Abstract
With a warming climate, rising Arctic temperatures have reduced the summertime extent of Arctic sea ice, expanding the presence, persistence, and role of the marginal ice zone (MIZ), where ice is freer to drift faster and exhibit complex motion (translation and rotation). Conventional ice motion characterization schemes are geared toward retrieving translational motions outside of the MIZ, at a large-scale, and when they do address complex MIZ ice motion, they rely on motion tracking techniques which track features that are unstable with time (floe edges), especially within the MIZ. At high-resolutions (<100 m), sea ice deformation values derived from motion products characterized with such techniques may include unreasonably large values where leads (fractures in the ice) form. Future investigations in the high-resolution of sea ice dynamics would also benefit from improved collocation between sensor platforms for the purposes of validation across datasets.
A hybrid feature-tracking and cross-correlation sea ice motion characterization scheme is proposed and used to characterize ice motion across a series of TerraSAR-X (TSX) synthetic aperture radar (SAR) images which span a week in September 2014, and are coincident with an on-ice instrument cluster deployed during the Office of Naval Research’s (ONR) MIZ Field Season as a part of their five-year MIZ Departmental Research Initiative (DRI). The hybrid algorithm, coined SURF/MCC, uses a feature tracking algorithm, Speeded-Up Robust Features (SURF), to initiate user-required parameters in a maximum cross-correlation (MCC) to retrieve motion across subsequent satellite acquisitions. Deformation values are derived from SURF/MCC-characterized ice motion fields and analyzed for erroneous deformation due to miscorrelations. Finally, on-ice deployed radar reflectors and high-precision GPS units (SATICE) are used to assess the geolocational accuracy of the polar-orbiting satellite TSX’s three ephemeris products, and their accuracies are compared to their reported error.
It is found that the SURF/MCC motion characterization scheme works well to retrieve fine-scale, high-resolution ice motion in the upper-latitude MIZ, and that SURF is an efficient and accurate feature tracker which can be used where MIZ motions are too complex for MCC-characterization. It is also found that, even when initiated with parameters provided by SURF, the MCC tends to miscorrelate along areas of extreme deformation (leads), and so deformation values derived from MCC-characterized fields must be interpreted carefully; it is found that this phenomenon is acute at high resolutions when using the MCC. In addition, on-ice deployed radar reflectors are demonstrated as a cheap and low-tech method of improving collocation between SAR imagery and coincident in situ instrumentation. Finally, comparison of radar reflector SAR signatures coordinates in TSX imagery with SATICE GPS positions reveal varying degrees of accuracy as reported by the German Aerospace Center (DLR) for their three orbital products, but that the Science ephemeris in particular, in the Arctic, does not meet its reported accuracy (<0.5 m).