Abstract and Figures

This work is focused on deformation activity mapping and monitoring using Sentinel-1 (S-1) data and the DInSAR (Differential Interferometric Synthetic Aperture Radar) technique. The main goal is to present a procedure to periodically update and assess the geohazard activity (volcanic activity, landslides and ground-subsidence) of a given area by exploiting the wide area coverage and the high coherence and temporal sampling (revisit time up to six days) provided by the S-1 satellites. The main products of the procedure are two updatable maps: the deformation activity map and the active deformation areas map. These maps present two different levels of information aimed at different levels of geohazard risk management, from a very simplified level of information to the classical deformation map based on SAR interferometry. The methodology has been successfully applied to La Gomera, Tenerife and Gran Canaria Islands (Canary Island archipelago). The main obtained results are discussed.
Content may be subject to copyright.
remote sensing
Article
A Methodology to Detect and Update Active
Deformation Areas Based on Sentinel-1 SAR Images
Anna Barra 1, *ID , Lorenzo Solari 2ID , Marta Béjar-Pizarro 3ID , Oriol Monserrat 1,
Silvia Bianchini 2, Gerardo Herrera 3ID , Michele Crosetto 1, Roberto Sarro 3,
Elena González-Alonso 4, Rosa María Mateos 5, Sergio Ligüerzana 4, Carmen López 4and
Sandro Moretti 2
1Centre Tecnològic de Telecomunicacions de Catalunya (CTTC/CERCA), Geomatics Division,
08860 Castelldefels, Spain; omonserrat@cttc.cat (O.M.); mcrosetto@cttc.cat (M.C.)
2Earth Sciences Department, University of Firenze, Via La Pira, 4, I-50121 Firenze, Italy;
lorenzo.solari@unifi.it (L.S.); silvia.bianchini@unifi.it (S.B.); sandro.moretti@unifi.it (S.M.)
3Geohazards InSAR Laboratory and Modeling Group (InSARlab), Geoscience Research Department,
Geological Survey of Spain (IGME), Alenza 1, 28003 Madrid, Spain; m.bejar@igme.es (M.B.-P.);
g.herrera@igme.es (G.H.); r.sarro@igme.es (R.S.)
4Centro Nacional de Información Geográfica, Instituto Geográfico Nacional, C/General Ibáñez de Ibero, 3,
28003 Madrid, Spain; egalonso@fomento.es (E.G.-A.); cnig.slr@fomento.es (S.L.);
clmoreno@fomento.es (C.L.)
5Geological Survey of Spain (IGME), Urb. Alcázar del Genil, 4-Edif. Bajo, 18006 Granada, Spain;
rm.mateos@igme.es
*Correspondence: anna.barra@cttc.cat or abarra@cttc.cat; Tel.: +34-93-645-2900
Received: 4 August 2017; Accepted: 21 September 2017; Published: 28 September 2017
Abstract:
This work is focused on deformation activity mapping and monitoring using Sentinel-1 (S-1)
data and the DInSAR (Differential Interferometric Synthetic Aperture Radar) technique. The main
goal is to present a procedure to periodically update and assess the geohazard activity (volcanic
activity, landslides and ground-subsidence) of a given area by exploiting the wide area coverage and
the high coherence and temporal sampling (revisit time up to six days) provided by the S-1 satellites.
The main products of the procedure are two updatable maps: the deformation activity map and
the active deformation areas map. These maps present two different levels of information aimed at
different levels of geohazard risk management, from a very simplified level of information to the
classical deformation map based on SAR interferometry. The methodology has been successfully
applied to La Gomera, Tenerife and Gran Canaria Islands (Canary Island archipelago). The main
obtained results are discussed.
Keywords: SAR; DInSAR; deformation; measurement; landslide; subsidence; risk management
1. Introduction
This paper is focused on geohazard activity mapping and monitoring using Sentinel-1 (S-1) data
and the DInSAR (Differential Interferometric Synthetic Aperture Radar) technique. In the last 25 years,
the mapping and monitoring of geohazard phenomena have received an important contribution from
the DInSAR technique. This approach was firstly proposed in 1989, using data from the L-band
Seasat sensor [
1
]. Since then, the technique has experienced a continuous growth mainly related
to two main components. The first one is the important research and development effort made in
this period, which has generated a wide number of data processing and analysis tools and methods.
They include the classical single-interferogram DInSAR methods (e.g., see [
2
4
]), the DInSAR stacking
techniques [
5
] and several implementations of the so-called Persistent Scatterer Interferometry (PSI)
Remote Sens. 2017,9, 1002; doi:10.3390/rs9101002 www.mdpi.com/journal/remotesensing
Remote Sens. 2017,9, 1002 2 of 19
and Small Baseline Subsets (SBAS) methods [
6
10
]. A review of all these advanced methods, which
are sometimes referred to as Advanced DInSAR (A-DInSAR) or Time Series Radar Interferometry
(TSInSAR), is provided in [
11
]. In the last years, the number of DInSAR users has increased thanks
to the availability of free platforms and software, such as Grid Processing on Demand (G-POD) and
Sentinel Application Platform (SNAP), provided by ESA, which have widened the range of potential
users [12,13].
The second component is satellite data availability, which has increased in terms of number of
satellites with different spatial and temporal resolutions. Most of the DInSAR and PSI developments
have been based on C-band data acquired by the sensors on-board the satellites ERS-1/2, Envisat
and Radarsat. The available imagery collected by these satellites cover long periods of time (starting
from 1992), a key aspect to guarantee a long-term deformation monitoring and to make historical
studies [
14
]. DInSAR and PSI have experienced a major step forward since 2007, with the advent of very
high-resolution X-band data [
15
] of TerraSAR-X and COSMO-SkyMed. This includes the capability
to generate a dense sampling of Persistent Scatterers (PS), a high sensitivity to small displacements
and a remarkable quality improvement of the time series with respect to the C-band [
16
,
17
]. Those
improvements have had an important impact on the geohazard applications, improving the analysis at
different scales and allowing the combination of the results from different satellites [
18
24
]. A review
over the available satellite SAR sensors and their potentialities for landslide application can be found
in [
25
,
26
]. A significant further improvement is given by the new C-band sensor on-board the S-1A
and B satellites, launched on 2014 and 2016, respectively [
27
]. S-1 has improved the data acquisition
throughout and, compared to previous sensors, has increased considerably the DInSAR and PSI
deformation monitoring potential [
4
,
28
] allowing to make long-term geohazard monitoring planes
over regional areas [29].
This work is aimed at exploiting the wide area coverage and the high coherence [
4
] and temporal
sampling (revisit time up to six days) provided by the S-1 satellites to generate and periodically update
regional-scale deformation activity maps for the geohazard management. The proposed methodology
has been developed in the framework of the ongoing European ECHO (European Civil Protection
and Humanitarian Aid Operations) project “Safety—Sentinel for geohazards regional monitoring
and forecasting”, which aims at providing Civil Protection Authorities (CPA) with the capability of
periodically evaluate and assess, at regional scale, the potential impact of geohazards (volcanic activity,
landslides and subsidence) on urban areas.
The interpretation of the DInSAR derived maps (e.g., velocity maps) can be complex, mostly for
users who are not familiar with radar data [
30
,
31
]. This is more evident working at regional scale,
where the high number of PSs can difficult the analysis and in some cases misinterpret the real scenario.
Several authors have shown different approaches to address this issue [
25
,
32
]. This work presents a
procedure to generate clear products that can be easily exploited by the authorities involved in the
geohazard and risk management chain. The main output is the so-called Active Deformation Areas
(ADA) map. It is derived from the DInSAR Deformation Activity Map (DAM) by discriminating
the more reliable deforming areas. A further step, which is the integration of the products of the
methodology (DAM and ADA maps) in the Civil Protection risk management activities, is described
in [33].
The procedure is illustrated through its application over the Canary Islands, a Spanish volcanic
archipelago located in the Atlantic Ocean, northwest of Africa, which is one of the test sites of the
Safety project. Canary Islands present different types of geohazards, including landslides, earthquakes
and volcanic activity.
The paper starts with the description of the procedure (Section 2), then the results of the active
deformation maps obtained over the Canary Islands test site are described (Section 3). This is followed
by the discussion of the results by emphasizing the main advantages and main challenges of the
proposed approach (Section 4). Finally, the conclusions of the work are drawn (Section 5).
Remote Sens. 2017,9, 1002 3 of 19
2. Methodology
In this section, the procedure to derive the DAM and the ADA maps is described. The proposed
procedure can be applied to the data acquired by any satellite SAR sensor. However, it provides the
best performances with the S-1 characteristics.
The general scheme of the procedure, shown as a flowchart in Figure 1, is divided in two main
blocks (Figure 1):
1.
Raw Deformation Map (RDM) generation: This includes all the PSI processing steps to estimate
the annual linear velocities and the time series of deformation (TS). The RDM is an intermediate
product that is not delivered to the final users.
2.
Deformation Activity Map (DAM) generation and Active Deformation Areas (ADA) extraction:
In this block, the two final products of the procedure are generated. It includes a filtering of the
RDM and all the steps to generate the ADA map. These two products are easily readable and
thus exploitable by the risk management decision makers.
Remote Sens. 2017, 9, 1002 3 of 19
The general scheme of the procedure, shown as a flowchart in Figure 1, is divided in two main
blocks (Figure 1):
1. Raw Deformation Map (RDM) generation: This includes all the PSI processing steps to estimate
the annual linear velocities and the time series of deformation (TS). The RDM is an intermediate
product that is not delivered to the final users.
2. Deformation Activity Map (DAM) generation and Active Deformation Areas (ADA) extraction:
In this block, the two final products of the procedure are generated. It includes a filtering of the
RDM and all the steps to generate the ADA map. These two products are easily readable and
thus exploitable by the risk management decision makers.
Figure 1. Flow chart of the proposed procedure.
All the deformation values included in the output maps are estimated along the satellite Line of
Sight (LOS) direction. The procedure is designed to be periodically processed to have a continuous
update of the products and thus a continuous input of regional-scale deformation maps for the
authorities to detect potential hazards or to decide more focused analysis in critical areas.
2.1. Raw Deformation Map Generation
The main goal of this block is to derive the deformation scenario of an area of interest from the
SAR data. The output is a deformation map that consists in a set of selected points with both the
information of the estimated LOS velocity and the accumulated displacement at every satellite
acquisition. The main input is a set of SAR images acquired at different times. Several Persistent
Scatterer Interferometry (PSI) techniques have been developed in the last decade. The main common
steps to generate a deformation map are: the interferogram network generation, the selection of
points, the phase unwrapping, the Atmospheric Phase Screen (APS) estimation and removal and the
estimation of the velocities and/or deformation time series (TS). The choice between the different
techniques depends on many factors like the radar sensor characteristics, the target of the study or
the characteristics of the test site (geology, land use, topography, etc.). In particular, for this research,
the maps have been generated using an approach of the Persistent Scatterer Interferometry chain of
the Geomatics (PSIG) Division of CTTC (PSIG) described in [32]. The main steps of the processing are
briefly described in the following lines (Figure 2):
Interferogram network generation: This step consists of the generation of the interferogram
network. S-1 uses a sophisticated data acquisition procedure, the TOPS (Terrain Observation by
Progressive Scan) imaging mode [34], which is key to achieve the wide area coverage. The
drawback is that, compared to other sensors, the S-1 data require extra processing. The key step
is the image co-registration, which needs to be very accurate [35].
Since a fundamental aspect of the PSIG chain is the redundancy of the network of
interferograms and images, all the possible interferogram pairs are generated. The selection of
the interferogram network is done by statistically evaluating the coherence of the study area.
Figure 1. Flow chart of the proposed procedure.
All the deformation values included in the output maps are estimated along the satellite Line of
Sight (LOS) direction. The procedure is designed to be periodically processed to have a continuous
update of the products and thus a continuous input of regional-scale deformation maps for the
authorities to detect potential hazards or to decide more focused analysis in critical areas.
2.1. Raw Deformation Map Generation
The main goal of this block is to derive the deformation scenario of an area of interest from
the SAR data. The output is a deformation map that consists in a set of selected points with both
the information of the estimated LOS velocity and the accumulated displacement at every satellite
acquisition. The main input is a set of SAR images acquired at different times. Several Persistent
Scatterer Interferometry (PSI) techniques have been developed in the last decade. The main common
steps to generate a deformation map are: the interferogram network generation, the selection of
points, the phase unwrapping, the Atmospheric Phase Screen (APS) estimation and removal and the
estimation of the velocities and/or deformation time series (TS). The choice between the different
techniques depends on many factors like the radar sensor characteristics, the target of the study or
the characteristics of the test site (geology, land use, topography, etc.). In particular, for this research,
the maps have been generated using an approach of the Persistent Scatterer Interferometry chain of
the Geomatics (PSIG) Division of CTTC (PSIG) described in [
32
]. The main steps of the processing are
briefly described in the following lines (Figure 2):
Interferogram network generation: This step consists of the generation of the interferogram
network. S-1 uses a sophisticated data acquisition procedure, the TOPS (Terrain Observation
Remote Sens. 2017,9, 1002 4 of 19
by Progressive Scan) imaging mode [
34
], which is key to achieve the wide area coverage.
The drawback is that, compared to other sensors, the S-1 data require extra processing. The key
step is the image co-registration, which needs to be very accurate [35].
Since a fundamental aspect of the PSIG chain is the redundancy of the network of interferograms
and images, all the possible interferogram pairs are generated. The selection of the interferogram
network is done by statistically evaluating the coherence of the study area. This analysis provides
key inputs for the network like the maximum temporal baseline to be used as well as the presence
of periods characterized by low coherence (e.g., snow periods in mountain areas). As example,
in the Canary Islands test site, the selected maximum temporal baseline was 156 days.
Point Selection: Even if a single S-1 frame contains millions of pixels, only a small portion of them
is exploitable for deformation purposes. There are different statistical criteria used to discriminate
the noisier pixels from those with low level of noise [
11
]. However, the use of very restrictive
thresholds can result in a critical loss of spatial coverage. The general purpose of this step is
to find a good compromise between the quality of the selected points (little affected by noise)
and a good spatial coverage. Hence, for each case, different criteria are evaluated in order to
find the best trade-off. For example, in the Canary Islands test site, the selection of points was
based on the Dispersion of Amplitude (DA) [
6
]. Only points with a DA value lower than 0.5 have
been selected.
2+1D phase unwrapping: This is a two-step spatial-temporal phase unwrapping [
32
]. The approach
starts with a spatial phase unwrapping (2D) performed over the selected set of points and for
each interferogram of the network. Then, in a second phase, a phase unwrapping consistency
check (1D) is performed. This check is done point wise, exploiting the temporal component of the
SAR images stack. It is based on an iterative least squares method (LS) and the analysis of the LS
residuals at each iteration. For each pixel, the main outputs are: (i) the temporal evolution of the
phases (TEP) with respect to a reference image; and (ii) some statistical parameters used to assess
the quality of the LS inputs.
APS (Atmospheric Phase Screen) estimation and removal: The APS is estimated using spatial-temporal
filters [
36
]. The main input is the TEP estimated in the previous step. The estimated APS is removed
from the TEP. The remaining phases are then transformed into deformations, obtaining the final
deformation time series (TS).
Deformation velocity estimation: This is the last step of the deformation map generation block.
It consists of an estimation of the deformation velocity from the obtained time series. The used
method is a robust regression line estimation.
Remote Sens. 2017, 9, 1002 4 of 19
This analysis provides key inputs for the network like the maximum temporal baseline to be
used as well as the presence of periods characterized by low coherence (e.g., snow periods in
mountain areas). As example, in the Canary Islands test site, the selected maximum temporal
baseline was 156 days.
Point Selection: Even if a single S-1 frame contains millions of pixels, only a small portion of
them is exploitable for deformation purposes. There are different statistical criteria used to
discriminate the noisier pixels from those with low level of noise [11]. However, the use of very
restrictive thresholds can result in a critical loss of spatial coverage. The general purpose of this
step is to find a good compromise between the quality of the selected points (little affected by
noise) and a good spatial coverage. Hence, for each case, different criteria are evaluated in order
to find the best trade-off. For example, in the Canary Islands test site, the selection of points was
based on the Dispersion of Amplitude (DA) [6]. Only points with a DA value lower than 0.5
have been selected.
2+1D phase unwrapping: This is a two-step spatial-temporal phase unwrapping [32]. The
approach starts with a spatial phase unwrapping (2D) performed over the selected set of points
and for each interferogram of the network. Then, in a second phase, a phase unwrapping
consistency check (1D) is performed. This check is done point wise, exploiting the temporal
component of the SAR images stack. It is based on an iterative least squares method (LS) and the
analysis of the LS residuals at each iteration. For each pixel, the main outputs are: (i) the temporal
evolution of the phases (TEP) with respect to a reference image; and (ii) some statistical
parameters used to assess the quality of the LS inputs.
APS (Atmospheric Phase Screen) estimation and removal: The APS is estimated using spatial-
temporal filters [36]. The main input is the TEP estimated in the previous step. The estimated
APS is removed from the TEP. The remaining phases are then transformed into deformations,
obtaining the final deformation time series (TS).
Deformation velocity estimation: This is the last step of the deformation map generation block.
It consists of an estimation of the deformation velocity from the obtained time series. The used
method is a robust regression line estimation.
Figure 2. Flow chart of the Row Deformation Map (RDM) estimation.
Figure 2. Flow chart of the Row Deformation Map (RDM) estimation.
Remote Sens. 2017,9, 1002 5 of 19
The final output of this block is a raw deformation map (RDM) including, for each point,
the deformation velocity and the accumulated deformation at each acquisition time (TS). The estimated
deformations are in the satellite LOS direction.
It is worth noting that the described approach can slightly change depending on the site. A frequent
variation is to perform the deformation velocity estimation before the 2+1D phase unwrapping.
The deformation velocity is estimated over the wrapped phases and then removed from them before
the phase unwrapping [
5
]. This is done to ease the 2D phase unwrapping step in areas strongly affected
by deformation.
2.2. Deformation Activity Map and Active Deformation Areas Extraction
This block is aimed at obtaining both the final Deformation Activity Map (DAM), which is the
filtered version of the raw deformation map (RDM), and the Active Deformation Areas (ADA) map,
which is the main product of the procedure. The main goal is to identify and monitor, over wide
areas, the most critical deformations to provide the Civil Protection authorities with the capacity to
perform prevention and mitigation actions. Therefore, the three main aspects that have to characterize
the final maps are: (i) the readability; (ii) the reliability; and (iii) the regional-to-local scale. The main
constraining factors to achieve these goals are the spatio-temporal noise of the deformation map and
the high number of PSs which in some cases can lead to wrong interpretations.
A key parameter of this block is the assessment of the general noise level (sensitivity) of the RDM.
In this research, the sensitivity has been evaluated using the standard deviation (
σmap
) of the RDM
velocity values. A stability threshold of 2
σmap
is set to distinguish the active points, those where we
measure movements, from those we do not. A point is considered moving if |v|>2
σmap
, where
|v| is the absolute velocity value of the point. It is worth to underline that the points classified as
“stable” can be truly stable as well as instable points, with a not detectable movement. To simplify the
readability, we call “stable” all the points with the absolute LOS velocity below the stability threshold.
As example, in the Canary Island test site, the stability threshold has been set as ±4.7 mm/year.
This block can be summarized in three main steps (Figure 3): (i) filtering of the RDM; (ii) automatic
extraction of the more reliable and relevant active areas (ADA); and (iii) Quality Index (QI) attribution
to each ADA.
(i) Filtering of the raw deformation map
This action aims to filter the RDM obtained in block 1. The final point selection has been based
on two different criteria: (i) the standard deviation of the 2+1D phase unwrapping residues (
σres
);
and (ii) a spatial criterion based on the variability of a point with respect its neighbors. The first one
is used to remove points susceptible to be affected by phase unwrapping errors. The used threshold
for the
σres
is 2.4 rad (approximately 1 cm). We have selected this relatively high threshold in order
to keep the maximum number of measurements. Regarding the spatial criterion, it is used to clean
sparse measurements (isolated points) and points with strong discrepancy with respect its neighbors
(outliers). The filtering is window based. The used window has a radius of 2 times the data resolution
(e.g., around 80 m in Canary Island).
The filtering criteria are: (i) eliminate points without neighbors inside the window; and
(ii) eliminate moving points without more than one moving neighbors inside the window. It is
worth noting that the eliminated points can be real moving points related to a geohazard, like for
example a landslide. However, an isolated point can be related to several factors including noise. In
this context, we accept to lose information about few phenomena in order to highly reduce the general
level of noise and simplify the readability of the map.
Remote Sens. 2017,9, 1002 6 of 19
Remote Sens. 2017, 9, 1002 6 of 19
Figure 3. Flowchart of the Deformation Activity Map (DAM) and the Active Deformation Areas
(ADA) maps generation.
(ii) Automatic extraction of the more reliable and relevant active areas (ADA)
The aim of the ADA map generation is to perform a rapid identification of the most reliable
active deformation areas. The final map has to represent a clear input to be validated and integrated
with other data (e.g., geohazard inventories, ground truth information, etc.) in order to determine the
nature of the deformation and thus to generate the Geohazard Activity Map.
The ADA map has been by using an evolution of the approaches proposed by [37–40]. The main
input is the filtered deformation velocity map obtained in Step (i). Only the moving points (with |v|
> 2σ
map
) are selected. Then, from this subset of points (PSm), groups of at least five neighbor PSs,
sharing their influence area, are aggregated in polygons representing the Active Deformation Areas
(ADA). To define the influence area of every PS we consider the approximated footprint of the PSs.
For example, in Canary Island test site, the PS area is of 28 m by 40 m. Then we calculate the radius
of the circle inscribing the PS area (40 m by 40 m in the Canary Island Site) and we multiply it by a
factor of 1.3 to ensure that neighboring pixels are selected. If the grouped PSs are less than five, they
are considered to represent a non-significant deformation for a regional scale map.
Finally, for each ADA, the following parameters are estimated:
- Number of aggregated active points (APs).
- Mean, maximum and minimum values of the APs velocities.
- Mean value of the APs accumulated deformations. To avoid strong influence of
atmospheric or digital elevation model error effects, we estimate the final accumulated
Figure 3.
Flowchart of the Deformation Activity Map (DAM) and the Active Deformation Areas (ADA)
maps generation.
(ii) Automatic extraction of the more reliable and relevant active areas (ADA)
The aim of the ADA map generation is to perform a rapid identification of the most reliable
active deformation areas. The final map has to represent a clear input to be validated and integrated
with other data (e.g., geohazard inventories, ground truth information, etc.) in order to determine the
nature of the deformation and thus to generate the Geohazard Activity Map.
The ADA map has been by using an evolution of the approaches proposed by [
37
40
]. The main
input is the filtered deformation velocity map obtained in Step (i). Only the moving points (with |v|
> 2
σmap
) are selected. Then, from this subset of points (PSm), groups of at least five neighbor PSs,
sharing their influence area, are aggregated in polygons representing the Active Deformation Areas
(ADA). To define the influence area of every PS we consider the approximated footprint of the PSs.
For example, in Canary Island test site, the PS area is of 28 m by 40 m. Then we calculate the radius
of the circle inscribing the PS area (40 m by 40 m in the Canary Island Site) and we multiply it by a
factor of 1.3 to ensure that neighboring pixels are selected. If the grouped PSs are less than five, they
are considered to represent a non-significant deformation for a regional scale map.
Finally, for each ADA, the following parameters are estimated:
- Number of aggregated active points (APs).
- Mean, maximum and minimum values of the APs velocities.
Remote Sens. 2017,9, 1002 7 of 19
-
Mean value of the APs accumulated deformations. To avoid strong influence of atmospheric
or digital elevation model error effects, we estimate the final accumulated deformation as
the average of the accumulated values of the last four acquisition times of all the APs of
the ADA.
-
Velocity class, which is a classification of the ADA as a function of its maximum velocity
(vm). The class is 1 if |vm| > 1 cm/year or 0 if 2σmap < |vm| < 1 cm/year.
- Quality Indexes, which are explained in the following lines.
(iii) Quality index attribution to each ADA
Although the ADA map is based on filtered data, the automation of the process needs a final
quality assessment for each single ADA. The noise level of the TSs and thus the robustness of the
deformation estimations vary significantly pointwise. This step describes an implemented Quality
Index (QI) that provides an estimation of the noise level of the ADA. This QI is a key parameter to
properly interpret the ADA map.
The QI is based on the evaluation of two parameters for each ADA: the noise of each AP time
series is evaluated and the spatial homogeneity of the estimated deformations in time is considered
(i.e., consistency between AP time series). Hence, each ADA is characterized by a temporal noise index
(TNI) and a spatial noise index (SNI) that are used to derive the final QI.
Temporal Noise Index (TNI)
To attribute the TNI, for each APs the first order autocorrelation (
ρt,t1
) of its TS is evaluated.
The first order autocorrelation allows evaluating the temporal noise degree for both linear and
nonlinear deformation trends. The autocorrelation coefficient ranges from 0 to 1, where 0 means the
prevalence of noise over the deformation trend. The temporal index (TNI) is a four-class classification
of the ADA based on the median value (Med(
ρ
)) between the TS autocorrelation coefficients, it varies
from 1 to 4, where 1 corresponds to a high Med(ρ)and 4 to a very low Med(ρ).
To find a relationship between the autocorrelation coefficient and the noise level, a simulation
was performed on a set of 20 TSs characterized by a linear trend (b) and random noise ():
Y=bx+
where b = velocity in (mm/day), x = 0, 12, 24, 36,
. . .
, 468 (days) is the 12 days spaced time
series. The random noise was characterized by a normal distribution. For each velocity, we tested
different levels of noise by setting the standard deviation of the simulated random noise. Since the
autocorrelation also depends on the number of sampling (i.e., on the length of the TS and the revisit
time), our simulation was calculated over 468-day time series (almost one year and half) with temporal
steps of 12 days. This period corresponds to the temporal window used to generate the deformation
maps in our test site (Canary Islands). Figure 4shows the results of the two simulations performed
with the velocities of 5 and 10 mm/year. The plotted values represent the median value of the set of
TSs autocorrelation coefficients calculated for each tested noise level. By expressing the noise level in
terms of percentage of the velocities, the relationship can be approximated by the linear regression of
the whole data resulted by the simulation.
It is worth mentioning that this is a simplified model, helpful to evaluate the physical meaning of
the thresholds. Note that, for equal noise level, the autocorrelation coefficient changes with temporal
sampling. Therefore, the thresholds can change, depending on the study case, if both the Sentinel-1A
and Sentinel-1B images are used (six-day revisit time).
The
ρt,t1
values of 0.53, 0.70 and 0.84, respectively, corresponding to the 35%, 25% and 15% of
noise with respect to the velocities (Figure 4and Table 1), were chosen as thresholds for the four classes
of TNI.
Remote Sens. 2017,9, 1002 8 of 19
Figure 4.
Plot of autocorrelation coefficient vs. level of noise (%). The level of noise is defined by the
standard deviation of the simulated normal distributed random values. The blue lines indicate the
selected thresholds for the Temporal Noise Index (TNI) classification.
Table 1. Final classification of TNI.
Med(ρ)Noise-Velocity Ratio (%) Class
<0.53 >35 4
0.53–0.70 35–25 3
0.70–0.84 25–15 2
>0.84 <15 1
Spatial Noise Index (SNI)
The aim of the SNI estimation is to evaluate the spatial consistency of the detected ADA, i.e.,
to quantify how the PSs composing an ADA evolve with a similar trend. We assume that all the TSs of
the same ADA belong to the same deformation phenomena. Thus, we expect different magnitude of
the detected movements, but a spatial correlation between their temporal evolutions. With this aim, for
each ADA the correlation coefficient between all the possible pairs of TSs are calculated (CORR(Xi,Yj),
where i,jrepresents all the possible pair combinations of the APs and Xi and Yj are their respective
TSs. The spatial index (SNI) is a four-class classification of the ADA based on the median value
(Med(CORR)) of all the ADA’s TSs pairs correlation values. It varies from 1 to 4 where 1 corresponds to
a high Med(CORR) and 4 to a very low Med(CORR). The classification thresholds (Table 2) have been
set on the statistical distribution of the results, specifically the values corresponding to the quartiles
that have been chosen.
Table 2. Final classification of the Spatial Noise Index (SNI).
Med(ρ)Cumulative Frequency (%) Class
<0.53 <2 4
0.53–0.7 2–25 (1st quantile) 3
0.7–0.84 25–75 (2nd and 3rd quantiles) 2
>0.84 >75 (4th quantile) 1
ADA Quality Index (QI)
The global QI is derived from the combination of both the TNI and the SNI, and measures the
degree of reliability of each detected ADA. The numerical classes (QI) assigned to each TNI-SNI
Remote Sens. 2017,9, 1002 9 of 19
combination are represented by the matrix shown in Figure 5. The QI ranges from Class 4, which is the
noisiest one, to Class 1, which represents the ADA characterized by very high quality time series (TS).
More in detail: Class 4 represents a not reliable ADA; Class 3 means reliable ADA but TS that cannot
be exploited; Class 2 means reliable ADA, but a further analysis of the TS is recommended; and Class 1
means reliable ADA and TS.
Remote Sens. 2017, 9, 1002 9 of 19
cannot be exploited; Class 2 means reliable ADA, but a further analysis of the TS is recommended;
and Class 1 means reliable ADA and TS.
Figure 5. Quality Index (QI) matrix representing the derivation of the QI from the combination of the
Spatial Noise Index (SNI) and the Temporal Noise Index (TNI) is generated.
3. Canary Island Results
In this section, the results of the above application of the procedure over the Canary Islands and
some related technical aspects are discussed and presented.
The explained procedure has been applied to three islands: La Gomera, Tenerife and Gran
Canaria (Spain). The test-site, covering a total land surface of around 4000 km
2
, allows testing the
regional scale potentialities of the procedure.
3.1. Dataset Description
The three islands are covered by a single Sentinel-1 frame. In particular, three swaths and 18
bursts have been processed. In Table 3, the main characteristics of the used dataset are described. The
used image dataset consists of 64 Sentinel-1 Wide Swath images covering around a two years and a
half period, with the first acquisition time in November 2014 and last acquisition time in March 2017.
In this study only images from Sentinel-1A satellite have been used, thus the minimum temporal
sampling is 12 days, while the maximum temporal sampling, which is defined by the images
availability, is 48 days. Table 4 shows the list of all the acquisition times of the processed images.
As explained in the introduction, the aim of the procedure is to generate and periodically update
deformation activity maps. With this aim, the dataset has been divided in two parts and processed
separately to produce and compare two versions of the ADA map: version V1 and the temporally-
updated version V2. The temporal windows covered by the two processing iterations and the number
of the processed images are resumed in Table 4: for the first iteration, 51 images covering a period of
around two years have been processed; and, for the second iteration, 42 images covering a period of
one and a half years have been processed, the overlapping period between the two iterations is one-
year long. Furthermore, considering the specific radiometric characteristics of the test-site, using
temporal windows of one and a half years, to be processed each six months (i.e., with an overlapping
period of six months) is the ideal way of generating and updating the maps.
The SRTM Digital Elevation Model provided by NASA has been used to process the
interferometric products [41].
Figure 5. Quality Index (QI) matrix representing the derivation of the QI from the combination of the
Spatial Noise Index (SNI) and the Temporal Noise Index (TNI) is generated.
3. Canary Island Results
In this section, the results of the above application of the procedure over the Canary Islands and
some related technical aspects are discussed and presented.
The explained procedure has been applied to three islands: La Gomera, Tenerife and Gran Canaria
(Spain). The test-site, covering a total land surface of around 4000 km
2
, allows testing the regional
scale potentialities of the procedure.
3.1. Dataset Description
The three islands are covered by a single Sentinel-1 frame. In particular, three swaths and 18 bursts
have been processed. In Table 3, the main characteristics of the used dataset are described. The used
image dataset consists of 64 Sentinel-1 Wide Swath images covering around a two years and a half
period, with the first acquisition time in November 2014 and last acquisition time in March 2017. In this
study only images from Sentinel-1A satellite have been used, thus the minimum temporal sampling is
12 days, while the maximum temporal sampling, which is defined by the images availability, is 48 days.
Table 4shows the list of all the acquisition times of the processed images.
As explained in the introduction, the aim of the procedure is to generate and periodically
update deformation activity maps. With this aim, the dataset has been divided in two parts and
processed separately to produce and compare two versions of the ADA map: version V1 and the
temporally-updated version V2. The temporal windows covered by the two processing iterations and
the number of the processed images are resumed in Table 4: for the first iteration, 51 images covering a
period of around two years have been processed; and, for the second iteration, 42 images covering a
period of one and a half years have been processed, the overlapping period between the two iterations
is one-year long. Furthermore, considering the specific radiometric characteristics of the test-site, using
temporal windows of one and a half years, to be processed each six months (i.e., with an overlapping
period of six months) is the ideal way of generating and updating the maps.
The SRTM Digital Elevation Model provided by NASA has been used to process the interferometric
products [41].
Remote Sens. 2017,9, 1002 10 of 19
Table 3. Main characteristics of the processed data.
Satellite Sentinel-1A
Acquisition mode Wide Swath
Period November 2014–March 2017
Minimum revisit period (days) 12
Wavelength (λ) (cm) 5.55
Polarization VV
Full resolution (azimuth/range) (m) 14/4
Multi-look 1 ×5 resolution (azimuth/range) (m) 14/20
Multi-look 2 ×10 resolution (azimuth/range) (m) 28/40
Orbit Descending
Incidence angle of the area of interest 36.47–41.85
Table 4. List of the acquisition dates of the dataset. The intersection between both periods is in bold.
Image Date Image Date Image Date Image Date
1 5 November 2014 18 8 August 2015 35 28 February 2016 52 13 October 2016
2
17 November 2014
19 20 August 2015 36 11 March 2016 53 25 October 2016
3
29 November 2014
20 1 September 2015 37 23 March 2016 54 6 November 2016
4 11 December 2014 21
13 September 2015
38 4 April 2016 55
18 November 2016
5 23 December 2014 22
25 September 2015
39 16 April 2016 56
30 November 2016
6 4 January 2015 23 7 October 2015 40 28 April 2016 57 12 December 2016
7 16 January 2015 24 19 October 2015 41 10 May 2016 58 24 December 2016
8 28 January 2015 25 31 October 2015 42 22 May 2016 59 5 January 2017
9 9 February 2015 26
12 November 2015
43 3 June 2016 60 17 January 2017
10 21 February 2015 27
24 November 2015
44 15 June 2016 61 29 January 2017
11 5 March 2015 28 6 December 2015 45 9 July 2016 62 22 February 2017
12 17 March 2015 29 18 December 2015 46 21 July 2016 63 6 March 2017
13 29 March 2015 30 30 December 2015 47 2 August 2016 64 18 March 2017
14 9 June 2015 31 11 January 2016 48 14 August 2016
15 3 July 2015 32 23 January 2016 49 7 September 2016
16 15 July 2015 33 4 February 2016 50
19 September 2016
17 27 July 2015 34 16 February 2016 51 1 October 2016
3.2. Deformation Activity Maps
To derive the deformation maps, we have generated a network of 398 interferograms in the first
iteration and 481 interferograms in the second iteration. The maximum temporal baseline used is
156 days. This threshold has been derived by statistical analysis of the coherence with respect to the
temporal baseline. The reference points used for the processing are located in the historical centres
of the three capitals of the islands (Figure 6a): San Sebastián de La Gomera; Santa Cruz de Tenerife;
Las Palmas de Gran Canaria.
Due to the geologic and land cover settings, mainly of sparse vegetation and rocky surfaces,
Canary Islands show a good radar response in terms of coherence. This characteristic results in a
deformation map characterized by both a high coverage of points and a low spatio/temporal noise.
Figure 6shows the high density of points of the velocity map: only the few zones covered by forest
show absence of points (northern humid flanks of the islands). The noise level of the map (i.e.,
the sensitivity) was estimated as two times the standard deviation of the velocity of all the measured
points and is equal to 4.7 mm/year for both iterations. This value (see Section 2) also represents the
stability threshold, i.e., the value that separates the moving points from the points with no-detected
movement (stable points).
As explained in Section 2, three filters have been applied to the raw deformation map in order to
reduce the spatio-temporal noise and thus to improve the readability and the reliability of the map
measurements. Figure 6shows the velocity map resulted from the first iteration before and after the
spatial filtering. After the filtering, the number of measured points of the Deformation Activity Map
(DAM) is 1,060,750 for the first iteration and 1,036,328 for the second iteration. The total number of
points identified as non-stable is 2358 in the first iteration and 1859 in the second one, which represents
less than 1% of the total number of measured points.
Remote Sens. 2017,9, 1002 11 of 19
Remote Sens. 2017, 9, 1002 11 of 19
points identified as non-stable is 2358 in the first iteration and 1859 in the second one, which
represents less than 1% of the total number of measured points.
Figure 6. The velocity map of the first iteration (V1): before the raw deformation map (a); and after
the filtering (b). The latter one is the final Deformation Activity Map (DAM).
3.3. Active Deformation Areas (ADA) Map
To extract the ADA from the DAM, the methodology shown in Figure 3 and explained in Section
2 has been applied. Figure 7 shows an example of ADA extraction over the Pico Viejo-Teide area, in
the heart of Tenerife Island: only the active PSs are visualized, the red polygons are examples of
extracted ADA (five or more active contiguous PSs), the green polygons highlight the spatial outliers
PSs (one or two active isolated PSs), which are not included in the final DAM, while the orange
polygons are active PSs that are not extracted as ADA (three or four contiguous PSs) but are included
in the DAM for a local scale analysis. Figure 8 shows an example of two extracted ADA, located
southeast of Tenerife (Figure 8a), which are both subsidence phenomena related to the activities of a
waste dump. Figure 8b presents the DAM of the waste deposit area. There are four areas affected by
movements, two with subsidence (red) and two with uplift (blue), both related to the waste dump
activities.
Figure 8c shows the two ADA affected with subsidence in Figure 8b. Figure 8d shows the time
series of three PSs located, as indicated in Figure 8c, in one of the two ADA. The area is classified
Figure 6.
The velocity map of the first iteration (V1): before the raw deformation map (
a
); and after the
filtering (b). The latter one is the final Deformation Activity Map (DAM).
3.3. Active Deformation Areas (ADA) Map
To extract the ADA from the DAM, the methodology shown in Figure 3and explained in Section 2
has been applied. Figure 7shows an example of ADA extraction over the Pico Viejo-Teide area, in the
heart of Tenerife Island: only the active PSs are visualized, the red polygons are examples of extracted
ADA (five or more active contiguous PSs), the green polygons highlight the spatial outliers PSs (one or
two active isolated PSs), which are not included in the final DAM, while the orange polygons are active
PSs that are not extracted as ADA (three or four contiguous PSs) but are included in the DAM for a
local scale analysis. Figure 8shows an example of two extracted ADA, located southeast of Tenerife
(Figure 8a), which are both subsidence phenomena related to the activities of a waste dump. Figure 8b
presents the DAM of the waste deposit area. There are four areas affected by movements, two with
subsidence (red) and two with uplift (blue), both related to the waste dump activities.
Figure 8c shows the two ADA affected with subsidence in Figure 8b. Figure 8d shows the time
series of three PSs located, as indicated in Figure 8c, in one of the two ADA. The area is classified with
both the SNI and the TNI equal to 1, which means that the ADA is characterized by the highest spatial
and temporal quality. It is interesting to note that the SNI method evaluates the spatial noise in terms
Remote Sens. 2017,9, 1002 12 of 19
of spatial homogeneity of the ADA temporal evolution and is not affected by the spatial variation
of the deformation magnitude. In this case, for example, the ADA presents very different velocities,
following the subsidence spatial distribution, with the higher deformation rate in the centre of area
(Figure 8c,d, PS-2) and the lower ones in the peripheral zones (PS-1 and PS-3). This subsidence area is
active in both iterations, without a significant change in the mean velocity (less than the sensitivity of
the map): 41.4 mm/year in the first iteration and 40.4 mm/year in the second iteration.
For each ADA, the information resumed in Table 5is collected, forming the attribute table of
the corresponding polygonal shapefile. This include the velocity class, which allows enhancing the
visualization of the most critical ADA in terms of magnitude of deformation, and the QI class, which
is a fundamental information for the interpretation of the map. An example of ADA map visualization
of both the QI and the velocity class information is presented in Figure 9. In the first iteration, 72 ADA
has been extracted: 69 are localized on Tenerife Island and the other three on Gran Canaria. In the
second iteration, 120 ADA have been detected: 112 are on Tenerife, seven on Gran Canaria and one
on Gomera. In total, 68% of the ADA detected in the first iteration (V1) fall in the QI classes 1 and 2,
while, in the second iteration (V2), the percentage of 1 and 2 QI drops to 43% (see the Total columns
of Table 6). This reflects the higher general noise level of the second version (V2) of the DAM and
ADA maps.
To compare the two iterations, a simple intersection has been performed. Table 7and Figure 10
summarize the comparison between the two iterations. Table 7summarizes the global numbers of
both iterations. The total number of detected ADA is 192: 68 of them have QI 1 from which 43 are
present in both iterations (Intersect) and 25 only in one of them (No Intersect). Regarding the last
ones, it is worth noting that, even if they are not in both iterations, they are considered reliable ADA.
The reason an ADA is detected in some iterations, but not in others, can be due to different factors like
the loss of coherence or a different behavior of the phenomenon in different periods.
Remote Sens. 2017, 9, 1002 12 of 19
with both the SNI and the TNI equal to 1, which means that the ADA is characterized by the highest
spatial and temporal quality. It is interesting to note that the SNI method evaluates the spatial noise
in terms of spatial homogeneity of the ADA temporal evolution and is not affected by the spatial
variation of the deformation magnitude. In this case, for example, the ADA presents very different
velocities, following the subsidence spatial distribution, with the higher deformation rate in the
centre of area (Figure 8c,d, PS-2) and the lower ones in the peripheral zones (PS-1 and PS-3). This
subsidence area is active in both iterations, without a significant change in the mean velocity (less
than the sensitivity of the map): 41.4 mm/year in the rst iteration and 40.4 mm/year in the second
iteration.
For each ADA, the information resumed in Table 5 is collected, forming the attribute table of the
corresponding polygonal shapefile. This include the velocity class, which allows enhancing the
visualization of the most critical ADA in terms of magnitude of deformation, and the QI class, which
is a fundamental information for the interpretation of the map. An example of ADA map
visualization of both the QI and the velocity class information is presented in Figure 9. In the first
iteration, 72 ADA has been extracted: 69 are localized on Tenerife Island and the other three on Gran
Canaria. In the second iteration, 120 ADA have been detected: 112 are on Tenerife, seven on Gran
Canaria and one on Gomera. In total, 68% of the ADA detected in the first iteration (V1) fall in the QI
classes 1 and 2, while, in the second iteration (V2), the percentage of 1 and 2 QI drops to 43% (see the
Total columns of Table 6). This reflects the higher general noise level of the second version (V2) of
the DAM and ADA maps.
To compare the two iterations, a simple intersection has been performed. Table 7 and Figure 10
summarize the comparison between the two iterations. Table 7 summarizes the global numbers of
both iterations. The total number of detected ADA is 192: 68 of them have QI 1 from which 43 are
present in both iterations (Intersect) and 25 only in one of them (No Intersect). Regarding the last
ones, it is worth noting that, even if they are not in both iterations, they are considered reliable ADA.
The reason an ADA is detected in some iterations, but not in others, can be due to different factors
like the loss of coherence or a different behavior of the phenomenon in different periods.
Figure 7. Example of ADA extraction from the active PSs of the first iteration velocity map (V1). The
area includes the Pico Viejo and Teide craters, the highest elevations on Tenerife Island. Only the
active PSs are visualized. The red polygons are the extracted ADA and the black numbers are the
associated Quality Indexes.
Figure 7.
Example of ADA extraction from the active PSs of the first iteration velocity map (V1).
The area includes the Pico Viejo and Teide craters, the highest elevations on Tenerife Island. Only the
active PSs are visualized. The red polygons are the extracted ADA and the black numbers are the
associated Quality Indexes.
Remote Sens. 2017,9, 1002 13 of 19
Remote Sens. 2017, 9, 1002 13 of 19
Figure 8. (a) The ADA V1 map of Tenerife, the blue square highlights the area that is showed in detail
in (b,c); (b) the DAM (velocity map) in correspondence of the blue frame in (a), which is an industrial
landfill area affected by subsidence (red points) and uplift (blue points); (c) two of the extracted ADA
(red polygons) of the landfill subsidence (the uplift ADA are not represented here); and (d)
deformation time series of points PS-1, 2 and 3.
Conversely, the total number of ADA with QI equal to 4 are 69, where the majority (62) have no
intersection between the two iterations. Looking only at the intersecting ADA, 53 out of 66 (80%) falls
in the first and second QI class. Summarizing, the ADA with worst QI (3 or 4) have a low probability
to be detected in more than one iteration because they are highly affected by noise and thus they are
less reliable. This fact is evidenced in Figure 10, where it can be observed that most of the non-
intersecting ADA have a low QI. This fact can be considered as an indicator of the significance of the
QI information. Among the intersecting ADA, some of them present a change of the QI: the QI values
are mainly lower in the second iteration. This is due to the noise level of the DAM, which is slightly
higher in the second iteration. All but one intersecting ADA are localized on Tenerife. The remaining
one is localized on Gran Canaria.
Table 5. The attributes associated to each ADA.
Field Description Units
Join Count Number of unstable points grouped in the hotspot -
Fi WGS84 Geographic Latitude (average of the grouped PSs) °
Lambda WGS84 Geographic Longitude (average of the grouped PSs) °
E WGS84 UTM zone 32N—East (average of the grouped PSs) m
N WGS84 UTM zone 32N—North (average of the grouped PSs) m
H SRTM Height (average of the grouped PSs) m
Acc. Defo. Accumulated deformation (average of the grouped PSs) mm
Velocity mean Mean velocity of the hotspot (average of the grouped PSs) mm/year
Velo max Maximum velocity of the PSs grouped in the hotspot mm/year
Velo min Minimum velocity of the PSs grouped in the hotspot mm/year
QI Quality index of the ADA -
Class Classification of the hotspots based on the Velo max. -
Figure 8.
(
a
) The ADA V1 map of Tenerife, the blue square highlights the area that is showed in detail
in (
b
,
c
); (
b
) the DAM (velocity map) in correspondence of the blue frame in (
a
), which is an industrial
landfill area affected by subsidence (red points) and uplift (blue points); (
c
) two of the extracted ADA
(red polygons) of the landfill subsidence (the uplift ADA are not represented here); and (
d
) deformation
time series of points PS-1, 2 and 3.
Conversely, the total number of ADA with QI equal to 4 are 69, where the majority (62) have no
intersection between the two iterations. Looking only at the intersecting ADA, 53 out of 66 (80%) falls
in the first and second QI class. Summarizing, the ADA with worst QI (3 or 4) have a low probability to
be detected in more than one iteration because they are highly affected by noise and thus they are less
reliable. This fact is evidenced in Figure 10, where it can be observed that most of the non-intersecting
ADA have a low QI. This fact can be considered as an indicator of the significance of the QI information.
Among the intersecting ADA, some of them present a change of the QI: the QI values are mainly lower
in the second iteration. This is due to the noise level of the DAM, which is slightly higher in the second
iteration. All but one intersecting ADA are localized on Tenerife. The remaining one is localized on
Gran Canaria.
Table 5. The attributes associated to each ADA.
Field Description Units
Join Count Number of unstable points grouped in the hotspot -
Fi WGS84 Geographic Latitude (average of the grouped PSs)
Lambda WGS84 Geographic Longitude (average of the grouped PSs)
E WGS84 UTM zone 32N—East (average of the grouped PSs) m
N WGS84 UTM zone 32N—North (average of the grouped PSs) m
H SRTM Height (average of the grouped PSs) m
Acc. Defo. Accumulated deformation (average of the grouped PSs) mm
Velocity mean Mean velocity of the hotspot (average of the grouped PSs) mm/year
Velo max Maximum velocity of the PSs grouped in the hotspot mm/year
Velo min Minimum velocity of the PSs grouped in the hotspot mm/year
QI Quality index of the ADA -
Class Classification of the hotspots based on the Velo max. -
Remote Sens. 2017,9, 1002 14 of 19
Remote Sens. 2017, 9, 1002 14 of 19
Table 6. Summary of the ADA extracted in the V1 (left) and V2 (right). In the Total column, the
percentages for each QI class refer to the total number of the detected ADA. The Intersect column
refers to the ADA that are detected in both iterations and the percentages are relative to each QI class.
The No Intersect column refers to the ADA that are not detected in the other iteration and the
percentages are relative to each QI class. In V1, 42% of the ADA are also detected in V2; of this, 80%
are classified in the higher QI class (1). In V2, 31% of the ADA are also detected in V1; of this, 51% are
classified in the higher QI class (1).
V1 V2
Total Intersect No Intersect Total Intersect No Intersect
QI n° % n° % n° % QI n° % n° % n° %
1 36 50% 24 33% 12 17% 1 32 27% 19 16% 13 11%
2 13 18% 3 4% 10 14% 2 19 16% 7 6% 12 10%
3 6 8% 1 1% 5 7% 3 17 14% 5 4% 12 10%
4 17 24% 2 3% 15 21% 4 52 43% 5 4% 47 39%
Total 72 100% 30 42% 42 58% Total 120 100% 37 31% 84 70%
Figure 9. ADA map of Tenerife (Iteration 1). Both the QI (colors) and the velocity classes (white
numbers) are represented. This visualization allows a rapid identification of the most critical and
reliable deformations.
Figure 9.
ADA map of Tenerife (Iteration 1). Both the QI (colors) and the velocity classes (white
numbers) are represented. This visualization allows a rapid identification of the most critical and
reliable deformations.
Table 6.
Summary of the ADA extracted in the V1 (left) and V2 (right). In the Total column, the
percentages for each QI class refer to the total number of the detected ADA. The Intersect column refers
to the ADA that are detected in both iterations and the percentages are relative to each QI class. The No
Intersect column refers to the ADA that are not detected in the other iteration and the percentages are
relative to each QI class. In V1, 42% of the ADA are also detected in V2; of this, 80% are classified in
the higher QI class (1). In V2, 31% of the ADA are also detected in V1; of this, 51% are classified in the
higher QI class (1).
V1 V2
Total Intersect No Intersect Total Intersect No Intersect
QI n% n% n% QI n% n% n%
1 36 50% 24
33%
12 17% 1 32 27% 19 16% 13 11%
2 13 18% 3 4% 10 14% 2 19 16% 7 6% 12 10%
3 6 8% 1 1% 5 7% 3 17 14% 5 4% 12 10%
4 17 24% 2 3% 15 21% 4 52 43% 5 4% 47 39%
Total 72
100%
30
42%
42 58% Total
120 100%
37 31% 84 70%
Remote Sens. 2017,9, 1002 15 of 19
Table 7. Summary of the detected ADA in both iterations. The Intersection column refers to the ADA
that are detected in both iterations. The No Intersection column refers to the ADA that are detected in
only one iteration. See also Figure 10.
V1 and V2 ADA Summary
QI Class Tot Intersection No Intersection
1 68 43 25
2 32 10 22
3 23 6 17
4 69 7 62
Total 192 66 126
Remote Sens. 2017, 9, 1002 15 of 19
Table 7. Summary of the detected ADA in both iterations. The Intersection column refers to the ADA
that are detected in both iterations. The No Intersection column refers to the ADA that are detected
in only one iteration. See also Figure 10.
V1 and V2 ADA Summary
QI Class Tot Inte
r
section No Intersection
1 68 43 25
2 32 10 22
3 23 6 17
4 69 7 62
Total 192 66 126
Figure 10. A representation of Table 7. The blue bars (Intersection) represent, for each QI class, the
percentage of the ADA that have been detected in both the iteration. The red bars represent, for each
QI class, the percentage of the ADA that have been detected in only one iteration. The purple line
represents the QI percent of the total the detected ADA. The graphic shows that the majority (63%) of
the ADA with a high Quality Index (1) are detected in both iterations, while the majority of the ADA
with the lower Quality Index (4) are detected in only one iteration. This confirms the significance of
the QI that permits to detect the noisier and not reliable ADA.
4. Discussion
In this section, some key aspects, as well as the strengths and limitations of the presented
methodology are commented.
The presented methodology is aimed at generating and periodically updating geohazard
activity maps over wide areas, using the satellite Sentinel-1 data. The main challenge is to generate
rapidly and semi-automatically a product to be easily exploited in the geohazard management by the
Civil Protections and the Geological Surveys. The regional scale potentialities of the methodology
have been presented through its application over the three islands of Tenerife, La Gomera and Gran
Canaria (in Spain, Figure 6). The main output of the methodology is the ADA map, which localizes
only the most important detected active areas, summarizing and simplifying the significant
information of the areas.
The methodology can be applied by using as input every type of satellite SAR images.
Nevertheless, the best performances of the methodology are obtained using Sentinel-1 satellite data. On
the one hand, S-1 acquires data with a 250 km swath at 4 m by 14 m spatial resolution (full resolution),
allowing a wide area (regional scale) monitoring. On the other hand, the short revisit time of the S-1,
varying 6–12 days depending on the area, reduces the temporal decorrelation of the interferometric
pairs and, together with the regular worldwide acquisition, it highly improves the monitoring
potentialities. In other words, it allows making long-term monitoring planning throughout. Moreover,
Figure 10.
A representation of Table 7. The blue bars (Intersection) represent, for each QI class, the
percentage of the ADA that have been detected in both the iteration. The red bars represent, for each
QI class, the percentage of the ADA that have been detected in only one iteration. The purple line
represents the QI percent of the total the detected ADA. The graphic shows that the majority (63%) of
the ADA with a high Quality Index (1) are detected in both iterations, while the majority of the ADA
with the lower Quality Index (4) are detected in only one iteration. This confirms the significance of the
QI that permits to detect the noisier and not reliable ADA.
4. Discussion
In this section, some key aspects, as well as the strengths and limitations of the presented
methodology are commented.
The presented methodology is aimed at generating and periodically updating geohazard activity
maps over wide areas, using the satellite Sentinel-1 data. The main challenge is to generate rapidly
and semi-automatically a product to be easily exploited in the geohazard management by the Civil
Protections and the Geological Surveys. The regional scale potentialities of the methodology have been
presented through its application over the three islands of Tenerife, La Gomera and Gran Canaria (in
Spain, Figure 6). The main output of the methodology is the ADA map, which localizes only the most
important detected active areas, summarizing and simplifying the significant information of the areas.
The methodology can be applied by using as input every type of satellite SAR images.
Nevertheless, the best performances of the methodology are obtained using Sentinel-1 satellite data.
On the one hand, S-1 acquires data with a 250 km swath at 4 m by 14 m spatial resolution (full
resolution), allowing a wide area (regional scale) monitoring. On the other hand, the short revisit
time of the S-1, varying 6–12 days depending on the area, reduces the temporal decorrelation of the
Remote Sens. 2017,9, 1002 16 of 19
interferometric pairs and, together with the regular worldwide acquisition, it highly improves the
monitoring potentialities. In other words, it allows making long-term monitoring planning throughout.
Moreover, all the Sentinel-1 satellite data are free of charge, improving the long-term sustainability of
the methodology from the point of view of the costs. For a qualitative estimation of the costs in terms
of people and time needed by the methodology application, refer to [33].
The methodology results are influenced by intrinsic limitations of the SAR satellite data.
Apart from the theoretical maximum and minimum measurable deformations, which depend both on
the sensor characteristics and on the revisit time as described in [
11
], there are two other aspects that
spatially influence the possibility of detecting movements: the acquisition geometry and the coherence.
The last one, for equal acquisition characteristics (sensor and revisit time) and meteorological
conditions, is mainly determined by the land cover. In forested areas, for example, it is very difficult to
obtain reliable PSs, and thus ADA, with the PSI analysis. The geometrical limitation is determined by
the geometry of SAR acquisition with respect to both: (a) the main deformation direction; and (b) the
terrain topography. The InSAR techniques are capable of measuring only the LOS direction component
of the real movement: it measures a percentage of the real movement that is zero if the deformation
direction is perpendicular to the satellite LOS. The terrain topography, with respect to the radar
wave-front angle, causes a slope-dependent ground spatial resolution variation with a consequent
variation in the capability of detecting movement. Two extreme examples are the so-called shadow
zones, where the slope is not seen by the radar beam (no radar visibility), and the foreshortening
zones, where the slope is almost parallel to the wave-front. All those aspects, among others, belong to
a fundamental knowledge background that is necessary for a correct interpretation of a PSI derived
deformation map. As an example, an important aspect to underline is that the presence of “stable”
(green) PSs, does not always means ground stability. In this context, the ADA map, reporting only
the active detected areas (no ADA only means no information), is a strong product that is more easily
read and interpreted by not-expert final users. On the contrary, the interpretation of the Deformation
Activity Map (DAM) is not straightforward and the real scenario can be misinterpreted by non-expert
users. Moreover, the ADA map overcomes the problems of the noisy information and of the huge
amount of measures (millions of points) to be managed, by localizing only the detected areas and
summarizing the most important attributes of each area. Those aspects are fundamental for a regional
scale overview. Nevertheless, the DAM is an important tool that can be used for a more detailed
(local scale) spatio-temporal analysis of each ADA. To partially overcome the geometrical limitations,
a parallel processing of both the ascending and descending datasets is highly recommended, if the
images are available. The double geometry processing allows not only to improve the coverage, but
also to have additional and independent information that is very important for the interpretation of
the deformation phenomena.
An important aspect of the methodology is that it is a reproducible work flow, adaptable to
each case study or final user’s needs. Depending on the site characteristics and on the main target of
the monitoring, specifically the spatial and temporal magnitudes of the expected deformations, the
methodology can be applied by changing for example the DInSAR processing technique, the minimum
number of points to extract the ADA or the stability threshold. In addition, the temporal window
between successive iterations for a regular updating of the ADA map is an aspect that has to be tuned
on the base of the target deformation velocities (e.g., longer periods for slower deformations) and
monitoring aims. For what concerns the temporal window to be processed, we have evaluated that a
minimum of one year and a half is necessary in order to get acceptable results in terms of noise level.
It is worth underlining that the ADA map can be used to periodically update geohazard
inventories and to drive or support the risk management activities. A step forward is the use of
ADA map to rapidly evaluate the impact of the detected deformations on buildings and infrastructures,
as explained in [33].
Remote Sens. 2017,9, 1002 17 of 19
5. Conclusions
This paper aims at explaining the implemented methodology to generate and periodically
update Geohazard Activity Maps, over wide areas, using the DInSAR technique and S-1 data.
The methodology has been developed in the framework of the ongoing European ECHO (European
Civil Protection and Humanitarian Aid Operations) project Safety, “Sentinel for geohazards regional
monitoring and forecasting”. The aim was to find a way to simplify and summarize the SAR satellite
derivable information in order to be exploited by any not radar-expert final user, specifically by the
Civil Protection Authorities in the risk management activities. The outputs of the methodology are
the Deformation Activity Map (DAM), in terms of velocity map and deformation time series, and the
Active Deformation Areas (ADA) map. The last one is the main product that can be exploited to update
the existing geohazard inventories. All the main steps of the methodology have been explained, starting
from the PSI processing, the raw deformation map filtering to generate the DAM and the subsequent
ADA extraction. Then, a methodology to evaluate the reliability of each ADA has been implemented
and explained: a Quality Index is assigned to each ADA based on the temporal and spatial noise of
its time series. The application and the results of the methodology over the islands of Gran Canaria,
La Gomera and Tenerife (Spain) have been presented and discussed. Two temporally-displaced
iterations have been processed to test the updating potentialities of the ADA map. A total of 72 ADA,
in the first iteration, and 120 ADA, in the second iteration, have been detected over the three islands.
The majority of the ADA that have been detected in both iterations, are classified as highly reliable
according to the QI, demonstrating the significance of the QI information. The results exhibit the
regional scale monitoring potentialities of the methodology.
Acknowledgments:
This work has been funded by the European Commission, Directorate-General Humanitarian
Aid and Civil Protection (ECHO), through the project Safety (Ref. ECHO/SUB/2015/718679/Prev02).
Author Contributions:
Anna Barra (CTTC) wrote the paper and was involved in the data processing and
analysis; Lorenzo Solari, Silvia Bianchini and Sandro Moretti, from UniFi, Marta Béjar-Pizarro, Gerardo Herrera,
Roberto Sarro and Rosa María Mateos, from IGME, Elena González-Alonso, Sergio Ligüerzana and Carmen López,
from IGN, provided inputs in the analysis of the detected ADA and in the Canary Island site characterization;
Michele Crosetto and Oriol Monserrat (CTTC) contributed in the SAR data processing and analysis of the results.
Conflicts of Interest: The authors declare no conflict of interest.
References
1.
Gabriel, A.K.; Goldstein, R.M.; Zebker, H.A. Mapping small elevation changes over large areas: Differential
radar interferometry. J. Geophys. Res. 1989,94, 9183–9191. [CrossRef]
2.
Massonnet, D.; Feigl, K.L. Radar interferometry and its application to changes in the Earth’s surface.
Rev. Geophys. 1998,36, 441–500. [CrossRef]
3.
Rosen, P.A.; Hensley, S.; Joughin, I.R.; Li, F.K.; Madsen, S.N.; Rodriguez, E.; Goldstein, R.M. Synthetic
aperture radar interferometry. Proc. IEEE 2000,88, 333–382. [CrossRef]
4.
Barra, A.; Monserrat, O.; Mazzanti, P.; Esposito, C.; Crosetto, M.; Scarascia Mugnozza, G. First insights on
the potential of Sentinel-1 for landslides detection. Geomat. Nat. Hazards Risk
2016
,7, 1874–1883. [CrossRef]
5.
Biescas, E.; Crosetto, M.; Agudo, M.; Monserrat, O.; Crippa, B. Two radar interferometric approaches to
monitor slow and fast land deformation. J. Surv. Eng. 2007,133, 66–71. [CrossRef]
6.
Ferretti, A.; Prati, C.; Rocca, F. Permanent scatterers in SAR interferometry. IEEE TGRS
2001
,39, 8–20.
[CrossRef]
7.
Berardino, P.; Fornaro, G.; Lanari, R.; Sansosti, E. A new algorithm for surface deformation monitoring based
on small baseline differential SAR interferograms. IEEE TGRS 2002,40, 2375–2383. [CrossRef]
8.
Mora, O.; Mallorqui, J.J.; Broquetas, A. Linear and nonlinear terrain deformation maps from a reduced set of
interferometric SAR images. IEEE TGRS 2003,41, 2243–2253. [CrossRef]
9.
López-Quiroz, P.; Doin, M.P.; Tupin, F.; Briole, P.; Nicolas, J.M. Time series analysis of Mexico City
subsidenceconstrained by radar interferometry. J. Appl. Geophys. 2009,69, 1–15. [CrossRef]
Remote Sens. 2017,9, 1002 18 of 19
10.
Hooper, A. A multi-temporal InSAR method incorporating both persistent scatterer and small baseline
approaches. Geophys. Res. Lett. 2008,35. [CrossRef]
11.
Crosetto, M.; Monserrat, O.; Cuevas-González, M.; Devanthéry, N.; Crippa, B. Persistent scatterer
interferometry: A review. ISPRS J. Photogramm. Remote Sens. 2016,115, 78–89. [CrossRef]
12.
De Luca, C.; Cuccu, R.; Elefante, S.; Zinno, I.; Manunta, M.; Casola, V.; Casu, F. An on-demand web tool for
the unsupervised retrieval of earth’s surface deformation from SAR data: The P-SBAS service within the
ESA G-POD environment. Remote Sens. 2015,7, 15630–15650. [CrossRef]
13.
Cignetti, M.; Manconi, A.; Manunta, M.; Giordan, D.; De Luca, C.; Allasia, P.; Ardizzone, F. Taking advantage
of the esa G-pod service to study ground deformation processes in high mountain areas: A valle d’aosta case
study, Northern Italy. Remote Sens. 2016,8, 852. [CrossRef]
14.
Pepe, A.; Sansosti, E.; Berardino, P.; Lanari, R. On the generation of ERS/ENVISAT DInSAR time-series via
the SBAS technique. IEEE Geosci. Remote Sens. Lett. 2005,2, 265–269. [CrossRef]
15.
Crosetto, M.; Monserrat, O.; Iglesias, R.; Crippa, B. Persistent Scatterer Interferometry: Potential, limits and
initial C- and X-band comparison. Photogramm. Eng. Remote Sens. 2010,76, 1061–1069. [CrossRef]
16.
Monserrat, O.; Crosetto, M.; Cuevas, M.; Crippa, B. The thermal expansion component of Persistent Scatterer
Interferometry observations. IEEE Geosci. Remote Sens. Lett. 2011,8, 864–868. [CrossRef]
17.
Fornaro, G.; Reale, D.; Verde, S. Bridge thermal dilation monitoring with millimeter sensitivity via
multidimensional SAR imaging. IEEE Geosci. Remote Sens. Lett. 2013,10, 677–681. [CrossRef]
18.
Bovenga, F.; Wasowski, J.; Nitti, D.O.; Nutricato, R.; Chiaradia, M.T. Using COSMO/SkyMed X-band and
ENVISAT C-band SAR interferometry for landslides analysis. Remote Sens. Environ.
2012
,119, 272–285.
[CrossRef]
19.
Bovenga, F.; Nitti, D.O.; Fornaro, G.; Radicioni, F.; Stoppini, A.; Brigante, R. Using C/X-band SAR
interferometry and GNSS measurements for the Assisi landslide analysis. Int. J. Remote Sens.
2013
,34,
4083–4104. [CrossRef]
20.
Motagh, M.; Wetzel, H.U.; Roessner, S.; Kaufmann, H. A TerraSAR-X InSAR study of landslides in southern
Kyrgyzstan, Central Asia. Remote Sens. Lett. 2013,4, 657–666. [CrossRef]
21.
Kiseleva, E.; Mikhailov, V.; Smolyaninova, E.; Dmitriev, P.; Golubev, V.; Timoshkina, E.; Hanssen, R. PS-InSAR
monitoring of landslide activity in the Black Sea coast of the Caucasus. Procedia Technol.
2014
,16, 404–413.
[CrossRef]
22.
Calò, F.; Ardizzone, F.; Castaldo, R.; Lollino, P.; Tizzani, P.; Guzzetti, F.; Manunta, M. Enhanced landslide
investigations through advanced DInSAR techniques: The Ivancich case study, Assisi, Italy. Remote Sens.
Environ. 2014,142, 69–82. [CrossRef]
23.
Del Ventisette, C.; Righini, G.; Moretti, S.; Casagli, N. Multitemporal landslides inventory map updating
using spaceborne SAR analysis. Int. J. Appl. Earth Obs. Geoinf. 2014,30, 238–246. [CrossRef]
24.
Costantini, M.; Ferretti, A.; Minati, F.; Falco, S.; Trillo, F.; Colombo, D.; Rucci, A. Analysis of
surface deformations over the whole Italian territory by interferometric processing of ERS, Envisat and
COSMO-SkyMed radar data. Remote Sens. Environ. 2017, in press. [CrossRef]
25.
Wasowski, J.; Bovenga, F. Investigating landslides and unstable slopes with satellite Multi Temporal
Interferometry: Current issues and future perspectives. Eng. Geol. 2014,174, 103–138. [CrossRef]
26.
Wasowski, J.; Bovenga, F. Remote sensing of landslide motion with emphasis on satellite multitemporal
interferometry applications: An overview. In Landslide Hazards, Risks and Disasters; Academic Press: Boston,
MA, USA, 2015; pp. 345–403.
27.
Rucci, A.; Ferretti, A.; Monti Guarnieri, A.; Rocca, F. Sentinel 1 SAR interferometry applications: The outlook
for sub millimeter measurements. Remote Sens. Environ. 2012,120, 156–163. [CrossRef]
28.
Huang, Q.; Crosetto, M.; Monserrat, O.; Crippa, B. Displacement monitoring and modelling of a high-speed
railway bridge using C-band Sentinel-1 data. ISPRS J. Photogramm. Remote Sens.
2017
,128, 204–211.
[CrossRef]
29.
Tang, P.; Chen, F.; Guo, H.; Tian, B.; Wang, X.; Ishwaran, N. Large-area landslides monitoring using advanced
multi-temporal InSAR technique over the giant panda habitat, Sichuan, China. Remote Sens.
2015
,7,
8925–8949. [CrossRef]
30.
Bouali, E.H.; Oommen, T.; Escobar-Wolf, R. Interferometric stacking toward geohazard identification and
geotechnical asset monitoring. J. Infrastruct. Syst. 2016,22, 05016001. [CrossRef]
Remote Sens. 2017,9, 1002 19 of 19
31.
Barboux, C.; Delaloye, R.; Lambiel, C. Inventorying slope movements in an Alpine environment using
DInSAR. Earth Surf. Process. Landf. 2014,39, 2087–2099. [CrossRef]
32.
Devanthéry, N.; Crosetto, M.; Monserrat, O.; Cuevas-González, M.; Crippa, B. An approach to Persistent
Scatterer Interferometry. Remote Sens. 2014,6, 6662–6679. [CrossRef]
33.
Solari, L.; Barra, A.; Herrera, G.; Bianchini, S.; Monserrat, O.; Bejar, M.; Crosetto, M.; Sarro, R.; Salviati, P.;
Moretti, S. Vulnerable Elements Activity Maps based on Sentinel-1 InSAR. Nat. Hazard Earth Sci.
2017–2018
.
submit.
34.
Yague-Martinez, N.; Prats-Iraola, P.; Rodriguez Gonzalez, F.; Brcic, R.; Shau, R.; Geudtner, D.; Eineder, M.;
Bamler, R. Interferometric Processing of Sentinel-1 TOPS Data. IEEE Trans. Geosci. Remote Sens.
2016
,54,
2220–2234. [CrossRef]
35.
Prats-Iraola, P.; Scheiber, R.; Marotti, L.; Wollstadt, S.; Reigber, A. TOPS interferometry with TerraSAR-X.
IEEE Trans. Geosci. Remote Sens. 2012,50, 3179–3188. [CrossRef]
36. Hanssen, R. Radar Interferometry; Kluwer Academic Publishers: Dordrecht, The Netherlands, 2001.
37.
Meisina, C.; Zucca, F.; Notti, D.; Colombo, A.; Cucchi, A.; Savio, G.; Giannico, C.; Bianchi, M. Geological
Interpretation of PSInSAR Data at Regional Scale. Sensors 2008,8, 7469–7492. [CrossRef] [PubMed]
38.
Bianchini, S.; Cigna, F.; Righini, G.; Proietti, C.; Casagli, N. Landslide HotSpot Mapping by means of
Persistent Scatterer Interferometry. Environ. Earth Sci. 2012,67, 1155–1172. [CrossRef]
39.
Herrera, G.; Gutiérrez, F.; García-Davalillo, J.C.; Guerrero, J.; Notti, D.; Galve, J.P.; Fernández-Merodo, J.A.;
Cooksley, G. Multi-sensor advanced DInSAR monitoring of very slow landslides: The Tena Valley case study
(Central Spanish Pyrenees). Remote Sens. Environ. 2013,128, 31–43. [CrossRef]
40.
Notti, D.; Herrera, G.; Bianchini, S.; Meisina,C.; García-Davalillo, J.C.; Zucca, F. A methodology for improving
landslide PSI data analysis. Int. J. Remote Sens. 2014,35, 2186–2214.
41.
Farr, T.G.; Rosen, P.A.; Caro, E.; Crippen, R.; Duren, R.; Hensley, S.; Seal, D. The shuttle radar topography
mission. Rev. Geophys. 2007,45, 1–33. [CrossRef]
©
2017 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).
... This is due to the following reasons: (i) A-DInSAR processing is very fast (less than 24 h); (ii) it does not require an expert user; (iii) it allows obtaining results over large areas in the order of tens of thousands of km 2 . Likewise, the active deformation areas (ADA) procedure [17][18][19] was designed to facilitate the analysis and interpretation of LOS mean deformation velocity maps on a regional scale. The ADA procedure considerably reduces the analysis and interpretation of the results, focusing the study on specific areas with active ground motion. ...
... Additionally, in this work, the LOS mean deformation velocity maps are presented in mm year −1 , as well as a time series in mm. To represent these velocity maps, the stability range (or the threshold for discriminating stable and unstable points) was estimated as two times the standard deviation of the velocity of all the measured points [17]. Therefore, the stability range was set between 6 and −6 mm year −1 for the ascending orbit processing and between 7 and −7 mm year −1 for the descending orbit. ...
... We have applied the ADA procedure (software ADAFinder v 1.1.31 ® ), designed by CTTC (Castelldefels, Catalunya, Spain) and described by [17][18][19]. This methodology performs the identification of ADAs, allowing us to focus the analysis and interpretation in those areas with active ground motion. ...
Article
Full-text available
The Lisbon metropolitan area (LMA, central-west of Portugal) has been severely affected by different geohazards (flooding episodes, landslides, subsidence, and earthquakes) that have generated considerable damage to properties and infrastructures, in the order of millions of euros per year. This study is focused on the analysis of subsidence, as related to urban and industrial activity. Utilizing the A-DInSAR dataset and applying active deformation areas (ADA) processing at the regional scale has allowed us to perform a detailed analysis of subsidence phenomena in the LMA. The dataset consisted of 48 ascending and 61 descending SAR IW-SLC images acquired by the Sentinel-1 A satellite between January 2018 and April 2020. The line-of-sight (LOS), mean deformation velocity (VLOS) maps (mm year−1), and deformation time series (mm) were obtained via the Geohazard Exploitation Platform service of the European Space Agency. The maximum VLOS detected, with ascending and descending datasets, were −38.0 and −32.2 mm year−1, respectively. ADA processing over the LMA allowed for 592 ascending and 560 descending ADAs to be extracted and delimited. From the VLOS measured in both trajectories, a vertical velocity with a maximum value of −32.4 mm year−1 was estimated. The analyzed subsidence was associated to four ascending and three descending ADAs and characterized by maximum VLOS of −25.5 and −25.2 mm year−1. The maximum vertical velocity associated with urban subsidence was −32.4 mm year−1. This subsidence is mainly linked to the compaction of the alluvial and anthropic deposits in the areas where urban and industrial sectors are located. The results of this work have allowed to: (1) detect and assess, from a quantitative point of view, the subsidence phenomena in populated and industrial areas of LMA; (2) establish the relationships between the subsidence phenomena and geological and hydrological characteristics.
... InSAR based techniques allow processing areas from regional/national scale up to very detailed scale such as single buildings, providing a high number of displacement measurements at low cost [1,2,3]. However, the outputs provided by such techniques are usually not easy to understand, requiring an expert to interpret those results, which might turn out to be a time-consuming task for users who are not familiar with radar data [4]. ...
... The presented work is as an example of multi scale (medium to large) application of InSAR for geohazard applications exploiting the ADA (Active Displacement Areas) tools [4] developed with the aim of facilitating the management, use and interpretation of InSAR-based results. The velocity of deformation map and the displacement time series have been estimated over the test area of Granada County (Spain) by processing Sentinel-1 (A and B) Synthetic Aperture Radar (SAR) images. ...
... The velocity of deformation map and the displacement time series have been estimated over the test area of Granada County (Spain) by processing Sentinel-1 (A and B) Synthetic Aperture Radar (SAR) images. From these initial InSAR ouputs a semiautomatic extraction of the most significant Active Displacement Areas (ADAs) is carried out using the ADAFinder tool [4]. The application of the ADA tool to the Riskcoast project test site encompassing the coast of Granada (Spain) is shown. ...
Conference Paper
Full-text available
Geohazard prone areas require continuous monitoring to detect risks, understand the phenomena occurring in those regions and prevent disasters. Satellite interferometry (InSAR) has come to be a trustworthy technique for ground movement detection and monitoring in the last few years. InSAR based techniques allow to process large areas providing high number of displacement measurements at low cost. However, the results provided by such techniques are usually not easy to interpret by non-experienced users hampering its use for decision makers. This work presents a set of tools developed in the framework of different projects (Momit, Safety, U-Geohaz, Riskcoast) and an example of their use in the Granada Coastal area (Spain) is shown. The ADA (Active Displacement Areas) tool have been developed with the aim of easing the management, use and interpretation of InSAR based results. It provides a semi-automatic extraction of the most significant ADAs through the application ADAFinder tool. This tool aims to support the exploitation of the European Ground Motion Service (EU-GMS), which will provide consistent, regular and reliable information regarding natural and anthropogenic ground motion phenomena all over Europe.
... The ADA map is the main dataset to focus deeper analysis on the most significant moving areas, which have a potential impact in terms of risk. Active Deformation Area map has been generated using the ADA Finder tool, which makes a semi-automatic extraction from the A-DInSAR results of the most significant active areas, based on e methodology described in Barra et al., 2017. Each ADA polygon is a grouping of some adjacent points with a displacement velocity higher than a threshold. ...
... Each ADA polygon is a grouping of some adjacent points with a displacement velocity higher than a threshold. (For more information see Barra et al., 2017 andBarbier et al., 2021). ...
Article
Full-text available
Monitoring ground deformation at national and regional level with millimetre-scale precision, nowadays, is possible by using Advanced Differential Interferometric SAR (A-DInSAR) techniques. This study concerns the results of the European Ground Motion Service (EGMS), part of the Copernicus Land Monitoring Service, which detects and measures land displacement at European scale. This Service provides reliable and consistent information regarding natural ground motion phenomena such as landslides and subsidence. The ground motion is derived from Synthetic Aperture Radar (SAR) time-series analysis of Sentinel-1A/B data. These data, which provide full coverage of Europe from two different observation geometries (ascending and descending) every six days, are processed at full resolution. The paper is focused on the exploitation of the basic product of EGMS for both regional and local purposes. Analysing the slope and aspect of the deformation field is the novelty of this investigation. In particular, the focus is put on the generation of wide-area differential deformation maps. Such maps indicate the gradient of the deformation field. The obtained information is not only beneficial for monitoring anthropogenic phenomena but also vital for urban management and planning. Most of the significant damages to manmade structures and infrastructures are associated with high deformation gradient values. Thus, monitoring the temporal and spatial variations of deformation gradient is essential for dynamic analysis, early-warning, and risk assessment in urban areas. Although EGMS productions are prepared for monitoring at regional level, their resolutions are high enough to investigate at local level. Therefore, this paper considers the local deformations that affect single structures or infrastructures. Local differences in such deformation can indicate damages in the corresponding structures and infrastructures. We illustrate these types of analysis to generate differential deformation maps using datasets available at CTTC.
... ADAfinder is the implementation of the methodology explained in [Barra et al., 2017]. It is used to identify the main areas where a displacement has been measured by the Persistent Scatterers Interferometry (PSI) processing, relying on the information contained in the input deformation map and assessing the quality of the time series information (i.e., spatial-temporal noise) of each ADA. ...
... As inputs, ADAfinder takes either an ESRI shapefile or a comma-separated values (CSV) text file containing the deformation map-which amounts, for each PS, for the coordinates of the point, its average velocity and the related deformation time series measuring the movement it underwent. Additionally, a series of key parameters controlling how the detection process takes place must be provided (see again [Barra et al., 2017] for details on such parameters). ...
... Using the same threshold to extract the deformation from different regions leads to inaccurate results, and the use of different thresholds by different researchers results in landslide identification results that are considerably influenced by the subjectivity. Invariant thresholds are no longer used to extract deformation regions and identify landslides; instead, adaptive thresholds based on statistical characteristics of the phase are used [40,41]. Given the need for landslide identification in large areas, some automatic procedures have been proposed according to statistical indicators to automatically identify and classify landslides [42][43][44]. ...
Article
As a type of earth observation technology, interferometric synthetic aperture radar (InSAR) is increasingly widely used in the field of geological disaster detection. However, the application of InSAR in low-coherence areas, such as alpine canyon areas and vegetation coverage areas, is subject to considerable limitations. How to accurately identify landslides from InSAR measurement data in these areas remains the subject of several challenges and shortcomings. Based on statistical analysis and spatial cluster analysis, in this paper, we propose an automatic landslide identification and gradation method suitable for low-coherence areas. The proposed method combines the small baseline subset InSAR (SBAS-InSAR) method and the interferogram stacking (stacking-InSAR) method to obtain a deformation map in the study area, using statistical analysis and spatial cluster analysis to extract deformation regions and landslide polygons to propose a landslide screening model (LSM) based on multivariate features to screen landslides and reduce the interference of noise in landslide identification, in addition to proposing a landslide gradation model (LGM) based on signum function to grade the identified landslides and provide support to distinguish landslides with different deformation degrees. The method was applied to landslide identification in the upper section of the Jinsha River basin, and 47 potential landslides were identified, including 15 high-risk landslides and 13 landslides endangering villages. The experimental results show that the proposed method can identify landslides accurately and hierarchically in low-coherence areas, providing support for geological hazard investigation agencies and local departments.
... Using the same threshold to extract the deformation from different regions leads to inaccurate results, and the use of different thresholds by different researchers results in landslide identification results that are considerably influenced by the subjectivity. Invariant thresholds are no longer used to extract deformation regions and identify landslides; instead, adaptive thresholds based on statistical characteristics of the phase are used [40,41]. Given the need for landslide identification in large areas, some automatic procedures have been proposed according to statistical indicators to automatically identify and classify landslides [42][43][44]. ...
Article
Full-text available
As a type of earth observation technology, interferometric synthetic aperture radar (InSAR) is increasingly widely used in the field of geological disaster detection. However, the application of InSAR in low-coherence areas, such as alpine canyon areas and vegetation coverage areas, is subject to considerable limitations. How to accurately identify landslides from InSAR measurement data in these areas remains the subject of several challenges and shortcomings. Based on statistical analysis and spatial cluster analysis, in this paper, we propose an automatic landslide identification and gradation method suitable for low-coherence areas. The proposed method combines the small baseline subset InSAR (SBAS-InSAR) method and the interferogram stacking (stacking-InSAR) method to obtain a deformation map in the study area, using statistical analysis and spatial cluster analysis to extract deformation regions and landslide polygons to propose a landslide screening model (LSM) based on multivariate features to screen landslides and reduce the interference of noise in landslide identification, in addition to proposing a landslide gradation model (LGM) based on signum function to grade the identified landslides and provide support to distinguish landslides with different deformation degrees. The method was applied to landslide identification in the upper section of the Jinsha River basin, and 47 potential landslides were identified, including 15 high-risk landslides and 13 landslides endangering villages. The experimental results show that the proposed method can identify landslides accurately and hierarchically in low-coherence areas, providing support for geological hazard investigation agencies and local departments.
... Using the same threshold to extract the deformation from different regions leads to inaccurate results, and the use of different thresholds by different researchers results in landslide identification results that are considerably influenced by the subjectivity. Invariant thresholds are no longer used to extract deformation regions and identify landslides; instead, adaptive thresholds based on statistical characteristics of the phase are used [40,41]. Given the need for landslide identification in large areas, some automatic procedures have been proposed according to statistical indicators to automatically identify and classify landslides [42][43][44]. ...
Article
Full-text available
As a type of earth observation technology, interferometric synthetic aperture radar (InSAR) is increasingly widely used in the field of geological disaster detection. However, the application of InSAR in low-coherence areas, such as alpine canyon areas and vegetation coverage areas, is subject to considerable limitations. How to accurately identify landslides from InSAR measurement data in these areas remains the subject of several challenges and shortcomings. Based on statistical analysis and spatial cluster analysis, in this paper, we propose an automatic landslide identification and gradation method suitable for low-coherence areas. The proposed method combines the small baseline subset InSAR (SBAS-InSAR) method and the interferogram stacking (stacking-InSAR) method to obtain a deformation map in the study area, using statistical analysis and spatial cluster analysis to extract deformation regions and landslide polygons to propose a landslide screening model (LSM) based on multivariate features to screen landslides and reduce the interference of noise in landslide identification, in addition to proposing a landslide gradation model (LGM) based on signum function to grade the identified landslides and provide support to distinguish landslides with different deformation degrees. The method was applied to landslide identification in the upper section of the Jinsha River basin, and 47 potential landslides were identified, including 15 high-risk landslides and 13 landslides endangering villages. The experimental results show that the proposed method can identify landslides accurately and hierarchically in low-coherence areas, providing support for geological hazard investigation agencies and local departments.
... At present, active landslides are identified by the InSAR technique generally in terms of the following characteristics: (1) surface deformation (e.g., Xie et al. 2016;Dong et al. 2018;Wang et al. 2019;Rehman et al. 2020), (2) surface deformation and topography (e.g., Zhao et al. 2012;Liu et al. 2018;Guo et al. 2021), (3) surface deformation, geomorphology, and topography (e.g., Zhao et al. 2018a, b;Zhang et al. 2020;Dun et al. 2021), and (4) surface deformation and geology (e.g., Solari et al. 2020). Furthermore, prior knowledge of landslides is generally required to assist in identifying active deformation regions (regions characterized by surface mass movement) (Barra et al. 2017) in InSAR products, and few studies conducted landslide identification with no use of their prior knowledge (Bekaert et al. 2020). Bekaert et al. (2020) adopted the time-series InSAR technique and pixel clustering to identify slow-moving landslides in the Himalayas centered on the Trishuli River catchment, Western Nepal, with no prior knowledge of their location. ...
Article
Full-text available
More than 70% of catastrophic landslides were previously unknown and brought tremendous losses to human life and property in urban regions; therefore, there is an urgent need for early identification of active landslides to eliminate landslide risk at the early stage. However, early identification of landslides has always been a worldwide challenge due to high concealment, steep topography, inaccessible location, and sudden onset. This work suggests a new set of comprehensive criteria for the early identification of landslides by integrating surface deformation, geological, topographic, geomorphological, and disaster-failure features. This set of criteria is universally applicable with no use of the prior knowledge of landslide locations (blind identification) and is successfully validated by a field survey. This work selects the Xuecheng region, a hard-hit area of landslides, as the study area and employs multisource data (seismic, geological, topographic, meteorological, SAR, and optical remote sensing data) and time-series InSAR technology to identify active landslides and reveal their deformation rules. Some new viewpoints are suggested. (1) The new comprehensive criteria synthesize the surface deformation, disaster-controlling, and disaster-inducing characteristics and achieve relatively high accuracy by field validation. (2) Forty-seven active landslides are identified in Xuecheng with no use of the prior knowledge of landslides. The soft rocks or soft-hard interbeddings, tectonic movement, fluvial undercutting and eroding, precipitation, earthquakes, and human engineering activity control or induce the development of these active landslides. (3) Two giant landslides that significantly threaten human lives and properties and exhibit different movement modes are selected to highlight the deformation rules of active landslides under the coupled action of poor lithologic condition, tectonic movement, river erosion, precipitation, and human engineering activity. The suggested new criteria can be applied to other landslide hard-hit urban regions and contribute to the timely and effective prevention and control of catastrophic landslides, reduction of enormous disaster losses, and rational management of the environment.
... Landslides are common natural geological disasters that cause a lot of property loss and personal injury every year [1,2]. The landslide early warning system (LEWS), including real-time monitoring methods such as remote sensing (RS), the global position system (GPS), and the geographic information system (GIS), is an effective approach to identify the early displacement and development trend of landslides [3][4][5]. The landslide surface displacement prediction based on real-time monitoring data is an important part for LEWS [6,7]. ...
Article
Full-text available
Landslide displacement prediction is an essential base of landslide hazard prevention, which often needs to establish an accurate prediction model. To achieve accuracy prediction of landslide displacement, a displacement prediction model based on a salp-swarm-algorithm-optimized temporal convolutional network (SSA-TCN) is proposed. The TCN model, consisting of a causal dilation convolution layer residual block, can flexibly increase the receptive fields and capture the global information in a deeper layer. SSA can solve the hyperparameter problem well for TCN model. The Muyubao landslide displacement collected from a professional GPS monitoring system implemented in 2006 is used to analyze the displacement features of the slope and evaluate the performance of the SSA-TCN model. The cumulative displacement time series is decomposed into trend displacement (linear part) and periodic displacement (nonlinear part) by the variational modal decomposition (VMD) method. Then, a polynomial function is used to predict the trend displacement, and the SSA-TCN model is used to predict the periodic displacement of the landslide based on considering the response relationship between periodic displacement, rainfall, and reservoir water. This research also compares the proposed approach results with the other popular machine learning and deep learning models. The results demonstrate that the proposed hybrid model is superior to and more effective and accurate than the others at predicting the landslide displacement.
Chapter
One of the main characteristics of landslides is the variability of their kinematics over time. Satellite remote sensing has proven to be very useful in detecting and monitoring slow‐moving landslides over wide and sometimes remote areas. This detection/monitoring is based on two main techniques: image correlation applied to synthetic aperture radar (SAR) amplitude images or optical images and interferometric synthetic aperture radar (InSAR). Image correlation techniques use the recognition of patterns in two successive orthorectified images, based on their similarity and evaluated through the optimization of a cost‐function calculated with a small window size. The correlation of SAR amplitude images, also known as offset tracking can be applied to landslides when patterns, due to persistent features or to correlated speckle, exist and are visible in the diachronic SAR pairs. Since the first launches of civil SAR satellites in the 1990s, InSAR techniques have been applied to detect and monitor landslides.
Article
Full-text available
Abstract Bridge displacement monitoring is one of the key components of bridge structural health monitoring. Traditional methods, usually based on limited sets of sensors mounted on a given bridge, collect point-like deformation information and have the disadvantage of providing incomplete displacement information. In this paper, a Persistent Scatterer Interferometry (PSI) approach is used to monitor the displacements of the Nanjing Dashengguan Yangtze River high-speed railway bridge. Twenty-nine (29) European Space Agency Sentinel-1A images, acquired from April 25, 2015 to August 5, 2016, were used in the PSI analysis. A total of 1828 measurement points were selected on the bridge. The results show a maximum longitudinal displacement of about 150 mm on each side of the bridge. The measured displacements showed a strong correlation with the environmental temperature at the time the images used were acquired, indicating that they were due to thermal expansion of the bridge. At each pier, a regression model based on the PSI-measured displacements was compared with a model based on in-situ measurements. The good agreement of these models demonstrates the capability of the PSI technique to monitor long-span railway bridge displacements. By comparing the modelled displacements and dozens of PSI measurements, we show how the performance of movable bearings can be evaluated. The high density of the PSI measurement points is advantageous for the health monitoring of the entire bridge.
Article
Full-text available
This paper describes a new approach to Persistent Scatterer Interferometry (PSI) data processing and analysis, which is implemented in the PSI chain of the Geomatics (PSIG) Division of CTTC. This approach includes three main processing blocks. In the first one, a set of correctly unwrapped and temporally ordered phases are derived, which are computed on Persistent Scatterers (PSs) that cover homogeneously the area of interest. The key element of this block is given by the so-called Cousin PSs (CPSs), which are PSs characterized by a moderate spatial phase variation that ensures a correct phase unwrapping. This block makes use of flexible tools to check the consistency of phase unwrapping and guarantee a uniform CPS coverage. In the second block, the above phases are used to estimate the atmospheric phase screen. The third block is used to derive the PS deformation velocity and time series. Its key tool is a new 2+1D phase unwrapping algorithm. The procedure offers different tools to globally control the quality of the processing steps. The PSIG procedure has been successfully tested over urban, rural and vegetated areas using X-band PSI data. Its performance is illustrated using 28 TerraSAR-X StripMap images over the metropolitan area of Barcelona.
Article
Full-text available
This paper presents a methodology taking advantage of the GPOD-SBAS service to study the surface deformation information over high mountain regions. Indeed, the application of the advanced DInSAR over the arduous regions represents a demanding task. We implemented an iterative selection procedure of the most suitable SAR images, aimed to preserve the largest number of SAR scenes, and the fine-tuning of several advanced configuration parameters. This method is aimed at minimizing the temporal decorrelation effects, principally due to snow cover, and maximizing the number of coherent targets and their spatial distribution. The methodology is applied to the Valle d'Aosta (VDA) region, Northern Italy, an alpine area characterized by high altitudes, complex morphology, and susceptibility to different mass wasting phenomena. The approach using GPOD-SBAS allows for the obtainment of mean deformation velocity maps and displacement time series relative to the time period from 1992 to 2000, relative to ESR-1/2, and from 2002 to 2010 for ASAR-Envisat. Our results demonstrate how the DInSAR application can obtain reliable information of ground displacement over time in these regions, and may represent a suitable instrument for natural hazards assessment.
Article
Full-text available
This paper illustrates the potential of Sentinel-1 for landslide detection, mapping and characterization with the aim of updating inventory maps and monitoring landslide activity. The study area is located in Molise, one of the smallest regions of Italy, where landslide processes are frequent. The results achieved by integrating Differential Synthetic Aperture Radar Interferometry (DInSAR) deformation maps and time series, and Geographical Information System (GIS) multilayer analysis (optical, geological, geomorphological, etc.) are shown. The adopted methodology is described followed by an analysis of future perspectives. Sixty-two landslides have been detected, thus allowing the updating of pre-existing landslide inventory maps. The results of our ongoing research show that Sentinel-1 might represent a significant improvement in terms of exploitation of SAR data for landslide mapping and monitoring due to both the shorter revisit time (up to 6 days in the close future) and the wavelength used, which determine an higher coherence compared to other SAR sensors.
Article
Full-text available
Sentinel-1 (S-1) has an unparalleled mapping capacity. In interferometric wide swath (IW) mode, three subswaths imaged in the novel Terrain Observation by Progressive Scans (TOPS) SAR mode result in a total swath width of 250 km. S-1 has become the European workhorse for large area mapping and interferometric monitoring at medium resolution. The interferometric processing of TOPS data however requires special consideration of the signal properties, resulting from the ScanSAR-type burst imaging and the antenna beam steering in azimuth. The high Doppler rate in azimuth sets very stringent coregistration requirements, making the use of enhanced spectral diversity (ESD) necessary to obtain the required fine azimuth coregistration accuracy. Other unique aspects of processing IW data, such as azimuth spectral filtering, image resampling, and data deramping and reramping, are reviewed, giving a recipe-like description that enables the user community to use S-1 IW mode repeat-pass SAR data. Interferometric results from S-1A are provided, demonstrating the mapping capacity of the S-1 system and its interferometric suitability for geophysical applications. An interferometric evaluation of a coherent interferometric pair over Salar de Uyuni, Bolivia, is provided, where several aspects related to coregistration, deramping, and synchronization are analyzed. Additionally, a spatiotemporal evaluation of the along-track shifts, which are directly related to the orbital/instrument timing error, measured from the SAR data is shown, which justifies the necessity to refine the azimuth shifts with ESD. The spatial evaluation indicates high stability of the azimuth shifts for several slices of a datatake.
Article
Full-text available
The region near Dujiangyan City and Wenchuan County, Sichuan China, including significant giant panda habitats, was severely impacted by the Wenchuan earthquake. Large-area landslides occurred and seriously threatened the lives of people and giant pandas. In this paper, we report the development of an enhanced multi-temporal interferometric synthetic aperture radar (MTInSAR) methodology to monitor potential post-seismic landslides by analyzing coherent scatterers (CS) and distributed scatterers (DS) points extracted from multi-temporal l-band ALOS/PALSAR data in an integrated manner. Through the integration of phase optimization and mitigation of the orbit and topography-related phase errors, surface deformations in the study area were derived: the rates in the line of sight (LOS) direction ranged from −7 to 1.5 cm/a. Dozens of potential landslides, distributed mainly along the Minjiang River, Longmenshan Fault, and in other the high-altitude areas were detected. These findings matched the distribution of previous landslides. InSAR-derived results demonstrated that some previous landslides were still active; many unstable slopes have developed, and there are significant probabilities of future massive failures. The impact of landslides on the giant panda habitat, however ranged from low to moderate, would continue to be a concern for conservationists for some time in the future.
Article
Full-text available
Persistent Scatterer Interferometry (PSI) is a powerful remote sensing technique able to measure and monitor displacements of the Earth’s surface over time. Specifically, PSI is a radar-based technique that belongs to the group of differential interferometric Synthetic Aperture Radar (SAR). This paper provides a review of such PSI technique. It firstly recalls the basic principles of SAR interferometry, differential SAR interferometry and PSI. Then, a review of the main PSI algorithms proposed in the literature is provided, describing the main approaches and the most important works devoted to single aspects of PSI. A central part of this paper is devoted to the discussion of different characteristics and technical aspects of PSI, e.g. SAR data availability, maximum deformation rates, deformation time series, thermal expansion component of PSI observations, etc. The paper then goes through the most important PSI validation activities, which have provided valuable inputs for the PSI development and its acceptability at scientific, technical and commercial level. This is followed by a description of the main PSI applications developed in the last fifteen years. The paper concludes with a discussion of the main open PSI problems and the associated future research lines.
Article
Full-text available
This paper presents a web tool for the unsupervised retrieval of Earth's surface deformation from Synthetic Aperture Radar (SAR) satellite data. The system is based on the implementation of the Differential SAR Interferometry (DInSAR) algorithm referred to as Parallel Small BAseline Subset (P-SBAS) approach, within the Grid Processing on Demand (G-POD) environment that is a part of the ESA's Geohazards Exploitation Platform (GEP). The developed on-demand web tool, which is specifically addressed to scientists that are non-expert in DInSAR data processing, permits to set up an efficient on-line P-SBAS processing service to produce surface deformation mean velocity maps and time series in an unsupervised manner. Such results are obtained by exploiting the available huge ERS and ENVISAT SAR data archives; moreover, the implementation of the Sentinel-1 P-SBAS processing chain is in a rather advanced status and first results are already available. Thanks to the adopted strategy to co-locate both DInSAR algorithms and computational resources close to the SAR data archives, as well as the provided capability to easily generate the DInSAR results, the presented web tool may contribute to drastically expand the user community exploiting the DInSAR products and methodologies.
Article
Interferometric processing of series of data acquired over time by synthetic aperture radar (SAR) satellites makes it possible to measure millimetric deformations (typically due to landslides, subsidence and earthquake or volcanic phenomena) and to monitor the stability of terrain and infrastructures. Despite the unique capability to observe very large areas, this technology has been typically applied to the analysis of relatively small sites or specific geophysical phenomena. In this work, we present the first application of this technology to a national scale project, which required the processing, through advanced persistent scatterer interferometry (PSI) techniques, of about 20,000 SAR images acquired from 1992 to 2014 over the whole Italian territory by the ERS, Envisat, and COSMO-SkyMed satellites. The obtained results provide a huge database of surface deformation measurements covering the last 20 years and the whole Italian territory, which represents an extremely useful and pioneering service for geo-hazards mapping and prevention.
Article
Geotechnical assets are those found along transportation environments and are made up of or supported by earth materials. A geotechnical asset management program aims to achieve life-cycle performance goals by maintaining the geotechnical assets in a safe, costeffective, and timely manner. Interferometric synthetic aperture radar, a remote sensing technique that derives ground deformation information from satellite-based radar data, is proposed as a monitoring method in the context of geotechnical asset management. A total of 90 ERS-1, ERS-2, and ENVISAT radar images, acquired from 1992 through 2010, are processed using two interferometric stacking techniques to measure ground deformation along a railroad corridor. A local-scale study was first conducted over a known slide (S-1) where total displacements up to 6 cm were measured. Then a regional-scale study was performed to locate hazard zones. Six potential geo-hazard zones were identified along the railroad corridor based on five criteria: slope distance, slope height, slope angle, average downslope velocity, and total downslope displacement. Interferometric stacking is shown to be quite useful for geotechnical asset management, as this remote sensing technique allows for an indirect method of quantifying asset condition and allows for the long-term monitoring of geotechnical assets that a management program aims to achieve.