Fig.5.Differential interferograms obtained by ERS/SAR.
(a) Pair-1 (Jan.21 - May 6, 1999).
(b) Pair-2 (May 6 - Sep.23, 1999).
(c) Pair-3 (Jan.21 - Sep.23, 1999).
(d) Phase difference patterns between (b) and (c).

Extraction of Land Displacement by Differential Interferogram
We attempted to generate differential interferograms from the first three pairs indicated in Table 1.Figure 5 (a),(b),and (c)shows all of the obtained differential interferograms. The phase patterns in these interferograms indicate the phase differences due to land displacement in the slant range direction under the assumption that the phase patterns caused by orbit difference (orbital fringes)and topography (topographic fringes)are removed perfectly. For the confirmation of removal of both fringes, the phase patterns (a)by pair-1 is a key result, because this pair does not include the earthquake occurrence and so it is hardly possible to get the phase patterns due to land displacement. In fact, the phase patterns in the interferogram by pair -1 are almost flat, which supports that both fringes are almost completely removed. Another important key is the consistency of the phase patterns between the results from different data combinations. This consistency can be investigated by subtracting the interferogram by pair-2 from that by pair -3.The result is shown in Figure 5 (d).The phase patter ns in the interferogram (d) are also almost f lat and consistent with those by pair-1,which interval corresponds to the difference of the intervals by pair-3 and pair-2.

The investigation described above on all of the obtained intergerograms clearly supports that the differential interferometry by ERS/SAR succeeded to extract the phase patterns relat ed to the land displacement caused by Chi-chi earthquake.owever,the usable phase patterns are only available in flat or semi-flat areas.In mountainous areas,the interferograms are almost noisy and they can not bring any information about land displacement.This result clearly comes from poor coherence by ERS/SAR inter- ferometry in mountainous areas as pointed out in the beginning of this section.Therefore the application of differential interferometry by ERS/SAR is actually limited to the areas where topography is flat or semi-flat and the vegetation cover is relatively less like urban or agricultural areas. 

Fig.4 Results of evaluation of SAR intensity changes (left)and coherence changes (right)due to the earthquake in the sample urban areas in Taichung and Nantou Prefectures (except Changhua).(*denotes severely damaged urban group).


Fig.6.Displacement patterns obtained from the differential interferoframs in Figure 5. [(a):Pair-1,(b):Pair-2,(c):Pair-3 ]

From the above results, the land displacement patterns due to Chi-chi earthquake are possible to extract from the differential interferograms shown in Figure 5.The amount of displacement in a slant range direction is a half of the wavelength (a half of 5.6 cm for ERS/SAR) for one cycle of phase patterns. The direction of the displacement is toward satellite if the phase values decrease toward satellite, and backward satellite if the phase values increases toward satellite. As all test ERS/SAR data were acquired in a descending orbit, he radar signals were illuminated from the right side of the images. In the interferograms of Figure 5 (b)and (c),as the phase decreases in the right direction, namely toward satellite, the directions of the displacement are all toward satellite if the left-side edge portion (actually the coastal line of the left part of the images)does not move.

Figure 6 (a),(b)and (c)show the displacement patterns obtained from the three differential interferograms in Figure 5 (a),(b)and (c)respectively.As the amount of displacement is only computed in the areas where phase unwrappi ng was succeeded,the displacement patterns were not obtained in almost of mountainous areas where the coherence was very poor and so the displacement fringes could be hardly obtained.In addition,the displacement patterns in Figure 6 were computed as the relative displacement inside the whole areas where phase unwrappi ng was succeeded because the displacement is possible to compute only from the relative changes of phase patterns.

As the displacement values in Figure 6 indicate the displacement in a slant range direction, it is necessary to convert the displacement values to those in a vertical or a horizontal direction, although it is impossible to decide either vertical or horizontal only by the interferograms. The conversion is done by the multiplication of the inverse of cos denotes incidence angle of SAR)for a vertical displacement and sin for a horizontal displacement respectively. As is approximately 23 deg. for ERS/SAR, the multiplication factor is 1.09 for a vertical and 2.56 for a horizontal direction respectively. As the maximum relative displacement value seen in Figure 6 (b)and (c)is about 22 cm in a slant range, the maximum displacement value is about 24 cm for a vertical and about 56 cm for a horizontal direction. As the result of the comparison between the displacement patterns in Figure 6 and those by GPS observation (Nat ional Cheng-Kung Univ.,1999),the displacement patterns by SAR were proved to be fairly compatible to those by GPS in both of direction and amount. 

Detection of Landslide Areas Using Coherence and Intensity
We also attempted to detect landslide areas in mountainous regions caused by the earthquake using the change of coherence and intensity. For the extraction of coherence changes, the coherence images by pair-1 and pair-4 were compared each other. The reason why these two pairs were used was that there might be some possibility for coherence increase in landslide areas after the earthquake because forest vegetation was almost lost due to the landslide. However, the result was that coherence by pair-4 (after the earthquake)was extremely low in mountainous regions and still same as that by pair-1 (before the earthquake).This result indicates some limitation of InSAR by C-band SAR in steep mountainous regions. On the other hand, it was found that the intensity by SAR multi-look images partly decreased in landslide areas after the earthquake. Therefore, some of the landslide areas were possible to be detected using these intensity changes, although the landslide areas in the slopes facing to radar illumination were hardly detected due to a big foreshortening effect in ERS-2/SAR images. This result indicates that backscattering intensity is partly available to detect landslide areas even in mountainous regions. 

Conclusion
As the first result of this study, the coherence information was proved to be an effective parameter to detect damaged urban areas by the earthquake, which has been also verified by another case study for The 1999 Great anshin-Awaji Earthquake (Yonezawa and Takeuchi,1999).The consistency between the two case studies seems quite important and valuable because it clearly enhances that SAR observation in which a good interferometreic condition is maintained is significantly important for disaster monitoring by SAR data.The second result suggested that differential interferometry by ERS/SAR is possibly effective to extract land displacement by an earthquake together with some limitation. In flat or semi-flat areas where main land covers are occupied by urban or agricultural lands, the differential interferogframs obtained by ERS/SAR brought reasonable displacement information caused by the earthquake. The third result suggested an essential limitation of C-band InSAR in steep mountainous regions. The first two successful results clearly enhance the importance of interferometric SAR for disaster or environmental monitoring using space-borne SAR data.

References
Chen A.J.and Chien Ying Wang.2000.Using SPOT I mageri es to Monitor Landslides in Chi-chi earthquake.Pr esented At I nter national Wor kshop 2000 in I T on Gener ation of Advanced Earth Environmental Information.

Kokusai Kogyo Co.,LTD.1999.Record of Great Taiwan Earthquake on September 21,1999 (in Japanese). 

National Cheng-Kung University.1999.Satellite Geoinformatics Research Center,921 Chi-chi earthquake damaged area satellite control points displacement map,http://www.sgrc.ncku.edu.tw,

Yonezawa,C.and S.Takeuchi.1999.Detection of Urban Damage Using Interferometric SAR Decorrelation.Proceeding of IGARSS'99,pp925-927.