In this work, we show a proof-of-concept that deep learning can parametrise the subgrid-scale for sea-ice dynamics. Although, this have been already proven useful for other components of the Earth ...