Author(s): Hajime Shirozu; Koji Asai

Linked Author(s): Hajime Shirozu

Keywords: SAR; Satellite imagery; Remote sensing

Abstract: In July 2018, the heavy rainfall in western Japan caused severe damage to Mabi Town, Kurashiki City, Okayama Prefecture. A large area of 12 km2 was flooded in a short time after collapse of the embankment of Odagawa River. While there is concern about the intensification of heavy rainfall disasters, it is required to grasp the situation of inundation quickly and widely. This is expected to use earth observation satellites that equip with Synthetic Aperture Radar (SAR), and the implementation of the emergency observation immediately after the disaster and the development of base map are in progress. In this research, we propose a method to estimate the inundation depth easily by detecting the flood zone automatically using SAR satellite data of emergency observation and base-map data of the same season in the near past. The analysis was performed on the area surrounded by the west bank of the Takahashigawa River and the north bank of the Odagawa River. We used ALOS-2/PALSAR-2 satellite images observed on July 7 and April 14, 2018, which were provided by JAXA in the framework of SENTINEL ASIA, an international joint disaster prevention project, immediately after the disaster. It is recorded as a number and represented on the screen as a luminance distribution. At the water surface, microwaves are reflected specularly and their backscattering is weakened, so they appear darker. On the other hand, it appears brighter on the rough land and structures due to the stronger backscattering to the satellite. This feature can be used to identify the inundation area in the area. Although there is room for improvement in estimation accuracy due to layover and resolution limitations, it seems to be an effective measure for emergency disaster response. Seasonal conditions are important factors in finding and identifying flooded areas. For this reason, it is desirable to have multiple pre-disaster satellite images as a base map in the analysis of the flooded area. In this study, the actual working time was about 12 hours, including trial and error on the threshold setting. Other processing times, except for manual threshold setting, can be done within one hour. In practice, the goal is to obtain data within 30 minutes after obtaining them, so automation of the operation is necessary in the future.

DOI: https://doi.org/10.3850/IAHR-39WC252171192022302

Year: 2022