Polarimetric Properties of Burned Forest Areas at C- and L-band
Tanase, Mihai1; Santoro, Maurizio2; Aponte, Cristina1; de la Riva, Juan3
1The University of Melbourne, AUSTRALIA; 2Gamma Remote Sensing AG, SWITZERLAND; 3The University of Zaragoza, SPAIN
Fully polarimetric C- and L-band synthetic aperture radar (SAR) data have being investigated to determine the relationship between polarimetric target decomposition components and forest burn severity over two sites located in a Mediterranean environment. The dependence of the polarimetric decomposition metrics on SAR acquisition geometry and environmental conditions was also analysed at C-band. Multiple linear regression models with interactions were used to quantify burn severity retrieval accuracy. For steep SAR acquisition geometries C-band polarimetric components related to surface scattering mechanisms showed increased sensitivity to burn severity levels while for datasets acquired with more grazing geometries the polarimetric components related to volume scattering and dihedral scattering mechanisms were more correlated with burn severity levels. At L-band only volume and dihedral scattering related decomposition components provided significant relationships with burn severity levels. Relatively low burn severity estimation errors were obtained for all datasets with L-band data presenting the highest sensitivity to fire effects. Using a single regression model provided sufficient accuracy for burn severity estimation when taking into account the local incidence angle and its interactions with the SAR metrics. The use of fully polarimetric data improved the estimation accuracy of forest burn severity with respect to backscatter intensities by only a small margin.