IEEE Geoscience and Remote Sensing Magazine - March 2020 - 137

- etect deformation precursors to catastrophic landslides
d
and that early warnings can be achieved with real-time, in
situ observations. We propose a novel and exciting framework that employs EO technologies to build an operational landslide EWS.
INTRODUCTION
Landslides, when soil or rock moves down a slope, have
been shaping mountainous regions for millennia, but today they pose a destructive hazard to people and infrastructure that results in hundreds of deaths and billions of dollars in damages every year [1]. The combination of a rapidly
increasing global population and the intensifying weather
extremes associated with recent climate change suggests
that landslide risk will dramatically increase over the next
decade. Landslide deformation can be extremely slow (a
few millimeters per year) or involve sudden failure [2], so
their hazards include both enduring damage to manmade
structures and catastrophic destructive events.
While small landslides make up the vast majority of
landslide events in any given year, large landslides tend to
be responsible for most damage and loss of life [3]. Current
landslide risk mitigation strategies tend to reduce exposure,
the likelihood that someone or something is impacted by
a landslide, primarily by moving to, or locating infrastructure in, less hazardous locations. However, asset relocation
is not feasible for most people. In these situations, shortterm evacuation is often the most attractive or only option.
Therefore, improved landslide forecasts and early warning
capabilities are expected to be crucial in managing landslide risk for many individuals and communities.
Although major landslide triggers (e.g., rainfall and seismic shaking) and the basic physics governing landslide initiation are well known, predicting where and when landslides
will occur remains a challenge, primarily due to the difficulty in forecasting the triggering factors themselves as well
as the spatial variations in earth materials and slope conditions. Existing forecasting methods generally involve functional relationships between trigger-factor intensity (e.g.,
precipitation history and peak seismic ground acceleration)
and landslide probability. However, the connection between
triggers and landslides is complex; some landslides occur
without an identifiable trigger and others with significant
delay. For example, the 2006 Leyte landslide, which killed
more than 1,100 people in the Philippines, occurred five
days after a large rainstorm. Although the population was
initially evacuated, they had returned to their homes before
the landslide occurred [4]. Displacements recorded over time
could provide critical additional information for predicting
the possible timing of impending slope failure [5].
Based on conventional in situ survey methods, the concept of landslide EWSs has been proposed for several years
[6]-[12]. These works often result in suggested warning criteria for specific locations. Successful early warning cases,
in which a clear warning was given prior to catastrophic
slope failure, have been very limited due to the inadequate
MARCH 2020

IEEE GEOSCIENCE AND REMOTE SENSING MAGAZINE

temporal and spatial precision of ground observations
[13]. Building trustworthy real-time EWSs that can identify
when to prompt short-term evacuations with suitable spatial and temporal precision is important but difficult.
Spaceborne synthetic aperture radar (SAR) sensors emit
radar signals and record the amplitude of the backscattered
signal as well as the phase from which the changes in range
between the satellite and Earth's surface can be inferred [14].
Interferometric SAR (InSAR) is a powerful tool for measuring Earth's surface motion over large regions [15]-[17], in all
weather conditions, at meter resolution. InSAR also offers
the capability to remotely monitor unstable slopes [18]-[21].
Recent studies have demonstrated that conventional InSAR
and related time-series techniques (e.g., persistent scatterer InSAR and small baseSUCCESSFUL EARLY
line InSAR) can identify, map,
WARNING CASES HAVE
and monitor active landslides
BEEN VERY LIMITED DUE TO
[22]-[26] and even detect precursory deformation signals
THE INADEQUATE
prior to their eventual failure
TEMPORAL AND SPATIAL
[27]-[29]. Note that spacePRECISION OF GROUND
borne InSAR currently has a
OBSERVATIONS.
minimum repeat cycle of six
days for Sentinel-1, one day for
COSMO-SkyMed [30], 11 days
for TerraSAR-X, and longer for other satellites, which represents a major limitation of spaceborne InSAR for EWSs.
In situ global navigation satellite system (GNSS) monitoring can measure 3D landslide motion at a very high temporal frequency (e.g., 20 Hz) and spatial accuracy (2-4 mm in
plan and 4-8 mm in vertical) [31]. Other in situ monitoring
methods include extensometers, inclinometers, and pore
water pressure sensors. However, these methods provide
point-based measurements only at sensors, which are
costly to install and maintain. Thus, in situ obser vations
are limited by the number of sensors that can be deployed
at key locations and may not capture the spatial variations
in landslide motion prior to failure. There are two obvious
hurdles to deploying ground-based monitoring techniques:
sites with potential landslides should be detected prior to
their failure, and key monitoring locations in the landslide
bodies should be identified.
Spaceborne InSAR and in situ sensors are complementary tools to monitor surface displacements given InSAR's
high spatial resolution (meters to tens of meters) over a
wide region (e.g., 250 km × 250 km for Sentinel-1) but are
limited by temporal resolution (constrained by the frequency of satellite overpasses) and in situ sensors' fine temporal
resolution at their locations. We suggest that it is now both
feasible and timely to combine these EO technologies to
build an integrated landslide EWS. In this article, the 2017
Xinmo landslide in Sichuan, China, is used to demonstrate
the ability of spaceborne InSAR to identify precursory landslide deformation, while the 2017 Dangchuan 4# landslide
in Heifangtai (Gansu, China) is used to demonstrate the
137

IEEE Geoscience and Remote Sensing Magazine - March 2020

Table of Contents for the Digital Edition of IEEE Geoscience and Remote Sensing Magazine - March 2020