Vegetation Identification and Change Detection with Radarsat-2 Polarimetric Imagery
(1) Royal Military College of Canada, PO Box 17000 Stn Forces, Kingston, ON K7K 7B4, Canada
There is a long history of vegetation identification and monitoring studies using polarimetric scatterometers and airborne fully polarimetric synthetic aperture radars. The recent launches of ALOS/Palsar, Terrasar-X and Radarsat 2 allow us to extend these analyses into near-earth orbit. In comparison to ground-based or airborne platforms, spaceborne polarimetric SARs have the advantage of larger image areas at more constant incidence angles, and consistent orbital paths to allow the collection of operational time series over selected sites. Their principal disadvantages are steeper incidence angles, fixed revisit times and higher noise floors.
Since June 2008, Radarsat 2 fine quad-pol imagery has been collected over Canadian Forces Base Shilo, a large, relatively natural area in Southern Manitoba, Canada, containing region of tall grass prairie, deciduous and coniferous forests, wetlands and sand dunes. Polarimetric analyses of these images were carried out with PolSARPro. ENVI/IDL was used for all other analyses. The distribution of dominant scattering mechanisms throughout the region is entirely consistent with the known vegetation patterns established through ground truth and Landsat imagery. Both Wishart H/A/alpha and Freeman classifications are well correlated with optically derived image classifications. The imagery is being collected principally as a 24 day repeat sequence at an incidence angle of 38 degrees. Temporal changes in the dominant scattering type and in other polarimetric parameters are consistent with the annual life cycle of vegetation in the region, especially in the wetlands, where the changes are most pronounced,
A second set of imagery has been acquired over CFB Petawawa and the Petawawa Research Forest primarily to explore the effects of incidence angle on polarimetric analysis. These images, collected in association with the Canadian Forest Service, show that there is a significant change in the values of some polarimetric parameters associated with changes in incidence angle throughout the range of 22 to 38 degrees. Hence it appears necessary to maintain a specific imaging geometry for the creation of meaningful time series of polarimetric imagery.