Amazing what you can do with tree rings

A new Nature article reports a cooling trend for the last 2,000 years suggested by tree ring data.  The cooling trend is statistically correlated with orbital mechanics but uncorrelated with CO2 levels.   Commenters at Watts Up With That quickly pointed out that the sample was regional (Finland) and that tree-ring data articles published to date have been notoriously unreliable.  As one put it:

The divergence problem is a mathematical artifact of calibration.  Formally known as “selection on the dependent variable”, it is a statistical flaw in the methodology that creates bias in the results.  This bias leads to divergence at the calibration boundaries, and misleading results over the proxy period. 

In other words, it isn’t the trees that are at fault.  It is the knuckleheads looking at the tree cores that have improperly applied amplifier technology to statistics, thinking they were inventing a better way to look at noisy data.  What they invented instead was a way to amplify noise, while making it look like signal.  They fooled not only themselves, but most of the world as well.

"Divergence," in the climatology context, refers to the discrepancy between thermometer data and tree-ring temperature inferences in periods when we have access to both.  Thermometer and tree-ring data agreed reasonably well during the last 150 years until about 20 years ago, when they began to diverge sharply.  AGW skeptics infer from this that the thermometer readings are being jiggered, a charge that involves controversies over climate scientists' doctoring of data, refusal to reveal basic data, cherry-picking of temperature sensor sites, and location of too many sites near urban heat islands.  AGW believers infer from the same divergence that tree-ring data are less accurate than thermometer data, and therefore that past warming periods deduced from ancient tree-ring records may have been overestimated.  Part of the excitement over the new Nature article (in both AGW camps, skeptic and believer) is that it does not suffer from the divergence problem; its tree-ring data match up well with thermometer data in the recent period.