Validity Study of EDES Application to Taiwan Chi-Chi Earthquake Disaster
Masayuki Kohiyama, Haruo Hayashi, Norio Maki, Shin Hashitera
Earthquake Disaster Mitigation Research Center, RIKEN
2465-1 Mikiyama, Miki, Hyogo, 673-0433, Japan
Tel: (81)-794-83-6623 Fax: (81)-794-83-6695
E-mail: kohiyama@miki.riken.go.jp
Keywords
Damaged Area Estimation, DMSP/OLS, Radiance Calibration, Taiwan Chi-Chi Earthquake Disaster, Electric Load
Abstract
The authors are developing the Early Damaged Area Estimation System (EDES),
which provides the information as to the estimated impacted areas after any significant earthquake based on nighttime lights observed by the Defense Meteorological Satellite Program (DMSP) Operational Linescan System (OLS). The estimation method is based on the detection of significant reductions or loss of lights in nighttime images following the event, for it can be expected that city lights will observably decrease after a large earthquake. In the 1999 Taiwan Chi-Chi Earthquake Disaster, there was a countrywide power cut due to the damage of the transmission facilities in Chungliao and Tienlun. The post-event image seems to reflect this influence in appearance compared with a pre-event image, but the further investigation is needed because the gain setting of OLS changes along with lunar phase and elevation, and visible-near infrared (VNIR) band imagery only provides relative light intensity. This paper presents a new method to estimate damaged areas based on a radiance-calibrated image and a method to estimate gain and radiance of observed VNIR imagery, and the validity of the estimation result of EDES application to the Chi-Chi Earthquake is discussed using the electric load data before and after the earthquake.
1. Introduction
In order to support emergency response after an earthquake, we are developing the Early Damaged Area Estimation System (EDES), which disseminates spatial information of estimated damaged areas using Defense Meteorological Satellite Program (DMSP) Operational Linescan System (OLS) nighttime imagery (Hayashi et al., 2000). The estimation method is basically based on significant test of lights decreases more than normal fluctuation. But the sensor gain of OLS is controlled to generate consistent imagery of clouds at all scan angles for air navigation of United States Air Force, and the gain values are not open to the public. OLS visible-near infrared (VNIR) images are observed with different gain values in the different moon situations, and their digital numbers (DNs) only provide relative light intensities in the same observed scene. For all that, it can be assumed that the gain setting is almost the same in a similar moon situation, and we proposed image selection criteria to compare a pre-event image (PreI) and post-event image (PostI) (Kohiyama et al., 2000).
Elvidge et al. (1997) created the stable lights image (SLI), which is a map of VNIR emission sources based on cloud-free observations from 1994 to 1995 under low lunar illumination conditions. We improved the damaged area estimation method using artificial PreI derived from SLI in order to prevent cloud influence in PreI (Kohiyama et al., 2000). Elvidge et al. (1999) produced the radiance calibrated nighttime lights image (RCI), which was assembled observation data of three levels of gain setting from 1996 to 1997. It is revealed that calculated radiance from RCI is highly correlated to population and electric power consumption of states in the United States because of broader dynamic range and higher expressive power of urban areas than SLI, which has saturated data in many urban areas. In this paper, we propose a new method of damaged area estimation using artificial PreI derived from RCI and a method to estimate gain and radiance of observed VNIR imagery, and verify the estimation result based on the relation of electric load and radiance of nighttime imagery before and after the Chi-Chi, Taiwan Earthquake of September 21, 1999.
2. Nighttime Images before and after Chi-Chi Earthquake
Heavy typhoon clouds covered Taiwan Island around the earthquake onset day, and we chose the image on September 23, 1999 (lunar age: 13.6), which has little cloud influence, to detect lights reductions. As a-month-before images all have cloud contamination, the July-24-1999 (lunar age: 11.4) image was selected as a PreI. DMSP passed over the center of Taiwan (23.8 o N) at 20:47 on July 24 and at 20:09 on September 23, respectively. Fortunately, this 38- minute time lag makes lunar elevations of two images more similar. Figures 1 and 2 show the VNIR PreI and PostI ranging from 15 o N to 30 o N latitude and from 115 o E to 130 o E longitude. Thermal infrared (TIR) band images are shown in figures 3 and 4.
Figure 1 Pre-event VNIR image (July 24, 1999)
