IEEE Geoscience and Remote Sensing Magazine - March 2020 - 146

accurate continuous monitoring in time and space for
EWSs ARE MOST EFFECTIVE
all hazardous landslides in a
WHEN THEY ARE BUILT IN
region, integrating these two
COLLABORATION WITH
systems while minimizing
THOSE AT RISK, RATHER
the associated costs by limTHAN IMPOSED FROM
iting the number of in situ
sensors. High-rate (e.g., 1 Hz)
OUTSIDE.
raw in situ observations (e.g.,
GNSS and crackmeters) can be
transmitted to a data center via
wireless communication infrastructure and are processed
in real time with short baselines in a kinematic mode.
Recent experiments with GNSS suggest that ~2-4 mm
horizontal and 4-8 mm vertical accuracy is possible
at 1 Hz [63], [65]. Real-time monitoring is particularly
important since existing observations on tertiary creep
-suggest that the timescale for this phase ranges from
minutes to months [44], [65], [66]. Thus, the data should
be transmitted back to the data center in real time and
processed automatically. However, these in situ observations are not only useful for identifying the onset of
tertiary creep; they can also be used in the secondary
phase to determine the sensitivity of landslide motion
to external factors at a higher resolution and precision
than was possible in Stage 1 [23], [56]. The mechanical
models introduced in Stage 1 can be refined and cali-

TABLE 1. COMMONLY USED TECHNOLOGIES FOR
LANDSLIDE MONITORING.
OBSERVATION
TYPES
TECHNOLOGY

PRECISION

EXAMPLES

Displacement

mm-cm [68]

[21], [25], [69],
[70]

Pore pressure

Precipitation

Spaceborne InSAR
Airborne InSAR

mm-cm [71]

[71], [72]

Ground-based InSAR

mm-cm [73]

[66], [73], [74]

UAV photogrammetry

~6cm [75]

[75], [76]

GNSS

mm-cm [77]

[70], [84]

Optical image matching cm-m [78]

[78], [79]

Crackmeter

[81], [82]

Extensometer

~3 mm [84]

[84], [85]

In-place inclinometer

~8 mm [68]

[10], [86], [87]

Tiltmeter

~0.1° [13]

[13], [82], [90]

Total station

~±1 ppm [80]

[80], [88]

Terrestrial Lidar

~0.2-0.5 m [83] [83], [89]

Shape acceleration
array

±1.5 mm/
30 m [90]

[13], [83], [90]

Active waveguides

mm [91]

[13], [91]

Seismometer

-

[92], [93]

Piezometer

-

[13], [94], [95]

TDR

-

[96], [97]

Tensiometer (soil
hygrometer)

-

[57], [97]

Rain gauge

-

[82], [98]

TDR: time domain reflectometry.

146

mm-cm [80]

brated by monitoring environmental factors and geological-geotechnical parameters such as pore pressure
in soils (Table 1) [13], [67].
◗◗ Step 4: The ultimate objective of an EWS is to communicate through timely and useful warnings to the people
in local communities who are exposed to a landslide
hazard. Thus, engagement and communication with
local communities should be a key feature of an effective landslide EWS. A large body of work on the social
science of early warning already exists that provides
useful insights, explanations for unexpected EWS failure and potential secondary disasters, and examples
of good practice. Experience from past disasters worldwide suggests that emergency preparedness, planning,
and response are some of the weakest elements in many
existing EWSs [99]. In particular, the link between the
technical capacity to issue a warning and the public's
capacity and commitment to respond effectively to the
warning is often weak, which limits the warning's ability to t- rigger an a- ppropriate and effective response from
the community. Warning systems that mainly focus on
technical aspects and ignore social factors generally
do not work effectively because the warnings do not
prompt effective action due to a lack of community buyin and to poor engagement and operation results. Both
academics and practitioners widely agree that EWSs are
most effective when they are built in collaboration with
those at risk, rather than imposed from outside.
OUTLOOK
There are three big questions for landslide forecasting and
early warning to address:
Big question 1: Where are potential landslides? We are entering an exciting new era of EO data. Recent advances in
satellite radar and in situ sensors (e.g., GNSS) have allowed
us to collect high-quality measurements to quantify Earth's
surface displacements and then address this question over
entire mountain ranges, at space and time scales that are
finer than ever before and at a relatively low cost. In an
EO-based landslide EWS, the relatively short repeat cycles
of current SAR missions still represent InSAR's limitation
to detect potential landslides. However, the Geosynchronous Continental Land-Atmosphere Sensing System, one
of three ideas for Earth Explorer accepted by ESA's Program
Board for EO to compete as the tenth Earth Explorer mission,
might provide a solution. Considerable work has been done
to interferometrically process massive SAR data sets in an
automatic way [100], but more should be done to investigate how to detect potential landslides from big SAR data
in a consistent, reliable, and smart manner. Machine learning technologies have been widely implemented in the
field of computer science and remote sensing [101], [102],
where statistical techniques are employed to learn specific
and complex tasks from given data. Recent studies report
that machine learning can identify signals associated with
geohazards from large data sets [103], which suggests that
IEEE GEOSCIENCE AND REMOTE SENSING MAGAZINE

MARCH 2020
IEEE Geoscience and Remote Sensing Magazine - March 2020

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