Sharper Resolution of Arctic Sea Ice Dynamics With Non‐Conforming Finite Elements in FESOM2

Abstract

Sea ice dynamics in numerical models are discretized on a spatial grid, with variables defined at grid cell vertices, edges, or centers. The CD‐grid discretization places the velocity vector on element edges and tracers on vertices. On a triangular grid, the number of edges is three times the number of nodes, therefore placing velocity on edges increases the number of degrees of freedom and effectively enhances the spatial resolution on the same underlying mesh, compared to the vertex‐based placement of velocity of the A‐grid. In this study, we present the first application of the CD‐grid in a realistic, high‐resolution, fully coupled sea ice–ocean simulation using FESOM2, with a horizontal resolution of 4.5 km in the Arctic. We show that the CD‐grid produces a more sharply resolved sea ice field with enhanced small‐scale structure relative to the A‐grid configuration, while preserving large‐scale sea ice characteristics. We further propose a new stress‐subcycling formulation within the mEVP method, in which the divergence of stress is iterated instead of the stress tensor itself. This reformulation is necessary to stabilize the CD‐grid, and, for other discretizations, reduces computational cost while producing identical results. An analysis of the parallel performance of the CD‐grid across a wide range of core counts shows that, within the context of a coupled sea ice–ocean model, the total runtime is only 5%–8% higher than for the A‐grid configuration.