The Carbon Dioxide Thermometer:

by Jarl Ahlbeck

D.Sc. (Chem.Eng.), docent (env.sci)
lecturer (Mathematical Statistics and Env. Technology)
Abo Akademi University (The Swedish University of Finland)
Process Design Laboratory
Biskopsgatan 8 FIN-20500 ABO-Finland
phone: +358-400-899226


Updated numbers 1979-2000 for satellite measured (MSU) lower troposphere hemispheric and global temperature anomalies, corresponding surface temperature anomalies, atmospheric CO2 concentrations, and emission data for fossil fuels and cement production were investigated by correlation analysis and regression analysis. A high correlation was found between the MSU and the surface record for the Northern Hemisphere. For the Southern Hemisphere, however, the correlation between MSU and surface was statistically insignificant. There was no significant correlation neither between the surface record and the variations in the increase rate of atmospheric carbon dioxide concentration, nor between the emission rate and the increase rate of atmospheric carbon dioxide. But the MSU record explained the variations in the increase rate of atmospheric carbon dioxide concentration to a great extent. A possible explanation to the sensitivity of CO2 concentration to global temperature is the temperature-dependent CO2(atm) - CO2(water) equilibrium. This explanation is supported by the statistical analysis. The MSU record therefore seems to be a much better measure of the real surface temperature anomaly than the surface record, especially for the Southern Hemisphere.

The original data matrix is shown in Table 1.:

Table 1. Six temperature anomaly records: MSU/Global, MSU/Northern Hemisphere, MSU/Southern Hemisphere, Surface/Global, Surface/Northern hemisphere, and Surface/Southern Hemisphere together with atm. CO2 -concentration, increase rate of atm. CO2 and finally emissions from fossil fuels and cement production for 1979-2000. References 1. 2. and 3.

The increase rate of atmospheric CO2 concentration has been 1.533 ppm/year and the CO2 vs. time curve is completely linear as the second order (quadratic) coefficient could improve the multiple correlation coefficient only from 0.998399 to 0.99845. The percentage increase rate has therefore decreased from 0.455 %/yr in the beginning of the 1980:s to 0.415%/yr in the end of the 1990's.

From the original data matrix, the correlation matrix is calculated and shown in Table 2

Table 2. Correlation matrix from Table 1. Interesting high correlations (statistically significant) are marked by * and insignificant correlations by #.

The Northern Hemisphere surface temperature [Jones2] correlates to the Satellite-MSU Northern Hemisphere temperature of the lower troposphere [Christy1] to a great extent, r = 0.8.

For the Southern Hemisphere the correlation between MSU and surface is, however, not statistically significant for the Southern Hemisphere at a 1% significance level because r=0.5. Obviously, both the MSU record by Christy1 and surface record by Jones2 are measures of about the same climate change on the Northern Hemisphere, but these records differ substantially on the Southern Hemisphere.

None of the surface records gives any statistically significant correlation to the increase rate of CO2 (dCO2). The maximum correlation coefficient was r=0.4. The variations of the emission rate do not either explain the variations of the increase rate of CO2 as r=0.1. However, the variation coefficient for the emissions was only 9.7% (stdev/mean*100%) when the variation coefficient for the increase rate of atmospheric CO2 was as high as 33%.

All three MSU records could explain the variations of increase rate of CO2 to a high extent as the correlations coefficients were r= 0.7, 0.7 and 0.6 for global, Northern Hemisphere and Southern Hemisphere MSU vs. yearly increase rate of atmospheric CO2. The Fischer’s-value, a measure of the statistical reliability of the correlation, was 17.5, that should be compared to the critical F-value of 8.1 for p=0.01 (20 degrees of freedom).

As all these three MSU records are strongly intercorrelated, it is not possible to say directly from these three correlation coefficients which of the two hemispheres that is the best explainer. But if the effect would have been merely related to the Northern Hemisphere (biospheric temperature sensitivity), we would probably have seen a strong correlation between the surface Northern Hemisphere temperature anomaly and the dCO2 too because the surface record correlates strongly to the MSU record for the Northern Hemisphere that in turn correlates to the dCO2. But this correlation, r=0.4 is not statistically significant. And we also know that the MSU and the surface record do not correlate significantly for the Southern Hemisphere.

Therefore the MSU-CO2 correlation must originate physically mainly from the Southern Hemisphere ocean region. It also proves that the surface record, especially for the Southern Hemisphere, is flawed because CO2 variations certainly originate from something happening in the ground or in the water, not in the air. This means that the the Global Warming trend in the surface record is wrong.

If the increase of CO2 concentration for one year, dCO2 (ppm), is measured (Mauna Loa), the global "Carbon Dioxide Thermometer" temperature anomaly for one year can be estimated by the regressional formula:

estimated MSU = CDT = 0.23*(dCO2 - 1.53) ± 0.2°C.

where CDT means "Carbon Dioxide Thermometer" reading. Unfortunately, the thermometer has a residual standard deviation of 0.2°C, but it is still better than the global anomaly by Jones2 as you at least now may get the sign right for 64% of the years.

See Table 3. for a comparison between CDT obtained from carbon dioxide measurements and measured global MSU data.

Table 3. MSU lower troposphere anomaly (Global) and Carbon Dioxide Thermometer reading.


There is a trap hidden in these calculations. If the increase of atmospheric carbon dioxide partly is a creeping effect upwards due to enhanced degassing as a result of a real Global Warming according to the surface record, this analysis may give wrong results. This could easily be tested by nullifying the possible creeping effect by replacing the original surface anomaly vector by the residual vector (deviations) around a linear regressional warming trend. Using this residual vector instead of the original surface anomaly vector, did not, however, change the finding at all.


1. John Christy: Monthly Means of Lower Troposphere Channel 2LT.D.

2. Phil Jones et al. Monthly Annual Temperature Anomalies. (2001).

3. C.D. Keeling and T.P.Whorp: Atmospheric CO2 Concentrations at Mauna Loa.

4. Gregg Marland and Tom Boden: Global CO2 Emissions from Fossil Fuel Burning, Cement Manufacture, and Gas Flaring: 1751-1998. (2001).

Return to Climate Change Guest Papers page

Return to `Still Waiting for Greenhouse' main page

FastCounter by bCentral