PAGES2K (2013) unequivocally stated that the Arctic was “warmest during the 20th century”:
The Arctic was also warmest during the twentieth century, although warmer during 1941–1970 than 1971–2000 according to our reconstruction.
McKay and Kaufman 2014 did not withdraw or amend the above statement, instead reporting that the revision amplified the cooling trend prior to the 20th century and had only a “fairly minor impact on the relative variability” of the reconstruction . However, in the corrected reconstruction, the 20th century is only the third warmest. (I do not conclude from their data and methods that this is necessarily true, only that the assertion in the original article is not supported by the revised reconstruction.
Impact of Amendments
Although McKay and Kaufman provided a comprehensive archive of proxy data as used, they did not archive their actual revised reconstruction, making it impossible to directly plot or check the revised reconstruction at the time of my original post on the amendments (see here). I asked Nature to require Kaufman to archive the revisions. Initially, Nature asked me to try to obtain the data from the authors. In response, I asked Nature to enforce their own (commendable) policies on data archiving. A couple of weeks ago, McKay and Kaufman quietly archived the revised reconstruction (see here). None of the parties notified me and the availability of the data only came to my attention a few days ago when I was checking the NOAA archive for additions.
Below is an excerpt from the Arctic panel of the original Figure S2, with annotations. I’ve overplotted (in yellow) the 30-year average of the PAGES-2013 reconstruction, and overplotted the corresponding 30-year average for the PAGES-2014 corrected version in green. In the corrected version, the 11th and 5th century averages are both higher than the 20th century average.
Figure 1. Excerpt from PAGES2k-2013 SI showing the PAGES2K Arctic reconstruction (yellow – 30 year averages), with the corrected version (30 year averages) shown in green.
In the original article, the following was highlighted about the Arctic reconstruction:
The Arctic was also warmest during the twentieth century, although warmer during 1941–1970 than 1971–2000 according to our reconstruction.
This is not true of the corrected version, in which the 20th century is only the third warmest (after the 5th and 11th centuries). I’ve shown code in the comments supporting this statement. Update (Oct 28). Reader MikeN observed that the above statement does not say that the 20th century is the warmest, but that it was warmest in the 20th century. Re-examining the paragraph, the authors were discussing 30-year periods, which are also the points shown in the graphic illustrated above. However, even with this interpretation, the PAGES2K assertion is no longer valid after the changes, though the authors would have been entitled to say the following:
According to our reconstruction, the Arctic was warmer during 1941-1970 than 1971- 2000, with AD1941-1970 values being only slightly lower than the record maximum in AD381-410 and AD1971-2000 values being only slightly lower than in the medieval periods AD981-1010 and AD1011-1040.
I do not conclude anything from these squiggles in terms of the actual past history, only that the above claim is untrue on their data. Because the Arctic reconstruction had the largest Stick of any of the components of the overall PAGES2K reconstruction, the overall average will also be impacted.
Discussion
Although the relative position of the 20th century to earlier centuries was highlighted in the original article, McKay and Kaufman did not retract or amend the highlighted assertion. Instead, they chose to report that the revisions have a “fairly minor impact on the relative variability” of the reconstruction – true enough, but this was not a point of particular import – and that the cooling trend was “amplified” – also true enough, but evasive of the required correction:
Overall, the database revisions have a fairly minor impact on the relative variability in the reconstruction, but they do affect the long-term trend (Figure 2). The primary change is a relative increase in reconstructed temperatures for most of the record, especially between AD 1-1300. This results in an amplified long-term cooling trend that preceded 20th century warming; 0.47 °C/kyr in the revised reconstruction compared to 0.29 °C/kyr in the original. Decadal-scale variability in the revised reconstruction is quite similar to that determined by Kaufman et al. 7; however, the variability is about twice as great in the revised PAGES Arctic 2k reconstruction (Figure 2d). This is likely due the averaging and scaling procedures used in the earlier study [7].
I will discuss their observation that their decadal-scale variability is “quite similar” to that of Kaufman et al 2009 in a forthcoming post, in which I will show that the assertion is tautological, rather than containing fresh information. It turns out that 16 of the 18 most heavily weighted proxies were previously used in Kaufman et al 2009 and that overall 80% of the weight of the PAGES2K reconstruction comes from Kaufman 2009 proxies. Thus, no wonder that their decadal variability is similar.
By far, the largest contribution to the very large changes comes from a single proxy – Hvitarvatn, the orientation of which was reversed in the corrections. (I’ll look at relative contributions in a forthcoming post.) I drew attention to this problem within days of publication of PAGES2K. (While I was writing this post, I noticed a contemporary ClimateBaller sneer at me for “obsessing with little details of individual proxies, ignoring the whole picture as usual”. However, the “little detail” of using Hvitarvatn upside-down had a dramatic impact on the Arctic reconstruction, quite aside from the importance of Hvitarvatn in establishing perspective on both Iceland and regional temperature history.)
Thus far, Kaufman has failed to issue a corrigendum, instead choosing (and being permitted by Nature so far) to report errors in a fresh publication. I can understand why academics prefer to have a new paper on their CV, rather a Corrigendum, but failing to correct errors in their place of publication distorts the research record. Mann got away with this in connection with the flawed no-dendro reconstruction of Mann et al 2008, where errors were acknowledged deep in the SI of a different paper (thus providing cover if challenged), but without any record of the error at the source, thereby permitting results known by the author to be erroneous to continue to be cited even in EPA documents.
In addition, McKay and Kaufman only reported the new paico reconstruction, but did not re-do the other PAGES2K Arctic reconstructions (e.g. “basic reconstruction”). Because the Arctic region is relatively highly weighted in the overall average and because it has the largest Stick, changes to the Arctic reconstruction will have knock-on impact on the overall average, which likewise needs to be recalculated by the authors. I’ve written to Nature, asking them to require a Corrigendum. We’ll see what happens.