phase) or only from the ERS-1/ERS-2 tandem mission (1 day 
apart) are suitable. This time difference between the data 
acquisitions leads to a temporal decorrelation, which among 
other factors depends on the weather conditions during the data 
acquisition. Until now, the simultaneous data acquisition to 
avoid the temporal decorrelation is only implemented on 
airborne systems; however, the concept of an implementation 
for satellites by means of a tethered system is given by Moccia 
and Vetrella (1992). Besides the derived phase difference which 
mainly depends directly on the topography of the observed site, 
the accuracy of the baseline estimation is an important 
influencing factor for the quality of the results. 
The length of the baseline in connection with the time 
difference between the data acquisition determines the 
coherence which is a standard measure of quality of an 
interferogram. At the critical baseline length (e.g., about 1100 
m for ERS-1) there is a complete loss of coherence. According 
to Schwäbisch and Winter (1995) there are several other factors 
which lead to a decreasing amount of coherence: 
e thermal noise, 
e different atmospheric conditions during the data 
acquisitions, 
e phase errors due to the processing, 
e changes in the object phase between the data acquisition, 
« slightly different viewing positions. 
Another problem is the influence of the topography and the 
weather conditions. Quantifying these parameters in SAR 
interferometry is a rather complex task. The topography and its 
backscattering behaviour directly cause changes in the phase 
difference contained in the interferogram. Besides the 
geometric distortions well known in any radar imagery such as 
layover, foreshortening and shadow, the slope angle has a direct 
impact on the quality of the phase unwrapping. In areas with 
steep slopes the 2m-ambiguity of the phase cannot be solved 
without introducing additional information. The volume 
scattering of any object leads to a time delay in the reflection of 
the signal. The result is a distortion in the geometry; i.e., the 
signal is received at another position. The direction of the slope 
has a direct impact on the angle of the phase gradient. 
Wind, snow and temperature are parameters which have a direct 
impact on the topography as well as on the coherence of the 
images. The coherence is quite sensitive to temporal changes, 
e.g. the change of the soil moisture by freezing (Askne and 
Hagberg, 1993). 
Finally, each sensor works with a signal which is determined by 
its frequency, polarisation and bandwidth. These parameters 
have a specific impact on the data depending on the conditions 
at the data acquisition such as atmosphere, acquisition time or 
weather. The frequency limits the depth of penetration of an 
object. While a sensor using X-band receives the backscattered 
signal from the top layer of a forest, the sensor which uses a 
wavelength in P-band is able to receive additional information 
from the ground. Depending on the structure of the surface, the 
response from the objects varies with the polarisation used by 
the system (e.g. VV for ERS-1/ERS-2). This is used as 
additional information for the interpretation of radar imagery; 
however, it has no direct influence on the geometry. 
The atmosphere needs to be investigated because of its known 
influence on SAR interferometry (Tarayre and Massonnet, 
1994). For radar imagery in general the influence of refraction 
in the ionosphere and effects caused by the troposphere might 
be negligible. However, for the accuracy requirements in SAR 
interferometry they have to be taken into account. The 
compensation for this effect is still a challenging task and part 
of the current research. 
2. DATA PROCESSING 
The data processing of interferometric imagery is another field 
which has a significant impact on the quality of the derived 
products such as coherence maps, interferograms and digital 
elevation models. The necessary steps of the data processing 
are well understood and implemented in several ways. 
However, it is still a challenging task to develop a software 
package for SAR interferometry to an operational status. 
According to De Fazio and Vinelli (1993) the processing 
scheme for SAR interferometric data includes in general (1) the 
registration of the single look complex images, (2) the 
formation of the interferograms, (3) the phase unwrapping, and 
(4) the reconstruction of the digital elevation model. In this 
paper the first two steps of the data processing are analysed in 
more detail. 
Assuming a sufficient accuracy of the used complex data sets, 
the quality of the interferometric results depends on the 
performance of each single processing step. For an accurate 
registration of the images a precise knowledge of the shift 
between the two scenes is needed. This is done by measuring 
control points in both scenes. After performing the coarse 
registration a first fringe image should be calculated to check 
the presence of a sufficient number of fringes. The coarse 
registration is already very sensitive to changes in the shifts. 
  
Figure 1: Interferogram with a shift of -3/-5 (Image courtesy of 
University Freiburg) 
Figures 1 and 2 show an example from two ERS-1 scenes 
acquired on 3 and 6 March 1994 at the northern part of the 
McClary Glacier near the Argentinean station on the Antarctica 
peninsula. There has been a significant amount of surface 
changes between data acquisitions. This temporal decorrelation 
reduces the number of fringes rapidly. The variation between 
the images is caused by choosing shifts in range direction 
which differ by a single pixel. 
108 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B2. Vienna 1996 
  
Figur 
Univ 
Assu! 
can b 
main 
impr: 
Thes: 
point 
contr 
with. 
SIR-( 
To re 
filter 
techn 
infor] 
techn 
were 
new 
to an 
the c« 
Base 
as th 
coheı 
perfo 
litera 
whicl 
requi 
There 
SAR 
and a 
For t 
parar 
of th 
bette; 
Data 
with 
differ