On the variability and forced response of atmospheric regime systems

The nonlinear, eddy-driven dynamics of the Euro-Atlantic troposphere are notoriously complex. The regime hypothesis – the idea that the continuous range of atmospheric flow states can be meaningfully partitioned into a small number of large scale ‘regimes’ – provides a conceptual framework from which to understand the impacts of internal variability and external forcing on the Euro-Atlantic. In reanalysis and realistic model output, regime structure is noisy and often difficult to detect, confounded by the high dimensionality of the system and by model deficiencies. Therefore much of our understanding of regime systems comes from simple, conceptual models, where the mechanisms of regime dynamics can be studied cleanly. This dialogue across the hierarchy of models has been very fruitful, but it can be challenging to translate the behaviour of simple systems into testable hypotheses for the behaviour of the vastly more complex systems studied by the climate modeller or weather forecaster. This thesis contributes to bridging this divide by working across the model hierarchy to compare the behaviour of simple and complex regime systems. In Chapter 1 we explain the phenomena of stochastically induced regime persistence using the Charney deVore model, and frame our results in physical terms. To assess the impact of stochasticity on more complex regime structure, we find we must first address the large sampling variability in regime statistics. We introduce a novel regime framework in chapter 2 which provides very stable regime structure, and use it to compare stochastic and deterministic climate model simulations. In chapter 3 we characterise historical regime variability and assess the regime response of European climate change in CMIP6. We explicitly test the long-standing hypothesis that anthropogenic forcing will primarily change the frequency of regime occurrence rather than the regime patterns themselves. We show the hypothesis broadly holds and then extend the simple regime forcing model to better capture climate model trends.