Professional Documents
Culture Documents
Engineering Geology
journal homepage: www.elsevier.com/locate/enggeo
a r t i c l e
i n f o
Article history:
Received 9 July 2013
Received in revised form 6 December 2013
Accepted 1 March 2014
Available online 15 March 2014
Keywords:
Multi Temporal Interferometry
Satellite
Landslides
Unstable slopes
Technical constraints
Interpretation issues
a b s t r a c t
Multi Temporal Interferometry (MTI) stands for advanced synthetic aperture radar differential interferometry
(DInSAR) techniques, which include Permanent/Persistent Scatterers Interferometry PSInSAR/PSI and
similar methods, as well as Small Baseline Subset SBAS and related/hybrid approaches. These techniques are
capable to provide wide-area coverage (thousands of km2) and precise (mmcm resolution), spatially dense information (from hundreds to thousands of measurement points/km2) on ground surface deformations. New MTI
application opportunities are emerging thanks to i) greater data availability from radar satellites, and ii) improved capabilities of the new space radar sensors (X-band Cosmo-SkyMed, C-band RADARSAT-2, TerraSAR-X)
in terms of resolution (from 3 to 1 m) and revisit time (from 11 to 4 days for X-band acquisitions). This implies
greater quantity and quality information about ground surface displacements and hence improved landslide detection and monitoring capabilities. Even though the applicability of MTI to regional and local-scale investigations
of slow landslides has already been demonstrated, the awareness of the MTI utility and its technical limitations
among landslide scientists and practitioners is still rather low. By referring to recent works on radar remote sensing, many regional and local scale MTI application examples from the geoscience literature and our own studies,
we present an up-to-date overview of current opportunities and challenges in this eld. We discuss relevant
technical constraints and data interpretation issues that hamper the use of MTI in landslide assessment. Then
guidelines on how to mitigate MTI technical limitations and avoid erroneous interpretations of radar-derived
slope surface deformations are presented for the benet of users lacking advanced knowledge in SAR applications. Finally, in view of the upcoming radar satellite launches, future perspectives on MTI applications are
outlined and recommendations for applied research priorities are suggested. We foresee that with regular
globe-scale coverage, improved temporal resolution (weekly or better) and freely available imagery, new
radar satellite background missions such as the European Space Agency's Sentinel-1 will guarantee ever increasing and more efcient use of MTI in landslide investigations. Furthermore, thanks to the improved temporal and
spatial resolutions of the new generation radar sensors, signicant breakthroughs are expected in detailed slope
instability process modeling (e.g. kinematic and geotechnical models), as well as in the understanding of spatial
and temporal patterns of landslide movement/activity and their relationships to causative or triggering factors
(e.g. precipitation, seismic loading).
2014 Elsevier B.V. All rights reserved.
Contents
1.
2.
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Background on DInSAR and Multi Temporal Interferometry . . . . . . . . .
2.1.
DInSAR and MTI . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2.
MTI techniques . . . . . . . . . . . . . . . . . . . . . . . . . .
2.3.
Technical and practical issues of MTI applied to landslide investigations .
2.3.1.
CT selection, coherence threshold, number of images . . . . .
2.3.2.
CT density and articial reectors (AR) . . . . . . . . . . .
2.3.3.
CT geocoding accuracy . . . . . . . . . . . . . . . . . .
2.3.4.
Different types of CT . . . . . . . . . . . . . . . . . . .
2.3.5.
Deformation model . . . . . . . . . . . . . . . . . . . .
http://dx.doi.org/10.1016/j.enggeo.2014.03.003
0013-7952/ 2014 Elsevier B.V. All rights reserved.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
104
105
105
107
107
107
108
108
108
109
104
2.3.6.
Aliasing problem and maximum detectable velocity . . . . . . . . . . . . . . . . . . . . .
2.3.7.
Reference point selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.3.8.
3D deformations vs. LOS measurements . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.3.9.
Geometrical distortions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.3.10. SAR missions and data characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.3.11. Performance of L/C/X band MTI in terms of movement monitoring . . . . . . . . . . . . . . .
2.3.12. High resolution X-band data and new application possibilities . . . . . . . . . . . . . . . . .
2.3.13. MTI measurement precision . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.4.
Quality, reliability and validation of MTI products (assessments using cross-comparisons and ground truth)
2.4.1.
General . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.4.2.
Factors that inuence the quality of MTI products . . . . . . . . . . . . . . . . . . . . . .
2.4.3.
Assessments of MTI results quality using cross-comparisons and ground truth . . . . . . . . . .
3.
MTI applications in landslide investigations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.1.
Feasibility assessments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.1.1.
General . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.1.2.
A priori assessments of MTI applicability . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.1.3.
Land cover/use, lithology and CT density . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.1.4.
Sensitivity to down-slope movements . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.1.5.
Example of landslide visibility and sensitivity to down-slope movement maps . . . . . . . . . .
3.2.
Regional scale MTI investigations of landslides . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.2.1.
General . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.2.2.
Examples of regional scale ERS-based MTI studies of landslides . . . . . . . . . . . . . . . .
3.2.3.
Applicability potential of MTI based on ERS and other C-band data . . . . . . . . . . . . . . .
3.3.
Catchment-wide and local scale (slope specic) MTI investigations of landslides . . . . . . . . . . . . .
3.3.1.
Example of catchment scale MTI investigation of landslides in the US . . . . . . . . . . . . . .
3.3.2.
Examples of catchment scale MTI investigations of landslides in Europe . . . . . . . . . . . . .
3.3.3.
Example of catchment scale MTI investigation of landslides in Gansu, China . . . . . . . . .
3.4.
Local scale (slope specic) MTI investigations of landslides . . . . . . . . . . . . . . . . . . . . . . .
3.4.1.
Madesimo case study, Central Alps, Italy . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.4.2.
BindoCortenova case study, Central Italian Pre-Alps . . . . . . . . . . . . . . . . . . . . .
3.4.3.
Saint Moritz case study, Swiss Alps . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.5.
Monitoring infrastructure stability in landslide-prone areas . . . . . . . . . . . . . . . . . . . . . .
4.
Limitations of MTI applications in landslide investigations and user guidelines . . . . . . . . . . . . . . . . .
5.
Summary discussion and future research priorities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1. Introduction
Although the rst application of space-borne Synthetic Aperture
Radar Differential Interferometry (DInSAR) to a landslide investigation
dates back to the mid-1990 (Fruneau et al., 1996), it was only in the
following decade that this technique captured the attention of the landslide community. This was prompted by i) the availability of sufciently
long series of radar images acquired in particular by the European Space
Agency (ESA); ii) the development of the Multi Temporal Interferometry (MTInSAR or MTI) methods (e.g. Permanent Scatterers Interferometry PSInSAR, Ferretti et al., 2001, Small Baseline Subset SBAS,
Berardino et al, 2002; Usai, 2003), which overcome some of the limitations of conventional differential SAR interferometry (DInSAR) and
iii) the presentation of several successful examples of MTI applications
to landslide investigations (e.g. Colesanti et al., 2003a; Hilley et al.,
2004; Colesanti and Wasowski, 2006).
Indeed, MTI techniques are attractive, in that with wide spatial
coverage and regular satellite re-visits over the same areas they offer
the possibility of continuous, long-term monitoring (several years). Furthermore, retrospective studies are possible by exploiting radar imagery
archives now spanning over 20 years.
Referring to the exploitation of ERS satellite medium resolution data,
in a review of PSInSAR applications for landslide assessment Colesanti
and Wasowski (2006) pointed out the following advantages of the
technique:
Cost-effectiveness for wide-area applications (hundreds and thousands
of km2)
High precision (mm) of measurements (comparable to or better
than GPS)
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
109
109
109
109
110
110
110
111
111
111
111
111
112
112
112
114
115
115
116
117
117
117
118
118
119
119
120
123
123
125
127
131
131
132
136
136
further signicant advances in the eld of MTI-based landslide investigations and monitoring.
Nevertheless, the engineering geology community appears to be not
yet fully aware of the actual potential and limitations of MTI as applied
to landslide investigations. Some persisting skepticism, especially
among practitioners and end users, may have resulted from the initial
tendency to overlook the difculties in the practical exploitation of the
technique for the detection and monitoring of unstable slopes. A more
balanced picture is needed considering also that scientic literature is
generally dominated by examples of successful studies, often based on
large quantities of radar imagery and regarding favorable environmental settings. This work addresses the above issues by providing an up
to date critical review of current opportunities and challenges of
MTI-based landslide investigations. We assume that by now many
landslide scientists are (or could easily become) familiar with
space-borne SAR, DInSAR and PSInSAR, since relevant information
has already been provided in two Special Issues of Engineering
Geology focused on remote sensing of landslides and unstable slopes
(e.g. Colesanti et al., 2003a; Colesanti and Wasowski, 2006). Nevertheless, numerous new, up-to-date references are made to remote
sensing literature that covers theoretical and practical aspects of
radar interferometry, as well as to the applied earth science literature. We draw also from the outcomes of international conferences
(e.g. Fringe workshops; ENVISAT symposia) focused on the science
and applications of SAR interferometry, which are regularly organized
by the European Space Agency (ESA).
The paper begins with a very brief presentation of principles of radar
interferometry. This is followed by an updated, thorough overview of
the recent advances in MTI methods, considering both technical and
practical issues that are relevant in the applications focused on slope
instability investigations. Then, we present different examples of
regional to local scale MTI applications to landslide investigations in various geomorphologic and climatic settings; here we rely both on our
own investigations conducted in Mediterranean and central Europe, as
well as in China, and also on a review of relevant experiences reported
in literature by other research groups. The cases reported from literature
rely on the use of different advanced processing algorithms, whereas our
own examples are based on the MTI results derived using the SPINUA algorithm (Bovenga et al., 2005). To better highlight the current limitations in the MTI applicability and relevant interpretation issues, rather
than presenting our ideal case studies with straightforward analyses
and conclusions, we mainly show and discuss those examples where
auxiliary data (detailed geology/geomorphology, ground monitoring
data) and/or local/direct knowledge are limited and the assessment
of MTI results is challenging. Such situations will likely be common in
practice, considering that MTI is best exploited (and represents an attractive investigation opportunity) in wide-area applications where
ground monitoring data are typically scarce or unavailable.
Afterwards, focusing on practical aspects, of interest to engineering
geology community, we re-assess i) the limitations of the MTI techniques, as well as their potential to provide, under suitable conditions,
valuable reconnaissance and site-specic information on slope surface
deformations, ii) the challenges in inferring the exact cause(s) of slow
displacements (mmcm/year) commonly registered on radar targets
and the risks of misinterpretation. Furthermore, specic recommendations and a set of guidelines are offered on how to mitigate limitations
of MTI applied to landslide and slope instability investigations and
how to avoid erroneous interpretations of ground deformation measurements obtained from SAR data.
It is argued that to foster a more protable use of these techniques in
landslide investigations additional progress needs to be made in our understanding of small ground surface deformations detected from SAR
data and in their integration with ground-based information. This is
one of the research topics that deserve more attention and effort from
the engineering geology community. Finally, with reference to the upcoming radar satellite launches, future MTI application opportunities
105
Fig. 1. Sketch of two SAR acquisitions occurring at successive satellite passes: R is the
sensor-target geometrical distance, h is the target elevation, dh is the target displacement,
and Ratm is the additional path induced by the atmosphere.
106
InSAR phase data can be used to produce 3D images of the Earth surface. Furthermore, by isolating the topographic contribution from the
interferometric phase by using independent elevation model, it is possible to detect and measure possible ground surface movements that
occurred between the SAR image acquisitions along the LOS (dRdisp);
this is the basic principle of DInSAR (Hanssen, 2001). From Eq. (1) the
phase sensibility to the movement is /2 thus meaning that, since the
wavelength of the different SAR bands ranges from about 3 to 20 cm
Table 1
Selected characteristics of principal Synthetic Aperture Radar (SAR) sensors.
Satellite
mission
Wavelength (cm)
Life status
Resolution
az./range (m)
Repeat
cycle (days)
Swath
width (km)
Max. vel.
(cm/yr)
Incident
Angle(o)
C-band
ERS-1/2
5.6
19922001
6/24
35
ENVISAT
5.6
20032010
6/24
35
100
14.6
23
100
14.6
1944
20.4
2050
20.4
2050
85
3046
163.2
2055
45 (fine)
100 (strip)
200 (scan)
10 (spot)
40 (strip)
200 (scan)
RADARSAT-1
5.5
1995
830
24
RADARSAT-2
5.5
2007
3/3
8/8
26/25
24
Sentinel-1
5.6
20142024
520
6, 12
250
RADARSAT
Constellation
Mission
(3 Sat)
5.5
20182026
550
3, 12
30350
L-band
J-ERS
23.5
19921998
18
44
75
48.7
35
ALOS
PALSAR
23.6
20062011
5/788
46
4070
46.8
860
22.9
20132017
1/3
310/310
100/100
14
25 (spot)
5070 (strip)
350 (scan)
149.2
870
23.5
20142021
1050
8, 16
20150
268
2050
ALOS PALSAR-2
SAOCOM
(2 Sat)
X-band
COSMOSkyMED
(4 Sat)
3.1
20072014
2.5/2.5
1.0/1.0
2, 4, 8, 16
10 (spot)
40 (strip)
200 (scan)
17.7
35.4
70.7
141.4
2060
TerraSAR-X
3.1
20072018
3.3/2.8
1.0/1.0
11
10 (spot)
30 (strip)
100 (scan)
25.7
2055
COSMOSkyMED-2
(2 Sat)
3.1
20152023
13
1040
TerraSAR-X-2
3.1
20152018
0.54
1040
107
108
depends mainly on: i) the discrepancy between the model adopted for
the displacement estimation and the actual deformations; and ii) the
reliability of the estimates based on this model which, in turn, is inuenced by both the number and the spatial distribution of CT, and the
number of images.
A low number of CT, or their non-uniform distribution over the radar
image, can hamper a reliable estimation of atmospheric signal and/or
displacements eld, even though the model assumptions are correct.
A density 5 CT/km2 can be considered as a minimum requirement
(e.g. Colesanti and Wasowski, 2006) even when using algorithms optimized for scarcely urbanized areas, which adopt robust estimation of
atmospheric signal over small patches of land (Bovenga et al, 2005).
Given a particular conguration of spatial and temporal baselines, a
certain probability exists that a fraction of selected CT pixels do not actually correspond to coherent objects, because random phase values
fortuitously give an estimated higher than the threshold. This false
alarm probability is inversely proportional to the threshold and to
the number of acquisitions available in the stack, although the particular
distribution of spatial and temporal baselines also inuences the results.
Given the conguration of spatial and temporal baselines and the expected average density of CT, the value of threshold which guarantees
an acceptable false alarm probability can be evaluated by simulations
(Colesanti et al., 2003a; Bovenga et al., 2006). Higher threshold values
imply lower false alarm rates (higher reliability), but also lower absolute numbers of selected CT. Moreover, a minimum number of images
is required to set a reliable threshold. This number depends on the
sensor parameters, on the precision required for the displacement
estimation (Bovenga et al., 2012a), and on the adopted MTI algorithm.
For PSI approaches relying on medium resolution SAR missions
(e.g. ENVISAT) the minimum number of images is about 1520
(Colesanti et al., 2003b; Hilley et al., 2004), while for higher resolution
sensors e.g. CosmoSkyMed (CSK), TerraSAR-X (TSX), it can be decreased (Bovenga et al., 2012a). The issue of proper selection of CT
was recently discussed by Riddick et al. (2012) with reference to the
application of StaMPS algorithm. They presented a method for optimizing the selection of CT, with focus on the minimization of false positives
and negatives in different natural terrains.
vb
vmax :
4 dt
109
110
Fig. 2. Acquisition geometry in case of descending pass and slopes facing east.
111
variety of factors call for very careful and transparent quality assessment of MTI results.
Following Hanssen (2005) and Mahapatra et al. (2012), the factors
that inuence the quality of MTI products can be grouped in four
major categories:
i) Original radar data and their products (CT), and specically:
number and quality of images
temporal distribution (homogeneity/gaps)
temporal extent of image stack
number and quality of scatterers (CT)
spatial distribution of CT (homogeneity/gaps)
ii) Characteristics of the physical process
temporally/spatially smooth signal versus the presence of
abrupt variations in time/space
temporal and spatial extent
iii) Processing technique (PSI, SBAS, hybrid)
iv) A priori assumptions concerning:
choice of deformation model (e.g. linear)
CT detection method
atmosphere dynamics.
Given the large number of the variables involved, it is not surprising
that, even when exploiting the same radar datasets, different MTI approaches can lead to somewhat different results in terms of number
(and quality) of CT, their spatial distributions, and temporal trends of
displacements. In general, better quality MTI products (and CT) can be
obtained with higher number of images, homogenous temporal distribution of acquisitions and longer temporal extent of image stack. The
number and quality of CT (and hence quality of MTI products) depends
directly on the coherence () threshold used for their selection; thus
represents a useful quality indicator. While high CT density is desirable
to assure better coverage of AOI, there is a trade-off between the number and quality of CT (lowering coherence thresholds leads to greater
amount of CT, but their quality decreases). In addition, due to natural
variability of land cover and environmental conditions, coherence and,
therefore, quality of CT varies spatially within a given AOI. End users
have to be fully aware of this and other quality/precision issues in
order to avoid possible over interpretation of the MTI results.
2.4.3. Assessments of MTI results quality using cross-comparisons and
ground truth
In summary, depending on the type of investigation and availability
of different SAR imagery and ground data, the following consistency
checks of the MTI measurement results can be attempted:
i) using both dual mode (ascending and descending) data from the
same SAR sensor (acquired over the same area); with such independent data even simple qualitative comparisons of the results
are very useful; with a sufcient number of ascending and descending CT, ideally corresponding to the same targets on the
ground (e.g. the same buildings), a quantitative assessment of
product quality can be attempted (Bovenga et al., 2013)
ii) comparisons of MTI results similarity using data from different
sensors (e.g. Bovenga et al., 2012a, 2013; Duro et al., 2012);
iii) intercomparisons involving independent results based on different MTI processing techniques (e.g. Wasowski et al., 2007a;
Lauknes et al., 2010; Bovenga et al., 2013)
iv) validation of MTI results using ground truth e.g. GPS, leveling
data (e.g. Raucoules et al., 2009 and references therein).
These various ways of assessing and validation of MTI results are illustrated by Bovenga et al. (2013), who recently studied the Assisi landslide, Italy. In this case a series of the result checks was possible thanks
to the availability of different sensor data, including both ascending and
descending C-band ENVISAT imagery, the Stripmap CSK, the independent MTI results obtained through the use of two different processing
algorithms (PSI-like SPINUA and SBAS-like TSIA), and high quality GPS
112
data. In order to compare the precision of the results from different sensors, the mean velocity standard deviations were computed, resulting in
about 0.2 mm/yr for the ENVISAT dataset and about 0.4 mm/yr for both
the CSK datasets. The lower precision of the CSK results was expected
because of the limited number of images and the narrower time span.
By coupling results from descending and ascending CSK it was also
possible to derive horizontal and vertical displacement maps (Figure
4C and D), which were compared to the GNSS 3D measurements. In
particular, a strong horizontal component of the movement, following
approximately the maximum slope, was consistently shown by both
MTI and GNSS.
Fig. 3. Distribution of the average line of sight (LOS) velocities obtained by processing ENVISAT data (dots) and GNSS records (squares). Symbol colors are representative of the velocity
according to the color bar (top of gure). The landslide limits shown in red.
After Bovenga et al. (2013). Background image from Google Earth.
113
Fig. 4. Distribution and mean LOS velocity of coherent targets (CT) detected through SPINUA algorithm by processing X-band CSK data acquired along descending (A) and ascending passes
(B). Distribution of the eastward (C) and upward (D) components of movement.
Background image from Google Earth.
114
characterized by shallower slopes, fewer rock outcrops (clay-rich lithologies) and a greater proportion of cultivated land. Furthermore, only
three broad land cover/use classes with decreasing MTI applicability
potential were distinguished:
i) urbanized/industrial areas (63 CT/km2) and bare rocks;
ii) cultivated areas and bare soils (18 CT/km2);
iii) inland water and vegetated areas (6 CT/km2).
Note that these CT densities refer to low-resolution (80 80 m) SBAS
processing of ERS data (19952000) and, therefore, are much lower than
those derived from PSInSAR results of Colombo et al. (2006).
Furthermore, by combining the three land cover/use classes, characterized by different MTI applicability potential, with the sensor acquisition geometries (ascending and descending), topographic data and
landslide inventory information, Cascini et al. (2009) generated a priori
DInSAR landslide visibility maps. Accordingly, the study area was
divided into the following three broad classes:
i) visible;
ii) visible with difculty (low sensitivity to measure slope
displacements)
iii) not visible.
Low sensitivity in this case refers to the difculty to measure translational movements on moderately inclined slopes facing approximately north or south (Cascini et al., 2010). On such oriented slopes it is still
possible to measure displacements in the vertical direction, and therefore the movements characterizing rotational landslides.
The utility of the a priori landslide visibility maps was tested using
CT obtained via low-resolution SBAS processing of ERS descending
dataset. Note that here the term visibility represents a comprehensive
index of MTI suitability incorporating the radar visibility (i.e. lack of
shadowing and layover effects), as well as the qualications of different
measurement capability with reference to CT density.
Another interesting example of a predictive analysis of the likely CT
spatial distribution and density was presented by Notti et al. (2010),
who considered a 47 km2 catchment area in the Spanish Pyrenees. The
area studied, though relatively small, includes considerable amounts of
rocky outcrops forming steeper slopes (2535), as well as shale
(clay-rich) formations with moderate slopes (15), and as such,
perhaps shows intermediate characteristics with respect to the
previously described cases from the Alps and Apennines. As in the
above described studies the sensor geometry and slope topography
were considered to identify areas visible to radar through the so-called
R-index ranging from 0 (no visibility) to 1 (best visibility). To test the
results of radar visibility assessment, Notti et al. (2010) compared
their R-index based zonation of slopes to the actual CT densities obtained by processing 14 TSX images (acquired between May and October
2008 in the descending, STRIPMAP mode). A good correlation was
found in terms of increasing CT densities with increasing R-index.
Subsequently, following the approach of Colombo et al. (2006), on
the basis of empirical evidence, several land cover/use classes were
ranked according to the probability of locating CT (expressed through
LU-index). Then an overall MTI feasibility index (CR-index) was
obtained through the integration of the R-index based zonation with
LU-index. At the end 21% of the landslides in the study area resulted
suitable for MTI monitoring.
A revised version of the CR-index (with weighted average of R- and
LU indexes) has recently been proposed and its utility tested in the
Piedmont Alps, north-western Italy using RADARSAT and ERS data
(Notti et al., 2012b). In general, the comparisons of the actual CT distributions with the study area zonation based on CR-index values indicated
coherent results. However, the authors also acknowledged the difculties
in estimating LU-index for some terrains, problems with the quality and
resolution of land cover/use data and uncertainties in selecting a proper
weighted average to combine R- and LU-index.
115
116
Fig. 5. Map of landslide visibility and sensitivity to down-slope movement for a high-resolution CSK-based MTI application in the Southern Gansu Mountains, China. A) Landslide inventory
superimposed on the visibility and sensitivity to down-slope movement map; B) CT and landslide distributions superimposed on the visibility and sensitivity to down-slope movement
map (derived according to Eq. (3)). Legend: Layover = green); Shadowing = blue; Low sensitivity = white. CT velocity map derived by processing 3 m resolution descending CSK
imagery through SPINUA algorithm (GAP srl).
Background image from Google Earth.
117
118
by Guzzetti et al. (2009). In this case ERS imagery was processed using
SBAS technique (Berardino et al., 2002) and the landslide vs CT statistics
considered here refer to the centers of low resolution (80 80 m
ground pixel). The most relevant aspects and outcomes of the work
are summarized below:
study area ~6500 km2 (Umbria region Apennines: hills, mountains
with large open valleys);
93% vegetation cover (41% forest, 10% grassland, 42% cultivated);
~ 33,500 mapped failures (landslide density ~ 5.2/km2; ~ 10% of
mountainous/hilly areas);
49 and 24 descending and ascending ERS1/2 images (period: 1992
2000);
1.2% and 0.9% of landslides with 1 CT, respectively for descending
and ascending data;
0.2% (44) of landslides with 1 CT imposing the availability of data
from both orbits (suitable for retrieval of EW and vertical deformation components).
Guzzetti et al. (2009) noted the lower percentage of landslides containing at least one CT with respect to the results of Farina et al. (2006).
Considering the similarity of landslides types and geology in Umbria
and the Arno River Basin, the differences were ascribed to the somewhat higher percentages of forest and cultivated land in the former region and to specic processing approaches. Another limitation noted by
Guzzetti et al. (2009) was the difculty in establishing the exact location
of landslides suitable for MTI based investigation.
3.2.3. Applicability potential of MTI based on ERS and other C-band data
The outcomes of the above-discussed three different regional scale
studies appear largely comparable. They clearly point out that in settings comparable to those of central and northern Italy, on the average,
only a small percentage (10%) of landslides will benet from MTI information that can be derived from ERS data. This is further conrmed
by the results of another ERS based PSInSAR application regarding a
nearly 1100 km2 landslide-prone area in the north-western Apennines,
Lombardia region, Italy (Meisina et al., 2006). With over 90,000 CT
identied (density 82/km2) only 7% of the inventoried landslides had
one or more CT.
Although such low percentage gures indicate a low performance or
applicability of ERS based MTI, with the tens of thousands of landslides
mapped in many Italian regions, the implication is that potentially useful space-based information can always be obtained for a signicant
number of slope failures (on the order of thousands). And this information may often be the only one available considering the generally high
costs of in situ monitoring. Importantly, the percentage of urban
landslides suitable for MTI investigations will be considerably higher
because of the general availability of CT in urbanized areas.
Furthermore, it is clear that the MTI applicability statistics may
depend also on the predominant type of slope movements and their
state of activity. This seems to be the case of about 490 km2 area in
south-central Italy investigated by Cascini et al. (2010). Using ERS1/2
based SBAS results and focusing on rotational slides, earth ows and rotational slides-earth ows, they showed that 33% of the total number
(553) contained MTI information. This percentage is about ten times
higher with respect to the results of Guzzetti et al. (2009), who also
used low resolution SBAS approach and ERS data, but examined a
much larger area with a greater variety of landslide types (and greater
vegetation cover). Signicantly, most of the landslides studied by
Cascini et al. (2010) were dormant.
In general, extremely slow to very slow, deep creeping landslides
seem the most suitable for MTI investigations (e.g. Colesanti and
Wasowski, 2006; Colesanti et al., 2006; Delgado et al., 2011; Bovenga
et al., 2012a). Furthermore, in case of dormant landslides or those
with very low displacement rates and greater depths, the ground surface disturbance is null or limited and probability of preserving good
radar target coherence increases.
Literature provides also few examples of regional scale landslide investigations based on ENVISAT MTI (e.g. Righini et al., 2011; Bianchini
et al., 2012). They indicate a similar or even better performance with respect to ERS data. This is not surprising considering the nearly identical
characteristics of the two sensors and the incidence angle exibility of
ENVISAT (Table 1).
In particular, Bianchini et al. (2012) exploited ENVISAT-based MTI
results to detect landslides and characterize their activity in the central
Calabria (Southern Italy), a region prone to diffuse slope movements. In
this case radar imagery was processed using the method of Persistent
Scatterers Pairs (PSP) (Costantini et al., 2008). The salient aspects and
outcomes of the work are summarized below:
study area ~ 4470 km2 (90% hilly, mountainous terrain with
Hercynian crystalline basement rocks and overlying Mesozoic
Tertiary sedimentary units);
4102 mapped failures (landslide density ~ 0.9/km2; ~ 7% of total
area); predominantly slow roto-translational (53%) and deep-seated
(21%) slides;
landslide activity: 78.5% dormant, 21% active, 0.5% stabilized
108 ascending ENVISAT images (period: 20032009); LOS 23
~349,000 CT (78 CT/km2)
24% of landslides with sufcient number of CT (68% and 8%, respectively with no CT or insufcient number of CT)
64 new landslides detected (1.5% of the total of the pre-existing
inventory).
This case study from Southern Italy would suggest a signicantly
better coverage of landslides with ENVISAT (24%) with respect to ERS
(typically 10%). However, the statistics regarding MTI performance
based on ERS data refer to the regions from Central and Northern
Italy, which are generally characterized by higher precipitation (and periods with snow cover) and vegetation density.
In theory, even better applicability can be inferred for MTI based on
RADARSAT-1/2 data. This is related to the shorter revisit time of the
satellite (24 days), which implies lesser coherence loss problems. Indeed, the literature provides encouraging, albeit limited, examples of
MTI applications based on RADARSAT-1/2 imagery. Of interest is work
by Lu et al. (2012), who used RADARSAT-1 processed by PSInSAR to
investigate landslides in about 6300 km2 area covering about 70% of
the Arno River Basin, which was earlier studied through ERS-based
MTI by Farina et al. (2006). Using both ascending and descending
datasets (respectively 54 and 48 images), Lu et al. (2012) obtained displacement information for 3203 landslides (14.9%). This suggests moderate improvement assuming that number of CT could be doubled in
this case, and considering also that Farina et al. (2006) exploited a
signicantly larger number of descending ERS imagery (350).
Finally, the comparative analysis of the ndings of MTI landslide investigations based on ERS, ENVISAT and RADARSAT data provides useful
indications on the potential performance of future C-band sensors. In
particular, with reference to the technical parameters of the upcoming
Sentinel-1 mission, and specically to the shorter re-visit time and exible incidence angle geometry (Table 1), one can expect much better
practical utility of this sensor for landslide monitoring.
3.3. Catchment-wide and local scale (slope specic) MTI investigations
of landslides
While the literature indicates that regional scale landslide applications of MTI have so far been restricted to the Italian territory, this is
not the case of the catchment and especially larger scale studies. Nevertheless, it is apparent that at these scales the majority of published
examples of MTI-based landslide investigations regard European countries. Indeed, the MTI techniques have been rst developed in Europe,
where they also gained earlier acceptance by landslide scientists and
end users.
119
120
121
Fig. 6. Zhouqu study area in the Southern Gansu Mountains (China): A) General view showing distribution of 25 landslides (in red) mapped using Google Earth optical imagery and 3D
tools. The town of Zhouqu (right of center) develops on the northern side of the Bailong River Valley. The 4 km long Suortou landslide represents the largest slope failure in the area;
B) Distribution and average velocities of radar targets (over 46,000 CT) superimposed on the GE image with 25 landslides. Velocity values are saturated 20 mm/yr and represent LOS
displacements. Dashed and continues line arrows indicate, respectively, radar azimuth and LOS. Red and blue dots indicate, respectively, CT movements away and toward the satellite
sensor. CT velocity map derived by processing 3 m resolution descending CSK imagery through SPINUA algorithm (GAP srl).
hazard to the local population and infrastructure (Figures 6, 7). Monitoring these long-lived landslides is important especially when, as in
the case of the Bailong catchment, they are known to undergo periods
of increased activity resulting in river damming and disastrous ooding.
The MTI results relevant to two of these major landslides are
presented in more detail in Fig. 7, which includes also time series of
selected radar targets. In both cases over 100 CT were available per landslide. The majority of the radar targets that fell within the slides' limits
showed considerable LOS displacement rates, with the maxima exceeding 60 mm/yr in the case of the larger movement (about 4 km long
Suortou landslide), and 140 mm/yr in the other case (Figure 7). Significantly, despite the very similar land cover/use and lithology, the
122
Fig. 7. A) Google Earth view of two major landslides located to the west of Zhouqu showing distribution and average velocities of radar targets. Velocity values are saturated 20 mm/yr
and represent LOS ground displacements; numbers 1 and 2 indicate CT whose time series are shown, respectively, in B and C; B) and C) show time series of CT with high displacement rates
from the Suortou (1) and another landslide (2) located further west. CT velocity map and time series derived by processing 3 m resolution descending CSK imagery through SPINUA
algorithm (GAP srl).
landslide with faster moving radar targets had more gaps in CT coverage. This may be related to greater ground surface disturbance at higher
movement rates. Furthermore, in areas with good CT coverage it was
also possible to detect and interpret local variations in the direction of
movement with reference to the CSK satellite LOS. These are well visible
in the middle-lower portions of both landslides (e.g. predominantly SEand SSE-ward movements diverging in the fan-shaped toe areas
towards S and SSW).
The presence of complex kinematic patterns in large landslides in intensely sheared materials is not surprising. The same can be expected
for many other and also smaller failures in soil, for which a coherent
block-like motion model may not be inappropriate. All this calls for
much caution when introducing simplifying assumptions for landslide
assessments based on MTI displacement results. For example, landslide
activity classications based on the average velocity of all CT falling
within the slide limits (assuming unidirectional movement), can easily
Southern Gansu, the MTI-derived displacements offer unique information, which, following expert judgment, can be used for preliminary
wide-area assessments of hazards linked to landslide activity.
The relatively high percentage of the landslides covered by CT is consistent with the outcomes of the radar visibility assessment (Figure 5,
Section 3.1.5). The high mountain setting investigated here seems favorable for high resolution X-band MTI because of i) scarce vegetation;
ii) abundant rock exposure, and iii) relatively dry climate. In this respect
the Gansu study area resembles the Tena Valley, which is also characterized by very similar rock assemblages and comparably high proportion
of large, slow landslides.
In addition, very good performance of 3 m resolution CSK-based MTI
in the Alpine environment has been documented by Bovenga et al.
(2012a), who investigated slow deep-seated landslides in Switzerland
and Northern Italy. Similarly, Dehls et al. (2012) reported very good results from a MTI study of three unstable rock slopes in Norway based on
3 m resolution TSX (Stripmap) data. Nevertheless, it is likely that in
some other, more vegetated settings, and where predominant type
of slope failure is different (e.g. faster landslides), the application of
X-band MTI to landslide monitoring will be less successful.
3.4. Local scale (slope specic) MTI investigations of landslides
Scientic literature, including Engineering Geology journal, offers
numerous examples of MTI applied to local scale landslide investigations. Here we refer to our own case studies from Europe that rely
on ENVISAT and CSK data processed using SPINUA algorithm. The
123
Fig. 8. Distribution and average line of sight (LOS) velocity of radar targets (marked by color dots) in the Madesimo area; the town is visible in lower right. Mount Mater facing the town
and another unstable peak are located, respectively, right and left of center; velocity saturated at 10 mm/year for visualization purposes. CT velocity map derived by processing ENVISAT
descending imagery through SPINUA algorithm (GAP srl).
Background optical image is from Google Earth.
124
Fig. 9. A) Distribution and average LOS velocity of radar targets (marked by color dots) in the upper part of the Mount Mater; velocity saturated at 10 mm/year for visualization purposes.
Numbers 1 and 2 refer to CT, whose times series are shown in B) and C). Note the scarp near the mount summit likely linked to deep-seated slope deformations and fresh debris indicative
of active mass wasting. B) and C) Two time series showing LOS displacement rates exceeding 20 mm/yr; temporal gaps in measurements refer to winter months with snow coverage. CT
velocity map derived by processing ENVISAT descending imagery through SPINUA algorithm (GAP srl).
Background optical image is from Google Earth.
125
debris or simply iii) creep of talus overlying stable bedrock. This calls
for caution when inferring slope instability hazards from remotely
sensed ground surface displacements. Again, simple averaging of velocities of many different CT without distinguishing the origins of their
motion may lead to misleading results. If eld controls are not feasible,
a careful interpretation of very high resolution optical imagery can help
avoid erroneous assessments.
3.4.2. BindoCortenova case study, Central Italian Pre-Alps
This example again regards rock slope instability in Italy. The
exploited SAR dataset was the same as in the case of Madesimo (32
ENVISAT descending images from June 2005 till August 2010).
The interpretation of the MTI results was also in this case facilitated
by the high resolution optical data and 3D visualization tools available
from GE. We had no direct knowledge of the study area, but beneted
from the geological and geomorphologic descriptions and the detailed
analysis of the damaging landslide events that occurred in Bindo in
2002 (Crosta et al., 2006).
In particular, the area is characterized by the metamorphic basement
rocks of Variscan age with granitic intrusions. The sedimentary cover is
represented by the Permian age formation made of conglomerates and
Fig. 10. Distribution and average LOS velocities of radar targets (marked by color dots) in the upper part of the unstable slope located north of Madesimo; velocity saturated
at 10 mm/year for visualization purposes. White circles indicate CT from rock debris with the highest velocities: 39 mm/yr (marked A) and 2527 mm/yr (B). CT velocity
map derived by processing ENVISAT descending imagery through SPINUA algorithm (GAP srl).
Background optical image is from Google Earth.
126
Fig. 11. (A) LOS velocity results from ENVISAT data (period 20052010) superimposed on a Google Earth view of a large, apparently relict deep-seated slope failure (in the center) in the
BindoCortenova area; velocity saturated at 10 mm/year for visualization purposes. Note a 500 m long and 300 m wide debris avalanche (left of center) that in 2002 destroyed the SE
part of the Bindo village, and few coeval debris ows along the left lateral portion of the main old failure. The presence of moving radar targets on the slope affected by the debris avalanche
suggests post-failure instability. Yellow mark indicates location of GPS measurement point whose monitoring results are shown in Fig. 12; B) Distribution and average LOS velocity of CT in
the middle-upper part of the slope affected by the 2002 Bindo debris avalanche. Close spatial association of moving and apparently stable CT may suggest lack of coherent deformation
pattern. Note presence of abundant rock debris, including some large blocks. Arrow indicates CT whose displacement time series is shown in Fig. 12. CT velocity map derived by processing
ENVISAT descending imagery through SPINUA algorithm (GAP srl).
avalanche that reactivated part of the toe of the old deep-seated slope failure (paleo-landslide), destroying part of the Bindo village (Figure 11).
In addition, few debris slides mobilized the lateral portions of the old
failed slope.
127
A possible explanation stems from the fact that the two measurements are distant about 200 m from each other and may reect local
differences in slope deformations. However, phase unwrapping errors
occurring due to high displacement rates cannot be excluded: the
phase trend derived from the CT phase values by adding 2 jumps
shows a good agreement with the GNSS measurements (Figure 12).
The Bindo case illustrates some difculties that can be encountered
while interpreting CT movement results:
i) CT data limited only to the 2002 debris avalanche that mobilized
portion of the toe of a much larger, old slope failure covered by
dense vegetation;
ii) Complex post-failure displacement pattern perhaps in part related
to independent motion of rock debris overlying stable or unstable
body of the pre-existing large slide, or to possible unwrapping
errors, or both;
iii) Very limited ground monitoring results, regarding a GPS monitored site spatially distant (about 200 m) from the nearest CT;
this confers much uncertainty to any attempt of data integration.
Ideally, in similar cases careful and timely (during radar data acquisition period) in situ inspections would be needed to provide necessary
constraints for more reliable data interpretation. Today, over 10 years
after the debris avalanche event, the site conditions are different and
re-vegetation hampers detailed scale retrospective assessments.
3.4.3. Saint Moritz case study, Swiss Alps
This case has already been examined by Bovenga et al. (2012a), who
conducted a comparative study of MTI performance based on ENVISAT
and high resolution CSK data. Here we focus on the complementarities
of the two datasets and on the difculties in interpreting CT displacement
results. The results discussed below refer to the following datasets: i) 28
images acquired between March 2009 and October 2010 by CSK in
Stripmap mode along ascending pass with mean incident angle 26.6;
and ii) 35 images acquired between April 2004 and June 2010 by
ENVISAT along ascending pass with mean incident angle 22.8.
With reference to the 2006 Report of the Department of Civil, Environmental and Geomatic Engineering of Swiss Federal Institute of
Fig. 12. Monitoring results of GNSS measurement point (red circles) and displacement time series of CT (blue squares) located, respectively, upslope from and within the Bindo debris
avalanche (see Fig. 11 location). Comparison of the data may suggest the presence of unwrapping errors due to improper phase ambiguity resolution: the phase trend (black crosses)
obtained by adding 2 jumps to the CT original phase values is in good agreement with the GNSS measurements.
128
Fig. 13. A) MTI results, obtained by processing ENVISAT data acquired on ascending pass, superimposed on Google Earth view of the slopes overlooking the town of Saint Moritz. Velocity
values are saturated 10 mm/yr and represent ground displacement along the LOS. Note complex landslide on the SE facing slope above St. Moritz. The failure is composed of the Gianda
Laret rockfall (upper zone) and Brattas landslide (lower zone), whose foot movement is constrained by the via Maistra rock ridge. Black letter C and arrow indicate location of the site
shown in Fig. 15C. B) Close-up showing distribution and average LOS velocities of CT in the Gianda Laret rockfall area. Note lack of CT in the zone of most active movements. Complex
displacement pattern could be related to independent motion of rock debris overlying unstable bedrock, low radar sensitivity to movements in NS direction, as well as to possible
unwrapping errors.
Background optical image is from Google Earth.
129
Fig. 14. LOS velocity map of the peri-urban part of the unstable slope area at Saint Moritz obtained by processing ENVISAT data acquired on ascending pass (A), and CSK SAR data acquired
on ascending pass (B). Numbers 1 and 2 indicate, respectively, the Brattas landslide (foot) and a small unstable slope area in the western periphery of the town. Black letters A and B with
arrows indicate location of the sites shown, respectively in Fig. 15A and B.
Background optical image is from Google Earth.
130
Fig. 15. Examples of LOS displacement times series of the radar targets (pointed by arrows) from three sites of ground instability in the Saint Moritz area: foot of the Brattas landslide characterized, respectively, by down slope (A) and upward (B) displacements, and the unstable slope area to the west of the Gianda Laret rockfall (C). CT displacement maps refer to the CSK
dataset acquired on ascending pass ((A) and (B)), and to ENVISAT dataset acquired on ascending pass (C). Note that displacements of the CT in gures (A) and (B) are, respectively, away
from and toward the satellite sensor.
Background optical images are from Google Earth.
slope of 20 (Figure 13). The upper zone shows signs of instability clearly visible on Google Earth optical image from September 2010
(Figure 13). The Brattas landslide affects directly urban area. The slide
is few tens of meters thick and its 600 m wide foot is blocked by the
Via Maistra rock ridge (D-BAUG, 2006). It is apparent that the Gianda
Laret rock fall and the Brattas landslide belong to one large complex
slope failure.
CT displacement results from MTI processing of ENVISAT data conrmed the overall instability of the SE facing slope that overlooks
Saint Moritz, even though due to tree cover no information was obtained for the middle part of the slope (Figure 13). The available high resolution CSK images covered only the lower part of the slope, but their
processing provided nine times greater density of CT for the urbanized
foot of the Brattas landslide with respect to that derived from C-band
data (Bovenga et al., 2012a). This can be visually appreciated in
Fig. 14, which also illustrates the complex deformations in the foot
area of the Brattas landslide, with down slope and upward moving CT.
Examples of time series of two CT showing such deformations are
included in Fig. 15. Signicantly, the in situ evidence of active compression at the foot of the Brattas landslide has recently been reported by Puzrin et al. (2012), who proposed a geotechnical model of the
phenomenon.
The MTI also revealed the presence of a group of moving CT (average
LOS velocities ranging from a few to several mm/yr) in the via Laret
area, to the west of the town center (Figure 14). We have no direct information about this site, but its instability has recently been of concern
for the Saint Moritz Municipality (A. Puzrin, personal communication,
2012). Although conditions of instability are indicated by the results obtained from the CSK (ascending) and ENVISAT (ascending) radar data,
only the latter allow the actual recognition of the unstable area, which
appears to be over 100 m long and wide. Indeed, in our previous assessment of MTI results from CSK data, this slope instability has been ignored (Bovenga et al., 2012a). The instability shows up better through
the ENVISAT-based CT, which form a more homogeneous displacement
pattern (Figure 14). This may be related to the longer observation period of ENVISAT (from April 2004 to June 2010). It is possible that with
the much shorter temporal coverage (from March 2009 to October
2010), CSK-based MTI was not capable to highlight displacements of
some very slow CT. However, in the absence of ground data we cannot
also exclude the presence of faster movements in the earlier monitoring
period (before March 2009) covered only by ENVISAT data. This example also suggests that in urbanized settings MTI based on medium resolution C-band data can locally furnish even better information with
respect to MTI from X-band data that currently are generally characterized by short temporal coverage.
Signicantly, the MTI results derived from ENVISAT data revealed
the presence of slope instability also to the west of the Gianda Laret
rock fall site (Figure 13). Time series of CT from that area show presence
of apparently gradual displacements (Figure 15). Even though the LOS
velocities result to be low (generally within 10 mm/yr), the overall CT
displacement pattern is complex. This could again be related to independent motion of rock debris overlying unstable bedrock, low radar
sensitivity to movements in NS direction, as well as to possible
unwrapping errors.
In spite of the above interpretation uncertainties, we expect that CT
displacement information obtained for the Saint Moritz area can be of
much help for the positioning of some ground monitoring points in
the upper part and other zones of the unstable slope. This case study
also shows how the combined use of data from different sensors (here
ENVISAT and CSK imagery) can help securing complementary results.
3.5. Monitoring infrastructure stability in landslide-prone areas
We present two examples of MTI-based detection of road network
instability associated with slope movements. For comparative reasons
one case regards the application of medium resolution ENVISAT data
131
and the other relies on high resolution (3 m) CSK data. The intention
is to show the advantages of MTI applications relying on higher resolution (and shorter re-visit time) acquisitions by new generation X-band
radar sensors.
The rst example is from a landslide-prone region in the Southern
Apennines, Italy. It illustrates the detection of local instabilities based
on 47 ENVISAT images acquired between November 2002 and January
2010. Fig. 16 shows that the majority of CT located along the road of interest is stable. However, two radar targets exhibit average LOS displacements of about 6 mm/yr. One of these coincides with the road
portion adjacent to a landslide active during late 20022003 period.
The site has been known to the rst author, who inspected it in early
2003 following the occurrence of a 100 m long landslide (Figure 16).
This local knowledge helped associate the moving CT with the specic
slope failure event.
However, in the absence of site-specic information, the use of information from isolated CT to infer infrastructure instability and its causes
may lead to erroneous assessment. The interpretation uncertainties can
be much reduced where higher density of CT is available, as illustrated
through the second example from the Calabria Region mountains in
Southern Italy.
In this case MTI processing of 53 CSK images acquired between April
2009 and April 2012 resulted in good CT coverage of the area of interest
which includes A3 Highway (Figure 17). In particular, thanks to the
presence of numerous CT it was possible to delimit in detail the unstable
portion of the highway corresponding to a viaduct. Furthermore, high
resolution optical imagery from Google Earth help associate the structural instability with the presence of about 200 m long slope failure
(Figure 17). Signicantly, thanks to the higher temporal sampling of
CSK, it was also possible to reveal the presence of non-linear deformations (Figure 17B, C).
The available literature data (e.g. Bovenga et al., 2012a and references therein) indicate that with respect to medium resolution MTI,
from few to over 10 times greater CT densities can now be obtained
with the higher resolution X-band radar data. As illustrated in the two
examples above, high resolution implies increased feasibility of sitespecic or even engineering scale assessments of infrastructure stability
in landslide-prone areas.
4. Limitations of MTI applications in landslide investigations and
user guidelines
As reported in this review, many successful MTI case studies can be
found in scientic literature concerned with landslides. The difculties
commonly encountered in MTI applications to landslide investigations,
though discussed in some works (e.g., Bovenga et al., 2006; Colesanti
and Wasowski, 2006; Delacourt et al., 2007; Wasowski et al., 2007b;
Notti et al., 2010), have received less attention.
However, landslides often occur in environmental settings that are
harsh for the MTI applications (e.g. vegetated slopes, steep and rough
topography). Furthermore, with the exception of urban and bare rock
slopes, elsewhere the density of radar targets suitable for interferometric measurements could be low and this implies difculties in MTI analysis, as well as introduces considerable uncertainties in the assessments
of ground motions. Also, as emphasized by Colesanti and Wasowski
(2006), the interpretation of the exact geological/geotechnical signicance of millimeter to centimeter (per year) displacements can be
very challenging, because i) very slow ground surface deformations
may arise from a wide variety of causes and, therefore, their presence
on slopes may not always reect shear movements or occurrence of
landslides, and ii) many CT correspond to man-made objects (e.g. buildings) and thus their structural state/behavior (and groundfoundation
interactions) should be taken into account.
From a practical point of view, it seems useful to go beyond the scientic discussions of MTI applicability and offer some specic recommendations and guidelines on how to mitigate the limitations of the
132
Fig. 16. Distribution and LOS velocities of radar targets (ENVISAT ascending data) along roads crossing a landslide-prone area of the Southern Apennine Mountains. One of the two slowly
(6.5 mm/yr) moving CT coincides with a road portion adjacent to a 100 m long landslide, whose movement direction is indicated by black arrow.
Background optical image is from Google Earth.
Given the recent opportunity of multi-sensor and multi-band investigative approaches, it is apparent that to be most effective MTI will have
to be tailored to the specic region/site conditions, landslide types, depending on the primary objectives of the investigation. For example,
in case of critical facilities at risk such as dams or bridges, of urban landslides or potentially unstable slopes in urban/peri-urban areas, as well
as in case of slopes traversed by lifelines and important engineering
structures, the cost of acquiring and processing high resolution radar
data can easily be justied. Conversely, in wide-area regional investigations, the use of medium resolution imagery will be more appropriate
and the most cost-effective.
Apart from the satellite data availability, the success of any MTI application in landslide investigation will depend on the two following
key factors:
suitability of the area of interest in terms of sufcient coverage by
radar targets (CT);
adequate knowledge of the limitations involved and the user ability
to correctly interpret MTI results and to extract useful information
for landslide assessment.
In particular, in local scale assessments typical of engineering geology practice, MTI results will be critically constrained by the availability
of a sufcient density of CT and their suitable distribution within and
in the neighborhood of the features to be investigated or monitored,
i.e. potentially unstable slopes and landslides. Depending on the objectives of the investigation, feasibility study can be recommended, especially when contemplating MTI applications in harsh environments
(e.g. dense vegetation), particularly when the focus is on obtaining
site-specic information.
The MTI feasibility study can be conducted by rstly constructing
simple radar visibility maps with reference to sensor geometry and
133
Fig. 17. A) Distribution and LOS velocity of CT showing instability of a viaduct (A3 Highway) crossing a landslide area in Calabria, Southern Italy; B) and C) Time series of two CT showing
the presence of non-linear deformations. CT velocity map and time series derived by processing 3 m resolution ascending CSK imagery through SPINUA algorithm (GAP srl).
Background optical image is from Google Earth.
ascending and descending acquisitions is recommended as this can mitigate the shadowing and layover effects. Secondly, based on a consistent
number of MTI case studies conducted in different environmental settings, CT availability can be anticipated by exploiting land cover/use
information. However, due to the great variability/heterogeneity of
134
Table 2
Technical issues in satellite MTI applications to landslide investigations and guidelines for users.
Technical issues
User guidelines
Lack of natural, coherent radar targets (CT) or low spatial density (b5 CT/km2) in areas
characterized by scarce degree of urbanization, limited rock outcrops/bare ground,
dense vegetation cover, land use changes (e.g. seasonal crop agriculture, ongoing
engineering constructions) and steep and adversely oriented slopes with respect to
satellite Line of Sight (LOS) geometry
Use both ascending and descending radar datasets; select data from other sensors with
suitable radar LOS geometry (e.g. L-band works better in vegetated areas; greater CT
density from data acquired by new generation high resolution satellites CSK, TSX,
RADARSAT-2); consider different MTI processing options/algorithms (e.g. lower
coherence threshold compromise), combined Persistent Scatterer Interferometry (PSI)
and small baseline approaches; use articial reectors: Corner Reectors (CR) or Compact
Active Transponders (CAT)
Can anticipate likely distribution of areas with good, limited, very poor CT coverage by
taking into account slope geometry and radar LOS geometry (to generate sensor visibility
maps), and land cover/use
CT positioning errors, usually within radar image pixel size (e.g. 20 m for medium
resolution sensors), can represent a serious limitation in site-specic or local scale
engineering investigations
Conduct in situ inspections and use highest resolution optical imagery (also Google Earth)
or detailed maps to associate CT with actual objects on the ground; with high resolution
sensors positioning errors between 1 and 3 m
Can detect and monitor only very slow, coherent landslides with a limited range of
measurable ground displacement velocities (usually up to several cm/yr)
With shorter satellite revisit time (CSK constellation, TSX), the range of detectable
displacement velocities can be increased up to several cm/month (see Table 1)
MTI usually unable to detect strong non-linear deformation signals and high velocity
displacements (e.g. exceeding few tens of cm/yr)
If available, non-linear deformation model based on ground data can be used to support
MTI analysis; higher temporal sampling (days) by new generation satellites should help
Signicant differences between average displacement rates from MTI and landslide
movement rates from ground monitoring (e.g. with CT detect only fraction of the
horizontal component of ground motion); almost impossible to retrieve slope
movements in the radar satellite azimuth direction
Account for variable LOS sensitivity of current radar systems to measure displacements
(for incidence angles between 20 and 50, obtain 613%, 3475% and 9464%,
respectively, of the north, east and vertical components of motion); can use
speckle-tracking method to retrieve the deformation in the azimuth direction, but
impractical for small deformation elds. MTI and ground data tied to different reference
systems and hence cannot be directly compared; GNSS 3D motion data very helpful
MTI displacement results are relative in time and space; choice of suitable reference point Advice MTI experts on the selection of reference points; use detailed in situ observations and,
if available, ground monitoring data (e.g. GNSS) to verify the stability of reference point; if
often difcult; stability of reference point uncertain; more than one reference point
needed, change reference point; avoid structures subject to signicant thermal effects (e.g.
needed for large areas (or in tectonically complex, active regions)
metal roofs), settings with high groundwater level variations, with clay-rich materials (volumetric changes) and compressible sediments; rock sites generally most suitable
Need time (usually several months or more) to build up new radar imagery data stack
required for MTI processing; precludes rapid response in case of emergency situations
With shorter revisit time of new (e.g. CSK, TSX) and planned satellites the acquisition time
can be reduced to few months (see Table 1); MTI useful in post-failure phase that may
require long-term monitoring
Complex issues concerning quality, reliability and validation of MTI products (average
displacement velocity maps and time series); different measurement precision
estimates in literature (from several to a few mm/yr for a single measure, to 12 mm/yr
or better for average annual rates); limited appreciation of underlying assumptions and
a priori information, uncertainties and errors involved in different types of MTI analysis
Quality assessment difcult and often site-, data- and method-specic; a priori
assumptions (e.g. displacement model) and case-specic MTI measurement precision
(e.g. in terms of Standard Deviation with respect to model) should be clearly stated; to
mitigate noisy results set signicant velocity thresholds consistent with measurement
precision (e.g. 12 mm/yr); lower threshold for X-band (1 mm/yr)
depending on the site-specic variations of land cover/use and topography (local changes in slope aspect and inclination). In this case, detailed
DEM and land cover/use information is necessary to provide reliable a
priori feasibility assessment. For local scale investigations one may
also consider the use of high resolution radar sensors which can offer
from 3 to over 10 times greater densities of CT with respect to medium
resolution MTI (cf. Bovenga et al., 2012a).
Furthermore, the highly variable sensitivity of SAR systems to downslope displacements can represent a considerable limitation in landslide
investigations. However, under certain conditions (e.g. predominantly
translational movement mechanism), the LOS displacement (or velocity) values can be converted to slope velocities, thereby facilitating interpretation of MTI surface deformation monitoring results.
Ultimately, however, the extent to which MTI can be protably used
in landslide investigation will depend on the users' background knowledge of the technique (and of slope processes), and on the interpretation capability of the CT displacement results. The idea beyond the
exploitation of MTI-derived ground surface displacement data is conceptually very simple, in that moving and non-moving CT denote, respectively, conditions of instability or stability. However, we argue
that LOS displacement data must be carefully interpreted using expert
judgment before they can be condently exploited as an index of landslide state of activity and slope instability. This will be especially important in case of extremely slow surface deformations that can have
different origins.
135
Table 3
Interpretation issues in satellite MTI applications to landslide investigations and guidelines for users.
Interpretation issues
User guidelines
If possible, use both ascending and descending data to reconstruct 3D deformation eld or
combine MTI results with ground data (e.g. GNSS, borehole inclinometers); for translational slides can use MTI displacement values resulting from projection along maximum
slope direction
Field checks necessary; use independent measurement data; examine site/slope history
(e.g. recent changes involving construction, local failures); with densities b5 CT/km2
(minimum threshold for successful processing), MTI product quality questionable; results
quality spatially variable (the higher CT density the better quality)
Use detailed geological data and in situ observations to associate CT to actual targets in the
eld; detailed scale topographic maps along with CT height information help differentiate
between radar signals from buildings and ground surface; distinguish between CT on rock
and soil slopes, with the latter being often subjected to seasonal creep and localized
deformations under loading; distinguish between CT from rock debris and bedrock
Lack of homogeneous spatial pattern (or CT clusters with similar velocities) can be
indicative of local deformations (e.g. settlements) or autonomous CT motions (e.g. rock debris
or unstable man-made structures), can result from non-unidirectional movements or reect
local variations in slope inclination/aspect or different deformation mechanisms (compression, extension) within a landslide; use maximum velocity of a consistent cluster of CT
Focusing mainly on warm (or cold) color moving CT (instability) and over-looking
non-moving CT, which offer valuable indications of stability conditions
Exploit jointly non-moving and moving CT to provide a general distinction between stable
and unstable ground and then focus on specic slope instabilities or landslides
Using surface displacement information from MTI for maping and assessing spatial
landslide hazard (susceptibility), updating landslide inventories (e.g. identifying new
landslides, revising boundaries of known landslides)
Although very useful in regional and local scale hazard assessments, MTI provides spatially
incomplete information (good CT density/coverage for a portion of slopes/landslides of interest); improved CT coverage with high resolution sensors and via multi-sensor approach;
pay attention to soil creep phenomena and minor ground deformations often present in
slope areas adjacent to active or recently active landslides (unloading/loading effects)
Capturing signals of potential slope failure (or landslide reactivation) from MTI time series
of ground deformations and exploiting it for landslide hazard forecasting and early
warning purposes; in theory MTI has good potential, but deformations preceding failure
could be very localized, extremely small and strongly non-linear, and hence difcult to
detect and measure reliably
Examples of pre-failure strain detection via MTI are lacking and records of very slow CT
motions seem to preferentially capture creep of deep slides and post-failure ground
instability (settlements, volumetric changes); easier to capture seasonal accelerations of
large landslides; distinguish deformations of rock and soil materials (and brittle vs
non-brittle behavior); shorter revisit time (better temporal sampling) and high resolution
(higher CT density) of new generation radars should help; X-band data seem best suited,
because of higher measurement precision
136
Acknowledgments
The CSK imagery for the Zhouqu site, China was provided by the
Italian Spatial Agency (ASI) within the COSMO-SkyMed AO Project ID
1820. ENVISAT and CSK data for St. Moritz (Switzerland), Madesimo,
Bindo and Assisi (Italy) were provided respectively by the European
Space Agency (ESA) and ASI in the framework of the MORFEO project
funded by ASI (Contract no. I/045/07/0). CSK imagery for the test site in
Calabria Region (Italy) provided by ASI (Agenzia Spaziale Italiana) in
the framework of the project Landslide Monitoring and Mapping
System CAR-SLIDE (PON 01 00536). ENVISAT data for Volturino site
provided by ESA under CAT-1 project ID 2653 in the framework of Puglia
Region project FRANE PUGLIA Rilevamento di deformazioni al suolo
con tecniche satellitari avanzate. We thank Raffaele Nutricato and
Davide Oscar Nitti of GAP srl for providing MTI processing through
SPINUA algorithm. Comments by two anonymous reviewers and suggestions by the journal's Editor Hsein Juang helped us improve this work. We
also thank Qu Shannon (ELS-BEI) for soliciting this review article.
References
Agenzia Spaziale Italiana, 2007. COSMO-SkyMed System Description & User Guide, Doc
No: ASI-CSM-ENG-RS-093-A. ASI, Agenzia Spaziale Italiana, Roma, Italy.
Apuani, T., Arosio, D., De Finis, E., Stucchi, E., Zanzi, L., Ribolini, A., 2012. Preliminary seismic survey on the unstable slope of Madesimo. Proceedings of SAGEEP (Symposium
on the Application of Geophysics to Engineering & Environmental Problems), 2012,
Tucson, Arizona.
Bally, P., 2012. Scientic and Technical Memorandum of the International Forum on Satellite EO and Geohazards, 21-23 May 2012, Santorini, Greece. http://dx.doi.org/10.
5270/esa-geo-hzrd-2012.
Bamler, R., Eineder, M., 2005. Accuracy of differential shift estimation by correlation and
split-bandwidth interferometry for wideband and delta-k SAR systems. IEEE Geosci.
Remote Sens. Lett. 2 (2), 151155.
Bamler, R., Eineder, M., Adam, N., Gernhardt, S., Zhu, X., 2009. Interferometric potential of
high resolution spaceborne SAR. Photogrammetrie Fernerkundung Geoinformation
(PGF), 5/2009 407419.
Barboux, C., Delaloye, R., Strozzi, T., Collet, C., Raetzo, X., 2012. TSX InSAR assessment
for slope instabilities monitoring in alpine periglacial environment (Western Swiss
Alps, Switzerland). Proceedings of Fringe 2011 Workshop, September 19-23, 2011,
Frascati, Italy. ESA Special Publication, SP-697. ISBN: 978-92-9092-261-2 (January
2012, CD).
Berardino, P., Fornaro, G., Lanari, R., Sansosti, E., 2002. A new algorithm for surface deformation monitoring based on small baseline differential SAR interferograms. IEEE
Trans. Geosci. Remote Sens. 40, 23752383.
Berardino, P., Constantini, G., Franceschetti, G., Iodice, L., Pietranera, L., Rizzo, V., 2003. Use
of differential SAR interferometry in monitoring and modelling large slope instability
at Matera (Basilicata, Italy). Eng. Geol. 68 (12), 3151.
Berger, M., Moreno, J., Johannessen, J., Levelt, P., Hanssen, R., 2012. ESA's sentinel missions
in support of Earth system science. Remote Sens. Environ. 120 (2012), 8490.
Bianchini, S., Cigna, F., Righini, G., Proietti, C., Casagli, N., 2012. Landslide hotspot mapping
by means of persistent scatterer interferometry. Environ. Earth Sci. 67 (4), 11551172.
http://dx.doi.org/10.1007/s12665-012-1559-5.
Bovenga, F., Rece, A., Nutricato, R., Guerriero, L., Chiaradia, M.T., 2005. SPINUA: a exible
processing chain for ERS/ENVISAT long term interferometry. Proceedings of ESAENVISAT Symposium, September 610, 2004, Salzburg, Austria. ESA Special Publication, SP-572 (April 2005, CD. ISBN 92-9092-883-2, ISSN 1609-042X).
Bovenga, F., Nutricato, R., Rece, A., Wasowski, J., 2006. Application of multi-temporal
differential interferometry to slope instability detection in urban/peri-urban areas.
Eng. Geol. 88 (34), 218239.
Bovenga, F., Wasowski, J., Nitti, D.O., Nutricato, R., Chiaradia, M.T., 2012a. Using Cosmo/
SkyMed X-band and ENVISAT C-band SAR interferometry for landslide analysis.
Remote Sens. Environ. 119, 272285. http://dx.doi.org/10.1016/j.rse.2011.12.013.
Bovenga, F., Rece, A., Pasquariello, G., 2012b. Using corner reectors and X-band SAR interferometry for slope instability monitoring. Proceedings of Tyrrhenian Workshop
on Advances in Radar and Remote Sensing (TyWRRS), September 1214, 2012,
Naples, Italy, pp. 114120. http://dx.doi.org/10.1109/TyWRRS.2012.6381114.
Bovenga, F., Nitti, D.O., Fornaro, G., Radicioni, F., Stoppini, A., Brigante, R., 2013. Using C/Xband SAR interferometry and GNSS measurements for the Assisi landslide analysis.
Int. J. Remote Sens. 34 (11), 40834104. http://dx.doi.org/10.1080/01431161.2013.
772310.
Cascini, L., Fornaro, G., Peduto, D., 2009. Analysis at medium scale of low-resolution
DInSAR data in slow-moving landslide affected areas. ISPRS J. Photogramm. Remote
Sens. 64 (6), 598611. http://dx.doi.org/10.1016/j.isprsjprs.2009.05.003.
Cascini, L., Fornaro, G., Peduto, D., 2010. Advanced low- and full-resolution DInSAR map
generation for slow-moving landslide analysis at different scales. Eng. Geol. 112
(14), 2942.
Casu, F., Manzo, M., Lanari, R., 2006. A quantitative assessment of the SBAS algorithm
performance for surface deformation retrieval from DInSAR data. Remote Sens.
Environ. 102, 95210.
137
Ferretti, A., Prati, C., Rocca, F., 2001. Permanent scatterers in SAR interferometry. IEEE
Trans. Geosci. Remote Sens. 39 (1), 820. http://dx.doi.org/10.1109/36.898661.
Ferretti, A., Savio, G., Barzaghi, R., Borghi, A., Musazzi, S., Novali, F., Prati, C., Rocca, F., 2007.
Submillimeter accuracy of InSAR time series: experimental validation. IEEE Trans.
Geosci. Remote Sens. 45, 11421153. http://dx.doi.org/10.1109/TGRS.2007.894440.
Ferretti, A., Fumagalli, A., Novali, F., Prati, C., Rocca, F., Rucci, A., 2011. A new algorithm for
processing interferometric data-stacks: SqueeSAR. IEEE Trans. Geosci. Remote Sens.
49 (9), 34603470. http://dx.doi.org/10.1109/TGRS.2011.2124465.
Fruneau, B., Achace, J., Delacourt, C., 1996. Observation and modeling of the SaintEtienne-de Tine'e landslide using SAR interferometry. Tectonophysics 265 (34),
181190.
Fornaro, G., Reale, D., Serano, F., 2009. Four-dimensional SAR imaging for height estimation and monitoring of single and double scatterers. IEEE Trans. Geosci. Remote Sens.
47 (1), 224237.
Fornaro, G., Paciullo, F., Reale, D., 2011. A null-space method for the phase unwrapping of
multitemporal SAR interferometric stacks. IEEE Trans. Geosci. Remote Sens. 49 (6),
23232334.
Fornaro, G., Reale, D., Verde, S., 2012. Adaptive spatial multilooking and temporal
multilinking in SBAS interferometry. IEEE Proceedings of Geoscience and Remote
Sensing Symposium (IGARSS), July 2227, 2012, Munich. http://dx.doi.org/10.1109/
IGARSS.2012.6351137.
Froese, C.R., Poncos, V., Skittow, R., Mansou, r M., Martin, D., 2008. Characterizing complex
deep seated landslide deformation using corner reection InSAR: Little Smoky Landslide, Alberta. Proceedings of the 4th Canadian Conference on Geohazards: From
Causes to Management, May 2024, 2008, Quebec City. Presse de l'Universit Laval,
Quebec.
Gostelow, P., Wasowski, J., 2004. Ground surface changes detectable by Earth observation and their impact on the stability of slopes. Pol. Geol. Inst. Spec. Pap. 15,
4352.
Guzzetti, F., Manunta, M., Ardizzone, F., Pepe, A., Cardinali, M., Zeni, G., Reichenbach, P.,
Lanari, R., 2009. Analysis of ground deformation detected using the SBAS-DInSAR
technique in Umbria, Central Italy. Pur. Appl. Geophys. 166, 14251459.
Hanssen, R., 2001. Radar Interferometry: Data Interpretation and Error AnalysisKluwer
Academic Publishers, Dordrecht, The Netherlands.
Hanssen, R., 2005. Satellite radar interferometry for deformation monitoring: a priori
assessment of feasibility and accuracy. Int. J. Appl. Earth Obs. Geoinf. 6, 253260.
Hanssen, R., van Leijen, F.J., van Zwieten, G.J., Dortland, S., Bremmer, C.N., Kleuskens, M.,
2008. Validation of PSI results of Alkmaar and Amsterdam within the Terrarma
validation project. Proceedings of Fringe 2007 Workshop, November 26-30, 2007,
Frascati, Italy. ESA Special Publication, SP-649 (February 2008, CD. ISBN 92-9291213-3, ISSN 1609-042X).
Herrera, G., Gutirrez, F., Garca-Davalillo, J.C., Guerrero, J., Notti, D., Galve, J.P., FernndezMerodo, J.A., Cooksley, G., 2013. Multi-sensor advanced DInSAR monitoring of very
slow landslides: the Tena Valley case study (Central Spanish Pyrenees). Remote
Sens. Environ. 128 (2013), 3143.
Hilley, G.E., Burgmann, R., Ferretti, A., Novali, F., Rocca, F., 2004. Dynamics of slow-moving
landslides from permanent scatterer analysis. Science 304, 19521955.
Hooper, A., Zebker, H., Segall, P., Kampes, B., 2004. A new method for measuring deformation on volcanoes and other natural terrains using InSAR persistent scatterers.
Geophys. Res. Lett. 31, L23611.
Hooper, A., 2008. A multi-temporal InSAR method incorporating both persistent scatterer
and small baseline approaches. Geophys. Res. Lett. 35, L16302.
Hooper, A., Bekaert, D., Spaans, K., Arkan, M.T., 2012. Recent advances in SAR interferometry time series analysis for measuring crustal deformation. Tectonophysics 514517,
113. http://dx.doi.org/10.1016/j.tecto.2011.10.013.
Kampes, B.M., 2005. Deformation Parameter Estimation Using Permanent Scatterer Interferometry. (PhD thesis) Delft University of Technology, Delft.
Kropatsch, W.G., Strobl, D., 1990. The generation of SAR layover and shadow amps from
digital elevation models. IEEE Geosci. Remote Sens. 28 (1), 98107.
Lauknes, T.R., Piyush Shanker, A., Dehls, J.F., Zebker, H.A., Henderson, H.C., Larsen, Y., 2010.
Detailed rockslide mapping in northern Norway with small baseline and persistent
scatterer interferometric SAR time series methods. Remote Sens. Environ. 114
(2010), 20972109.
Leroueil, S., 2001. Natural scope and cuts: movement and failure mechanisms. Geotechnique
51 (3), 197243.
Lu, P., Casagli, N., Catani, F., Tofani, V., 2012. Persistent Scatterers Interferometry Hotspot
and Cluster Analysis (PSI-HCA) for detection of extremely slow-moving landslides.
Int. J. Remote Sens. 33, 466489.
Mahapatra, P., Samiei-Esfahany, S., Hansen, R., 2012. Towards repeatability, reliability and
robustness in time-series InSAR. Proceedings of Fringe 2011 Workshop, September
1923, 2011, Frascati, Italy. ESA Special Publication, SP-697 (January 2012, CD. ISBN
978-92-9092-261-2, ISSN 1609-042X).
Marinkovic, P., Ketelaar, G., van Leijen, F., Hanssen, R., 2008. InSAR quality control: analysis of ve years of corner reector time series. Proceedings of Fringe 2007 Workshop,
November 2630, 2007, Frascati, Italy. ESA Special Publication, SP-649 (February
2008, CD. ISBN 92-9291-213-3, ISSN 1609-042X).
Meisina, C., Zucca, F., Fossati, D., Ceriani, M., Allievi, J., 2006. Ground deformation monitoring by using the permanent scatterers technique: the example of the Oltrepo Pavese
(Lombardia, Italy). Eng. Geol. 88, 240259.
Meisina, C., Zucca, F., Notti, D., Colombo, A., Cucchi, A., Savio, G., Giannico, C., Bianchi, M.,
2008. Geological interpretation of PSInSAR data at regional scale. Sensors 11,
74697492.
Meisina, C., Notti, D., Zucca, F., Ceriani, M., Colombo, A., Poggi, F., Roccati, A., Zaccone, A.,
2013. The use of PSInSAR and SqueeSAR techniques for updating landslide inventories. In: Margottini, C., et al. (Eds.), Landslide Science and Practice, vol. 1. SpringerVerlag, Berlin Heidelberg, pp. 8187.
138
Monserrat, O., Crosetto, M., Cuevas, M., Crippa, B., 2011. The thermal expansion component of persistent scatterer interferometry observations. IEEE Geosci. Remote. Sens.
Lett. 8 (5), 864868. http://dx.doi.org/10.1109/LGRS.2011.2119463.
Mora, O., Mallorqui, J.J., Broquetas, A., 2003. Linear and nonlinear terrain deformation
maps from a reduced set of interferometric SAR images. IEEE Trans. Geosci. Remote
Sens. 41, 22432253.
Ng, C.W.W., Zhan, L.T., Bao, C.G., Fredlund, D.G., Gong, B.W., 2003. Performance of an
unstaurated expansive soil slope subjected to articial rainfall inltration.
Geotechnique 53 (2), 143157.
Nitti, D.O., Nutricato, R., Bovenga, F., Rece, A., Chiaradia, M.T., Guerriero, L., 2009a.
TerraSAR-X InSAR multi-pass analysis on Venice (Italy). Proceedings of SPIE. 7477,
74771I (2009). Image and Signal Processing for Remote Sensing, XV. http://dx.doi.
org/10.1117/12.832047 (31 August - 03 September 2009, Berlin, Germany).
Nitti, D.O., Bovenga, F., Rece, A., Wasowski, J., Conte, D., Nutricato, R., 2009b. L- and Cband SAR interferometry analysis of the Wieliczka salt mine area (UNESCO heritage
site, Poland). Proceedings ALOS PI 2008 Symposium, 37 Nov., 2008, Rhodes,
Greece. ESA, SP-664 (January 2009).
Nitti, D.O., Bovenga, F., Nutricato, R., Intini, F., Chiaradia, M.T., 2013. On the use of COSMO/
SkyMed data and weather models for interferometric DEM generation. Eur. J. Remote
Sens. 46, 250271. http://dx.doi.org/10.5721/EuJRS20134614.
Notti, D., Davalillo, J.C., Herrera, G., Mora, O., 2010. Assessment of the performance of Xband satellite radar data for landslide mapping and monitoring: Upper Tena valley
case study. Nat. Hazards Earth Syst. Sci. 10, 18651875.
Notti, D., Meisina, C., Zucca, F., Crosetto, M., Montserrat, O., 2012a. Factors that have an
inuence on time series. Proceedings of Fringe 2011 Workshop, September 1923,
2011, Frascati, Italy. ESA Special Publication, SP-697 (January 2012, CD. ISBN 97892-9092-261-2, ISSN 1609-042X).
Notti, D., Meisina, C., Zucca, F., Colombo, A., 2012b. Models to predict persistent scatterers
data distribution and their capacity to register movement along the slope. Proceedings of Fringe 2011 Workshop, September 19-23, 2011, Frascati, Italy. ESA
Special Publication, SP-697 (January 2012, CD. ISBN 978-92-9092-261-2, ISSN
1609-042X).
Parizzi, A., Brcic, R., 2011. Adaptive InSAR stack multilooking exploiting amplitude statistics: a comparison between different techniques and practical results. IEEE Geosci.
Remote Sens. Lett. 8 (3), 441445.
Pepe, A., Lanari, R., 2006. On the extension of the minimum cost ow algorithm for phase
unwrapping of multitemporal differential SAR interferograms. IEEE Trans. Geosci.
Remote Sens. 44 (9), 23742383.
Perissin, D., Rocca, F., 2006. High-accuracy urban DEM using permanent scatterers. IEEE
Trans. Geosci. Remote Sens. 44 (11), 33383347.
Perissin, D., Ferretti, A., 2007. Urban-target recognition by means of repeated spaceborne
SAR images. IEEE Trans. Geosci. Remote Sens. 45 (12), 40434058.
Plank, S., Singer, J., Minet, C., Thuro, K., 2012a. Pre-survey evaluation of the differential
synthetic aperture radar interferometry method for landslide monitoring. Int. J. Remote Sens. 33 (20), 66236637.
Plank, S., Singer, J., Thuro, K., 2012b. Estimation of the persistent scatterer density using
optical remote sensing data and land cover data. Proceedings of Fringe 2011 Workshop, September 1923, 2011, Frascati, Italy. ESA Special Publication, SP-697 (January
2012, CD. ISBN 978-92-9092-261-2, ISSN 1609-042X).
Plank, S., Singer, J., Thuro, K., 2013. Assessment of number and distribution of persistent
scatterers prior to radar acquisition using open access land cover and topographical
data. ISPRS J. Photogramm. Remote Sens. 85, 132147.
Prati, C., Ferretti, A., Perissin, D., 2010. Recent advances on surface ground deformation
measurement by means of repeated space-borne SAR observations. J. Geodyn. 49,
161170.
Puzrin, A.M., Schmid, A., Schwager, M.V., 2012. Case studies of constrained creeping
landslides in Switzerland. In: Eberhardt, E., Froese, C., Turner, A.K., Leroueil, S. (Eds.),
Proceedings of the 11th International and 2nd North American Symposium on Landslides, Banff (Canada), 38 June, 2012. Landslides and Engineered Slopes, vol. 2. CRC
Press/Balkema, Leiden, The Netherlands, pp. 17951800.
Raucoules, D., Bourgine, B., De Michele, M., Le Cozanet, G., Closset, L., Bremmer, C.,
Veldkamp, H., Tragheim, D., Bateson, L., Crosetto, M., Agudo, M., Engdahl, M., 2009.
Validation and intercomparison of persistent scatterers interferometry: PSIC4 project
results. J. Appl. Geophys. 68 (3), 335347.
Reale, D., Nitti, D.O., Peduto, D., Nutricato, R., Bovenga, F., Fornaro, G., 2011. Post-seismic
deformation monitoring with the COSMO/SKYMED constellation. IEEE Geosci.
Remote Sens. Lett. 8 (4), 696-00. http://dx.doi.org/10.1109/LGRS.2010.2100364.
Riddick, S.N., Schmidt, D.A., Deligne, N.I., 2012. An analysis of terrain properties and the
location of surface scatterers from persistent scatterer interferometry. ISPRS J.
Photogramm. Remote Sens. 73 (2012), 5057.
Righini, G., Pancioli, V., Casagli, N., 2011. Updating landslide inventory maps using
Persistent Scatterer Interferometry (PSI). Int. J. Remote Sens. 33 (7), 20682096.
Rucci, A., Ferretti, A., Monti Guarnieri, A., Rocca, F., 2012. Sentinel 1 SAR interferometry
applications: the outlook for sub millimeter measurements. Remote Sens. Environ.
120, 156163. http://dx.doi.org/10.1016/j.rse.2011.09.030.
Sansosti, E., Casu, F., Manzo, M., Lanari, R., 2010. Space-borne radar interferometry techniques for the generation of deformation time series: an advanced tool for Earth's
surface displacement analysis. Geophys. Res. Lett. 37 (20), L20305.
Tang, C., Rengers, N., van Asch, T.W.J., Yang, Y.H., Wand, G.F., 2011. Triggering conditions
and depositional characteristic of a disastrous debris ow even in Zhouqu city, Gansu
Province, northwestern China. Nat. Hazards Earth Syst. Sci. 11, 29032912. http://dx.
doi.org/10.5194/nhess-11-2903-2011.
Usai, S., 2003. A least squares database approach for SAR interferometric data. IEEE Trans.
Geosci. Remote Sens. 41 (4), 753760 (part 1).
Terrarma, 2007. Terrarma project. http://terrarma.eu.com/Terrarmavalidation.htm.
Van der Kooij, M., Hughes, W., Sato, S., Poncos, V., 2006. Coherent target monitoring at
high spatial density, examples of validation results. Proceedings of Fringe 2005 Workshop, November 28December 2, 2005, Frascati, Italy. ESA Special Publication, SP-610
(February 2006, CD. ISBN 92-9092-261-9, ISSN 1609-042X).
Wasowski, J., Casarano, D., Bovenga, F., Conte, D., Nutricato, R., Rece, A. Berardino P.,
Manzo, M., Pepe, A., Zeni, G., Lanari, R., 2007a. A comparative analysis of the DInSAR
results achieved by the SBAS and SPINUA techniques: a case study of the Maratea
valley, Italy. Proceedings of the Envisat Symposium, April 2327, 2007, Montreux,
Switzerland. ESA Special Publication, SP-636 (July 2007, CD).
Wasowski, J., Ferretti, A., Colesanti, C., 2007b. Space-borne SAR interferometry for long
term monitoring of slope instability hazards. Proceedings of 1st North American
Landslide Conference, 38 June, 2007, Vail, U.S.A. , pp. 234243.
Wasowski, J., Bovenga, F., Nitti, D.O., Nutricato, R., 2012. Investigating landslides with Persistent Scatterers Interferometry (PSI): current issues and challenges. In: Eberhardt,
E., Froese, C., Turner, A.K., Leroueil, S. (Eds.), Proceedings of the 11th International
and 2nd North American Symposium on Landslides, Banff (Canada), 38 June,
2012. Landslides and Engineered Slopes, vol. 2. CRC Press/Balkema, Leiden, The
Netherlands, pp. 12951301.
Wenxue, F., Huadong, G., Qingjiu, T., Xiaofang, G., 2010. Landslide monitoring by corner
reectors differential interferometry SAR. Int. J. Remote Sens. 31 (24), 63876400.
Werner, C., Wegmuller, U., Strozz1, T., Wiesmaenn, A., 2003. Interferometric point target
analysis for deformation mapping. IEEE Proceedings of Geoscience and Remote
Sensing Symposium (IGARSS), 2125 July 2003. Toulouse, France, 7, pp. 43624364.
http://dx.doi.org/10.1109/IGARSS.2003.1295516.
Wright, T.J., Parsons, B.E., Lu, Z., 2004. Toward mapping surface deformation in three
dimensions using InSAR. Geophys. Res. Lett. 31, L01607. http://dx.doi.org/10.1029/
2003GL018827.