Partial Differential Equation Models and Deep Learning for the Sea Ice Concentration Field #
Delaney Mosier
14:10 Tuesday in 2Q42.
Part of the Sea ice modelling session.
Abstract #
The area fraction of ocean surface covered by ice is an important satellite data product called the sea ice concentration field, which has played an important role in observing the significant impact of planetary warming on Earth’s polar ice packs. We develop several mathematical models for the concentration field in both space and time, with increasing levels of complexity and realism. Motivated by prior use of Laplace equation models, we consider both Poisson equation and advection diffusion equation models of ice concentration. Further, we propose a deep learning approach for inferring advection diffusion coefficients based on satellite data. Our investigation of PDE models of the concentration field and its evolution gives us novel tools to mathematically analyze the changes in the polar sea ice covers under global warming, and to more efficiently represent sea ice in climate models.