Drivers of (non-)linear Eocene surface warming in the DeepMIP ensemble

The early Eocene greenhouse represents the warmest interval of the Cenozoic and therefore provides a unique opportunity to understand how the climate system operates under elevated atmospheric CO₂ levels similar to those projected for the end of the 21st century. Early Eocene geological records indicate a large increase in global mean surface temperatures compared to present day (by ~14°C) and a greatly reduced meridional temperature gradient (by ~30% in SST). However, reproducing these large-scale climate features at reasonable CO₂ levels still poses a challenge for current climate models. Recent modelling studies indicate an important role for shortwave (SW) cloud feedbacks to drive increases in climate sensitivity with global warming, which helps to close the gap between simulated and reconstructed Eocene global warmth and temperature gradient. Nevertheless, the presence of such state-dependent feedbacks and their relative strengths in other models remain unclear.

In this study, we perform a systematic investigation of the simulated surface warming and the underlying mechanisms in the recently published DeepMIP ensemble. The DeepMIP early Eocene simulations use identical paleogeographic boundary conditions and include six models with suitable output: CESM1.2_CAM5, GFDL_CM2.1, HadCM3B_M2.1aN, IPSLCM5A2, MIROC4m and NorESM1_F. We advance previous energy balance analysis by applying the approximate partial radiative perturbation (APRP) technique to quantify the individual contributions of surface albedo, cloud and non-cloud atmospheric changes to the simulated Eocene top-of-the-atmosphere SW flux anomalies. We further compare the strength of these planetary albedo feedbacks to changes in the longwave atmospheric emissivity and meridional heat transport in the warm Eocene climate. Particular focus lies in the sensitivity of the feedback strengths to increasing global mean temperatures in experiments at a range of atmospheric CO₂ concentrations between x1 to x9 preindustrial levels.

Preliminary results indicate that all models that provide data for at least 3 different CO₂ levels show an increase of the equilibrium climate sensitivity at higher global mean temperatures. This is associated with an increase of the overall strength of the positive SW cloud feedback with warming in those models. This nonlinear behavior seems to be related to both a reduction and optical thinning of low-level clouds, albeit with intermodel differences in the relative importance of the two mechanisms. We further show that our new APRP results can differ significantly from previous estimates based on cloud radiative forcing alone, especially in high-latitude areas with large surface albedo changes. We also find large intermodel variability and state-dependence in meridional heat transport modulated by changes in the atmospheric latent heat transport. Ongoing work focuses on the spatial patterns of the climate feedbacks and the implications for the simulated meridional temperature gradients.