ing. Fur- 
ms. They 
ferencing 
g can be 
the real 
linates of 
R images 
fact into 
and the 
e. 
ne using 
anges in 
\R simu- 
age. The 
ences be- 
and the 
1ges. Al- 
c 
he wave- 
ages are 
systems. 
d optical 
cially the 
0 consid- 
d optical 
range 
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B7. Istanbul 2004 
  
In Figure 1, the SAR geometry is shown schematically using a 
single house as an example. For SAR processing, flat terrain is 
assumed, therefore the house is wrongly positioned in the im- 
age. Point A is imaged correctly in A’, whereas the points B and 
C are imaged in B' and C', closer to near range, as their real po- 
sition. This effect is called layover and is caused by the run-time 
geometry of the SAR sensor. The range between A and E is the 
shadow area. Because of the displacement of point C in C' the 
shadow is starting at point A and not at point D. 
In dense urban environments the situation is getting more com- 
plex. As shown in Figure 2, occlusions and ambiguities make 
the interpretation of the data nearly impossible. In the RADAR 
shadow, no data is acquired at all. The layover area of buildings 
also occludes a lot of information. In fact it is hard, sometimes 
impossible, to tell which information is reflected from which 
building, thus making the interpretation of SAR images in urban 
areas very complicated. 
  
Figure 2. SAR geometry of an urban environment in range 
3. SAR SIMULATION 
A SAR simulator is a key tool for the interpretation of SAR 
images (Leberl & Bolter, 2001). Using SAR in urban environ- 
ments, the simulator is also useful during mission planning, for 
choosing the optimal SAR acquisition parameters and avoiding 
occlusions in the area of interest (Sórgel et al, 2003). 
For the presented approach, the SAR simulator is the key ele- 
ment. The complex interaction of different effects in SAR can- 
not be totally understood, but a SAR simulator helps under- 
standing SAR images by simplifying the reality. Therefore SAR 
simulators are used for training purposes and are also quite 
helpful in change detection applications. 
  
KP. E C. + 
Z1 of 
Figure 3. Subset of a DOSAR image of Karlsruhe 
  
  
Figure 4. Simulated SAR image 
Figure 3, shows a DOSAR image of the area of Karlsruhe. DO- 
SAR is the multifrequency polarimetric airborne SAR system of 
the EADS Dornier GmbH (Hoffmann & Fischer, 2002). The 
flight direction is 90.05?, the off-nadir angle is 70°, the pixel 
resolution is 0.26m and the 3dB-resolution is about 0.57m. In 
Figure 4, the result of the SAR simulation of the 3D-city model 
of Karlsruhe is shown. Comparing Figure 3 and Figure 4, the 
great differences between the simulation and the reality are get- 
ting obvious. 
  
Figure 5. Model of the St. Bernhardus church 
In Figure 5, the model of the St. Bernhardus church from the 
3D-city model of Karlsruhe is presented. This model was recon- 
structed from LIDAR data and ground plans (Haala & Brenner, 
1999). Due to problems during the data acquisition, the spire is 
not correctly modelled and the whole building is wrongly 
shaped. The error regarding the footprint of the building is 
around Im in different directions. Anyhow such a model is 
good enough for the presented approach. 
Using the SARView Light SAR simulator (basis version) of the 
EADS Dornier GmbH, this dataset was simulated according to 
the SAR parameters of the real SAR data acquisition flight (see 
Figure 11). 
3D-city models are generalised and simplified representations 
of the reality. In addition they can be erroneous and incomplete. 
Every simulation based on these models can therefore be in- 
complete and wrong. Furthermore the SAR simulator is not able 
to handle all the possible SAR effects and even the real SAR