PRELIMINARY EVALUATION OF RADARSAT-2 POLARIMETRIC SAR DATA FOR FOREST APPLICATIONS

摘要

Radarsat-2 polarimetric SAR data for Natural Resources Canada became available in May, 2008. In this paper we evaluate analysis methods and investigate potential forest applications with Radarsat-2 polarimetric SAR data. The Radarsat-2 Fine Quad-polarization data were studied and the Cloude H/A/Alpha and Freeman three-component decompositions were performed. We found that Radarsat-2 FQ data has a Faraday rotation angle estimated at less than 0.01 degree. The georeferencing information embedded in the FQ metadata file provided good pixel positioning accuracy. However, polarization orientation shifts introduced by terrain topography slopes in azimuth direction were difficult to be sensed. Various polarimetric parameterizations of Radarsat-2 FQ data are useful for identifying different landcover types and detecting changes in forested areas. The Cloude entropy, dominant scattering alpha and eigenvalues (λ_2+λ_3) were used to create a HSV coding image for better differentiation of water surface, vegetation and forest and to reduce topographic effects. Multitemporal time series of the Freeman decompositions showed that the volume and double bounce scattering were as expected. The surface scattering was seen to be stronger in winter than in summer.