Climate sensitivity estimated from LGM ensemble simulations

Recent studies have shown that it is not possible to reduce the uncertainty range of climate sensitivity (CS) by focusing on last Century warming or on present day climatology. Alternative approaches of estimating the sensitivity of the Earth system comprise the use of paleo-data, ideally based on a period with a radiative forcing and global mean temperature pronouncedly different to modern day climate, such as the Last Glacial Maximum (LGM, 21kyrs B.P.). Yet the forcing of the glacial climate is of different nature compared to the forcing causing future climate changes. Thus it had been questioned to what extent the past is a good analogue for the future. Purely proxy-data based estimates of CS turn out to be biased in case the sensitivity of the climate system to past and future forcings is pronouncedly different. Model-based studies do account for differences in the feedback behaviour between simulated glacial cooling and 2xCO2 warming, but currently do not give a consistent picture of the asymmetry in the feedback strengths. In our study we combine the knowledge from LGM paleo-data with model simulations of present day, 2xCO2, and glacial climate. For this purpose we used a fully-coupled model of intermediate complexity (CLIMBER-2) that is fast enough to allow for running various ensemble experiments (order of 1000 model runs). By having simultaneously perturbed a set of 11 model parameters that strongly affect the model's fast feedbacks (water vapour, clouds, albedo and lapse rate) we could cover a broad range of climate sensitivities. We then have used LGM paleo-data from the tropics and from Antarctica to constrain the set of models being consistent with the glacial climate. Our results suggest that very high estimates of CS (larger than 5° C) are hard to reconcile with proxy-evidence from the LGM. We especially address the question of model dependence of our results by analyzing the simulated feedback strengths. We compare these key model characteristics with recent GCM results and discuss the issue of asymmetry in the feedback strengths between glacial and 2xCO2 conditions.