Identification of Damaged areas Due to the 1995 Hyogoken-Nanbu Earthquake Using Satellite Optical Images
Characteristics of the DN in the Earthquake-Damaged Area
The pixels that represent the area of liquefaction, burning, heavy damaged, slightly damage and no damage were selected from the images to examine the characteristics of the DN in the earthquake-damage area. The liquefied class contained pixels included in the liquefaction data. The pixels for the burned class were selected from the city blocks where all buildings were determined to have the severe or moderate structural damage and slightly damage, and categorized into heavy and slightly damage classes, respectively.
The pixels used for DN correction were those in the nondamaged class of damaged class. The 300 and 500 pixels for each class of damage were taken to analyze from the LANDSAT images and the SPOT images, respectively.
Relationship between LANDSAT Images and Earthquake Damage: The characteristics of the DN in each classified damage area due to the earthquake are shown in Fig.2. The mean value of the DN and the standard deviation in the damage classes are shown in Table 1. The DN are larger for all bands, other than band 6, in the liquefied area in comparison with the nondamaged area. The DN of the burned area is smaller than that of the nondamaged area, but the DN of the heavy damaged area is similar to that of the nondamaged area. The mean value of the DN and the standard deviation in the damage classes before the earthquake are shown in Table 2,and the difference in the DN in classified damage area before and after the earthquake are shown in Fig. 3 and Table 3. like the liquefied area, the DN is high for all bands other than band 6. for the burned area, the DN in the blue light range of band 2 was especially low.
Figure.2 : Spectral features of LANDSAT/TM bands for damaged area after the earthquake.
Figure.3: Change in digital numbers of LANDSAT/TM bands for damaged areas before and after the earthquake
These results are in good agreement with the damage survey performed by airborne multispectral scanner remote sensing (Mitomi and Takeuchi, 1995). The DN in the heavy liquefied area, did not greatly differ from the values in the nondamaged area. However, the DN of band 1 in the blue light range was a somewhat high in comparison with the nondamaged area.
The DN in the liquefied area was high in the range from the visible to mid-infrared bands, because of the higher reflectance of sand than the surface of asphalt. Also, it is conceivable that the DN was raised in infrared bands in the heavy damage due to the exposure of soil under mud walls and roofing tiles upon the collapse of old wooden dwelling. The increase of reflection in the blue light band region is due to the roofs of the building being covered with the blue vinyl canvas sheets in the building damage areas.
| Classified damaged area |
Number of pixels |
Digital number: Average (sd.) |
| 1 |
2 |
3 |
4 |
5 |
6 |
7 |
| Liquefied |
300 |
74.5(6.6) |
31.9(4.5) |
40.3(7.6) |
32.2(6.5) |
48.9(12.4) |
97.6(2.0) |
22.7(6.6) |
| Burned |
300 |
66.0(3.9) |
25.3(2.5) |
30.2(4.0) |
23.1(3.7) |
33.6(6.1) |
98.9(1.0) |
17.8(3.5) |
| Heavy Damage |
300 |
71.8(5.9) |
27.5(3.1) |
32.1(4.8) |
26.4(4.6) |
35.7(9.7) |
98.1(1.5) |
18.8(8.7) |
| Slight Damage |
300 |
71.6(6.6) |
27.5(3.8) |
31.9(6.2) |
27.3(5.8) |
37.4(9.5) |
98.2(1.7) |
19.2(5.5) |
| No Damage |
300 |
69.8(5.6) |
27.0(3.6) |
31.4(5.9) |
26.5(5.6) |
36.8(9.7) |
98.5(1.9) |
19.1(5.6) |
Table 1: Average digital number corresponding to TM bands for the Classified damaged areas after the earthquake
| Classified damaged area |
Number of pixels |
Digital number: Average (sd.) |
| 1 |
2 |
3 |
4 |
5 |
6 |
7 |
| Liquefied |
300 |
71.0(6.8) |
28.5(5.4) |
34.5(9.2) |
27.2(8.3) |
39.0(15.5) |
100.0(1.7) |
19.5(6.6) |
| Burned |
300 |
69.1(4.0) |
26.1(2.5) |
29.9(3.9) |
23.1(3.5) |
31.2(5.7) |
99.2(0.5) |
17.5(3.1) |
| Heavy Damage |
300 |
68.6(4.9) |
25.9(3.0) |
29.5(4.9) |
23.6(4.6) |
31.6(7.2) |
99.0(0.9) |
16.7(4.2) |
| Slight Damage |
300 |
69.96(5.0) |
26.9(3.4) |
31.3(5.8) |
25.8(5.8) |
36.1(9.9) |
98.7(1.4) |
18.7(5.4) |
| No Damage |
300 |
69.6(5.0) |
26.9(3.2) |
31.3(5.3) |
26.5(5.2) |
36.9(8.8) |
98.5(1.8) |
19.1(5.0) |
Table 2: Average digital numbers corresponding to TM bands for the Classified damaged areas before the earthquake
| Classified damaged area |
Number of pixels |
Digital number: Average (sd.) |
| 1 |
2 |
3 |
4 |
5 |
6 |
7 |
| Liquefied |
300 |
3.5(7.6) |
3.5(5.4) |
5.8(9.0) |
5.0(8.3) |
9.9(14.5) |
-2.5(2.2) |
3.2(7.7) |
| Burned |
300 |
-3.1(4.2) |
-0.8(2.3) |
0.3(3.8) |
0.0(3.3) |
2.4(5.2) |
-0.3(1.1) |
0.3(3.5) |
| Heavy Damage |
300 |
3.0(4.9) |
1.6(2.8) |
2.5(4.5) |
2.8(4.4) |
4.1(9.2) |
-0.9(1.4) |
2.0(8.6) |
| Slight Damage |
300 |
2.0(6.3) |
0.5(3.4) |
0.6(5.1) |
1.5(5.1) |
1.3(8.5) |
-0.5(2.1) |
0.5(5.1) |
| No Damage |
300 |
0.2(5.8) |
0.1(3.4) |
0.2(5.3) |
0.0(5.0) |
-0.1(8.7) |
0.0(2.3) |
0.2(5.2) |
Table 3: Digital number corresponding to TM bands for the classified Damaged areas
Relationship between SPOT Images and Earthquake Damage: The DN after the earthquake and the DN difference before and after the earthquake in the classified damage areas are shown in Table 4 and Table 5, respectively, for SPOT images. The changes of the DN in earthquake damaged areas showed the same pattern as that in the LANDSAT images; the DN became high in the liquefied area. However the change of the DN in the burned or building damage area did not show such great difference compared with the nondamaged area. This coincides with results of previous research (Hosokawa and Zama, 1998). Because the sensor mechanism in LANDSAT is different from that in SPOT, there is the possibility that the sensitivities of the two sensors differ. Detailed examination will be necessary for further study.
