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

Lagrangian Data Assimilation and Parameter Estimation of an Idealized Sea Ice Discrete Element Model

“…The displacement l E x and the angular displacement l E  of the ice floes are often available from the high-resolution satellite imageries. See for example a recently developed floe tracking algorithm (Lopez-Acosta et al, 2019). These data can be used to recover the velocity l E v and the angular velocity l E  of the floes as well as the ocean flow field ˆo E u from the coupled system 19.…”

“…Our basic model formulation is motivated by the data assimilation task to be performed on relatively small timescales (days to weeks) in regions where floes could be unambiguously identified from freely available satellite data. This restricts out attention to marginal ice zones, specifically in summers when the ice is illuminated by the sun and the reflectance data is available but also because there we one could find sufficiently large floes from the breakup of winter sea ice such that they could be identified from satellite 10.1029/2021MS002513 4 of 27 imagery (Lopez-Acosta et al, 2019). Since the sea ice is not overly packed in marginal ice zones, the internal deformation of individual floes could be neglected and the main floe dynamics could be, at least partially, reduced to simplified collision rules that we describe below.…”

“…In addition, the ice trajectories determined from satellite tracking can be accommodated in a natural way for data assimilation using the DEM models. In fact, sea ice tracking techniques via automatic matching based on satellite-derived surface features have been developed recently using reflectance images at a resolution of 250 m, providing the observations of the locations and angular displacements for floes larger than about 5 km (Lopez-Acosta et al, 2019). Using the higher-resolution Synthetic Aperture Radar data with 20-60 m pixel size (Kozlov et al, 2020), it might be possible to track floes an order of magnitude smaller, but such datasets have not yet been developed but are expected to appear in near future.…”

“…Apart from the atmospheric and ocean stresses, the floes are also interacting with each other via collision rules that are not perfectly known and contribute to model error. In addition, datasets of floe trajectories (e.g., Lopez-Acosta et al, 2019) are inherently sparse as they are obtained from sequential satellite images that are separated by hours to days and commonly affected by clouds and other types of instrumental noise leading to only a relatively small subset of floes being observed in any given domain of interest. It is these challenges that motivate our data assimilation objectives.…”

Partial Shape Recognition for Sea Ice Motion Retrieval in the Marginal Ice Zone from Sentinel-1 and Sentinel-2

“…There are two shape-based ice drift algorithms in the literature. One uses a suite of geometrical ice floe parameters (perimeter, major and minor axes, centroid position, and surface and convex area) to identify and track ice floes [15], and another uses the Euclidean distance of the ice floe centroid to the nearest boundary pixel in n directions as a descriptor for floe matching [16]. However, the shape of the same ice floe in two sequenced images may be altered by melting, breaking off, or joining/rafting with surrounding floes, in which case the sea ice motion retrieval requires recognition of partial shape similarity.…”