From: Gifford Miller To: Darrell Kaufman Subject: Re: Fwd: Your Science manuscript 1173983 at revision Date: Tue, 26 May 2009 16:23:05 -0400 Cc: David Schneider , Nick McKay , Caspar Ammann , Bradley Ray , Keith Briffa , Miller Giff , Otto-Bleisner Bette , Overpeck Jonathan Darrell (from AGU Toronto): Great news from Science! A quick comment on Amplification and signal to noise issues (comment 1 below). It think you meant that the referee felt that Arctic amplificaton did not translate to a more robust signal because the noise would be equally amplified. I don't know that we can challenge the "climate noise" but we can make the case that the "proxy noise", that is, the uncertainty in proxy calibration, is, as far as I know, the same in the Arctic as in lower latitudes. Consequently, the larger temperature signal expected in the Arctic can be more reliably detected by our proxies because it is more likely to exceed the sensitivity limits of our proxies. If we assume the "climate noise" is more or less gaussian, then we should be better able to detect the relatively subtle temp changes of the Holocene in the Arctic than elsewhere. Giff >Co-authors: >I just received the reviewers' comments and editor's decision on our >SCIENCE manuscript (attached). The decision isn't final, but it >looks like good news, with very reasonable revisions. Reviewer #1 >had nothing substantial to suggest. Reviewer #2 was rather thorough. >I think I can address his/her suggestions but could use some help >with three: > >(1) The reviewer challenged our assertion that, because climate >change is amplified in the Arctic, the signal:noise ratio should be >higher too. We don't have more than 1 sentence to expand on the >assertion in the text. We could plead the case to editor and hope >that it doesn't trip up the final acceptance, or we could omit it >from the text. Suggestions? > >(2) The reviewer suggested that, if we are concerned about outliers >influencing the mean values of the composite record, we should >attempt a so-called "robust" regression procedure, such as median >absolute deviation regression. Does anyone have experience with this? > >(3) The reviewer was concerned that we overestimated the strength of >the relation between temperature and insolation in the long CCSM >simulation. Namely s/he criticized the leveraging effect of the one >outlier in the model-generated insolation vs temperature plot (Fig. >4b), and suggested that we use 10-year means instead of 50 year. >Dave: you up for this, please? > >Please forward any input to me and I'll compile them, and let you >all have a look before I submit the final revisions. I'm hoping we >can turn this around this week. > >Thanks. >Darrell > -- Gifford H. Miller, Professor INSTAAR and Geological Sciences University of Colorado at Boulder