Climate Extremes and Global Warming: A Statistician's Perspective

Many papers in the climatological literature have documented increases in the frequency of extreme events such as heavy rain storms and hurricanes. Since similar increases in extreme events are also observed in climate models, it is widely assumed that there is a causal connection, in other words, that changes in the overall climate because of increasing greenhouse gases are directly responsible for the observed increases in extreme events. My purpose in this talk is to look at these questions critically, from the point of view of a statistician. The first question is, is there really an increase in extreme events in the observed record? This leads us into questions of how we would determine statistical significance of a trend given both spatial and temporal correlations in the data. The second question is, how should we decide whether the patterns in observational and model-generated data are “similar”? I will illustrate these points by looking at correlations between tropical cyclones (typically the precursor of hurricanes) and sea surface temperatures in the Atlantic Ocean, and also by examining the spatial-temporal pattern of precipitation extremes in observations and models. Although both of these topics are the subject of ongoing research, my overall theme is that it would be premature to jump to conclusions about anthropogenic cause and effect.