Uncertainty in climate response to carbon dioxide and implications for mitigation policy

Global mean surface air temperature (GMST) change, due to increases in atmospheric carbon dioxide concentrations, is a fundamental measure of human induced climate change and its associated impacts. Estimates of the magnitude of GMST response to a doubling of atmospheric carbon dioxide (equilibrium climate sensitivity or ECS) remains uncertain over a broad range between 1.5-4.5K. Expected economic damages associated with climate change are strongly sensitive to this broad uncertainty, creating challenges in constructing mitigation policies to limit peak warming. In this thesis I explore uncertainty in the climate response through two methods. I show that potential constraints on the climate response using observations of a recent decade and planetary energy balance models are consistent with a low climate response that is not sampled by multi-model or perturbed physics ensembles of general circulation models (GCMs). I therefore subsequently set out to explore the physicality of the lower bound of climate response uncertainty in GCMs by conducting an emulator-driven perturbed physics ensemble search for low ECS models. I find a set of GCMs with ECS between 1.5K and 2K, driven primarily by negative feedbacks in tropical low cloud with GMST warming. Achieving low ECS is associated with reduced simulation fidelity relative to the standard version of the GCM, but fidelity reductions are judged to be insufficient to assuredly rule out an ECS between 1.5-2K in the real climate system. This experiment highlights the difficulty in further reducing climate response uncertainty in the near future. I therefore propose a new framing of mitigation policy that focuses on using physical methods to index the future evolution of policy variables to emergent climate change. Such a framing could lead to adaptive mitigation policies, aimed at limiting peak warming, that are demonstrably more robust under currently irreducible physical climate response uncertainty than conventional mitigation scenarios.