Short-tailed Temperature Distributions over North America and Implications for Future Changes in Extremes

Tuesday, 16 December 2014: 9:00 AM
Paul Loikith, Jet Propulsion Laboratory, California Institute of Technology, Pasadena, CA, United States and J David Neelin, University of California Los Angeles, Los Angeles, CA, United States
Probability distributions of daily surface temperatures can exhibit marked departures from Gaussianity in the tails of the distribution. Locations exhibiting shorter-than-Gaussian tails are prone to experiencing a relatively greater change in the frequency of extreme temperature exceedances, under a simple shift in the distribution towards warmer temperatures, than are locations with Gaussian or long tails. Under such conditions, places with a short warm tail would see a larger increase in extreme warm events and places with a short cold tail would see a larger decrease in extreme cold events than places with longer tails. The potential impact of this effect is quantified using gridded observations over North America by uniformly shifting the distribution at each grid cell by one standard deviation and assessing the change in frequency of extreme temperature threshold exceedances relative to the current climate. In winter, portions of the Pacific Coast of Canada and southern Alaska, where warm tails are notably short, temperatures that are exceeded 5% of the time in the current climate occur up to 60% of the time under such a shift. In summer, much of Mexico and the central and eastern United States exhibit short warm tails and also experience large increases in the frequency of warm exceedances. Temperatures that are below the 5th percentile of the distribution in the current climate no longer occur over central and eastern Canada and northern Alaska in the winter and throughout much of the Pacific Northwest in summer, due to short cold tails there. While actual changes in temperature may be less uniform across the distribution, these results demonstrate how important it is for climate models to be able to reproduce observed distribution tails and the key physical and dynamical processes that govern distribution shape.