Around minute 11:24, Bradley puts on a slide with the MBH99 title page. For background, here’s what the original press release of March 3, 1999 said:
Researchers at the University of Massachusetts who study global warming have released a report strongly suggesting that the 1990s were the warmest decade of the millennium, with 1998 the warmest year so far”⤮ "Temperatures in the latter half of the 20th century were unprecedented," said Bradley. “⤠The latest reconstruction supports earlier theories that temperatures in medieval times were relatively warm, but "even the warmer intervals in the reconstruction pale in comparison with mid-to-late 20th-century temperatures," said Hughes.
Here’s a comment from Mann in IPCC TAR:
Self-consistent estimates were also made of the uncertainties… The uncertainties (the shaded region in Figure 2.20) expand considerably in earlier centuries because of the sparse network of proxy data. Taking into account these substantial uncertainties, Mann et al. (1999) concluded that the 1990s were likely to have been the warmest decade, and 1998 the warmest year, of the past millennium for at least the Northern Hemisphere.
Here’s how Bradley now represents the matter (I’ve tried to make the transcript accurate and apologize for any inaccuracies):
Here we come to the famous hockey stick. Someone characterized the variations of the last 1000 years as a hockey stick. Here’s one example. It’s often cited. It was used in the IPCC report many times. It’s a simple icon, or a representation of what has gone on. It shows temperatures declining from about 1000 AD, then this abrupt warming, obviously this abrupt warming in the last 100 years or so The warmest period in this record was in the 12th century 1146-1195. The coldest period was in the 15 century, latter half of 1400s, and also early 19th century. This warmer period has been described as the Medieval Warm Period; this period in here as the heart of the Little Ice Age. The graph ended in 1998, but in fact I’ll show you “⤠
By the way, I just want to point out this graph which has caused so much discussion: This was the title of the paper: Northern Hemisphere temperatures during the past millennium: inferences, uncertainties and limitations. The whole point of this paper, the whole purpose of the paper was to point out how difficult, how uncertain these reconstructions can be. This was not the statement that says that this is exactly how it was for the last 1000 years. This is a working hypothesis put out there in the literature. People can take potshots at it, as they did. People can improve on it, the normal scientific process. We’re not wedded to this graph as though this was truth which must be defended at all costs.
I certainly don’t get the idea that they were pointing out "how difficult, how uncertain" these reconstructions were to the public. I would say that their public promotions had exactly the opposite effect – that of minimizing the uncertainties.
There’s some statements by Crowley that I find quite ironic “€œ Crowley says “Show me the data.”‘? Both he and Bradley emphasized how "big" the prospective warming was and he illustrated it by showing a large boulder in the middle of the prairies deposited by water. For someone who mostly minimizes the extent of natural variation, this seemed a pretty strange image. Crowley made some remarks about M&M in the question period. Before discussing Crowley in question period, here’s some correspondence – my entire file is here. In March 2005, Crowley asked me for a digital version of our alleged reconstruction. I explicitly told him that we did not purport to make a "reconstruction" and the figure is intended to illustrate the non-robustness of MBH98. Crowley replied on March 21 asknowledging this, but pointing out that some people were "misinterpreting" this and he wanted a copy for his records (which I provided him on April 6.)
Mar. 19, 2005 Crowley to McIntyre Hello, may I get a copy of your time series showing the prominent warming in 1400s? tom crowley
Mar. 19, 2005 McIntyre to Crowley [with attachments] Enclosed are our recent articles. If you read the articles themselves, including the E&E article,you will see that our position is that the MBH98 reconstruction lacks statistical significance – we estimate the R2 in the 15th century step in their version to be 0.0 and we show that their RE statistic is spurious. Mann et al. have refused to disclose their R2 statistic or the digital version of their 15th century step and have only replied with a critique of the R2 statistic, so it’s pretty clear that our estimation of their R2 statistic is pretty close. We also show that their PC method mines for hockey stick shaped series, putting the controversial bristlecone pine series in dominaiton of their PC1. We show that, contrary to their claims of robustness, their results are not robust to the presence/absence of bristlecone pines (showing the effect of the bristlecones in the PC4 if the PC results are calculated correctly.) We show that the high 15th century values result from MBH98-type calculations using the archived version of the Gasp” series (rather than the editing in MBH98), correct principal components calculations and the roster of retained PCs used in MBH98. We do not ourselves argue that 15th century temperatures were high – only that Mann et al. are not entitled to claim under their methods that 20th century values are robustly higher than 15th century temperatures. Mann et al have argued that they can retain 5 PCs in their calculation under a variation of Preisendorfer’s Rule N applied to tree ring networks, which was not mentioned in MBH98 and which cannot be reconciled with retained PC rosters in other networks. Be that as it may, their own current position shows the lack of robustness which we point out. I hope that this clarifies your question. Regards, Steve McIntyre
Mar. 21, 2005 Crowley to McIntyre Steve, thanks for the clarification. would it still be possible to obtain the temperature reconstruction showing the large warming – it is my understanding that some people are misinterpreting that series and i would like to have a copy for my records, tom ps it was not clear to me from your description whether the R2 statistic using your approach is different than 0.0 for the 15th c. – can you clarify? tom crowley
So here’s what Crowley said in the quesiton period (starting about 1:14):
With respect to McIntyre, he calculated that temperatures were warmer in the 1400s. You look at the raw data and it’s just not there. ,,,because McIntyre’s got a lot more press attention with this … Let me tell you. You see the raw data. You do not see a warming in the 1400s, yet he sees a giant warming”⤠
The analogy that I’d like to use is an Index of Leading Indicators. Suppose someone told you that there was a 6% rise in the index in the last year or so, one of the things that you’d want to know is what factors were responsible for that rise: heavy industry, agriculture whatever. Suppose you then look and find that for none of the indices was the rise more than 2%. Then you’d say this is crazy. how can you can get a 6% rise when you see no more than 2% in the individual indices? The only way to explain that is a methodological flaw. And McIntyre has made some points with respect to the Mann et al. reconstruction, but you can’t just focus on Mann et al. When you go back and you look at the raw data, there’s no way that you can crank up a warming in the 1400s that is consistent with the raw data. And so McIntyre himself has made a much more egregious methodological error by estimating this temperature rise that is completely out of whack with the original data. There’s this phrase in the bible that says: criticize not the speck in your neighbor’s eye before you remove the plank from your own. McIntyre is banging away at Mann and there are some legitimate reasons for thinking that Man can be improved now … we actually. ..as Ray says”⤮ but if you lock on Mann you distort the discussion as to what you have in the other evidence.
Given our emails of March 19 and 21, this is a pretty astonishing misrepresentation of our position. But then this is the same guy who accused me of "threatening" him. It’s interesting to contrast Crowley’s assertion that I made a “much more egregious methodological error”‘? with Ammann’s comments a little bit later (1:17:20), shown here:
Let me add one more thing I think that it is very important for the McIntyre and McKitrick research. This has focused on this reconstruction – a very complex statistical reconstruction of Mann and people have reported that they can’t reproduce this thing. With McIntyre, they have found some criticisms where they would think that there is a presetting.of a hockey stick and also a very warm 15th.century. What we are going to put out at the end of this week, it’s going to be submitted, is a tool for the research community, which does reproduce exactly Mann, Bradley, Hughes. It doesn’t say that this is the best reconstruction, it just gives a tool that everyone can apply. It’s going to be very simple, you can run on it your PC, Mac, Linux, whatever you have. Everybody can do this. Everyone can do the McIntyre-McKitrick exercises that they propose in these papers. The only thing that I can tell you here, what you find out is that you can generate this 15th century that is warm. But you’ll find that the statistical model has no skill in prediction. You have statistical measures of how skillful a tool is, based on a particular network that you are looking at – the 1400s, 1400-1449. That’s when they get these warm temperatures. If you look at their papers, they never report what the skill is, because the skill is not there for this particular period. So the warmth that they get in this framework of MBH given the model and data, it has no skill so you should not use it. You can calculate almost anything, but this is something that is not meaningful in a climatological sense. And this tool will be out and everyone here can reproduce this and see these statistical measures. I think that is going to be a good important contribution that we test not only Mann, but in general all the climate reconstructions. That’s where they wanted to point to.
A few points on Ammann. First he says that he has “exactly”‘? reproduced MBH. This is simply untrue. He’s obtained a result that is similar to MBH (as we did in our emulation using their erroneous PC series series and edited Gasp” series). But he doesn’t replicate their residuals for the 15th century for example. It’s not an exact replication. I really wish that these guys would not misrepresent these matters. Secondly, Ammann recognizes that you can get high 15th century results with the MBH98 method with very slight tweaks to the proxy network “€œ as we reported. So when Crowley says that a high 15th century result (which we simply presented as a reductio ad absurdum) is an “egregious methodological error”‘?, Ammann should have stepped in and contradicted Crowley saying that you can get high 15th century results using MBH98 methods.
Ammann realizes that you can get high 15th century without the bristlecones; he argues that you need to leave them in to get adequate statistical skill. This is a quite different argument. But not using bristlecones is hardly an "egregious methodological error"; using them might be if they are flawed proxies. Ammann criticizes us for not reporting verification statistics for "our" reconstruction. However, we were simply showing non-robustness, not asserting a warm 15th century. Ammann is hten really cure about statistics- he talks about statistical measures, but then withholds his R2 statistic, which, as readers of this site know, the MBH98 reconstruction fails. He certainly does not tell the Washington audience that the MBH98 reconstruction has a catastrophic failure of hte R2 statistic. Ironically, if you look at the statistic used to describe reconstruction success elsewhere in this presentation, it is always the r2 statistic: be it the resemblance between the Luterbacher and Mann reconstructions or the relationship of coral àŽⳏ18 to SST – so they would assume, to the extent that they thought about it, that Ammann was calculating statistics similarly to Crowley and Bradley, rather than using a highly questionable and almost certainly spurious RE statistic of MBH98.