NCAR: Solar cycle linked to global climate

Basil Copeland and I also found linkages between surface temperature and solar cycles in two articles we published in the last year. We were roundly criticized and ridiculed by warmists mainly due to a statistical error in the first essay, but the base premise remained and the second essay was improved due to that error. I’m pleased to see that NCAR has found other solar to earth linkages, such as this one in ENSO. This is exciting news, but by no means a complete solution to the climate puzzle. There is much more to be learned about this. This is but one connector of the hydra-like patch cable that Dr. Jack Eddy imagined – Anthony

http://www.physorg.com/newman/gfx/news/hires/solarcycleli.jpg

Scientists find link between solar cycle and global climate similar to El Nino/La Nina. Credit: NCAR

Establishing a key link between the solar cycle and global climate, research led by scientists at the National Science Foundation (NSF)-funded National Center for Atmospheric Research (NCAR) in Boulder, Colo., shows that maximum solar activity and its aftermath have impacts on Earth that resemble La Niña and El Niño events in the tropical Pacific Ocean.

The research may pave the way toward predictions of temperature and precipitation patterns at certain times during the approximately 11-year solar cycle.

“These results are striking in that they point to a scientifically feasible series of events that link the 11-year solar cycle with ENSO, the tropical Pacific phenomenon that so strongly influences climate variability around the world,” says Jay Fein, program director in NSF’s Division of Atmospheric Sciences. “The next step is to confirm or dispute these intriguing model results with observational data analyses and targeted new observations.”

The total energy reaching Earth from the sun varies by only 0.1 percent across the solar cycle. Scientists have sought for decades to link these ups and downs to natural weather and climate variations and distinguish their subtle effects from the larger pattern of human-caused global warming.

Building on previous work, the NCAR researchers used computer models of global climate and more than a century of to answer longstanding questions about the connection between solar activity and global climate.

The research, published this month in a paper in the Journal of Climate, was funded by NSF, NCAR’s sponsor, and by the U.S. Department of Energy.

“We have fleshed out the effects of a new mechanism to understand what happens in the tropical Pacific when there is a maximum of solar activity,” says NCAR scientist Gerald Meehl, the paper’s lead author. “When the sun’s output peaks, it has far-ranging and often subtle impacts on tropical precipitation and on weather systems around much of the world.”

The new paper, along with an earlier one by Meehl and colleagues, shows that as the Sun reaches maximum activity, it heats cloud-free parts of the Pacific Ocean enough to increase evaporation, intensify tropical rainfall and the trade winds, and cool the eastern tropical Pacific.

The result of this chain of events is similar to a La Niña event, although the cooling of about 1-2 degrees Fahrenheit is focused further east and is only about half as strong as for a typical La Niña.

Over the following year or two, the La Niña-like pattern triggered by the solar maximum tends to evolve into an El Niño-like pattern, as slow-moving currents replace the cool water over the eastern tropical Pacific with warmer-than-usual water.

Again, the ocean response is only about half as strong as with El Niño.

True La Niña and El Niño events are associated with changes in the temperatures of surface waters of the eastern . They can affect weather patterns worldwide.

The paper does not analyze the weather impacts of the solar-driven events. But Meehl and his co-author, Julie Arblaster of both NCAR and the Australian Bureau of Meteorology, found that the solar-driven La Niña tends to cause relatively warm and dry conditions across parts of western North America.

More research will be needed to determine the additional impacts of these events on weather across the world.

“Building on our understanding of the solar cycle, we may be able to connect its influences with weather probabilities in a way that can feed into longer-term predictions, a decade at a time,” Meehl says.

Scientists have known for years that long-term solar variations affect certain weather patterns, including droughts and regional temperatures.

But establishing a physical connection between the decadal and global climate patterns has proven elusive.

One reason is that only in recent years have computer models been able to realistically simulate the processes associated with tropical Pacific warming and cooling associated with El Niño and La Niña.

With those models now in hand, scientists can reproduce the last century’s solar behavior and see how it affects the Pacific.

To tease out these sometimes subtle connections between the sun and Earth, Meehl and his colleagues analyzed sea surface temperatures from 1890 to 2006. They then used two computer models based at NCAR to simulate the response of the oceans to changes in solar output.

They found that, as the sun’s output reaches a peak, the small amount of extra sunshine over several years causes a slight increase in local atmospheric heating, especially across parts of the tropical and subtropical Pacific where Sun-blocking clouds are normally scarce.

That small amount of extra heat leads to more evaporation, producing extra water vapor. In turn, the moisture is carried by trade winds to the normally rainy areas of the western tropical Pacific, fueling heavier rains.

As this climatic loop intensifies, the trade winds strengthen. That keeps the eastern Pacific even cooler and drier than usual, producing La Niña-like conditions.

Although this Pacific pattern is produced by the solar maximum, the authors found that its switch to an El Niño-like state is likely triggered by the same kind of processes that normally lead from La Niña to El Niño.

The transition starts when the changes of the strength of the trade winds produce slow-moving off-equatorial pulses known as Rossby waves in the upper ocean, which take about a year to travel back west across the Pacific.

The energy then reflects from the western boundary of the tropical Pacific and ricochets eastward along the equator, deepening the upper layer of water and warming the ocean surface.

As a result, the Pacific experiences an El Niño-like event about two years after solar maximum. The event settles down after about a year, and the system returns to a neutral state.

“El Niño and La Niña seem to have their own separate mechanisms,” says Meehl, “but the solar maximum can come along and tilt the probabilities toward a weak La Niña. If the system was heading toward a La Niña anyway,” he adds, “it would presumably be a larger one.”

Source: National Science Foundation (news : web)

h/t to Leif Svalgaard

0 0 votes
Article Rating

Discover more from Watts Up With That?

Subscribe to get the latest posts sent to your email.

202 Comments
Inline Feedbacks
View all comments
Paul Vaughan
July 19, 2009 8:47 pm

Re: Leif Svalgaard (19:53:35)
These comments are constructive. What concise labeling of the wavelet transform plot would you deem ‘acceptable’? – & balanced (given the potential for misunderstanding due to widespread perceptions that are beyond our control)?

Paul Vaughan
July 19, 2009 8:51 pm

Re: Leif Svalgaard (20:10:59)
To reiterate:
I only bother to produce wavelet images when I know intuitively (from conceptual understanding) that they will provide a concise summary of something I already see.

July 19, 2009 9:09 pm

Paul Vaughan (20:38:33) :
Re: Leif Svalgaard (20:10:59)
You appear awfully eager to quote that OUT-OF-CONTEXT.
I think the context is very appropriate. We once had a student that discovered the usefulness of a Chree-analysis [superposed epoch analysis]. As you did, she tried it with every data set she could find [about a hundred] and did turn up a few correlations at the 95% level.
Paul Vaughan (20:47:28) :
Re: Leif Svalgaard (19:53:35)
These comments are constructive.
I strive hard to make ALL comments constructive.
What concise labeling of the wavelet transform plot would you deem ‘acceptable’? – & balanced (given the potential for misunderstanding due to widespread perceptions that are beyond our control)?
No labels at all would be acceptable to me, and minimize misunderstanding.

July 19, 2009 9:17 pm

Paul Vaughan (20:36:12) :
that I’m not running “crap-shoot” “shot-in-the-dark” analyses
Then, since you claim no causation and only correlation, what are the criteria for selecting a correlation to try [knowing ‘exactly” what to look for]. I would ordinarily only even look for correlation if I had an idea of causation, then the correlation might or not confirm the mechanism I was toying with, thus causation comes first. To me it makes no sense to look for correlation without a mechanism in mind, then it has degenerated to [as you say] crap shots.

Paul Vaughan
July 19, 2009 10:06 pm

Leif Svalgaard (21:17:19) “[…] you claim […] only correlation […]”
Yet more distortion – note that the plot is univariate.

Leif Svalgaard (21:09:38) “[…] did turn up a few correlations at the 95% level.”
You can do that with a random series. You are (possibly) missing the point that it is healthy for students to learn from experience. In today’s age of heavily-algorithmic & computationally-intensive analysis, it is important that students (& teachers) learn a “feel” for the performance of (& issues with) the algorithms (& software, hardware, etc.).
You may also be missing another point: Real series are not necessarily random, so you would not expect 5 in 100. (That logic holds only under the assumption of randomness.) Think about what PCA & factor analysis are all about. Those are legitimate tools.
Anyway, I think you and I both understand and that this exchange has degenerated into a show for the public (who you are worried will misunderstand something).

Paul Vaughan
July 19, 2009 10:07 pm

A physicist brought the following complaint to a judge:
“The statistician is investigating physics. This is unacceptable.”
The statistician then counter-charged:
“The physicist is using statistics without first becoming formally recognized as a member of the Statistical Society.”
And so the wise judge passed a rule requiring an army of administrators to ensure that light-bulbs be changed by electricians. Lawyers delighted in their profits.
Meanwhile – in a society with less in-fighting – problems were solved efficiently.

tallbloke
July 20, 2009 2:02 am

It’s great to see you two doing your level best to understand each other’s angle.
In support of Paul’s approach, I do think it’s a worthwhile exercise to investigate statistical correaltions regardless of a lack of obvious causative link for two reasons.
1)It may evoke a realisation that known forces not previously thought to be effective in certain ways are.
2) It may lead to the discovery of new forces. If physicists can have ‘dark energy’ and string, and climatologists can have ‘teleconnections’, then statisticians can have ‘unknown causitive links’ to chuck over the fence for physicists and climatologists to think about.
In support of Leif’s approach, a belief in the sufficiency of known forces and their effects enables progress in theory development, even if it turns out to be wrong or inadequate later. The realisation that the theory is wrong or inadequate leads us back to the a priori consideration of new frameworks for aspects of physics. Newton -> Einstein, Dalton ->Bohr, etc.
Let the physicists do physics, let the statisticians do stats, and let them both talk to each other in non-judgmental and collaborative ways.
Well done guys.

July 20, 2009 3:38 am

Paul Vaughan (22:06:51) :
Yet more distortion – note that the plot is univariate.
what has that to do with anything? speak English, please.
You can do that with a random series
A series with random data is not the same as a series chosen at random.

July 20, 2009 10:10 am

Paul Vaughan (22:06:51) :
Yet more distortion – note that the plot is univariate.
Here is how a physicist sees and solves a multivariate problem:
http://www.leif.org/research/suipr699.pdf
You may want to try your hand on that problem. The data is the am-index and the OMNI dataset.

Anders L.
July 20, 2009 2:35 pm

The solar cycle is also very clearly linked to the number of Republican senators in the US Congress. The mechanism for this is just as well established as the link between the solar cycle and climate. (Completetly unknown, that is.)

Paul Vaughan
July 20, 2009 5:17 pm

Univariate = 1 variable
The plot is univariate.
It is a plot of one variable.
No claim has been made.
No correlation has been presented.
[Correlation with what??? There is only one variable!!!]

Paul Vaughan
July 20, 2009 5:19 pm

tallbloke,
Thanks for the Erl Happ link. I’m making my way through it…

July 20, 2009 6:49 pm

Paul Vaughan (17:17:27) :
[Correlation with what??? There is only one variable!!!]
y = f(x) is a function of one variable. The functional form, e.g. the set of pairs (x,y), is [especially if errors are present] essentially, a correlation of y with x.

Paul Vaughan
July 20, 2009 10:43 pm

Re: Leif Svalgaard (18:49:01)
Thanks for another laugh.

July 21, 2009 5:36 am

Paul Vaughan (22:43:53) :
Re: Leif Svalgaard (18:49:01)
Thanks for another laugh.
No problem; you provide me with good material for that.

Paul Vaughan
July 21, 2009 2:38 pm

The only substantive issue you have raised is what the title of the graph should be – and you fail to offer a satisfactory alternative.

July 21, 2009 3:02 pm

Paul Vaughan (14:38:03) :
The only substantive issue you have raised is what the title of the graph should be – and you fail to offer a satisfactory alternative.
Is this is your idea of substance? I don’t care what you call it. Having no title is perfectly satisfactory as the substance is not in the title but in the graph.

Paul Vaughan
July 21, 2009 4:06 pm

Leif Svalgaard (15:02:27) “Having no title is perfectly satisfactory as the substance is not in the title but in the graph.”
Many will irritatedly disagree, but I agree with you ^here.
The assumption of stationarity (e.g. in unwindowed-FFT) is untenable.

The substance of the wavelet transform is SOLID.
http://www.sfu.ca/~plv/SunspotCyclePeriod.PNG
Arguments suggesting otherwise are a foolish waste of time (possibly aimed at pulling wool over the eyes of the innumerate &/or those lacking deeper conceptual understanding of wavelet methods).
My guess is that you are willing to use the weight of your stature in this community to deceitfully distort because you are worried others will misuse the result. If others misuse the result, your issue is with them.
The result on its own is purely descriptive. (No inference has been made.)
Challenging the result is as foolish as challenging a verifiable claim that a red car is red (which is a long way from claiming red paint on the car causes x & y & z).

If others misuse the result, I might be inclined to join you in calling them on it – but let’s cross that bridge if & when it arises.

July 21, 2009 4:40 pm

Paul Vaughan (16:06:06) :
If others misuse the result, your issue is with them.
and if you misuse the result?
As I’ve said the (your?) concept of the length of a solar cycle is already dubious, no matter by what method you come up with a number. For another purpose I have followed the practice [because others did and I compare with them] of defining a cycle ‘length’, two ways: max-max and min-min, just using the ‘official’ sunspot numbers. Within the accuracy of the data and the [much more troubling] fuzziness of the concept, my plot of the ‘length’ as a function of time:
http://www.leif.org/research/Cycle%20Lengths%20and%20Temperatures.png if identical to yours [except extends usefully further on both sides], so there is no additional [meaningful] information in the wavelet plot, so I will agree that the result is solid.

July 21, 2009 4:46 pm

Paul Vaughan (16:06:06) :
My guess is that you […] deceitfully distort
Your comment shows that you fit the old Danish proverb: “a thief thinks that everybody steals”.

Mark T
July 21, 2009 4:58 pm

Leif Svalgaard (18:49:01) :
y = f(x) is a function of one variable. The functional form, e.g. the set of pairs (x,y), is [especially if errors are present] essentially, a correlation of y with x.

Um, no.
Mark

bill
July 21, 2009 6:19 pm

Paul Vaughan (22:33:48) :
I’m very curious to hear what you might propose to explain the reversed phase relationships of the 1800s.
Paul Vaughan (20:51:53) :
I only bother to produce wavelet images when I know intuitively (from conceptual understanding) that they will provide a concise summary of something I already see

Paul I have plotted phase/period/FFTs of the chandler wobble and I see little of which you speak.
1. There are only 20 data points per year from 1890-present and 10 points prior to this date.
2. The phase/period anomalies occur during the minimum amplitude where accuracy disappears
http://img39.imageshack.us/img39/5355/chandlerwobblephaseperi.jpg
3. There is a beat frequency betwen 365.25 days and the chandler period which causes the minima.
http://img24.imageshack.us/img24/4861/chandlerperiodfft.jpg
I am rather surprised that data is recorded on a 18.26 day period. Why?
in your plot
http://www.sfu.ca/~plv/ChandlerPeriodAgassizBC,CanadaPrecipitationTimePlot.PNG
you seem to get a chandler period with a 1 day accuracy – how is this possible from a 18 day measurement period?
The above plot shows 420 day period in 1930 My 10 year averaged chandler period in 1930 is ~340 days
http://img339.imageshack.us/img339/8189/chandlerwobbleaveragepe.jpg
Notice also that this dip is not unusual.

July 21, 2009 7:13 pm

Mark T (16:58:12) :
“y = f(x) is a function of one variable. The functional form, e.g. the set of pairs (x,y), is [especially if errors are present] essentially, a correlation of y with x.”
Um, no.

Well, in my book it is, “Um” or not. Example, geomagnetic activity is to first order a function of solar wind speed. A = k V^n. Correlating A vs. V one finds n = 2. I’m not limiting myself to linear correlation, as one would do in statistics. It is more useful to broaden the concept to any functional relationship. One can then be specific and say linear correlation, if there is chance of confusion. One can also calculate the Spearman’s rank correlation coefficient as long as the function is monotonic, but not necessarily linear. In physics and in general applications it makes sense to broaden the concept. Of course, if a statistician wears blinders he cannot get out of his box.

Paul Vaughan
July 21, 2009 10:34 pm

Re: bill (18:19:12)
Don’t confuse polar motion (in general) with the Chandler wobble. The annual wobble beats with the Chandler wobble to produce the ~6.5a polar motion group waves. (It seems you have figured this out.)
That the Chandler wobble reversed phase in ~1931 is not in dispute – see the literature. For example:
Jan Vondrak (1999). Earth rotation parameters 1899.7-1992.0 after reanalysis within the hipparcos frame. Surveys in Geophysics 20, 169-195.
http://www.yspu.yar.ru/astronomy/lib/Rotation.pdf
[See particularly section 3.2.]
J. Vondrak & C. Ron (2005). The great Chandler wobble change in 1923-1940 re-visited. In: H.-P. Plag, B. Chao, R. Gross, & T. Van Dam (eds.), Forcing of polar motion in the Chandler frequency band: A contribution to understanding interannual climate variations, Cahiers du Centre Europeen de Geodynamique et de Seismologie 24, 39-47.
Even just looking at the raw polar motion series, one can see that the group wave envelopes lengthen in period on either side of ~1931.
As for the sampling frequency: They started out with 10 per year & then doubled it to 20 per year in 1890.
To reiterate past comments:
Unwindowed-FFT is misleading (sometimes extremely so) for nonstationary time series. I recommend wavelet methods. I posted the instructions for the analysis in an earlier thread:
http://wattsupwiththat.com/2009/07/01/another-paper-showing-evidence-of-a-solar-signature-in-temperature-records/
See Paul Vaughan (19:14:38) [July 7, 2009].
Note that the variable I worked with is an index derived from the raw polar motion data.
I think you may want to read Gross (2005) and then review some of your plots.
Gross, R. S. (2005). The observed period and Q of the Chandler wobble. In: H.-P. Plag, B. Chao, R. Gross, & T. Van Dam (eds.), Forcing of polar motion in the Chandler frequency band: A contribution to understanding interannual climate variations, Cahiers du Centre Europeen de Geodynamique et de Seismologie 24, 31-37.
http://www.sbl.statkart.no/literature/plag_etal_2005_editors/gross_1_CWTQ_final.pdf

Paul Vaughan
July 21, 2009 10:50 pm

Leif Svalgaard (16:40:48) “[…] so there is no additional [meaningful] information in the wavelet plot […]”
You appear to have missed the point made above (in response to tallbloke’s original inquiry) that it provides monthly estimates.
As indicated above, I’ve (previously) done the analysis the way you did.
You are eager to label “solar cycle period” as a “dubious” term and yet you cannot provide a suitable alternative that will be understood by a general science audience.
Nonetheless, you agree that the results are solid – so case closed.