Model reasoning about climate change?

The NYTimes reprinted an Op-Ed piece that Sarah Palin published there in January of this year. (Here's a link.) The message was that though polar bears are worth protecting, they shouldn't be placed in the endangered species list because they aren't actually endangered. Not knowing the science, I won't pretend to say whether Gov. Palin was right or wrong, though it's interesting that the recommendation to list the polar bear as endangered came from an administration not known for being alarmist about environmental issues. But there was one line in the piece that caught my attention: "The possible listing of a healthy species like the polar bear would be based on uncertain modeling of possible effects. This is simply not justified."

One word in this passage stands out for me: modeling. I finally began to notice (late to the party, I suspect) that this word comes up a lot when people want to call climate research into question. For another example, have a look at this passage from Charles Krauthammer:

Predictions of catastrophe depend on models. Models depend on assumptions about complex planetary systems -- from ocean currents to cloud formation -- that no one fully understands. Which is why the models are inherently flawed and forever changing. The doomsday scenarios posit a cascade of events, each with a certain probability. The multiple improbability of their simultaneous occurrence renders all such predictions entirely speculative.

Scientists who read this must have scratched their heads. Modeling is central to what science does. And virtually all models of anything complex depend on probabilities. And yes: if the assumptions fed into the model are too far off, the model is in trouble. But scientists understand this perfectly well. A simple example may help make things clearer. Start with the case of flipping 100 fair coins. There's no predicting what will happen on any individual flip. However, it's 95% certain that the number of heads will be between 40 and 60. And its 98% certain that the number will be between 30% and 70%. A large-scale pattern emerges from the individually chancy events. But there's more: good models are robust: they don't depend on getting the individual assumptions exactly right. We assumed that the coins being flipped were fair. Suppose the coins aren't all fair: suppose that some are biased toward heads, some toward tails. If we know something about the distribution of these biases, we can still say precise things about the overall tendency, and if the distribution is "normal" (a bell-curve), then the overall prediction will be much the same. The large-scale pattern that emerges doesn't depend on getting each of the details right,

Real modeling, of course, is more complex. The coin-flip example merely illustrates the point that even if we're very uncertain about individual events, a good model may help us see overall patterns that are highly likely to emerge.

Krauthammer, we noticed, tells us that there's too much uncertainty, and that as a result, the models are "entirely speculative." Two points, however.

First, saying this doesn't make it so. My point isn't that Krauthammer is wrong; it's that he's in no position at all to know that he's right, particularly since his description of modeling suggests that he doesn't understand it very well, and last time I checked, he wasn't an expert on climate science.

Second, and more important, whether or not Krauthammer is guilty of treating all modeling as mere speculation, it's not unusual to hear remarks of that sort when someone wants to challenge climate science. But the fact is that modeling is a standard tool of science, without which we wouldn't know where to look for oil, wouldn't know how to figure out the geology of the planets we spend large sums sending probes to explore, wouldn't know how to think about epidemics, or complex biological processes, or any number of other topics that we have to think about if we're going to make policy.

In fact, what we seem to have here is a variation on the "lies, damned lies and statistics" theme. People who don't understand probability view statistical reasoning with suspicion. But to give up probabilistic reasoning would be to give up most of the benefits of the science we all routinely rely on. Say it again: the fact that probabilistic reasoning rests on probabilities does not mean that it can't produce secure conclusions. It simply is not a criticism of climate science that it uses models. If we decide to scrap models in figuring out how to plan for contingencies, we might as well throw up our hands.

Is the conclusion that global warming is a fact? If it were, it would be a huge non sequitur; nothing said here mentions any of the evidence. My own view that there's no serious doubt, but you can disagree with that if you like. The moral here is to be very suspicious of climate change skeptics who mutter darkly about models. Those are people who are bound to stay in the dark not just about the climate, but a good deal more as well.