The Atmosphere and Climate Modeling - Introduction to Carbon Capture and Sequestration - page 92

While philosophers would thus argue that it is preposterous to claim
that our models simply tell the Truth, they would not argue that we should
therefore dismiss science as our most reliable source of information
about the world, the next best thing to Truth that we have. The main
argument for this is known as the “no miracles argument.” It points to the
extraordinary success of science: our best theories are pretty good at
predicting and explaining stuff. That success would be a miracle if those
theories were completely wrong. Indeed, climate models are based on
the same theories that we use to build airplanes. If we trust these theories
enough to get in a plane and fl y over the Atlantic, we might give some
credence to their predictions on climate, too. Against that background,
we might say that, if we are going to base crucial policy decisions on the
predictions made by our best theories, our problem is not whether our
models are absolutely True in some abstract philosophical sense: our
problem is to be clear why the methods we use to accept claims are
good enough to warrant the uses to which we put them [2.17].
Now the scientifi c sense of “ We're not sure ” becomes important. In
the case of climate models, most of the underlying physics and chemistry
is well understood, and we can trust that the algorithms, given the input
parameters, generate the correct solutions.
However, these algorithms do require basic data on the interactions
of molecules in the atmosphere, the fl ow of air over mountains, the effect
of clouds, etc. In practice we will never have experimental data from all
temperatures and conditions that occur to put into climate simulations.
So these models rely on all kinds of clever schemes to estimate these
input parameters. In addition, we cannot use an infi nitely accurate grid
(see Figure 2.5.2 ). The use of fi nite grid points results in a loss of informa-
tion. To strengthen the no-miracles argument above: given all these
assumptions and simplifi cations, it is impressive to see the accuracy with
which climate models could predict the path of superstorm Sandy in
October 2012. Is Sandy enough evidence that our current climate models
give a suffi ciently realistic picture of the climate that we can ask people
to invest billions of dollars in CO 2 emission reductions?
Figure 2.5.12 gives an intriguing result: the uncertainty in our climate
change predictions over the years has not decreased! We have tried to
impress you with the fact that increased computer power has increased
our resolution and complexity of climate models. The reason that this did
not result in less uncertainty in the predictions is that each additional