How'd They Do That?

Take a look at today’s puzzle.

On the left, I’ve plotted the three EIV reconstructions for the NH hemisphere using the infilled CRU series as a “target”. The data is straight from Mann’s website. One shows the “full” global network, one shows the “full” NH network (both of which I take to really be series selected as discussed in my Full Network post, but it’s always hard to be sure) and the third one is the “screened” NH network. As I understand it, SH proxies are used in the “GL full” network to teleconnect with NH reconstruction and this is the difference with the NH “full”, but again I’m not 100% sure.

The right panel shows the differences between the series, showing the differences in a cycle.

Take a careful look at both panels because lots of questions jump out at me.

First, all three reconstructions go to 2006 even though there aren’t any proxies that go to 2006. How’d they do that? (Of course, we can’t tell by consulting the EIV code, because, contrary to the representaton by PNAS, working EIV code isn’t provided.) Did they splice the instrumental record? Could be. It’s hard to tell.

Second, all three reconstructions are identical in the calibration period. How’d they do that? Is this from splicing or overfitting?

Third, look at the the year 600 in the right hand panel. The black and red series essentially flip orientation. How’d they do that? It looks like there’s a reconstruction step and the regression sign for a series flips over when the composition of the network changes. Whatever it is, it’s not a good sign,

Fourth, what about the green series in the left hand panel. What’s going on in the early portion? How’d they do that?

Just to show that I’m not making this up. Here’s a graphic from the SI in which I think that I discern some of these patterns:

Stranger and stranger.


  1. Posted Oct 7, 2008 at 2:43 PM | Permalink

    Hone your skills Steve I think it is time for the PALEO RECON CHALLENGE! SUDAY SUNDAY SUNDAY! See the best and brightest of the Paleo climate reconstruction team face the evil geologist from the cold North country. In a no holds barred cage match. Can McIntyre pin the Mann? Will Mann say, “Duh, what do you mean inversely indicative?

    This Sunday at a Mosh Pit near you!

  2. Posted Oct 7, 2008 at 3:07 PM | Permalink

    I didn’t use the reconstruction EIV composite data in the calculations on my latest post or my back calculation post because it looked to me like the measured curve was added probably with RegEM tape into the signal. I thought this was done in more than one place.

    I found the CPS curves weren’t extended like the EIV curves and stopped at 1995. This made more sense for the fact that there’s no data in 06. I then tried to back calculate to the CPS curve. It was actually frustrating trying to find a good reconstruction to work with.

    If there was any question, you are definitely not making this up 🙂

  3. Pat Keating
    Posted Oct 7, 2008 at 3:07 PM | Permalink

    What most struck me is that, in the left panel, the red (NH full) is going down, if anything, when aborted. Why is it aborted, if the green (NH screeened) continues on to present-day?

  4. Jean S
    Posted Oct 7, 2008 at 3:08 PM | Permalink

    First, all three reconstructions go to 2006 even though there aren’t any proxies that go to 2006

    Second, all three reconstructions are identical in the calibration period.

    Yes, that’s the way it should be — it’s the same iCRU instrumental series by construction.

  5. Posted Oct 7, 2008 at 3:14 PM | Permalink

    They all have the same shape as you get when you sort and scale red noise by r value. From right to left a downslope followed by a slow recovery.

    The sudden drop to -.4 in the green series in the top right looks a bit like something went pretty wrong. They should have looked at that one a bit more before publishing it.

    • Posted Oct 7, 2008 at 3:41 PM | Permalink

      Re: Jeff Id (#5), Of course something went wrong. First trees grow under optimum conditions. Too hot, too cold, too dry. too wet, too bad. Second high frequency resolution for multi-decadal reconstructions is tilting wind mills. The Sargasso Sea proxy is calibrated for temperature. How well? Who knows, but that was the purpose. Is it indicative of global temperatures? Of course not. It is a sound study that should be considered. Low frequency proxies calibrated for temperature have inherently more reliability than high frequency precipitation proxies in determining past temperatures. They also have natural smoothing. Steve will snip me for this, but low frequency proxies are the best gauge of the potential accuracy of the high frequency proxies.

  6. Henry
    Posted Oct 7, 2008 at 4:23 PM | Permalink

    Third, look at the the year 600 in the right hand panel. The black and red series essentially flip orientation. How’d they do that? It looks like there’s a reconstruction step and the regression sign for a series flips over when the composition of the network changes. Whatever it is, it’s not a good sign,

    The right hand key and left hand graph might explain it, though there may be errors in one (or both) of the keys.

    What you have on the left hand side is green (“NH screened”) flatish from 0 to 600 (almost proxiless?) at which point it jumps over the other two lines. Meanwhile black (“GL full”) and red (“NH full”) are close (perhaps since NH makes up most of GL?).

    Moving to the righthand side, the black line is described as “GL full – NH full” i.e. left hand black-red though it looks to me more like left hand red-green (or perhaps black-green), while the right hand red line is described as “NH full – NH screened” i.e. left hand red-green though it looks to me more like left hand green-black (or perhaps green-red), and the right hand green line is described as “NH screened – GL full” i.e. left hand green-black though it looks to me more like black-red.

    Let’s suppose my eyes are correct and the lines have been mislabelled. Then up to 600, the right hand black line is a volatile positive line minus a lower flatish line while the right hand red line is the low flatish line minus a similar volatile positive line. So they are almost mirror images. At 600 the left hand flatish line leaps up over the other two, so the signs of the differences shown on the right hand black and red lines reverse. In around 1000-1400 the left hand green line is again the lowest so the right hand black line above and reflecting the red line pattern re-establishes itself.

  7. Carl Gullans
    Posted Oct 7, 2008 at 6:06 PM | Permalink

    Steve, re: your comments about 0-600 flipping of the red and black series… this actually clearly continues until 1500 or so, after which it is difficult to tell. I’d be curious what the correlation is between these two series between 0-1500 (it may be close to -1) and then in more and more recent intervals to find the approximate year in which the correlation breaks.

  8. Ross McKitrick
    Posted Oct 7, 2008 at 6:55 PM | Permalink

    Steve, puzzle #3 looks like the following. In your left panel you have 3 lines; B, R and G. In the right panel you have 3 lines, b=(B-R), r=(R-G) and g=(G-B). Suppose G and B are highly correlated after 600, as they seem to be. Then b approximately = -r. That’s why they mirror each other in the right panel.

  9. Gary
    Posted Oct 8, 2008 at 9:45 AM | Permalink

    Hello, I am a long time reader of CA but have never posted here before. If I am slightly off topic I am sorry, but I would like to make two comments. The first is on Mann 08 and the second on Climate “Science” in general. I have read the Mann 08 study, and as the rest of you, was quite surprised. If I look at the graphs on page 13254 the simple eyeball test indicates that the proxies (NH CPS) and the CRU temperature data do not approximate each other in any way. I also see that the proxies indicate a -0.15 to 0.15 degree C temperature increase since 1850, not the 0.7 that we are supposed to have. Then still on page 13254 Dr. Mann say this.

    Interestingly, although the elimination of all tree-ring data from the proxy dataset yields a substantially smaller divergence bias, it does not eliminate the problem altogether (Fig. 2B). This latter finding suggests that the divergence problem is not limited purely to tree-ring data, but instead may extend to other proxy records.

    In other words, Ozone or other tree growth factors can not be the cause for divergence. And there must be something either wrong with many the proxies or something wrong with the CRU temperature data. Then he goes on to say this.

    Interestingly, the problem is greatly diminished (although not absent—particularly in the older networks where a decline is observed after 1980) with the EIV method, whether or not tree-ring data are used (Fig. 2 C and D).

    At this point he, and the reviewers, should realize that the manipulation with the EIV formula has done something that has unnaturally altered the original data. It has changed not just the slope of the data but the direction of change over time ahnd changed the data points on the graph by 300%. But apparently he doesn’t see that. Of course there are a lot of other problems like using the IPCC test for statistical significance, 66% to 90% without specifically stating in the abstract or summary that the findings did not reach statistical significance.

    My comment on Climate “Science” in general is a result of my exposure to other scientific research. I am in the medical field and have previously published original research, and read over 20 scientific articles a week. The practice of Infilling or Backfilling is some thing I have not seen in other fields. This would be like I was studying blood pressure and was supposed to take a patients pressure once a week but had not done so on several weeks. Would I be able to say that the pressure would have been about the same as the other pressures that I actually measured. So I can use a formula to create the data that did not exist? Of course not. Or comparing temperatures to the surrounding 500 miles to see if the data needs to be adjusted to make it better. Any one on the west coast knows that temperature can vary 30 degrees in 20 miles so this makes no sense. Or using made up temperatures from sites which are not reporting by averaging surrounding sites. This would be the same as if one of the patients we were measuring blood pressure on died and we said “well the pressure would have been the same as some of the other patients was, so use that data”. And then to use made up data to statistically support your conclusion? Yikes! I just can’t imagine that this is scientific.

    Keep up the good work Steve and the other contributors to this site.

  10. TerryB
    Posted Oct 8, 2008 at 10:35 AM | Permalink

    Where should I post this? This eems as good a place as any.

    Is “Proceedings of the National Academy of Sciences” really publishing a paper that seems to indicate a MWP and LIA – but now suggesting it was caused by CO2 variances?????????

    Oh lordy me.

    Steve: No, it’s not as good a place as any. It doesn’t have anything to do with Mann et al 2008. Please revive some other thread.

  11. tty
    Posted Oct 8, 2008 at 11:45 AM | Permalink

    These are data that have been published before. However to use them to “explain” the MWP (MCA?) and the LIA seems rather daring since they show a strong CO2 minimum ca 1200 AD and a maximum in the early 1300’s. If anything I would say that they tend to indicate that the MWP was not due to CO2 forcing.

    By the way these, like essentially all Stomata-based CO2 measurements are moderately but consistently higher than the ice-core data. This is handled rather discreetly in the paper. The CO2 values are “normalized” in Figures C and D. The absolute values (292-319 ppm) are mentioned in the text, but without any comment, except that they are in accord with other stomata-based record. There is one sentence in the material and methods section though: “It should be noted that, in general, CO2 data derived from stomatal frequency analysis have higher average values (ca 300 ppmv) compared with the IPCC baseline”. Period.

  12. ad
    Posted Oct 8, 2008 at 3:29 PM | Permalink

    How about:

    Although some of the preindustrial CO2 changes are at least
    temporally associated with anthropogenic influences on the
    environment, the amount of carbon needed to cause a shift of 34
    ppmv would far exceed the size of potential carbon sources and
    sinks in the terrestrial biosphere. It is likely that, analogous to
    early Holocene CO2 changes (25–28), depletion and restoration
    of atmospheric CO2 between A.D. 1000 and 1500 was driven
    mainly by short-term perturbations of sea-surface temperature
    and/or salinity.

    So they are saying temperature changes drove CO2 changes.

    • Dave Dardinger
      Posted Oct 8, 2008 at 4:29 PM | Permalink

      Re: ad (#13),

      snip – this sort of issue has been discussed elsewhere. Please do not hijack this thread.

  13. TerryB
    Posted Oct 9, 2008 at 2:07 AM | Permalink

    Steve, John A : yes sorry guys, will do

  14. Richard
    Posted Oct 9, 2008 at 3:36 AM | Permalink

    Steve, are you planning to bring any of this work together as a publication (or several publications)? The internet is good but people tend to be dismissive of anything discussed on the internet however well prepared and documented it is.

    Perhaps a publication which demonstrated the effect of filtering out negative correlation samples would be a good way of showing what is going on without necessarily appearing too aggressive?

    If I have understood things correctly, you now have most of the base data used by Mann 08. Perhaps you could propose a new paper with the same figures without the “Mannian” filters.

    Sorry if this post is a little off topic.

  15. Edouard
    Posted Oct 9, 2008 at 5:33 AM | Permalink


    This is not at all off topic. ?Millions? are reading climateaudit, but noone knows what it really means.

    To win a Web award, we should get real information about what is happening.

    Whom can we trust? What can we believe? What do we know for sure?

    This could be one of the best web sites, but noone seems to care about us??? :-(((

  16. Posted Oct 9, 2008 at 8:28 AM | Permalink

    This thread is drifting a bit.

    In reply to 17, on my blog I have just demonstrated that CPS sorting not only creates a HS but demagnifies the historic pre-calibration trends. After it is reviewed thoroughly I think it should be used as a serious warning sign to stop this kind of data sorting.

    I have had some comments about red noise matching and other things, from my many hours of work the noise level in proxies is more than sufficient to cause a substantial distortion of the data. I believe the long term and HS shape of the graphs above are created by the sorting technique, not by the proxies.

    Please snip if this is too off topic.

  17. masmit
    Posted Oct 16, 2008 at 10:58 AM | Permalink

    Sorry to all if this is OT, but in the spirit of what I think this site is about, I’ve created a petition at the Number 10 site:

    We the undersigned petition the Prime Minister to Require
    government funded research into climate change to meet minimum
    standards of honesty.

    In order to ensure that public policy is guided by the best
    possible scientific knowledge, it should be required of
    research bodies such as the Hadley Centre that their published
    research meet at least the same standards of disclosure and
    transparency as financial and mining prospectuses, such that
    failure to meet such standards should disqualify research from
    consideration in setting public policy.

    Any brits (I assume only brits can vote), the URL is:

%d bloggers like this: