Global Shoreline Mapping from an Airborne Polarimetric SAR: Assessment for RADARSAT 2 Polarimetric Modes

A project between two agencies, DREO and NIMA was initiated in 1998 to study shoreline extraction from imagery. DREO researched SAR applications with an emphasis on Radarsat products. The focus of this report is the extraction of shorelines from polarimetric SAR imagery in preparation for the launch of Radarsat 2. The desired vector shoreline product for this project is the Mean High Water Line (MHWL). This specification implies that consideration of tidal models are required. For this project in addition to radar considerations, geomatics, topographic and oceanographic issues were considered part of the investigation. Polarimetric classification methods were used to detect the shoreline region. Contrast enhancement methods were implemented to emphasize the shoreline edge. The cross-polarized channels provided velocity information which was used for discriminating between non-stationary (i.e.; ocean) and stationary surfaces (i.e.; land). One method for extracting shore slopes was studied here to determine if it is feasible for C-band SAR data. Two experimental trials acquired polarimetric SAR image data for each study site at both the high and low tide extremes. In situ ground truthing was conducted while the image data was collected. The trial locations were at Fundy, NS, and the North Carolina (NC) coast. Comparisons were made with either UPS vector data collected at the waterline during the acquisition time, or a surveyed MHWL vector estimate. Phenomenological differences between the shore types were observed and are documented here. One method to extract shore slopes was studied and did not produce reasonable results for the North Carolina site and reasonable results for the NS site. Further work is required here. A mean error estimate between a waterline extracted from an image, acquired at low tide, and a surveyed MHWL was 6.3 m over a 2 km vector. Original contains color plates: All DTIC reproductions will be in black and white. Text in English; abstract in English and French.