Investigation of hybrid polarimetric features for tsunami-induced damage assessment of urban areas

Abstract

ABSTRACTIn this study, a polarimetric scattering investigation was performed by using hybrid polarimetric (HP) synthetic aperture radar (SAR) data to determine the optimum HP feature for urban damage assessment related to the Tohoku earthquake and subsequent tsunami disaster on 11 March 2011. To minimize the effect of uncertainty on the feature selection techniques, quantitative analysis of HP features derived from Advanced Land Observing Satellite (ALOS)/Phase Array type L-band SAR (PALSAR) data was conducted. First, preliminary selection was performed based on the Jeffries–Matusita (JM) distance, which was applied to evaluate the separability between damaged and undamaged areas. Second, random forest (RF) was implemented to improve the reliability and determine the most relevant feature based on importance score analysis. Accordingly, comparative analysis was performed and compared with quad polarimetric (QP) SAR data in terms of urban damage levels. The experimental results demonstrate that the circular polarization ratio is the most informative feature for the identification of urban damage and can be used to discriminate different damage levels, even in areas with large polarization orientation angle (POA).