by Dana Nuccitelli, via Skeptical Science
In the interest of full disclosure, many Skeptical Science team members are big fans of Nate Silver’s FiveThirtyEight blog at The New York Times. Silver runs a model which uses polling results and various other input factors (such as economic indicators) to predict election outcomes in the USA, with an impressive track record of accuracy.
Thus we were intrigued to hear that Silver had included a chapter on climate change in his newly-published book The Signal and the Noise: Why So Many Predictions Fail-but Some Don’t, particularly since we at Skeptical Science are often forced to explain the difference between signal and noise. Having great respect for the work and climate-related opinions of Michael Mann (who Silver consulted in writing the book), we were also concerned to see his criticisms of Nate Silver’s climate chapter.
Nevertheless, Mann recommended that people read the book for themselves, praising much of the content. So I did just that, and overall I believe that if we take Silver’s analysis a step further, we can learn a lot about the accuracy of climate models. It’s also important to remember that, as Silver himself notes in the chapter, our basic understanding of how the climate works and how much it will warm in response to our greenhouse gas emissions is not just dependent on models.
Correlation is not Causation without Physical Connection
Silver’s climate chapter starts out very well, noting that correlation does not necessarily imply causation, and that determining climate change causation requires a physical understanding of the climate system.
“…predictions are potentially much stronger when backed up by a sound understanding of the root causes behind a phenomenon. We do have a good understanding of the cause of global warming: it is the greenhouse effect.”
Failing to consider physics in trying to determine the cause of global warming has been the pitfall for many a climate contrarian, for example Roy Spencer, Craig Loehle, Nicola Scafetta, Syun-Ichi Akasofu, and many others, so Silver’s point is an important and relevant one. It is easy to fall into the curve fitting trap.
Silver goes on to explain some of that fundamental physics as discussed in the IPCC report – that atmospheric CO2 has increased steadily and rapidly, that this CO2 increase will in turn increase the greenhouse effect and cause global surface warming (which we’ve known for well over a century), and that water vapor will amplify that global warming as a feedback effect, ultimately noting “The greenhouse effect isn’t rocket science.”
Healthy Skepticism or Noise?
After this good start, the chapter then proceeds to discuss what Silver considers the healthy form of scientific skepticism, noting that
“In climate science, this healthy skepticism is generally directed at the reliability of computer models used to forecast the climate’s course.”
Silver then discusses J. Scott Armstrong as an example of this type of healthy skeptic of science who is concerned about the accuracy of climate model predictions. Armstrong is basically used to establish the ‘skeptic’ criticisms of climate models, though his arguments are very weak, basically boiling down to ‘climate models are too complex to be accurate.’ Armstrong also tends to focus on short-term noise rather than long-term trends, which Silver does eventually point out toward the end of the chapter. After establishing Armstrong’s criticisms, Silver moves on to the more interesting part of the chapter, evaluating the accuracy of past climate models.
Testing Hansen’s 1988 Model Accuracy
Silver attempts to evaluate the accuracy of climate models by examining the model projections made by James Hansen in 1988 and the IPCC in 1990 and 1995. We should note here that Skeptical Science has evaluated many other temperature projections going back as far as Wallace Broeker’s 1975 paper in the Lessons from Past Predictions series, with the results summarized in Figure 1 (though not all of these are based on climate models). Note that most of the accurate predictions have come from mainstream climate scientists and models, while the least accurate predictions have come from various ‘skeptics’.