How the good (global-mean surface temperature) and the bad (surface temperature patterns and top of the atmosphere radiative imbalance) conspire to the ugly (radiative feedbacks)

Using large initial condition ensembles, I show how GCMs over the last 24 years simulate global-mean temperature well but fail to reproduce the observed top of atmosphere radiative imbalance. It follows that GCMs do not correctly represent the observed radiative feedback, which is defined as a change in radiation with a change in global-mean temperature. This is due to both an erroneous representation of SST patterns and atmospheric physics: GCMs indeed consistently misrepresent the surface warming pattern for many, but not all, timescales over the last 70 years. In addition, AMIP simulations do not necessarily represent recent TOA radiation trends correctly.
I will discuss two implications of this particular good-bad-and-ugly combination: First, I will quantify how a continuation of an erroneous representation of SST patterns will affect the radiative imbalance and global-mean temperature in the near future. Second, I will discuss how the research on the pattern effect points to an ubiquitous model problem in the magnitude of ocean heat uptake and to large-scale erroneous ocean-atmosphere interactions in general.