Use this URL to cite or link to this record in EThOS:
Title: A Bayesian approach to the interpretation of climate model ensembles
Author: Demetriou, M.
ISNI:       0000 0004 7970 5977
Awarding Body: UCL (University College London)
Current Institution: University College London (University of London)
Date of Award: 2016
Availability of Full Text:
Access from EThOS:
Full text unavailable from EThOS. Please try the link below.
Access from Institution:
This thesis is concerned with uncertainty quantification when interpreting ensembles of climate models (MMEs), to learn about the true climate. The work improves a recent Bayesian framework for MME interpretation and applies it for first time on real data. The original framework accounts for shared simulator discrepancy from reality. Inference for the true climate is provided through a posterior distribution. The posterior is obtained based on a computationally efficient implementation which assigns estimates to the covariance matrices expressing variability due to different sources of uncertainty in the MME structure. Three framework improvements are considered. Firstly, an improved estimator of the covariance matrix expressing vari- ability due to shared simulator discrepancy from reality is proposed. It relates future to historical discrepancy using bootstrapping from earlier data, instead of subjectively setting a parameter to relate them. The second improvement incorporates prior in- formation about the framework's covariance matrices to account for uncertainty in their values, leading to a fully Bayesian implementation. Two such implementations are introduced, to explore sensitivity of the true-climate posterior to different priors. The third improvement proposes an extended framework which accounts for simu- lator grouping. A random effects model is also proposed to estimate within-group variability when singleton groups exist. The improvements are assessed through an application on global surface air temperature, using observations and simulator out- puts. Results suggest that ignoring simulator grouping and uncertainty in the values of the framework's covariances does not seriously underestimate the uncertainty in inference for real climate. A simulation study is also presented, to compare the performance of the extended framework relative to the simpler. Results suggest that given small shared discrepancy and a well-defined grouping structure, accounting for grouping improves uncertainty quantification in inference for true climate, provided the bias in estimation of shared discrepancy is small.
Supervisor: Chandler, R. Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available