Climate change scenarios and risks

I enjoyed reading Heather Graven’s article “The carbon cycle in a changing climate” (Physics Today, November 2016, page 48 ) and the other articles that month on modeling different aspects of climate change. After reading those and other articles on the subject, I have an uneasy feeling that people are missing the point when it comes to dealing with climate change.

The optimal strategy almost certainly depends far more on the very small probability of extreme changes in climate than on the most likely changes. People seem to focus only on modeling the most likely changes in climate, and maybe the one-standard-deviation error bars, which are likely to be almost irrelevant in choosing the best strategy. The general consensus of climate models, such as those discussed by Graven, is for average global temperatures to rise by a few degrees Celsius over the next several decades, which will have serious effects on agriculture, water resources, coastal flooding, species diversity, and human migration. People argue over justification of spending today to avoid the costs of those effects over the next several decades.

However, if one assumes that climate changes are 100% certain to resemble the shorter-term model consensus, then the best strategy for minimizing net costs might be to spend very little now and instead wait 20 or 30 years. By that time some future technology, perhaps involving intelligent robots, may be able to accomplish anything we want, virtually for free, to reverse whatever climate change has meanwhile occurred, even if much more drastic measures are needed then than would be needed now. Freeman Dyson makes a similar point about discounting future costs because of advancing technology. ¹

The problem with the wait-and-see approach is that we cannot be 100% certain that something like these consensus climate models is correct. In particular, we cannot exclude the possibility that a runaway greenhouse effect will become unstoppable in less than 20 or 30 years and eventually leave Earth uninhabitable. After all, it did happen on Venus, which started out similar to Earth but with about twice the solar forcing. If that is a real—even if unlikely—possibility, then the optimal strategy might be completely different: Spend a lot of money now to try to prevent the runaway greenhouse effect.

Since our optimal strategy depends very much on unlikely, but still possible extreme scenarios, it seems to me we should be devoting a significant portion of our research budget to modeling how those scenarios might occur, rather than modeling only what is likely to occur.