Wednesday, January 15, 2020 - 3:30am to 4:30am
- Broida 1640
Title: Spatio-temporal statistical models for glaciology
The purpose of this talk is to review a variety of Bayesian statistical models and associated methodologies developed in the context of glaciology. Specifically, I will describe a Bayesian hierarchical model (BHM) for glaciers that incorporates a physical model known as the shallow ice approximation (SIA), and will introduce the physical-statistical framework introduced by Berliner  and Wikle et al. , upon which the BHM is based. I will also describe a few methods to improve the run-time of computing with the statistical model, with particular emphasis on log-likelihood evaluation and emulation of the SIA model. Results of applying the statistical model and associated methodology will be presented from both simulation studies and real glaciological data collected at Langjökull, one of Iceland’s main glaciers. Additionally, results are presented for modeling glacial surface velocity data, which can aid in the inference of glaciological parameters such as ice viscosity and basal sliding. Finally, the talk will discuss current research directions that are motivated by the described work.
 L. Mark Berliner. Hierarchical Bayesian time series models. In Kenneth M. Hanson and Richard N. Silver, editors, Maximum Entropy and Bayesian Methods, pages 15–22, Dordrecht, 1996. Springer Netherlands.
 Christopher K. Wikle, L. Mark Berliner, and Noel Cressie. Hierarchical Bayesian space-time models. Environmental and Ecological Statistics, 5(2):117–154, Jun 1998.
January 13, 2020 - 7:59am