Ss 
  
SEE EEE EEE EEE A eT i ET ——— emm 
ee Ne ES NOTES E 2 E EN E Eis 
iud 
p 
| 
Hi 
Yueqin Zhou 
  
  
  
      
  
  
  
  
pi^ NN b: 
N « 
§ Ny 
^s NA ^ 
^ S 
4 N 
# Min & 
A “a À 4% 
N a, FR 
Wy N 
y" NN S 
RSS a Sixt 
Y AN Boundary of study area 
ignis e N Levelling points 
wk 
i + 
am Kilometers 
  
  
Figure 2. Location of test site and subsidence Figure 3. Levelling network 
Three scenes of ERS-1/2 SAR SLC data were used in this paper. The relevant information is listed in Table 1. 
  
Temporal | Spatial baseline to 23481 
No. | System Orbit Date baseline 
; to 23481 | Pam. B, Perp. B, 
  
  
1 ERS-1 23481 | 11/01/96 
ERS-2 | 03808 | 12/01/96 1 day 102 42 
3 ERS-2 12826 | 03/10/97 | 21 months -76 -37 
Table 1. Spatial and temporal baselines of ERS SAR data 
  
  
  
  
  
  
  
  
  
  
  
4.2 Levelling data processing 
Since the acquisition time of SAR data is different from that of levelling measurements, the temporal interpolation is 
required to calculate the height values of the levelling points at the time of SAR data acquisitions. Assume the temporal 
height change at an y point can be modelled by a polynomial: 
2 
Zi = 200) = bp + Dj + bat” + bar” + (8) 
Where, 7 denotes the time period with respect to a reference time, b,, b, ‚b, oss +++ are polynomial coefficients. At the 
given levelling point, the coefficients are calculated using F7 
the linear regression by means of the known levelling 
measurements. The order of the polynomial is automatically [eo 4 
determined through the significance test. Then using the 
calculated coefficients, the height at the time of SAR scene 
acquisition is interpolated by (8). 
  
  
40 + 
20 Std. Dev = .14 
Mean = .43 
AN = 437.00 
  
  
   
MY CY uo o 
Using the above method, the height values of total 437 |^! T 
levelling points within the study area are calculated for the UN MR UT ra a i RNS valves te 
acquisition time of scene 23481 and 12826. The average 
RMS value is 0.43 cm, while the standard derivation of 
RMS value is 0.14 cm. Figure 4 shows the histogram of the RMS at all points. It is, therefore, reasonable to use the 
polynomial to do temporal modelling. 
  
  
  
Figure 4. Histogram of RMS of temporal modelling 
From the interpolated height values, the height change between the acquisition time of scene 23481 and that of 12826 is 
calculated at each levelling point, and then the subsidence map is generated, shown in Figure 5. 
4.3 InSAR data processing 
Atlantiss EarthView InSAR software is used to process the InSAR data. This software provides *DEM mode" to generate 
DEM, and "Differential mode" with ‘two -pass”, ‘three -pass”, or ‘four -pass” to make change models. In our experiment, 
the DEM mode is firstly used to generate the reference DEM from the tandem pair 23481/03808. Then using the reference 
DEM as “External InNSAR DEM", the three -pass differential SAR interferometry is applied to generate the differential 
interferogram as well as the height change model from the image pair 23481/1282 6. A subarea with a size of 25km x 23km 
(seen in Figure 3), which covers the urban and suburban area of Tianjin, is selected for the test. The geocoded differential 
interferogram and the height change model are illustrated in Figure 6 and Figure 7, respect ively. 
  
356 International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part Bl. Amsterdam 2000. 
Even t 
differe 
Each f 
Figure 
4.4 Fa 
In ord: 
by D-I 
into th 
using t 
SAR i 
area, 
transfo 
points. 
change 
The cr 
as the 
226 le 
and ar: 
Based 
betwee 
(7), an 
square 
residu: 
Figure