On July 17, 2007, Rob Wilson archived British Columbia measurement data obtained in 1997-1998 and used in Wilson and Luckman 2002, 2003 (20 ring width; 7 mxd series). The data is here: ftp://ftp.ncdc.noaa.gov/pub/data/paleo/treering/updates. Wilson et al (JGR 2007), a new article, reports on a new tree ring reconstruction from 1750 on, which Rob claims to mitigate the divergence problem. This includes a very slight SI ftp://ftp.agu.org/apend/jd/2006jd008318 , which contains a new version of the BC reconstruction from Wilson and Luckman 2002, 2003, discussed on an earlier occasion here.
On an earlier occasion, Rob made it clear that any decision to archive would owe nothing to climateaudit (a position expressed with a tinge of spite rather uncharacteristic of Rob.) Nonetheless, this is the first measurement data archived by Rob and I appreciate that this has been done. It is too bad that Rob’s friend, Jan Esper, continues to wilfully withhold measurement data used in various publications.
From the recently archived measurement data, I’ve calculated Jacoby-type and RCS-type site chronologies for the 20 RW datasets and the 7 MXD datasets and compared them to the version used in the Wilson et al 2007 divergence article. As you will see, if you take simple averages of chronologies calculated from the archived data, there is a marked ring width divergence problem with values reaching a maximum around 1940 (as with the Schweingruber data considered by Briffa.) The MXD data shows little trend whatever.
Wilson et al 2007 purport to analyze the divergence problem as follows:
We attempt to address the divergence problem’ noted in large-scale reconstructions by developing a new independent TR based reconstruction of ENH temperatures using published local/regional reconstructions and newly updated/sampled TR chronologies that show no divergence against local temperatures …
I’ve done two separate chronology calculations for the B.C. network of 20 RW and 7 MXD datasets. Figure 1 shows a “Jacoby-type” chronology, calculated by attempting to fit a generalized negative exponential (RW), negative line (MXD) or mean to individual cores. There is a marked divergence problem in the RW average, while the MXD average shows little trend. The series carried forward in Wlson et al 2007, in contrast, has a marked 20th century trend.
Next, here are corresponding averages from an RCS-type chronology calculation. When I tried to fit negative exponentials to the data, this failed in nearly all cases. I presume that Wilson got similar results. I accordingly fitted the RCS calculation with a spline and used this for the calculation below. These chronologies show more pronounced centennial variability, but the RW average still has a major divergence problem (with a maximum occurring in the 1930s) and there is negligible trend in the MXD data. Neither shows the 20th century trend used in Wilson et al 2007.
I presume that the difference between the Wilson et al 2007 IBC series and the RW and MXD averages comes from the principal components and inverse regression of temperature against several orthogonal series – a highly questionable procedure (common in dendro circles) criticized in connection with an earlier discussion of Wilson and Luckman 2002, 2003. Wilson et al 2007 perpetuates this methodology for the B.C. series as follows:
Multiple linear regression of three orthogonal principal components (derived from 12 RW and 7 MXD chronologies) were used to reconstruct each climate parameter separately. Calibration explained 64% (Tmax) and 39% (Tmin) of the variance in the instrumental climate record (1895—1991). In this study, we utilize the same data set as Wilson and Luckman , except we removed the Harts Pass (Washington State) RW chronology (1585—1991) from the data set.
These statistical operations are not ones that one sees in econometrics or non-dendro statistics; in my opinion, they have negligible intrinsic validity and to be extremely prone to spurious fits and overfits.
The NRC panel recommended the use of averages in carefully selected tree ring networks as an alternative to principal components, I think that Wilson et al 2007 would have done well to have followed this recommendation. By not following the NAS recommendation, they have simply produced an IBC reconstruction that has little to recommend itself to a third party. Worse, had they followed the NAS recommendation, the resulting RW series would have provided further evidence of the divergence problem, evidence that has been swept under the carpet by their principal components-regression methodology. JGR reviewers should have observed this faulty methodology and required Wilson et al to correct it.
References: Rob Wilson’s publications are listed at http://www.geos.ed.ac.uk/homes/rwilson6/Publications/