| Popis: |
Current operational sea ice models solve primitive equations on a grid and treat sea ice as a continuum with smoothly varying properties. This is the same method that is used in ocean models. The continuum assumption is unrealistic for sea ice which consists of separate rigid ice floes. The assumption works best for length scales much larger than typical floe size, and worst for very small length scales.Winter shipping in finnish ports depends on timely sea ice information on the Baltic Sea. Due to climate change, the yearly ice covered area and thermodynamic ice growth are decreasing. However, sea ice is also becoming more mobile and dynamic, especially in the Bay of Bothnia which lies in the north end of the Baltic Sea.A particle-based granular approach is more realistic in the length scales of individual ice floes. Such models have been developed (eg. by Mark Hopkins and Agnieszka Herman) and used successfully in limited scales, such as fjords. For larger horizontal scales, they have been computationally too expensive. Using modern GPU acceleration techniques, discrete element simulation of sea ice is becoming possible in the scale required for Baltic sea basins.This work presents an ongoing project for building a granular sea ice model for forecasting ice dynamics. This includes ice movement and deformation and describes ridge and lead formation and similar phenomena. Existing accelerated solvers are examined, and the most suitable is adapted for Baltic sea ice and applied for the Bay of Bothnia. |
