Changes in U.S. Temperature Extremes under Increased CO2 in Millennial-scale Climate Simulations

Abstract:

Changes in extreme weather may produce some of the largest societal impacts from anthropogenic climate change: present-day weather damages are dominated by rare events that happen only every several decades or more. However, it is intrinsically difficult to estimate changes in those rare events from the short observational record. We therefore look for insight to climate models, where we can generate long simulations. In this work we use millennial runs from the Community Climate System Model version 3 (CCSM3) in equilibrated pre-industrial and future (700 and 1400 ppm CO2) conditions. We examine both how extremes change using 1000-year timeseries, and how well these changes can be estimated based on shorter pieces of these runs. We estimate changes to distributions of future annual temperature extremes (wintertime minima and summertime maxima) in the contiguous United States by fitting generalized extreme value (GEV) distributions using the block maxima approach. Our results show that the magnitude of summer warm extremes largely shifts in accordance with mean shifts in summertime temperatures, and their distribution does not otherwise change significantly. In contrast, winter cold extremes warm more than mean shifts in wintertime temperatures, with changes in spread and skewness at inland locations that lead to substantial changes in tail behavior. We then examine uncertainties that result from using shorter model runs. In principle, GEV modeling allows us to predict infrequent events using timeseries shorter than the recurrence frequency of those events. To investigate how well this approach works in practice, we estimate 20-, 50-, and 100-year extreme events, first in the full 1000-year model timeseries and then using segments of 20 and 50 years. We find that even with GEV modeling, timeseries that are of comparable length or shorter than the return period of interest can lead to very poor estimates. These results suggest caution when attempting to use short observational timeseries/model runs to infer infrequent extremes. For evaluating changes in extremes, future model intercomparison exercises may need to include longer runs or more extensive ensembles.