The powerful earthquakes occurred in Kumamoto at the highest magnitude 7.3 hitting this prefecture on April 16, 2016 at 01.25 A.M. local time. This study proposed a method to utilize Sentinel-1A images to detect earthquake damaged areas. There were two images being used.One was before-earthquake images (on March 3, 2016 and March, 27 2016) and another was an after-earthquake image (on April 20, 2016). The method operated on interferometric coherence of sentinel-1A image. In order to estimate the damaged areas, the coherence change between the before-earthquake and during-earthquake pairs were processed. The damage map was produced by concerning into different classes comprised of the urban damaged areas, none and less damaged of the urban areas, and non-urban areas. This approach was compared with the ground-truth data, which provides high overall accuracy at 88% (kappa coefficient is 0.82). 24 Consequently, Sentinel-1A could contribute the insightful geospatial information of the earthquake situation and support the disaster management.
KeywordsDamage Assessment, Sentinel-1A, Interferometric Coherence, Kumamoto Earthquake * 1 1 * 2 2 * Sentinel-1A could be appropriately applied in the earthquake damage detection. Also, the interferometric technique was efficient enough for detecting the earthquake displacement, and the damage assessment. The interferogram shows the phase change induced by earthquake from satellite looking dimension while the coherence analysis result presents the possibility of the large damaged areas without the interference of cloud coverage, which still performs the high accuracy. Moreover, the Sentinel-1A could provide the accelerated observation within 5 days after the disaster occurred. However, the further discussions are still concerned for the future study. The result was included some effects by the processing method and atmospheric conditions because Sentinel-1A provides C-band information. The wavelength is short about 5-6 centimeters, could have more effects than the longer wavelength. The spatial resolution was roughly generated with the default resolution at 100 meters operated by the software. It was rougher than using the higher resolution such as ALOS-2, which was investigated on the earthquake damage assessment (Tamkuan & Nagai, 2017). However, the spatial resolution depends on the SAR observation modes. For damage assessment, our study considered only the damaged, and none to less damaged areas. The multi-level of damages will be considered in further study. The spatial resolution is important for the damage classification. The density of urban and damaged areas could lead to the quality of the damage map. Some damaged buildings with 100x100 pixel of non-damaged areas may unfortunately be missing information.This point should be focused for next study to improve the accuracy for the earthquake detection.
ConclusionThis study proposed the method to detect the earthquake damaged areas using multitemporal Sentinel-1A images. The damaged areas could be rapidly detected i...