In this applied talk, we present a methodology for assessing climate models' ability to simulate extreme events. Assessing model output is not straightforward, due to the different spatial and temporal scales between climate model output and observations; and this difference is exacerbated when assessing extreme behavior. There has been some work on downscaling the extremes of model output. We ask a more fundamental question of does downscaling make sense by examining the tail dependence between model output and observations. We employ bivariate extremes methods to examine the strength of daily correspondence of extreme precipitation events between observations and the output of both regional climate models (RCMs) and the driving reanalysis product. We analyze winter precipitation on the Pacific Coast and summer precipitation in the US Midwest.
Further focusing on the Pacific Coast and one particular RCM, we construct a bivariate model of the dependence between precipitation from the RCM output and observations. In an attempt to better understand and quantify the processes which lead to Pineapple Express events, we develop a daily "PE index" based on mean sea-level pressure fields. We show this index to be tail dependent to the observed precipitation data. We investigate the possible behavior of Pineapple Express events as produced by climate models' future projections.
|
ASA Meetings Department
732 North Washington Street, Alexandria, VA 22314
(703) 684-1221 • meetings@amstat.org
Copyright © American Statistical Association.