No, it’s not about the NHL lockout. It’s about a February article in the journal, Geophysical Research Letters, that questions what has become the traditional graph of the rise in global temperature: the hockey stick. The hockey stick refers to the shape of the temperature line, which is approximately unchanging (straight) from 1000 to 1900AD and then spikes from 1990 to 2000. The spike is particularly severe in the 1990s. The hockey stick came from a computer model created by a team led by Michael Mann and appeared in 1998 in a Nature article authored by Mann, Raymond Bradley and Malcolm Hughes.
(this image is from the IPCC 3rd assessment report, chapter 2 and describes “Millennial Northern Hemisphere (NH) temperature reconstruction (blue) and instrumental data (red) from AD 1000 to 1999, adapted from Mann et al. (1999). Smoother version of NH series (black), linear trend from AD 1000 to 1850 (purple-dashed) and two standard error limits (grey shaded) are shown.”)
To calculate modern temperature, the authors used instrument readings. To calculate historical temperatures, Mann et al. could not use instruments: there were no thermometers in 1200AD. Instead they relied on data from tree rings, ice cores, corals, as well as historical accounts of temperature.
According to a report in the BBC, it was the assumptions in how this “proxy” data was modelled that posed the problem. The authors Stephen McIntyre and Ross McKitrick in Geophysical Research Letters asserted that the statistical treatment of the tree ring data favoured a hockey stick shape for the temperature curve. The strong bias essentially “flipped the entire analysis”.
(For a detailed explanation of the technique and the controversy, see this realclimate.org post.)
BTW, the controversy has a Canadian connection. Stephen McIntyre is from a company called Northwest Exploration in Toronto and Ross McKitrick hails from the Department of Economics, University of Guelph.
In a subsequent post, I’ll go over the recent political controversy surrounding this computer modelling conflict. For the moment, it shows how model results can hinge on a single data set, whether primary or secondary (“proxy”), and on a single assumption. This isn’t to say that assumptions are bad–they are a necessary component of any model. In this instance the statistical technique chosen is well-respected (couldn’t interpret remote sensing data without it) and extensively backed-up by the literature. And it demonstrates that conflict is a healthy part of scientific advancement. The challenge is whether the scientists in the community are entrenched in a position or open to questioning the assumptions.
Updated post with graph directly from the IPCC report.