37 
n earthquake 
ce. 
im (see figure 
e image as is 
)d of only 35 
1 markedly in 
lg mountains, 
tands out as a 
coherence, 
coherence of 
igions in the 
learly are not 
; earthquake, 
interferogram 
[ une 11, 2003 
lg the time of 
ce in the city 
lerence map 
. 3. Areas of 
city of Bam 
s outside the 
cant changes 
Figure 4 Analysis of the coherence change index 
Areas 
f",.. 
f Mcc 
K 
Area 1 
0.4683 
0.2044 
0.3923 
Area 2 
0.4921 
0.2019 
0.4182 
Area 3 
0.4112 
0.1746 
0.4039 
Area 4 
0.4095 
0.2615 
0.2206 
Area 5 
0.4975 
0.3100 
0.2322 
Area 6 
0.4925 
0.4662 
0.0274 
Area 7 
0.4078 
0.4252 
-0.0209 
Table 1. Analysis of the coherence change index 
4. EXPERMENTAL RESULTS 
The damage mapping of the Bam city reported by International 
Centre for Geohazards (ICG) is shown in figure 5. According to 
coherence change index k , the damage grade of Bam city was 
defined as: 
k < 0.14 , light or none damage areas 
0.14 < /c < 0.25 , moderate damage areas 
k > 0.25 , severe damage areas 
The coherence change index map for the city of Bam shown in 
figure 6 appears very speckled. This may result partly from 
noise in one or both of the coherence maps used to calculate the 
index. However, there is actually a high level of heterogeneity 
in the damage field, that is represented in the k -image. It 
should be noted though, that the coherence is computed on a 
neighborhood sample around each image pixel. Individual 
damaged structures will consequently affect several 
neighboring pixels and cannot usually be resolved at the image 
resolution of the radar. Nevertheless, images like figure 6 are 
too detailed to be useful in emergency response operations after 
the earthquake. To address this we analyzed the average 
coherence change index in small areas on a city-block level 
(figure 6), that were defined based on the IKONOS image [6]. 
In figure 7, the red areas were severe damage areas, the green 
areas were moderate areas and the blue areas were light or none 
damage areas. The severe damage areas located in the east and 
south areas of Bam city, all the areas had dense buildings. 
Moreover, in the rectangle of figure 6, the coherence change 
index were large, figure 7 showed no damage. The reason was 
related to the earthquake are expected. Regions of increased 
coherence are also seen at several locations in the image, most 
clearly in the ephemeral river and related runoff patterns. This 
indicates a loss of coherence due to rain between June and 
December that did obviously not affect the coseismic 
interferogram. 
Figure.3 Map of the coherence change index 
From the damage mapping of the Bam city reported by 
International Centre for Geohazards (ICG) [1], seven areas were 
chosen to analyse coherence change index (shown in figure 3). 
Areas 1, 2 and 3 are severe damage areas. Areas 4 and 5 are 
moderate damage areas. Areas 6 and 7 are light or none damage 
areas. The results are shown in figure 4 and table 1. Before 
earthquake, the coherence coefficients of these seven areas 
were in the same level, namely in the level of 0.4. After 
earthquake, the coherence coefficients of the areas 1 to 5 
decreased remarkably. But the coherence coefficients of the 
areas 6 and 7 changed inconspicuously. The coherence change 
index of areas 1, 2 and 3 were the largest, namely all greater 
than 0.25; the index of areas 4 and 5 take second place, between 
0 and 0.2; the index of areas 6 and 7 were smallest, namely 
around zero. 
Above results by the coherence change index of Bam 
earthquake verifies that the magnitude of coherence change 
index is relative to urban damage level, i.e. severe damage areas 
have largest coherence change index, moderate damage areas 
take second place and the coherence change index of light or 
none damage areas are smallest. As a result, it is feasible using 
coherence change index to detect earthquake-induced urban 
damage. 
Area 1 
0.6 
Area 7\ , 
Area 2 
—♦— Preseismic 
coherence 
Area 6 
a 
I 
[sy Area 3 
—Coseismic 
coherence 
*■ Coherence 
change index 
Area 5 
Area 4