Case Studies: Enhancing Sea-Ice Monitoring with RADARSAT - 2

Through the Earth Observation Application Development Program (EOADP) the Canadian Space Agency (CSA) awarded a project to MacDonald Dettwiler (MDA) to investigate the use of multi-polarimetric synthetic aperture radar (SAR) products for operational sea- ice monitoring. The key objectives of this project were to:

• identify how multi-polarimetric data from RADARSAT-2 can be used to enhance Canada's operational capability for sea-ice monitoring
• specify multi-polarimetric derived products for sea-ice monitoring
• identify operational improvements to the RADARSAT-2 mission enhancing delivery of sea-ice information

Polarimetric Data Analysis

Data from the Shuttle Imaging Radar (SIR) was used for the project; specifically C-band imagery of the west coast of Newfoundland, acquired April 18, 1994. The SIR data was acquired in fully- polarimetric mode, which permitted multi-polarimetric and polarimetry analysis. A three-channel merge (HH+VV+HV) shows the presence of sea ice, open water, and land (Figure 1). Variability in sea-ice types is evident by the different colours and tones in the image.

Single channel data (HH and HV) is shown in Figures 2 and 3. The HH polarization is the same as that currently available from RADARSAT-1, and the HV polarization will be available from RADARSAT- 2. Inspection of the image shows that the cross-pol channel (HV) provides good discrimination between the ice-edge and the sea, and the co-pol channel (HH) provides good ice-type information. Therefore, when acquiring data in the RADARSAT-2 Selective Polarization mode (i.e. acquisition of one co-pol and one cross-pol channel), the pairing of HH and HV polarization should provide optimal information for sea ice applications. In addition, the channel ratios also provide valuable information. For example, the ratio of HV:HH provides a relative measure of volume scattering (HV) versus surface scattering (HH); the different scattering mechanisms provide input for classification schemes.

In addition to Selective Polarization modes, RADARSAT-2 offers Polarimetry modes, or quad-pol. When data is acquired in quad-pol mode, both the amplitude and phase information are retained; the use of this information provides input for classification algorithms. For this project, the H/A/ minimum distance classification algorithm based on Maximum Likelihood was implemented. The H term in the H/A/ is the entropy, the A term is anistropy and the term is the alpha angle. (H) is a measure of the dominance of a single scattering mechanism (H=0) or a random mixture of scattering mechanisms (H=1). For example, scattering from a relatively smooth surface exhibits an H- value close to 0. The H/A/ minimum distance classification uses an iterative approach; the classification results after five iterations are shown in Figure 4. Although operationally limited by swath width (25 km), the RADARSAT-2 polarimetry mode will provide improved ice information, and will be applicable in areas where the maximum information over a reduced spatial area is required, e.g. key shipping lanes.

Operational Improvements

Three general categories of operational improvements flow to and from User Needs. These categories include Space Systems, Ground Systems, and Information Extraction. The Space System deals with the satellite and includes components such as the orbit and payload. The Ground System deals with data after downlinking and includes data processing throughputs and image quality. Finally, the Information Extraction category deals with the use of SAR data, specifically polarimetric data, and the software tools optimized for ice applications and information delivery.

In terms of the Space System, the re-configurable capability of RADARSAT-2 allows the design of new imaging modes either before or after launch. In particular, and based on user needs, new polarimetric modes providing increased swath width are viable (there are other trade-offs if the swath width is increased). Ground System issues, image quality for example, will be improved. In particular, scalloping and banding will be significantly reduced, and Automatic Gain Control effects significantly reduced due to the use of Block Adaptive Quantization techniques on RADARSAT-2. Finally, algorithms, designed to enhance Information Extraction are under development. These algorithms will provide the platform for the development of sea-ice information products, and add to the commercial value of RADARSAT-2 data and information.