Impact of subgrid-scale ice thickness distribution on heat flux on and through sea ice

We evaluated the impact of subgrid-scale ice thickness distribution on the heat flux on and through sea ice in a numerical model. An ice-ocean coupled model with a subgrid-scale ice thickness distribution scheme, COCO4.5, is forced by an atmospheric climatology to simulate the present state of the sea ice and ocean. The modeled climatology reproduces the ice cover reasonably well with a realistic ice thickness distribution. at flux on and through the sea ice is established using the grid-averaged sea-ice and snow-on-ice thickness from the results of the simulation. When the grid-averaged thickness is calculated as a weighted arithmetic mean, the conductive heat flux through the ice and snow is underestimated compared with that actually driving the model. This underestimation becomes smaller in magnitude when either a weighted harmonic mean or a weighted arithmetic mean with a modification based on the ratio of these two types of means is used. Rearrangement of the ice categories shows that the flux bias decreases with an increase in the number of categories. We also perform a sensitivity experiment in which the model is forced by the biased heat flux identified using the arithmetic mean of the ice thickness. A significant decrease in ice volume is found, notably in the Arctic Ocean. These results suggest that sea-ice models without an ice thickness distribution scheme underestimate the conductive heat flux through ice, and thereby the resultant sea-ice thickness, because the ice thickness from these models typically corresponds to the weighted arithmetic mean thickness.