## Thursday, December 27, 2007

### Patrick Michaels & Ross McKitrick: decontaminating climate data

In this weekly dose of peer-reviewed skeptical literature about the climate, we look at a paper called "Quantifying the influence of anthropogenic surface processes and inhomogeneities on gridded global climate data" by Patrick Michaels and Ross McKitrick that was published two weeks ago in Journal of Geophysical Research - Atmospheres.
Full text in PDF
Abstract at the server of JGR
Patrick Michaels's summary
Ross McKitrick's explanation in the Financial Post
Climate Audit discussion
...
The question is whether the land surface temperature data are reliable.

The authors choose a McKitrickish strategy whose earlier versions I found kind of amusing but this particular paper looks rather ingenious to me (even though it is an improved version of their 2004 paper).

They start with the following thesis. If the temperature data really measure the climate and its warming and if we assume that the warming has a global character, these data as a function of the station should be uncorrelated to various socioeconomic variables such as the GDP, its growth, literacy, population growth, and the trend of coal consumption. For example, the IPCC claims that less than 10% of the warming trend over land was due to urbanization.

However, Michaels and McKitrick do something with the null hypothesis that there is no correlation - something that should normally be done with all hypotheses: to test it. The probability that this hypothesis is correct turns out to be smaller than 10-13. Virtually every socioeconomic influence seems to be correlated with the temperature trend. Once these effects are subtracted, they argue that the surface warming over land in the last 25 years or so was about 50% of the value that can be uncritically extracted from the weather stations.

Moreover, as a consistency check, after they subtract the effects now attributed to socioeconomic factors, the data from the weather stations become much more compatible with the satellite data! The first author thinks that it is the most interesting aspect of their present paper and I understand where he is coming from.

These correlations might look irrational but there exist pretty convincing mechanisms that can explain these correlations: the urban heat islands become just one example. If you want to hear another example, a poor country cannot afford to paint the stations too often. They get dark and absorb an increasing amount of light which warms them up.

A part of the claimed correlations may turn out to be accidental or results of spatial autocorrelation but the general lesson that the data from thermometers inside particular stations don't just naively measure the "real" temperature in the troposphere above the region but many other things is certainly a correct and important one.