Kaufman and McKay recently and quietly issued an Arctic2K correction file at NOAA xls here (now here) that concedes yet another upside-down series previously pointed out to them at Climate Audit. Once again, they used information from Climate Audit without acknowledgement or credit (see NSF definition of plagiarism here).
Previously, Kaufman and McKay had plagiarized Climate Audit commentary on Hvitarvatn (used upside down by PAGES2k), Igaliku (contaminated data used by PAGES2K) and Kepler Lake (same d18O behaviour as rejected Mt Logan) in the preparation of McKay and Kaufman 2014 (Nature Geoscience). Kaufman had previously plagiarized Climate Audit commentary in his corrigendum to Kaufman et al 2009.
As part of my commentary on McKay and Kaufman 2014, I observed that, even after correcting upside-down use of Hvitarvatn, ocean sediment series P1003 (d18O) was used upside-down to the orientation published by the author (which accorded with usual paleoclimate interpretation). The new list of corrections states of P1003: “Relation between water temp and d18O is inverse; change from postive to negative”
Because the PAICO method used in Arctic2K (see CA here) gives very slight weight to series with few data points (such as P1003), upside-down P1003, unlike upside-down Hvitarvatn, has relatively little impact on the Arctic2K reconstruction. However, as shown previously at CA, the impact of the errors conceded in McKay and Kaufman 2014 is very substantial: a corrigendum to PAGES2K should have been issued and, to borrow a phrase, it’s a “travesty” that it hasn’t been issued.
In previous discussion of the Igaliku series contaminated by modern agriculture but still used by PAGES2K (first identified at CA almost immediately), ClimateBaller Nick Stokes contested the contamination, claiming that pollen accumulation rates were unimpacted by modern agriculture – see here). Even the original authors reject Stokes claim. The list of corrections contains the following pers.comm. from the Igaliku authors:
the massive use of N fertilizers has a strong effect on pollen productivity since the 1980’s. The two top values are clearly outliers and should be removed (mid depth 1.2cm and 3.1 cm). The value just below at 5.5 cm (~ AD 1966) occurred just at the beginning of the effects of recent farming (slight soil erosion due to grazing) but before the major ecological shift at AD 1988. I think we can keep this data point.” C Massa, pers commun, March 11, 2014
It appears to me that the value at 5.5 cm is also contaminated by modern agriculture. I am not reassured by the statement: “I think that we can keep this data point”. In my original discussions, I had argued that singleton proxies should not be used in assessment reports, since there was insufficient knowledge of their behaviour: a population of like proxies ought to be required in order to carry out assessments. the Igaliku example simply shows the wisdom of such a policy. Because the Igaliku series has few data points, it is lightly weighted in the PAICO reconstruction (though not necessarily in other variations) but it ought never have been used.
Climate Audit 
21 Comments
I’m confused.
I thought that I’d identified two key points in this scenario.
(1) That Kaufman, McKay and Nick were members of the climate consensus tribe, and
(2) that anyone questioning or debating the consensus position would automatically be labelled as anti-science!
So, which of the aforementioned triumvirate are pro-science and which are anti-science?
Enquiring minds need to know.
” Nick Stokes contested the contamination, claiming that pollen accumulation rates were unimpacted by modern agriculture”
That wasn’t my claim. I said that the authors had said of the pollen accumulation rate “pollen accumulation appears to document climatic changes of the last millenia nonetheless”. Kaufman had stated ex ante rules to the effect that proxies would be included if the authors stated that they were proxies for temperature, and I supported this decision, and criticised CA’s inconsistency in abjuring ex post decisions elsewhere, while making a clearly ex post criticism here. That was all rehashed here. The Igaliku change mentioned here is the one covered there.
I see that Kaufman has stuck to his rules. In the XLS file linked above, each change is supported, not by his interposed judgment, but by reference to material from the original authors.
Steve: Nick, this is old ground and you again are making false accusations, as you do elsewhere. Ex post screening by temperature correlation has been criticized here for valid reasons, but the issue here is not using contaminated data. That you are obtuse to this is nothing more than obtuseness on your part. If the criteris developed such as Kaufman end up including contaminated data, that means that their criteria were stupid – an obvious point that you should clearly endorse. But racehorse away. One way to avoid stupidity is to avoid the use of singleton proxies in assessments – something that I’ve consistently urged and which, in this case, would have avoided the problem on other grounds.
Nick,
Do you think contaminated proxies should ever be used in a reconstruction? It sounds an awful lot like you are suggesting any pre-stated rule, even one which leads to the inclusion of bad data, is always OK. Perhaps a better rule would be ‘never include contaminated or upside down data in a reconstruction, and discount any reconstruction that does’.
SteveF,
A rule has to incorporate a decision process. If you look hard enough, you can find some element of contamination in just about anything. Who decides?
The PAGES rule delegates the decision to the author and journal process. They think about contamination too. The rule says in effect that no-one else is likely, in general, to make better decisions. I think that is plausible.
Steve: yawn. once again, any system that results in use of contaminated data and data upside down is not a good system and, in the real world, would be laughed at. If you think that this is really good, you might as well get your clown suit and red nose on.
Nick,
Who has more reason to err on the side of thinking that a questionable proxy is OK than an original author who uses that proxy? I can think of no WORSE a rule for making a decision, unless your goal is to uncritically include doubtful proxies. If someone is going to go to the trouble of preparing a comprehensive review, they should at least be *more* critical than original study authors. Otherwise, you could just as well write a list of original study references and be done with it. Excluding rubbish proxies should be a consistent objective of every study; if the original authors don’t get it right, and it is clear they sometimes don’t, then someone ought to.
SteveF,
“If someone is going to go to the trouble of preparing a comprehensive review, they should at least be *more* critical than original study authors. “
So would you prefer Kaufman to offer as an ex ante rule:
“I’ll read through all the papers in the area and decide which ones seem OK to me”?
You seem to forget that it isn’t just the author’s judgment. The papers have already been reviewed.
I’m not sure this is that cut-and-dried, nor that SM would say it is either…
After all, for example, when listening to a cockpit voice recorder or something, there can be evidence that is clear and then later garbled, but that’s no reason to toss out the whole thing. Like-wise, there could be a set of emails with evidence on it, but perhaps not the 100% full conversation of all participants… And other examples. I do not think there’s disagreement that a data stream with contamination is automatically all bad data.
However, in this particular example, it seems quite questionable to try to weave in and out of a range of contaminated samples to identify something that might remain reasonable, especially if it’s a proxy. Yes, obviously in the literature it becomes an originating author’s judgment call (one that specialists will later avail themselves)… but it can also be one that specialists can/have criticize/set-aside too.
It would be a shame if what determines your approach is what the data looks like after you make your decision.
Have you ever ‘endorsed’ proxies or made a list of those you like? Might be useful…
I know you are asking Steve, but there aren’t any good proxies for determining temp that I know of. If there were any good ones, they have probably been so contaminated by scientists choosing preferred ones and pre-filtering that I doubt we would ever really know. Even those ‘good’ ones would probably only be good enough to make the most general claims like – the trees were happier in this timeframe than in an immediately adjacent one. The data is not of good quality but it would be awesome if it were. What a cool window into the past.
Hasn’t the problem been that through Mannomatics they’ve been a cool window into the past all through? 🙂
Even the old thermometer readings have been altered…
Jeff Id wrote:
“…there aren’t any good proxies for determining temp…”
There aren’t any good ways to verify that a proxy is adequately tracking temperature over long time intervals without resorting to circular reasoning, ex ante selection, and/or unsupported assumptions. Some proxies show impressive fidelity to local temperature over periods of a few decades.
I am not really up on the history of temperature measurement, but it is probably sufficient to say that the behavior of a column of mercury in a capillary “thermometer” was originally DEFINED to be the measurement of temperature, and then other measurement methods, such as thermocouples, were calibrated with respect to the mercury column. I mean, duh, that’s the only way to do it. So now, how well do tree rings, lake sediments, etc., calibrate against mercury thermometers? Well, they do NOT calibrate “straight up”, if you get my drift. Maybe, if we could collect rigorous data on each of these “proxies” for a few more hundred years, during the mercury thermometer period, MAYBE a calibration relationship might emerge, no assumptions and no data-manipulation or data-neglecting required. Maybe. Wake me when that happens.
Thank you for the link to the NSF’s definitions of academic misconduct. They define plagiarism as “…the appropriation of another person’s ideas, processes, results or words without giving appropriate credit.” I can believe that Kaufman and McKay have appropriated your ideas; can you actually prove it? Or should I be asking them whether they can prove that they haven’t done so?
Pledjurism inna litchurchur? Time for tea with the Dean.
=============
“Kaufman and McKay recently and quietly issued an Arctic2K correction file at NOAA xls here that concedes yet another upside-down series previously pointed out to them at Climate Audit.”
Steve : do you keep a list of the papers withdrawn and corrigendums issued due to your work at CA.
If they use matrix operations of whatever sort, they themselves can combine the series upside down or maybe even sideways, so does it matter how they turn them in the input?
Please Steve, can you put some light on this matter.
I am not expert in the matter but have played a little with http://bluebit.gr/matrix-calculator/ to see what a SVD did.
Steve: the PAICO method used in the Arctic2K does not use principal components (SVD) and the prescribed orientation matters. As I’ve said on many occasions, I strongly oppose multivariate methods in which the orientation/sign is not prescribed/known ex ante. Ex ante knowledge of the sign ought to be a prerequisite of something being a “proxy”.
I am not really up on the history of temperature measurement, but it is probably sufficient to say that the behavior of a column of mercury in a capillary “thermometer” was originally DEFINED to be the measurement of temperature.
Steve,
After all your efforts finding problems/errors (and others I assume) have Pages2k updated their paper and corrected their graphs??
I was quiet surprised that the MWP and LIA seemed to be neutralised in the Pages2k graph with an upkick in the 20th century like Manns hockey stick. Have the MWP and LIA been neutralised?? please send a link if the graph has been redone.
Thanks,
Col
No. I asked Nature to require them to issue a corrigendum, since errors had been admitted in McKay and Kaufman, but Nature refused to require a corrigendum and the original PAGES2K data continues in use.
Nature willingly diverging from nature.
==============
2 Trackbacks
[…] also notes that scientists he has criticized in the past have been sneaking corrections into their work–corrections Mac has called for. He notes that they manage to use his corrections without […]
[…] https://climateaudit.org/2015/03/17/another-upside-down-pages2k-non-corrigendum/#more-20711 […]