Like a Model on the Cover of a Magazine

by Michael Tobis     Department: Earth

Climate science depends on climate models. There is nothing odd about that. All science depends on models.

Many words, notably "theory" and "energy" and even "global warming", have very precise meanings to scientists, and relatively vague meanings to nonscientists. People who intend to muddy the waters often take advantage of this, to extract statement from scientists that sound as if they mean or imply something different than is actually the case.

As the title to this essay will remind you, the word "model" also has multiple meanings. The situation is complicated by the fact that this is a word that the climate sciences are quite informal with.  This doesn't mean we don't know what we are talking about, or what the formal meaning of the word "model" is.

We use the word informally because it is a very central concept in what we do. The formal name for what climate people often call a "model", what the press usually calls a "climate model" is a "coupled general circulation model"
or a CGCM, which is a very specific kind of a beast. You'll admit the name is quite a mouthful, too. Even the acronym "CGCM" doesn't exactly roll off the tongue. Many of us spend much of our time wrestling with these beasts, and so naturally we call them "models", and indeed they are central to much of the enterprise of climate science.

There are lots of meanings of the word "model", though. Understanding the role of the "climate model" starts with understanding how models are crucial to science.

Consider the somewhat contentious statement by the great physicist Ernest Rutherford: "In science there is only physics; all the rest is stamp collecting." What did he mean by this?

What distinguishes the science that is like physics from the science that is like stamp collecting?

By stamp collecting (a pursuit that has little representation among scientists today) Rutherford meant the collection of facts. Biological field work that identifies species and populations as a goal in itself was very common in Rutherford's day (a century ago), and it certainly has some value. To Rutherford, though,
it was unimpressive. and in fact, little research biology is conducted in that way nowadays.

The alternative approach to science is mathematical.

To do physics is to identify and test some mathematical structure that describes reality. Indeed, the usual test of such a structure is that it describes reality so well that the outcome of some experiment or measurement
can be predicted in advance of the actual event. That mathematical structure is called a "model" or a hypothesis. Once the model shows some utility it becomes part of the "theory" of a field. Scientists never formally claim perfect certainty about anything, but sometimes these theories are enormously powerful.

Consider being a test pilot in a modern airplane. When you get into that cockpit, you are investing a great deal of confidence  into the models and theories of aerodynamics. (It is in fact reasonable to put far more confidence into such theories than into political or economic theories, because these theories are far more testable and far
better tested.)

In the past, before a pilot flew an airplane, tests were done on scale model mockups in "wind tunnels". There are mathematical laws, which once were models and now are establlished theory, that relate the behavior of the scale model to the behavior of the full-sized vehicle. Nowadays, wind tunnel tests are rare. Aerodynamic engineers prefer to use big, complicated computer models. The experience base for these models is sufficient that test pilots habitually rely on the results.

Weather models are the same way, and indeed (since aerodynamics and meteorology share some of the same physical principles) are in some ways similar to computer models of airplanes.

So there is nothing fundamentally new about the idea of building models.  

In the past, our models easily exceeded our ability to do the calculations that the models implied. Fluid dynamics (which includes aerodynamics and much of meteorology) was especially out of reach. The predictions that could be made were limited by the calculations that could be done. Many corners had to be cut, and as a  
result science often had to settle for qualitative rather than quantitative predictive abilities even if the mathematics of the system was well-established and precise.

Weather models have made great strides. When I was a boy, five day weather predictions were unheard of. Now they are common and mostly reliable.

Here's a picture of hurricane Katrina that was made four days before it strengthened to a category 4, while it was approaching the tip of Florida as a relatively modest  hurricane. This prediction turned out to be almost exactly correct. Regardless of whether the warning implicit in this picture was heeded well enough, the science
behind this kind of a result is an amazing achievement.




We say this is the output of a "computer model" of the weather.  This way of speaking is confusing. We in the field sloppily confuse the computer code with the "model" sometimes. The computer code embodies the model. It isn't actually the model: it is the representation of the model on a calculating machine.

So there's nothing inherently suspicious about a model: models are a central component of science. Computers are new tools that help us study the implications of our model.

There's a ways to go yet in this essay series before we get a deeper understanding of what climate models actually do, and how much or how little we should rely on them. For now, though, I just want to establish that the idea of building a computer model is perfectly natural for studying a complex system.

Tags: climate, models, weather

TrackBack URL for this entry: http://www.pbs.org/mt4/mt-tb.cgi/108