20th century portion of our paleotemperature stack is not statistically robust, cannot be considered representative of global temperature changes, and therefore is not the basis of any of our conclusions.
Otherwise, their response is pretty much a filibuster, running the clock on questions that have not actually been asked and certainly not at issue by critics. For questions and issues that I’ve actually raised, for the most part, they merely re-iterate what they already said. Nothing worth waiting for.
Nor do they discuss the difference between the results that had (presumably) been in the submission to Nature (preserved as a chapter in Marcott’s thesis).
Nor did they discuss the implausibility of their coretop redating of MD95-2043 and MD95-2011 as discussed here.
Instead of dealing with actual questions, for the most part, their “FAQ” filibusters about questions that no one was asking in connection with this study.
Eventually, they get to a question about their redating:
Q. Why did you revise the age models of many of the published records that were used in your study?
But here too they filibuster. They spend 3.5 paragraphs pontificating about radiocarbon recalibration – a topic that was not an issue. Indeed, on several occasions, including a comment on a similar filibuster by William Connolley, I clearly distinguished between radiocarbon recalibration and coretop redating, taking issue only with coretop redating. Nonetheless, Marcott et al adopted the same filibuster that Connolley had used.
When they eventually get to coretop redating, they say only:
In geologic studies it is quite common that the youngest surface of a sediment core is not dated by radiocarbon, either because the top is disturbed by living organisms or during the coring process. Moreover, within the past hundred years before 1950 CE, radiocarbon dates are not very precise chronometers, because changes in radiocarbon production rate have by coincidence roughly compensated for fixed decay rates. For these reasons, and unless otherwise indicated, we followed the common practice of assuming an age of 0 BP for the marine core tops.
This is nothing more than a re-iteration of the statement in their SI that I had quoted in my Core Tops post:
Core tops are assumed to be 1950 AD unless otherwise indicated in original publication.
Nor did they directly answer questions about the questionable uptick. They pose the question for themselves as follows:
Q: What do paleotemperature reconstructions show about the temperature of the last 100 years?
They give a “short” and “technical” answer. First they stated:
A: Our global paleotemperature reconstruction includes a so-called “uptick” in temperatures during the 20th-century. However, in the paper we make the point that this particular feature is of shorter duration than the inherent smoothing in our statistical averaging procedure, and that it is based on only a few available paleo-reconstructions of the type we used. Thus, the 20th century portion of our paleotemperature stack is not statistically robust, cannot be considered representative of global temperature changes, and therefore is not the basis of any of our conclusions. Our primary conclusions are based on a comparison of the longer term paleotemperature changes from our reconstruction with the well-documented temperature changes that have occurred over the last century, as documented by the instrumental record. Although not part of our study, high-resolution paleoclimate data from the past ~130 years have been compiled from various geological archives, and confirm the general features of warming trend over this time interval (Anderson, D.M. et al., 2013, Geophysical Research Letters, v. 40, p. 189-193;http://www.agu.org/journals/pip/gl/2012GL054271-pip.pdf).
Although they now apparently concede that “the 20th century portion of our paleotemperature stack is not statistically robust, cannot be considered representative of global temperature changes, and therefore is not the basis of any of our conclusions”, this was definitely not the impression left by the authors when the article was published, when it was hailed as supposed confirmation of the Hockey Stick.
Andy Revkin, an extremely experienced reporter, certainly didn’t interpret Marcott et al as taking the position that their reconstruction wasn’t “statistically robust” in the modern portion: he titled his article “Scientists Find an Abrupt Warm Jog After a Very Long Cooling”. (Note: see Revkin’s fresh commentary here).
Further, their present admission that their reconstruction has no statistical skill in its modern portion completely refutes the previous claim that it had somehow vindicated Mann’s Hockey Stick, the distinguishing characteristic of which was supposed to be its “statistical skill” in reconstructing temperatures since 1850 (its calibration and verification periods.) If as Marcott et al now admit, their reconstruction is “not statistically robust” during the Mannian calibration and verification periods, it cannot “confirm” the Mann Hockey Stick.
later in the article, they take another run at this question as follows:
A: Here we elaborate on our short answer to this question above. We concluded in the published paper that
“Without filling data gaps, our Standard5×5 reconstruction (Figure 1A) exhibits 0.6°C greater warming over the past ~60 yr B.P. (1890 to 1950 CE) than our equivalent infilled 5° × 5° area-weighted mean stack (Figure 1, C and D). However, considering the temporal resolution of our data set and the small number of records that cover this interval (Figure 1G), this difference is probably not robust.”
This statement follows from multiple lines of evidence that are presented in the paper and the supplementary information:
(1) the different methods that we tested for generating a reconstruction produce different results in this youngest interval, whereas before this interval, the different methods of calculating the stacks are nearly identical (Figure 1D),
(2) the median resolution of the datasets (120 years) is too low to statistically resolve such an event,
(3) the smoothing presented in the online supplement results in variations shorter than 300 yrs not being interpretable, and
(4) the small number of datasets that extend into the 20th century (Figure 1G) is insufficient to reconstruct a statistically robust global signal, showing that there is a considerable reduction in the correlation of Monte Carlo reconstructions with a known (synthetic) input global signal when the number of data series in the reconstruction is this small (Figure S13).
Again, this is adding little to nothing to what was known more or less on day one. In Marcott’s original email to me, he had referred to the above paragraph in the article, arguing that the above sentence (which stated that the differencebetween reconstruction versions was not robust constituted a statement that the reconstruction itself in the modern portion was not statistically robust. I certainly do not read the original sentence this way and do not agree that it says what the authors now contend that it says. Nor does it clearly and unequivocally say what the authors now say:
the 20th century portion of our paleotemperature stack is not statistically robust, cannot be considered representative of global temperature changes, and therefore is not the basis of any of our conclusions.
Nor do the above caveats of non-robustness properly deal with the dependence of their 20th century uptick on their deletion of 20th century data points from critical series and the importation of earlier data points into the 20th century by unjustifiable coretop re-dating.
As a closing point, in Simonsohn’s widely publicized discussions of academic fraud last year e.g. here, Simonsohn stated that compliant articles ought to disclose “failed” calculations, as well as the ones reported (see Simmons et al 2011 here.) In the SI to Marcott et al, the authors reported on a wide variety of sensitivity studies, most of which will be of little interest to any reader.
However, they conspicuously did not report the results prior to the coretop redating that have been preserved in Marcott’s thesis and were presumably the ones in the article rejected by Nature. This version lacked the distinctive Marcott uptick. Under Simonsohn’s standards, disclosure of this “failed” calculation would have been obligatory. Comparing the two results shows sensitivity to Marcott core re-dating – a topic that ought to have been brought to the attention of reviewers by Marcott et al themselves. The authors are then free to argue for the merits of their coretop redating.
Although the authors now argue that the coretop redating doesn’t “matter”, it clearly did cause a difference between the present results and the results submitted to Nature and clearly did “matter”.
This is not to say that this issue affects “everything” in the paper, which, as the authors now argue, dealt with other issues. However, if these other issues are the ones that really “matter’, as they now say, then they should not have embarked on the coretop redating that resulted in a spurious uptick.
I think that their decision to take guidance from Realclimate was exceedingly unwise. It would have been better to simply respond to questions in a natural way when they were initially raised.
Update 1:51 pm) Andy Revkin has a comment on the FAQ here, pointedly commenting:
there’s also room for more questions — one being how the authors square the caveats they express here with some of the more definitive statements they made about their findings in news accounts.