Abstract:  Sea ice is a defining component of polar environments and impacts nearly all maritime activities in the Arctic.   Like many processes in the Arctic, climate change is causing fundamental changes in the seasonal behavior of sea ice.  New summer shipping routes are becoming available and many nations are preparing for more extensive operations in a thinning sea ice pack.   With this additional activity comes a need for high resolution forecasts of sea ice dynamics.   This talk will discuss an ongoing effort to enable the efficient calibration of particle-based sea ice models from satellite-based remote sensing imagery.   A key to our approach is the use of optimal transport (OT) metrics, like the Wasserstein distance, to form misfit functions and likelihood functions in deterministic and statistical inverse problems.   Various interpretations of balanced and unbalanced OT metrics will be discussed and preliminary model calibration results will be provided.  The efficient computational implementation of semi-discrete OT will also be presented. 

Bio:  Matt is a former ACDL'er who received his Ph.D. in Computational Science and Engineering in 2015.   Since leaving MIT, he's been working on a variety of modeling and uncertainty quantification projects at the US Army Corps of Engineers Cold Regions Research and Engineering Laboratory (CRREL).   His research focuses on algorithms and software for solving Bayesian inverse problems arising in sea ice modeling, structural health monitoring, and remote sensing.