A couple of days ago, Greg Laden published a response from Malcolm Hughes to my recent Sheep Mountain article. In today’s post, I’ll show that the “response” was both unresponsive and absurd.
In my article, I had compared recent (“out-of-sample”) values of the Sheep Mountain chronology as shown in a detailed figure in Salzer et al 2014 to the corresponding values of Graybill’s Sheep Mountain chronology, a series discussed on numerous occasions here and in our 2005 articles, since it was the most dominant series in Mann’s North American PC1, which in turn was the most dominant series in the Mann et al 1998,1999. (Indeed, the contribution of all proxies other than bristlecones to the Mann et al reconstruction was indistinguishable from noise,)
In making this comparison, I emphasized the importance of out-of-sample testing as a means of validating a proxy reconstruction. I showed that, after 1980, the bristlecone chronology declined dramatically, while NH temperatures went up.
In response, Laden wrote as follows:
Climate science denialist Steve McIntyre has also weighed in on Salzer et all’s research. His post is truly mind numbing, as he treats Salzer et al as a climate reconstruction paper, and critiques it as such, but the paper examines the methodology of tree ring proxy use and the ecology of tree rings. McIntyre shows the same figure I show above (Figure 5 from that paper) and critiques the researchers for failing to integrate that figure or its data with Mann et al’s climate reconstructions. But they shouldn’t have. That is not what the paper is about. Another very recent paper by the same team is in fact a climate reconstruction study (published in Climate Dynamics) but McIntyre manages to ignore that.
First, I reject the idea that it is somehow an error, let alone a “mind numbing error” to take tree ring chronology information from one article on Sheep Mountain and compare the information to earlier articles. If the authors neglected to do the comparison themselves, I’m entirely entitled to do so.
Nor do the results shown in the Climate Dynamics paper change one iota of my criticism. So it was very misleading for Laden and Hughes to imply that it does. I’ll show this step-by-step.
First, here is the chronology shown in Salzer et al 2013 (Clim Dyn).
Figure 1. Salzer et al 2013 Figure 3 Top Panel.
The above figure is considerably smoothed. Hughes, Salzer and coauthors did not archive the chronology from this figure, but did archive (August 2014) a temperature reconstruction that is a smoothed linear transformation of the above chronology. I estimated the change in scale from the graphic and show that the chronology in the figure and the archive of the smoothed reconstruction match up to linear transformation, other than a somewhat puzzling tail shown in the graphic, but not present in the archive.
Figure 2. Comparison of linear transformation of archived reconstruction (yellow) to Salzer et al 2013 Figure 3 detail.
In the next figure, I’ve overplotted onto the Salzer et al 2013 figure in a similar style as I had used for the Salzer et al 2014 figure. In the left panel, I’ve shown the Graybill Sheep Mountain chronology as used in Mann et al 1998. On the right, I’ve shown the Graybill Sheep Mountain chronology and the Salzer et al 2014 South chronology both in red, comparing them to the Salzer 2013 smooth (black-yellow) and the scaled HadCRU NH temperature that is the “target” of the Mann reconstruction.
Some points are obvious regardless of cavils from Laden or Hughes. Under the presumed relationship between temperature and the Graybill Sheep Mountain chronology, values of the chronology should be nearly triple their long-term mean. However, this hasn’t happened. There are two effects here.
First, even the Graybill chronology (to 1990) went dramatically down after 1980. While the Salzer chronology doesn’t show the dramatic decline of the Graybill chronology, this is because it doesn’t have the same dramatic increase. The Graybill chronology in 1980 is dramatically higher.
Third, the values shown in the archive are highly smoothed, diminishing the information on post-1980 performance. In addition, the archived version of Salzer et al 2013 ends in 2006, while the information shown in Salzer et al 2014 ends in 2010. This may or may not be relevant to the puzzling discrepancy between the downtick shown in Salzer et al 2013 Figure 3 and the archived data.
Figure 3. Annotated version of Salzer et al 2013 Figure 3 top panel. Left. 1000-1980 showing Salzer 2013 chronology (black-yellow) compared to Graybill chronology to 1980, as used in Mann et al 1998. Right – 1875-2010 detail. The Graybill chronology is shown up to 1990 together with the Salzer et al 2014 South chronology as digitized. Magenta- HadCRU NH scaled to the mean and standard deviation of the Graybill chronology over 1902-1980 calibration period.
The Mann reconstruction, either intentionally or by accident, ends in 1980, pretty much exactly at the top of the market of the Sheep Mountain chronology and did not show or discuss the divergence between the bristlecones and temperature that had already occurred in the 1980s. Instead, Mann has repeatedly asserted that his proxies tracked temperature to 1980 and were not impacted by the post-1960 divergence problem of the Briffa reconstruction. Laden repeats this talking point as follows:
Note that there wasn’t a “divergence problem” in Mann et al in the sense of Briffa et al. Mann et al match the observational record very well through 1980, which is the end of the calibration interval (owing to the fact that many proxies drop out after 1980). This is something else the deniers tend to get wrong; they try to conflate the Briffa et al post-1960 divergence problem Mann et al’s hockey stick work. There is no such issue with that work, in that there was no detectable divergence through the end of the calibration interval.
I, for one, have never conflated the issues of the Briffa reconstruction with the issues of the Mann reconstruction and indeed have, on numerous occasions, observed that the inconsistency between these two tree ring based reconstructions was an issue that specialists ought to have addressed more squarely long ago. (And perhaps would have done so had the discrepancy not been concealed in spaghetti graph comparisons which hid the decline in the Briffa reconstruction.) However, if the key proxies in the Mann reconstruction suffer from a post-1980 divergence problem that is even more severe than the post-1960 Briffa divergence problem, I don’t see that this offers any comfort to those pointing to non-divergence up to 1980. Quite the opposite.
In his peroration in Laden’s post, Malcolm Hughes stated:
Back in 1999 we (Mann et al) made the best available choices with the information and data we had. Now, more than 15 years later, with a Bristlecone Pine record that extends back 5000 years, the original results hold up remarkably well.”
I am as aware as anyone that it’s easy to criticize people with 20-20 hindsight. However, even at the time, there were many caveats attached to the validity of stripbark bristlecone chronologies and it is far from evident that the data met the assumptions of the methodology.
But even if that was the “best” that they could do at the time, the proxy has not performed out of sample. End of story. It is beyond ludicrous to say that a proxy that has gone dramatically down since 1980, while NH temperatures have gone sharply up has performed “remarkably well”. Those are the words of grant-seekers, not scientists.