Bias, ignorance and reality in climate science
The rains have returned to California, and the six-year drought appears to be largely over. We have heard countless assertions from journalists and politicians, ignorant of the weather history of California and the other western states, that the drought was a result of global warming.
In the January edition of Scientific American, there is a well-told story “California Megaflood: Lessons from a Forgotten Catastrophe” by B. Lynn Ingram, a professor of earth and planetary science at the University of California, Berkley. She notes: “Geologic evidence shows that truly massive floods, caused by rainfall alone, have occurred in California every 100 to 200 years. The only megaflood to strike the American West in recent history occurred during the winter of 1861-62. California bore the brunt of the damage. This disaster turned enormous regions of the state into inland seas for months, and took thousands of human lives. The costs were devastating: One quarter of California’s economy was destroyed, forcing the state into bankruptcy.” The floods followed “two exceptionally dry decades.”
People are endlessly surprised by some unusual weather, geological, political or economic event, often with the erroneous assumption that such a thing has never happened before. This lack of historical knowledge is not confined to the poorly educated, but often experts in some field or another do not know the history of their own discipline. With the advent of low-cost, powerful computers, mathematical model-building has become all the rage. I am all for model-building, provided the models are tempered with historical reality. A way of testing the predictive ability of a particular model is to compare its predictions against the observed data.
For instance, there had been a pause in global warming for nearly two decades, despite the rise in carbon-dioxide emissions, which none of the major climate models had predicted. Climate scientists Patrick Michaels and Chip Knappenberger of the Cato Institute compared observed warming rates from 1950 to predictions made by 108 models. In virtually all cases, at a statistically significant level, the models projected much higher rates of warming than actually occurred. The fact that models all erred in one direction indicates that they misspecified one or more major variables or they were subject to bias.