the Air Vent

Because the world needs another opinion

Mann 07 — Proxy Models Part 1

Posted by Jeff Id on July 15, 2010

I’ve been playing around with Mann07 which to my way of thinking is an important paper in Paleoclimatology.  This is a very first step in the process as I only have a few hours into the analysis.  The reason it’s important is that M07 is a demonstration of a lack of signal amplification I show in the hockey stick links above and many others have discussed.  VonStorch and Zorita, who are treated like skeptics for their work on the topic.

I learned more from the SI and code than from the paper.  Before we begin, those of us on the outside owe Mann a thanks for archiving everything for this work.  Of all the climate work you can dig into, Mann is currently setting the standard and his colleagues should follow.  There are people genuinely interested in the methods and data and in this case, he’s done another great job.   Unfortunately for Mike Mann, Steve McIntyre gets points again because he’s the one who forced the situation, we must give credit where due.

All information for this paper, including pdf’s, data and code can be accessed without the paywall at this link:

Figure 5 of the supplimentary info is shown below.  Mann is showing in this paper that  RegEM, an obscure form of what is basically MV regression, does not create signal amplification of his proxy information.  His contention is that this proves that his (and by proxy – some pun intended – paleoclimate) methods are extracting a consistent amplitude signal from the proxy data.

It’s an intriguing problem for me, in particular because the first time I realized what CPS is and later what these regressions are doing, I nearly blew a blood vessel.    It cost me hundreds of hours of my life blogging and reading.   Everything I found confirmed my initial belief that these methods de-amplify historic signal in relation to the modern one.


So, when I see Mann’s graph above, I ask – What the heck is going on?   How come his results are so different from my own.

This will disappoint some here, but I’m not planning on running the full RegEM version any time soon.  It’s a few hours of work with little payout.  CPS correlation methods or simple regression are enough to figure out what is going on here.  If there are enough of the original crowd still hanging around, you will get that both algorithms are methods of elimination and strengthening preferred information.  Preferred being “the most temperature-like”.

My own results from proxies, came up with something like this:

From the link to my hockey stick work in this post above, you know that I used one signal plus noise.   It’s sometimes a criticism of my post above that there isn’t any spatial difference in the signal, but I haven’t bothered doing a whole calculation on spatially noise filled signals because it makes no difference to the problem.  Mann, however, managed to use model data to demonstrate spatial differences in signal and then applied noise before running his methods.

I’m certain his results should be the same as my own, but don’t have proof yet as to why they are not. Of course, I do have an idea though.

But for today, I’ll leave this post with a plot of what is for Mann07’s purposes the noiseless model data of temperature.

My first impression of this temperature network was that it’s awfully tame.  Certainly not very proxy like but it shouldn’t be proxy-like if you believe tree rings are temperature.  This plot is pure temperature so the ‘noise’ of the tree ring proxy needs to be added in, and that’s exactly what Mann07 correctly does.  –The black line is the average.

14 Responses to “Mann 07 — Proxy Models Part 1”

  1. AMac said

    > I’ll leave this post with a plot of what is for Mann07′s purposes the noiseless model data of temperature… [figure]…This plot is pure temperature so the ‘noise’ of the tree ring proxy needs to be added in, and that’s exactly what Mann07 correctly does.

    Some background concepts would be helpful, maybe they are in the links you supplied?

    For the real world, prior to 1850 there is no temperature record, hence the efforts to discover, qualify, and analyze proxy series.

    So “the noiseless model data of temperature” shown in the figure pre-1850 cannot be actual data from actual proxies, somehow stripped of noise. They can only be model proxies, ie pseudoproxies. And the pseudoproxies are constructed according to the investigator’s notions of what signal (temperature signal) looks like (there should be plenty of noise, so constructing signalless pseudoproxies shouldn’t be as difficult).

    But that seems to me to get at a central conceptual problem with this approach. Prior to starting this exercise, the investigator has to make assumptions on what the temperature signal would have looked like in the proxies in question, in the 13th, 14th, 15th, etc. centuries.

    Suppose the investigator examines trees of a certain species (larch, say) at location X (halfway up a Siberian mountainside, say). Location X experiences local conditions (temperature, precipitation, growing season length, wind, shade, soil nutrients, etc.) such that predefined criteria can be used to select larches whose tree rings yield data with a temperature signal for recent times. Suppose that in the modern period, a 0.1 C rise (compared to last year) correlates to a 0.1 mm increase in tree ring width, over a range of average summer temperatures from 8 C to 12 C, say. The investigator can use the principle of uniformitarianism to assert that this relationship held true in the 13th etc. centuries as well, for larches in X-like locations.

    But how does the investigator evaluate whether the particular 13th century trunks s/he is sampling come from trees that were growing at locations that had X-like conditions in the 13th century?

    If the 13th century is much like the present, then X-like conditions probably obtain halfway up the mountainside, as today. But if the 13th century was colder, X-like conditions might have been limited to near the bottom of the slope. If the 13th century was warmer, they might have existed nearer the top of the mountain.

    If the investigator can establish that a 13th century stand of larches were growing in X conditions, it becomes possible to extract signal that says “the 1320s were on average 1C cooler than the 1310s” etc. It seems very challenging to put absolute numbers to such comparisons: distinguishing {11 C in the 1320s, 12 C in the 1310s}(top of the range of X-qualified temperatures) from {8 C in the 1320s, 9 C in the 1310s} (bottom of the range). They’d look the same with respect to tree ring data.

    Sorry this is so long and on such basic questions. Probably a link to on of your earlier posts will provide a response.

  2. kim said

    Yellow bird up high in the cherry tree,
    Who needs stats when data is so plastic?

  3. kim said

    dang, ‘plastique’.

  4. Kenneth Fritsch said

    SI Figure 5 with its reconstruction and uncertainties to the instrumental period, as I have noted previously, shows, by eyeballing, nothing of particular interest over the past 1000 years in the way of past and historic temperature fluccuations. I am not clear on what part of the time series that covers the instrumental period are instrumentally measured temperatures that are tacked onto the reconstruction or are the reconstruction itself. I need to review that article.

  5. John F. Pittman said

    I believe that MM03 showed that psuedo proxies did NOT have the characteristics of the real proxies. I think it was Zorita that used such in their reply to MM, and MM showed that the amplitude problem was not properly discernable using psuedo proxies. IIRC. I just was reading something similar. I believe the Bishop would know, or Steve, of course.

  6. Layman Lurker said

    The explanation must have something to do with the shape / amplitude / variability of the pre-calibration period relative to the calibration period.

  7. Layman Lurker said

    Perhaps if Mann had constructed an unprecedented historical MWP scenario with the same added noise, he would have seen the signal de-amplified.

  8. timetochooseagain said

    Here’s a crazy idea-Mann doesn’t get de-amplification because zero divided by any number is still zero. In other words, as long as you presuppose that the climate didn’t vary much in the past (which is generally what the model he uses shows) then there is no signal to deamplify in the first place.

    If anyone thinks I am way off base, let me know.

  9. Jeff Id said

    #8 I haven’t had much time to play today but the big spikes just prior to calibration didn’t get deamplified. I would expect a dampening of those.

    It would be cheating to tell you guys as to what I’m guessing the problem is. I’ve already started working on it though.

  10. The volcanic signatures are an interesting feature.

  11. stumpy said

    A clear issue I can see, is when mashing all those different proxy series together, if the spikes in the data really are temperature but dont quite line up, they will be averaged out, and I know the timings of those spikes wont be that accurate. This for me is one of the key issues with these kinds of “reconstructions” and really makes them meaningless, unless all the variation observed for each records occurs in the exact year, just a 1 year error between different series in enough to dampen the signal – regardless of the different smoothings and statistical methods employed – its simply a factor of the uncertainty in the dating of the tree rings etc… the patterns will never quite line up, and like a million random ripples, they will all cancel each out to form tiny movements, but for the temperature record where the dates are very accurate, the ripples line up and a clear wave is formed.

    Are they still using 30 year smoothings in the proxy data and 5 year smooths on the temperature record like previous studies? That doesnt help either!

  12. […] Volcanic Sulfur 2010 July 15 Leave a comment Go to comments Jeff Id had a post up at tAV that was taking a preliminary look at Mann 2007 SI Fig 5. I’m no paleo-guy, but there was […]

  13. Bill Stoltzfus said


    I was wondering if anyone had done reconstruction work using a fake instrumental period record–one with no 20th century trend, or possibly even a negative one (like by flipping the instrumental record upside down). Would those reconstructions resemble the now fake temperature trend? I think that would certainly provide some interesting work if you could show that the proxies will match whatever instrumental record you input.

  14. Jeff Id said

    #13, See the hockey stick posts in the header bar above.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: