cc: Martin Juckes , anders@misu.su.se, Eduardo.Zorita@gkss.de, hegerl@duke.edu, esper@wsl.ch, k.briffa@uea.ac.uk, m.allen1@physics.ox.ac.uk date: Thu, 01 Feb 2007 13:59:11 +0000 from: Nanne Weber subject: Re: mitrie: revision to: Tim Osborn Hi Martin, started to look at your replies. Below follow my suggestions for the reply to referee 1. I think that it could be more precise and also stress more that we do present new results (this person is over-presumptuous about the Jones/Mann and Mann papers, which are purely review and repeat each other). I was also wondering about Tim's point. Just 1 referee is below CPD standards. What did the editor say? That is all for now, Nanne Tim Osborn wrote: > Hi Martin -- will look in detail at this soon. I looked on COPD and > could find comments from only one solicited reviewer and no feedback > from the editor, plus a number of unsolicited comments. Did you get > more reviews and/or editors comments directly to you? Apologies if > you did and already circulated them, but I couldn't find them on COPD > or in my email. With regards climate2003.com, I suspect this may be > just temporary downtime as it has disappeared at times in the past > only to subsequently re-appear. I don't think it's worth making much > of an issue over it. Cheers, Tim > ============================================================================== REPLY to referee #1. The format is point of criticism, followed by reply to that point. 1) First of all, in my opinion this paper does not provide much new results Sections 2 and 3 are purely intended as reviews of recent reconstructions and the criticism of the IPCC consensus, respectively. As such, they do not contain new results. We do take a line of approach which is different from earlier reviews by Jones and Mann (2004) and Mann (2007) and for this reason our review is included in the paper. In section 4 we compare the impact of varying methods versus varying data collections. We are not aware that this issue is covered by the review papers of Jones and Mann (2004) and Mann (2007) or in any other published study. 2) Actually the introduction until page 20 is quite interesting and gives a nice overview, however, also here there is not much new. It does possibly not need to be so as it should be a review. See reply to 1) 3) The authors selected two reconstruction techniques and compared them with each other. That choice is arbitrary and the results and interpretations are not convincing. What are the arguments that one method should be used in favor of the other? The two selected methods (inverse regression and scaled composites), or variants thereof, are used in all reconstructions considered except those based on low-resolution records alone (HPS2000 and OER2005). Scaled composites is used by JBB, ECS, MSH and HCA (?), inverse regression by MBH (?). A trawl through a sample of text books on statistical theory will reveal a huge range of potential techniques, some of which are listed by the reviewer. However, they have not been used in NH temperature reconstructions up till now and therefore we do not discuss (or evaluate) them. The two reconstruction techniques represent different assumptions about the quality of the data. This needs to be explained more clearly in the manuscript (change [1] below). In our comparison it appears that, for millennial reconstructions, the simplest technique, which makes the smallest number of assumptions about the data, works best. What about filtered uncertainties??? 4) The second related point to check whether a specific method seems to perform 'better' would be to use coupled paleo runs. There are two 1000 year long runs which can be used. The difficulty with using model paleo runs to test the method is that we do not have a comprehensive predictive model of how the proxies respond to temperature. Such tests typically model the proxies by prescribing a linear dependence on temperature and adding random (white or red) noise. Reality is clearly more messy. There are, however, important issues which can be addressed with such paleo runs (see eg Mann, 2007). As explained in the Appendix, it is possible to construct a situation in which one or the other method is optimal. 5) I suppose, the authors show annual averaged temperature data? That is not clear. Further, I have my doubts about the choice of the predictor data in the new union reconstructions. Yes, we show annual averages (now clarified in the introduction). What about the choice of predictor data??? 6) Other issues that might be worth addressing are: Sensitivity to the calibration period, whether to detrend or not as well as the color of noise related to the model data. We do not aim to address all issues involved in millennial temperature reconstructions. That would indeed be repeating work that has been done by others (and references are given in our paper). Sensitivity to calibration period: we refer to one sensitivity test, extending the calibration period to 1985. This issue has been dealt with in more detail in a paper by Zorita, Gonzalez-Rouco and von Storch which is now accepted for publication in the Journal of Climate. 7) It would also help if the authors could come up with some recommendation concerning the use of those methods for different applications. Also, how do those methods perform if, like in the case of Moberg et al. (2005, Nature) proxies with different temporal resolutions are combined? Could the authors say anything about the methods that aim at sub-hemispheric reconstruction, resolve seasons and other climate parameters such as rainfall? We do give a clear recommendation on the choice of method (eg in the abstract). What about the combination of high/low resolution proxies??? We do not intend to talk about sub-hemispheric reconstructions or rainfall reconstructions in this study. There is clearly valuable work being done in that direction.