A sensitivity study of the sea ice simulation in the global coupled climate model, HadGEM3

• We perform a sea ice sensitivity study using a coupled atmosphere–ocean-ice model.
• Test sensitivity to sea ice parameter changes, and to atmosphere and ocean changes.
• Arctic sea ice is most sensitive to snow albedo and ice/snow thermal conductivities.
• In the Antarctic, effects of changes in the atmosphere and ocean dominate.
• Changes result in improved representation of Arctic sea ice in the model.

We present the results of a wide-ranging sea ice sensitivity study, performed with a fully-coupled global atmosphere–ice-ocean climate model. We investigate sensitivity to a selection of sea ice parameters, varied within the range of observational uncertainty, and additionally study the effect on the sea ice of increased resolution in the atmosphere and ocean-ice models, as well as dynamics and physics changes in the atmosphere. In the Arctic, we find that the sea ice thickness is most sensitive to the albedo of the overlying snow layer (because of its influence on surface melt) and the thermal conductivities of ice and snow (because of their role in regulating heat flux from the ocean to the atmosphere through the ice). We find the winter Arctic ice extent to be sensitive to the resolution of the ocean-ice model, through increased sea surface temperatures in the Labrador Sea at higher resolution. The Arctic ice extent is reduced under increased atmospheric resolution, because of increased poleward heat transport. In the Antarctic, the sensitivity to sea ice parameters is weaker, and atmosphere and ocean forcing dominate; in particular, increased resolution of the atmosphere and ocean-ice models leads to the enhancement of a warm bias in the Southern Ocean, which has a large impact on sea ice thickness and extent. Inclusion of a selection of these parameters in combination, together with changes to the atmosphere and ocean models, leads to significant improvements in representation of Arctic sea ice extent, thickness and volume in a new global coupled model configuration.