On the ability of CMIP6 OMIP models to simulate the seasonalcycle of the ocean mixed layer depth in the central Arctic Ocean

In this study, we assess the ability of the ocean-sea ice general circulation models that participated in the CMIP6 Ocean Model Intercomparison Project (OMIP) to simulate the seasonal cycle of the ocean mixed layer depth in the area of the Arctic Ocean covered by multiyear sea ice. During summertime, all models understimate the mixed layer depth by about 20 m compared to the MIMOC (Monthly Isopycnal/Mixed layer Ocean Climatology) observational data. The origin of this systematic bias is unclear. In fall and winter, differences of several tens of meters are noticed between the models themselves and between the models and the observational data. Some models generate too deep mixed layers, while others produce too shallow mixed layers. Since the mixed layer deepening in ice-covered regions during these seasons is largely controlled by the brine rejection associated with ice growth, the discrepancies between models might be related to differences in the modelled sea ice mass balance. However, a detailed model comparison reveals that this is not the case, all models simulating more or less the same sea ice mass balance and thus salt flux into the ocean during sea ice freezing. By applying to model outputs the analytical model developed by Martinson (1990), that allows in particular to determine the main processes responsable for maintaining stablility in polar oceans, it is finally found that most of the disagreement between models can be explained by the accuracy with which the Arctic halocline is reproduced by those models. This feature is simulated generally poorly and quite differently from one model to another, and models with less stratified halocline generally lead to deeper mixed layers. It now remains to identify the model deficiencies responsible for this situation.