Tree-rings: Proxies for Temperature or CO2?

Recall step 2 of the  five super-simple steps for building a hockey stick:

Step 2. Select those time series that fit the instrumental (measured) temperature record of choice. Assume that since these time series match the measured temperature in some way, then they are, in fact, temperature proxies.

This step begs the question (in the classical logical sense) about the usefulness of tree-rings as proxies for temperature.  Surely tree-ring width is not solely dependent on temperature, is it? 

What about drought conditions?  Usually we think of droughts as coinciding with high temperatures.  Would higher temperatures cause tree-rings to be thicker even when the tree is being stressed or dying due to lack of water?  Of course not.

What about the abundance of CO2? 

Back when I was a college student I worked at Phytofarms of America in DeKalb, Illinois, USA, which grew the highest quality leafy vegetables in a giant indoor,  innovative, artificially lighted, hydroponic facility.  One of the keys to this industrial sized facility was elevated CO2.  Huge tanks of CO2 pumped up the indoor level to about 1000 ppm, or about 4 times the pre-industrial level.  The resulting veggies were expensive, but they were the best money could buy.

Is it possible that tree-rings are better proxies for atmospheric CO2 than for temperature?   As a simple test, I selected all the tree-ring proxies used for Mann’s 2008 version of the hockey stick and did a simple correlation to the Northern Hemisphere instrumental temperature record and to the atmospheric CO2 record.  The tree-ring data and the instrumental temperature record came from the NCDC archive for Mann’s paper. 

 The CO2 data is a combination of Mauna Loa data (1959 to present) and the Siple Station ice Core (1744 to 1953).  The Siple data was not in yearly increments, so I interpolated.   I also interpolated between the end of the Siple data (1953) and the beginning of the Mauna Loa date (1959).  The Mauna Loa data was truncated beyond 2006 so that the CO2 data would cover the same time domain as the instrumental record used by Mann.

Results

The first graph below (click to enlarge) shows the 30 tree-ring time series with the best correlations to the instrumental temperature record.  Each of these correlations has been matched with the correlation to the CO2 level.  The first thing to jump out is that for 23 out of 30 cases, the CO2 correlation is better than the temperature correlation!

The next graph is the reverse situation.  It shows the 30 tree-ring time series with the best correlations to CO2 in descending order.  As above, each of these correlations is matched with the correlation to the instrumental temperature record.  This time the thing that jumps out is that the correlation to CO2 is better in every single case.  And these correlations to CO2 are not just a little better, they are a lot better.