Calibration of ice sheet models using neural Bayes estimators
Author
Bao Anh Vu
Co-author
-
Andrew Zammit Mangion
-
David Gunawan
-
Felicity McCormack
-
Noel Cressie
-
-
-
-
Conference
Format: IPS Abstract
Keywords: climate-and-ice-sheet-modelling, ensemble-kalman-filter, ice-sheet-model, missing data, neural-bayes-estimator, neural-network, posterior-approximation, spatio-temporal
Session: Invited Session 1B - Statistics for Securing Antarctica's Environmental Future
Monday 2 December 11 a.m. - 12:30 p.m. (Australia/Adelaide)
Abstract
Ice sheet models are used routinely to quantify and forecast an ice sheet's contribution to sea level rise. In order for an ice sheet model to generate valid forecasts, its parameters must first be calibrated using observational data; this is challenging due to the model's high degree of nonlinearity, its complex parameterisation, and limited data availability. This study leverages the emerging field of neural posterior approximation for efficiently calibrating ice sheet parameters, namely the bed elevation and basal friction, based on velocity and surface elevation data. Samples from the approximate posterior distribution are then used with an ensemble Kalman filter to infer ice thickness and velocity in a 1D Shallow-Shelf Approximation (SSA) model. We apply our approach to infer the bed elevation and basal friction along a transect in Thwaites Glacier, Antarctica. This work is joint research with Andrew Zammit-Mangion (UOW), David Gunawan (UOW), Felicity McCormack (Monash University), and Noel Cressie (UOW).