Ice Floe Tracker: An algorithm to automatically retrieve Lagrangian trajectories via feature matching from moderate-resolution visual imagery

Satellite observations of sea ice along marginal ice zones suggest a strong coupling between sea ice transport and the underlying ocean turbulent eddy field. Moderate Resolution Imaging Spectroradiometer (MODIS) satellite imagery spanning over almost two decades of daily observations at a resolution of up to 250 m provides a good resource for deriving long-term ocean kinematics from sea ice dynamics. In this paper, we present a newly developed automatic algorithm to retrieve dynamic measurements of sea ice from these images. We describe the methodology by presenting results acquired along the East Greenland Current (ECG) for 6.5 weeks in the spring of 2017. During this period, our ice floe tracker was used to identify and track ice floes with length scales ranging from 8 to 65 km. By effectively filtering atmospheric conditions from MODIS images, ice floes were tracked for up to ten consecutive days, and a total of 1061 trajectories were retrieved. A southward mean sea ice flow associated with the ECG was observed along with deviations in both direction and magnitude, suggesting the effect of an underlying turbulent eddy field. The absolute position and tracking errors associated with our method are 255 m and 0.65 cm/s, respectively, each derived from a comparison between manually and automatically identified ice floes. Going forward, our methodology will be employed to process longer time sequences to analyze nonlinear interactions between drifting ice floes and the upper ocean turbulent eddy field in the ECG as well as to investigate other prominent regions of the Arctic Ocean. [Display omitted] •Processing of MODIS imagery by adaptively filtering atmospheric conditions•Sea ice tracking via automatic matching based on surface features•Estimation of sea ice velocity fields calculated from retrieved trajectories