101
MAPPING GROUND DEFORMATION BY RADAR INTERFEROMETRY BASED ON
PERMANENT-SCATTERER NETWORK: ALGORITHM AND TESTING RESULTS
Guoxiang Liu 3, *, S. M. Buckley b , Xiaoli Ding c , Qiang Chen 3 , Xiaojun Luo 3
a Dept. of Surveying Engineering, Southwest Jiaotong University, Chengdu, China - rsgxliu@swjtu.edu.cn
b Centre for Space Research, The University of Texas at Austin, Austin, Texas, USA - sean.buckley@mail.utexas.edu
c Dept. of Land Surveying and Geo-Informatics, The Hong Kong Polytechnic University, Hong Kong, China -
lsxlding@polyu.edu.hk
KEY WORDS: Ground Deformation, Radar Interferometry, PS Networking, Atmospheric Signal, Empirical Mode Decomposition
ABSTRACT:
The full operational capability of synthetic aperture radar (SAR) interferometry in deformation monitoring has not been achieved yet
due to the negative influences of spatio-temporal decorrelation and atmospheric delay. With the use of time series of SAR images,
deformation extraction can be however improved by only tracking some objects with steady radar reflectivity, generally referred to
as permanent scatterer (PS). This paper presents an attempt to explore a PS-networking approach to isolate deformations from other
effects such as atmospheric signals and topographic errors. The deforming process in time and space is modelled and estimated with
a very strong network which is formed by connecting adjacent PSs. The linear deformations and topographic errors are estimated by
optimizing objective functions and by adjusting the network via weighted least squares (LS) solution. The time series of nonlinear
deformations and atmospheric signals are computed by singular value decomposition (SVD) and empirical mode decomposition
(EMD). To validate the algorithm, 39 ERS C-band SAR images acquired over Phoenix in Arizona (USA) from 1992 to 2002 are
used to detect land subsidence caused by the excessive groundwater withdrawal.
1. INTRODUCTION
Existing studies have shown that there are two major limitations
in conventional differential SAR interferometry (DInSAR) for
land deformation monitoring, i.e., spatio-temporal decorrelation
and atmospheric artifacts (e.g., Zebker & Villaseno, 1992;
Buckley, 2000; Liu, 2003; Ding et al., 2004). To mitigate such
drawbacks, many research efforts have been made in recent
years to explore various techniques for detecting the temporal
evolution of deformations using time series of SAR images.
A strategy proposed early is to stack the multiple
interferograms (Sandwell & Price, 1998). Ground deformation
analysis can be therefore improved by enhancing fringe clarity
and suppressing atmospheric effects. Afterwards, a very generic
approach, called permanent scatters (PS) technique, was
proposed to extract both linear and nonlinear deformations from
a set of interferograms by isolating atmospheric effects and
topographic errors (Ferretti et al., 2000, 2001). Since PSs are
usually some hard objects such as buildings and rocks, they can
remain temporal coherent radar reflectivity, and thus facilitating
deformation extraction on the basis of PSs’ phase measurements
with high signal-to-noise ratio (SNR). Subsequently, another
effective approach, called small-baseline subset (SBAS) method,
was developed to further decrease the negative influences due
to decorrelation noise and bias (Berardino et al. 2002).
PS technique suffers from spatial decorrelation as some long-
spatial baselines may result in by sharing a unique master image
in forming interferometric combinations, while SBAS technique
suffers from errors caused by full-resolution phase unwrapping.
However, the two techniques can complement each other (Mora
et al. 2003). Combining the merits of PS and SBAS technique,
this paper presents an improved algorithm to isolate and extract
deformations, topographic errors and atmospheric signals with a
very strong network formed by freely connecting neighbouring
PSs. To maximize coherence of all the SAR dataset, the spatial
and temporal baseline thresholding are applied when forming .
interferometric combinations. The phase modelling is based on
the network. The linear deformation velocities and topographic
errors are first estimated by optimizing an objective function of
each arc (a connection of two PSs) and adjusting the network by
LS solution. Time series of phase measurements at each PS is
then reconstructed by singular value decomposition (SVD) and
decoupled into nonlinear deformations and atmospheric signals
by a relatively new signal analysis method - empirical mode
decomposition (EMD), proposed by Huang et al. (1998). For
algorithm validation, some experiments have been carried out to
analyze subsidence evolution in Phoenix metropolitan with 39
ERS-1/2 C-band (wavelength of 5.6 cm) SAR images acquired
1992 through 2000.
2. DATA PREPROCESSING AND PS NETWORKING
2.1 Data Preprocessing
Suppose that deformation analysis is based on N+1 SAR images
acquired at the ordered times (t 0 ,t,,A ,t N ) over the same area.
An interferometric combination is acceptable if and only if its
temporal baseline is below a given threshold (e.g., 4 years) and
its spatial baseline is also below a given threshold (e.g., 120 m
for ERS SAR). Let us assume that M interferograms may be
formed in this way. Prior to further analysis, several procedures
are necessarily performed to compute differential
interferograms.
Since co-registering SAR imagery is a key prerequisite for any
* Corresponding author.