TCS Daily

TCS COP 10 Coverage: Model Matters

By Roy Spencer - December 15, 2004 12:00 AM

A recent report by the UK's Hadley Centre, presented this week at the COP10 meeting in Buenos Aires, outlines one view of the future state of the climate system. It is based upon an August 2004 research paper that describes the result of a climate model experiment involving a large ensemble of model runs. A new claim is that uncertainties in climate models have been quantitatively analyzed, leading to improved predictions of some major climate variables such as temperature and rainfall. The report, entitled "Uncertainty, Risk, and Dangerous Climate Change" (as well as the research paper) predicts warming after a doubling of carbon dioxide that is 90% likely to be within a range of 4.3 to 9.7 degrees F, with a best estimate of 6.3 degrees F. Since these estimates are on the high side of the average of a dozen or so climate models that are run in different countries, they will lead to somewhat more concern among those who are prone to believing in climate models. Can climate prediction uncertainty be quantitatively estimated, as claimed in this report? Are we now closer to a global warming prediction that we can have higher confidence in?

First let me say that I greatly admire the expertise at the Hadley Centre. I have friends there, and they were wonderful hosts for a recent meeting I attended. But as is often the case with reports predicting climate change, some critical assumptions have been left unmentioned. Implicit in the report's conclusions is that climate model predictions are dominated by random errors, and not biases. In other words, the different ways in which, say, clouds are treated in climate models are assumed to be correct in an average sense. Our understanding of clouds, water vapor, (or any of a variety of important processes in the climate system) is not good enough to make this assumption, in my view. Science can only deal with what we know, not what we don't know. The real situation might lie entirely outside the range of what we currently assume for some climate model parameter.

This is a problem that has always plagued climate model predictions. It is sometimes noted in climate model intercomparison efforts that there is a certain amount of peer pressure for individual models to be adjusted so that they will more closely match the average of all the other models. Outliers are assumed to have the most serious problems. This, I believe, is more of a human issue than a scientific one. While scientists will sometimes remind themselves that some of the greatest discoveries in science were once considered to be on the fringe, they then return once again to trying to satisfy their peers.

In a sense, the Hadley Centre report gives us the resulting range of "opinions" on global warming that result from including a range of possibilities in a variety of climate model simulations. This is certainly a useful exercise. But the implication still remains that this range is centered on the most realistic value, an assumption that involves more than a little faith.

The fact is that what we don't understand about the climate system can not be quantitatively analyzed -- precisely because we don't understand it! Thus, science is stuck with analyzing what we do understand. And we often forget the unstated assumptions that accompany our scientific conclusions. Given an assumed range for various physical processes in different climate model runs, and assuming that those physical processes are realistically handled by the models, and that there aren't important processes that have not yet been discovered and are being currently neglected, etc.,...then future climate change would probably be expected to lie within the predicted range. But believing in all those assumptions requires a level of faith I can not muster.

The report also mentions some potentially very serious climate change events, such as the collapse of the ocean's thermohaline circulation (as in "The Day After Tomorrow"), a complete melting of the Greenland ice sheet (maybe that will be in "The Day After That"?), and more intense heat waves like that experienced in the summer of 2003 in Europe. Fortunately, it doesn't specifically predict these event to occur. Only that, if they do, the climate impacts on humanity could (surprise) be quite large.

Also stated in the report is the view that climate models are our only way to predict future climate change. I agree. But that doesn't mean I must believe their predictions. Climate modelers have a daunting task. It is important, even necessary, work. I simply do not believe we are yet to the point where the critical components of the climate system are understood well enough to make quantitative predictions.


TCS Daily Archives