Abstract:
|
We extend the method of pattern scaling to analyzing extremes, and apply our approach to investigate future extreme precipitation under two climate change scenarios over the contiguous United States. We fit a non-stationary generalized extreme value model to annual maximum daily precipitation from a 30-member initial condition ensemble of transient CESM runs under the RCP8.5 scenario. Our pattern scaling method allows us to create a predictive distribution of annual maxima for any global mean temperature of interest, specifically for global mean temperatures under RCP4.5. We find these predictive distributions to be well calibrated with annual maxima actually produced by RCP4.5 runs. We compare impacts under RCP8.5 and RCP4.5 in terms of the 1% annual exceedance probability (AEP) level. Under RCP8.5, the 1% AEP level estimate increases 23.1% on average between 2005 and 2100, and up to 47.2% at some locations. Compared to RCP8.5 in the year 2080, RCP4.5 reduces the 1% AEP level by 6.9% on average, with reductions as large as 12.6%. We also investigate within-model variability by comparing parameter estimates from a single ensemble member to those obtained using the full ensemble.
|
ASA Meetings Department
732 North Washington Street, Alexandria, VA 22314
(703) 684-1221 • meetings@amstat.org
Copyright © American Statistical Association.