Reducing uncertainty in equilibrium climate sensitivity, commonly defined as the surface temperature response to an instantaneous doubling of atmospheric CO2 concentrations, has been an abiding goal of the climate science community for the past decades. The range of plausible climate sensitivities, derived from ever improving generations of GCMs, however, remains essentially unchanged.

Here we present a new approach to multi-model uncertainty quantification (UQ) using a reduced-form simple climate model, tuned to reproduce relevant physics of different global climate models (GCMs). Rather than focusing on the discrete projections made by individual GCMs, our simple model allows us to smoothly interpolate between the dynamics of the multi-model ensemble by forming a continuous probability distribution over a reduced model parameter space.

We will discuss an early version of the simple model, an idealized ocean-atmosphere energy balance model (EBM). The EBM is fit to surface temperature, ocean heat content and top-of-atmosphere radiative fluxes of GCMs participating in the Coupled Model Intercomparison Project version 5 by varying several parameters, including climate sensitivity and feedback. We obtain distributions of these parameters and update the uncertainties associated with them using observations of surface temperature, ocean heat uptake and top-of-atmosphere radiative flux in a Bayesian inference framework.