D’Arrigo, Wilson et al 2007 is an interesting article on the Divergence Problem, about which I plan to post from time to time. Since we’ve been discussing temperature data recently – with special discussions of GHCN data problems in Russia, China and Australia, I thought that it would be interesting to draw attention to an interesting aside in D’Arrigo, Wilson et al 2007 in which they reject GHCN adjusted data for Dawson, Yukon on the basis that it has inserted an unrealistically large trend.
The discussion comes up in their discussion of the D’Arrigo et al 2004 discussion of Twisted Tree Heartrot Hill, a site discussed previously at CA here and discussed in McIntyre and McKitrick 2003 (Mann used a version of this data which was obsolete even in 1997). D’Arrigo et al observe a “Divergence Problem” in which temperatures rise and ring widths decline in the late 20th century – for which they (plausibly in my opinion) postulate an upside-down U relationship:
The trees at this location, when compared with local climate data for Dawson, Yukon Territory, appeared to have reached a temperature threshold due to recent warming. The positive ring-width/temperature relationship weakened such that a pre-1965 linear model systematically overpredicted tree rings at this site from 1965—1999. A nonlinear model showed an inverted U shaped relationship between this chronology and summer temperatures.
In trying to mitigate the results of the D’Arrigo 2004 analysis, D’Arrigo, Wilson et al 2007 observe that there is a much lesser temperature trend in the Canadian data than in the GHCN adjusted data. They say (their figure below)
D’Arrigo et al. (2004a) utilized a version of the Dawson instrumental temperature record that had been adjusted for inhomogeneities (Global Historical Climate Network (GHCN), V. 2; Peterson and Vose, 1997)… However, the corrected Dawson station data from the Historical Canadian Climate Database (HCCN, Vincent and Gullett, 1999) have less pronounced positive trends than the GHCN version used in D’Arrigo et al. (2004a).
Caption: Fig. 2. Time series plots of mean July—August temperatures for the Dawson meteorological station, Yukon Territory. The data were taken from the homogeneity corrected Global Historical Climate Network (GHCN, Peterson and Vose, 1997) and Historical Canadian Climate Network (HCCN, Vincent and Gullett, 1999) data-sets. The lines denote the linear trend of each record over the 1900—2000 period.
The difference between the GHCN and Canadian trends are 4.42 deg C and 1.38 deg C! For comparison, here are plots of July-August temperatures for Dawson (71966000) as collated from GHCN Unadjusted and Adjusted Versions. The shape of the GHCN Adjusted version matches the plot in D’Arrigo et al 2007, but the version that I downloaded comes forward a little more to the present. The unadjusted version is different again from the Canadian version.
Jerry Brennan observed that GISS does its own adjustments. So here are the adjusted and unadjusted versions at GISS, which are virtually identical and similar to the GHCN unadjusted version. (The two overlay almost exactly so I’ve re-plotted one as dots as well.)
In 2007, D’Arrigo, Wilson et al didn’t think that the GHCN adjusted data was appropriate. We know that Rob Wilson’s a thoughtful guy and D’Arrigo’s a leading authority – who are we to argue with this decision.
One more oddity: all these various versions of the Dawson data end in 1989. You can find Dawson temperature up to the minute online here yet the GHCN and GISS versions don’t appear to have been updated in nearly 20 years.