Those who forget the past are condemned to repeat it. The SI for MBH (and Mann et al 2007) had incorrect geographic locations for numerous proxies.
The same error is repeated in Mann et al 2008, a defect encountered by Jeff Id and myself in trying to replicate reported correlations to gridcell temperatures. Nearly 100 “Schweingruber” MXD series have incorrect geographic locations in the SI; strangely occasional “schweingruber” MXD have correct geographic locations. It’s the usual dog’s breakfast.
Mann et al 2008
In the graphic comparing calculated correlations to reported correlations, there was a very high degree of inconsistency between my calculations and reported correlations for the “Schweingruber” gridded MXD series (relative to other proxy classes). To try to understand this, I attempted to examine this data set in more detail.
The Mann et al 2008 SI (SD1.xls) lists 105 Osborn MXD series (rows 910-1014 on the spreadsheet) together with latitudes and longitudes. The SI contains ID references such as “schweingruber_mxdabd_grid91″; unfortunately, although I’m as familiar as anyone with proxy IDs, these IDs do not correspond to any data version that I’ve examined. The SI contains latitudes and longitudes and I used these for calculating grid cell correlations. Because of the inconsistency, I went back and plotted these gridcell locations (as shown below).
These locations do not match any conceivable locations from the Briffa MXD network, or, for that matter, the location map in Mann et al 2008 Figure1 (shown below), where Briffa MXD locations are shown as diamonds. I guess the listing of site latitudes and longitudes in the PNAS Supplementary Information is the “wrong” data set. It would be nice if they archived the “right” SI.
Update: As suggested by Jean S below, I examined the lat-longs in rtable1209 and it turned out that these lat-longs, rather than those in the PNAS SI, are the “right” coordinates.
Provenance of the 105 ‘Schweingruber’ Series
The Mann et al 2008 SI contains no information on the provenance of these 105 series (provenance is sometimes given for other series.) The main article says of this data:
The gridded maximum density dataset was developed differently by Osborn et al. (http://www.cru.uea.ac.uk/~timo/datapages/mxdtrw.htm) [as used by Rutherford et al. (2)] with the specific goal of capturing century to multicentury variability. [2. Rutherford S, et al. (2005) Proxy-based Northern Hemisphere surface temperature reconstructions: Sensitivity to methodology, predictor network, target season and target domain. J Climate 18:2308–2329.]
The 105 gridded series are not actually referred to in the above webpage. The above webpage lists 300+ MXD series constructed by Schweingruber, but does not contain any reference or identification to the gridded data referred to here. (In passing, I had some role in the creation of this webpage, which is quite recent. Briffa did not list the sites or provide an SI listing the sites in any of his many papers on this network. I asked for this information, which they refused to provide. Eventually, I resorted to the FOI process, finally resulting in the identification of the sites. Most – but not all – of the sites have been archived by Schweingruber at ITRDB, with the incompleteness hampering replication efforts.
To my knowledge, the first mention of gridded MXD series occurs in Rutherford (Mann) et al 2005, which describes the construction of 115 (not 105) gridded MXD series as follows:
OSB therefore worked first with the traditionally standardized data at the individual chronology scale and gridded them to provide values in 115 5° by 5° grid boxes (26 available back to A.D. 1400) in the extratropical NH (Fig. 1b). They then developed temperature reconstructions by the local calibration of the MXD grid-box data against the corresponding instrumental grid-box temperatures.
Their Figure 1b, referred to above, is shown below and compares to the sites in Mann et al 2008 (though the match is not exact as 10 sites seem to have disappeared in the translation.)
Figure 3. Location map from Rutherford, Mann et al 2005. Original caption: FIG. 1. Distribution of proxies for the two networks used in this study. …(b) The age-banded MXD network of Briffa et al. (2001), where each dot corresponds to the center of one 5° x 5° grid box.
Although this caption certainly implies that the 115 grid boxes derive from Briffa et al 2001, that publication contains no mention whatever of 115 gridded series. Instead they report on 9 regional composites. As far as I can tell, the 115 gridded series are calculated in Rutherford et al 2005 and not by Briffa. Looking at the definition of “OSB” in Rutherford et al 2005, it is defined as follows:
The version of the MXD dataset used here was compiled using a combination of grid-box estimates based on traditionally standardized MXD records (with limited low-frequency information) and regional estimates developed to retain low-frequency information (OSB; the data in the MXD network are available online at http://fox.rwu.edu/~rutherfo/supplements/jclim2003a).
Unfortunately, the sentence on availability of data is untrue. I was blocked from this website for a long time, but the block seems to have been lifted and I re-visted this website today, which says today (as it has since it was first put up):
MXD Network Contact Tim Osborn
So the representation in the Journal of Climate was totally bogus. The data in the MXD network was never online at the referenced website. I brought this up with Journal of Climate, but, like subprime accounting offices, they simply blew me off. At this time, the 115 (or 105 or whatever) Rutherford, Mann et al gridded MXD versions simply do not exist anywhere online, other than in the versions at Mann et al 2008.
Here we see another inconsistency between the representation in the SI and the “original” data actually provided. For example, the SI states that the end data of schweingruber_mxdabd_grid10 series is 1980. However the “original” data provided at WDCP has truncated values after 1960. This is not mentioned in the main text, but is sort of mentioned in the SI. (In my opinion, this mention hardly justifies the truncation, let alone the removal of the data from the “original” data, particularly when there is no external provenance for the “original” data.
Because of the evidence for loss of temperature sensitivity after ~1960 (1), MXD data were eliminated for the post-1960 interval. The RegEM algorithm of Schneider (9) was used to estimate missing values for proxy series terminating before the 1995 calibration interval endpoint, based on their mutual covariance with the other available proxy data over the full 1850–1995 calibration interval.
The 1960 truncation of the Briffa data has been a sore point here at Climate Audit, being discussed in many posts, going back to the truncation of the inconvenient bits in IPCC 2001. It’s disappointing that it’s still being done. What makes this incident particularly bizarre is that, after Mann deletes the post-1960 values of the Briffa data, he then substitutes “infilled” values and then calculates correlations using the imaginary data. Out of the 484 series that “pass” Mann’s correlation test, no fewer than 95 are these spliced Briffa MXD series which exemplify the divergence problem without splicing; another 71 are the Luterbacher instrumental series.
I’ve just located an old post in which I calculated the average of all Schweingruber MXD series. I’ll do a comparison of that the Mannian RegEM versions in my next post.