of weather-related events; examples of such decisions include the ratings of catastrophe bonds or weather derivatives; calculations that statisticians use to adjust data for seasonal effects; and public and private decisions made in agriculture, insurance, reinsurance, health (e.g., disease), fire control, and ecosystem management. Many of the decisions are based on statistical calculations that are appropriate for stationary climates. One well-known example is the widespread use of “30-year normals” in deriving climate data for individual locations.

On the whole, these assumptions are reasonable, if imperfect, rules of thumb to use when the variability of weather is small and climate is stationary. If some parameter related to climate, such as runoff, follows normal distributions with known and constant means and standard deviations, businesses and governments can reasonably use current practices. However, in light of recent findings related to nonstationary and often highly skewed climate-related variables, current practices can be misleading and result in costly errors. An event that is estimated to be a “500-year flood” or a “500-year fire” can be found to occur every 100 or 50 or 20 years if the averages and the variabilities are significantly misestimated.

The potential for abrupt climate change and the existence of thresholds for its effects require revisions of our statistical estimates and practices. We have not attempted to compile a list of all the places where society estimates or uses the assumption of a stationary and stable climate; such an effort would be enormous, and lies outside the committee charge. Rather, we believe that the time is ripe to examine this issue, develop improved methods, and develop improved estimates of the likelihood of extreme events.