Estimating Missing Observations of Ice Floes Using Data Assimilation

Satellite observations are increasingly important to the study of Arctic sea ice. Imagery from satellites can be used to form datasets of Lagrangian observations of ice floe positions and orientations over time. However, such images are often partially obscured by weather, leading to missing observations in the dataset. Using an ensemble Kalman smoother with a simple sea ice-ocean model, we can estimate missing data by interpolating floe trajectories between the available observations. In contrast to linear interpolation, these interpolated observations better reflect floe dynamics, as well as incorporating internal estimates of ocean features and ice thickness from the model.