Skip to main content

Tropical cloud feedbacks estimated from observed multi-decadal trends

Emily
van de Koot
University of Oxford
Michael Byrne, University of St. Andrews
Tim Woollings, University of Oxford
Poster
Significant challenges in modelling clouds limit our understanding of how changes in cloud radiative effects (CREs) influence long-term warming and circulation trends. Previous observational estimates of cloud feedbacks rely on the CERES dataset, but this begins in 2000 so estimates largely focus on inter-monthly or inter-annual co-variability between CREs and surface temperatures. In this study, we instead use top-of-atmosphere radiative fluxes from the DEEP-C dataset, which spans 1985-2020, allowing estimation of tropical cloud feedbacks from multi-decadal trends in surface temperature and top-of-atmosphere radiation. We perform the same analysis using ERA5, which does not assimilate top of atmosphere radiative fluxes, and compare the results, gaining insight into the impact of sub-gridscale parameterisations on the reanalysis. Furthermore, we compare the trend-based feedback estimates to feedbacks calculated from inter-monthly variability in each dataset.

Following Bony et al. (2004), we separate cloud feedbacks into a wide range of circulation regimes to explore the extent to which cloud feedbacks are circulation-driven. We then map out the spatial distribution of the dynamic and non-dynamic cloud feedbacks, to gain further insight into the physical processes driving the changes. We find that dynamically-driven cloud feedbacks, primarily associated with the narrowing and strengthening of tropical ascent, form a critical component of local feedbacks while averaging to zero in the tropical-mean. Thus, although local magnitudes are similar, the non-dynamic component dominates the tropical mean. We also show that both the dynamic and non-dynamic components of the variability-derived feedbacks differ from the trend-derived feedbacks. Finally, we find that ERA5 has a strong negative shortwave cloud feedback on multidecadal timescales, likely associated with biases in the CRE climatology, which could in turn influence the simulated trends in large-scale circulation.
Poster thumbnail