154 
The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part Bl. Beijing 2008 
Note that the congestus clouds were masked due to the lack of 
the information of water content of such clouds and geometrical 
depth of the clouds. 
QFE (uncorrected from height effects) was used to estimate the 
air pressure error. As the wind speed low and the area of interest 
were not so much vast, a central station was chosen to derive 
the air pressure. Another station was used to validate the 
amounts. Air pressure amounts were assumes like Table2 and 
exracted from Mashhad Meteorological synoptic data. 
Parameter 
12-09-2005 
08-08-2005 
Air Pressure (mb) 
903.2 
894.6 
Table 2: QFE amounts of images. 
Air pressure zenith delay was calculated from equation 4. 
For each pixel, the amount of error was extracted by nearest 
neighbor interpolation of error map and 0 was calculated by 
using the SLC header parameters like near range and height. 
After implementing the error algorithm, plane fit strategy was 
chosen to flatten the interferogram. Final Corrected 
interferogram could be seen in Figure 5. 
ZHD = 10 -6 — P s 
§ m 
(3) 
Where g m is the local gravity at centre of atmospheric column. 
As could be seen, by using the values of Table 2 around 2 
centimeter error was derived which is too much high. If a plane 
fit algorithm is used for flattening this error will be almost 
removed. But if a calculative method is implemented, this error 
may be ruined the results. 
The total error map is shown in Figure 4. 
To implement this map to the interferogram, the Eq.4 was used 
(Zebkeretal, 1997). 
4tt ZTD 
 Cos 6' 
(4) 
In this equation, ZTD is the Zenith total delay and 0 is the 
incident angle at each point. A point to point algorithm was 
implemented to th interferogram for total atmospheric 
correction. 
Figure5: Corrected and flattened interferogram 
4. RESULTS 
GPS time series points were used in this stage to validate the 
results. 3 points in approximately middle parts of this 
interferogram and near the subsidence area (where is the 
research area of interest) was used for this purpose. The 
distance between the points is around 20 kilometers. The 
comparison of the values could be seen in Table 3. (The GPS 
values were obtained from the National Cartographic centre of 
Iran). The uncertainty of GPS values are around a millimeter. 
The total RMS Error of these points is around 8.6 Millimeters 
which is acceptable in comparison with the other carried out 
activities. 
Points 
InSAR (mm) 
GPS (mm) 
Mashhad 
-8 
-2 
Tous 
-52 
-41 
Torqabeh 
5 
-3 
Table 3: GPS and InSAR deformation 
Figure4: Total Error map 
5. CONCLUSION 
As it could be seen in validation part and the result map, the 
atmospheric artifacts was reduced in the correction process. The 
RMS error of just flattened interferogram was about 1.9 
centimeter. Furthermore if the water vapor amounts differed 
more in location, this error would be intensified. 
Implementing the flattening model after atmospheric correction 
helps the process to obtain better results. By using the flattening 
model before the correction, some unreal systematic error 
would be imported to the interferogram; because the specific 
model which used for flattening will omit the linear errors and 
linear parts of other influences. This would cause overreduction 
on the interferogram. Mitigating the air pressure, implementing 
the total error map to the interferogram and flattening in respect 
showed around 9.1 millimeters disagreement between InSAR 
and GPS. This shows that the linear errors like Ionosphere could 
be neglected if a fit method for flattening was chosen.