On the strongly negative cloud feedback over the satellite period implied by observational SST reconstructions

Clouds strongly modulate Earth's radiative budget, and uncertainties in numerical model simulations of the global cloud field contribute substantially to uncertainties in future warming. In coupled atmosphere-ocean General Circulation Model (GCM) simulations, the global cloud field and its radiative effect are well correlated with global average surface temperature. However, GCM simulations with prescribed Sea Surface Temperatures (SSTs) from observational SST reconstructions over the historical period show time-varying relationships between the cloud field and average surface temperature (known as the "pattern effect"). We show that CERES/EBAF observational data confirms the presence of a second mode (in addition to mean SST) in particular in low cloud amount (and correspondingly SWCRE) that is consistent with variations in tropical atmospheric stability in ERA-Interim reanalysis data. This second mode in observations is tied to ENSO, and evolves in quadrature to ENSO indexes. It arises from differences in surface temperature change between regions of tropical deep convection and the tropical (or global) average. In contrast to the multidecadal trends over the full historical period, trends in this second mode since the year 2000 are small. The PCMDI/AMIPII SSTs recommended for CMIP6 stand out as having the largest trend over the full historical period. Different SST reconstructions agree on a trend over the satellite period - specifically the 1980s-90s - that is much larger than what coupled GCM simulations show: In forced coupled GCM simulations the regions of deep convection warm order 10% more than the tropical average, whereas over the satellite period the amplification is order +50%  in the AMIP simulations and in estimates using rainfall observations to identify regions of deep convection.