Science-based risk assessments for rare events in a changing climate

Abstract:

History shows that substantial investments in protection against any specific type of natural disaster usually occur only after (usually shortly after) that specific type of disaster has happened in a given place. This is true even when it was well known before the event that there was a significant risk that it could occur. Presumably what psychologists Kahneman and Tversky have called “availability bias” is responsible, at least in part, for these failures to act on known but out-of-sample risks. While understandable, this human tendency prepares us poorly for events which are very rare (on the time scales of human lives) and even more poorly for a changing climate, as historical records become a poorer guide.

A more forward-thinking and rational approach would require scientific risk assessments that can place meaningful probabilities on events that are rare enough to be absent from the historical record, and that can account for the influences of both anthropogenic climate change and low-frequency natural climate variability. The set of tools available for doing such risk assessments is still quite limited, particularly for some of the most extreme events such as tropical cyclones and tornadoes. We will briefly assess the state of the art for these events in particular, and describe some of our ongoing research to develop new tools for quantitative risk assessment using hybrids of statistical methods and physical understanding of the hazards.