December 21, 2009
The Perverse Economics of Climate Modeling
The Perverse Economics of Climate Modeling
By Bill Frezza
As the dust continues swirling around emails purloined from the climate research unit at the University of East Anglia, global warming activists keep insisting that computer climate models predicting disaster represent "settled science." How can that be when climate models aren't science at all?
Laws are science. Models are engineering.
Scientists conduct controlled experiments, collect observable data, and construct testable hypotheses. In this case, they compare and discuss the accuracy of various sets of temperature measurements, ice core drillings, or tree ring observations. The peer review process, when properly administered, helps establish a body of accepted facts that both scientists and engineers can work from.
Given the credible allegations that the peer review process may have been corrupted and that some of the data has been cherry picked, it's in everyone's best interest to have a thorough public review of the quality and comprehensiveness of the data collected. This should give all of us a higher degree of confidence about when, where, and how much the earth's climate has actually changed. Such a review can be done without taking a position on potential causes of climate change or proposed solutions. After the data are scrubbed, it might be correct to claim that climate measurement science is "settled."
That's step one.
When testable, repeatable relationships are observed on a set of data that can be expressed in the language of mathematics, scientists often say that they have discovered a new "law." Examples run from the straightforward F=MA and powerful E=MC2 to the numbingly complex Schrödinger equations.
If anyone can articulate a mathematical law that describes the relationships amongst the climate data collected thus far, let them step forward and publish it. Not just the predictions - show us the equation so scientists everywhere can test it to see how well it matches the "settled" climate measurement data.
This has not yet happened.
That doesn't mean we're helpless. It just means that it's time for the engineers to step in. Developing empirical computer models to forecast the behavior of complex, nonlinear, stochastic, and even chaotic systems is what engineers do all the time. Without such models we couldn't build computer chips, airplanes, inventory optimization systems, skyscrapers, cell phones, financial derivatives, and many other things that constitute our man-made world. We do this even if the science is not entirely understood.
Most of these models have to deal with incomplete knowledge and limited testability, and many have to accommodate high degrees of randomness and uncertainty. Models are never "settled" but through trial and error driven by profit and loss they keep getting better. While planes don't fall from the sky as often as they used to and computer chips can be manufactured at very high yield, you might have noticed that many financial derivatives models didn't do such a great job predicting the future. Why is that?
The nuance of modeling is that while scientists ask the objective question "is this true," engineers ask the subjective question "does this solve my problem"?
We know what problem the wizards of Wall Street were trying to solve. "How do I develop derivatives models that maximize my year-end bonus?" That approach delivered fantastic bonuses right up until it didn't, at which point the entire world economy was driven into a ditch.
And climate modelers? Scientists living on the public weal get compensated via a mix of government grants and scientific prestige. Their problem they are trying to solve is "How do I get my papers published in the most prestigious journals so I can maximize the size of my next grant?"
If you've ever done computer modeling you know that there are a thousand ways to make a set of curves fit retrospective data in underconstrained systems. And right now, both the government grant and scientific prestige markets are dishing out significant rewards for models that predict runaway climate disaster. So which curves do you think savvy modelers are going to pick?
Calling these climate models "science" and then having the audacity to call them "settled" is the same kind of Big Lie that allowed Congress to assure us that Freddie Mac and Fannie Mae were completely safe so we had nothing to worry about when they embarked on an orgy of sub-prime lending. Every computer model predicting the future behavior of the mortgage market fit all the historical data - right up until the moment that they didn't.
Before we let suspect computer models developed by a handful of people drive the entire world economy into a ditch, don't you think we should take the covers off and invest a little more time and effort to thoroughly examine how these models work? Hopefully this will include analytical critiques from a wider cast of characters than the self-serving cabal whose mendacity and ham-handed attempts to marginalize dissent were recently exposed. Perhaps an open process will help both "sides" focus on attacking weaknesses in the models themselves rather than attacking each other's tribal affiliations. Only then can we hope to get real value for all the money we taxpayers fork over to support these scientists.
Bill Frezza is a partner at Adams Capital Management, an early-stage venture capital firm. He can be reached at firstname.lastname@example.org. If you would like to subscribe to his weekly column, drop a note to email@example.com.