Error-Regulated Multi-Pass DInSAR Analysis
for Landslide Risk Assessment
Jung Rack Kim, HyeWon Yun, Stephan van Gasselt, and YunSoo Choi
Abstract
Landslide risk assessment based on Differential Interferomet-
ric
SAR
analyses (
DInSAR
) is associated with a number of error
effects. We here approach the problem of assessing landslide
risks over mountainous areas, where
DInSAR
observations are
often influenced by orographic effects and inaccurate base
topographies by employing a dedicated error compensation.
In order to obtain accurate information on surface deforma-
tion, we apply corrections for
DInSAR
interferograms using
high-resolution base topography and water vapor informa-
tion obtained from a satellite radiometer. We observe that
the corrected
DInSAR
output is in accordance with the en-
vironmental context as inferred by geological and geomor-
phological settings. It is feasible to better quantify landslide
monitoring schemes whenever high- accuracy atmospheric
error maps and a methodology to effectively compensate for
external errors in
DInSAR
interferograms are available. The
approach in this study can be further upgraded for future
SAR
-based assessments and various error correction ap-
proaches for even more precise landslide risk assessments.
Introduction
Remote sensing tools allow to efficiently monitor disasters
and assess potential natural hazards that might be caused as
consequence of global climate change and intensified anthro-
pogenic activities in areas that pose challenges for infra-
structure, building construction, or agriculture. Among those
techniques,
Differential Interferometry Synthetic Aperture
Radar
(
DInSAR
) has been effectively applied for natural and
artificial risk assessments including monitoring of seismic
activity (Wright
et al
., 2003; Fialko
et al
., 2005; Kobayashi
et
al
., 2011), land subsidence (Galloway
et al
., 1998; Osmanoğlu
et al
., 2011), and landslides (Ye
et al
., 2004; Yin
et al
., 2010;
Zhao
et al
., 2012). Since Zebkar and Goldstein (1986) have
shown the possibility of tracing topographic deformation
using
SAR
phase-angle difference,
DInSAR
techniques, includ-
ing recent time-series analyses such as
Permanent Scatterers
(
PS
), Ferretti
et al
., 2000) and
Small Baseline Subset
(
SBAS
),
Berardino
et al
., 2002; Lanari
et al.
, 2004), have evolved to
provide high accuracies. Landslide monitoring using
DInSAR
techniques is challenging considering that only limited moni-
toring activities are conducted by local authorities despite the
considerable threat to population and buildings.
For our
DInSAR
survey, we investigated cases of rock falls
and wall-rock failures near the coastline of the eastern Korean
peninsula. The failures are mainly associated with wall-rock
cuts introduced by extensive road and building construction.
Wall rock is commonly stabilized by rock bolt arrays while
constructions are protected against rock-fall by prevention
nets, gabions, or extensive shotcrete measures.
Wall-rock failure and slope instabilities usually occur in
areas of steep and high-frequency slopes. When processing
DInSAR
interferograms for the assessment of locations of poten-
tial failure, such topographies pose considerable challenges as
chances of introducing errors due to inaccurate Digital Eleva-
tion Models (
DEM
) are amplified. In addition, weather effects,
in particular local variations of atmospheric humidity, can
cause significant signal errors in remote sensing which are not
assessable in a straightforward way due to their complex char-
acteristics. Such conditions cause additional challenges when
trying to decouple surface movement preceding a landslide
event, as measured from the observed
DInSAR
phase angle.
In this study, we explore the possibility of landslide risk
monitoring using a conventional two-pass
DInSAR
technique
with an error regulation method. We compare our results
to the
Stanford Method for Persistent Scatterers
(
StaMPS
),
(Hooper, 2008), which is based on sequences of techniques
tracing the stable scatterers to extract ground displacement
more precisely.
Landslide Monitoring Using Remote Sensing Data
Landslide forecasting and hazard assessments have common-
ly been conducted using three major approaches:
1. probabilistic analysis of climate and seasonal effects, in
particular effects of precipitation (Lee and Talib, 2005;
Jibson
et al
., 2000),
2. dynamic modeling of potential landslide sites (Hungr,
1995; McDougall and Hungr, 2004), and
3. synoptic spatial analyses of potentially relevant data sets
(Liu
et al
., 2004 and 2012).
Remote sensing data have been frequently employed as
information complementary to field data, in order monitor
landslide risks and locations. Information about larger-area
topography, vegetation, and soil can be effectively collected
from spaceborne sensors and exploited by landslide risk as-
sessment. Li
et al.
(2014) showed that the automatic detection
of changes in remotely sensed imagery can be used to map
landslides. In that context, using
LIDAR
data seems to be a
highly promising tool to assess landslide risk, as shown by
both Chu
et al
. (2015), who used a variety of interpolation
Jung Rack Kim, YunSoo Choi and Stephan van Gasselt,
University of Seoul, Department of Geoinfomatics,
Seoulsiripdae-ro, Dongdaemun-gu, Seoul, 02504, Republic of
Korea.
HyeWon Yun, Korean National Disaster Management
Research Institute, Disaster Information Research Division,
365 Jongga-ro, Ulsan, 44538, Republic of Korea.
Stephan van Gasselt, now at: National Chengchi University,
Department of Land Economics, No. 64, Sec 2, ZhiNan Rd.,
Taipei 11605, Taiwan
).
Photogrammetric Engineering & Remote Sensing
Vol. 84, No. 4, April 2018, pp. 189–202.
0099-1112/17/189–202
© 2018 American Society for Photogrammetry
and Remote Sensing
doi: 10.14358/PERS.84.4.189
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
April 2018
189
