HR: 14:10h
AN: GC23B-03    [Abstracts]
TI: Functional ANOVA Modeling of Regional Climate Model Experiments
AU: * Kaufman, C
EM: cgk@ucar.edu
AF: National Center for Atmospheric Research, P.O. Box 3000, Boulder, CO 80307, United States
AU: Sain, S
EM: ssain@ucar.edu
AF: National Center for Atmospheric Research, P.O. Box 3000, Boulder, CO 80307, United States
AB: One of the goals of the North American Regional Climate Change Assessment Program (NARCCAP) is to explore uncertainties in regional climate modeling over North America by nesting a set of regional models within a set of global models. As a diagnostic tool to compare model similarities across spatial regions, we propose a functional ANOVA approach, in which we decompose the climate response into a common mean response, effects due to regional model, effects due to global model, and interactions. We model these functional effects using Gaussian process priors, and use the posterior distribution to compare sources of variability in model output. We illustrate the method on an existing data set from the Prudence Project.
DE: 0520 Data analysis: algorithms and implementation
DE: 1637 Regional climate change
DE: 3252 Spatial analysis (0500)
DE: 3275 Uncertainty quantification (1873)
SC: Global Environmental Change [GC]
MN: 2007 Fall Meeting