Springen naar inhoud

Joep engels in trouw


  • Log in om te kunnen reageren

#1

Hans Erren

    Hans Erren


  • >250 berichten
  • 411 berichten
  • Ervaren gebruiker

Geplaatst op 23 april 2007 - 23:10

Trouw zaterdag 21 april 2007

Het was geen fraaie vertoning. Toen twee jaar geleden de 'hockeystick” onder vuur kwam te liggen, een icoon in de klimaatwetenschap, beet de geestelijke vader ervan fel van zich af. Michael Mann diskwalificeerde zijn critici. hij weigerde inzage te geven in zijn eigen gegevens en hij verweet het wetenschappelijk tijdschrift dat het een flutartikel had gepubliceerd. Dat had hij allemaal beter niet kunnen doen. Zijn reactie bevestigde het beeld dat velen van de klimatologenwereld hebben: een gesloten bastion dat meer politiek bedrijft dan wetenschap beoefent, en dat geen kritiek toestaat. Begrijpelijk was het wel. Na een jarenlange achtervolging hadden zijn critici hem eindelijk te pakken gekregen. In 1997 had Michael Mann in het vakblad Nature een reconstructie gepubliceerd van de temperatuur op het noordelijk halfrond in de laatste duizend jaar. Aan de hand van boomringen, ijskernen en andere historische bronnen, en met behulp van ingewikkelde statistiek had Mann een grafiek gefabriceerd die liet zien hoe de temperatuur sinds de vroege Middeleeuwen gestaag daalde om na 1900 plots te stijgen.

Verontwaardiging
De grafiek, vanwege zijn vorm de 'hockeystick' genoemd, kreeg een prominente plaats in het vorige rapport van het Intergovernmental Panel on Climate Change (IPCC) uit 2001. In een hoofdstuk dat mede geschreven was door diezelfde Michael Mann. Dat maakte hem tot een interessant doelwit: als je de hockeystick onderuithaalde, schoot je een flink gat in het IPCC-rapport. Twee Canadezen, de econoom Ross McKitrick en de statisticus Steven McIntyre, plozen de reconstructie van Mann na en ontdekten dat ze met kleine aanpassingen een heel andere grafiek konden krijgen. Manns conclusie dat de huidige opwarming uniek is in dit millennium, leek helemaal niet zo hard. De twee leurden een paar jaar met hun artikel, geen enkel respectabel blad wilde het plaatsen. De verontwaardiging hierover zwol aan op het internet totdat Geophysical Research Letters in 2005 besloot de Canadese kritiek te publiceren. Tot groot verdriet van Michael Mann.

Geuzen
Het was wel volgens goed wetenschappelijk gebruik. Nu de kritiek openbaar was, kon iedereen nagaan wat er niet deugde aan het werk van Mann en of de kritiek van M&M - zoals de Canadezen in de wandelgangen werden genoemd – hout sneed. Na een jaar was het beeld duidelijk. Inderdaad, de hockeystick was niet helemaal correct in elkaar gezet. Maar ook M&M hadden wat al te gretig naar hun antwoord toegewerkt. En het belangrijkste: de hoofdconclusie dat de huidige opwarming uniek is voor dit millennium, bleef overeind.


Joep Engels had zich beter wat dieper in de materie kunnen verdiepen, hier is een samenvatting van de Hockeystickaffaire:
http://www.climateau...#comment-105953

Can someone explain just what is going on here? Did MM actually “remove” 80% of the proxy data for 15th century”? I would assume that in “Rutherford, S., Mann, M.E., Osborn, T.J., Bradley, R.S., Briffa, K.R., Hughes, M.K., Jones, P.D., Proxy-based Northern Hemisphere Surface Temperature Reconstructions: Sensitivity to Methodology, Predictor Network, Target Season and Target Domain, Journal of Climate, in press (2005)” in which RealClimate claims has debunked MM by peer reveiw, are actually the same guys who did the work MBH9X?? and other contested documents in this blog? I mean aren’t all these guys the ones who have worked together using similar assumptions and should not state “peer-review”, but say “in defense of” their work…etc.


The answer is no. The history of the dispute is obviously convoluted, but Rutherford et al 2005 is an intentional misrepresentation of the situation. Our first article observed that the MBH98 principal components were incorrectly calculated. MBH had stated that they had used conventional principal components but this was untrue on a variety of counts. Instead of using a principal components algorithm, they standardized the data on a short segment and used SVD on the short segment. They also used an unreported stepwise procedure. They claimed that their results were “robust” to the presence/absence of all dendroclimatic indicators.

In our 2003 article, we noted that there were many missing values in the tree ring networks and calculated principal components over the maximum period for which all records in the network were available. (Prior to doing so, we had asked Mann for clarification of the procedures used, which he refused to provide). We illustrated the problems with the PC calculations in our 2003 article on the Australian network where the changing availability of records was not as much of an issue as the North American network, showing that the Mann PC calculation yielded a HS, whereas a conventional calculation did not. In our 2003 article, we observed problems but were not able to fully diagnose them.

After our 2003 article, Mann made much information available and we were able to diagnose the biased PC methodology that had led to the HS shaped PC1s. This was subsequently described in our two 2005 articles. After our 2003 article, Mann also disclosed the stepwise PC methodology - a methodology that is hardly “conventional” and we implemented this methodology in our 2005 articles (published in January, February.) In our two articles, we describe a variety of permutations and combinations of results, all turning on the weighting of bristlecone pines in the various reconstructions. It’s not whether “80% of the North American” tree ring records are in or out, it’s whether bristlecones are heavily weighted or not. Mann’s PC methodology weighted bristlecones heavily so that the HS is essentially made of bristlecone.

Mann had done (but did not report) a CENSORED calculation in which he calculated the results without bristlecones, obtaining a non-HS result as we reported (and as Ammann and Wahl grudgibgly confirm under such circumstances). Knowing that their results were not robust to the presence/absence of bristlecones, they nonetheless claimed (see Mann et al 2000) that their results were robust to the presence/absence of all dendroclimatic indicators.

The Rutherford et al article that you mention was published in late 2005 and did not discuss either of our 2005 articles, which were in print long before Rutherford et al was published. In addition, the key findings of our 2005 articles were known to Mann in early 2004. I objected in writing to Andrew Weaver, editor of Journal of Climate, in January 2005 (based on an online version posted up a year before publication) that Rutherford et al was misleading on a number of counts, not least of which was their failure to consider the 2005 articles.


The claim that we got results by omitting 80% of the data is a pure misrepresentation. The germ of truth in it is this: in our 2003 E&E article we couldn’t figure out how Mann dealt with segments of missing data in his PC analysis, so we truncated some matrices at points that allowed the analysis to go through (though it was not 80% of the data). We later found out that Mann had used a splicing procedure where he recalculated principal components on different sub-networks and spliced the results together. There was no way to reproduce this methodology from his disclosure. However, after much experimentation with Mann’s newly-released archive (following our 2003 paper), Steve worked out the most likely splicing sequence, and as of early 2004 we have used the entire network in all replication work. Everything that has been debated since then has been on the basis of both teams using the entire data base. The best coverage of this is our 2005 E&E paper in which we explain the exact steps needed to go between the hockey stick of Mann and the non-hockey stick shape we had earlier obtained. As Steve mentions, to get a hockey stick, Mann needs to include the small group of bristlecone pine data, which (for reasons detailed in the 2005 E&E paper) are likely contaminated and do not belong in a temperature reconstruction model–a point upheld by the NAS panel. Remove the bristlecones, and there is no hockey stick shape left, regardless of any other methodological decision, regardless of data inclusions or exclusions, etc. Mann had discovered this in his own sensitivity analysis, which was posted on his FTP site but which hadn’t been reported in his paper. And with or without the bristlecones, the results lack statistical significance, a point we discuss in our 2005 GRL paper and exchanges with Huybers and von Storch-Zorita.

Don't just talk, show your work.

Dit forum kan gratis blijven vanwege banners als deze. Door te registeren zal de onderstaande banner overigens verdwijnen.




0 gebruiker(s) lezen dit onderwerp

0 leden, 0 bezoekers, 0 anonieme gebruikers

Ook adverteren op onze website? Lees hier meer!

Gesponsorde vacatures

Vacatures