Richard S. Courtney writes in comments on the Mann and misrepresentations thread…
In the same week as MBH98 was published I wrote an email on the ‘ClimateSkeptics’ circulation list. That email objected to the ‘hockeystick’ graph because the graph had an overlay of ‘thermometer’ data over the plotted ‘proxy’ data. This overlay was – I said – misleading because it was an ‘apples and oranges’ comparison: of course, I was not then aware of the ‘hide the decline’ (aka “Mike’s Nature trick”) issue.
Unknown to me, somebody copied my email to Michael Mann and he replied.
‘Climategate’ revealed that email from Michael Mann and it can be read here:
Mann’s response consists solely of personal abuse against me and, importantly, it does not address the issue which I had raised immediately upon seeing the ‘hockeystick’ graph. Hence, I am certain that the graphical malpractice of the ‘hockeystick’ was both witting and deliberate.
I’ve reproduced the email below, the redactions were in the linked content that Courtney cites. Mann’s claims about dataset splicing are laughable, as even the Muir Russell investigation (for the later version which appeared in the IPCC TAR) labeled it as such, as McIntyre notes:
Here are Muir Russell’s comments on the IPCC 2001 incident (of which Mann was Lead Author), which they somewhat conflated with the WMO 1999 incident of the “trick” email:
In relation to “hide the decline” we find that, given its subsequent iconic significance (not least the use of a similar figure in the TAR), the figure supplied for the WMO Report was misleading in not describing that one of the series was truncated post 1960 for the figure, and in not being clear on the fact that proxy and instrumental data were spliced together. We do not find that it is misleading to curtail reconstructions at some point per se, or to splice data, but we believe that both of these procedures should have been made plain – ideally in the figure but certainly clearly described in either the caption or the text.
Here is the email Courtney speaks of:
date: Tue, 18 Jul 2000 16:41:12 +010 ???
from: Phil Jones <???@uea.ac.uk>
subject: Re: Global Surface Record Must Be Wrong
>Date: Tue, 18 Jul 2000 10:29:15 -0400
>From: “Michael E. Mann” <???@virginia.edu>
>Subject: Re: Global Surface Record Must Be Wrong
>Cc: ???@geo.umass.edu, ???@uea.ac.uk
>This guys email is intentional deceipt. Our method, as you know, doesn’t
>include any “splicing of two different datasets”-this is a myth perptuated
>by Singer and his band of hired guns, who haven’t bothered to read our
>papers or the captions of the figures they like to mis-represent…
>Phil Jones, Ray Bradley, and Malcolm Hughes dispelled much of the mythology
>expressed below years ago.
>This is intentional misrepresentation. For his sake, I hope does not go
>public w/ such comments!
>>Date: Tue, 18 Jul 2000 08:38:35 +0100 ???(BST)
>>To: Chick Keller <???@lanl.gov>
>>From: COURTNEY <???@courtney01.cix.co.uk>
>>Subject: Re: Global Surface Record Must Be Wrong
>>X-MIME-Autoconverted: from quoted-printable to 8bit by
>holocene.evsc.virginia.edu id DAA27832
>>Your past performance demonstrates that your recent piece to Peter Dietze is
>>unworthy of you. Smears and inuendoes are not adequate substitutes for
>>evidence and reasoned argument. You say;
>>”As to Michael Mann’s “hocky stick” paleo-temperature graph, I realize why
>>many attack it for it puts the nail in the coffen of the argument that
>>recent natural variability is as large as what has been observed in the 20th
>>No ! People attack the ‘hockey stick’ because it is uses an improper
>>procedure to assess inadequate data as a method to provide a desired result.
>>I have defended Mann et al. from accusations of scientific “fraud” because I
>>am willing to accept that this was done in naive stupidity, but I am not
>>willing to accept that is good science. As you say, “people like Mann,
>>Briffa, Jones, etc.” have conducted “careful work”, but doing the wrong
>>thing carefully does not make it right.
>>The ‘hockey stick’ is obtained by splicing two different data sets. Similar
>>data to the earlier data set exists for up to near the present and could
>>have been spliced on, but this would not show the ‘hockey stick’ and was not
>>Also, it is not true to say, as you have;
>>”But, it’s going to take more than rhetoric about Europe’s Little Ice Age
>>and Medieval Warming to get around the careful work of people like Mann,
>>Briffa, Jones, etc.”
>>Nobody in their right mind is going to place more trust in the proxy data of
>>”Mann, Briffa, Jones, etc.” than in the careful – and taxed – tabulations in
>>the Doomesday Book. The Medieval Warm Period is documented from places
>>distributed around the globe, and it is not adequate to assert that it was
>>”not global” because it did not happen everywhere at exactly the same time:
>>the claimed present day global warming is not happening everywhere at the
>>exactly the same time. Indeed, you say;
>>”recent temperature anomalies show that, while the tropics is cooler than
>>usual due to La Niña, the rest of the world is pretty much still as warm as
>>It is historical revisionism to assert that the Little Ice Age and Medieval
>>Warming did not happen or were not globally significant. It will take much,
>>much more than analyses of sparse and debatable proxy data to achieve such a
>>dramatic overturning of all the historical and archaelogical evidence for
>>the Little Ice Age and the Medieval Warm Period. Those who wish to make
>>such assertions should explain why all the historical and archaelogical
>>evidence is wrong or – failing that – they should expect to be ridiculed.
>>All the best
>>>In a recent message to Tom Wigley you wrote:
>>>>”Nowadays, what is measured is mostly quite correct. This holds for the
>>>>counts of frogs, butterflies and for the MSU measurements as well as for
>>>>the ground station readings. What is seriously flawed, are the biased
>>>>*interpretations*. So the surface record may be not wrong at all and
>>>>part of the warming is indeed anthropogenic. Wrong is only the paradigm
>>>>that ground warming is mostly caused by CO2 – and that this warming has
>>>>to show up in the lower troposphere as well. It is striking how the
>>>>ground warming grid pattern coincides with winter heating (Vincent Gray)
>>>>- if the warming was caused by CO2 it should rather be evenly
>>>>distributed over the globe, MSU-detected and only being modified by
>>>>meteorological conditions. Note that this energy caused warming only
>>>>depends on our energy demand and does hardly increase with CO2
>>>>concentration. So this warming should neither be allocated to the CO2
>>>>increment nor be misused with future CO2 projections.”
>>>I have been looking at NCDC plots of global temperature anomalis divided
>>>into three regions- tropics (20N–20S) and the rest of the
>>>globe–(20N–90N) and (20S–90S). When looked at that way, recent
>>>temperature anomalies show that, while the tropics is cooler than usual due
>>>to La Niña, the rest of the world is pretty much still as warm as in 1998.
>>>This is particularly true of northern subtropics and southern subtropical
>>>oceans. The most recent data in fact show the following: for the period
>>>March-May 2000, the northern subtropics are the warmest march-may ever, and
>>>the southern subtropics are essentially as warm as in 1998. Note that this
>>>is not in the winter for either hemisphere. Thus, it would seem to be
>>>important not to make too much of the winter-only observations.
>>>As to Michael Mann’s “hocky stick” paleo-temperature graph, I realize why
>>>many attack it for it puts the nail in the coffen of the argument that
>>>recent natural variability is as large as what has been observed in the
>>>20th century. Gene Parker in the most recent Physics Today just pushed
>>>that point of view citing 20 year-old work as his only support. But, it’s
>>>going to take more than rhetoric about Europe’s Little Ice Age and Medieval
>>>Warming to get around the careful work of people like Mann, Briffa, Jones,
>>>etc. And more recently , Tom Crowley’s article in last week’s Science!!!
>>>Their work includes those acknowledged regional events (LIA and MWP) and
>>>still shows the 20th cent. to be anomalous. (I might add here that it also
>>>calls into question suggestions that solar variability has an additional
>>>indirect forcing amplification since that should have come out of the data.
>>>Instead most published studies show a significant solar influence but a
>>>moderate one.) And so the only way around recent thousand year paleo
>>>studies is for more comprehensive hemispheric and global studies that fill
>>>in acknowledged gaps and in addition show that climate variability is
>>>larger than recent studies show.
>>> Perhaps a more fruitful approach would be to ask what the magnitude
>>>of regional variations has been in the past 150 years. If there are no
>>>regions whose temperature variations were very far from the global average,
>>>then one could legitimately ask how clear anomalies such as the little ice
>>>age could have been sustained in the face of the larger hemispheric
>>>climate. As one example I might cite the eastern United States and perhaps
>>>a large region to the north east since 1940. It clearly has not
>>>participated in the global trend, so much so that urban heat island fans
>>>cite it as an example of how good records (the US) don’t show as much
>>>warming as bad records (the rest of the world).
>>>Charles. “Chick” F. Keller,
>>>Institute of Geophysics and Planetary Physics/University of California
>>>Mail Stop MS C-305
>>>Los Alamos National Laboratory
>>>Los Alamos, New Mexico, 87545
>>>Every thoughtful man who hopes for the creation of a contemporary culture
>>>knows that this hinges on one central problem: to find a coherent relation
>>>between science and the humanities. –Jacob Bronowski
> Professor Michael E. Mann
> Department of Environmental Sciences, Clark Hall
> University of Virginia
> Charlottesville, VA 22903
>e-mail: ???@virginia.edu Phone: (804)??? FAX: (804)???
Prof. Phil Jones
Climatic Research Unit Telephone +44 ???
School of Environmental Sciences Fax +44 ???
University of East Anglia
Norwich Email ???@uea.ac.uk
UPDATE: Steve McIntyre responded in comments, saying he didn’t think the “splicing” issue in MBH98 was a substantial issue for him, and I responded to him, giving my reasons for why I disagree.
For the sake of completeness to discussing this issue, I’m elevating his comment and my response to the body of the post. – Anthony
Anthony, this post has numerous errors, none of which should be made by people interested in this topic. It is very disappointing to read such material.
In Mann et al 1998 (as Jean S first figured out), to calculate the smooth, Mann padded the MBH98 proxy reconstruction after its 1980 end point with instrumental data. Mann only used the smooth up to 1980. This was “Mike’s Nature trick”. Jean S observed the irony of this procedure, given Mann’s protestations against splicing, but the effect was relatively subtle. Contra Courtney’s conflation of “hide the decline aka Mike’s Nature trick, Mike’s Nature trick applied in Mann et al 1998 had NOTHING to do with “hide the decline” – which was an issue with the Briffa reconstruction.
Further, in Courtney’s 1998 email, he said:
The ‘hockey stick’ is obtained by splicing two different data sets. Similar data to the earlier data set exists for up to near the present and could have been spliced on, but this would not show the ‘hockey stick’ and was not done.
In the Mann et al 1998 diagram criticized in Courtney’s email, the proxy reconstruction and the observed data are distinguished by being plotted in different colors or different line type. In other words, they were not “spliced” in the diagram. In Courtney’s recent email to Anthony, he says that the above email “objected to the ‘hockeystick’ graph because the graph had an overlay of ‘thermometer’ data over the plotted ‘proxy’ data. This overlay was – I said – misleading because it was an ‘apples and oranges’ comparison: of course,” I, for one, would never have guessed that this was the criticism being made in the original email. While Mann’s response was marred by his all-too-typical invective, I can well understand why he rejected the allegation in Courtney’s email.
In Courtney’s recent covering email to Anthony, he now characterizes his earlier objection as an objection to proxy reconstructions being plotted on the same graph as observations as follows:
That email objected to the ‘hockeystick’ graph because the graph had an overlay of ‘thermometer’ data over the plotted ‘proxy’ data. This overlay was – I said – misleading because it was an ‘apples and oranges’ comparison: of course, I was not then aware of the ‘hide the decline’ (aka “Mike’s Nature trick”) issue.
While, as noted above, it would have been very difficult, if not impossible, for a contemporary reader to discern this meaning, this criticism is equally invalid in my opinion. I, for one, absolutely do not take issue with plotting a proxy reconstruction on the same scale as observations. I and others take issue with the “divergence problem” precisely because when one plots the Briffa reconstruction against observed temperatures in the 20th century, the two plots diverge. According to Courtney’s criticism, it would be invalid to do such a plot. This is absurd. This does not mean that I endorse the muddiness of Mann’s graphics or other defects. Only that I, for one, do not take issue with plotting a reconstruction and observations on the same scale. On the contrary, it is something that I’ve done on many occasions. As I said to Courtney at CA on this point, if I’m unconvinced on this issue, I can’t imagine why a judge or jury would be convinced.
In the WMO 1999 graphic, Jones deleted values of the Briffa reconstruction after 1950 or so (the decline), spliced instrumental temperature to the end of the record, smoothed the combination and plotted the spliced version (without peeling back to 1950 as in Mike’s Nature trick.)
Muir Russell criticized the truncation and splicing of data in WMO1999 as follows:
the figure supplied for the WMO Report was misleading in not describing that one of the series was truncated post 1960 for the figure, and in not being clear on the fact that proxy and instrumental data were spliced together.
However, he did not take issue with plotting proxy reconstructions and observations on the same graphic. (Not that Muir Russell would be definitive on this.)
There are important issues in connection with the Mann corpus. This is not one of them. Too often, Mann’s opponents make irrelevant and easily rebutted criticisms. Unfortunately, this makes it easier for Mann to avoid more substantive criticisms. For a full explication of the differences between the various incidents, I refer people to the following CA post: http://climateaudit.org/2011/03/29/keiths-science-trick-mikes-nature-trick-and-phils-combo/
REPLY: Thanks for your opinion and clarifications Steve. Bear in mind that Courtney wrote this before the “trick” and truncation was known. While I often defer to your superior knowledge on the subject of MBH98 it is my respectfully differing opinion that plotting the two datasets together (proxy reconstruction and instrumental temperatures) is indeed problematic and misleading in that both techniques have different samplings and sensitivities to temperature.
Instrumental temperature is much more sensitive than tree ring derived proxy temperature, which has a long time domain and is not exclusively a representation of temperature, due to equal if not greater sensitivity to other variables, as I pointed out here: http://wattsupwiththat.com/2009/09/28/a-look-at-treemometers-and-tree-ring-growth/
While Courtney’s complaint is most certainly incomplete in today’s perspective, we shouldn’t just say that plotting two dissimilar datasets on the same chart without proper caveats is a proper practice.
An analog to the spliced combination plot of Mann’s MBH98 graph in today’s climate arena might be this: suppose somebody wants to argue that hurricanes in the NH are becoming more frequent, and they are the more frequent now than in the last 1000 years.
One way to do this is to look at historical reports of hurricanes in literature, newspapers, magazines and other historical writings. These would be a “proxy” for the actual frequency of hurricanes in a given year. Suppose that the researcher was able to find enough reports to to make what looks like a viable dataset, but that instead of using historical writings to determine frequency of hurricanes in the 20th century, the actual record of named hurricanes (essentially observations) was used, such as this graph, which has a nice “hockey stick” shape implying that hurricanes frequency had increased dramatically in the late 20th century.
Arguably, that’s incomplete, showing only the Atlantic, but it’s the best I can do on short notice before I head to work this morning.
The combination of the two datasets, historical literature accounts, plus named storms in the north Atlantic might very well look much like Mann’s flat section of the hockey stick up to about 1925…mostly flat, maybe a slight increasing trend. It would likely look a lot like this graph you plotted in the CA discussion of Besonen et al 2008 (which has other issues independent of this discussion, I’m only using it as an example of what such a graph for this discussion might look like).
To the layman and even to some scientists, they might take such a construct of hurricane historical accounts (proxy) and named storms (observations) as being proof that hurricane frequency is indeed dramatically increasing in the 20th century.
But the issue is sampling and sensitivity. As you’ve pointed out many times, low sampling and/or selected sampling of proxies leads to spurious results when extrapolated to a larger scale (regional to global for example).
From a sensitivity standpoint, since human literature is less frequent as a we go back in time, we’d expect any dataset of historical hurricane accounts to have lower sensitivity to the actual number of hurricanes in any given year simply due to population density and the lack of communications. Many storms would go unreported.
Even in the 20th century data, as shown in the Pew graph above, this effect is likely, due to the early part of the century having lower population, and less ability to observe hurricanes due to a lower level of technology. I talk about this effect in the reporting bias of “extreme weather” here: http://wattsupwiththat.com/2011/04/19/why-it-seems-that-severe-weather-is-getting-worse-when-the-data-shows-otherwise-a-historical-perspective/ So even the Pew graph would almost certainly have a lower representation for named storms in the pre-satellite era.
So, for the purposes of my exercise, knowing that the two datasets for hurricane frequency would have different samplings and sensitivities to actual hurricane frequency in the NH, would it be proper to put these two datasets together into a single graph to argue that hurricane frequency in the NH is the “highest ever” at the endpoint of the graph?
From my viewpoint it would not be, because these two datasets have significantly different samplings and sensitivities to actual hurricanes. The layman doesn’t likely know this, and many of the media that might seize on such a graph probably wouldn’t note this as they often work from press releases. A press release about this hurricane frequency “paper” probably wouldn’t trumpet the fact that the two datasets are greatly dissimilar, and that as you go back in time, the sampling is less, and the sensitivity in the last part of the graph to hurricanes is dramatically higher than any part of the record.
And that’s why I see the splicing in MBH98 as another “trick”. Putting the two dissimilar datasets together implied they have equal sampling and sensitivity to temperature, when they clearly don’t, and the public and the media ran with that visual almost entirely without questioning it, because even though the colors were different, many newspapers back then didn’t reproduce in color, and many people simply take the graph’s “total shape” at face value, without realizing the differences between the two datasets.
(added, here is what a newsprint version of MBH98 might look like…note the dataset delineations disappear, laymen and politicians certainly wouldn’t be able to see beyond the total graph shape in B&W))
To me, that’s just as wrong as the truncation and the overlay issues.
Plotting/splicing two similar datasets of equal sampling and sensitivity in my mind is not an issue. Plotting two greatly dissimilar datasets with unequal sampling and sensitivity, is an issue.