Probabilistic uncertainty quantification and simulation for climate modelling

Hauser, Tristan Paul (2014) Probabilistic uncertainty quantification and simulation for climate modelling. Doctoral (PhD) thesis, Memorial University of Newfoundland.

[English] PDF (Migrated (PDF/A Conversion) from original format: (application/pdf)) - Accepted Version Available under License - The author retains copyright ownership and moral rights in this thesis. Neither the thesis nor substantial extracts from it may be printed or otherwise reproduced without the author's permission. Download (21MB) [English] PDF - Accepted Version Available under License - The author retains copyright ownership and moral rights in this thesis. Neither the thesis nor substantial extracts from it may be printed or otherwise reproduced without the author's permission. (Original Version)

Abstract

This thesis addresses probabilistic approaches to uncertainty quantification, within the context of climate science. For the results of climate studies to be appropriately understood and applied, it is necessary to quantify their relation to the observable world. Probability theory provides a formal approach that can be applied commonly to the encountered uncertainties. Three studies are presented within. The first addresses the Bayesian calibration of climate simulators. This method quantifies simulation uncertainties by taking into account inherent model and observation uncertainties. Here an alternative method for the fast statistical emulators of model parameter relationships is tested, as well as a rigorous approach to quantifying model limitations. The second examines probabilistic methods for identifying regional climatological features and quantifying the related uncertainties. Such features serve as a basis of comparison for climate simulations, as well as defining, to some extent, how we view evolution of the modern climate. Here typical patterns are recreated using an approach that quantifies uncertainty in the data analysis. As well, temporal shifts in the distribution of these features and their relation to ocean variability is explored. The third study experiments with approaches to regional stochastic weather generation. There is an inherent residual between climate simulations and large scale features, and regional variability seen on daily timescales. Weather generators provide an error model to quantify this uncertainty, and define features and variability underrepresented in global simulations. A method is developed which allows for regional, rather than site specific, simulation for the North Atlantic, a region of very active and varied atmospheric activity. In total, the work presented within covers the range of uncertainty types that must be considered by climate studies. The individual articles addresses contemporary questions concerning appropriate methods and implementation for their probabilistic quantification.

Actions (login required)

View Item