Climate models, extremes and the role of statistics.

Abstract:

It is well accepted that some of the greatest impacts of a changing climate will occur through changes in the distribution of extreme events. This talk will survey the range of current statistical methods that are available to estimate the tails of a distribution and to quantify the uncertainty in those estimates. A particularly useful recent development is the combination of extreme value distributions with (Bayesian) hierarchical models that can include the dependence of parameters across spatial locations, trends over time, and the effects of other variables. These techniques are applied to pattern scaling as an example of extending model experiments beyond mean quantities to lesscertain climate statistics such as extremes.