REPLY BY MANN, BRADLEY AND HUGHES TO MATERIALS COMPLAINT

   1.      The disclosure of data used in MBH98 at the Nature Supplementary Information ("SI") appears to be materially incorrect. We have identified no fewer than 38 cases in which the series the series listed in the Nature Supplementary Information were not actually used in MBH98 (as evidenced at the UVA ftp site). This is obviously something that a reviewer would have been unable to identify since the UVA site was unavailable until at least the summer of 2002.

This isn't correct. There were, as indicated above, 34 series that weren't used but were listed in our original posted supplementary information and 2 series that were used but not indicated. These differences are detailed in the text file to be posted in our revised supplementary information (mbh98datasummaryadd-delete.txt) that lists the data series that were either mistakenly omitted from the Supplementary Materials (2 series), or mistakenly included (34).

 

2.        There is evidence that some of these series were intentionally deleted. In this respect, we have the text of an email message in which an MBH98 co-author proposed the deletion of a series (arge030) as being "better for our purposes". This series was deleted in the actual calculations, but was shown in Nature SI as being included. Under the circumstances, a full explanation is surely required as to the "purposes" involved.

This is a very distasteful. The email in question (which was mistakenly placed in one of our data directories) describes results of a screening process of the available data used by Malcolm Hughes, based on considerations of mean segment length, replication of chronologies, and other technical details, that was used to determine what tree-ring chronologies met our standards for inclusion in the analysis (see below). The chronologies in question were never used in our analysis (they were simply listed incorrectly in the supplementary data, along with several other series, as indicated in the corrected data list) and to suggest that they were in some way 'deleted' or that there was some ulterior motive to not including them, is deeply offensive. Specifically, the 34 series listed in the file 'mbh98datasummaryadd-delete.txt' mentioned above, were excluded either because they did not meet the criteria listed below, or because, as in the case of ARGE030 they were duplicated by more recent data meeting all criteria and sent to us directly by colleagues. In the case of ARGE030, the replacement series had been standardized more conservatively than ARGE030, and so better served these clearly stated purposes.

The precise standards for our filtering of the ITRDB database were described in detail in the followup publication, Mann, M.E., Gille, E., Bradley, R.S., Hughes, M.K., Overpeck, J.T., Keimig, F.T., Gross, W. , Global Temperature Patterns in Past Centuries: An interactive presentation, Earth Interactions, 4-4, 1-29,2000.
including the specific following discussion:

As noted by MBH98 (and references therein) non-climatic influences related to intrinsic tree growth trends, difficulty in their removal, and the combining of different tree segments, make the lowest frequency (ie, century and longer timescale) variations potentially somewhat suspect in studies that rely heavily on dendroclimatic indicators.

In selecting data from the International Tree-Ring Data Bank (ITRDB), MBH98 set criteria designed to minimize these problems. Starting with the full data bank available in 1997, they identified 1589 site chronologies, each representing a unique combination of species, variable (e.g. ring width or maximum latewood density) and location. Only the 251 chronologies that met the following criteria were retained:

·         Reliable information on the methods used to remove biological trend was available;

·         The median length of the individual segments used to build the chronology was greater than 150 years;

·         The mean correlation of these individual segments with the site chronology was greater than 0.5;

·         The first year of the chronology was before AD 1626, and it contained at least 8 segments by 1680;

·         The last year was after 1970, and there were still 8 segments after 1960.

Of the 251 ITRDB chronologies that met these criteria, 229 were ring-width chronologies, and 22 maximum latewood density.

This article is available online, http://www.ngdc.noaa.gov/paleo/ei/ with the specific discussion above here:
http://www.ngdc.noaa.gov/paleo/ei/ei_nodendro.html

 

3.      MBH98 states that "the dendroclimatic data used were carefully screened for conservative standardization and sizeable segment lengths." We have the text of an email in which the term "wild guess" was used to describe the inclusion or exclusion of certain series. In this case, the deletion of various series is recommended, but this recommendation is not implemented in the work shown at the FTP site.

At the time this work was done (1996, 1997) few  tree-ring data meeting these criteria were available in the ITRDB for Siberia. However, as we were preparing the dataset, we obtained a 1996 book by three of Dr Hughes’ Russian colleagues, containing listed mean tree-ring chronologies for 61 sites in northern Siberia. This was, until very recently the way Russian colleagues made their data publicly available. These are the “Vaganov” data mentioned above. Given Dr Hughes’ long-standing collaboration in field and laboratory with the authors of the Russian book, his detailed knowledge of their working methods, and his own direct experience of working with the same species, with these colleagues, in the same region, he recommended that we fill this major spatial gap by digitizing these data, and also made recommendations on which should be used.. The complainants quote Malcolm Hughes out of context from an email pre-dating our publication. Our procedure was in fact as  objective and  rigorous as possible.

 

4.      MBH have deleted portions of at least 3 series without explanation to the reader. In each case, the FTP site contains a parallel series with the correct data. The deletions include the first 70 years of the Central England historical temperature series, the first 25 years of the Central Europe historical series and the first 2 years of chin04. In the first 2 series, the issue is not merely that early decades were removed but that summer (JJA) data was substituted for annual data without notice to the reader.

Each of these claims is false or disingenuous, and again they are distasteful. The claim that MBH98 unjustifiably eliminated the earlier part of the Central England temperature (CET) record is unfounded. The decision by MBH98 not to use this record  prior to 1730 is in keeping with previous studies (e.g. Bradley and Jones, 1993) and is based on the existence of inhomogeneities that exist in the record prior to the 1730s (Manley, 1953;1974). Manley (1974), for example, describes a gap from 1707 to 1722 in daily data from the region of Central England he was seeking to cover, and used an adjusted version of data from the Netherlands to fill this gap. Since 1722 falls within the stepwise reconstructions for 1700-1980 and 1730-1980, 1730-1980 is the first interval of our step-wise reconstruction that includes the indicator.

The use of the summer mean version of these series was in keeping with the use of these series by Bradley and Jones (1993). The MBH98 reconstruction, furthermore, is completely insensitive to the use (as in Bradley and Jones, 1993) of summer (rather than annual) mean CET and Central European temperature series. We have updated the references for these two series in the revised supplementary information to "Bradley and Jones, 1993" to avoid confusion, and have indicated that summer mean, rather than annual means, of the series were used by MBH98. We find it peculiar that the complainants would raise this particular complaint, when the interval over which their putative reconstruction differs from ours (the 15th century) predates either of these two series in any case.

 

5.      MBH have recently stated that the UVA FTP site is the repository of data actually used in MBH98. This site contains at least 10 series not used in MBH98, including an instrumental temperature series with decreasing 20^th century temperatures.

Nowhere have we indicated that every single individual series in those directories were used in our analysis. The directories include the tree-ring data used to construct PC series of the various proxy networks, as well as the networks of the PC series themselves, and there are some proxy data that are not listed in the supplementary information, and were in fact never used. We do not see why this is a valid point of criticism. We must stress that we were never under any known requirement to post these data in the first place, and only did so as a service to our scientific colleagues. If  Nature had asked for all of the proxy data used at the time we published our paper, we would have, of course, provided all those we were at liberty to pass on.

6.      Some series in the North American PC region occur in duplicate versions. There is no rational description of the reasons for inclusion or exclusion as individual proxies.

This claim is incorrect with one exception. An analysis of the data file of all 415 individual series used (available through the revised supplementary information) reveals two series that are indeed the same: the Jacoby et al Northern Treeline series #11 (as provided to the authors by Gordon Jacoby) and North American ITRDB series 'cana036.txt'. The latter was only used by MBH98 after AD 1450, and elimination of the series from the North American ITRDB dataset used by MBH98 yields no detectable influence on the MBH98  reconstruction (Mann et al, manuscript in review).  Two other series directly provided by David Stahle (series 'swmxdfew11.dat' and 'swmxdfew01.dat') are the same prior to AD 1488. The MBH98 reconstruction, again, is robust to the exclusion of either or both series from the data set (as shown elsewhere).

Indeed, the MBH98 method is, by its very construction,  robust with respect to whether a moderate amount of redundant information is used in the predictor network, so the criticisms here are irrelevant to the question of whether or not the complainants claimed 'correction' of the MBH98 reconstruction is valid (which, as we have shown elsewhere, it most certainly is not).

7.      The SWM region used in MBH98 includes a site (Spruce Canyon CO), which is not in the site roster of the original article and which is in the roster of the NOAMER region. This apparently small point has a significant impact on 15th century values.

The corrected list of proxy data has been provided with our revised supplementary information. Our reconstruction is in no way, despite the claim made by the complainants, sensitive to whether or not this particular tree-ring series is used (as shown elsewhere). Any sensitivity to whether these data are used in the analysis performed by McIntyre and McKitrick almost certainly arises from the fact that the authors have censored approximately 80% of the proxy data used by MBH98 prior to AD 1450 and 70% prior to AD 1500 in their analysis, yielding in principle much greater sensitivity to the small number of remaining data actually used by them over that time interval.

 

8.      Some data versions used in MBH98 were obsolete when the paper was published. Many more versions are now obsolete. The SI refers to WDCP as a reference for tree ring data, but this reference is inaccurate given the material differences between the version used in MBH98 and the version actually archived. Again, the differences can be material. There are other inaccuracies in the references of MBH98: for example, the citation Bradley-Jones 1992 does not contain several instrumental sites referred to in MBH98.

This statement is absurd. We listed the specific data used by us (albeit with some typos, and incorrect references, as noted) in the supplementary information, and provided all of the data on our data site. We did not indicate there, or elsewhere, that all of the tree ring data used were available in the NOAA databank. The authors appear to be unaware of the fact that not all proxy data shared between fellow scientists are placed in the public data archives, nor are same versions of the proxy data always made available publicly. In general, there is no requirement for scientists to post their proxy data in these public archives, although they are encouraged to do so.

In at least one instance that we have pointed out, the complainants have confused certain unrelated data from the NOAA website with data that were in fact never made available by the original authors to the NOAA website, but which were made available to us. The complainants, based on a thorough misunderstanding of how PC series of tree-ring data were calculated, and other mistakes, have replaced our network with a highly depleted version of the network, in which roughly 80% of the proxy indicators used by MBH98 prior to AD 1600 were eliminated by them in their supposed reproduction of our analysis.

The Bradley and Jones (1992) chapter was used as the best available reference for long instrumental temperature and precipitation records. That doesn't mean that every single instrumental record used (and described in our supplementary information) was shown in the article. We have updated this reference with a description of the specific source of the long instrumental data series and means by which they were gridded, in the revised supplementary information.

 

9.      The disclosure of methodology for calculating temperature principal components is inaccurate. MBH98 describe their methodology as "conventional". Conventional principal components fail with missing data. The underlying data set contains much missing data and some other procedure was necessarily used.

This represents the complainants’ fundamental mistake of using a different version of the CRU surface temperature data site from the one that we used. The pattern of missing data in the two datasets is different, owing to the use of different reference periods. Small gaps in the instrumental record we used (an older CRU surface temperature dataset which goes back to 1856) were filled by linear interpolation, as described in our revised supplementary information.

 

10.      The disclosure of methodology for calculating tree ring principal components is inaccurate. Again MBH98 methodology is not "conventional". In this case, the FTP site contains computer programs which show that the data was transformed in ways not disclosed in MBH98. These undisclosed transformations have a material impact on the final results.

Each of these statements in incorrect. A conventional PCA was indeed used. The authors apparently failed to take note of the stepwise procedure used by us, and described in our paper. This procedure allows  PC series to be calculated independently for each sub-interval (e.g. 1820-1980, then 1780-1980, ..., 1400-1980) to allow for the use of an increasing number of data in the different sub-networks increasingly later in time. The misunderstanding of this procedure led to them eliminating roughly 80% of the proxy indicators used by us prior to AD 1600, the primary reason for the spurious result that they have reported. Precise details regarding how the data were standardized are provided in the revised supplementary information. We have shown elsewhere that the MBH98 reconstruction is in fact entirely robust with respect to whether or not the proxy series were standardized by the detrended or raw calibration period variance.

Once again, all of the original proxy data used, and all of the PC series used, were available on the public ftp site from July 2002,  though the complainants did not download and use the correct data. The new, revised ftp site provides the data and listings of data in a thoroughly documented manner such that similar mistakes should not be possible in the future.