One of the very first contributions to realclimate was an FAQ from Jeff Severighaus on Dec 3, 2004. A year earlier, Severinghaus attempted (unsuccessfully) to get an explanation of the “divergence” problem from Mann and the rest of the Team. Severinghaus had become interested in the question following a presentation by Tom Karl of NOAA in which he had used a figure from Briffa and Osborn 2002, in which he wondered about the “flat” response of the tree ring proxies in the last half of the 20th century.
In nearly all defences of the deletion of the decline in spaghetti graphs that yield a rhetorical effect of coherence between the Briffa and other reconstructions in the last half of the 20th century, it’s been argued that the divergence problem was fully disclosed in a couple of 1998 Briffa articles and that this disclosure in the original technical literature constituted sufficient disclosure – a point that I contested long before Climategate.
The Severinghaus exchange is highly pertinent to this issue. Severinghaus was a climate scientist who was not a specialist in the area who asked specifically about a diagram in which the decline had been hidden (though Severinghaus was unaware that the decline had been hidden.)
Severinghaus was concerned merely by the flattening of proxy response. One can only imagine how the exchange would have read had Severinghaus been aware that the Briffa reconstruction actually declined sharply. Read and see whether Mann, Jones and/or Briffa drew Severinghaus’ attention to the early articles in which the divergence problem was disclose.
On the afternoon of Feb 1, 2003 California time (emails -2545, 19, 4355 Feb 2, 2003 00:15 GMT), Severinghaus wrote to Tom Karl of NOAA about his presentation at the MIT Global Change Forum the previous day. Severinghaus asked about the “flat” response of tree rings to late 20th century warmth, referring to an article by Briffa and Osborn in Science (2002). The diagram in question would be the following:
Figure 1. Briffa and Osborn (Science 2002) Figure 1.
Severinghaus observed that this lack of response is an “embarrassment” and that it “casts doubt on the integrity of the proxy”:
Subject: tree rings and late 20th century warming
Date: Sun, 2 Feb 2003 16:15:04 -0800
From: Jeff Severinghaus
Dear Dr. Karl,
I enjoyed your presentation yesterday at the MIT Global Change forum. You may recall that I asked about the failure of tree rings to record the 20th century warming. Now that I look at my records, I realize that I remembered this wrongly: it is the LATE 20th century warming that the tree rings fail to record, and indeed, they do record the early 20th century warming.
If you look at the figure in the attached article in Science by Briffa and Osborn, you will note that tree-ring temperature reconstructions are flat from 1950 onward. I asked Mike Mann about this discrepancy at a meeting recently, and he said he didn’t have an explanation. It sounded like it is an embarrassment to the tree ring community that their indicator does not seem to be responding to the pronounced warming of the past 50 years. Ed Cook of the Lamont Tree-Ring Lab tells me that there is some speculation that stratospheric ozone depletion may have affected the trees, in which case the pre-1950 record is OK. But alternatively, he says it is possible that the trees have exceeded the linear part of their temperature-sensitive range, and they no longer are stimulated by temperature. In this case there is trouble for the paleo record. Kieth Briffa first documented this late 20th century loss of response.
Personally, I think that the tree ring records should be able to reproduce the instrumental record, as a first test of the validity of this proxy. To me it casts doubt on the integrity of this proxy that it fails this test.
Severinghaus obviously didn’t know that the Briffa and Osborn diagram had deleted the post-1960 decline from the Briffa reconstruction. Had they shown the actual data, the diagram would have looked more like the one shown below.
Had Briffa and Osborn shown the decline, instead of Severinghaus wondering about the “flat” response, he would presumably have been asking about the sharp divergence between the Briffa reconstruction (based on a very large population of sites identified ex ante as temperature limited) and the Mann reconstruction (based on bristlecones).
At 12:49 GMT (7:49 am Eastern, emails 19,2545,4355) Karl forwarded Severinghaus’ message to an email list that ultimately involved (during the single day) Chris Miller of NOAA, Verardo of NSF, Mann, Bradley, Hughes, Briffa, Jones, Wally Broecker, Ed Cook and R.F. Weiss of UCSD. Karl asked:
Correct me if I am wrong, but I always thought the failure was a lack of tree cores subsequent to the 1980s. Please correct me if I am wrong, and if Jeff is correct, then indeed we have a significant implication.
At 14:15 GMT (emails 19, 2545, 4355), Mann denied admitting that the divergence problem was an “embarrassment”, referring back to an EGS meeting of 2002. Mann stated that they had discussed the issue “time and again in our own work” [SM – offhand, I can’t recall a clear discussion of the issue in Mann’s work to that date: dig here].
Have no fear, Jeff has still got his facts wrong, even after going back and checking once…
First off, I never made any such comment to Jeff–he clearly misunderstood comments that I made at EGS a year ago in response to a question he asked. Of course, it is well know that there are a number of competing explanations [this is what I said–to quote this as offering “no explanation” is a bit unfair Jeff, don’t you think? As I recall, I even invited Tim Osborn in the audience to add his own comments–but he had little to say] for the fact that *high latitude*, primarily *summer responsive*, tree-ring *density* data have exhibited a noteable decline in the past few decades in the amplitude of their response to temperature variability. We have discussed this issue time and again in our own work, and Keith Briffa, Malcolm Hughes, and many others have published on this, w/ competing possible explanations (stratospheric ozone changes, incidentally, is the least plausible to me of multiple competing, more plausible explanations that have been published). See e.g.: Vaganov, E.A., M.K. Hughes, A.V. Kirdyanov, F.H. Schweingruber, and P.P. Silkin, Influence of Snowfall and Melt Timing on Tree Growth in Subarctic Eurasia, Nature, 400 (July 8), 149-151, 1999.
Mann re-iterated that the MBH reconstruction tracked temperature well to 1980 and blamed Severinghaus for “only looking” at Briffa and Osborn 2002 which “doesn’t properly align the 20th century means of the various reconstructions and instrumental record”.
It should *also* be noted that we used essentially none of these data in the multiproxy Mann/Bradley/Hughes (MBH) reconstruction, and that the MBH reconstruction tracks the instrumental record quite well through the very end of our calibration interval (1980–it stops then because there are far fewer paleo records available after 1980). This was shown in our 1998 Nature article quite clearly, and of course remains true today. Jeff made the mistake of only looking at the Briffa & Osborn paper, which doesn’t properly align the 20th century means of the various reconstructions and instrumental record.
As so often in Team discussions, Mann’s answer was unresponsive to the issue: Severinghaus was wondering about the failure of late 20th century tree rings to track increased temperature – this has nothing to do with alignment of the mean, but with the slope of the various curves. Mann continued with a reference to the “recent high-latitude decline issue” – note that Severinghaus had not actually mentioned the decline – saying (incorrectly) that these had been “discussed in IPCC” (they hadn’t – quite the opposite).
An appropriate alignment of all the records is provided in IPCC, and in the attached Science perspective from last year. This shows how well the Mann et al reconstruction (and several model-based estimates) track the entire instrumental record. There are some good reasons that some of the other purely tree-ring based reconstructions differ in their details, in addition to the greater influence of the recent high-latitude density decline issue, and these are discussed in IPCC and the Science piece.
Mann then asserted that MBH98 had provided “provided detailed calibration and verification statistics that establish the skill in our reconstruction” together with uncertainties calculated using “rigorous analysis of the calibration and cross-validation residuals”.
Of course, we have in, our own work provided detailed calibration and verification statistics that establish the skill in our reconsruction in capturing the details of both the modern instrumental record, and independent, withheld earlier instrumental data (19th century and, more sparsely, 18th century), and we publish uncertainties that are based on rigorous analysis of the calibration and cross-validation residuals. I know that Jeff has seen me talk on this many times, and probably has read our work (I would hope), so I’m frankly a bit disappointed at the comments. I would have liked to think that he would have approached us first, before broadcasting a message full of factual errors.
Please let me, or any of the others know, if we can provide any further information that would help to clarify (rather than obscure!) the facts,
At the time, it was not known that Mann had failed to “provide” the verification r2 and CE statistics that were then standard practice in the tree ring reconstruction community and that the MBH reconstruction failed these tests in its early stages. In addition, the methodology for Mann’s calculation of uncertainties was undecipherable prior to Climategate. (There is some Climategate information that I haven’t parsed yet and may provide clues.)
At 16:36 GMT, Jones endorsed Mann’s unresponsive reply, adding more inaccurate information:
Mike’s answer is a fair response. Jeff has mixed some facts up and this is maybe because we’ve never explained them clearly enough. There are two facts:
1. There are few tree-core series that extend beyond the early 1980s. This is because many of the sites we’re using were cored before the early 1980s. So most tree-ring records just don’t exist post 1980.
2. The majority of the recent warming is post-1980, so no proxy would pick this up. This warming has been large and it would be good to go back and see if the trees have picked it up. It would give more faith in tree-ring reconstructions, but any reconstruction method is being pushed to the limit by the rate of temperature rise over the late 20th century. Applies to other proxies but you have to note the following:
It is important to remember that locally few regions exhibit statistically significant warming. Highly significant at the hemispheric level, but not great at the local level due to high level’s of variability. The spatial scales are important and this is difficult to get across.
The supposed lack of updated tree ring sites was a common meme in 2003 and was one that I encountered when I first entered the field. Mann argued in the early days of realclimate that the proxies were in difficult locations and it would be virtually impossible to replicate the heroic work of the 1970s. In 2005, I wrote an op ed urging climate scientists to update the proxies to prove that they were out-of-sample responsive and satirized Mann’s claim that the proxies would require undue effort to update.
At 17:55 GMT, in the wake of Mann’s counterattack, Severinghaus backed off, even apologizing for asking such a sensible question. However, he (quite correctly) didn’t understand what alignment of means had to do with the issue at hand and (quite correctly) pointed out that the Esper results went up to 1993 (thus the data didn’t stop in 1980 as Jones and Mann had proposed):
Please accept my apologies if I have gotten the story wrong. I am not a specialist in the tree-ring field, and was simply reporting what I saw in the Briffa and Osborne paper, several other papers, and what several tree-ring people have told me in conversations. I agree, we need to keep the level of misinformation out there down to a minimum! regret adding to it.
I am still confused, however, about Mike’s explanation for the Briffa and Osborne paper’s curve appearing flat after 1950 AD. Can you try explaining this again, Mike, please? I don’t understand how aligning could change the slope of a curve. The curves appear to continue to 1990 AD or so, and the Esper et al. curve continues to 1993. So the explanation that the records only go up to 1980 doesn’t seem to hold in this case. The dashed black line is the instrumental record for warm-season >20 N latitudes and it does indeed diverge from the tree-ring records in the 1980s. Can you help me out here?
At 19:16 GMT, Mann sent another lengthy and unresponsive reply. Mann re-iterated that the MBH reconstruction had the same amplitude as the increase in temperature over the calibration period and ignored the flattening out in the proxies that was worrying Severinghaus. Mann didn’t mention that Briffa and Osborn had deleted post-1960 values of their reconstruction nor did he suggest that Severinghaus consult Briffa’s 1998 article in which he had reported a noticeable decline in both tree ring density and widths in the last half of the 20th century.
Jeff, Choice of aligning has no influence on the slope of the curve, it simply changes the mean baseline for comparison. The Mann et al reconstruction has the same amplitude increase as the full Northern Hemisphere annual mean instrumental record over the calibration interval (1900-1980). On this simple point, there is no debate. And this seems to be the origin of your misunderstanding of the issues involved.
Briffa & Osborn use a slightly different convention from that used elsewhere (e.g. IPCC and in the attached Science piece which I’ve re-sent for the benefit of your expanded recipient list), and by their convention the instrumental record is observed to lie ever-so-slightly above the MBH reconstruction over the entire interval available for comparison (mid 19th century-> 1980). This difference is actually quite small, so I’m not sure why we’re even discussing it in the first place. It, however, does not in any case impact a comparison of the trends in the two series, which match remarkably well over that same interval. This is despite the fact that the MBH reconstruction represents the entire Northern Hemisphere (which gets half of its contribution from the tropics i.e., latitudes 30N) while the instrumental series shown by Briffa & Osborn is only the extratropics north of 20N. This is old stuff, and I would guess that the others cc’d in on this message (Ray, Malcolm, Keith, Phil) are not interested in re-hashing these old discussions.
The state of the the science here has moved well beyond these semantic and/or conventional arguments, focusing instead on detailed intercomparisons of methods and data (employing rigorous diagnostics of reconstructive fidelity (collaborative between Bradley/Briffa/Hughes/Jones/Mann/Osborn/Rutherford). There is little disagreement between us on the broad trends when seasonal and spatial sampling issues, and differing conventions for e.g. defining reference periods, have been taken appropriately into account. I hope you find that the above information clarifying Jeff. Due to other demands on my time, I have to sign out now on this series of exchanges.
best regards, Mike Mann
Mann’s statement that his reconstruction got “half of its contribution from the tropics” was untrue. The reconstruction was heavily weighted towards strip bark bristlecones. It only had a few tropical proxies. Briffa and Cook satirized Mann’s claims of tropical skill [-cite].
At 20:03 GMT (4355), Mann conceded that Esper reconstruction was also flat after 1950, “defying” the instrumental record. Although the Esper reconstruction was built on ring width data, Mann said that flatlining “generally doesn’t appear to be a problem with tree ring width data, at least those available through 1980″.
One final point I didn’t respond to, upon re-reading your previous email: My comments about the baseline period issue only refers to comparisons of the instrumental record against the MBH reconstruction (as shown in the Briffa & Osborn piece).
Unlike the MBH reconstruction, which tracks the instrumental record well through the end of the calibration interval (1980), the Esper et al reconstruction indeed doesn’t show any warming after 1950 or so, which defies evidence from the instrumental record. This is similar to what has been noted, as discussed in the previous emails, with high-latitude summer-temperature sensitive maximum latewood tree-ring density chronologies (e.g. Briffa and coworkers) and it may relate to the same factors that have been discussed in that context. This generally doesn’t appear to be a problem with tree ring width data, at least those available through 1980.
Once again, the wisest approach is to make use of all annually-resolved proxy information available…
That’s my final word on this, promise…
In fat, although there had been negligible recent discussion in the “peer reviewed literature” of divergence of ring widths, it had been mentioned in Briffa’s original 1998 articles. Mann’s assertion here was untrue.