Validity Study of EDES Application to Taiwan Chi-Chi Earthquake Disaster
Figure 2 Post-event VNIR image (September 23, 1999)
Figure 3 Pre-event TIR image (July 24, 1999)
Figure 4 Post-event TIR image (September 23, 1999)
3. Damaged area Estimation using Radiance Calibrated Image
The relationship between the OLS VNIR band gain G (dB), digital numbers DV, and observed radiances R (W/cm 2 /sr/?m) derived from the preflight sensor calibration are depicted by Elvidge et al. (1999). The following equations are derived from the relationship graph:
Equations (1)
Equations (2)
Equation (2) shows the relation between RCI DN DR and radiances R (W/cm 2 /sr/µm). Kohiyama et al. (2000) proposed a method to evaluate a PreI using SLI, but the estimation result corresponds only qualitatively to the result based on an observed PreI. We replicate the method using RCI instead of SLI. The relation DV-DR can be derived from equations (1) and (2):
Equations (3)
We conducted regression analysis to evaluate the DV-DR function using the same sample dataset that was selected in the function evaluation using SLI. The regression curve DV = 0.171DR 1.4 has the correlation factor 0.749, which is greater than that of regression curve using SLI, 0.631. The estimated damaged areas based on the observed PreI, SLI and RCI are shown in figures 5, 6 and 7, respectively. In the figures, red and yellow areas indicate exceeding 97.5 % and 85 % probability points of the normal distribution derived from histograms of DV differences between PreI and PostI. Gray areas show VNIR band saturation in PostI. Estimated results using RCI resembles to that based on observed PreI better than results using SLI especially around Koahsiung.
Figure 5 Estimated damaged areas based on observed PreI
Figure 6 Estimated damaged areas based on SLI
Figure 7 Estimated damaged areas based on RCI
4. Electric Load Data Before and After Chi-Chi Earthquake
We obtained electric load (EL) data in Taiwan on July 23 and from September 21 to 23, 1999. The date of pre-event data is different from that of satellite imagery (July 24, 1999), but the difference is enough short to be assumed negligible. The time series data of three aggregation areas are provided as post-event data. As for pre-event data, the time series data of the total system load and peak loads of three aggregation areas are provided. satellite-crossing-time load of each area based on the peak load ratio of each area. Figures 8 and 9 show the EL data before and after the earthquake. The estimated loads are plotted in Figure 8. The aggregation areas of pre-event data
shown in figure 10 are different from those of post-event data shown in figure 11.
Figure 8 Electric load data on July 23, 1999
