Urban Flood Detection using SAR Interferometric Coherence of Multidirectional Observations

Abstract:
This paper reports a highly accurate inundation area estimation method using interferometric synthetic aperture radar (SAR) coherence data observed from multiple directions. In recent years, it has become possible to observe and perform interferometric analysis for an identical area from different orbits due to the increase in the number of satellites and the improvements in maneuvering technology. We considered that we can take both high spatial resolution and detection accuracy for flood detection by combining those interferometric pairs. We applied a method with a support vector machine (SVM) to the data from the 2015 heavy rainfall disaster in Kanto and Tohoku area, Japan, and found a separating hyperplane which can distinguish flooded region more accurately than a single orbit case.