- Frequency and polarization study
- Multisource data analysis
- Integration and validation of new methodology
Frequency and polarization study
Previous studies of oil spill detection with multi-frequency SAR show that there is significant difference in the detection capability for C-band, X-band and L-band SAR. Studies in the Black Sea using ENVISAT ASAR images with dual polarization during field experiments with oil spills suggested that dual polarisation SAR can be beneficial for detecting slicks and discriminate real oil spills from lookalikes (Sandven et al., 2008).
The project will carry out case studies where SAR images with different wavelengths (mainly X-, and C-band) and polarisation modes are compared. This requires that images are obtained near simultaneously from RADARSAT-2 or ENVISAT and Terra-SAR or CosmoSkymed, in specific test areas where oil spills are assumed to be present. A controlled experiment with oil spills where all these SAR images are obtained during the experiment will be performed. It will also be useful to monitor areas with high ship traffic where the probability of finding oil spills is the best. A number of dual-polarisation and multi-frequency SAR images will be obtained and can be compared regarding oil spill detection, and the best operational SAR algorithm will be determined.
Multisource data analysis
Discriminating oil spills from lookalikes that have similar SAR signature is a complex task and as much prior information about the problem as possible should be utilized. The expected number of oil spills and lookalikes will depend on the wind level and sea state condition. The wind level can roughly be estimated from the SAR image, or be extracted from wind forecasts. Detection of ships or platforms can can be done directly in the SAR image (bright targets).
Our aim is to establish an operational methodology that combine the information extracted from the SAR image with other sources of information, including algae information in terms of optical imagery or recent algae forecasts, vessel identification, ocean-current information from other sources, oil spill statistics, and databases of major ship lanes, oil installations, pipelines, etc.
The spectral dimension of optical observation has been investigated and it is known that spectral signature of oil spill are observable in the range 300-500 nm (for instance by impacting Raman-line response). The present missions (MODIS/MERIS) allow observations at 412, 443 and 490 nm that may be utilized with respect for detection of larger algae blooms. We will investigate the optical reflection combined with the SAR backscatter in order to distinguish between oil spills, low wind areas and algae blooms more successfully than from SAR backscatter alone.
Integration and validation of new methodology
A very important part of determining the usefulness of an oil spill detection approach is to compare the performance of automatic detection to manual inspection of SAR images, and, furthermore, to compare satellite-based detection to aircraft verifications. By using SAR data that have been used in operational oil spill monitoring, including manual inspection of SAR images followed by immediate aircraft inspection, one has a de facto baseline for comparison with an automatic algorithm.
The benchmarking will focus on testing the combined algorithm, using all available sources of information as indicated above, including dual- or quad-polarisation SAR, algae information, ocean-current information, presence of ice (if relevant), and an oil spill statistics and context.