Abstract
Building damage detection technique which we had developed, has been successfully applied to past earthquakes such as the 1995 Kobe, the 2001 India, and the 2003 Bam, by using the compound index, z-value, which is the value derived from the correlation and difference in intensities between pre- and post-event SAR images. This technique was applied to the affected areas due to the 2004 Niigata-ken Chuetsu earthquake on October 23, 2004, by using one pair of Radarsat images taken before and after the earthquake. However, it was not possible to identify any significant distribution of damaged buildings. In this study, we examined the reason and proposed a new technique by using two pairs (pre-seismic and co-seismic) of SAR images, to identify smaller building damage ratios in less densely built-up areas compared to previous techniques. The main idea is to minimize the effect of signal noise and tempral changes of the earth's surface, on building damage estimation, by calculating the difference values from the two pre-event images and one post-event image. In a macroscopic point of view, the distributions of the both deference values of z-value and correlation coefficient in built-up areas were in good agreement with damage by survey reports. In Yamakoshi village, located in highland, we also could identify large-scale landslides with accuracy as good as interpretation from aerial photos.
