Saturday, December 05, 2009

Climate Science: "Even the most rigorous statistical methodology will generate estimates with large margins of error."

Stephen F. Hayward, writing in The Weekly Standard:
It has long been thought that over the last thousand years the earth experienced two significant natural climate cycles: the "medieval warm period" (MWP) centered around the year 1000 and the "little ice age" (LIA) from about 1500 to 1850 or so. The first report of the IPCC in 1992 displayed a stylized thousand-year temperature record showing that the MWP was warmer than current global temperatures, but this was mostly conjecture. Yet it was a huge problem for the climate campaigners: If the medieval warm period was as warm as today, as some scientists believe, it would mean that today's temperatures are arguably within the range of normal climate variability, and that we could not yet confirm greenhouse gas emissions as the sole cause of recent increases or rely on computer climate models for predictions of future climate apocalypse. There had long been rumors that leading figures in the climate community believed they needed to make the medieval warm period go away, but until the CRU leak there was no evidence besides hearsay that scientists might be cooking the books.

The evidence for the medieval warm period and the little ice age is mostly anecdotal, since there were no thermometers in the year 1000. Is there a way we could determine what the temperature was a thousand years ago? Calculating the average temperature for the entire planet is no simple matter, even today. This is where the paleoclimatologists at the CRU enter the picture. The CRU circle set out to "reconstruct" past temperature history through the use of "proxies," such as variations in tree rings, samples of centuries-old ice drilled out of glaciers and polar ice caps, lake sediment samples, and corals from the ocean. Using a variety of ingenious techniques, it is possible for each of these proxies to yield a temperature estimate at a particular location. Tree rings are thought to be the best proxy, because we can count backwards and establish the exact year each ring formed, and by its width make temperature estimates. But tree ring data are very limited. There are only a few kinds of trees that live a thousand years or more, mostly bristlecone pines in the western United States and a few species in Siberia. The thousands of data points that emerge from these painstaking efforts are not self-explanatory. They need to be adjusted and calibrated for latitude, altitude, and a number of other factors (such as volcanic activity and rainfall during the period). Even the most rigorous statistical methodology will generate estimates with large margins of error. One of the striking features of the CRU emails is how much time the CRU circle spent discussing with each other the myriad problems with processing these data and how to display them to a wider world. On the one hand, this is typical of what one might expect of an evolving scientific enterprise. On the other hand, these are the selfsame scientists who have insisted most vehemently that there is a settled consensus adhered to by all researchers of repute and that there is nothing left to debate.

No comments: