Bristlecone "Adjustment" #2

The bristlecone pine "adjustment" for CO2 fertilization in MBH99, while genuflecting to Graybill and Idso [1993], is completely different as seen in Figure 1 below.
Figure 1. Bristlecone Pine Adjustments. Black solid - Graybill and Idso; red dashed - MBH99

Figure 1. Bristlecone pine CO2 adjustments. black solid – Graybill and Idso strup-bark; red dashed – MBH99

Graybill and Idso [1993] hypothesized that tree ring "chronologies" for strip-bark forms would respond to higher CO2 concentrations in proportion to the total biomass response (benchmarked on their sour orange experiments), while "chronologies" for entire-bark forms would respond in proportion to trunk cross-sectional area (also benchmarked on sour orange experiments). Their Figure 4 shows the difference between strip-bark and entire-bark chronologies at Sheep Mountain and associates these differences with the above factors.

In order to detect this hypothesized response, strip-bark trees "were the primary focus of investigation wherever possible" in Graybill and Idso. (see page 86). In MBH99, a non-climatic effect on bristlecone pines was noted and an adjustment proposed for the 1000-1399 period. As I mentioned before, there is no adjustment in any results after 1400 in either MBH98 or MBH99. In passing, the existence of a non-climatic factor is contrary to one of the major assumptions of MBH98 and presumably should have occasioned a corrigendum to MBH98 at that time.

The adjustment developed in MBH99 bears no relationship to the Graybill and Idso analysis, as shown in Figure 1 below, where the two versions are compared. Graybill and Idso associate rising CO2 levels of the 20th century with enhanced ring widths; MBH99 argues that rising CO2 levels in the 19th century led to increased growth, but that the effect became saturated in the 20th century. I’m looking at MBH99 in more detail right now and will post up some details on how the MBH99 adjustment works (or doesn’t work). Since the MBH99 is so different from the Graybill and Idso [1993] effect, it seems to me like there should have been some contemporary discussion as to the pros and cons of the MBH99 adjustment. It’s easy to see that the Graybill-Idso adjustment completely eliminates the hockey stick pattern from the bristlecone pines.

In passing, Graybill and Idso[1993] pointed out that two bristlecone chronologies lacked the growth increase after the mid-19th century (San Francisco Peaks AZ and Mount Goliath).


  1. Tom Rees
    Posted Feb 11, 2005 at 10:50 AM | Permalink

    The MBH99 correction is not a correction for CO2. There is debate about what exactly causes the spike in the BCPs, but it certainly isn’t a simple, linear CO2 effect. You know this (at least, it was in the review you co-authored), and MBH know this. So the comparison in the figure above seems a bit futile.

    Their correction makes no assumptions about the cause of the effect, other than it doesn’t occur in their control series (north american treeline, iirc). My understanding is that PC1 is transformed to effectively chop off the upswing. Whether the magnitude or the precise form of the transformation is correct is, to some extent important. But the key that I would like you to clarify is how there is an upswing in the final data if 1) PC1 no longer has an upswing and 2) PC1 dominates the effect. That’s why I’m saying that you can’t simply transpose this particular argument from MBH98 to MBH99.

    Furthermore, your argument that MBH sought to with hold this non-temperature effect from their peers is not true, since they published a correction. It was published in MBH99.

  2. Steve McIntyre
    Posted Feb 11, 2005 at 12:40 PM | Permalink

    Tom, thanks for commenting on this issue, but I disagree with pretty much everything you’ve said here.

    The MBH99 adjustment is described as a CO2 correction. Look on page 760 e.g. “it is plausible that the divergence of the two series is related to a CO2 influence on the ITRDB PC#1 series. The residual is indeed coherent with rising atmospheric CO2 until it levels off in the 20th century, which we speculate may represent a saturation effect whereby a new limiting effect is established at high CO2 levels”. I think that it’s entirely reasonable to plot their adjustment against the adjustment of Graybill and Idso [1993], who actually studied these trees.

    MBH99 does not articulate the debate about what causes 20th century BCP growth, but we tried to articulate this evenhandedly in MM05(E&E). I’m not saying that it was CO2 or nitrate – only that people need to know what it is before relying on this data for world climate history with or without pretty ad hoc adjustments. I was simply illustrating the difference in two adjustment methods. Maybe a 3rd method is better. But this is a highly relevant adjustment and needs to be explored.

    Your udnerstanding about chopping off the PC1 uptick is incorrect. MBH99 adjusts 19th century data, but keeps the 20th century uptick relative to the 19th century (or even increases it). Look how flat the MBH99 adjustment is in the 20th century. It’s a crazy adjustment if it’s supposed to be for CO2 fertilization. In any event, the adjustment does not get rid of the uptick – plot out the “pc1-fixed” and you’ll see. Because it doesn’t get rid of the uptick, a hockey stick shaped PC1 will remain in MBH99. I’ve tried to explain how the PC1 imprints the MBH98 reconstruction in MM05(E&E). The effect is probably similar in MBH99, but I haven’t worked through MBH99 in the same detail yet. In part, I’ve had trouble figuring out exactly how this MBH99 adjustment worked (but I’ve now been able to emulate the adjustment and will post up some details). It’s pretty weird.

    I disagree strongly with your characterization of MBH99 as being in any sense a “correction” of MBH98. Not one post-1400 number was changed in MBH99. To me, a “correction” would have required (1) explicit notice that some of the proxies in MBH98 were affected by non-climatic factors, contrary to the representation in MBH98; (2) recall of the post-1400 MBH98 numbers and replacement with an adjustment for the non-climatic factors. That’s what you’d have to business. They didn’t do this. Not only did they not do this, they perpetuated the problem by carrying forward the MBH98 numbers into MBH99 and splicing them into MBH99.

    But the more fundamental disclosure problem relates to the non-disclosure of doing AD1400 calculations without the bristlecone pines. Mann did these calculations; they give high 15th century values; he just didn’t report them. In a securities offering, you’d have to disclose this; you could argue that they don’t matter, but the reader has the right to be fully informed and decide for himself. This would be serious stuff in the real world.

    As to your point about BCPs and the 20th century uptick, I think that you’re grabbing the wrong end of the stick, as it were. I’m persuaded that the 20th century is warmer than the 19th century. The $64 question in our articles is whether Mann’s data and methods allow him to say that the 20th century is warmer than the early 15th century. This is entirely dependent on bristlecone pines: the early part of the series “wants” to go up and Mann has to work at it to drag it down. The editing of the Gaspe series to include it in the AD1400 roster was a gross example of this. (Compare the updated Gaspe version and you’ll see why the Hockey Team haven’t published the updated Gaspe version, especially now that it has a knock-on effect on MBH98.)

  3. Tom Rees
    Posted Feb 14, 2005 at 7:27 AM | Permalink

    Hi Steve. MBH speculte that the anomalous growth is due to CO2. But they point out that “However, it suffices that the residual is non-climatic in nature”. They don’t attempt to covary for CO2, or use a metabolic model or anything like that, they just assume that it’s not climate. My point about PC1 is that it is very much different after the transformation than before. According to fig 1a (corrected ITRB millenial PC#1), the value at 1400 is pretty much what it is in the 14th century. That said, I don’t see how it can explain the trend in MBH99. Also, it seems to me that MBH99 values post 1400 must be different from pre-1400. You said you plotted out the data. In which case what you’ve plotted may be different from what’s shown in fig 1a.

  4. Steve McIntyre
    Posted Feb 14, 2005 at 2:53 PM | Permalink

    Hi, Tom, I think that what you’re missing is the effect of the stepwise splicing. I’m not saying that there isn’t an increase from the 19th to the 20th century. However, if you do the calculations without the bristlecones or using a correct PC calculation, you get high 15th century values under Mann’s methodology. The 15th century PC1 is not adjusted. You’ve distinguished clearly that the appropriateness of the adjustment and the existence of the adjustment are different issues. In my opinion, if at the time of MBH99, Mann’s view was that there needed to be an adjustment to bristlecones, then he should have adjusted the 15th century calculations as well instead of reproducing the unadjusted calculations. I assure you that if you compare archived reconstructions from MBH98 with the post-1400 archived values of MBH99, there is no difference.

  5. Ed Snack
    Posted Feb 16, 2005 at 3:11 PM | Permalink

    One of the basic assumptions behind all proxy studies (including MBH98 &99) is that there is a linear relationship between the proxy and temperature. The relationship can be derivative, that is the proxy actually reflects precipitation but if precipitation is lineraly related to temperature the overall linear relationship holds. Alternatively, if the relationship is not linear but is known, an adjustment can be applied to produce a linear relationship. Unless such a relationship can be established, a proxy is unusable as a temperature indicator, or so I assert. The linearity would need to be established, preferably by a comparison between the proxy and measured temperatures over some period, although with lower certainty a theorectical coupling can be asserted.

    In the case of the disputed BCP series, surely a direct linear relationship does not hold. There is a time period for which the proxy series can be measured against local temperatures, and it must be against local temperatures as any attempt to measure against non-local temperatures is surely an absurdity, and the relationship is definitely not linear. One could therefiore apply an adjustment, take the temperature measurement and calibrate the proxy over the period, such an adjustment for the last century would look like the Gray & Idso adjustment. There would also need to be serious consideration of adjustments going back on the assumption that the current behaviour was representative and always applied.

    It would seem that the MBH99 adjustment does not meet the objective criteria, and is clearly a theorectical adjustment based on some understanding of the biological mecanisms, but this appears to completely fail to explain the non-temperature related response of the proxy over the 20th century. This is to me, obviously problematic, and the only reasonable response would be to omit such a proxy record from the study. I would find any adjustment problematic, even one such as the G&I because this would seem to imply that the 20th century (and some of the 19th century) proxy results are the only non-linear results, and I don’t believe that there are sufficient grounds for this to be accepted.

    I’m not entirely dogmatic about this, but I would suggest that at this stage, there is insufficient knowledge about the temperature response of at least some of the BCP sites to enable their use in a proxy study, and therefore they should be excluded.

    Steve’s comments: These points are in agreement with our views. Note that Mann studied the effect of excluding the BCPs in the BACKTO_1400-CENSORED file, which we discuss in MM05(GRL) and which I’ve referred to in this blog. Without the BCPs, the 15th century values under MBH methods are higher than 20th century values.

  6. Ed Snack
    Posted Feb 21, 2005 at 3:48 AM | Permalink

    Tom, any chance of a response:
    (i) to Steve’s assertion that the BCP’s are in MBH98 & MBH99 with any adjustments having minimal differences;
    (ii) to my assertions on the suitability of the using the BCP measurements at all; and
    (iii) to whether the files in the BACK_TO_1400 etc directory on Mann’s ftp site are what you get if the BCP’s are omitted, as Steve claims to have shown to at least a reasonable degree of certainty.


%d bloggers like this: