ks, the 
he field 
is been 
; a very 
imetric 
lapping 
sions in 
ost due 
derable 
rdware 
tensive 
ves of 
Premalatha Balan 
  
THE EFFECT OF BASELINE ESTIMATION ON ACCURACY OF INTERFEROMETRICALLY 
DERIVED DIGITAL ELEVATION MODELS 
Premalatha BALAN 
University of Nottingham, 
Nottingham, England 
balan@geography.nottin .ac.uk 
KEY WORDS: Interferometric Processing, SIR-C - IfSAR, Digital Elevation Model. 
ABSTRACT 
Interferometric SAR (InSAR) is being studied extensively to evaluate its potential to generate high resolution and high 
accuracy Digital Elevation Models (DEM). The accuracy of derived height information depends on many factors, 
including processing parameters, which need critical evaluation. Among the many processing parameters, baseline 
estimation requires a detailed evaluation of the methods used, as any inaccuracy in baseline estimation introduces a 
residual slope in the final DEM. Methods using orbital information to estimate the baseline are appropriate if the orbital 
information is precisely known. In the absence of such knowledge, other methods of estimating the baselines are 
required, and these are discussed in the present study. Fringe rate method depend on the location and size of the 
estimation window, as this method assumes that the fringes within the estimation window are due entirely to the phase 
trend on a flat earth. Baseline estimate using the fringe rate method or image offset parameters also fail to provide an 
accurate estimation of the baseline when the perpendicular and parallel components are estimated using the same 
method. However, fringe rate methods provide the best estimate for the perpendicular baseline component when 
compared to other methods studied. It is found that the use of the fringe rate method to estimate the perpendicular 
component of the baseline and the use of image offset parameters to estimate the parallel component of the baseline 
provides better results. Other methods gave ^tilted" DEMs due to the presence of residual slope. 
1 INTRODUCTION 
Interferometric Synthetic Aperture Radar (InSAR) is, in theory, a source for accurate Digital Elevation Model (DEM) 
generation. In practical applications, the achievable accuracy varies depending on many factors. It is necessary to 
evaluate the factors governing the accuracy of the InSAR derived DEM to bring the technology to operational use. 
The backscattered radar signal carries information about the target in the form of amplitude and phase, where amplitude 
is the strength of the signal and phase is the angular distance travelled by the signal from the antenna to the target and 
from the target back to the antenna. Information about the elevation of the target is contained in the phase content of the 
signal. To retrieve this information, two images of the same area collected from different observation points are 
processed by a series of steps. The DEM derived from InSAR processing and the expected (theoretical) InSAR DEM do 
not always match. The differences are due to the effects of different factors during processing. Estimation of baseline 
(the distance between the two observation points) is one of those factors which seriously affects the accuracy of the 
InSAR-derived DEM. Zebker et al. (1993) point out that inaccurate baseline estimation as one of the two main error 
sources in interferometrically-derived DEMs. The other major error source is phase noise. Though many researchers 
have mentioned the importance of accurate baseline estimation, the problem has not been addressed in detail. The 
reason could be that most of the evaluation experiments that have been conducted were dealing either with less sloping 
areas or relatively flat terrain where some inaccuracy in baseline estimation could have gone unnoticed. As InSAR is a 
potential source for high accuracy and high resolution DEMs, it is essential to address the effect of every processing 
parameter that affects the quality of the derived DEM. Significance of baseline estimation on the accuracy of InSAR 
derived DEM is addressed in this study. 
Graham (1974) introduced the use of interferometry for topographic mapping in 1974. Later, Zebker et al. (1986) 
demonstrated the potential of the technique with practical results. They achieved a vertical accuracy of 10m using 
airborne data. The main source of error in their study was aircraft attitude. Since then many publications have reported 
different figures of height accuracy under different conditions. It is necessary to understand the causes of error and the 
conditions under which the accuracy was achieved (wavelength used, baseline separation, terrain nature etc.), otherwise 
the single figure reported as the accuracy of the InSAR-derived DEM is misleading. Other than system parameters and 
terrain characteristics that affect the achievable accuracy of the DEM, InSAR processing also affects the quality of the 
final derived DEM significantly. 
  
International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part Bl. Amsterdam 2000. 29