Understanding the Statistics of Monthly Climate Extremes

There is great scientific and public interest to understand and anticipate extreme climate events. These may include, for example, heat waves, like the one that occurred in Russia in July 2010, or long wet periods such as occurred in the Ohio River Valley in Spring 2011. These kinds of persistent extreme weather events are often well represented by the statistics of monthly means. As a first step to understand the statistics of extremes, we investigate the changes in moments of the probability distribution function (PDF) of monthly surface temperature and precipitation over the 20th century in the observed climate, a simulated 20th century climate, and a set of experiments with different combinations of sea surface temperatures and radiative forcing. Preliminary results indicate that large model errors make it difficult to accurately describe the moments of the distribution, much less how the tails of the distribution (i.e. extremes) change over the 20th century. Bias correction is used to determine if the model's sensitivity to changes in radiative forcing are realistic, and may still provide useful information on changes in extremes, despite the model biases. We also identify regions where changes in PDF moments appear to be significant over the 20th century and attempt to understand the physical and dynamical mechanisms for these changes. Future projections are also explored to determine whether similar mechanisms may play a role in future changes in climate extremes.