Estimation of Snow Surface Dielectric Constant From Polarimetric SAR Data

Abstract

A novel methodology is proposed in this paper for the estimation of snow surface dielectric constant from polarimetric SAR (PolSAR) data. The dominant scattering-type magnitude proposed by Touzi et al. is used to characterize scattering mechanism over the snowpack. Two methods have been used to obtain the optimized degree polarization of a partially polarized wave: 1) the Touzi optimum degree of polarization given by Touzi et al. in 1992. The maximum $(p_{\rm max})$ and the minimum $(p_{\rm min})$ degree of polarizations are obtained along with the optimum transmitted polarizations $(\chi _{t}^{\rm{opt}},\psi _{t}^{\rm{opt}})$ . 2) The adaptive generalized unitary transformation-based optimum degree of polarization $m_{E}^{\rm{opt}}$ proposed by Bhattacharya et al. in 2015. This optimum degree of polarization is obtained either by a real or a complex unitary transformation of the 3 $\times$ 3 coherency matrix. These two degrees of polarizations are used and compared in this study as a criterion to select the maximum number of pixels with surface dominant scattering. These pixels were then used to invert the snow surface dielectric constant. It has been observed that the $m_{E}^{\rm{opt}}$ have increased the number of pixels for inversion by $\approx \text{9--10}\%$ compared to the original data. On the other hand, it was observed that the Touzi maximum degree of polarization $p_{\rm max}$ has increased the number of pixels for inversion by $\approx 2\%$ compared to that of $m_{E}^{\rm{opt}}$ . The proposed methodology is applied to Radarsat-2 PolSAR C-band datasets over the Indian Himalayan region. It is observed that the correlation coefficient between the measured and the estimated snow surface dielectric constant is 0.95 at 95% confidence interval with a root mean square error of 0.20.