Perturbing Parameters to Understand Cloud Contributions to Climate Change
A perturbed parameter ensemble (PPE) samples the uncertainty associated with poorly constrained parameters in a model. Here 45 parameters are varied in 262 experiments with CAM6. Each experiment includes a simulation forced by current climate conditions and another with a uniformly warmed sea surface.
The results show that estimates of climate feedbacks and climate sensitivity from the PPE are spread out nearly as much as an ensemble of completely different climate models. This hints that choosing parameter values has a major impact on a model’s climate sensitivity, and could be as important as the choices made in designing the model.
The PPE is scrutinized in terms of the top-of-atmosphere radiation balance, and a few members are excluded because their climates are inconsistent with observations. After member selection, 206 experiments remained. The cloud feedbacks in each experiment are quantified using the well-established method of cloud radiative kernels. This method relies on using satellite simulator software that provides cloud binned by optical depth and cloud-top pressure. The resulting feedback values are compared with similarly-calculated feedbacks from the CMIP6 models. Feedbacks are further decomposed into seven components following the WCRP climate sensitivity assessment. In general, the CAM6 PPE shows spread that comparable to the CMIP6 ensemble. CAM6 has high-cloud altitude feedback that appears stronger than most models. Several parameters are identified that have especially pronounced impacts on the feedbacks, but they are spread across the physics parameterizations (deep convection, moist turbulence, and microphysics). Analysis shows that the change in sensitivity from CESM1 to CESM2 is likely tied to structural changes (e.g., the move to CLUBB) rather than parametric changes made to tune the model. This analysis also uncovered a decrease in climate sensitivity between the versions of CAM6 used for CMIP6 and the PPE.