Sunday, December 18, 2011


From: David Rind <>
To: Stefan Rahmstorf <>
Subject: Re: [Wg1-ar4-ch06] Comments on Section 6.3
Date: Wed, 20 Jul 2005 12:11:20 -0400

Dear Stefan,

The distinction here is that GCMs attempt to calculate from first
principles the zeroth and first order processes that dominate the
problem they are studying, whereas EMICs parameterize many of those
processes. The fact that EMICs can reproduce GCM results suggest that
their parameterizations have been tuned to do so - but this does not
in any way imply that if one alters the forcing or boundary
conditions outside of a small range, or apply them to completely
different problems, that the two types of models will react
similarly. In fact, there is a history of this - the first "EMICs"
had a very large sensitivity to a 2% solar insolation change; then
they had to be re-tuned to prevent that from happening. EMICs are
used for paleo-problems because of their ability to take large
time-steps, but there is no free lunch - in doing so, they sacrifice
calculating the fundamental physical processes the way the real world
does it. GCMs have storms, they have real water vapor transports,
they have winds calculated from solving the conservation of momentum
equation, etc. etc. There is a quantum difference between the
fundamental approaches - it is not a continuum, in which there are no
real differences, everything is simply a matter of opinion, there is
no such thing as truth - that's the argument that greenhouse
skeptics use to try to make science go away.

Because we can't use GCMs for long-time scale problems, we do the
best we can - we use these heavily parameterized models. If we could
use GCMs for those problems, EMICs could then be tuned to produce the
GCM results on those time-scales as well. But in this case we have no
way to validate the EMIC results - and since the first principles are
not being used, we cannot know whether they represent a physically
consistent solution or not. Therefore all they can do is suggest
interactions among processes, a useful though not definitive addition
to the field.


ps - concerning CLIMBER-2, I asked a number of leading climate
scientists to read the model description paper. Peter Stone was the
only person I asked who thought the model was at all useful for
studying the types of problems we are discussing. And it was not only
GCM scientists. If you want to hear further cogent arguments
concerning its inapplicability, consider contacting Bill Rossow (the
recent winner of a major honor as a leading climate scientist) but
make sure your email program or telephone accepts unexpurgated text.

At 4:22 PM +0200 7/20/05, Stefan Rahmstorf wrote:
>Dear David,
>I take from your response that you consider all models that
>parameterise an important first-order process "conceptual models". I
>can live with that - but then there are only conceptual climate
>models around. Any coupled climate GCM that I know of parameterises
>oceanic convection (and in a very crude way), hence it is a
>conceptual model in your terms, and there is no fundamental
>distinction of category between your model and our model.
>To me the scientific question is not whether an important process is
>parameterised (many are in GCMs) - it is how well this
>parameterisation works, for the task at hand. We have tested the
>feedbacks in great detail (e.g., the cloud, water vapour, lapse rate
>and snow/ice albedo feedbacks for 2xCO2) in our model and they
>perform quantitatively within the range simulated by various GCMs.
>The same is true for many other diagnostics - the model has taken
>part in model intercomparisons with GCMs and always falls within the
>range of different GCMs, in a quantitative way. To repeat that
>point, the quantitative differences between different GCMs are
>larger than the typical difference between our model and a GCM. So I
>see no basis for your claim that this model can only "suggest orders
>of magnitude". That's just plain wrong from all the evidence that I
>have seen (a lot). If you have concrete evidence to the contrary,
>other than just knowing one person who happens to agree with you,
>please come forward with it.
>To reach me directly please use:
>(My former addresses are read by my assistant Brigitta.)
>Stefan Rahmstorf

Wg1-ar4-ch06 mailing list

No comments:

Post a Comment