Using CryoSat-2 estimates to analyse sub-grid scale sea ice thickness distribution in HadGEM3 simulations for CMIP6

A sub-grid scale sea ice thickness distribution (ITD) is a key parameterization to enable a large-scale sea ice model to simulate winter ice growth and sea ice ridging processes realistically. Recent sophisticated developments, e.g. a melt pond model, a form drag parameterization, a floe-size distribution model, fundamentally depend on the ITD. In spite of its importance, knowledge is poor about the accuracy of the simulated ITD. Here, we derive the ITD from individual Arctic sea ice thickness estimates available from the CryoSat-2 (CS2) radar altimetry mission during ice growth seasons since 2010. We bin the CS2 data into 5 ice thickness categories used by the sea ice component CICE of HadGEM3 climate simulations: (1) ice thickness h < 60 cm, (2) 60 cm < h < 1.4 m, (3) 1.4 m < h < 2.4 m, (4) 2.4 m < h < 3.6 m, (5) h > 3.6 m. Our analysis includes historical simulations and future projections with the HadGEM3-GC31 model as well as forced ocean-ice and standalone ice simulations with the same model components NEMO v3.6 and CICE v5.1.2. The most striking difference occurs regarding the annual cycle of area fraction of ice in the thickest category (> 3.6 m). According to CS2, in the Central Arctic the fraction is below 2% in October and increases to 15-40% in April. In contrast the annual cycle is weak in all simulations. The magnitude of the area fraction differs between the simulations. For simulations which agree best with CS2 for grid cell mean ice thickness, the area fraction of thick ice is around 5% constantly throughout the whole year. Potential reasons for the discrepancy are discussed and sensitivity experiments presented to study the impact of sea ice settings on the simulated ITD, e.g. ice strength parameter, parameter for participating in ridging, heat transfer coefficients.