Note: This is my analysis of a new paper by Joe D’Aleo, I’ve tried to simplify and explain certain terms where possible so that it can reach the broadest audience of readers. You can read the entire paper here.
Joe D’Aleo, an AMS Certified Consulting Meteorologist, one of the founders of The Weather Channel and who operates the website ICECAP took it upon himself to do an analysis of the newly released USHCN2 surface temperature data set and compare it against measured trends of CO2, Pacific Decadal Oscillation, and Solar Irradiance. to see which one matched better.
It’s a simple experiment; compare the trends by running an R2 correlation on the different data sets. The result is a coefficient of determination that tells you how well the trend curves match. When the correlation is 1.0, you have a perfect match between two curves. The lower the number, the lower the trend correlation.
|
Understanding R2 correlation |
|
| R2 Coefficient | Match between data trends |
| 1.0 | Perfect |
| .90 | Good |
| .50 | Fair |
| .25 | Poor |
| 0 or negative | no match at all |
If CO2 is the main driver of climate change this last century, it stands to reason that the trend of surface temperatures would follow the trend of CO2, and thus the R2 correlation between the two trends would be high. Since NCDC has recently released the new USHCN2 data set for surface temperatures, which promises improved detection and removal of false trends introduced by change points in the data, such as station moves, it seemed like an opportune time to test the correlation.
At the same time, R2 correlation tests were run on other possible drivers of climate; Pacific Decadal Oscillation (PDO), Atlantic Multidecadal Oscillation (AMO), and Total Solar Irradiance (TSI).
First lets look at the surface temperature record. Here we see the familiar plot of temperature over the last century as it has been plotted by NASA GISS:
The temperature trend is unmistakeably upwards, and the change over the last century is about +0.8°C.
Now lets look at the familiar carbon dioxide graph, known as the Keeling Curve, which plots atmospheric CO2 concentration measure at the Mauna Loa Observatory:
A comparison of the 11year running mean of the USHCN version 2 annual mean temperatures with the running mean of CO2 from CDIAC. An r-squared of 0.44 was found.
The results were striking to say the least. An R2 correlation of only 0.44 was determined, placing it between fair and poor in the fit between the two data sets.
Now lets look at other potential drivers of climate, TSI and PDO.
Scafetta and West (2007) have suggested that the total solar irradiance (TSI) is a good proxy for the total solar effect which may be responsible for at least 50% of the warming since 1900. To test it, again the same R2 correlation was run on the two data sets.
In this case, the correlation of TSI to the surface temperature record is better than with CO2, producing an R2 correlation of 0.57 which is between fair and good.
Finally. Joe ran the R2 correlation test on PDO, the Pacfic Decadal Oscillation. He writes:
We know both the Pacific and Atlantic undergo multidecadal cycles the order of 50 to 70 years. In the Pacific this cycle is called the Pacific Decadal Oscillation. A warm Pacific (positive PDO Index) as we found from 1922 to 1947 and again 1977 to 1997 has been found to be accompanied by more El Ninos, while a cool Pacific more La Ninas (in both cases a frequency difference of close to a factor of 2). Since El Ninos have been shown to lead to global warming and La Ninas global cooling, this should have an affect on annual mean temperature trends in North America.
This PDO and TSI to surface temperature connection has also been pointed out in previous post I made here, for former California State Climatologist, Jim Goodridge. PDO affects the USA more than the Atlantic cycle (AMO) because we have prevailing westerly wind flow.
Here is how Joe did the data correlation:
Since the warm modes of the PDO and AMO both favor warming and their cold modes cooling, I though the sum of the two may provide a useful index of ocean induced warming for the hemisphere (and US). I standardized the two data bases and summed them and correlated with the USHCN data, again using a 11 point smoothing as with the CO2 and TSI.
This was the jackpot correlation with the highest value of r-squared (0.83!!!).
An R2 correlation of 0.83 would be considered “good”. This indicates that PDO and our surface temperature is more closely tied together than Co2 to surface temperature by almost a factor of 2.
But he didn’t stop there. He also looked at the last decade where it has been commonly opined that the Top 11 Warmest Years On Record Have All Been In Last 13 Years to see how well the correlation was in the last decade:
Since temperatures have stabilized in the last decade, we looked at the correlation of the CO2 with HCSN data. Greenhouse theory and models predict an accelerated warming with the increasing carbon dioxide.
Instead, a negative correlation between USHCN and CO2 was found in the last decade with an R or Pearson Coefficient of -0.14, yielding an r-squared of 0.02.
According to CO2 theory, we should see long term rise of mean temperatures, and while there may be yearly patterns of weather that diminish the effect of the short term, one would expect to see some sort of correlation over a decade. But it appears that with an R2 correlation of only 0.02, there isn’t any match over the past ten years.
As another test, this analysis was also done on Britain’s Hadley Climate Research Unit (CRU) data and MSU’s (John Christy) satellite temperature data:
To ensure that was not just an artifact of the United States data, we did a similar correlation of the CO2 with the CRU global and MSU lower tropospheric monthlies over the same period. We found a similar non existent correlation of just 0.02 for CRU and 0.01 for the MSU over troposphere.
So with R2 correlations of .01 and .02 what this shows is that the rising CO2 trend does not match the satellite data either.
Here are the different test correlations in a summary table:
And his conclusion:
Clearly the US annual temperatures over the last century have correlated far better with cycles in the sun and oceans than carbon dioxide. The correlation with carbon dioxide seems to have vanished or even reversed in the last decade.
Given the recent cooling of the Pacific and Atlantic and rapid decline in solar activity, we might anticipate given these correlations, temperatures to accelerate downwards shortly.
While this isn’t a “smoking gun” it is as close as anything I’ve seen. Time will give us the qualified answer as we have expectations of a lower Solar Cycle 24 and changes in the Pacific now happening.
References:
US Temperatures and Climate Factors since 1895 , Joeseph D’Aleo, 2008
Persistence in California Weather Patterns, Jim Goodridge, 2007
Phenomenological reconstructions of the solar signature in the Northern Hemisphere surface temperature records since 1600 Scafetta and West, 2007
The USHCN Version 2 Serial Monthly Dataset, National Climatic Data Center, 2007
Discover more from Watts Up With That?
Subscribe to get the latest posts sent to your email.
Many thanks!
This is most interesting, and I look forward to reading the entire paper.
While normally, my inclination is to throw out any r-squared under .90, I have to admit that the PDO-AMO vs USHCN2 plot is breathtaking.
BTW, Anthony, I am in the middle of a “Vista Experience” you wouldn’t believe. If I ever resolve it, I’ll write it up.
Well.
I am not getting the PDO correlation. Is that a problem on the other end? I need to see that graph. Especially from 1980 to present. Because it leads to the next question that is strongly hinted by the Solar graph.
Note how the solar indication dips under the recent temperatures a bit right near the end.
Consider, Rev, how this delineates the contradiction of recent microsite violations and the official record. Is it the same with the PDO ADO map? If the recent temps were lower woulf it all correlate even better?
What if your being right and their being right makes the perfect fit?
Enquiring Minds want to know!
REPLY: I don’t understand what you don’t get about it. Are you saying you can’t see the graphics or that you don’t understand what you are seeing?
The former. I’m getting an X and a white square.
REPLY: well try refresh,clear cache, or different browser, they are all there
This gun smokes.
Larry
Okay. Netscape to the rescue.
So.
Two observations:
1.) When you look at the solar coordination, you see a dip at the end where Solar goes under the most recent readings.
2.) When you look at the ADO/PDO correlation you see a slight dip where ADO/PDO goes under.
But (big BUT) you see the ADO/PDO ‘way over the temp. mark in the 30’s! So why is the temp lower in the 30s in relation to post-1990? Shouldn’t there be the same separation in the 30s as the post-90s?
I.e., what would the righthand scale on the ADO/PDO look like if it were physically lowered by a .25 notch? To bring the 30s in line, but BELOW the current temp measurements. See my drift?
So my previous question applies: Does this imply that the match may be even closer than it seems IF THE RECENT TEMPERATURES WERE ADJUSTED DOWNWARDS TO ACCOUNT FOR RECENT MICROSITE VIOLATIONS?
If so, this would seem to be an analogous confirmation of the hypothesis that recent temperatures have been exaggerated in relation to recent times.
Any thoughts on this from the Rev or the others?
“recent temperatures have been exaggerated in relation to recent times.”
Sorry.
I mean: “recent temperatures have been exaggerated in relation to LESS recent times.”
Evan thats a very valid question. The assembly line of TOBS, FILNET, and other USHCN1/USHCN2 adjustments to raw data does tend to create such offsets.
This would be a perfect test case for removal of questionable sites from the data set too.
Anthony,
This is one of the most significant (and interesting) posts you’ve had since I started reading your blog last June.
And be sure to run them numbers incorporating the preliminary “Watts Adjustment” for site violations averaging, oh, around -0.02C per year since 1980! (To be on the conservative side.)
Perhaps starting around -0.015C (1980) and winding up at -0.025C (2000) . . .
[…] Warming Trend: PDO And Solar Correlate Better Than CO2 […]
Evan, might I suggest that you can test that somewhat with the satellite data from the lower forty eight? RSS:
http://www.remss.com/pub/msu/monthly_time_series/RSS_Monthly_MSU_AMSU_Channel_TLT_Anomalies_Land_and_Ocean_v03_1.txt
(as a warning, I haven’t checked whether the recent corrections made to RSS have effected these images)
On the right. The satellite data of course indicate less warming in the US since 1979. I had a post on my blog, so I have an image here:
http://i23.photobucket.com/albums/b370/gatemaster99/ustemp1.png
And another from the last nine years:
http://i23.photobucket.com/albums/b370/gatemaster99/uslastten.png
A quick look seems to confirm this suspicion.
My goodness, Anthony, don’t confuse people with facts. The Warmists will respond to these new findings by saying that CO2 is driving the temp changes in the oceans (while ignoring the cooling phases).
Religion depends on faith and these pesky cool cycles will be explained away.
Sam, if they were being honest, they’d know that AMO and PDO have any GW signals REMOVED before hand!
Also, in case Evan was wondering, no, the RSS error doesn’t effect those images, but there is not a perfect match with UAH:
http://i23.photobucket.com/albums/b370/gatemaster99/RSScorrectUAHUS.png
Anthony, I, like many, believe natural Pacific Ocean oscillations–ENSO and PDO/IPO–affect global climate, temperature, and temperature records more than climatologists are willing to accept. I’ve graphed PDO and ENSO against global temperature data numerous ways and my results agree with Mr. D’Alea’s for the most part. And I’ve included the AMO in many of these investigations as well. The PDO is responsible for most of the dip in global temperature during the mid 20th century, leading one to question why aerosols are used to duplicate that temperature drop in GCMs. It doesn’t help GCM credibility.
I do, however, have a problem I can’t resolve. Possibly you can help, Anthony. I cannot duplicate Mr. D’Aleo’s graph of the PDO and AMO on Page 6 of the linked ICECAP report, which is your graph above with the title “PDO+AMO vs USHCN2”. As a reference, a graph of PDO data is here: http://jisao.washington.edu/pdo/ Note how the PDO dips drastically in the late 1980s. This drop to negative values is not reflected in the D’Aleo graph. Adding AMO data doesn’t help. Refer to the AMO graph here. http://www.aoml.noaa.gov/phod/d2m_shift/inline_fig.jpg Note the temperature scale of the AMO—tenths of a degree. Compare it to the scale of the PDO—degrees. Since the global area covered by both indices is approximately the same, adding the AMO to the PDO doesn’t raise the sum enough to make the drop in the PDO disappear. Smoothing it with an 11-year running average filter doesn’t make it go away either. For my own investigations, I would love to be able to make that drop in the PDO disappear, but I can’t.
REPLY: Replication is important. I’ll pass this on to Joe so he can provide reference steps.
Anthony,
What is the R2 correlation between the PDO+AMO and TSI? Given that have similar curves, how far out of phase is the PDO+AMO to TSI???? If you were to remove that phase difference by shifting over the curves, what would be the R2 correlation then??? My guess is that due to the thermal mass of the ocean, the PDO should lag the TSI change.
Furthermore, what is the R2 correlation between PDO and AMO? Any phase difference? If PDO and AMO are highly correlated it might suggest that both have the same driver. What is the R2 correlation between the PDO and TSI and same for AMO and TSI? Given the Pacific Ocean is sizably bigger than the Atlantic, I would expect the thermal mass of the Pacific to lag that of the smaller Atlantic. Two different thermal masses will have two different lag times. The interaction of those lag times will cause regional temperature differences.
More importantly do the Southern Hemispheric oceans have their own oscillation frequencies separate to that of the NH?
REPLY: Spot on. I’d suggested the phase lag problem to Joe earlier today, and he’s going to run some checks on that and other suggestions, I’ll put up more when I have it from him.
I think I just might hunt up some more dead trees to use for heating in the next few years. Looks like there might be a climate change on the way!!!!!
Nice Job Anthony and Joe.
Bill
Here is Joe’s answer:
Anthony and Bob
I could not agree with you more. Replication is key to science and if what i did can’t be replicated, I need to know now before I go further.
I downloaded the PDO Mantua data and the AMO unsmoothed Long Term data extended back to 1900 from the CDC sites below. The AMO went back even further than the PDO but I ignored it since Mantua PDO started in 1900. CDC linked to Mantua’s page for the PDO.
I standardized the two sets and then did a 11 point (year) smoothing and then added the two.
I have attached the excel file used. The math and addition was done on the PDO page.
Download the Excel file here
http://www.cdc.noaa.gov/ClimateIndices/
http://www.cdc.noaa.gov/Correlation/pdo.data
http://www.cdc.noaa.gov/Correlation/amon.us.long.data
The final sum was taken into another excel file with HCSN and TSI and CO2 similarly smoothed data for graphing.
Let me know please if you find an error. I have checked this numerous times to be sure.
Joe
P.S.
The difference is that the PDO values are standardized by Mantua which exaggerates the small basin wide values and the AMO values from CDC are unstandardized (raw, unsmoothed so smaller).
By standardizing AMO, I get a similar apples to apples data set to compare. This allows the AMO to offset the PDO dip in the late 1980s.
Standardizing the PDO was not really necessary and as should be the case produced very little change.
Joe: Thanks for the clarification. My problem was assuming the PDO was calculated like the AMO. I ran a quick comparison graph of the raw and standardized AMO data to check the effect of standardizing. It does, in fact, exaggerate the data. That took care of my concerns.
Thanks again.
Thank you for a very interesting article.
The truism that “a correlation does not prove a cause” raises some questions with me. Is it not true that there is a proven causal relationship between CO2 and global warming (even if the size of factor etc is a matter of debate)? As I understand it most of the more reputable sceptics acknowledge that CO2 is a GHG – but argue about the scale. Would it be possible to put in various realistic estimates for the amount of global warming caused by the increase of CO2 which the science really does tell us should happen – and then see how other possible factors might be involved.
Looking at the XL data for PDO Mantua, it appears that the last four months of 2004 are in disagreement with the jisao.washington.edu/pdo/PDO.latest webpage. Also, the Dec 2007 data point is there now.
Cheers,
Bill N
Interesting discussion here. Re. PDO, suggest a look at some of the proxy PDOs from tree rings; e.g. Franco Biondi et at. They, I, and others find a problem with Mantua’s PDO in 1920s-30s. Might be some of the 1936 peak would change if you looked at those proxies. Trees do pretty pretty well as weather stations, if their response can be understood. Have a look at Hamilton 2005 Polar Geography.
OT I have been reading some articles by the Space and Science Research Center. They agree with this article and if I remember correctly reference it.
My Question before I put to much confidence in their conclusions, They being a newly formed organization with limited funding is this. Do they have any degree of reliability? I know that sometimes it is hard to make judgements but I would like to know.
Thanks for your answer
Bill
Martin, very true, Correlation doesn’t prove causation, but tell that to Al Gore! Causation requires to other things: A mechanism demonstrated by experiment or observation (like the Greenhouse Effect) and it also requires changes in one to precede the other. Okay, in this case the connection would be that changes in the ocean influence over-sea weather patterns, which are carried over to the US etc. by wind patterns. The specifics, you may have more difficulty with.
Additionally, “As I understand it most of the more reputable sceptics acknowledge that CO2 is a GHG – but argue about the scale.” Exactly, my friend. The issue really comes down to how the feedbacks add up, or what, if any, effects of GW would slow down or speed up the process. Evidence would seem to suggest that the catastrophe that comes from feedbacks that sum to large positive numbers doesn’t make sense. Given that there are many factors at work in climate, it is important for us to figure out the value of “Climate Sensitivity”, or the response of the climate to doubling CO2. If some of recent warming is natural, then that necessarily pushes down the value of Climate Sensitivity. Even if it isn’t natural at all, you’d need to argue that aerosols, which are not so well understood compared to GHG’s, are substantially cooling the Earth in order to get higher sensitivities than about 1.2 C. Looking at this chart:
http://www.sciencebits.com/files/pictures/climate/IPCC-Forcings.png
I see two things: The first is a “low” level of scientific understanding for most forcings, the second is that there are ways to add them up to negative totals. The aerosol uncertainty is troubling. Additionally, aerosols are concentrated in the NH which has seen more warming, not less. Also troubling. The Oceans, I’ve heard, are supposed to warm more slowly, and SH has more Ocean, I think, but this isn’t why either becuase oceanic trends in the SH are also less than those in the Northern Hemisphere. For these reasons, I find it hard to justify the range of climate sensitivities used in the models, that is from about 2 to 5 C (note that this is not the same as the range that the IPCC thinks the sensitivity falls in, which is 1.5 to 4 C. How odd.)
Bill, the SSRC is a hoax.
[…] sehr interessanten Bericht über die CO2-These fand ich auf diesem Blog: wattsupwiththat Leider in Englisch. Aber ein sehr interessant Beitrag zur Klimahysterie und CO2 – der meine Meinung […]