Flood Area Observation Using Quad-polarimetric SAR Data Based on Model-based Decomposition

Abstract

This paper proposes an effective usage of the model-based decomposition by considering the behavior of the Degree of Polarization (DoP) to improve the accuracy of observation using quad-polarimetric SAR data for flooded residential area. The nonnegative eigenvalue decomposition (NNED) is used as the decomposition method. For the volume scattering model in the decomposition, we utilize the Yamaguchi volume scattering model, in which three kinds of distributions are used in relation to physical distribution of the vegetation target under observation. Along with the decomposition, we also examine the variation of the DoP before and after a flood event. By considering the time variations of both the double bounce scattering and the DoP, damage situation of the flood around residential area can be estimated. We show some experimental results of ALOS-2/PALSAR-2 data including flooded residential area before and after a flood, to confirm the validity of the proposed usage of the decomposition and the DoP.