Here is a longer excerpt from the July 19, 2000 Raymond Bradley Climategate email posted earlier today:
[......At this point Keith Alverson throws up his hands in despair at the ignorance of non-model amateurs...]
But there are real questions to be asked of the paleo reconstruction.
First, I should point out that we calibrated versus 1902-1980, then “verified” the approach using an independent data set for 1854-1901. The results were good, giving me confidence that if we had a comparable proxy data set for post-1980 (we don’t!) our proxy-based reconstruction would capture that period well. Unfortunately, the proxy network we used has not been updated, and furthermore there are many/some/ tree ring sites where there has been a “decoupling” between the long-term relationship between climate and tree growth, so that things fall apart in recent decades….this makes it very difficult to demonstrate what I just claimed. We can only call on evidence from many other proxies for “unprecedented” states in recent years (e.g. glaciers, isotopes in tropical ice etc..).
But there are (at least) two other problems — Keith Briffa points out that the very strong trend in the 20th century calibration period accounts for much of the success of our calibration and makes it unlikely that we would be able be able to reconstruct such an extraordinary period as the 1990s with much success (I may be mis-quoting him somewhat, but that is the general thrust of his criticism). Indeed, in the verification period, the biggest “miss” was an apparently very warm year in the late 19th century that we did not get right at all. This makes criticisms of the “antis” difficult to respond to (they have not yet risen to this level of sophistication, but they are “on the scent”).
Furthermore, it may be that Mann et al simply don’t have the long-term trend right, due to underestimation of low frequency info. in the (very few) proxies that we used. We tried to demonstrate that this was not a problem of the tree ring data we used by re-running the reconstruction with & without tree rings, and indeed the two efforts were very similar — but we could only do this back to about 1700.
Whether we have the 1000 year trend right is far less certain (& one reason why I hedge my bets on whether there were any periods in Medieval times that might have been “warm”, to the irritation of my co-authors!). So, possibly if you crank up the trend over 1000 years, you find that the envelope of uncertainty is comparable with at least some of the future scenarios, which of course begs the question as to what the likely forcing was 1000 years ago. (My money is firmly on an increase in solar irradiance, based on the 10-Be data..).
Another issue is whether we have estimated the totality of uncertainty in the long-term data set used — maybe the envelope is really much larger, due to inherent characteristics of the proxy data themselves….again this would cause the past and future envelopes to overlap.
In Ch 7 we will try to discuss some of these issues, in the limited space available. Perhaps the best thing at this stage is to simply point out the inherent uncertainties and point the way towards how these uncertainties can be reduced. Malcolm & I are working with Mike Mann to do just that.
In an earlier post, I criticized the repugnant attitude in which Bradley sneered at “antis” who had not yet reached the “level of sophistication” sufficient to disentangle adverse results that Bradley and coauthors had failed to report, but were “on the scent”.
In the present post, I wish to focus on a different point, the one referred to in the following paragraph:
Furthermore, it may be that Mann et al simply don’t have the long-term trend right, due to underestimation of low frequency info. in the (very few) proxies that we used. We tried to demonstrate that this was not a problem of the tree ring data we used by re-running the reconstruction with & without tree rings, and indeed the two efforts were very similar — but we could only do this back to about 1700. [my bold]
It seems certain to me that Bradley is here referring to the analysis in Mann et al 2000( Earth Interactions), which had an online version at NOAA here. The published version of this article states that it was “Received 11 May 1999; accepted 31 May 2000. (in final form 15 June 2000) ]”. Thus it was very recent (to say the least) at the time of Bradley’s July 10, 2000 email. It is my surmise (though this is just a surmise) that this article combined with the interactive web presentation was what Bradley was referring to in the closing paragraph of his July 20, 2000 email (though nothing turns here on the correctness of this surmise):
Perhaps the best thing at this stage is to simply point out the inherent uncertainties and point the way towards how these uncertainties can be reduced. Malcolm & I are working with Mike Mann to do just that.
MBH98 had stated that “the long-term trend in NH is relatively robust to the inclusion of dendroclimatic indicators in the network”. No supporting evidence is given for this statement in MBH98. The only support appears to come from the Mann et al 2000 article, which gives the following summary statement:
We have also verified that possible low-frequency bias due to non-climatic influences on dendroclimatic (tree-ring) indicators is not problematic in our temperature reconstructions.
Note that this statement is much broader and much more categorical than the one in Bradley’s email, where he admitted to insiders that their results not hold up for the pre-1700 networks.
This statement in Mann et al 2000 linked to a NOAA webpage http://www.ncdc.noaa.gov/paleo/ei/ei_nodendro.html, which has itself attracted considerable commentary over the past few years. It stated:
MBH98 argued, furthermore, that biases unique to a particular type of proxy indicator (e.g., tree-ring widths) are less problematic for “multiproxy”-based reconstructions that make use of the complementary information in a diverse proxy network. MBH98 found through statistical proxy network sensitivity estimates that skillful NH reconstructions were possible without using any dendroclimatic data, with results that were quite similar to those shown by MBH98 based on the full multiproxy network (with dendroclimatic indicators) if no dendroclimatic indicators were used at all. We show this below for annual-mean reconstructions of Northern Hemisphere mean temperatures.
They illustrated this “finding” with a graphic showing their AD1760 reconstruction. They continued with a similar statement as above, but this time include the qualification that the graphic only shows the similarity for “the period in question”.
Also shown is the reconstruction based ONLY on dendroclimatic indicators (ie, no coral, ice core, or historical or instrumental indicators). Again, the primary features of the reconstruction are very similar. Whether we use all data, exclude tree rings, or base a reconstruction only on tree rings, has no significant effect on the form of the reconstruction for the period in question.[my bold] This is most probably a result of the combination of our unique reconstruction strategy with the careful selection of the natural archives according to clear a priori criteria. Furthermore, we note that Jones et al. (1998), get similar results for the recent changes using an almost completely different tree-ring network based on wood density from high latitude trees. These comparisons show no evidence that the possible biases inherent to tree-ring (alone) based studies impair in any significant way the multiproxy-based temperature pattern reconstructions discussed here.
The above statements, both individually and collectively, give an entirely different impression to the reader than Bradley’s email to Oldfield.
The argument in http://www.ncdc.noaa.gov/paleo/ei/ei_nodendro.html has continue to be cited, most recently in Hughes’ evidence to the Muir Russell Committee in early 2010. Hughes stated that it was “it was possible to make this comparison only for the period AD 1750 to 1980″ – a point obviously contradicted by Bradley’s email noting (adverse) results for earlier networks:
We published a comparison of our results including and excluding tree rings in the online journal Earth Interactions in 2000 http://www.ncdc.noaa.gov/paleo/ei/ei_nodendro.html showing that the decadal course of reconstructed temperatures was largely insensitive to the inclusion or exclusion of tree-ring data, including after 1960. At that time (in 2000), it was possible to make this comparison only for the period AD 1750 to 1980 because many of the records we used ended by 1980, and because, although there were plenty of tree-ring records before 1750, that was not true of the non-tree-ring data at that time.
The webpage http://www.ncdc.noaa.gov/paleo/ei/ei_nodendro.html was cited in Mike Mann’s well-discussed Sep 22, 1999 email 136. 0938018124.txt, leading up to hide the decline (though no attention was paid to this point in my earlier discussions of IPCC and the Trick). This was the email in which Mann noted “everyone in the room at IPCC was in agreement that this [the Briffa reconstruction] was a problem and a potential distraction/detraction from the reasonably concensus viewpoint we’d like to show w/ the Jones et al and Mann et al series”, later adding that he didn’t want to give “fodder” to the skeptics.
Mann supported his email with a reference to the supposed nodendro argument as follows:
One other key result with respect to our own work is from a paper in the press in “Earth Interactions”. An unofficial version is available here: http://www.ngdc.noaa.gov/paleo/ei/ei_cover.html
The key point we emphasize in this paper is that the low-frequency variability in our hemispheric temperature reconstruction is basically the same if we don’t use any dendroclimatic indicators at all (though we certainly resolve less variance, can’t get a skillful reconstruction as far back, and there are notable discrepancies at the decadal and interannual timescales). A believe I need to add a sentence to the current discussion on this point, since there is an unsubstantiated knee-jerk belief that our low-frequency variability is suppressed by the use of tree ring data.
We have shown that this is not the case: (see here: http://www.ngdc.noaa.gov/paleo/ei/ei_datarev.html and specifically, the plot and discussion here: http://www.ngdc.noaa.gov/paleo/ei/ei_nodendro.html
Ironically, you’ll note that there is more low-frequency variability when the tree ring data *are* used, then when only other proxy and historical/instrumental data are used!
SO I think we’re in the position to say/resolve somewhat more than, frankly, than Keith does, about the temperature history of the past millennium.
Tim Lambert and others have attempted to argue that this is adequate disclosure. They point to the qualification in the commentary that the result applies only for the “period in question”, disregarding the failure of Bradley and others to report the opposite results in earlier periods. Due to the Climategate email, we now know that Bradley and coauthors were aware of the adverse results in earlier periods (notwithstanding Hughes’ evidence to the Muir Russell panel.) Once again, we see the differences between “full, true and plain disclosure” and Team practices. In a “full, true and plain disclosure” regime, Bradley and co-authors would be obliged to disclose the adverse results for the networks prior to AD1700.