Earth's Fidgeting Climate

From a statistical point of view, no single scientific result based on real-life data ever deserves absolute confidence. There always lingers the possibility -- however small -- that the apparent results are due to chance patterns in the data, i.e., "noise."

In the case of Crowley's study, statistical tests show that the probability of his results being due to chance is less than 1 percent. Usually, anything less than 5 percent is considered credible.

"This is not mathematics where you can prove something and write Q.E.D. at the bottom of the page," Crowley said. "This is geoscience. It's a dirtier field, and usually you make statistical arguments."

The abbreviation stands for the Latin phrase quod erat demonstrandum, which means, "which was to be demonstrated." Mathematicians use "Q.E.D." to indicate the end of the written proof of a theorem, which, if correct, is considered absolute.

In addition to the caveats inherent to statistics, conclusions from studies like Crowley's that are based on computer simulations of the world's climate are plagued by questions of how well computer models portray the real thing.

To prove causation, scientists must perform experiments under controlled conditions on the system being studied, manipulating the system to understand what causes what. Other scientists repeat the experiments to show that the explanation is reliable.

Since the Earth's climate is beyond the reach of such experimentation, scientists instead run computer simulations of global climate. These models are always much simpler than the Earth's climate itself. In fact, it's theoretically impossible to create a "perfect" model of climate that includes all the detail of the real system.

"The climate system is too complex," Mosley-Thompson said. "Even the most complex climate model doesn't get it right. And why is that? Because who writes the climate models? Humans. What is a climate model? It's a set of equations that describes what we think we know. If you're not cognizant of a particular phenomenon, then how can you incorporate it into a climate model?"

The fact that different computer models often produce different forecasts doesn't offer much reassurance. For example, one model predicted that the Southeastern U.S. would become more jungle-like in the next century, while another model predicted the same region would become a dried-out savannah, according to Dr. John Christy, a professor of atmospheric science at the University of Alabama in Huntsville.

However, scientists can establish some degree of confidence in their computer models by seeing if the model can accurately "predict" past climate patterns that are known to science.

"Models in isolation may not be believable, but when ... a model can simulate a number of different observed climate responses, the results have more weight than mere calculation," Crowley said. "That still doesn't prove the point, but it minimizes the value of the argument, 'It's only a model.'"

Putting the pieces together

Ultimately, the verdict from science about the extent and cause of global climate change may not come from one particular study or observation.