Atmospheric Effects on InSAR Measurements in Southern China 
and Australia: A Comparative Study 
X L Dine". L Ge Z, W. Li* C, Rizos” 
"Dept. of Land Surveying and Geo-Informatics, Hong Kong Polytechnic University, Hung Hom, Hong Kong, China - 
(Isxlding, zhiwei.li) 9 polyu.edu.hk 
"School of Surveying and Spatial Information Systems, University of New South Wales, Sydney, NSW 2052, Australia - 
(l.ge, c.rizos) €? unsw.edu.au 
KEY WORDS: InSAR, Radon, Gaussianity, Bispectrum, Power spectrum 
ABSTRACT: 
Satellite synthetic aperture radar (SAR) signals are often seriously contaminated by atmospheric delays. The state of the atmosphere, 
especially the atmospheric water vapor, varies significantly both in space and with time. It is necessary to understand the 
characteristics of the atmospheric variations in order to devise appropriate means for the mitigation of the effects. 
We use interferograms generated from some ERS tandem pairs covering Hong Kong, Shanghai of southern China and New South 
Wales of Australia to study the characteristics of the atmospheric effects in these areas. The anisotropy and Gaussianity oi the 
atmospheric effects are first examined with the method of Radon transform and Hinich test and then compared in terms of different 
regions and climate patterns. The spectral feagures of the atmospheric effects are then analyzed with the method of Fast Fourier 
Transform (FFT) and power spectrum. The implication of the results on practical Interferometric SAR (InSAR) measurements, 
especially on the modeling and correction of the ajmospneric effects are examined. 
1 Introduction 
Repeat-pass InSAR measurements can be significantly affected 
by the atmosphere (e.g., Massonnet and Feijl, 1995; Rosen et 
al., 1996; Tarayre and Massonnet, 1996; Zebker et al., 1997). It 
is useful to characterize the effects for different regions due to 
the highly variable nature of the effects in place and time. 
Beside enabling us to understanding the effect better, such 
studies will also offer useful results for the modeling and 
mitigaton of the results. In addition, the unprecedented spatial 
resolution of InSAR also makes it a useful too to study the 
atmosphere. 
This paper will first generate a number of differential SAR 
interferograms, covering Shanghai, Hong Kong of southern 
China and New South Wales of Australia that reveals the 
atmospheric signals in the regions. The atmospheric signals are 
then studied comparatively for the different regions to 
determine such parameters of the signals as isotropy, 
Gaussianity and spectral features. Finally the implications of 
the results on modeling and mitigating the atmospheric effects 
on InSAR measurements are discussed. 
2 Extraction of Atmospheric Signals from SAR 
Interferograms 
A SAR interferogram generated by complex conjugate 
multiplication of two SAR images is a superposition of 
topographic information, surface deformation and atmospheric 
propagation delay variations between the two acquisitions, and 
noise (e.g., Tarayre and Massonnet, 1996; Hanssen et al., 1999; 
Li et al., 2003). The contribution from the topography can be 
removed by using a reference elevation model. That from the 
surface deformation can be neglected or removed if the 
deformation of the Study area between the two 
ma image 
acquisitions 1s insignificant or the deformation is known, 
Besides, multi-looking operation and careful interferometric 
processing are also needed to Suppress the noise. The 
70 
atmospheric signatures can then be obtained by removing the 
topographic and deformation terms (Hansen et al., 1999). 
We choose a tandem pair covering Shanghai, Hong Kong and 
New South Wales respectively in this study (see Table 1). It is 
safe to assume that there is no ground deformation phase in the 
SAR interferograms over such a short time interval. The 
topographic phases are removed based on external DEMs. For 
the Shanghai study region, the SRTM 3 arc-second DEM is 
used. Its accuracy is about 10m and it can potentially introduce 
a phase error of less than 0.05 cycles for an ambiguity height of 
216 m (See Table 1). For Hong Kong region, a DEM from the 
Lands Department of the Hong Kong Government with an 
accuracy of about 10 m is used. It can cause a phase error of 
about 0.1 cycles. For the New South Wales region, the accuracy 
of the DEM is about 15 m and the phase error caused is less 
than 0.1 cycles. Therefore, for the three interferograms, the 
phase errors from DEM errors are negligible. 
Precise ERS1 and ERS2 orbits from The Delft University of 
Technology are used in the processing to reduce errors from co- 
registration and flat earth phase removing (Scharroo et al. 
1998). To suppress noise, the SAR images are processed by 
multi-looking operation with 10 pixels in azimuth and 2 pixels 
in range directions to get a final resolution of about 40 m by 40 
m. Since the residue satellite orbit error and ionosphere will 
possibly introduce a linear phase ramp in interferogram 
(Tarayre and Massonnet, 1996: Hanssen, 1998), careful 
baseline refining and linear phase ramp removing procedures 
are also included in the interferometric processing (Atlantis, 
2003). The differential interferograms and differential 
atmospheric delay signals in mm (mapped from unwrapped 
phases) for the three study regions are shown in Figure 1, 2 and 
3, respectively. 
International 
  
  
1 
Figure |. Dif 
atmospheric i 
square regior 
  
  
Figure 2. D 
atmospheric 
A square reg