IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing (Jan 2018)
Insights Into Polarimetric Processing for Wetlands From Backscatter Modeling and Multi-Incidence Radarsat-2 Data
Abstract
We have observed unexpected results using the Freeman-Durden (FD) and other polarimetric decompositions in Radarsat-2 quad-pol data from many swamps in Eastern Ontario. In particular, the decompositions reported minimal backscatter from the double-bounce mechanism in a situation where there was compelling evidence that double-bounce backscatter contributed substantially to the return. This led to a hypothesis that the FD and similar models give erroneous results because of the physics of Fresnel reflection of wood, a lossy dielectric material, that makes up the vertical reflecting surfaces in swamps. We found some support for this hypothesis in the literature, and now report on an extensive theoretical and observational investigation. This work has shown that the Freeman-Durden decomposition, and other decompositions that use the same logic, will often mistake double-bounce backscatter as single-bounce backscatter in wetlands. This is a consequence of the fundamental physics of Fresnel reflection. It is important for users to be aware of this pitfall. Double-bounce backscatter from natural surfaces can be identified without recourse to polarimetric decomposition. The simplest, and most reliable, indicator of double-bounce backscatter is a high return in HH polarization. Double-bounce backscatter will generally produce higher return in HH than any other scattering mechanism. If both HH and VV polarizations are available, a high HH/VV intensity ratio is also a strong indicator of double-bounce backscatter. Additional modeling efforts are expected to provide further insights that can lead to improved applications of polarimetric data.
Keywords