My thanks to Peter Berényi for this short-sharp and to the point wake up call to climate modellers worldwide. Kevin Trenberth would be wise to take heed, given the Nature article he wrote in 2010. Peter is a mathematician by training, with a firm background in physics. He used to
work in acoustics, where he designed, supported and performed experiments. 12-14 years ago he switched to IT to make a living. His earlier article here at the talkshop on Earth’s energy balance is well worth a read too.
The fundamental issue with computational climate models is an epistemological one. Fitting multiple models, and computational ones of high Kolmogorov complexity at that, to a single run of a unique instance is not science, never was and never will be. The very paradigm of climate modelling, as it is practiced in the last several decades is flawed, and should be abandoned immediately.
The proper approach is to seek a simple theory, that fits many experimental runs of multiple physical instances, but GCMs are as far away from this requirement, as anything ever can get.
Therefore it should be realized, there is no such thing as “climate science” as an autonomous field, independent of the general and as yet unresolved problems of physics.
Non-equilibrium thermodynamics of complex systems (with a vast number of non-linearly coupled degrees of freedom) in the presence of significant radiant heat is one of the few areas of semi-classical physics, where little progress is seen, basically because of difficulties around proper lab models of such systems. That is, we still
do not understand on a general level what’s going on.
But terrestrial climate is just one example of such systems. Why would
anyone at her right mind expect to understand it better than the general
Go back to the lab, guys and gals, and do actual experiments on real physical
objects. Not on a simulacrum of Earth of course, because that’s impossible.
Study other objects, filled with semi-transparent fluids of complex
chemical composition, on a rotating table to induce as much turbulence as
possible. Send a vigorous flow of free energy through it with a high rate
of entropy production, isolate it from its environment in all respects
except radiative coupling. Put it into a vacuum chamber whose walls are
kept at a low temperature, by liquid nitrogen perhaps. Have its effective
temperature as high as permitted by construction materials (at several
hundred centigrade at least). It’s even better if some component of the
fluid has phase transition close to the operating temperature of the device.
As soon as such a system is understood adequately, that is, you can get
into a position when you are able to construct a computational model of it
based on full theoretical insight, that can predict (not project!) its
behavior reliably in multiple experimental runs, even if it is perturbed in
any number of ways, notably by changing the optical depth of fluid filling
it, in various spectral bands, then, and only then, you can return to
That’s the way science is done, not the other way around.
Please note that this requirement is not applicable to collecting adequate
climate data. That’s a must, because later on, even with more insight,
measurements missed in the past would still be missing, forever.
On lack of proper experiments & theory, wiki is your
in this case.
There is much theoretical work going on in the field, of course,
unfortunately with little experimental backup. But it is ignored
by *mainstream* climate science anyway.
For a review see:
Invited contribution to
*Variational and Extremum Principles in Macroscopic Systems*,
H. Farkas and S. Sieniutycz, eds.,
Amsterdam: Elsevier Science, 2004
*The Nonequilibrium Thermodynamics of Radiation
*Christopher Essex, Dallas C. Kennedy, Sidney A. Bludman*
I could imagine some promising directions for theory to go, like SOC
(Self-Organized Criticality), SAD (Sandpile Avalanche Dynamics), MEPP
(Maximum Entropy Production Principle) and the like, but with no
experiments theory can only go so far. We definitely need more constraints
to be able to handle such systems properly.
Why, I don’t even think we can do a detailed computational model of a
boiling pot on a stove yet.