Michele Crosetto 
  
  
  
  
  
  
  
  
  
  
  
  
Processing Terrain Mean Error Standard Deviation 
Type type [m] [m] 
Without hilly/flat 0.25 15.18 
Atmospheric mountainous —441 22.71 
Correction 
entire area - 1.21 18.14 
i 2 
Atmospheric Correction hilly 0:29 10.54 
with mountainous 1.08 18.47 
Radargrammetry Data : 
entire area 0.54 13.75 
Atmospheric Correction hilly/flot 029 342 
with mountainous -0.21 15.93 
Optical Data 
entire area 0.13 11.36 
  
  
  
  
  
  
Table 2: Ascending InSAR DEM results. 
2.1.1 Image coherence. The coherence is a good indicator of the interferometric phase quality for DEM generation. 
Low coherence causes both the degradation of the precision of the InSAR point positioning and problems in the phase 
unwrapping. Therefore, in low coherence areas there is an important decrease of the DEM quality. In Figure 1 it is 
shown an example of low coherence area (along the Ebro River) with the corresponding large errors in the InSAR T 
DEM. The decrease of DEM quality due to coherence is illustrated in Table 3, where the statistics of the height s 
differences (InSAR versus reference DEM) computed for different coherence classes are reported. The standard e 
deviation increases from 5 m (coherence 0.81) to 18 m (coherence 0—0.1). 
The last row of Table 3 refers to the pixels of the geocoded coherence map that are not associated to any coherence i 
value. These pixels are consequence of the slant range nature of SAR images that makes the terrain sampling very 
irregular. To these pixels (named “Interpolated” in Table 3) corresponds a large standard deviation (23.5 m). 
2.1.2 Terrain Topography. There is a strong correlation between the type of terrain topography and the 
corresponding quality of the IDSAR generated DEMs. Assumed to process images with a high mean coherence (e.g. 
bigger than 0.5), InSAR DEMs have quite good precision over areas characterised by gentle terrain variations. Dealing 
with complex terrain topography, some effects related to the SAR image distortions (foreshortening, layover and 
shadow) make difficult the phase unwrapping and result in spacing irregularities (i.e. “holes”) in the generated grids. 
0 1.5 km - 40m 
N 
Om 
+40 m 
  
Figure 1: Ascending InSAR DEM. Geocoded coherence image (left) and the corresponding map of the height 
differences — InSAR versus reference DEM (right). 
  
48 International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part Bl. Amsterdam 2000.