Mann's Screw Up #8 - Hypocrisy

So far this series has focused on Michael Mann's original hockey stick. In 2003, Willie Soon and Sallie Baliunas published a paper disagreeing with it. Mann's response is... worth examining:

This is crap of the worst kind–it was written explicitly for political purposes; there is no science there at all

I kid. He said that about the Soon and Baliunas paper, but he also published a scientific paper of his own criticizing it. His first argument in it says:

it is essential to assess proxy data for actual sensitivity to past temperature variability.... The existence of possible underlying dynamical relationships between temperature and hydrological variability should not be confused with the patently invalid assumption that hydrological influences can literally be equated with temperature influences in assessing past climate

That's a lot of words to say it is essential not to mix up temperature and precipitation proxies. In other words, when we look at the data file for a temperature reconstruction, we shouldn't see precipitation proxies listed like:

Southeast U.S-N. Carolina  Dendro ring widths   precip            36N   80W   1005     Stahle et al 1988
Southeast U.S-S. Carolina  Dendro ring widths   precip            34N   81W   1005          "
Southeast U.S-Georgia      Dendro ring widths   precip            33N   83W   1005          "

And we absolutely, positively, should not see instrumental precipitation records like:

Station Precipitation             "             precipitation   12.5N 82.5E   1813          "
        "                         "                    "        17.5N 72.5E   1817          "
        "                         "                    "        37.5N 77.5W   1809          "

Only papers which are "crap of the worst kind" would say those can "literally be equated with temperature" data. In other words, Mann's original hockey stick is "crap of the worst kind." After all, anyone familiar with this series will remember that's where those entries were taken from.

Mann's second argument in his paper says:

It is essential to distinguish... between regional temperature anomalies and anomalies in hemispheric mean temperature, which must represent an average of temperature estimates over a sufficiently large number of distinct regions

This makes sense. We shouldn't conflate regional information with hemispheric information. Data from one area should not be taken to represent temperatures for an entire hemisphere. That's why nobody should ever say:

one such indicator PC #1 of the ITRDB data is found to be essential

For the results of their temperature reconstruction. After all, if one proxy (indicator) is essential for results, there's no distinguishing "regional temperature anomalies and anomalies in hemispheric mean temperature." And only someone who is "crap of the worst kind" would fail to do that. "Crap of the worst kind" like Michael Mann, discussing his original hockey stick.


But okay. Even if Michael Mann is completely hypocritical, having done exactly what he condemns Soon and Baliunas for doing, that's okay. Science progresses. People change their minds. It's not like Mann would promote his work which does this while criticizing Soon and Baliunas for doing it:


Wait, what? Does that figure really say Mann et al, 1999? As in, the paper which contained Mann's original hockey stick? The hockey stick which used precipitation series as temperature series and used proxies from one region as representing temperatures for an entire hemisphere? The hockey stick which did exactly what he condemns Soon and Baliunas for doing?



  1. What bothers me with the graphs is that I cannot clearly identify where the "zero" line comes from. Is it from a 10K data set? A 100K data set? Wouldn't the "zero" line change based on the duration of the history therefore not really represent anything of value to the reader?

  2. The "zero" line is arbitrary. The choice of it doesn't matter because the graphs are showing anomalies. Whether we say something went from -1 to 0 or 0 to 1 doesn't change anything. It's still an increase of 1.

    That said, there is an issue when you display multiple lines together. In effect, both line has its own "zero" line. Displaying them together requires you set those lines equal to one another. Since the lines are arbitrarily chosen, the values you set equal to one another are arbitrary as well. In effect, you can shift each line up or down independently of the rest. How well the lines align with one another is largely a matter of how you choose to lay them over one another.

  3. Hi Brandon,

    The question of the zero line also raises, for me, a question about the error ranges, both accuracy and precision.

    Let's simplify by using a hypothetical example. We study human body temperature and find a typical (normal, average, mean, whatever) value of 37 degree C, plus or minus one degree C. We claim fever for an individual when this value is 2 or more degrees above 37, or an anomaly value of +2. We study cats and find the typical value is 39 degrees. But we find that the range of variation in cats is plus or minus 1.5 C, and so we don't declare a cat's body temperature "high" until we get above 40.5. On the other hand, we might decide the evidence of a disease kicks in for cats at 41.5 -- a ONE degree "anomaly" above the range of normal values. (I'm only about 90% convinced my own "simpler example" is so simple after all...)

    ANYHOW, if I chart the comparisons of fever in cats and humans in anomaly degrees C with some similar condition such as bronchitis, do I get some sort of meaningful comparison?

  4. Pouncer, your example is odd as there's no particular reason we'd expect cats and humans to have body temperature variations of the same magnitude (and many reasons we wouldn't). Also, I'm pretty sure we'd consider normal variations in determining how high a body temperature needs to be before deciding there's a fever. That is, the +2 anomaly would account for the +/-1.5 variations you mention.

    I get what you're saying though. The answer is, maybe. It may be possible to get a meaningful comparison between in a situation like that, but it'd require examining a number of things, not just plotting the two in the same figure.

  5. By the way, I'm currently debating on what my next post should be about. There's a lot of history with the three papers I've discussed in this series so far (MBH98, MBH99 and the "rebuttal" of S&B) I could cover. There's all sorts of things which were said about the hockey stick once Steve McIntyre and Ross McKitrick came on the scene, and the controversy over the S&B paper had a lot of stuff happening behind the scenes. The problem is I don't see any of that as really showing big screw ups by Michael Mann. They're mostly just little things he and others did to hide the things I've already discussed.

    Any thoughts? I kind of want to jump straight to Mann's 2008 paper now so I can speed the series along. I just hate the idea of "ignoring" so many things he's done.

  6. Don't forget to add # 8 to your summary page with all the links (I waited a few days, expecting you to do so...)

  7. "By the way, I’m currently debating on what my next post should be about."

    I am not sure what your next post should be about as you have done a great job thus far. I did note that you intentionally skipped a detailed discussion of the problems with Mann's PCA analysis. For some of your readers who are interested in a not-too-technical description of the issues, I would commend this article: What is the ‘Hockey Stick’ Debate About? Ross McKitrick, Department of Economics, University of Guelph , April 4 2005, found here:

    The article gives a more complete description of the flaws in Mann's application of a non-standard PCA analysis, and how it effects the graph. He also covers the some of the same ground that you cover regarding the sensitivity of the hockey stick graph to certain questionable proxies and the fact that Mann knew that the graph was not robust to their deletion as he claimed as evidenced by the "censored" directories found on his servers.

    On a different tack, it would be interesting to compile a list of unanswered questions that Mann should be asked about when he is deposed, although I would concede that would be yet another large undertaking.

  8. David Jay, thanks for reminding me. So you know, if I don't remember to do something like that right away, odds are I've forgotten and won't do it until something reminds me.

    pauldd, thanks. And yes, that is a good piece. It's probably one of the best summaries I've seen. It's also interesting to read given it was written nearly a decade ago. Reading it shows how little has changed (though at least Mann has mostly released his code).

    Anyway, I should have the next post up on Thursday. I need to do a bit more outlining of how I want to address Mann's 2008 paper, but once I do, the posts should come out pretty quickly. That is, if I can stop being distracted by this card game I'm trying to design. I doubt I'll ever produce the game, but how could anyone not like the idea of D&D played with cards? It's much simpler and more streamlined.

  9. Brandon, I saw a page at ClimateAudit with counts of instrumental temperature, some as precipitation proxies, and I think some are instrumental precipitation. Also, there was I think 3 are temperature reconstructions, delimited in degrees, while others are tree-ring chronologies/proxies.

  10. MikeN, I gave such a count in an earlier post in this series. I didn't attempt to give a full account of things, but even going off just what Michael Mann himself said (in supplementary material), 22 of his 112 proxies were precipitation proxies. 22 of the 112 proxies were instrumental records. 11 of those overlap. That means 33/112 (29%) of his proxies were not what one would normally think of as temperature proxies for a study like this.

    And that's being as favorable toward Mann as possible. I didn't count several proxies I could have, and I didn't discuss any of the issues with proxies that'd require being familiar with the underlying work. The problems with those proxies require more effort to see than I felt was worth putting into that post. I figure if one in three proxies were inappropriate by what Mann himself discloses, people don't need me to find more examples.

    As for ones which were temperature reconstructions, I don't recall MBH using any of those. It's possible I'm forgetting some though. Reconstructions are used as proxies in some multiproxy temperature reconstructions. For example, there's a Yang reconstruction I've seen used as a proxy in at least half a dozen reconstructions. It was even displayed as a proxy in the 2003 EOS paper I discuss in this post. You can see it as the last proxy displayed in Figure 2 (labeled China) of the paper.

    If you're familiar with the sources for series, it's pretty easy to tell if they're temperature/precipitation proxies, reconstructions or what. It's just a pain to try to familiarize yourself with so many proxies.

Leave a Reply

Your email address will not be published. Required fields are marked *