Abstract

There is a possibility that the climate could undergo rapid and dramatic changes over the next century or so. It is believed that these abrupt changes have a low probability, but there is little evidence for what this probability could be. In this talk I discuss methods we could use for estimating these probabilities from deterministic climate models. These methods are based on the idea of an emulator. An emulator is a statistical approximation to the full deterministic climate model. The full climate model is expensive to run, so we are limited in the number of runs we can achieve. This means we cannot use Monte Carlo methods, for example. We run the climate model at a designed set of places and use these runs to build an emulator. This emulator can now be used instead of the full model since it takes an insignificant amount of computer time. Using the example of the collapse of the thermohaline circulation in the North Atlantic we illustrate these methods and show how they can be used to estimate the risk of low probability events.