Jeff on Steig RegEM

Jeff has made some progress on the basis that PC was applied to the AVHRR data before RegEM. See a good post here .

This is obviously inconsistent with claims at by Gavin Schmidt and Eric Steig, but we’ve known for a while that their explanations are not accurate.

This is a different exercise from assessing whether the method makes any sense and an effort that is required only because of inaccurate documentation, lack of data and continued obfuscation by PR Challengers Schmidt and Steig.


  1. Level_Head
    Posted Apr 5, 2009 at 8:18 PM | Permalink | Reply

    An image of Antarctica showing correlation of cloudiness (and thus infilled data) to higher temperatures would be interesting.

    ===|==============/ Level Head

  2. tetris
    Posted Apr 5, 2009 at 9:48 PM | Permalink | Reply


    I seem to recall that the Steig, Mann, et. al. paper purported to show – from a very convoluted statistical “calculations” by any stretch of the imagination- [ref: Trenberth's acerbic comment to the effect that one can not produce data from where none exist] a temperature increase in Antarctica of 0.1C per decade with a margin of error of +/- 0.3C. If that is what it boils down to, it goes right to the heart of the argument you made about the PR “packaging” of the Trouet paper. Forget about the actual data, it’s only the bumfluff that counts.

    PS: I think many would welcome your following up on your musings following your return from SE Asia, and put a proper [and certainly much needed] focus on the role of water vapour.

  3. dearieme
    Posted Apr 6, 2009 at 8:36 AM | Permalink | Reply

    It all makes work for the working man to do.

  4. woodNfish
    Posted Apr 6, 2009 at 12:33 PM | Permalink | Reply

    I know this is tedious, but I did check the acronym list before deciding to ask this question. Will someone please tell me what the acronym “PR” in “PR Challengers Schmidt and Steig” means? Thanks in advance.

  5. Posted Apr 6, 2009 at 1:03 PM | Permalink | Reply

    RE WoodNfish, #4,
    R is supposed to be for Reconstruction, as in Paleoclimate Reconstruction Challenge. See Ammann’s April Fool’s Joke thread and But Steve’s point is that after a year of NOAA funding, it is still no more than a “Public Relations” stunt.

  6. Posted Apr 6, 2009 at 1:49 PM | Permalink | Reply

    Jeff Id’s posts are looking pretty devastating for Steig & Co.

    Yesterday’s post Steig AVHRR Recon… (dated April 5) shows that using 3 PCs of cloudmaskedAVHRR.txt and regpar = 3, he can almost (if not exactly) replicate Steig’s ant_recon.txt. His continent-wide trend is 0.108 °C/decade vs 0.118 °C/decade using ant_recon.txt, but his Figure 7 shows that the match is very close. His Figure 8 and 9 show that he gets basically the same distribution of trends as in Steig’s cover image, with warming throughout W.Ant.

    Yet in his April 4 post, “What’s wrong with RegEM”, he finds that when he cranks up the permitted geographical detail, by using 20 PCs and regpar =10, the continental trend drops to 0.038 °C/decade, and the geographical distribution changes dramatically — the warming becomes concentrated in the peninsula itself, with the Ross shelf, much of W. Ant., and a sizeable sector of E. Ant. actually cooling.

    It’s not clear what the mechanism is yet, but it looks as though RegEM, when coupled with oversmoothed low-PC data, does somehow “smear” local events over a wider region, and thereby gives them undue importance, as Jeff has suspected for some time.

    Discussion continues over on Jeff’s site.

    • Jeff C.
      Posted Apr 6, 2009 at 4:17 PM | Permalink | Reply

      Re: Hu McCulloch (#6),

      It’s not clear what the mechanism is yet, but it looks as though RegEM, when coupled with oversmoothed low-PC data, does somehow “smear” local events over a wider region, and thereby gives them undue importance, as Jeff has suspected for some time.

      I think you are absolutely right Hu. Another thing that appears to be happening is temporal smearing. Most of the AVHRR data I have reviewed shows an early rise in temps, followed by a later, slightly smaller drop in temps. Steigs recon (and Comiso’s data) should look something like a fairly dramatic inverted “V”. Instead they have almost a continuous positive slope over time, giving the impression of an unrelenting warming trend.

      The upside down V has been flattened out, most likely the result of the “climatolgical mean” filter (the +/- 10 deg C threshold) and the reduction to 3 PCs. This may be the same phenomena as the spatial smearing, just viewed from a different domain.

  7. vg
    Posted Apr 6, 2009 at 10:44 PM | Permalink | Reply

    So would it be appropriate on this basis/stage ect to ask Nature to withdraw this work or at least publish a letter with the analysis that shows that Antarctica is not/has not recently warmed significantly as stated by Steig et al?

  8. Posted Apr 17, 2009 at 7:56 AM | Permalink | Reply

    I did the same reconstruction as above with the peninsula temp stations and satellite data removed.

Post a Comment

Required fields are marked *



Get every new post delivered to your Inbox.

Join 3,114 other followers

%d bloggers like this: