cc: ,k.briffa@uea.ac.uk date: Fri, 23 Jun 2006 09:32:23 +0100 from: Phil Jones subject: RE: Surface temps,, under embargo., seeking reaction, analysis to: "Wahl, Eugene R" ,,, Dear All, Won't have a chance to read this report in detail as off to Bergen for IPCC tomorrow. Had a quick scan last night. One thing that is very odd is their statement about less confidence is the warmth of the 1990s and the year 1998, cf their higher confidence in the warmth of the last few decades compared to the last 400 and 1000 years. The second statement wouldn't be possible without the 1990s and 1998. The greater errors on individual years cf decades and eventually centuries in the instrumental record is wrong. They haven't read Jones et al. (1997) properly. Some errors cancel but the biases don't. This will become clearer when Brohan et al. (2006) eventually appears in JGR. Apart from 2005, 1998 is warmer than all other instrumental years. Also they should have referenced Jones et al (2003) on the seasonal differences in long instrumental records and documentary series, when discussing Figure 2.2. Seems as though they have completely ignored the likelihood (certainty) of different seasonal rates of change in the past. Cheers Phil At 02:07 22/06/2006, Wahl, Eugene R wrote: >Hello Ray, Phil, Mike, Caspar: > >Here are three sets of thoughts I sent just now to Andy Revkin on the NAS >report. [I asked him to use these for his own purposes only for now, and >not to cite.] > >Hello Andy: > >Thanks for forwarding the pre-release to me. I have looked it over and >think you have done a fair job of characterizing the committee's work and >conclusions. > >The one thing I have particular issue with is somewhat technical, and >possibly not germane therefore to your article, but I want to give it to >you in any case. This has to do with the issue of "spurious Principal >Components" cited on pp. 86-87 of the report. The analysis there that >replicates the work of McIntyre and McKitrick in this regard I think is >insufficient. I do not argue what is given, but there is more information >that we have in my article with Caspar Ammann on this that I believe >should have been cited, and was left out. In short, we have showed that >whether one uses the MBH centering convention (over the period 1902-1980) >for Principal Component caluclation or over the entire length of a set of >proxy records (600 years in the example) makes virtually no difference in >the resulting reconstruction (on the order of 0.05 deg C) in the MBH >framework. All the different centering conventions do is to spread the >actual climate signal differently across the range of the principal >components, and as long as an adequate number of them are retained, then >the impact on the reconstructions is nearly zero. The issue raised by MM >in this regard actually hinges on whether the proxy data going into the >Principal Component caluclations are first normalized (divided by the >standard deviation) or not, given the method of Principal Component >calculation MM employed. If not, then a larger number of Principal >Components needs to be retained, but even in this case if the appropriate >number IS retained, then the resulting reconstruction is unchanged. > >Along with the abstract case the committee shows, they should also have >included this empirical information -- which was available to them in our >paper. They have, in this particular situation, helped maintain confusion >in this regard, rather than reduce it. That is a bit frustrating, >although overall I think they have done a decent job. > >Thanks for all your hard work on covering this issue so well. > > >Hi Andy: > >Two other thoughts. > >1) In my estimation, the committee also did something of a disservice by >not noting that, although the MBH-type reconstruction does have poor >year-to-year validation performance before the 19th century according to >the validation instrumental data used (it captures mean offsets from 20th >century temperatures though), other indirect tests of year-to-year >performance of the reconstruction show very good performance. This is >also shown in the Wahl and Ammann paper that the committee cited in a >number of places on other issues. For example, the MBH reconstructions >into the 18th century for Europe have very good r^2 relationship with the >reconstructions done by Juerg Luterbacher and colleagues for Europe, based >on a completely separate proxy data set mostly driven by instrumental >temperature records. Also, the MBH method as used by Von Storch et al. in >their model-based work shows extremely good year-to-year performance in >general, even though we (Wahl, Ritson, and Ammann, 2006) took some issue >with the results VS et al show in terms of lower frequency loss of >"amplitude" by MBH. > >I'm left with a vague impression that the Wahl and Ammann (in press) paper >was somewhat selectively (or possibly myopically) cited; in any case I >feel strongly this way concerning the Principal Component issue I wrote >about in my last message. > > >2) I think the committee also missed the mark to a degree by their >emplasis on the usefulness of the CE statistic over the RE >statistic. This is on pp. 88-91 of the text (same pages reference as for >(1) above). They don't emphasize that RE rewards detection of a >difference in mean between the calibration and validation periods, which >CE cannot -- by construction -- detect. A CE value near 0 in validation >DOES indicate that year-to-year tracking is very poor, but this does not >say whether the detection of the difference in means is occuring or >not. So, as with my last message, I don't so much argue with what the >committee did say, but rather find a sense of incompletion that I think is >unfortunate. > > >Peace, Gene >Dr. Eugene R. Wahl >Asst. Professor of Environmental Studies >Alfred University > >607-871-2604 >1 Saxon Drive >Alfred, NY 14802 > Prof. Phil Jones Climatic Research Unit Telephone +44 (0) 1603 592090 School of Environmental Sciences Fax +44 (0) 1603 507784 University of East Anglia Norwich Email p.jones@uea.ac.uk NR4 7TJ UK ----------------------------------------------------------------------------