Use this URL to cite or link to this record in EThOS:
Title: Quantifying uncertainty in projections of large scale climate change
Author: Rowlands, Daniel James
ISNI:       0000 0004 2721 4017
Awarding Body: Oxford University
Current Institution: University of Oxford
Date of Award: 2011
Availability of Full Text:
Full text unavailable from EThOS.
Please contact the current institution’s library for further details.
A systematic approach to quantifying uncertainty in climate projections is through the application of observational constraints to an ensemble of climate model simulations. In this thesis we investigate how large perturbed physics ensembles of atmosphere-ocean general circulation model (AOGCM) simulations can be used to represent modelling uncertainty in climate projections. We start by considering the challenges involved in ensemble design owing to the high dimensional parameter space in AOGCMs, introducing the technique of emulation that can be used to efficiently target regions of parameter space having properties of interest for a particular research question. We then present results from the BBC climate change experiment (BBC CCE), the first multi-thousand member coupled AOGCM ensemble exploring un- certainties in the transient response. We find a "likely" range (66% confidence interval) of 1.4-3K for global mean warming by 2050 relative to 1961-1990 under a mid-range emissions scenario. The range is larger than observed in multi-model ensembles of opportunity, especially at the upper end and is more consistent with the subjective estimate given in the Intergovernmental Panel on Climate Change Fourth Assessment Report (IPCC AR4). The results provide the first direct AOGCM evidence of high response worlds that are consistent with the recent observed warming, providing an additional set of physically coherent input scenarios that can support climate impact assessments. Application of observational constraints in calculating model error requires a met- ric to weight individual components, often through an inverse covariance matrix. We demonstrate how the sample covariance matrix is a poor estimator in many situations faced in climate research and follow up previous work introducing regularized covariance estimation, which is applied to the analysis of the BBC CCE. This stabilises uncertainty estimates removing the need for empirical choices of the truncation in the model error calculation. Finally we introduce objective Bayesian statistics as a methodology to address some of the difficulties faced when specifying prior distributions for AOGCM parameters. Results from applying Jeffreys' prior to a simplified energy balance model of the climate system suggest the approach can be applied to complex AOGCMs and therefore provides a complementary alternative to more traditional subjective Bayesian methods. However, we conclude that the probabilistic framework adopted in quantifying uncertainty should be partially motivated by the forecast of interest and strength of relevant observational constraints in the climate system.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available