Guest essay by Danley Wolfe
WUWT posted a piece I submitted last September titled ‘A look at carbon dioxide vs. global temperature’.
The main point I was trying to convey then is the “striking picture” of the actual data showing a complete lack of correlation between atmospheric CO2 and global mean temperature during the ongoing hiatus. The data set is NASA GISS global mean temperature and Mauna Loa/Keeling CO2, from 1959 through March, 2015.
The updated chart below (FIGURE 1) includes seven months of additional data from my last look. The recent months do not change the basic conclusion regarding the hiatus. But I feel there is more to learn by considering more deeply the implications of these data.
The crossplot of temperature versus CO2 [for the period 1999 to March 2015, commonly known as the “pause” or “hiatus”] reveals a shotgun scatter plot(Ref1) (FIGURE 2). Actually this figure says nothing at all about a relationship between atmospheric CO2 and global mean temperature, except that there is lack of any significant correlation (Ref2) That is a very important fact-data-based conclusion. This is the definition of the “hiatus”!
A first order fit of this data yields “an equation” relating temperature to (only) CO2, viz. T = .0024 * CO2 + 13.648; with an R-squared value = 0.033. You “could” use this equation to estimate the temperature rise from a doubling of CO2 (400 to 800 ppm), in this case 0.96oC. You also might be tempted to call this a “climate sensitivity” (in the sense commonly used), but it’s not. Actually it is just nonsense. So, what might be learned from this exercise?
The R squared(Ref3) of 0.033 prima facie tells you this correlation is, well … just meaningless. Therefore, using a 1st order regression is meaningless, as is any calculated climate sensitivity. The spread of data indicated by the standard deviation vs. min-max spread of the data shows the data are simply a scatter, no more.
To further illustrate the point, you might expand the temperature scale (vertical axis) (FIGURE 3). The 1st order regression fit equation of temperature to CO2 remains the same. I know this visual effect is “cheating”, but it helps in making the point.
The IPCC make a robust claim that climate change is “caused” by anthropomorphic / greenhouse gas causes – with a certainty at the “97% confidence level” (… never mind this is a social science Delphi polling of consensus hands, and not a fact-based probability. Having said that, they go on to say we are now “on track” (talking point phrase) for a temperature rise of 2oC, with range of 1.5-4.5oC (AR5) (Ref4), the self designated tipping point. So the obvious inference, therefore, is that AGW is what will be doing the “causing” of temperature to rise above the critical point leading to catastrophic damage to mother earth and all its inhabitants.
The actual data in the plot of temperature vs. CO2 during the hiatus is also shotgun scatter plot, except flatter. The accepted (by the consensus) hypothesis that global mean temperature (the dependent variable) can be explained by or is due to “mainly” a single variable, CO2 is patently false during the 18+ year hiatus. Did CO2 sensitivity go to sleep? Are other variables exactly canceling out the CO2 effect? It is also important to recognize that the Mauna Loa data includes manmade and non-manmade CO2. The policy prescriptions (and most of the agitation) are mainly directed towards reducing manmade CO2, although there is consideration on land use and burning of forests to plant palm plantations (as in Indonesia and elsewhere).
As I understand it, in a proper multiple regression analysis all the important “known” variables (say 6-7 in number) would be included in the regression model and their F stats would tell you the relative significance of each. Then you would adjust the model … eliminating variables to get the “best fit” with suspected variables … of course this doesn’t speak to “unknown variables” which is a different problem. Other variables would include solar incidence, water vapor, other GHGs, ocean temperature oscillation, etc. (A colleague pointed out it’s a little more complicated than this since “significance” in an econometric modeling sense also depends on degrees of freedom.)
We also know that the integrated assessment climate models (IAMs) are deterministic physical models of the climate with built in predetermined physical cause and effect structures. We can say they are wrong based on their ability to explain the data (facts) during this hiatus.
Nevertheless, the lousy R squared³ and apparent zero “fit” does allow us to conclude that during the hiatus, the assumption that CO2 is the major thing driving global mean temperature is not just a lousy hypothesis, it’s wrong and unsupported by the data (fact). We can also say that all of the variability (scatter) in the data is due to “not CO2.”
On April 15, 2015 the House of Representatives Committee on Science, Space and Technology held a hearing on the President’s UN Climate Pledge. I would like for someone to have made the points above with accompanying figures to Congressional types in explaining what the hiatus really means, and then watch to see any shock effect.
1. Engineering Statistics Handbook 184.108.40.206., Scatter Plot http://www.itl.nist.gov/div898/handbook/eda/section3/scatterp.htm
2. Engineering Statistics Handbook 220.127.116.11.1., Scatter Plot: No Relationship http://www.itl.nist.gov/div898/handbook/eda/section3/scatter1.htm
3. Duke University, What’s a good value for R-squared? http://people.duke.edu/~rnau/rsquared.htm
4. IPCC, Fifth Assessment Report (AR5) http://www.stopgreensuicide.com/Ch12_long-term_WG1AR5_SOD_Ch12_All_Final.pdf
NASA GISS global mean temperature http://data.giss.nasa.gov/gistemp/tabledata_v3/GLB.Ts+dSST.txt
Mauna Loa/Keeling CO2, from 1959 through March, 2015 ftp://aftp.cmdl.noaa.gov/products/trends/co2/co2_mm_mlo.txt