Seasonal transitions of Arctic sea ice over the satellite era in CMIP6 models

Projections of Arctic sea ice area show substantial model spread in CMIP3, CMIP5 and early results from CMIP6. Here we assess how simulated seasonal transitions in Arctic sea ice may be contributing to the large inter-model spread. For this we make use of CMIP6 models, the CESM Large Ensemble and the new Arctic Sea Ice Seasonal Change and Melt/Freeze Climate Indicators satellite dataset. Spring ice loss and fall ice growth can be characterized by various metrics (melt onset, break-up, opening, freeze onset, freeze-up, closing). By assessing numerous metrics of seasonal sea ice transitions, we evaluate a range of ice loss and gain processes in CMIP6 models, as well as biases that may contribute to the large spread in model projections of Arctic sea ice. We show that model biases in seasonal sea ice transitions can compensate for other unrealistic aspects of the sea ice, such as very low ice thickness, resulting in acceptable September sea ice areas for the wrong reasons. Furthermore, we find that the metrics of seasonal sea ice change, while often used interchangeably, are not related to ice area and thickness in the same ways.