Theory-based emergent constraints on climate sensitivities

Abstract

Earth system models show a wide variation of possible futures under climate change. To develop appropriate policy for curbing global warming and adapting to unavoidable change, better understanding of the climate system is crucial. One of the approaches to reduce uncertainty in climate models is the identification of emergent constraints. These are physically plausible empirical relationships between a particular simulated characteristic of the current climate versus future climate change from an ensemble of climate models, which can be exploited to reduce uncertainty using current observations. This thesis discusses various interpretations of this technique and includes a comparison with other methods that combine models and observations in climate. A mathematical theory based on linear response theory is developed for an important subset of constraints showing how nonlinear relationships appear from an interplay of system and forcing time scales. Several statistical issues are examined, such as the best way to deal with internal variability. This theory is applied to three emergent relationships. Decadal climate variability and climate sensitivity are found to be related in both conceptual climate models and in CMIP5 climate model ensemble. This results in compound risk: significant temperature surges on top of the long-term trend are more likely if climate sensitivity is high. Two additional emergent relationships are used to find a constraint on transient warming and climate sensitivity from observed warming. They exploit the fact that uncertain aerosol cooling is increasingly overshadowed by greenhouse gas warming. Many of the CMIP6 climate models show a warming inconsistent with observed trends. The thesis concludes with suggestions for future work.