Environmental Engineering Reference
In-Depth Information
effects . . . they agree that, if we continue to go down the same path that we are going down today, the
world as we know it will change—and it will change dramatically for the worse.
24
We'll get back to the 97 percent number in a minute, but if you press any climate scientist for an
ex-
planation,
he will explain (or admit) to you that there is nothing resembling absolute certainty about these
large positive feedback loops and the predictions associated with them. This is called the problem of de-
termining
climate sensitivity;
how much warming, in practice, in the full complexity of the atmosphere,
does
x
amount of CO
2
cause? How strong a
driver
of climate is CO
2
?
Those who speculate that CO
2
is a major driver of climate have, to their credit, made predictions based
on computer models that reflect their view of how the climate works. But fatally, those models have failed
to make accurate predictions—not just a little, but completely.
While everyone acknowledges that the climate is too complex to predict perfectly, the idea behind cata-
strophic climate change is that CO
2
is an overwhelming
driver
of the global climate system and thus that
its warming impact is predictable over time—in the same way that knowing the climate factors where I
live, in Southern California, allows you to predict that it will be mostly dry, even though you can't predict
exactly when it will rain.
Climate scientists who believe CO
2
is such a powerful driver feel confident in making
models
—simpli-
fications—of the global climate system that predict its future based on CO
2
emissions.
Just about every prediction or prescription you hear about the issue of climate change is based on mod-
els. If a politician talks about “the social cost of carbon,” that's based on model predictions. If an econom-
ist talks about “pricing fossil fuels' negative externalities,” that's based on model predictions. If we hear
dire forecasts of drought going forward, that's based on model predictions. Which means if the models
fall, they are invalid. Therefore we need to ask the experts advising us an obvious and essential question:
How good are the models at predicting warming or the changes in climate that are supposed to follow from
warming?
One pitfall in asking this question is that we have to make sure we have evidence of models
predicting
climate in advance
. Why do I say “in advance”? Because part of climate models involve “hindcasting” or
“postdicting”—that is, coming up with a computer program that “predicts,” after the fact, what happened.
There are reasons to do this—namely, it's important to see if your model could have accounted for the
past.
But a model is not valid until it makes real, forward predictions.
It's a truism in any field of math that
if you are allowed enough complexity, you can engage in “curve fitting” for any pattern of data with an
elaborate equation or program that will “postdict” exactly what happened in the past—but in no way does
that mean it will predict the future. (Many investors lose money doing this sort of thing.)
The best way to test a model is to see whether it can make accurate and meaningful predictions about
the future. In the last thirty years, the climate science community has had the opportunity to do that. Many
experts in modeling and in statistics thought this was an extremely dubious enterprise, given how complex
the climate is—at least as complex as the economic system, where failed computer models helped promote
policies that led to our recent Great Recession.
Consider perhaps the most famous model in the history of climate science, the 1988 model by James
Hansen, who has a reputation in the media as the world's leading climate scientist. At twenty-four years
old, the model has been given ample time to show its predictive accuracy. In the graph below, we can see
how Hansen's prediction compares with the actual temperature measurements Hansen subsequently repor-
ted; he dramatically overpredicted warming.
