Identification of Damaged areas Due to the 1995 Hyogoken-Nanbu Earthquake Using Satellite Optical Images
Masashi Matsuoka and Fumio Yamazaki
Earthquake Disaster Mitigation Research Center, RIKEN
2465-1 Mikiyama, Miki, Hyogo 673-0433, Japan.
Tel: (81)-794-83-6632 , Fax : (81)-794-83-6695
E-mail : mastuoka@miki.riken.go.jp
Abstract
Several earth observation satellites observed the Kobe area before and after the 1995 Hyogoken-Nambu, Japan Earthquake. Since a part of the damage survey results of this earthquake is maintained as GIS data, a quantitative analysis of the surface change in damage areas is possible. Spectral characteristics of the area damaged by the earthquake were investigated using LANDSAT and SPOT satellite optical images taken before and after the earthquake for examination the possibility of extracting the earthquake damaged distribution by satellite remote sensing. The damage distribution extract by discriminant analysis of the spectral characteristics agreed with the results of an actual damaged survey.
Introduction
Satellite remote sensing, which easily monitors a large area, could provide effective information at the time of recovery activity, e.g., forming a restoration plan, if it is able to know the distribution of the damage due to disasters at an early stage. Several satellites observed the Kobe area before and after the 1995 HYOGOKEN_Nanbu Earthquake (Sudo et al., 1995). Multispectral characteristics were different between images of the liquefied areas and burned areas taken by airborne remote sensing just after the earthquake occurred (Mitomi and Takeuchi, 1995). A study suggested the possibility of interpreting the damaged area based on the spectral pattern changes between satellite image taken before and after the earthquake (Inanaga et al., 1995; Yoshie and Tsu, 1995). However, no quantitative approach was found for examining the possibility of identifying the damage distribution from the relationship between the spectrum characteristics of the damaged area by satellite images and detailed survey results.
Since part of the damaged survey results of this earthquake was maintained as GIS data, a quantitative analysis of the surface in the damaged area is possible. In this study, the spectral characteristics of the damaged due to this earthquake were investigated using LANDSAT and SPOT satellite image taken before and after the earthquake. Then the relationships between damaged distribution extracted through discriminant analysis of satellite images and survey data were evaluated.
Data
Earthquake Damage Survey Data: Liquefaction and building damage were focused on as earthquake damaged in this study. Boiled sand deposits due to the earthquake are digitized on the 1/50,000 scale ground failure survey map (Hamada et al., 1995) and used as liquefied area data. The building damaged data based on detailed survey results compiled by AIJ (the Architectural Institute of Japan) and CPIJ (the City Planning Institute of Japan), and digitized by BRI (Building Research Institute, Ministry of Construction) were utilized as GIS data. In the GIS data, the building damage level was classified into the five categories: damage by fire, severe structural damage, moderate damage, slight damage and no damage, and the numbers of damaged building were totaled for each block in cities (Building Research Institute, 1996).
Satellite Image: The SPOOT/HRV and LANDSAT/TM observed the area of interest on January 20, and January 24, 1995, respectively. We used the images taken on March 22, 1990 by SPOT and on August 17,1994 by LANDSAT for data before the earthquake, and aimed to examine the change in spectrum characteristics for the damaged area. The region of the satellite images is shown in Fig.1.
Figure.1 : Area of this study
Correction of Images: Because the digitized values in the satellite images were different depending on the observation situation and the surface conditions, the digital number (DN) correction was required before starting this study. First of all, clouds and cloud shadows covering the area were removed by using proper threshold values of several spectral bands. Areas of vegetation were also excluded from
The images using NDVI (Normalized Difference Vegetation Index) because the reflection characteristics of the leaves of plants differ seasonally.
For DN correction, we selected pixels in nonliquefied and nondamaged areas from these images, using the liquefaction data and building damage data. Here, the selected pixels were considered to be the areas in which there was no change of the surface after the earthquake. Furthermore, the DN in the images taken before the earthquake for all areas were corrected so that the histogram of the selected pixel DN in this image fits that in the image taken after the earthquake.
