Trends in US Precipitation Extremes

• Activity Number: 630
• Type: Invited
• Date/Time: Thursday, August 2, 2012 : 10:30 AM to 12:20 PM
• Sponsor: ASA Advisory Committee on Climate Change Policy
• Author(s): Michael Wehner*+
• Companies: Lawrence Berkeley National Laboratory
• Keywords: Extreme value theory ; precipitation ; trends

Abstract

We examine observed and modeled trends in extreme precipitation over the continental United States by means of non-stationary extreme value methods applied to individual station data. A peaks over high threshold technique with time-varying coefficients in the associated generalized Pareto distribution is shown to fit the data well. We find widespread trends of increasing severity of long period return values in most of the daily US station data. Spatial coherence of these trends is also found, especially in the Northeast and Central states. Local likelihood techniques allow us to borrow strength from neighboring stations, and we use the bootstrap to estimate uncertainty. Finally, in conjunction with analysis of climate model output we examine the long-tailed behavior of precipitation extremes that is commonly seen in empirical analyses and explain some of this behavior based on seasonality and records with very low precipitation.