Mapping glacierized regions with quad-pol dual frequency LS-ASAR: insights for the NISAR mission

NISAR, a spaceborne L and S band radar mission, is set to launch in 2025. As part of mission testing, airborne L and S band radar sensors were employed to obtain dual-frequency quad-polarimetric (LS-ASAR) data in the United States. This study evaluates precursor NISAR data across the polarization channels of both L- and S-band radar frequencies for feature classification in glacierized regions. We use LS-ASAR data from July 2021, collected over Black Rapids Glacier and surrounding areas, to classify the region into snow, water, clean ice, debris-covered surface, rock, and vegetation. A Support Vector Machine classifier is applied to both backscatter-intensity and polarimetric decomposition channels. Polarimetric decomposition improves classification accuracy by 28%, highlighting the influence of surface, volume, and double-bounce scattering mechanisms on feature discrimination. Resampled data at 4 m and 8 m resolutions yield comparable results for intensity-based classification, while the accuracy of polarimetric decomposition based classification improves by 21.5% at 8 m. Combining all individual L- and S-band channels further enhances accuracy. Additionally, our analysis of polarization ratios, particularly the S to L intensity ratio in VH cross-polarization, helps in differentiating landslide-covered ice from the usual debris-covered ice. This study highlights the effectiveness of individual channels and the potential benefits of integrating dual-frequency L- and S-band polarimetric decomposition data to achieve higher accuracy in mapping glacierized regions. The NISAR data is expected to be publicly available, offering the opportunity to study glacierized regions globally at regular intervals, contributing to higher-level derived data products from NISAR.