Estimating fracture probability of Antarctic Peninsula ice shelves using Bayesian non-parametric statistical modelling

Closs, Marko (2022). Estimating fracture probability of Antarctic Peninsula ice shelves using Bayesian non-parametric statistical modelling. University of Birmingham. Ph.D.

Preview

Closs2022PhD.pdf
Text - Accepted Version
Available under License All rights reserved.
Download (107MB) | Preview

Abstract

Drainage of the Antarctica Peninsula is mainly through its floating ice shelves, which play a vital role in limiting the flow of grounded ice into the ocean via buttressing. The magnitude of the AP contribution to sea-level rise is dependent on the health of its ice shelves. Fractures are a key part of the calving process; however, observing or predicting fractures has proven difficult.

The annual extent of four AP ice shelves is delineated from 1990-2019. Quantile regression analysis draws inferences between retreat and contributing processes. The analysis suggests a statistically significant association between increasing mean annual 2 m air temperature and ice shelves that have retreated the most. Additionally, a reduction in summer and autumn sea ice concentration correlates with retreat.

Fracture probability is estimated for the Larsen C, Wilkins and George VI Ice Shelves using flow parameters as the predictors in a Bayesian Gaussian process mixed effect model. A hybrid Monte Carlo inference scheme enables Markov Chain Monte Carlo application while remaining computationally inexpensive. The fracture probability estimations matched observed fractures in some key regions. The fracture probability of 100-year transient flow simulations under possible climate scenarios is estimated. Under the higher emissions scenario, fracture probability is significantly higher on floating ice.

Type of Work:

Thesis (Doctorates > Ph.D.)

Award Type:

Doctorates > Ph.D.

Supervisor(s):