Manfred Sties 
  
the statistical evaluation of elevation differences between the resampled SAR and laser ’ground’ raster data sets is shown 
in Fig. 3. 13.753 differences were included, ranging from -5.0m up to +11.98 m. The average elevation difference 
amongst these samples was +0.45 m with a standard deviation of 2.02 m. In Tab. 1, it is shown that - for the same test area 
_ the differences between the SAR and the laser ’vegetation’ data sets amount to an average of -3,38 m with a standard 
deviation of 2,76 m. Evidently, the elevation measurements of the interferometric SAR are significantly lower than those 
of the laser scanner. This can be attributed to the upper edge of the vegetation cover which may also include roof tops. 
5.3 Intermap Results Comparing Co-located Elevation Rasters (Post-selected Bald-Earth) 
The radar and laser DEMs are presented as colorized, shaded-relief views in Fig. 4b and 4c, respectively, along with a 
common color scale bar. Due to the limited number of pages and space for this report, the quality of the figures appears 
to be degraded when compared to image resolution and colour depth available. Areas that are colorized as white are due 
   
   
  
um 
Figure 4a: STAR-3i Magnitude Figure 4b: STAR-3i DSM; Elevation Figure 4c: Laser DTM; Ele vation 
Image shown in meters shown in meters 
  
to the absence of data within the 15 m search radius. Inspection of these overview figures reveals immediately that while 
the general detail is very similar, the radar data (DSM) has retained effects of forests, buildings, objects on the surface. 
The laser DTM has successfully removed most of these, although some residual elevation effects remain, particularly in 
the urban areas. The difference surface (radar DSM minus laser D'TM) is shown in Fig. 4d along with its color scale bar. 
Figure Al; Dillerence Surlace 
(Radar minus Laser); Trillerence 
shown in meters 
  
The color scale bar has been selected to show all positive height differences greater than approx. 2m as green, while the 
micro-differences are colorized as shades of brown and cream. These micro-differences are due to a superposition of three 
effects: (1) systematic errors, (2) random noise errors, and (3) real terrain coverage differences (e.g., agricultural crops). 
To illustrate the magnitude of these three effects, statistics have been acquired for sub-areas denoted as A, B and C in 
Fig. 5, representing land-cover types 'bald-earth', 'agricultural crops', and "forest', respectively. The results are shown in 
Tab. 2. The mean offset for the bald-earth case is about 50 cm and is due to a combination of systematic errors from both 
radar and laser, and perhaps transformation errors. The 28 cm standard deviation of the bald-earth example is essentially 
the noise floor of the radar (assuming the noise floor of the laser to be significantly smaller). The mean difference in the 
crop sub-area is 0.66 m, but when the common systematic error is removed, the effective height of the crop is 1.13 m. 
The noise is similar to the bald-earth case. The effective height of the forest sample, on the other hand, is about 21 m, 
  
International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B3. Amsterdam 2000. 871