Title
Toward Ice Sheet Model Calibration with Spatial Data
Speaker
Murali Haran, Pennsylvania State University
Abstract
Predicting the future behavior of the Antarctic and Greenland ice sheets is of considerable interest because their melting rates can have an impact on future sea level rise. Making future projections based on a complex computer model of the ice sheet requires inference ("calibration") for key parameters of the model. This inference is performed using computer model output at various parameter settings and observational data regarding the past and present state of the ice sheet. Inference is challenging due to the nature and size of the spatial data and model output as well as complicated dependencies and uncertainties in the model and data. In this talk I will describe a computationally expedient Gaussian process-based inferential approach for computer models with high-dimensional spatial data, and a preliminary application of this method to an ice sheet model. This is joint work with Won Chang (Statistics, University of Chicago), Patrick Applegate, David Pollard (Penn State Earth and Environmental Systems Institute) and Klaus Keller (Geosciences, Penn State).
