Constraining cloud and convective parameterizations in GISS-E2.1 using paleoclimate data

Abstract:

Cloud and convective parameterization differences explain much of the spread in equilibrium climate sensitivity (ECS) estimates among climate models. Changes to these parameterizations directly influence water isotopes and therefore can be used as benchmarks for constraining model processes and identifying and correcting model biases. Large collections of water isotope measurements based on climate archives provide a unique opportunity to evaluate model performance across time periods drastically different from today. In this study, we evaluate a suite of isotope-enabled GISS-E2.1 simulations (n = 9), each of varying perturbations in cloud parameters, against δ¹⁸O compilations from speleothems and ice cores spanning the Last Glacial Maximum (LGM), mid-Holocene (MH) and pre-Industrial periods. The first-order spatial pattern of δ¹⁸O of precipitation (δ¹⁸O_p) is in excellent agreement between proxy data and all simulations across time periods (i.e., r² > 0.94, CE > 0.82). However, the magnitude of δ¹⁸O_p anomalies (i.e., LGM minus PI and MH minus PI) are consistently smaller in all simulations than those of the proxies. The simulated δ¹⁸O_p changes in the LGM are substantially overestimated (i.e., more positive) relative to the proxies, particularly over Greenland and most of Eastern Antarctica in all simulations. For the MH, the sign of simulated δ¹⁸O_p anomalies and direction of the mean offset repeatedly change depending on the model run, likely in response to changes in mean climate. The simulation showing the highest skill is not the same for LGM and MH changes, supporting the current approach of determining potential combinations of parameters to finetune future simulations, critical in refining ECS estimates.