In today’s post, I’m going to show Marcott-Shakun redating in several relevant cases. The problem, as I’ve said on numerous occasions, has nothing to do with the very slight recalibration of radiocarbon dates from CALIB 6.0.1 (essentially negligible in the modern period in discussion here), but with Marcott-Shakun core top redating.
Marcott et al re-dating appears to originate from the following:
Core tops are assumed to be 1950 AD unless otherwise indicated in original publication.
MD95-2043
The most recent radiocarbon date for this core was at 14 cm. The radiocarbon reading (1980) was converted by the authors using CALIB4.1to 1527BP. Marcott showed CALIB 6.0.1 upper and lower bounds, the average of which was 1535BP. Marcott showed a date of 1535BP for the alkenone sample at 14 cm. The differences between CALIB4.1 calibrated dates and CALIB6.0.1 calibrated dates in the last two millennia are negligible – single digit years.
The sedimentation rate between the final two radiocarbon points of MD95-2043 was 26.7 cm/kyr. The original authors (Cacho et al 1999;2001) dated samples above the most recent radiocarbon date by assuming a continuation of these sedimentation rates, thus interpreting the top 14 cm as covering the period from 1537BP to a coretop of 1007.6 BP. This implies core loss from the piston core of about 26 cm, a fairly typical result. (Box cores are used when recovery of surface sediments is desired.)
In contrast, Marcott et al 2013 dated the coretop to 0BP (1950 AD) and interpolated dates back to the radiocarbon date at 14 cm. In effect, Marcott et al presumed that the sedimentation had fallen to one-third of previously observed rates (“unprecedented”?). The difference is shown in the graphic below, which compares the age-depth according to the original authors to the age-depth of Marcott et al. It shows that there is negligible difference between published and Marcott radiocarbon dates.
The next graphic compares the temperature estimates using published dates to temperature estimates using Marcott et al dates. The two are essentially identical up to just before AD500. But whereas the original authors date the next few cm to late first millennium, Marcott et al dilate the data to go from AD500 to AD1950, reassigning the core top from AD943 to AD1950.
Marcott state in their archive – in my opinion, with unwarranted optimism – that there is zero uncertainty in the 0 BP dating of the core top.
Figure 2. MD95-2043 Temperature Estimates.
I do not take the position that non-specialists such as Marcott, Shakun, Clark and Mix are precluded from arguing that core dating by the original authors was incorrect and proposing their own alternative dating with the reasons laid out in the sunshine for relevant specialists to assess. However, that’s not what happened here. There was no listing of re-dated cores with the before and after core top dates. The only hint of anything going on with core tops was the following statement in the SI:
Core tops are assumed to be 1950 AD unless otherwise indicated in original publication.
This seems innocuous enough on the surface, but that’s not what Marcott et al actually did. For example, here is an excerpt from the NOAA MD95-2043 archive, where the core top is indicated to be 1007.6 BP.
MD95-2011
That core loss at surface in piston cores can be 26 cm or so should not be any surprise. Indeed, it seems to me to be a credit to the carefulness of the drillers that core losses are as low as this. Core loss in piston cores can be approximately estimated if there is a contiguous box core. (Box cores don’t go as deep as piston cores, but have much better recovery of near-surface sediments.)
MD95-2011, used twice by Marcott et al (#8 and #13), is another piston core, but in this case, there is a contiguous box core JM97-948-2A. These cores were discussed at CA in 2007 in connection with Loehle 2007. At the time, Richard Telford mentioned that he thought that alkenone results for JM97-948-2A had been measured but not published. They remain unpublished six years later: this is too bad as this would have provided an important addition to very scarce high-resolution alkenone records.
Pb210 dating confirmed the modernity of the top 10 cm of box core JM-948-2A, which was dated from 5-80 BP (1870-1945AD.) At 30.5 cm, box core JM-948-2A had a radiocarbon calibrated date of 579BP (radiocarbon 940BP), while MD95-2011 had a near contemporary date at 10.5 cm (601 BP – radiocarbon 980 BP). The data from the box core clearly indicates core loss of somewhat more than 20 cm at MD95-2011.
The original authors dated the 10.5 cm of MD95-2011 above the latest radiocarbon date to approximately 50 years, assigning the core top to 510 BP. In contrast, Marcott et al assigned the core top to 0 BP, using the same methodology as MD95-2043. This is illustrated in the age-depth comparisons shown below.
Figure 3. MD95-2011 and JM97-948-2A age-depth diagram.
As with MD95-2043, the Marcott re-dating dilates the time series after 560BP, so that data assigned by the original authors to 510BP is Marcott-dated to 0BP.
MD01-2421 Splice
The above two examples illustrate one face of Marcott et al re-dating. However, there is a second and even more puzzling aspect: their re-dating of cores which actually do have 20th century samples.
I started analysis of this in my post two days ago in which I drew attention to Marcott et al’s truncation of the final three (highly negative) values of the MD01-2421 splice. MD01-2421 is a piston core with considerable core loss. It was combined with contiguous gravity and box cores to yield a continuous record into the 20th century. The modernity of the top portion of the box core was unequivocally confirmed by the presence of a bomb spike, a commonly used marker. Out of all the ocean cores in the Marcott network, this series undoubtedly has the best dated modern material.
In the graphic below, I’ve compared the age-depth diagrams from the original information to Marcott’s re-dating. As noted yesterday, the three most recent values are truncated from Marcott calculations. Oddly the other recent values are dated somewhat younger by Marcott. The sample dated to 1922 AD by Isono et al is dated to 1939AD by Marcott (just missing inclusion in the 1940 roster.) I have no idea what they’re doing here. It is possible that some algorithm has gone awry (as opposed to Briffa’s manual deletion – commonly and incorrectly described as “MIke’s Nature trick”). Again, I am presently mystified. I emailed Marcott for an explanation yesterday, but thus far no response. In my next example, I’ll look at a related series for further clues.
OCE326-GGC300
OCE326-GGC300 was identified in an earlier post as one of two highly negative proxies (the MD01-2421 splice was the other) that had been disappeared from the 1940 network, the removal of which “explained” the 1940 uptick in the alkenone and dependent reconstructions.
This series was published by Julian Sachs, an accomplished specialist. Radiocarbon dates at 0, 2.5 and 12.5 cm all have “modern” radiocarbon dates – dates that are not further resolvable using radiocarbon. (In passing, Pb210 dating on this core would be very useful.) Sachs showed a date of 0 BP for all measurements between 0 and 12.5 cm in his archive here and in a summary here dated the coretop (0.5 cm) to 0BP, an assumption that on its face is identical to the declared methodology of Marcott et al.
However, once again, the Marcott calculation truncated the most recent values. The diagram below is an attempt to summarize the information. The black line shows the Sachs age-depth diagram from here with the coretop (0.5 cm) dated to 0 BP. The black + points show the sample measurements as reported (with unresolved “modern” samples). The red horizontal arrows show the radiocarbon upper and lower dates as archived by Marcott et al. Marcott’s archive also shows radiocarbon dates for 0.5 cm and 2.5 cm which have been set to NA (shown here in blue.) The most recent sample used in the Marcott calculation came from 12.5 cm, with the seven higher samples being excluded.
Figure ^. OCE326-GGC Age- Depth Diagram.
The next diagram compares the two versions as time-temperature plots, as shown below. (Note that Published is black this time and Marcott red.) If the coretop were dated at 0 BP and dates for samples between 0.5 and 16.5 cm interpolated, this would yield a series continuing to the present, with alkenone-indicated temperatures continuing to decline (for whatever reason.) Instead, Marcott’s series ends in the late 18th century and does not contribute to the 20th century network (which is primarily composed of coretops from different centuries and even millennia.)
Figure ^. OCE326-GGC30 Temperature Reconstruction Comparison. Black- published; red – Marcott; blue- excluded.
Summary
Both types of example result in serious errors, though of different types.
At this stage, I can more or less see what they did in the MD95-2043 class of examples. As noted above, if this was done intentionally and with knowledge of the effect of the re-dating on coretop dates, this should have been disclosed with red letter caveats, showing the effect of the redating on affected cores and relevant specialists should have been asked to review the reasonableness of the methodology. (I do not believe that a properly informed specialist would have signed off on this redating, let alone with no caveats.) Nor do I believe that the actual calculations are consistent with the methodological statement “core tops are assumed to be 1950 AD unless otherwise indicated in original publication”, since core tops were re-dated to 1950 regardless of what was indicated in the original publication. This seems just as serious to me as the problem with Gergis et al.
The problems with the cores with actual modern values are over and above this. I don’t think that they manually blanked out modern values (I hope not for their sakes.) I presume that some algorithm went awry, but right now I can’t picture what sort of algorithm would yield the observed truncations. I’m sure that we’ll find out in due course. As noted above, I’ve requested an explanation from Marcott and hopefully he will clarify things.
Not unexpectedly, William Connolley’s reaction to the problems with Marcott dating is the same sort of wilful obtuseness that characterized “professional” responses to Mann’s use of contaminated (and upside down) Korttajarvi sediments. Connolley pretended that this is nothing more than recalibration of radiocarbon dates.
If you think the “secret” of re-dating comes from “McIntyre’s latest analysis” you’ve been sold a pup. The re-dating is in the SOM itself: The majority of our age-control points are based on radiocarbon dates. In order to 67 compare the records appropriately, we recalibrated all radiocarbon dates.
So I’m curious – who sold you this pup? And why didn’t you bother check the original? -W]
On the hand, Paul Dennis, in a thoughtful comment at Bishop Hill, had no difficulty in understanding the difference between core top redating and radiocarbon calibration.
One can only guess at what Marcott et al were attempting to do when they made gross adjustments to core top dates. It is one thing to run a new, for example 14-C calibration, that will make small adjustments to age models but a completely different issue to redetermine core top dates by such gross margins.
As to the title of this post. I was reminded a few weeks ago of the “Junior Birdmen” ditty, which was sung at YMCA and Cubs and camps in the 1950s – a ditty that seemed especially apt for Upside Down Mann, who serves as inspiration for Marcott and Shakun. Its chorus goes:
Up in the air… the Junior Birdmen
Up in the air and upside down
UP in the air… the Junior Birdmen
Keep their helmets to the ground
The Junior Birdmen were apprentice pilots who were flying upside down. The verse talks about “box tops”, which slightly modified to the present circumstances:
And when they make… the grand announcement,
That things are worse than they’ve ever been…
You can be sure… the Junior Birdmen
Have bent their core tops in.Up in the air… the Junior Birdmen
Up in the air and upside down
UP in the air… the Junior Birdmen
Keep their helmets to the ground
