date: Fri, 04 Feb 2005 15:54:15 -0500 from: "Michael E. Mann" subject: Fwd: Re: FW: "hockey stock" methodology misleading to: Phil Jones , rbradley@geo.umass.edu, tom crowley , tom crowley , mhughes@ltrr.arizona.edu, rbradley@geo.umass.edu, Keith Briffa Date: Fri, 04 Feb 2005 15:52:53 -0500 To: Andy Revkin From: "Michael E. Mann" Subject: Re: FW: "hockey stock" methodology misleading Hi Andy, The McIntyre and McKitrick paper is pure scientific fraud. I think you'll find this reinforced by just about any legitimate scientist in our field you discuss this with. Please see the RealClimate response: [1]http://www.realclimate.org/index.php?p=111 and also: [2]http://www.realclimate.org/index.php?p=114 The Moberg et al paper is at least real science. But there are some real problems with it (you'll want to followup w/ people like Phil Jones for a 2nd opinion). While the paper actually reinforces the main conclusion of previous studies (it also finds the late 20th century to be the warmest period of the past two millennia), it challenges various reconstructions using tree-ring information (which includes us, but several others such as Jones et al, Crowley, etc). I'm pretty sure, by the way, that a very similar version of the paper was rejected previously by Science. A number of us are therefore very surprised that Nature is publishing it, given a number of serious problems: Their method for combining frequencies is problematic and untested: A. they only use a handful of records, so there is a potentially large sampling bias. B. worse, they use different records for high-frequencies and low-frequencies, so the bias isn't even the same--the reconstruction is apples and oranges. C. The wavelet method is problematic. We have found in our own work that you cannot simply combine the content in different at like frequencies, because different proxies have different signal vs. noise characteristics at different frequencies--for some records, there century-scale variability is likely to be pure noise. They end up therfore weighting noise as much as signal. For some of the records used, there are real age model problems. The timescale isn't known to better than +/- a couple hundred years in several cases. So when they average these records together, the century-scale variability is likely to be nonsense. D. They didn't do statistical verification. This is absolutely essential for such reconstructions (see e.g. the recent Cook et al and Luterbacher et al papers in Science). They should have validated their reconstruction against long-instrumental records, as we and many others have. Without having done so, there is no reason to believe the reconstruction has any reliability. This is a major problem w/ the paper. It is complicated by the fact that they don't produce a pattern, but just a hemispheric mean--that makes it difficult to do a long-term verification. But they don't attempt any sort of verification at all! There are some decades known to be warm from the available instrumental records (1730s, some in the 16th century) which the Moberg reconstruction completely misses--the reconstruction gives the impression that all years are cold between 1500 and 1750. The reconstruction would almost certainly fail cross-validation against long instrumental records. If so, it is an unreliable estimate of past changes. We're surprised the Nature Reviewers didn't catch this. E. They also didn't validate their method against a model (where I believe it would likely fail). We have done so w/ our own "hybrid frequency-domain" method that combines information separately at low and high-frequencies, but taking into account the problem mentioned above. This is described in: Rutherford, S., Mann, M.E., Osborn, T.J., Bradley, R.S., Briffa, K.R., Hughes, M.K., Jones, P.D., [3]Proxy-based Northern Hemisphere Surface Temperature Reconstructions: Sensitivity to Methodology, Predictor Network, Target Season and Target Domain, Journal of Climate, in press (2005). In work that is provisionally accepted in "Journal of Climate" (draft attached), we show that our method gives the correct history using noisy "pseudoproxy" records derived from a climate model simulation with large past changes in radiative forcing. Moberg et al have not tested their method in such a manner. F. They argue selectively for favorable comparison w/ other work: (1) Esper et al: when authors rescaled the reconstruction using the full instrumental record (Cook et al, 2004), they found it to be far more similar to Mann et al, Crowley and Lowery, Jones et al, and the roughly dozen or so other empirical and model estimates consistent w/ it. Several studies, moreover [see e.g.: Shindell, D.T., Schmidt, G.A., Mann, M.E., Faluvegi, G., [4]Dynamic winter climate response to large tropical volcanic eruptions since 1600, Journal of Geophysical Research, 109, D05104, doi: 10.1029/2003JD004151, 2004.] show that extratropical, land-only summer temperatures, which Esper et al emphasises, are likely to biased towards greater variability--so its an apples and oranges comparison anyway. (2) von Storch et al: There are some well known problems here: (a) their forcing is way too large (Foukal at al in Science a couple months back indicates maybe 5 times too large), DKMI uses same model, more conventional forcings, and get half the amplitude and another paper submitted recently by the Belgium modeling group suggests that some severe spin-up/initialization problems give the large century-scale swings in the model--these are not reproducible. (3) Boreholes: They argue that Boreholes are "physical measurements" but many papers in the published literature have detailed the various biases in using continental ground surface temperature to estimate past surface air temperature changes--changing snow cover gives rise to a potentially huge bias (see e.g. : Mann, M.E., Schmidt, G.A., [5]Ground vs. Surface Air Temperature Trends: Implications for Borehole Surface Temperature Reconstructions,Geophysical Research Letters, 30 (12), 1607, doi: 10.1029/2003GL017170, 2003). Methods that try to correct for this give smaller amplitude changes from borehole temperatures: Mann, M.E., Rutherford, S., Bradley, R.S., Hughes, M.K., Keimig, F.T., [6]Optimal Surface Temperature Reconstructions using Terrestrial Borehole Data, Journal of Geophysical Research, 108 (D7), 4203, doi: 10.1029/2002JD002532, 2003] [[7]Correction(Rutherford and Mann, 2004)] Most reconstructions and model estimates still *sandwich" the Mann et al reconstruction. See e.g. figure 5 in: Jones, P.D., Mann, M.E., [8]Climate Over Past Millennia, Reviews of Geophysics, 42, RG2002, doi: 10.1029/2003RG000143, 2004. Ironically, MM say our 15th century is too cold, while Moberg et al say its too warm. Hmmm.... To recap, I hope you don't mention MM at all. It really doesn't deserve any additional publicity. Moberg et al is more deserving of discussion, but, as outlined above, there are some real problems w/ it. I have reason to believe that Nature's own commentary by Schiermeier will actually be somewhat critical of it. I'm travelling and largely unavailable until monday. If you need to talk, you can possibly reach me at 434-227-6969 over the weekend. I hope this is of some help. Literally got to run now... mike At 02:14 PM 2/4/2005, Andy Revkin wrote: Hi all, There is a fascinating paper coming in Nature next week (Moberg of Stockholm Univ., et al) that uses mix of sediment and tree ring data to get a new view of last 2,000 years. Very warped hockeystick shaft (centuries-scale variability very large) but still pronounced 'unusual' 1990's blade. i'd like your reaction/thoughts for story i'll write for next thursday's Times. also, is there anything about the GRL paper forthcoming from Mc & Mc that warrants a response? I can send you the Nature paper as pdf if you agree not to redistribute it (you know the embargo rules). that ok? thanks for getting in touch! andy ______________________________________________________________ Professor Michael E. Mann Department of Environmental Sciences, Clark Hall University of Virginia Charlottesville, VA 22903 _______________________________________________________________________ e-mail: mann@virginia.edu Phone: (434) 924-7770 FAX: (434) 982-2137 [9]http://www.evsc.virginia.edu/faculty/people/mann.shtml ______________________________________________________________ Professor Michael E. Mann Department of Environmental Sciences, Clark Hall University of Virginia Charlottesville, VA 22903 _______________________________________________________________________ e-mail: mann@virginia.edu Phone: (434) 924-7770 FAX: (434) 982-2137 [10]http://www.evsc.virginia.edu/faculty/people/mann.shtml