Extreme Daily Temperature and Precipitation in a Weather@home Superensemble for the Western United States: Model Performance and Projections

Thursday, 17 December 2015
Poster Hall (Moscone South)
Sihan Li1, David E Rupp1,2, Philip Mote1,2, Neil Massey3 and Myles Robert Allen3, (1)Oregon State University, Corvallis, OR, United States, (2)Oregon Climate Change Research Institute, Corvallis, OR, United States, (3)University of Oxford, ECI/School of Geography and the Environment, Oxford, United Kingdom
Making credible projections of future changes in extreme events has been challenging because it requires not only running climate models at high resolution to faithfully reproduce impact-relevant extreme events, but also ensemble sizes on the order of 10³ and greater to obtain reliable statistics on the intensity and frequency of extreme events. Due to sparsity of high-resolution data, most studies have used fitted analytical probability distributions to produce statistics for extreme events, which in itself has limitations and uncertainties. Here we present results of a superensemble of simulations generated by weather@home, a citizen science computing platform, where Western United States climate was simulated for the recent past (1985-2014) and future (2030-2059) using a coupled regional/global model (HadRM3P/HadAM3P) at 25-km resolution. The very large number of simulations permits the detection of robust spatial patterns of anthropogenically forced change, amidst the “noise” of natural variability, in extremes in daily temperature and precipitation. We investigate to what extent extreme events change in frequency and intensity, relative to changes in the means. Also, the physical mechanisms underlying such changes are explored. We also compare projected daily extreme temperature and precipitation from weather@home with those from regional/global coupled model parings from the North American Regional Climate Change Assessment Program (NARCCAP), whereby statistics (e.g. 20-year, 50-year, etc., return values) are estimated from fitted extreme value distribution.