as always for your updates. Based on your rough estimate of 9.1 (which is almost right on heavywx's 9.19), we are at 7.8 sub 60N increase. The next closest of the years on your graph as of 10/31 is the 7.4 of 2009. So, as long as sub 60N doesn't fall back below 8.7 as of 10/31, 2014 will be positioned at least as high as 2009 for the largest sub 60N increase on your graph.

[cfbaggett]

Cfbaggett

Thanks as always for your updates. Based on your rough estimate of 9.1 (which is almost right on heavywx's 9.19), we are at 7.8 sub 60N increase. The next closest of the years on your graph as of 10/31 is the 7.4 of 2009. So, as long as sub 60N doesn't fall back below 8.7 as of 10/31, 2014 will be positioned at least as high as 2009 for the largest sub 60N increase on your graph.

All sounds great to me. IMO, qualitatively speaking, I think the snowcover has already done its dirty work at this point. Going forward, I think we now need to be looking for the upward propagating wave activity into the stratosphere which will eventually kill the polar vortex and allow the negative AO to take hold. According to Cohen's et al. 2014 paper, we should be looking for anamalous low heights in the mid-troposphere over the snow pack, which forces an upstream ridge over the Kara and Barents Sea. This ridge constructively interferes with the climatological ridge there which allows for easy wave activity propagation into the stratosphere.

[GaWx]

1) millwx,

I just read your write up about the various weeks, including making calendar related adjustments. Very interesting! Thanks for doing that. Regarding the correlation of SCE increase for the month to date to -AO in DJF, I also see the best correlation to be very near the end of the month rather than to days near the end of week 42.

2) das,

Thanks for explaining the reasons for the Rutgers high bias vs. Nesdis. The lower resolution explanation makes perfect sense. Also, doing it so as to be backwards compatible makes sense. Although not as accurate as Nesdis, the updated Rutgers numbers will still be important so as to allow the most apples to apples comparison to its past years on its database. I'm still looking for something over 19 msk for total Eurasian on Rutgers for around now. The very high side of estimates would be nearing 20. Anything over 19 would mean 2nd highest to 1976.

*edited*

[millwx]

Just refined my analysis further to find the peak single day for the snow cover correlations.  A few caveats:

 

1) I do realize that, obviously, one singular day is not driving the snow cover effects; I'm merely breaking it down this finely because there is probably some point at which the signal is the strongest.  That's what I'm after.  Obviously, I recognize that many days before and after are relevant.  I'm just trying to find the focus of the effect.

 

2) I'm assuming, based on my previous analysis of the Rutgers data, that their weekly numbers are a week end daily snapshot.  If I'm wrong about this... if it's a weekly average... then my dates below are all 3-4 days too late.

 

3) I'm merely doing a linear interpolation to get the daily data for every year (which, since we're not crossing any peaks or valleys, is probably fine in a bulk analysis like this... but there could be some small effect).

 

Results:

 

Eurasian snowfall to U.S. winter temperatures correlation peaks on Eurasian snow cover for October 24th.

Eurasian snowfall to DJF AO signal peaks on Eurasian snow cover for October 29th.

The correlation to U.S. winter temperatures drops to zero on November 18th (i.e., Eurasian snowfall by mid-November is irrelevant)

...note that the correlation to the AO doesn't drop as severely by then, but still drops off significantly

 

Still not sure why the AO signal would lag a few days.  Maybe it's just noise in the data.  But, if anything, I'd have expected the other way around.  Whatever the case... I tend to focus on end results.  So, I'm more interested in the correlation to actual weather than to the AO.  So, my focus is on about the third week in October Eurasian snowfall.  But it's not unreasonable to include late October snow cover in your considerations.  Beyond that (November Eurasian snow cover)... no... that's worthless.

[GaWx]

Millwx,

The Rutgers weekly numbers look like they pretty much have to be week ending daily snapshot numbers.

[das]

Millwx,

The Rutgers weekly numbers look like they pretty much have to be week ending daily snapshot numbers.

 

Rutgers uses a weekly number based on the Monday map.

[millwx]

Rutgers uses a weekly number based on the Monday map.

 

...which, from looking at their dates, would, in fact, be a daily snapshot from week end.  (Notice their last update, Week 40, is 9/30 - 10/6... the Monday in that 7-day period is 10/6.)

 

Thanks for confirming this.

 

You too, GaWx.

 

Glad I got that right... you can ignore that caveat in my post.  

[GaWx]

Rutgers uses a weekly number based on the Monday map.

 

1) das,

 Just to clarify: When I said week ending, I meant end of the Rutgers week, not end of calendar week and not "weekend". So, I agree it is based on the Monday map since that's their end of week.

 

2) millwx,

 You're welcome. The kind of weird thing is that there are actually subtle differences when you compare the daily Rutgers map for the Monday and the weekly map. It is almost as if it the weekly is a smoothed version of Monday in some way. Regardless, we all agree that the Monday maps are essentially matches.

[RaleighWx]

We really couldn't have asked for a better pattern for Eurasian October SCE. Notice how

close the 500 mb pattern from Eurasia to the Arctic was this month compared to to 1976.

 

14.jpg

 

76.jpg

 

Screen shot 2014-10-28 at 1.02.09 PM.png

 

 

Nice post.

 

Nice post

[DTWXRISK]

NO one with a brain  should be 

 

Anyone else concerned by the paucity of snow in NA? Usually we're filling up south of 60N, at least in Western Canada. It's great if we get a -AO and all, but if Canada is warm and snowless, it won't mean crap for the US.

[hosj III]

GFS shows some gains in southern Canada through 84 hours: 

 

 

[bluewave]

Nice post

 

Thanks. It will be interesting to see the final number for October.

[Mike.Ventrice]

Updated graphic with trend lines now. I decided to increase the northern latitude to 65N from 60N to see the result. Anyway, there does look like an impressive upward trend thus far for the month of October for this year.

 

[hosj III]

Updated graphic with trend lines now. I decided to increase the northern latitude to 65N from 60N to see the result. Anyway, there does look like an impressive upward trend thus far for the month of October for this year.

 

 

That really doesn't seem right. There is much more snow right now than in any other year.

[Mike.Ventrice]

That really doesn't seem right. There is much more snow right now than in any other year.

 

This is just what is being initialized in the GFS op (for 2014) compared to the ERA-interim reanalysis for the other years. Only quick way to get a real-time running index with the lack of snow cover satellite data. The data is what it is.

[GaWx]

Mike,

So, are you saying the other years are not using the same data source as 2014? If so, I'm strongly suspecting that that is the problem. I'm thinking that the 2014 line, itself, could be fine (overall steepness of rise and daily moves all seem to jibe with other sources) while perhaps the other years are all too high. One red flag: why is 2014 starting near the bottom with 2009 instead of near the top with 2013?

[GaWx]

heavy_wx,

In case you missed my post, would you mind tallying up SCE for Eurasia without the constraint of 0-60N latitude? Some of us also follow total Eurasian SCE.

[heavy_wx]

heavy_wx,

In case you missed my post, would you mind tallying up SCE for Eurasia without the constraint of 0-60N latitude? Some of us also follow total Eurasian SCE.

 

Do you mean for 0-90N or 25-85N and 0-180E?

 

EDIT: For 28 October, the GDAS indicates snow cover was 17.9 msk for both of those domains.

[GaWx]

Do you mean for 0-90N or 25-85N and 0-180E?

 

EDIT: For 28 October, the GDAS indicates snow cover was 17.9 msk for both of those domains.

Thank you. Did it drop a few hundred thousand msk since yesterday?

[GaWx]

From weatherdude in outage thread about an hour ago:

"Further the NESDIS snow report server is back on line and I am quite certain that unless Rutger's has data backlog issues, they should have a new report for snow cover out after Midnight."

Keep hope alive.

Edit: Weatherdude, are you saying that the Natice data will be available? I thought NESDIS data has been available per posts of maps here. Are these different?