Manfred Sties 
  
COMPARISON OF DIGITAL ELEVATION DATA FROM AIRBORNE LASER AND 
INTERFEROMETRIC SAR SYSTEMS 
Manfred Sties' and Susanne Krüger‘, J.Bryan Mercer? and S. Schnick? 
! [nstitute for Photogrammetry and Remote Sensing, University of Karlsruhe, Germany 
e-mail: sties@ipf.uni-karlsruhe.de 
?[ntermap Technologies Corp., Calgary, Canada 
e-mail: bmercer@intermap.ca 
KEY WORDS: laser; SAR; interferometry; DEM accuracy; lidar; IFSAR; DTM: DSM 
ABSTRACT 
Requirements for precise digital elevation data (i.e., Digital Elevation Models, or DEMS) of varying levels of detail are 
being formulated for different fields of application. The degree of detail found in a DEM is usually characterized by 
the horizontal sample spacing and by the vertical accuracy of the samples. Increased detail is normally associated with 
increased cost to the user, which fuels the debate over which technologies are suitable for which applications. In this 
paper, we present the results of comparative tests of two technologies from which are derived relatively detailed digital 
elevation data sets with sample spacings of a few meters. The two elevation data sets have been compared for a section 
of variable terrain in southern Germany, approx. 10km by 20 km in size. The data sets were produced by an airborne 
laser scanner and an airborne interferometric SAR sensor, respectively. The laser scanner data set consisted of two 
coordinated list sets of elevation measurements with a variable ground spacing ranging from approx. | m to 10 m. One 
list set contained measurement points of the earth's surface, whereas the second list set held measurement points of the 
vegetation surface. The interferometric SAR data set consisted of elevation data in a regular grid with a sample spacing of 
Sm. Using somewhat different approaches, the laser and radar data sets were assessed independently at the University of 
Karlsruhe and at Intermap Technologies, Corp. in Calgary, Canada. Three different comparative methods were applied to 
evaluate the elevation data sets. All comparisons came to similar results, namely, that both elevation data sets are in good 
agreement for bare soil or areas of low vegetation coverage (less than 0.5 m elevation difference), whereas larger elevation 
differences were found for all areas of substantial vegetation coverage as well as for settlement areas. 
1 INTRODUCTION 
A wide range of applications is now driving the requirements for increased detail in Digital Elevation Models (DEM). 
Detail in this instance is defined by the horizontal sample spacing and vertical accuracy of the measurement. Traditionally, 
DEMs have been produced by survey methods or by stereo photogrammetry. In this paper, we restrict ourselves to 
a comparison of DEMs from two active sensors, airborne scanning laser and airborne interferometric SAR (Synthetic 
Aperture Radar), which are capable of providing elevation samples every few meters and with sub-meter vertical accuracy. 
Focusing on this level of detail precludes inclusion of DEMs created from repeat-pass satellite SAR technology, as well 
as those from the recent NASA/DLR/ASI SRTM mission. The airborne laser scanner approach has been described in 
various papers, e.g. (Lohr 1999) or (Wever 1999) for the basics and (Kraus 1999, Knabenschuh 1999, von Hansen 1999) 
for applications. Several airborne interferometric SAR systems are currently in use. The interferometric process has been 
widely discussed in the literature, particularly for the case of repeat pass interferometry, e.g. (Zebker 1992), while some 
general issues associated with airborne interferometry have been discussed in (Gray 1993). 
Analysis of the accuracy of elevation measurements is often performed on the basis of a number of control points, the 
absolute coordinates of which can be determined very precisely using GPS techniques, refer to (Kleusberg 1999). While 
very relevant, this approach limits the ability to differentiate the effects associated with the fine spatial detail provided by 
the sensors considered here. In particular, it is difficult or impossible, using control points alone, to assess the accuracy of 
large area elevation measurements of regions with relief and various vegetation coverage. In this study, the laser-derived 
bald-earth DEM was assumed to represent "truth at the 15cm to 30cm accuracy level; the assumption was confirmed 
in part by (Kürbis 1998). It was therefore possible to compare the Radar DEM directly with this detailed, but wide- 
area, form of truth. Moreover, because aerial imagery and topographic maps were also obtained for the test area in 
Kraichgau (western part of Baden-Württemberg), the elevation differences could be assessed between the measurement 
sets in distinct areas of various relief or vegetation coverage such as urban areas, forests, water bodies, vineyards, bare 
soil areas, etc. In the following, DTM (Digital Terrain Model) is often used to denote the bald-earth DEM from which all 
surface vegetation and objects have been removed, while DSM (Digital Surface Model) references the scattering surface, 
including vegetation and surface objects. 
  
866 International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B3. Amsterdam 2000. 
(a) 
th 
as 
ac 
TM D (m C (5 £5 CT (p N "C UA 
NM