Jump to content

Welcome to our forums!

Sign In or Register to gain full access to our forums. By registering with us, you'll be able to discuss, share and private message with other members of our community.

Welcome!

Thanks for stopping by the Weather Forums! Please take the time to register and join our community. Feel free to post or start new topics on anything related to the weather or the climate.


Photo

Using Fall 2016 to predict Winter 2016-2017

- - - - -

  • Please log in to reply

#1
westcoastexpat

Posted 22 November 2016 - 07:25 PM

westcoastexpat

    Forum Contributor

  • Meteorologist
  • PipPipPipPip
  • 866 posts
  • LocationBlaine, WA

Disclaimer: not a climatologist so this is a real shot in the dark, for fun and almost certain future embarrassment

 

Disclaimer: outside of strong ENSO signals, I'm iffy on the usefulness of analogs as a predictor of seasonal temperature anomalies

 

Disclaimer: you could have a warmer than normal month with an epic snowstorm (and the opposite is true), so don't get too caught up in monthly anomalies 

 

Disclaimer: these views are purely my own and not affiliated with any other person(s)

 

 

Now that THAT'S out of the way, let's get to the good stuff.

 

Now that we're at the edge of our fall season, I decided to look at previous autumns to see if there was a correlation between a fall progression and an upcoming winter pattern.

 

I know that there are a lot of variables to consider for a seasonal forecast, from ENSO to solar activity to numerous other teleconnections. Although I remained cognizant of the ENSO mode and solar activity, the main driver of this "forecast" is to look at temperature anomalies for the Sep/Oct/Nov period and see what we can find!

 

First, let me start off by saying it's been a warm past 12 months across the continent:

 

cd138.229.24.136.326.16.28.24.prcp.png

 

 

 

 

And we know from November that the situation hasn't changed much.

 

But I have noticed a certain trend -- take a look at this crappy GIF I put together:

 

giphy.gif

 

 

 

 

You'll notice that since the summer, the "blob" of warm anomalies (pun intended) have slowly shifted westward. Does that mean a winter hotbox for the pro-pot west?

 

Let's look at November thus far:

 

Attached File  November.png   31.12KB   0 downloads

 

It does continue the trend of a westward progression of warm anomalies. But I don't think it's going to continue into December. Stay tuned.

 

I've looked through numerous fall seasons and have chosen the following analogs to be a "best fit" to the Fall of 2016:

 

2007

1998

2005

1983

1960

 

Here's a composite of the above 5 seasons:

 

cd138.229.24.136.326.20.14.3.prcp.png

 

 

 

 

And how does it measure up to the Fall of 2016?

 

cd138.229.24.136.326.20.13.56.prcp.png

 

 

 

 

Note that November couldn't be added to the above map, but we know that the warm anomalies were further west, so I would expect the warmest anomalies to shift slightly west. Not a perfect match, but pretty close.

 

I chose autumns that not only had very similar September to October to November progressions as in 2016, but also had weaker ENSO signals or years during quick transitions from Nino to Nina. I also noticed that a wet fall season was experienced in the PNW for most of these years, matching 2016.

 

Now the real question -- what does this tell us for the upcoming winter?

 

Take a look at December:

 

cd138.229.24.136.326.20.18.15.prcp.png

 

 

 

 

Quite the reversal. Going through many of the fall seasons, balmy falls almost always were followed by a cold December across the continent. I found that to be very interesting.

 

However, it's not all fun and games. Here's January:

 

cd138.229.24.136.326.20.31.54.prcp.png

 

 

 

 

And there's not a lot of confidence with February either:

 

cd138.229.24.136.326.20.31.24.prcp.png

 

 

 

 

This would go against my general prediction of a cooler winter overall, particularly with the core of cold being the upper central plains. Again, not a fan of analogs, but I thought it was interesting and wanted to share and discuss!


  • IbrChris likes this

#2
DJ Droppin

Posted 22 November 2016 - 07:53 PM

DJ Droppin

    Advanced Member

  • Members
  • PipPipPipPipPipPip
  • 4341 posts
  • LocationEast Portland, OR

Disclaimer: not a climatologist so this is a real shot in the dark, for fun and almost certain future embarrassment

 

Disclaimer: outside of strong ENSO signals, I'm iffy on the usefulness of analogs as a predictor of seasonal temperature anomalies

 

Disclaimer: you could have a warmer than normal month with an epic snowstorm (and the opposite is true), so don't get too caught up in monthly anomalies 

 

Disclaimer: these views are purely my own and not affiliated with any other person(s)

 

 

Now that THAT'S out of the way, let's get to the good stuff.

 

Now that we're at the edge of our fall season, I decided to look at previous autumns to see if there was a correlation between a fall progression and an upcoming winter pattern.

 

I know that there are a lot of variables to consider for a seasonal forecast, from ENSO to solar activity to numerous other teleconnections. Although I remained cognizant of the ENSO mode and solar activity, the main driver of this "forecast" is to look at temperature anomalies for the Sep/Oct/Nov period and see what we can find!

 

First, let me start off by saying it's been a warm past 12 months across the continent:

 

cd138.229.24.136.326.16.28.24.prcp.png

 

 

 

 

And we know from November that the situation hasn't changed much.

 

But I have noticed a certain trend -- take a look at this crappy GIF I put together:

 

giphy.gif

 

 

 

 

You'll notice that since the summer, the "blob" of warm anomalies (pun intended) have slowly shifted westward. Does that mean a winter hotbox for the pro-pot west?

 

Let's look at November thus far:

 

attachicon.gifNovember.png

 

It does continue the trend of a westward progression of warm anomalies. But I don't think it's going to continue into December. Stay tuned.

 

I've looked through numerous fall seasons and have chosen the following analogs to be a "best fit" to the Fall of 2016:

 

2007

1998

2005

1983

1960

 

Here's a composite of the above 5 seasons:

 

cd138.229.24.136.326.20.14.3.prcp.png

 

 

 

 

And how does it measure up to the Fall of 2016?

 

cd138.229.24.136.326.20.13.56.prcp.png

 

 

 

 

Note that November couldn't be added to the above map, but we know that the warm anomalies were further west, so I would expect the warmest anomalies to shift slightly west. Not a perfect match, but pretty close.

 

I chose autumns that not only had very similar September to October to November progressions as in 2016, but also had weaker ENSO signals or years during quick transitions from Nino to Nina. I also noticed that a wet fall season was experienced in the PNW for most of these years, matching 2016.

 

Now the real question -- what does this tell us for the upcoming winter?

 

Take a look at December:

 

cd138.229.24.136.326.20.18.15.prcp.png

 

 

 

 

Quite the reversal. Going through many of the fall seasons, balmy falls almost always were followed by a cold December across the continent. I found that to be very interesting.

 

However, it's not all fun and games. Here's January:

 

cd138.229.24.136.326.20.31.54.prcp.png

 

 

 

 

And there's not a lot of confidence with February either:

 

cd138.229.24.136.326.20.31.24.prcp.png

 

 

 

 

This would go against my general prediction of a cooler winter overall, particularly with the core of cold being the upper central plains. Again, not a fan of analogs, but I thought it was interesting and wanted to share and discuss!

Nice job on this. Thanks for doing all of the research and putting this together. It will be interesting to see how December-February shakes out.



#3
westcoastexpat

Posted 22 November 2016 - 08:42 PM

westcoastexpat

    Forum Contributor

  • Meteorologist
  • PipPipPipPip
  • 866 posts
  • LocationBlaine, WA

I also realize that the January map is a little misleading. The only signal that is really apparent is on the leeward side of the rockies. But the "normal" looking area for most of the continent is an average of all the years, so "0" could actually mean a mixture of cold and balmy Januaries. Let's take a closer look:

 

Jan 2008

cd138.229.24.136.326.21.42.8.prcp.png

 

 

 

Jan 1999

cd138.229.24.136.326.21.41.47.prcp.png

 

 

 

Jan 2006

cd138.229.24.136.326.21.41.14.prcp.png

 

 

 

Jan 1984

cd138.229.24.136.326.21.40.57.prcp.png

 

 

 

Jan 1961

cd138.229.24.136.326.21.40.34.prcp.png

 

A true tossup, but does seem to favor a warmer west.



#4
IbrChris

Posted 23 November 2016 - 08:02 AM

IbrChris

    Moderating Meteorologist

  • Meteorologist
  • PipPipPipPipPip
  • 1903 posts
  • LocationTigard, OR and Portland, OR (work)

I also realize that the January map is a little misleading. The only signal that is really apparent is on the leeward side of the rockies. But the "normal" looking area for most of the continent is an average of all the years, so "0" could actually mean a mixture of cold and balmy Januaries.

 

A true tossup, but does seem to favor a warmer west.

Yep it's important to consider spread...I would consider this a pretty low confidence forecast. I also share your skepticism on analog forecasts.


The Pacific Northwest: Where storms go to die.


#5
Front Ranger

Posted 23 November 2016 - 08:54 AM

Front Ranger

    Forum Fantastic

  • Members
  • PipPipPipPipPipPipPipPip
  • 10628 posts
  • LocationWestminster, CO

Yep it's important to consider spread...I would consider this a pretty low confidence forecast. I also share your skepticism on analog forecasts.

 

In my experience, analog-based forecasts (if they take into account the right factors) are the most successful method for seasonal forecasting. As always with LR forecasting, there's a relatively low degree of confidence compared to shorter term forecasts.


Cool anomalies soothe the soul.


#6
IbrChris

Posted 23 November 2016 - 08:58 AM

IbrChris

    Moderating Meteorologist

  • Meteorologist
  • PipPipPipPipPip
  • 1903 posts
  • LocationTigard, OR and Portland, OR (work)

In my experience, analog-based forecasts (if they take into account the right factors) are the most successful method for seasonal forecasting. As always with LR forecasting, there's a relatively low degree of confidence compared to shorter term forecasts.

 

You may want to apply for any openings at the CPC then.

Truth is a simple climatology forecast often performs better than a seasonal analog prediction, unless you take an objective statistical approach to verification your thinking is probably tainted by confirmation bias.


The Pacific Northwest: Where storms go to die.


#7
Front Ranger

Posted 23 November 2016 - 09:01 AM

Front Ranger

    Forum Fantastic

  • Members
  • PipPipPipPipPipPipPipPip
  • 10628 posts
  • LocationWestminster, CO

You may want to apply for any openings at the CPC then.

Truth is a simple climatology forecast often performs better than a seasonal analog prediction, unless you take an objective statistical approach your thinking is probably tainted by confirmation bias.

 

Of course.  :)


Cool anomalies soothe the soul.


#8
IbrChris

Posted 23 November 2016 - 09:09 AM

IbrChris

    Moderating Meteorologist

  • Meteorologist
  • PipPipPipPipPip
  • 1903 posts
  • LocationTigard, OR and Portland, OR (work)

Of course.  :)

 

In any seasonal forecast the neutral state is 33% chance of climatology, below climo or above climo (rounded). Looking at a CPC forecast they might put the PNW in a 40% chance of above normal temps...which means the other two options are only worth a combined 60%...though you don't know the weights...could be 30-30, could be 35-25. The advantage to an ensemble approach is that you can recognize potential disparate solution sets...maybe 40% are going warm, 35% are going cold and 25% are going climo. Not to mention it's quite arbitrary what defines "above" and "below" climo...does a 0.1 F deviation imply above/below normal? What about 3 F? What about 1 standard deviation from normal? If we took a poll we would all have different criteria.

In my industry seasonal forecasts are mostly used to gauge market direction and hedge against a potential costly long or short position when trading several months in advance of delivery.


The Pacific Northwest: Where storms go to die.


#9
westcoastexpat

Posted 23 November 2016 - 09:57 AM

westcoastexpat

    Forum Contributor

  • Meteorologist
  • PipPipPipPip
  • 866 posts
  • LocationBlaine, WA

In my experience, analog-based forecasts (if they take into account the right factors) are the most successful method for seasonal forecasting. As always with LR forecasting, there's a relatively low degree of confidence compared to shorter term forecasts.


Any thoughts on the analogs used here?

#10
Front Ranger

Posted 23 November 2016 - 10:36 AM

Front Ranger

    Forum Fantastic

  • Members
  • PipPipPipPipPipPipPipPip
  • 10628 posts
  • LocationWestminster, CO

Any thoughts on the analogs used here?

 

Well, personally I prefer to start with large scale factors like ENSO, PDO, QBO, solar, etc and then use fall pattern matches as a secondary factor. I just think the correlations are stronger. For that reason, it's hard for me to view 2007 and 1998 as good analogs to use. But I do understand why you picked them.

 

Also, I prefer to look at Northern Hemisphere pattern matches rather than just the U.S., but that's just me.


Cool anomalies soothe the soul.


#11
Front Ranger

Posted 23 November 2016 - 11:14 AM

Front Ranger

    Forum Fantastic

  • Members
  • PipPipPipPipPipPipPipPip
  • 10628 posts
  • LocationWestminster, CO

In any seasonal forecast the neutral state is 33% chance of climatology, below climo or above climo (rounded). Looking at a CPC forecast they might put the PNW in a 40% chance of above normal temps...which means the other two options are only worth a combined 60%...though you don't know the weights...could be 30-30, could be 35-25. The advantage to an ensemble approach is that you can recognize potential disparate solution sets...maybe 40% are going warm, 35% are going cold and 25% are going climo. Not to mention it's quite arbitrary what defines "above" and "below" climo...does a 0.1 F deviation imply above/below normal? What about 3 F? What about 1 standard deviation from normal? If we took a poll we would all have different criteria.

In my industry seasonal forecasts are mostly used to gauge market direction and hedge against a potential costly long or short position when trading several months in advance of delivery.

 

Good points. 

 

So applying this to my winter outlook, using the best match analogs and general temperature anomalies for the PNW, and giving double weight to my top 5 analogs.

 

Very warm: +3 and higher

Warm: +1 to +3

Average: -1 to +1

Cold: -1 to -3

Very cold: -3 and lower

 

DEC

 

Very warm: 2 (9%)

Warm: 3 (14%)

Average: 9 (43%)

Cold: 2 (9%)

Very cold: 5 (24%)

 

JAN

 

Very warm: 2 (9%)

Warm: 3 (14%)

Average: 6 (29%)

Cold: 7 (33%)

Very cold: 3 (14%)


  • IbrChris likes this

Cool anomalies soothe the soul.


#12
IbrChris

Posted 23 November 2016 - 11:25 AM

IbrChris

    Moderating Meteorologist

  • Meteorologist
  • PipPipPipPipPip
  • 1903 posts
  • LocationTigard, OR and Portland, OR (work)

Good points. 

 

So applying this to my winter outlook, using the best match analogs and general temperature anomalies for the PNW, and giving double weight to my top 5 analogs.

 

Very warm: +3 and higher

Warm: +1 to +3

Average: -1 to +1

Cold: -1 to -3

Very cold: -3 and lower

 

DEC

 

Very warm: 2 (9%)

Warm: 3 (14%)

Average: 9 (43%)

Cold: 2 (9%)

Very cold: 5 (24%)

 

JAN

 

Very warm: 2 (9%)

Warm: 3 (14%)

Average: 6 (29%)

Cold: 7 (33%)

Very cold: 3 (14%)

At least there's a tangible meaning there...I can say "This forecast is going for a 14% chance that the PNW will have a mean Jan temp -3 F or more below normal."

Rather than looking at a CPC forecast and saying "They're going for a near normal January." But we don't know if there's a 60% chance or a 40% chance it will be near normal or even the criteria for "near normal".


The Pacific Northwest: Where storms go to die.