a) Books
von Storch, H., S. Güss und M. Heimann, 1999: Das Klimasystem und seine Modellierung. Eine Einführung. Springer Verlag ISBN 3-540-65830-0, 255 pp(Die Darstellung des Buches ist weiterhin valide, selbst wenn die Entwicklungen des letzten Jahrzehnts nicht enthalten sind - diese sind aber im wesentlichen Verfeinerungen und Vertiefungen, die die Aussagen von 1999 nicht in Frage stellen. )
and
Washington, W.M. and C.L. Parkinson, 2005: An Introduction to Three-Dimensional Climate Modelling. 2nd edition, University Science Books, Sausalito, California, 354 pp.
b) Review articles: see block "Climate Models and Modeling" of WIREs Climate Change, in particularPeter Müller: Constructing climate knowledge with computer models
Published Online: Wed Jun 30 00:00:00 EDT 2010
DOI:10.1002/wcc.60
Julia C. Hargreaves: Skill and uncertainty in climate models
Published Online: Thu Jun 24 00:00:00 EDT 2010
DOI:10.1002/wcc.58
Susanne L. Weber: The utility of Earth system Models of Intermediate Complexity (EMICs)
Published Online: Mon Jan 11 00:00:00 EST 2010
DOI:10.1002/wcc.24
(I have been responsible domain editor for these articles; the function has been taken over recently by Eduardo Zorita). More review articles are in preparation.
Has there been any serious exclusion of spatio-temporal chaos effects in the climate system and models thereof?
ReplyDeleteThe assumption of all current models that spatio-temporal chaos averages out at large scales (space and time) is not at all trivial but hasn't been treated rigorously to my knowledge (perhaps you can help with refs?).
Isn't this one of the most central questions to the accountability of climate science?
See also current discussion over at http://judithcurry.com/2011/02/10/spatio-temporal-chaos/
Herr von Storch,
ReplyDeleteweil es im weiteren Sinne um Buchveröffentlichungen geht und die Zahl der Kommentare gerade recht übersichtlich ist, möchte ich mich mit einer Frage zur Regression an Sie (oder fachkundige Leser) wenden, die ich mithilfe von von Storch/Zwiers, "Statistical Analysis in Climate Research", Kap. 8.2ff nicht geklärt bekommen konnte.
Ich treffe in empirischen Untersuchungen immer wieder mal auf die Vorgehensweise, daß Trends mithilfe gewöhnlicher OLS-Regression quantifiziert werden, die Signifikanz dieses Trends dagegen mit zusätzlichen nichtparametrischen Tests (zB Mann-Kendall-Vorzeichentest) abgesichert wird.
Mir erscheint das intuitiv problematisch, einen Trend mit einem Modellverfahren zu quantifizieren, an dessen Anwendbarkeit man ja eigentlich nicht glaubt (sonst müßte man ja auch das OLS-Signifikanzresultat glauben) und die Bestätigung dieses eigentlich nicht vertrauenswürdigen Wertes dann durch andere, geeignetere Verfahren zu erreichen. Es besteht also nach meinem Empfinden die Gefahr, daß ein quantitativ unsicheres Ergebnis mit einem 'Gütestempel' versehen wird, den es eigentlich nicht verdient.
Leider habe ich auch in der Fachliteratur keine explizite Besprechung dieses Problems auftreiben können, zumindest keine, die ich als Ing. mit soliden Grundlagenkenntnissen als solche hätte identifizieren können. Auch die Veröffentlichungen selbst, die dieses Verfahren anwenden, verwiesen - wenn überhaupt - auf andere Veröffentlichungen, die dies ebenfalls taten, aber nicht auf die Grundlagen, die die Zulässigkeit dieser Vorgehensweise erörterten. Was mich schon etwas überrascht hat.
@ 1
ReplyDeleteIt is quite probable that important aspects of fluid dynamics in the climate context are still poorly known. We should keep in mind that we dont even have a satisfactory theory of turbulence, not only for climate science but in general. However, we should not just conclude that it is impossible to predict anything useful. In the end, the key question is the interplay between the external drivers and the internally chaotically generated variability. The climate is indeed complex but we all know that temperatures in June in the Northern Hemisphere are always warmer than the temperatures in December and so it is a sure bet to predict that this will happen in the future as well. This is obviously caused by the influence of the external forcing, and it seems that chaos and turbulence, are not very powerful to overwhelm this.
More specific to your question 'The assumption of all current models that spatio-temporal chaos averages out at large scales (space and time) '.. . I dont think this is true. Climate models do not make this assumptions, and by the same token climate models do not 'include' quasi oscillations like the NAO or ENSO. Climate models try to solve the dynamical equations that describe climate. If one of the product of these equations is spatio-temporal chaos or quasi oscillations, then ideally climate models would also show them. I would however refrain of accepting without a close critical look all theories with fancy names that are regularly proposed to explain climate phenomena, like regime shifts, stochastic resonance, etc, etc. These maybe attractive theories and for the more mathematically oriented even beautiful theories. Its relevance for climate is however far from proven.
It is sometimes surprising how quickly these fancy theories are grandly presented and cheered and on the other hand, the simple, old, measured greenhouse effect is dismissed as irrelevant. Not everything is greenhouse effect, for sure, but to assert that 'Isn't this one of the most central questions to the accountability of climate science?' is quite over the top.
@eduardo
ReplyDelete> If one of the product of these equations is spatio-temporal chaos or quasi oscillations, then ideally climate models would also show them.
Exactly, and that is the problem: This alone has the potential to destroy any long term prediction ability of climate models.
> It is sometimes surprising how quickly these fancy theories are grandly presented and cheered and on the other hand, the simple, old, measured greenhouse effect is dismissed as irrelevant. Not everything is greenhouse effect, for sure, but to assert that 'Isn't this one of the most central questions to the accountability of climate science?' is quite over the top.
Accountability of climate science builds on predictability of climate models. A prerequisite for predictable climate models is that spatio-temporal chaos averages out on large scales. I don't think that such a fundamental question should be dismissed as "being over the top". Our current inability to tackle that problem is no excuse for not addressing it.
Normally I do not respond to "anonymous" - and here it may be sufficient that we meet again a person who has not understood either the dimension of the "predictability"-concept in open, multiply non-linear high-dimensional systems, or the consequences of very many simultaneously acting chaotic processes. Instead we see empty talk, employing fancy words, in action.
ReplyDelete@ 4
ReplyDelete"Exactly, and that is the problem: This alone has the potential to destroy any long term prediction ability of climate models."
It may or may not have the potential. Previously, I mentioned an example where chaos does not destroy the predictability: we can always predict that summer temperatures are warmer than winter temperatures So this is a clear counterexample to your suggestion.
Whether or not the climate is predictable at longer time scales may be an open question that should be investigated, I didnt dismiss it. But what we know now is that when the forcing is strong enough, some aspects of climate changes can be predicted.
I think it was you who dismissed climate models in general, I dont think that you shown us convincingly why