99 
SURFACE DEFORMATION INVESTIGATED WITH SBAS-DINSAR APPROACH 
BASED ON PRIOR KNOWLEDGE 
HUANG Qi-huan a ’ *, HE Xiu-feng a 
a Institute of Satellite Navigation & Spatial Information System, Hohai University, Nanjing, P. R. China - (InSAR, 
xfhe)@hhu .edu.cn 
Commission III, WG 1/2 
KEY WORDS: Satellite remote sensing, SAR, Interferometric SAR (InSAR), Deformation analysis, Hazard mapping, Surface 
deformations 
ABSTRACT: 
Although Differential Synthetic Aperture Radar Interferometry (DInSAR) has been successfully used for deformation mapping, it is 
easily suffered from both spatial and temporal decorrelation which limits its application, especially for long-term deformation 
mapping. Based on the DInSAR algorithm, Small BAseline Subset (SBAS) approach, which generates interferograms only between 
two SAR images with small spatial baseline, can reduce both spatial and temporal decorrelation using a data set of SAR images 
acquired subsequently. SBAS-DInSAR approach allows us to generate mean deformation velocity maps and displacement time 
series. In this paper, the surface deformation of the Nanjing City, P. R. China was investigated with SBAS-DInSAR approach. In 
particular, a cubic behaviour for the time variation of the deformation phase signal was assumed as prior knowledge for the 
investigated area, and the deformation parameters obtained in an optimal least square(LS) way directly, simplify the SBAS algorithm 
which linking the interferograms using Singular value decomposition (SVD) method. Results of 13 differential interferograms 
generated by 8 SAR images in Nanjing from August 1996 to April 2000, demonstrate its efficiency. 
1. INTRODUCTION 
Differential Synthetic Aperture Radar Interferometry (DInSAR) 
is a relatively new technique that has been successfully used 
for generating large-scale surface deformation maps on a dense 
grid and with a centimetre to millimetre accuracy (C. Prati, 
1992), within several ten years development, the interest of the 
scientific community about DInSAR approach is now 
progressively moving from the study of a single deformation 
episode to the temporal evolution of the detected deformations, 
t 
ion of temporal and spatial decorrelation as well as atmospheric 
inhomogeneous which limit its accuracy for deformation 
detection; CR approach uses man-made feature called Comer 
Reflector to overcome the decorrelation phenomena at the 
expense of cost, 
therefore ,the most disadvantage of CR approach is its only 
showing pointwise information; The PS approach maximizes 
the number of the acquisitions used generates DInSAR 
interferograms, for each available acquisition even if the 
exploited data pair is characterized by a large baseline (even 
larger than the critical baseline), with respect to a common 
(master) image, and, therefore, affected by baseline 
decorrelation phenomena. It is evident that, in this case, the use 
of all the available data acquisitions is accomplished, but at the 
expense of imaged pixel density; indeed, only those targets that 
exhibit sufficiently high coherence values are considered, and 
their density may be in some cases rather low; this often 
happens, for instance, in nourban areas. 
Small BAseline Subset (SBAS) approach (P. Berardino, 2002) 
uses multiple small baseline (SB) acquisition subsets via an 
/ 
Consequently, the DInSAR algorithm is moving from the 
conventional DInSAR technique to the study of the detected 
high coherent pixels (Usai, S. 2001), such as small man-made 
feature called Comer Reflector (CR) (Xia Y. 2002) and 
Permanent Scatters(PS) (A.Ferretti, 200la,2000b), to generate 
deformation time-series that allows us to follow the evolution 
of the monitored deformations. 
Conventional DInSAR approach has the limita 
effective combination of all the available SB interferograms. 
The combination is based on a minimum-norm criterion of the 
velocity deformation via the application of the Singular Value 
Decomposition (SVD) method. The SBAS approach had two 
key advantages: (1) increased sampling rate by using all the 
acquisitions include in the SB subsets, and (2) preserved the 
capabilities of the system to provide spatially dense 
deformation maps, which being a key issue of conventional 
DInSAR interferometry (F. Casu, 2006). 
This paper investigates SBAS approach; in particular, we 
assume a cubic behaviour for the time variation of the 
deformation phase signal as prior knowledge for the study area, 
the City of Nanjing, P. R. China, this simplified the SBAS 
method; 8 ERS2 SAR images acquired from 1996 to 2000 in 
descending orbit was used to demonstrate the capability of the 
simplified SBAS algorithm. 
The paper is organized as follows: a short overview on the basic 
rationale of the SBAS algorithm is presented. Subsequently, we 
assume a cubic behaviour for the time variation of the 
deformation phase signal as prior knowledge to simplify the 
* School of Civil Engineering Hohai University, No.l Xikang Road, Nanjing, P.R.China;210098; Email: InSAR@hhu.edu.cn