Mathematician: Climate Models May Never Predict Accurately

In a great science story looking at the guts of most climate models in use today, James McWilliams, an Applied Math and Earth Sciences professor at UCLA, says that the models do not agree with one another in how the Earth's climate will change, and that they probably never will. The problem with the complex mathematical models is not even an initial values issue, which is already an established stumbling block with most models. McWilliams adds another problem he finds in the models: structural instability. Each model has a different set of approximations and calculations for how they handle physical effects. It's these approximations and calculations that scientists take to mean "theory." Each GCM in use today is a set of theories to be tested, and each one also has a different set of initial conditions, also based on theory, which the modelers plug in. Then they change a variable, the amount of carbon dioxide in the atmosphere to double in 50 years or whatever, and let the models crunch away.

What's great about the Science News article is that McWilliams does not present his findings as fact, but also a theory that needs to be tested. However, his theory is showing a strong evidence that climate models do exhibit structural instability. He goes on to say that the underlying theory that the models are based on, increasing carbon dioxide levels in the atmosphere will produce future warming, is sound science. Whether anthropogenic carbon dioxide is causing the present warming, or even whether the doubling of carbon dioxide levels from 1990 is a useful metric, these are things still being debated. None of this matters to McWilliams though.

McWilliams says the modeling community needs to grapple with this issue now to make sure that models are capable of providing answers to the kinds of questions being asked of them. He compares his argument with Kurt Gödel's proof that some mathematical statements are neither provably true nor provably false. Gödel's theorem, says McWilliams, "is understood as a strong cautionary result about making sure that you're asking the right questions before you exhaust yourself trying to answer them."

I like how mathematicians are lazy, some call it efficient, in tackling a problem: they like to know if they can solve it first.

Via GeekPress, who asks the proper follow-up question: should major public policy decisions be based on these models that cannot be proven to be valid?