Climate feedback uncertainty quantification and learning using reduced models

Physical uncertainty in global temperature projections depends in large part on uncertainty in the climate sensitivity to carbon dioxide due to system feedbacks. We investigate the application of reduced climate models, including energy balance models and low-order stochastic systems, to efficiently explore the space of uncertainty. Using Bayesian parameter estimation on simulated data, we investigate the expected rate-of-learning about climate sensitivity into the future as more observational data becomes available, thereby narrowing the range of uncertainty in climate projections.