Crossbreeding CMIP6 Earth System Models With an Emulator for Regionally Optimized Land Temperature Projections

The newest generation of the Coupled Model Intercomparison Project (CMIP6) exhibits a larger spread in temperature projections at the end of the 21st century than the previous generation. Here, a modular Earth System Model emulator is used to evaluate the realism of the warming signal in CMIP6 models on both global and regional scales, by comparing their global trends and regional response parameters to observations. Subsequently, the emulator is employed to derive large "crossbred" multimodel initial-condition ensembles of regionally optimized land temperature projections by combining observationally constrained global mean temperature trend trajectories with observationally constrained local parameters. In the optimized ensembles, the warmest temperature projections are generally reduced and for the coolest projections both higher and lower values are found, depending on the region. The median shows less changes in large parts of the globe. These regional differences highlight the importance of a geographically explicit evaluation of Earth System Model projections. Plain Language Summary The newest generation of state-of-the-art Earth System Models (ESMs) exhibits more uncertainty in temperature projections at the end of the 21st century than the previous generation. Here we use an ESM emulator, that is, a statistical tool which mimics the behavior of ESMs, to evaluate this newest generation of ESMs with respect to observations and to generate new ensembles combining the best performing features of these ESMs at both global and regional scales. The new ensembles consist of a large number of regionally optimized land temperature projections. At the end of the 21st century, the warmest regional temperature projections are generally lower in the optimized ensembles and the coolest projections are either reduced or enhanced, depending on the region. The temperature projections in the middle of the distribution are less affected in large parts of the globe. These regional differences emphasize the importance of also considering ESM performance in individual regions, not just the global mean, when evaluating climate change projections with current ESMs.