A Performance Portable Sea Ice Model Using the Discrete Element Method

  • Peterson, Kara (Sandia National Laboratories)
  • Turner, Adrian (Los Alamos National Laboratory)
  • Nikolov, Svetoslav (Sandia National Laboratories)
  • Bolintineanu, Dan (Sandia National Laboratories)
  • O'Connor, Devin (Sandia National Laboratories)
  • Clemmer, Joel (Sandia National Laboratories)

Please login to view abstract download link

The Discrete Element Model for Sea Ice (DEMSI) is a new sea ice model under development for use as a component in coupled Earth system models. DEMSI is based on the discrete element method, which models collections of ice floes as Lagrangian particles. In this method, contact forces between the particles are computed and the equations of motion for each particle are directly integrated in time to determine trajectories. This method enables the capture of the anisotropic, heterogeneous nature of sea ice deformation in response to ocean and atmospheric forcing. DEMSI utilizes the Large-scale Atomic/Molecular Massively Parallel Simulator (LAMMPS) library for the dynamics, which provides a high-performance base that relies on the Kokkos programming model for performance portability. We have implemented a Kokkos-enabled particle contact model for sea ice that includes both bonded and unbonded interactions to simulate ice fracture and pressure ridging. The Lagrangian motion of the particles results in significant convergence and ridging, which requires periodic remapping of sea ice variables from a deformed particle configuration back to an undeformed initial distribution. A conservative and bound-preserving remap of particle properties, based on a geometric remap algorithm, has be implemented to manage the particle convergence. We will describe the DEMSI model implementation including the contact model and remapping algorithm and demonstrate computational performance and mechanical behavior of the model on idealized test cases as well as more realistic Arctic regional simulations.