Spaceborne SAR in Sea Ice Monitoring: Algorithm Development and Validation for the Baltic Sea

Licentiatavhandling, 2011

The usefulness of new spaceborne synthetic-aperture radar (SAR) sensors at L-, C-, and X-band for operational sea ice monitoring in the Baltic Sea has been evaluated. It is concluded that the information content in C-band and X-band data is largely equivalent, although larger spatial coverage is an advantage of the current C-band satellites. L-band data provide complementary information and is especially useful to distinguish ice ridges, shear zones, and other deformation features. It is also concluded that cross-polarized data adds to the interpretation for wind conditions that make separation of open water and sea ice difficult in co-polarized data. A sea ice concentration algorithm has been developed that processes wide swath SAR images to estimate the ice concentration. The method is based on spatial autocorrelation of the images, and makes use of a neural network which is trained against operational sea ice charts produced by the Swedish Ice Service at the Swedish Meteorological and Hydrological Institute (SMHI). The algorithm is able to classify pixels as either open water or sea ice with an accuracy of about 90 percent. The sea ice concentration is determined with a root-meansquare error of less than 7 percentage points for a uniform distribution of sea ice concentrations after spatial averaging. A sea ice drift algorithm published by M. Thomas in 2008 has also been evaluated. The ice drift is computed from SAR images over the Bay of Bothnia and the result was compared with wind data from meteorological stations along the Swedish coastline. The evaluation was made with HH-polarized C-band data from the satellites ENVISAT and RADARSAT-2. The direction of the ice drift was in agreement with the flow direction of the wind. An improvement of the algorithm was suggested that will make tracking more robust in the vicinity of the shoreline, where the motionless solution often is favoured. A field campaign has been conducted in the Bay of Bothnia in spring 2010. An ice buoy was deployed in central Bay of Bothnia and its position was tracked. The drift data were compared with results from a further developed version of the sea ice drift algorithm using SAR images from ENVISAT. While still lying on the ice, the buoy successfully measured in situ ice drift between eleven image pairs. Thermometers mounted on the buoy provided basic data that made it possible to distinguish between water and ice underneath the buoy.