Jump to content

  • Member Statistics

    17,588
    Total Members
    7,904
    Most Online
    LopezElliana
    Newest Member
    LopezElliana
    Joined

And we begin.

weathafella

Recommended Posts

  • Replies 125
  • Created
  • Last Reply
WxUSAF

WxUSAF

Posted
Posted

 

Heres my guess... Between now and Saturday.

18z GFS generally agrees at least for Quebec.  Naso much Iowa and Minnesota. http://www.instantweathermaps.com/GFS-php/showmap-nh.php?run=2013102118&var=SNOWC_sfc&hour=102

 

It's also not fabulous for snowcover south of 60N in western Siberia. http://www.instantweathermaps.com/GFS-php/showmap-nh.php?run=2013102118&var=SNOWC_sfc&hour=228

Link to comment
Share on other sites
Wow

Wow

Posted
Posted

 Agreed. Ironically, the 2013 SAI is currently partially being hurt by there being so much snow near the end of Sep. and during the first few days of Oct. as that has caused the sloped line to start at almost 2 million sq. km. The other three years started at under 1/2 million sq. km. Had 2013 also started around 1/2 million sq. km, the 2013 current slope would have been much higher than it is and a good bit higher than 2007. Isn't it sort of counterintuitive that just starting the month with "too much" snow, via causing the SAI indicator to be lower than it otherwise would have been, will reduce the prospects for a good (-AO) winter? Does it make sense that above average late Sep. to very early Oct. snow hurts the SAI?

 

I wonder how much those slopes would change if you include the last week or so of Sept?

Link to comment
Share on other sites
Wow

Wow

Posted
Posted

 Agreed. Ironically, the 2013 SAI is currently partially being hurt by there being so much snow near the end of Sep. and during the first few days of Oct. as that has caused the sloped line to start at almost 2 million sq. km. The other three years started at under 1/2 million sq. km. Had 2013 also started around 1/2 million sq. km, the 2013 current slope would have been much higher than it is and a good bit higher than 2007. Isn't it sort of counterintuitive that just starting the month with "too much" snow, via causing the SAI indicator to be lower than it otherwise would have been, will reduce the prospects for a good (-AO) winter? Does it make sense that above average late Sep. to very early Oct. snow hurts the SAI?

 

Sept ended as 6th highest in snow coverage for the NH since 1966. What year was #1?  1972.  That winter featured a +A0, but overall temps were normal. Plus the snowstorm of Feb '73.

Link to comment
Share on other sites
Wow

Wow

Posted
Posted

Well I'm thinking the SAI connection to the AO is rather weak using an arbitrary time period as October.  I suppose it is the time period that resulted in the best correlation numbers when running a regressive model.  But I'd imagine it's going to lose it's effect when you have a situation where the snow advance is high in the last days of Sept rather than the first days of Oct.  The differences of the Oct SAI between these two scenarios would be significant.

 

There will always be a reliable correlation found for seasonal forecasting until it no longer correlates.  Just as Redskins games predicting presidential elections.

Link to comment
Share on other sites
HM

HM

Posted
Posted

Well I'm thinking the SAI connection to the AO is rather weak using an arbitrary time period as October.  I suppose it is the time period that resulted in the best correlation numbers when running a regressive model.  But I'd imagine it's going to lose it's effect when you have a situation where the snow advance is high in the last days of Sept rather than the first days of Oct.  The differences of the Oct SAI between these two scenarios would be significant.

 

There will always be a reliable correlation found for seasonal forecasting until it no longer correlates.  Just as Redskins games predicting presidential elections.

 

I mean, I see the point you are making and I do believe over time, r values do decrease usually in LR when they are very high to start. But, if you are going to bash the connection, you better supply us with something more than this. Believe me, I'm all about the bigger players and do believe the snow cover is a middle component here in cause/effect; but, it is a little crazy to call this connection "weak." The time period is not really arbitrary. October sees a transition from monsoonal flow/trough across E Asia to Mongolian High and the development of a full polar vortex. It is truly a transition month and Cohen simply measured how fast/slow we are transitioning with this metric. There is nothing wrong with that... 

Link to comment
Share on other sites
Snow_Miser

Snow_Miser

Posted
Posted

I mean, I see the point you are making and I do believe over time, r values do decrease usually in LR when they are very high to start. But, if you are going to bash the connection, you better supply us with something more than this. Believe me, I'm all about the bigger players and do believe the snow cover is a middle component here in cause/effect; but, it is a little crazy to call this connection "weak." The time period is not really arbitrary. October sees a transition from monsoonal flow/trough across E Asia to Mongolian High and the development of a full polar vortex. It is truly a transition month and Cohen simply measured how we fast/slow we are transitioning with this metric. There is nothing wrong with that... 

 

The October SAI correlation is good IMO. However, I can see Wow's point with regard to the SAI values for October being influenced by the rapid snow growth in late September. 

Link to comment
Share on other sites
HM

HM

Posted
Posted

The October SAI correlation is good IMO. However, I can see Wow's point with regard to the SAI values for October being influenced by the rapid snow growth in late September. 

 

Why? If September into early October is characterized by a warm-season regime and the rest of October sees a transition and then a full placement into a cold-season regime, why would late September be as big of a deal? The rapid onset of "winter" mid-late October will more often than not be correctly measured or implied by the advancement of the sustainable snow cover in this period.

Link to comment
Share on other sites
Mallow

Mallow

Posted
Posted

I'm with HM on this one. I'd add that until someone shows that a "better" correlation can be had by choosing a September start date for the SAI, any such speculation is just that--speculation.

 

I'd guess that it's probable there are start/end dates other than Oct 1st and Oct 31st that give higher correlation values, but the "better" start date could just as easily be Oct 5th as it could Sept 25th. The reason the dates were chosen by Cohen probably has to do with the fact that weekly snowcover data is available back to the 1970s, whereas daily snowcover data only goes back to the late 1990s or early 2000s. Moreover, his older snowcover index was based only on the month of October as well. So Dr. Cohen was probably just going for consistency.

Link to comment
Share on other sites
HM

HM

Posted
Posted

Yea I mean there are at least 3 posts in this thread explaining why October is an important time of the year and NOT an arbitrary period for the study. I'd hope people to at least read through the thread and/or the papers before they belittle the study in a post.

 

Yes, I agree. October also sees a transition from summer to winter in the form of "standing waves" which are unique to the NH. They, along with the polar vortex, become established this month. A lot of various things that work together have been simplified for us by Cohen with the SAI. Is it perfect? Nope... Are there larger-scale factors at play? Yes...

 

But in the end, that transition and full-scale beginning of the cold season that takes over mid to late October is definitely being picked up by the SAI.

Link to comment
Share on other sites
Wow

Wow

Posted
Posted

Yea I mean there are at least 3 posts in this thread explaining why October is an important time of the year and NOT an arbitrary period for the study. I'd hope people to at least read through the thread and/or the papers before they belittle the study in a post.

 

Ok, perhaps arbitrary is too strong of a word here.  Certainly October, typically the later half of it, is a crucial time for transitioning to the cold season in the Arctic - I think it's a good study... I just question this year's correlation since Oct began with an anomalously high snow cover value compared to recent years. I think this aspect will be modified in time as transition periods are unique to every year, where for some years the 31 day window will move a number of days in either direction.

 

But the study does make a great start in working away at a forecasting skill for something which most have deemed impossible to forecast.

 

Reliable correlations work great for long-term forecasting until they don't.   

Link to comment
Share on other sites
ORH_wxman

ORH_wxman

Posted
Posted

The October SAI correlation is good IMO. However, I can see Wow's point with regard to the SAI values for October being influenced by the rapid snow growth in late September. 

 

 

You would have to look at years where the SAI was reduced because of rapid snow growth in late September....and see if they behave like high SAI years.

Link to comment
Share on other sites
NJwinter23

NJwinter23

Posted
Posted

Ok, perhaps arbitrary is too strong of a word here.  Certainly October, typically the later half of it, is a crucial time for transitioning to the cold season in the Arctic - I think it's a good study... I just question this year's correlation since Oct began with an anomalously high snow cover value compared to recent years. I think this aspect will be modified in time as transition periods are unique to every year, where for some years the 31 day window will move a number of days in either direction.

 

But the study does make a great start in working away at a forecasting skill for something which most have deemed impossible to forecast.

 

Reliable correlations work great for long-term forecasting until they don't.   

 

 

The 0.86 correlation back to 1997 is probably too high, Cohen even admits it in one of his papers. But he also has weekly siberian snowfall data back to the 70s that he standardizes to a correlation of greater than 0.6 with the AO. While not as good a method as using the daily data (which we only have back to the late 90s) he admits it still captures the statistical significance and likely a more realistic toned down correlation coeff. 

Link to comment
Share on other sites
bluewave

bluewave

Posted
Posted

Yes, I agree. October also sees a transition from summer to winter in the form of "standing waves" which are unique to the NH. They, along with the polar vortex, become established this month. A lot of various things that work together have been simplified for us by Cohen with the SAI. Is it perfect? Nope... Are there larger-scale factors at play? Yes...

 

But in the end, that transition and full-scale beginning of the cold season that takes over mid to late October is definitely being picked up by the SAI.

 

He just recently expanded on some of the other larger-scale factors.

 

http://www.tos.org/oceanography/archive/26-4_cohen.pdf

Link to comment
Share on other sites
HM

HM

Posted
Posted

Ok, perhaps arbitrary is too strong of a word here.  Certainly October, typically the later half of it, is a crucial time for transitioning to the cold season in the Arctic - I think it's a good study... I just question this year's correlation since Oct began with an anomalously high snow cover value compared to recent years. I think this aspect will be modified in time as transition periods are unique to every year, where for some years the 31 day window will move a number of days in either direction.

 

But the study does make a great start in working away at a forecasting skill for something which most have deemed impossible to forecast.

 

Reliable correlations work great for long-term forecasting until they don't.   

 

Perhaps the SAI in some way does address the issue of anomalous snow cover to start off October. Maybe it is the stronger wave driving from later in the month with a strong snow cover advancement that is capable of disturbing the PV early winter. If you start things too soon, the PV will still be developing and the wave strength will not be as strong. We are seeing now the consequences of early wave 1 disturbances, actually aiding in the vortex intensification mostly since it is still in the formative stages.

 

Your analogy to the sun is ludicrous. The fields of meteorology and climatology are much further advanced than the sun. Solar physicists have to deal with actual poor data issues and no direct obs, to the point where everything is guess-work. It is not even comparable and there were a decent amount of solar physicists with a more progressive approach that correctly anticipated a weaker solar cycle 24 (and 25). They are in the infancy stages for prediction.

Link to comment
Share on other sites
Wow

Wow

Posted
Posted

Perhaps the SAI in some way does address the issue of anomalous snow cover to start off October. Maybe it is the stronger wave driving from later in the month with a strong snow cover advancement that is capable of disturbing the PV early winter. If you start things too soon, the PV will still be developing and the wave strength will not be as strong. We are seeing now the consequences of early wave 1 disturbances, actually aiding in the vortex intensification mostly since it is still in the formative stages.

 

Your analogy to the sun is ludicrous. The fields of meteorology and climatology are much further advanced than the sun. Solar physicists have to deal with actual poor data issues and no direct obs, to the point where everything is guess-work. It is not even comparable and there were a decent amount of solar physicists with a more progressive approach that correctly anticipated a weaker solar cycle 24 (and 25). They are in the infancy stages for prediction.

 

I wouldn't say AO prediction is not in its infancy stage.  Again, it's a start.  And a good one so far.

Link to comment
Share on other sites
ORH_wxman

ORH_wxman

Posted
Posted

FWIW, here are the ten highest snow cover years for the final week of September and a description of how SAI did that year in predicting the AO (weekly is used prior to 1997):

 

1977: AO was lower than predicted by SAI. Mean October snow cover was very accurate in this year predicting the lower AO.

1972: No data in the Cohen paper prior to 1973

1973: Snow cover still advanced rapidly and SAI predicted a strongly -AO along with overall Oct extent. AO was near neutral.

2013: ?????

2002: SAI predicted a neutral AO. Overall October snow cover predicted a negative AO. AO finished modestly negative.

2001: SAI and overall Oct snow cover both predicted near neutral AO. AO was modestly positive.

1989: SAI and overall Oct snow cover predicted near neutral AO and AO was solidly positive

2000: SAI predicted a negative AO and the AO was negative. Overall October snow cover predicted near neutral AO.

1995: SAI predicted negative AO and Oct snow cover predicted neutral AO. AO was solidly negative.

1986: SAI predicted slightly negative AO, while overall Oct snow cover predicted slightly positive. AO was negative.

 

 

 

This doesn't really give us a lot of info. The higher up on this list, it seems overall extent did better, the lower down on the list it seems SAI did better. Obviously tiny smaple size.

Link to comment
Share on other sites
skierinvermont

skierinvermont

Posted
Posted

The whole reason that the advance index correlates much better than the October mean is that it is unaffected by the starting amount of snow. It isolates the important variable (rate of change) and excludes the unimportant one (starting point and everything that happened before Oct 1). 

 

I wouldn't be surprised if a better correlation were found shifting the start date a few days.. but I doubt more than that and if I had to guess the correlation would be improved shifting the start date later rather than earlier.

 

If the last week of September mattered, the correlation would be stronger for mean Oct ext than for SAI. It isn't even close. Ignoring this fact is a poor attempt to justify a -AO this winter.

 

In addition, the OPI (discussed in the thread on the OPI) is indicating a strongly +AO this winter.

Link to comment
Share on other sites
skierinvermont

skierinvermont

Posted
Posted

I think it would be interesting to calculate the SAI including the rapid increase in late September, and excluding the rapid increase in late September, and see which predicted AO value based off of the two SAI values comes closer to the actual value this winter.  

 

We already know the answer to this... the correlation will be much better starting Oct 1. 

Link to comment
Share on other sites

Archived

This topic is now archived and is closed to further replies.

Go to topic listing

  • Recently Browsing   0 members

    • No registered users viewing this page.
×
×
  • Create New...