E 7 0 0O 3 T^ 
— 
) 
(0 = 
— ds my, €. CD CD 
- — | M = 
  
Manfred Sties 
  
2 BACKGROUND TO THE PROJECT 
In July, 1998, Intermap Technologies collected DEM and image data of the whole of the state of Baden-Württemberg 
(approx. 43,000 km?) using its STAR-3i airborne interferometric SAR system. A laser-derived DTM (Digital Terrain 
Model) for a test section of the state (approx. 150 km?) had been collected on behalf of the State Mapping Agency (LVA) 
in Jan., 1996, by the TOPSCAN company. As a result of a data swap between Intermap and LVA, the two sets of analyses 
were performed independently by the University of Karlsruhe and by Intermap Technologies in Calgary. As discussed 
below, the objectives and approaches were somewhat different but the results and conclusions were almost identical. 
The main objective from the Intermap point of view was to demonstrate the absolute and relative vertical accuracies of 
the STAR-3i DSM in ’bald-earth’ regions of the test area. The Karlsruhe university objective additionally included the 
assessment of vegetated and other classes of terrain, which in this comparative analysis, describes only indirectly the 
accuracies of the elevation measurement, but rather the response of the radar to vegetative classes. 
3 INPUT DATA DESCRIPTION 
The two prime data sets were those acquired by airborne scanning laser (an Optech scanning laser operated by TOP- 
SCAN) and by the STAR-3i interferometric airborne SAR (operated by Intermap). Some of the comparative operational 
characteristics of the two systems are described in (Mercer 1999). The TOPSCAN airborne laser campaign generated an 
irregular elevation data set with an average point spacing of approx. 4 m. The measurement points were geoid-referenced 
and projected into the Gauss-Krüger system. Multiple, overlapping flight paths yielded locally variable point density. 
The elevation data set was partitioned by Topscan into two disjunctive subsets. One subset, containing most of the data, 
was to represent the elevation of the ground surface (DTM, named 'ground' in the following) and the other subset was 
to represent the elevation of the top level of vegetation coverage (DSM, named 'vegetation' in the following). A detailed 
functionality of the selection method is not known. Hence, it is uncertain to which subset the measurement points e.g. on 
roofs of buildings have been allocated. The laser scanner data acquisition took place on Jan. 12 and 13, 1996, during the 
leaf-off and agriculturally dormant season. 
A second measurement campaign by the Intermap STAR-3i airborne interferometric SAR system generated a regular 
raster DEM with an element size of 5 m x 5 m. Radar image data were simultaneously collected and ortho-rectified during 
the processing activity. The data set was referenced horizontally and vertically to WGS-84 and projected into the UTM 
coordinate system. Data acquisition took place during the first half of July, 1998, during the leaf-on season and with crops 
well developed. In addition to the elevation data sets described above, panchromatic digital orthophotos of the test area 
with a ground resolution of 0.5 m, were obtained and used in the Karlsruhe analysis. The aerial imagery for generating 
these orthophotos was acquired on June 10, 1998. The determination of vegetation coverage for the various test fields was 
based on this date. It should be noted that the time difference between both elevation data acquisitions, and in particular 
the differing seasons, might influence the statistical evaluation of the elevation difference calculations unfavorably. 
4 ELEVATION ACCURACY ASSESSMENT 
The assessment of elevation differences between both, the laser scanner and the interferometric SAR data set was based 
on comparisons, the basis of the comparative study. Elevation differences were calculated (1) between the "original' mea- 
surements at nearly co-located positions and (2) between co-located elevation rasters which were derived by interpolation 
from the "original measurements. The differences were calculated for both, the ground and the vegetation data subset. 
Several features distinguish airborne radar from laser systems, and some of these are elaborated in (Mercer 1999). For 
purposes of this paper, two important distinctions should be noted: (1) Each radar elevation sample is an integrated re- 
sponse from an approx. 5m by 5m cell. For a homogeneous target over the cell area, the height is representative of the 
geometric center of the cell. On the other hand, the laser 'footprint is about 20cm in diameter, with a mean spacing 
between the irregularly spaced point samples of about 4 m. Therefore, the laser-derived DTM is an interpolated surface, 
and some inter-sample detail may be lost. (2) Some fractions of the laser pulses penetrate through forest canopy to the 
ground, particularly in the leaf-off condition, and these were used to create the assumed bald-earth truth’. Conversely, 
the radar only penetrates, depending on radar wave length and other parameters, partially into vegetation and the derived 
elevation represents the effects of volumetric scattering within the canopy. The effective scattering height is dependent on 
system and vegetation parameters. Comparison of the radar DEM/DSM vwith the laser derived bald-earth DEM/DTM in 
forested or crop-covered areas will, therefore, provide also information about the target response to the radar. 
4.1 Transformation Issues 
Before data from the two sets could be directly compared, it was necessary to have them in a common reference system. 
The laser data were in a local state plane system (Bessel ellipsoid, DHDN datum, Gauss-Krger projection, Denker geoid). 
The radar data were in a global system (WGS-84 ellipsoidally referenced, UTM projection). The Karlsruhe approach was 
  
International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B3. Amsterdam 2000. 867