Abstract

The wide range of Effective Climate Sensitivity (ECS) values in climate models are driven by inter-model spread in cloud feedbacks. Recent models (CMIP6) show an increase in both average ECS values as well as the appearance of very high ECS values (> 4.5 K) compared to the previous generation (CMIP5) which has been attributed to an increase in the strength of total cloud feedbacks in CMIP6. Constraining ECS and in particular the upper bound of ECS values is paramount for reliable predictions of future climate change.

The Last Glacial Maximum (LGM), is a recent paleoclimate and an out-of-sample climate for modern models. This work explores whether models with high ECS values are compatible with the LGM climate and whether we can use the LGM to constrain a plausible upper boundary of ECS. We create a single model ensemble with a wide range of ECS values by modifying cloud feedbacks in the MPI-ESM1.2 model. We simulate the LGM with this ensemble and compare it with multiple paleo-reconstructions. Our results indicate simulations with an ECS > 4 K are incompatible with the existing LGM climate reconstructions. This implies that models with large total cloud feedbacks and high ECS values are not plausible during the LGM.

This study highlights the value of using paleoclimates to benchmark models particularly in areas where existing validation techniques are not yet sufficient i.e. constraining cloud feedbacks.