IEEE Geoscience and Remote Sensing Magazine - March 2020 - 143

[39]. Since the Yellow River pumping irrigation project began in 1966, frequent landslides have occurred on the terrace
margins. The Dangchuan 4# landslide is in southwest-central Heifangtai near Guoxia, Yongjing County. Among the in
situ sensors, a crackmeter installed across the trailing head
scarp edge of Dangchuan 4# [Figure 3(a)] provided critical
displacement measurements in real time, which were used
in a successful 8-h early warning in 2017.
The crackmeter observations showed a clearly accelerated displacement rate at Dangchuan 4# on 23 August
2017 [Figure 3(b)]; hence, a yellow warning was issued to
the village leader and local government by text message,
informing them to "pay close attention to this slope and
prepare for disaster prevention." After a detailed field investigation, the local government confirmed the warning and
released an official landslide warning announcement to local communities on 23 September 2017 with several alert
boards posted around the landslide area [Figure 3(c)]. On
27 September 2017, the yellow warning was upgraded to
an orange warning due to the accelerating displacement
rate measured at the crackmeter. At 17:50 on 30 September
2017, the system (a geohazard real-time monitoring and
EWS [40] developed by SKLGP) automatically released a
red warning, which was confirmed by a panel of experts.
Three hours later, at 20:55 on 30 September 2017, an official red warning was issued to the local government [Figure 3(d)], prompting a government-led emergency response
and evacuation. The local government immediately started
its emergency response, and more than 20 villagers in the
landslide hazard zone were evacuated. At 05:00 on 1 October 2017, a landslide occurred [Figure 3(e)], damaging
several buildings, but there were no casualties owing to the
early warning [32].
This successful case clearly demonstrates the potential
importance of real-time displacement measurements and
the role that in situ sensors could play in EWSs. A preliminary retrospective InSAR study showed that InSAR with
L-band Advanced Land Observing Satellite-2 images were
able to capture the accelerated movements that occurred
15 days before the landslide (Figure 4).
DISCUSSION
FEASIBILITY AND COMPLEMENTARITY OF EARTH
OBSERVATION FOR LANDSLIDE EARLY WARNING
A range of laboratory, field, and theoretical studies have
identified prefailure creep acceleration of landslides and
suggest that it can be divided into three phases [41]-[47]:
primary creep, secondary creep, and tertiary creep [Figure 5(a)]. Primary creep is characterized by a decreasing
strain rate over time, which often lasts for a short period or can be even absent in some cases [42]. Secondary creep is characterized by slow movement at a nearly
constant rate, but with fluctuations in real slopes due
to the influence of external factors such as rainfall. The
duration of the secondary creep is difficult to estimate;
MARCH 2020

IEEE GEOSCIENCE AND REMOTE SENSING MAGAZINE

it can last for months, years, or even decades [42], [48]
despite continuous displacement during this phase. Tertiary creep is characterized by a rapid acceleration of displacement until final failure [49]. Although such speedups may be common prior to catastrophic failure events
[48], the number of actual observations of such speedup
behavior remains limited due to the absence of the right
EO technologies in the right locations at the right times.
Therefore, there are two primary challenges for landslide
early warning: 1) monitoring surface displacements over
a wide region with sufficient resolution and accuracy to
identify areas undergoing
secondary creep and 2) identifying when or under what
EXPERIENCE FROM PAST
circumstances a slow-movDISASTERS WORLDWIDE
ing landslide (i.e., in secondSUGGESTS THAT
ary creep phase) enters the
accelerated displacement of
EMERGENCY
a tertiary creep phase, leadPREPAREDNESS, PLANNING,
ing to rapid failure.
AND RESPONSE ARE SOME
Advances in EO offer the
OF THE WEAKEST ELEMENTS
potential to address these two
IN MANY EXISTING EWSs.
challenges. In the primary and
secondary phases, weekly to
monthly observations would
be enough to distinguish areas undergoing more rapid creep.
In the tertiary creep phase, subdaily sampling intervals are
needed to capture the accelerated creep [Figure 5(b)]. InSAR
currently has a shortest repeat cycle of 1-11 days while GNSS
and some other in situ sensors can provide high-rate (e.g.,
1-20-Hz) measurements. Only slow tertiary creep displacements (e.g., < 0.012 m/day over a distance of 100 m for Sentinel-1[50]) could potentially be captured by InSAR because its
measuring capability is limited by the spatial displacement
gradients. This limitation can be overcome using SAR pixel
offset tracking [19] or the range split-spectrum interferometry-assisted phase unwrapping method [50]; in situ sensors generally do not have such limitations [Figure 5(c)]. On
the other hand, InSAR offers extensive spatial coverage that
enables the detection of potential landslides in the primary
and secondary creep phases. To monitor a single slope in its
tertiary phase, InSAR and in situ sensors can provide complementary coverage in space and time.
EARTH OBSERVATION-BASED LANDSLIDE EWS
Figure 5 illustrates EO's ability to provide unprecedented
and encouraging opportunities for prefailure creep monitoring. However, the different technologies have their own
advantages and limitations, as illustrated by Xinmo and
Dangchuan's case studies. A single EO method is insufficient to capture all the signals in the different creep
stages, so multiple EO technologies should be combined
to develop a landslide EWS. Figure 6 shows the framework of an operational landslide EWS that relies on an
optimal combination of these EO technologies, detailed
as follows.
143

IEEE Geoscience and Remote Sensing Magazine - March 2020

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