Constraining CMIP6 sea ice simulations with ICESat-2

Abstract. This study evaluates sea ice simulations from the Coupled Model Intercomparison Project Phase 6 (CMIP6) using modern-era satellite measurements of sea ice area, total freeboard, and thickness. Current global climate models (GCMs) exhibit substantial uncertainties in simulating sea ice, with significant contributions from both model uncertainty and internal variability. In our study, simulated Arctic and Southern Ocean total freeboard and Arctic winter sea ice thickness are assessed with data from NASA’s ICESat-2 mission, providing an additional constraint beyond traditional passive-microwave sea ice area comparisons used extensively in previous studies. Freeboard comparisons benefit from accurate observations from satellite laser altimetry but motivate increased focus on bulk sea ice density estimates across models and observations. The short observational time period also increases the role of internal variability. We undertake a plausibility assessment where we account for both observational uncertainty and internal variability across our different metrics for both hemispheres. In general, we see more plausible metrics in the Arctic compared to the Southern Ocean, with important differences when analyzing annual means vs. March and September means. We provide an example of this plausibility assessment by producing constrained estimates of 2015–2035 seasonal sea ice volume, using model subsets constrained using either area metrics or the combined area, freeboard and thickness metrics, with freeboard and thickness providing important additional impacts in terms of the mean seasonal cycle and spread. Finally, we present regional comparisons and a composite analysis, with models showing systematic underestimation of thicker ice in the western Arctic and clear differences in the simulation of Eastern/Western Arctic sea ice. Overall, our study provides novel insights into sea ice model evaluation and emphasizes the potential benefits of integrating altimetry data from ICESat-2, as well as providing a discussion on the potential utility of these model constraints and future research priorities.