Sea ice dynamics prediction
Use Multi-marginal Optimal Transport for Sea Ice Dynamics Prediction
Joint with M. Parno [1], we develop a multi-marginal optimal transport (MMOT) framework for constructing continuous representations of discrete-in-time data. One natural application is in predicting the sea ice dynamics. Given a sequence of observations (the “marginals” in the MMOT framework) recorded at different time points, our method produces a continuous-time interpolation that captures the evolving motion of sea ice. This provides a solution to one stage in the Lagrangian Observation Mapping Challenges. The accompanying Python package and documentation are available in [2].
Using the SAR data (collected every 6 days) from Alaska Satellite Facility on the Robertson Channel (thanks to our group member J. Park at Dartmouth), the animation demonstrates predicted sea ice motion at two-day intervals. Remarkably, even finer temporal predictions can be generated within minutes on a standard personal computer.