The insurance industry is becoming increasingly exposed to the adverse impacts of climate variability and climate change. In developing policies and adapting strategies to better manage climate risk, insurers and reinsurers are therefore engaging directly with the climate modelling community to further understand the predictive capabilities of climate models and to develop techniques to utilise climate model output. With an inherent interest in the present and future frequency and magnitude of extreme climate-related loss events, insurers rely on the climate modelling community to provide informative model projections at the relevant spatial and temporal scales for insurance decisions. Furthermore, given the high economic stakes associated with enacting strategies to address climate change, it is essential that climate model experiments are designed to thoroughly explore the multiple sources of uncertainty. Determining the reliability of model based projections is a precursor to examining their relevance to the insurance industry and more widely to the climate change adaptation community. Designing experiments which adequately account for uncertainty therefore requires careful consideration of the nonlinear and chaotic properties of the climate system. Using the well developed concepts of dynamical systems theory, simple nonlinear chaotic systems are investigated to further understand what is meant by climate under climate change. The thesis questions the conventional paradigm in which long-term climate prediction is treated purely as a boundary value problem (predictability of the second kind). Using simple climate-like models to draw analogies to the climate system, results are presented which support the emerging view that climate prediction ought to be treated as both an initial value problem and a boundary condition problem on all time scales. The research also examines the application of the ergodic assumption in climate modelling and climate change adaptation decisions. By using idealised model experiments, situations in which the ergodic assumption breaks down are illustrated. Consideration is given to alternative model experimental designs which do not rely on the assumption of ergodicity. Experimental results are presented which support the view that large initial condition ensembles are required to detail the changing distribution of climate under altered forcing conditions. It is argued that the role of chaos and nonlinear dynamic behaviour ought to have more prominence in the discussion of the forecasting capabilities in climate prediction.
|Publication status||Published - Dec 2012|