Classification of Arctic Coastal Land Covers with Polarimetric SAR Data

Increased rates of permafrost degradation and erosion along Arctic coastlines are anticipated, as a result of changing climatic conditions. To characterize and quantify rates of change, baselines in terms of the land cover types present and the physical location of the land-water interface, need to be established. Earth observation data provides a logistical alternative to conventional mapping approaches, while Synthetic Aperture Radar (SAR) systems are particularly well suited to this application since images can be acquired through cloud cover and in the low lighting conditions observed during Arctic winters. The focus of this study will be to assess the potential of dual pol TerraSAR-X and quad-pol Radarsat-2 imagery for land cover mapping, using complimentary datasets acquired over two study areas in the southern Beaufort Sea, Northwest Territories, Canada. Novel polarimetric SAR processing techniques, including the extraction of Kennaugh matrix elements, have been used to generate inputs to the Maximum Likelihood Classifier. Results show good potential for general land cover mapping, and that overall accuracy increases when backscatter coefficients from both sensors are utilized (81.4% for eight classes), compared to TerraSAR-X HH/VV (68.9%) or Radarsat-2 HH/HV/VV (73.6%) alone. Similar accuracies were achieved using Kennaugh matrix elements, but slight decreases were observed when backscatter coefficients and Kennaugh matrix elements were combined.