A large-scale high-resolution numerical model for sea-ice fragmentation dynamics

Abstract

Forecasts of sea-ice motion and fragmentation are of vital importance for all human interactions with sea ice, ranging from those involving indigenous hunters to shipping in polar regions. Sea-ice models are also important for simulating long-term changes in a warming climate. Here, we apply the Helsinki Discrete Element Model (HiDEM), originally developed for glacier calving, to sea-ice breakup and dynamics. The code is highly optimized to utilize high-end supercomputers to achieve an extreme time and space resolution. Simulated fracture patterns and ice motion are compared with satellite images of the Kvarken region of the Baltic Sea from March 2018. A second application of HiDEM involves ice ridge formation in the Gulf of Riga. With a few tens of graphics processing units (GPUs), the code is capable of reproducing observed ice patterns that in nature may take a few days to form; this is done over an area of ∼100km × 100km, with an 8 m resolution, in computations lasting ∼10 h. The simulations largely reproduce observed fracture patterns, ice motion, fast-ice regions, floe size distributions, and ridge patterns. The similarities and differences between observed and computed ice dynamics and their relation to initial conditions, boundary conditions, and applied driving forces are discussed in detail. The results reported here indicate that the HiDEM has the potential to be developed into a detailed high-resolution model for sea-ice dynamics at short timescales, which, when combined with large-scale and long-term continuum models, may form an efficient framework for forecasts of sea-ice dynamics.