PROBABILISTIC INVERSION OF EXPERT ASSESSMENTS TO INFORM PROJECTIONS ABOUT ANTARCTIC ICE SHEET RESPONSES

Restricted (Penn State Only)
Author:
Fuller, Robert William
Graduate Program:
Geosciences
Degree:
Master of Science
Document Type:
Master Thesis
Date of Defense:
June 01, 2017
Committee Members:
  • Klaus Keller, Thesis Advisor
  • James F. Kasting, Committee Member
  • Sridhar Anadakrishnan, Committee Member
Keywords:
  • ice sheet
  • sea ice
  • antarctica
  • paleooceanography
  • ice shelf
  • ocean temperature
  • melting
  • cliffs
  • sea-level rise
  • calibration methods
  • overconfidence
  • uncertainty
  • deep uncertainty
  • bayesian
  • probabilistic inversion
  • markov chain
  • monte carlo
Abstract:
Anthropogenic greenhouse gasses are warming the planet and causing sea-level rise. Adaptation decisions in coastal areas depend crucially on sea levels. Projections are deeply uncertain. Much of the uncertainty stems from the contribution of the Antarctic ice sheet (AIS) to future sea levels. Expert assessment samples some of this uncertainty. Bayesian inversion of paleoclimatic and instrumental observations with an AIS model also samples some of this uncertainty. This thesis focuses on combining the intuition from expert assessments, observations, and an AIS model in order to quantify deep uncertainty. The research presented in this thesis illustrates a framework for combining expert assessments with instrumental and paleoclimatic observations in a coupled probabilistic-Bayesian inversion with a simple AIS model. The research demonstrates that probabilistic inversion of an expert assessment can be used to inform the prior probability distributions of model parameters. Adding instrumental and paleoclimatic observations to an expert assessment in a coupled probabilistic-Bayesian inversion sharpens the posterior probability distributions of model parameters. The results show that the different interpretations of the expert assessment are consistent with the instrumental and paleoclimatic observations as well as the simple AIS model. They further show that the projections are sensitive to the interpretation of an expert assessment, thereby demonstrating the deep uncertainty in future AIS contributions to sea levels.