Scattering characteristics of X-, C- and L-Band PolSAR data examined for the tundra environment of the Tuktoyaktuk Peninsula, Canada

• Published in: Applied Sciences
• Main Authors: Ullmann, Tobias, Banks, Sarah N., Schmitt, Andreas, Jagdhuber, Thomas
• Format: Article in Journal/Newspaper
• Language: English
• Published: 2017
• Subjects: ddc:526; Canada; Tuktoyaktuk; Tuktoyaktuk Peninsula; Tundra
• Online Access: https://opus.bibliothek.uni-wuerzburg.de/frontdoor/index/index/docId/15836; https://nbn-resolving.org/urn:nbn:de:bvb:20-opus-158362; https://doi.org/10.3390/app7060595; https://opus.bibliothek.uni-wuerzburg.de/files/15836/Ullmann_Applied_Sciences.pdf

In this study, polarimetric Synthetic Aperture Radar (PolSAR) data at X-, C- and L-Bands, acquired by the satellites: TerraSAR-X (2011), Radarsat-2 (2011), ALOS (2010) and ALOS-2 (2016), were used to characterize the tundra land cover of a test site located close to the town of Tuktoyaktuk, NWT, Canada. Using available in situ ground data collected in 2010 and 2012, we investigate PolSAR scattering characteristics of common tundra land cover classes at X-, C- and L-Bands. Several decomposition features of quad-, co-, and cross-polarized data were compared, the correlation between them was investigated, and the class separability offered by their different feature spaces was analyzed. Certain PolSAR features at each wavelength were sensitive to the land cover and exhibited distinct scattering characteristics. Use of shorter wavelength imagery (X and C) was beneficial for the characterization of wetland and tundra vegetation, while L-Band data highlighted differences of the bare ground classes better. The Kennaugh Matrix decomposition applied in this study provided a unified framework to store, process, and analyze all data consistently, and the matrix offered a favorable feature space for class separation. Of all elements of the quad-polarized Kennaugh Matrix, the intensity based elements K0, K1, K2, K3 and K4 were found to be most valuable for class discrimination. These elements contributed to better class separation as indicated by an increase of the separability metrics squared Jefferys Matusita Distance and Transformed Divergence. The increase in separability was up to 57% for Radarsat-2 and up to 18% for ALOS-2 data.