Spatial Hierarchical Modeling in Comparing Extreme Precipitation Generated by Regional Climate Models
Erin M. Schliep, M.S. Candidate, Department of Statistics, Colorado State University.
Tuesday, June 9, 2009
10:00 am, 223 Weber
Using a spatial Bayesian hierarchical model, we analyze precipitation output from six regional climate models (RCMs). The primary advantage of this approach is that the model is designed to borrow strength across location by means of a spatial model on the parameters of the generalized extreme value distribution. Being that the data we analyze have a relatively short time span for characterizing extreme behavior but have great spatial coverage, this is particularly important. The hierarchical model we employ is computationally efficient as we have data from nearly 12000 locations. The objective of our analysis is to compare the extreme precipitation generated by these RCMs.
Although the RCMs produce similar spatial patterns for the 100-year return level, our results show that their characterizations of extreme precipitation are quite different. We also found differences in the spatial patterns for the point estimates of the extreme value index. These differences, however, may not be significant due to the uncertainty associated with estimating this parameter
Dr. Dan Cooley, Advisor
Dr. Jennifer Hoeting, Committee Member
Dr. Scott Denning, Atmospheric Science, Outside Member