Multi-Resolution SAR Data Analysis for Automated Retrieval of Sea Ice Parameters
Zakharov, Igor; Power, Desmond; Bobby, Pradeep; Randell, Charles
C-CORE, CANADA

Sea ice monitoring is an important field of scientific research and relevant for operational applications including marine transportation, search and rescue operations, and oil and gas exploration and development. The outputs of sea ice monitoring include ice edge detection, thickness, ice classification and ice statistics. Depending on the application, this service must be provided in Near-Real-Time. Environmental services of different countries typically produce ice charts with low temporal and spatial resolution. In this case, the detailed ice parameters can be extracted using sensors with different resolution to achieve required coverage and revisit using multiple Synthetic Aperture Radar (SAR) sensors.
Results acquired over the Laptev Sea are demonstrated at the symposium. Russian Federal Service for Hydrometeorology and Environmental Monitoring produces the ice charts (maps) of the Arctic ice and freezing seas from the satellite data (visible, infrared range, radar images), ships, and research stations. During summer the charts are prepared monthly with the spatial resolution at scale 1:5,000,000. For safety and operational efficiency of offshore infrastructures the detailed parameters, such as, ice edge, surface, floe size, and thickness are important to estimate more frequently. This work examines the suitability of multi-resolution SAR data (ENVISAT, TerraSAR-X, TanDEM-X) for ice parameters retrieval using automated algorithms within a system of multisource data fusion.
Time series between August 24, 2010 and August 28, 2010 of C-band Envisat ASAR images of Wide Swath mode (150m by 150m resolution) HH polarization were acquired. Followed by image preprocessing (geocoding and landmasking), the analyst, working with the developed prototype of multisource data fusion system, is able to identify the areas of interest is sea ice relying on ice backscatter coefficient and texture which depends on multiple SAR parameters (incidence angle, polarization) and can significantly vary depending on the season, ice type, and ice surface. The important ice parameters retrieved from ENVISAT data using automated algorithms are ice edge, ice concentration, and estimation of ice thickness. Ice edge is effectively detected using image processing techniques based on textural parameters in combination with morphological operators. Ice thickness for pack and land fast ice is estimated using empirical relationships between ice thickness and backscattering coefficient. Time series of ENVISAT data allows monitoring the dynamic changes in ice and in addition it may assist with planning high resolution data. However, the high level of ocean clutter and the presence of ocean features, such as, currents and eddies in ENVISAT data limit efficiency of the automated algorithms. The implementation of expert system to improve efficiency of the fusion system is discussed at the symposium.
The unique TerraSAR-X/TanDEM-X constellation operates in monostatic mode and allows along track interferometric measurements with short time difference 2.6 sec. To monitor the area under study with sea ice, three pairs of SAR X-band images of the TanDEM-X constellation acquired between August 16, 2010 and August, 27, 2010 were used for analysis of sea ice parameters. Each image of Strip Map mode HH polarization Single Look Complex with the resolution 1.2m [slant range] and 3.3m [azimuth] covers area 30 km by 50 km.
One of the important tasks for the usage of high resolution data is to detect and to estimate size and geometry of ice floes and possible icebergs. The interesting result was found by analysing the interferogram. The interferogram of sea ice depends on two factors: (i) ice thickness, related to the surface elevation (Garcia et.al., 2012), and (ii) ice dynamic (Sheiber et.al., 2012). As the result, each individual floe has its own fringes pattern. Utilising this pattern we can accurately detect individual ice floe edge, which is problematic relying only on magnitude or coherence images.
It is shown that iceberg in sea ice exhibit interferogram which can be easily detected, which is an important task for operational applications when icebergs are buried in sea ice (Power et. al., 2011). Interferometric data processing can take more than two hours to extract and analyse digital elevation model. In addition the rotational fringes caused by ice dynamics have to be removed before the phase unwrapping to estimate icebergs surface. For certain areas, ice thickness estimated using TerraSAR-X data shows a good correlation with thickness estimated using ENVISAT data.
It is demonstrated that the several SAR sensors with multiple frequencies and resolution leads to a better understanding of ice conditions, including floe sizes, ice edge, thickness, and other ice features, such as, icebergs. In addition, the low resolution sensor data can assist with planning data from high resolution sensors. The future tasks will be validation of the developed algorithms with the ground truths data from planned field campaigns.
The authors would like to acknowledge ESA and DLR for data providing, Susan Carter for data acquisition, and Research & Development Corporation (RDC) of Newfoundland and Labrador for the financial support of the Ignite Grant.

References
Garcia, J.A., Eyssartier, K., Lopez-Dekker, P., Prats, P., De Zan, F., Krieger, G., Busche, T., 2012. Monitoring the Petermann ice island with TanDEM-X. IEEE International Geoscience and Remote Sensing Symposium (IGARSS-2012), pp.1912-1915.
Power D., Bobby P., Howell C., Ralph F., and Randell C. 2011. State of the art in satellite surveillance of icebergs and sea ice. Offshore Technology Arctic Technology Conference, February 2011, Houston, Texas, USA.
Scheiber, R., De Zan, F., Prats, P., Araujo, L.S., Kunemund, M., Marotti, L. 2011. Interferometric sea ice mapping with TanDEM-X: First experiments. IEEE International Geoscience and Remote Sensing Symposium (IGARSS-2011), pp.3594-3597, 24-29.