Effect of coherence on dems derived from sar interferometry:
A case study of mayon volcano, philippines
 |
B
r1
r2
h
a
R
S1
S2
Bc,Bn
a
f
q |
Baseline Magnitude
Reference slant-range
Second slant-range
Ground radial distance
Satellite radial distance
Center scene geocentric radius
Reference satellite
Second satellite
Baseline cross-track and normal component
Baseline orientation |
Figure 1 - Interferometric geometry for calculation
2.2 Investigating the effect of phase coherence on DEM accuracy
DEM derived from SAR interferometry above (DEMInSAR) was compared to DEMtopo with respect to the changes of coherence coefficient. Two-steps investigation was done. First, the profiles were taken from two DEMs and coherence image to give a figure of the trend. Then, coherence value was grouped with the interval of 0.01. With each group, the standard deviation of height difference was calculated. Last, we drew the graph of the relationship between coherence value and this standard deviation value.
2.3 Investigating the relationship between phase coherence and land cover type.
From geocoded Landsat TM images, Normalized Difference Vegetation Index (NDVI) was calculated. Like the above investigating, NDVI was grouped with the interval of 0.01. Then, the trend of coherence with respect to NDVI was investigated.
3. Results And Discussion
3.1 Study area
Mayon volcano is located in the Alabay province of Philippines, 300 km SE of Manila
(13°15.4'N - 123°41.1'E). The summit is about 2460 meters above sea level and base circumference is 62.8 km, encompassing the towns of Camalig, Malilipot and Sto. Domingo. Mayon has the classic conical shape of a stratovolcano. It is the most active volcano in the Philippines. The latest eruption is in February 2000 and on going.
3.2 Data processing
As mentioned above, two SAR data sets were used. The result and discussion here are demonstrated following the step in Method part.
Making DEM
The flattened interferograms are shown in Figure 2. In 1997 data, the fringe only appears in some area in the foothill of Mayon and completely disappears in the steep slope area. With 1996 data, the result is better and we can recognize that the fringe is worse in layover area. The estimated baseline of two data sets is introduced in Table 2.
Table 2 - Estimated baseline
|
Data | T | C | N | Length |
| 1996 baseline (TCN) (m) | 0 | -86.07 | 32.28 | 91.93 |
| 1997 baseline (TCN) (m) | 0 | 393.98 | 53.97 | 397.66 |
Let us consider coherence coefficient. The coherence images are depicted in Figure 3. Compare between Figure 2 and Figure 3, we know that the higher coherence coefficient, the clearer fringe.
Figure 2 - Flattened interferogram of 1996 data (left) and 1997 data (right)
Figure 3 - Coherence image of 1996 data (left) and 1997 data (right)
As mentioned above, region-growing algorithm was used for phase unwrapping. We have done successfully with 1996 data, but failed when trying 1997 data. The reason is that the high phase noise in 1997 interferogram. Therefore, only 1996 data was used to make DEM and compared to DEMtopo.
After automatically coregistration between DEMtopo and SAR by cross-correlation analysis, a number of GCPs was extracted for precise estimation of baseline geometry. The slant-range and ground-range elevation image are shown in Figure 4. It is noted that DEMtopo should be masked with DEMInSAR to eliminate these no-information areas in comparison.
Figure 4 - 1996 slant-range elevation (left) and ground-range elevation (right) images
