Application of discrete-element methods to approximate sea-ice dynamics

• Main Authors: Damsgaard, Anders, Adcroft, Alistair, Sergienko, Olga
• Format: Report
• Language: unknown
• Published: EarthArXiv 2018
• Subjects: Glaciology; Earth Sciences; Physical Sciences and Mathematics; Other Physical Sciences and Mathematics; Oceanography and Atmospheric Sciences and Meteorology; Sea ice
• Online Access: https://dx.doi.org/10.17605/osf.io/j6vpn; https://eartharxiv.org/j6vpn/

Lagrangian models of sea-ice dynamics have several advantages over Eulerian continuum models. Spatial discretization on the ice-floe scale as well as arbitrary concentrations are natural for Lagrangian models. This allows for improved model performance in ice-marginal zones. Furthermore, Lagrangian models can explicitly simulate jamming processes similar to sea ice movement through narrow confinements. Granular jamming is a stochastic process that occurs when the right grains arrive at the right place at the right time, and the jamming likelihood over time can be described by a probabilistic model. While difficult to parameterize in continuum formulations, jamming emerges spontaneously in dense granular systems simulated in a Lagrangian framework. Here, we present a flexible discrete-element framework for approximating Lagrangian sea-ice mechanics at the ice-floe scale, forced by ocean and atmosphere velocity fields. Our goal is to optimize the computational efficiency of mechanical ice-floe interaction relative to traditional discrete-element methods for granular dynamics. We demonstrate that frictionless contact models based on compressive stiffness alone are unlikely to produce jamming, and describe two different approaches based on Coulomb-friction and cohesion which both result in increased bulk shear strength of the granular assemblage. The frictionless but cohesive contact model can display jamming behavior which on the large scale is highly similar to the more complex model with Coulomb friction and ice-floe rotation, and is significantly simpler in computational cost.