Plain Language Summary

Abstract

General Circulation Model (GCM) simulations with prescribed observed sea surface temperature (SST) over the historical period show systematic global shortwave cloud radiative effect (SWCRE) variations uncorrelated with global surface temperature (known as “pattern effect”). Here, we show that a single parameter that quantifies the difference in SSTs between regions of tropical deep convection and the tropical or global average (Δconv) captures the time-varying “pattern effect” in the simulations using the PCMDI/AMIPII SST recommended for CMIP6. In particular, a large positive trend in the 1980s–1990s in Δconv explains the change of sign to a strongly negative SWCRE feedback since the late 1970s. In these decades, the regions of deep convection warm about +50% more than the tropical average. Such an amplification is rarely observed in forced coupled atmosphere-ocean GCM simulations, where the amplified warming is typically about +10%. During the post 2000 global warming hiatus Δconv shows little change, and the more recent period of resumed global warming is too short to robustly detect trends. In the prescribed SST simulations, Δconv is forced by the SST difference between warmer and colder regions. An index thereof (SST#) evaluated for six SST reconstructions shows similar trends for the satellite era, but the difference between the pre- and the satellite era is substantially larger in the PCMDI/ AMIPII SSTs than in the other reconstructions. Quantification of the cloud feedback depends critically on small changes in the shape of the SST probability density distribution. These sensitivities underscore how essential highly accurate, persistent, and stable global climate records are to determine the cloud feedback.