Cecilia Borghero

Student thesis, Advanced level (Master degree, one year), 30 HE Geomatics

Master Programme in Geomatics

Supervisor: Faramarz Nilfouroushan Assistant supervisor: Manhal Sirat

Examiner: Mohammad Bagherbandi Assistant examiner: Julia Åhlén

2017

Feasibility study of dam deformation monitoring in northern Sweden using

Sentinel1 SAR interferometry

FACULTY OF ENGINEERING AND SUSTAINABLE DEVELOPMENT

Department of Industrial Development, IT and Land Management

Abstract

Dams are man-made structures that in order to keep functioning and to be considered

structurally healthy need constant monitoring. Assessing the deformation of dams can be time

consuming and economically costly.

Recently, the technique of Interferometric Synthetic Aperture Radar (InSAR) has proved its

potential to measure ground and structural deformation. This geodetic method represents a

cost-effective way to monitor millimetre-level displacements and can be used as supplemental

analysis to detect movements in the structure and its surroundings.

The objective of this work is to assess the practicality of the method through the analysis of the

surface deformation of the Ajaure dam located in northern Sweden, in the period 2014-2017,

using the freely available Sentinel-1A images. The scenes, 51 in ascending and 47 in descending

mode, were processed exploiting the Persistent Scatterer (PS) technique and deformation

trends, and time series were produced.

Built in the 60’s, the Ajaure embankment dam is considered as high consequence, meaning that

a failure would cause socio-economic damages to the communities involved and, for this reason,

the dam needs constant attention. So far, a program of automatic measurements in situ has

been collecting data, which have been used partly to compare with InSAR results.

Results of the multi temporal analysis of the PS points on/around the dam show that the dam

has been subsiding more intensely toward the centre, where maximum values are of

approximately 5 ± 1.25 mm/year (descending) and 2 ± 1.27 mm/year (ascending) at different

locations (separated of approximately 70 m). Outermost points instead show values within -0.7

and 0.9 mm/year, describing a stable behaviour. The decomposition of the rate has furthermore

revealed that the crest in the observation period has laterally moved toward the reservoir.

It has been observed that the operation of loading and unloading the reservoir influence the

dam behaviour. The movements recorded by the PS points on the dam also correlate with the

air temperature (i.e. seasonal cycle).

The research revealed that the snow cover and the vegetation could have interfered with the

signal, that resulted in a relative low correlation. Therefore, the number of PS points on and

around the dam is limited, and comparison with the geodetic data is only based on a few points.

The comparison shows general agreement, showing the capacities of the InSAR method.

The study constitutes a starting point for further improvements, for example observation in

longer period when more Sentinel1 images of the study area are collected. Installation of corner

reflectors at the dam site and/or by use of high resolution SAR data is also suggested.

Keywords: SAR interferometry, Sentinel-1A, dam deformation monitoring

Acknowledgments

The Sentinel-1 images used in this thesis were provided by ESA according the free and open data

policy of the Copernicus program.

This thesis is finally real with the kind support of some people.

I would like to thank first of all Vattenfall for the given opportunity and in particular Dr. Manhal

Sirat and Peter Viklander, for the data provided and for the final revision.

My gratitude to Dr. Daniele Perissin and to his colleagues, who gave me technical support for

the software SARPROZ.

I wish to express my thanks to my supervisor, Dr. Faramaraz Nilfouroushan for his guidance in

this so challenging project, providing me all the information possible and encouraging me

constantly.

Thanks to my family and my partner Abraham, who supported me and who pushed me to give

my best.

Table of contents

List of Tables ................................................................................................................................. vii

List of abbreviations .................................................................................................................... viii

1. Introduction .......................................................................................................................... 1

1.2 Problem statement ............................................................................................................. 3

2. Aim ........................................................................................................................................ 4

3. Background ........................................................................................................................... 5

3.1 Embankment Dams ....................................................................................................... 5

3.2 Synthetic Aperture Radar (SAR) .................................................................................... 7

3.2.1 Synthetic Aperture Radar Interferometry (InSAR) ............................................... 13

3.2.2 Differential Synthetic Aperture Radar Interferometry (DInSAR) ......................... 17

3.2.3 Multi-Temporal InSAR (MT-InSAR) ....................................................................... 20

3.2.4 Quality of time series analyses ............................................................................ 23

3.3 Study area ................................................................................................................... 24

3.3.1 Climate and geological settings ........................................................................... 24

3.3.2 The dam ............................................................................................................... 26

4. Data and methods .............................................................................................................. 28

4.1 Data acquisition ........................................................................................................... 28

4.1.1 Sentinel 1A TOPS .................................................................................................. 28

4.1.2 In situ data............................................................................................................ 32

4.2 Data processing ........................................................................................................... 33

5. Results ................................................................................................................................ 40

5.1 InSAR deformation maps and time series ................................................................... 40

5.2 PSI compared to geodetic measurements .................................................................. 45

6. Discussion ........................................................................................................................... 49

6.1 Factors affecting the dam ........................................................................................... 49

6.2 Validation of the results .............................................................................................. 52

7. Conclusion .......................................................................................................................... 54

8. Bibliography ........................................................................................................................ 56

Appendix I – List of images .......................................................................................................... 64

Appendix II – Connection graphs ................................................................................................ 67

List of Figures

Figure 1 Location of the Ajaure dam, retrieved from www.lantamateriet.se. ............................. 4

Figure 2 Example of structure of zoned embankment dam. Modified from Söderlund (2011) ... 5

Figure 3 Microwave portion of the electromagnetic spectrum and its division into bands.

Retreievd from http://gsp.humboldt.edu/ ................................................................................... 8

Figure 4 Interaction of SAR signals with vegetation (Parker (2012)). ........................................... 9

Figure 5 SAR imaging geometry. Modified from Ray, Kartikeyan, & Garg (2016). .................... 10

Figure 6 Range resolution. P is the length of the pulse. Longer the pulse, coarser range

resolution. 1 and 2 are not separable. 3 and 4 instead will be recognized as two targets(Natural

Resources Canada, n.d.) .............................................................................................................. 11

Figure 7 Azimuth resolution. A is the angular width of the beam. The azimuth resolution

becomes coarser with increasing distance from the antenna. 1and 2 can be distinguished, while

3 and 4 will appear as one object (Natural Resources Canada, n.d.) .......................................... 11

Figure 8 Wave characteristics. The phase is the position of a point on a waveform cycle and is

measured as an angle in degrees or radians, as can be seen in the projected circle. Every cycle

goes from 0 to 2π. The amplitude is the height from the center line to the peak. Modified from

X. Zhou et al. (2009). ................................................................................................................... 12

Figure 9 Geometry of InSAR measurements. Two different signals are recorded at two different

times. The distance from SAR 2 to the target to the ground is a measure ∆R longer than the

distance from SAR 1 (so Rs,2 = Rs,1 + ∆R). This phase difference can be exploited to estimate the

height h. B ⊥and B// are the perpendicular and parallel component respectively of the spatial

distance, or baseline, between SAR1 and SAR2. From (Plank, 2014) ......................................... 14

Figure 10 InSAR basic processing ................................................................................................ 15

Figure 11 DInSAR principle, modified after Crosetto (2016). ..................................................... 18

Figure 12 Decomposition of LOS motion. The vertical and horizontal (East-West plane)

components can be extracted projecting the ascending and in the descending geometries. As

consequence, the real vector of displacement can be established. Sketch adopted from Tofani

et al., (2013). ............................................................................................................................... 19

Figure 13 Concept of Distributed Scatterer (a) and Persistent Scatterer (b). Whether distributed

individual scatters are present in a resolution cell then the back-scattered signal is random, since

it is affected by decorrelation. If one of the objects in the resolution cell is brighter than the

others, the variation in the received signal due to other scatters is minimal. Hence, the back-

scattered signal is stable and can be used for analysis. From Andrew Hooper, Bekaert, Spaans,

& Arikan (2012). .......................................................................................................................... 21

Figure 14 Precipitation and temperature data for the study area in the period 2010-2016. Data

retrieved from SMHI (http://luftweb.smhi.se/) .......................................................................... 25

Figure 15 Ajaure dam, the study area. The spillway structure is the gallery that allows the water

to flow out of the reservoir and the intake structure is where the water is captured to be sent

further to the hydroelectric powerplant. .................................................................................... 26

Figure 16 Lateral movement (on the right) and vertical movement (on the left) measured on the

crest since the construction. On the y axis the displacement expressed in mm. In the Ajaure dam

the horizontal displacements are larger than the subsidence values. Modified after Holm &

Nilsson (2007) .............................................................................................................................. 27

Figure 17 TOPS acquisition geometry. Tb is the busrt duration. ωr is azimuth beam rotation rate.

From (Meta, Prats, Steinbrecher, Scheiber, & Mittermayer, 2007) ........................................... 31

Figure 18 Plan of the Ajure dam with the measurement points used in this research for the

validation of the results. Modified after Johansson & Watley (2007) ........................................ 32

Figure 19 Interferometric configuration for ascending (top) and descending (bottom) dataset.

On the Y axis the perpendicular baseline expressed in meters, on the X axis the acquisition date

of the images in format yyyymmdd. ........................................................................................... 34

Figure 20 Top: Location of the footprints of the master images in Sweden. Bottom: closer look

to the footprints of the scene for descending (red) and ascending (blue) modes. The arrows

indicate the flight path (340 and 201 for ascending and descending respectively) and the LOS

direction for each dataset. The yellow circle represents the dam location. .............................. 36

Figure 21 Reflectivity maps of the study area. On the top: graphic representation of the direction

of the flight and of the radar beam in respect with the dam. On the bottom, right: ascending

mode and left: descending mode. The yellow rectangle indicates the dam location, whereas the

red points indicate the points selected as PS. The images use SAR coordinates: Samples indicate

range, Lines indicate azimuth. .................................................................................................... 38

Figure 22 Deformation rates for the PS points in the ascending mode for the period 2014-2017.

Positive values mean uplift, negative values mean subsidence. The pink triangle is the reference

point. The numbered points are the points selected for the time series analysis. GSD- Ortofoto,

1m © Lantmäteriet (2016) .......................................................................................................... 41

Figure 23 Deformation rates for the PS points in the descending mode for the period 2014-2017.

Positive values mean uplift, negative values mean subsidence. The pink triangle is the reference

point. The numbered point is the points selected for the time series analysis. GSD- Ortofoto, 1m

© Lantmäteriet (2016) ................................................................................................................ 42

Figure 24 Time series of the ascending selected points for the period 2014-2017 (see Figure 22

for their location). The table reassume the temporal coherence, the deformation trend and the

associated cumulative displacement for each point. .................................................................. 43

Figure 25 Time series of the descending selected points for the period 2014-2017 (see Figure 23

for their location). The table reassume the temporal coherence, the deformation trend and the

associated cumulative displacement for each point. .................................................................. 44

Figure 26 Projection of the deformation to the axis perpendicular to the dam (shown in

horizontal plane). ........................................................................................................................ 45

Figure 27 Results of the decomposition of the InSAR velocities. All the three couples show a

movement toward the reservoir (orange arrows). One couple shows vertical movement toward

up, while the other two toward down (red arrows). The arrows are not proportional to the

magnitude. .................................................................................................................................. 46

Figure 28 Comparison of InSAR data with geodetic data. In the x axis the time, in y axis the

displacement in mm. a) InSAR (couple 2 and 15) vs M44 b) InSAR (couple 3 and 14) vs M40 and

c) InSAR (couple 5 and 12) vs M32. The bias of the beginning of the graphs is due to the different

“year zero” of the two sets (for InSAR it is 2014, whereas for the geodetic data is 2001). ....... 48

Figure 29 InSAR projections in direction perpendicular to the dam for the couple ID 3 and 14 on

the dam (InSAR) against the variation of the water level of the reservoir. The periodic

fluctuations of the water level correlates well with the periodic jumps in the InSAR projections

shown by red ellipses. For the last ellipse, there is no data available for water level, but the

timing of jumps in INSAR time series is expected to be due to water level changes. ................ 50

Figure 30 InSAR vertical projections for the couple of points ID 3 and 14 on the dam (InSAR)

against the variation of the air temperature .............................................................................. 51

List of Tables

Table 1 Specifications of SAR missions. ........................................................................................ 8

Table 2 Characteristics of Sentinel 1. Information retrieved from (Sentinel-1Team, 2013) ...... 29

Table 3 Acquisition modes of Sentinel 1 (Sentinel-1Team, 2013). ............................................. 30

Table 4 Metadata of the images, both ascending and descending ............................................ 31

Table 5 Temporal coherence, deformation trend and cumulative displacement for the ascending

selected points ............................................................................................................................ 43

Table 6 Temporal coherence, deformation trend and cumulative displacement for the

descending selected points ......................................................................................................... 44

List of abbreviations

DEM Digital Elevation Model

DInSAR Differential Synthetic Aperture Radar Interferometry

ESA European Space Agency

GNSS Global Navigation Satellite System

InSAR Interferometric Synthetic Aperture Radar

LOS Line of sight

MT-InSAR Multi Temporal InSAR

PS Persistent Scatterer

PSC Persistent Scatterer Candidate

PSI Persistent Scatterer Interferometry

SAR Syntethic Aperture Radar

SLC Single Look Complex

SRTM Shuttle Radar Topography Mission

TOPS Terrain Observation with Progressive Scans

1

1. Introduction

Throughout history, barriers were built across rivers to create water reservoirs intended for

several human related activities and purposes, for example water supply, irrigation, navigation

and power generation. The construction techniques have been improved, evolving from the

rudimental walls made of stone bricks of the Ancient Egypt to the modern dams (Yang, Haynes,

Winzenread, & Okada, 1999).

Given the growth of the human population and the consequent needs of water and power,

the socio economic importance of these engineering structures has been increased globally in

the last few decades, with apex in the 50’s and 60’s (Altinbilek, 2002; Bosshard, 2010). The issue

of climate change provides also a reason for protecting water resources (Brown, Tullos, Tilt,

Magee, & Wolf, 2009) and dams have hence had a big role in the regulation of the distribution

of fresh water and in the keeping of an equilibrium while meeting the demands.

Despite the achievement of better living conditions and flood protection, the construction of

dams has some negative effects both for the environment, as for example impacts on species

and ecosystems, and for the society, as high financial costs and eventual dislocation of local

people (Kornijów, 2009; Tahmiscioğlu & Anul, 2007).

In addition to the issues mentioned above, some concerns are related to the maintenance

and the safety management of the infrastructures. Indeed, if from one side dams give benefits

to the society, from another represent potential risk hazard to the environment and to the

community if safety issues are in danger. Episodes of dam collapsing and consequent water

release have already happened in the past triggered by earthquakes, erosion, aging and heavy

rainfall, causing several damages downstream (Milillo, Perissin, et al., 2016; Roque, Perissin,

Falcão, Fonseca, & Maria, 2015). As consequence, continuous monitoring and analysis for the

detection of instabilities are demanded, aiming at accurately estimating and minimizing their

socio-economic risks impact.

Different procedures have been practiced for the observation and prediction of dam

behaviour and the monitoring of factors which might trigger failure, for instance ground water

pressure, stresses within the structure and surface displacements (Milillo, Perissin, et al., 2016;

Stewart & Tsakiri, 1993). Deformation in turn can be due to changes of ground water levels,

tectonic phenomena, and construction parameters (Emadali, Motagh, & Haghshenas Haghighi,

2017; Erol, Erol, & Ayan, 2004).

For the surface deformation monitoring, there are different techniques, namely geodetic

methods, which include total station, precise levelling, photogrammetry and Global Navigation

2

Satellite System (GNSS), non-geodetic methods, which include optical fibre sensors, piezometers

and inclinometers, and InSAR methods. Surveying geodetic approaches are usually used for

determining absolute or relative displacements of selected points with respect to some

reference points assumed to be stable and they are for example geodetic networks, performing

precise levelling and close-range photogrammetry. Approaches based on GNSS measurements

allow the estimation of movements by means of position data from GNSS receivers located on

the dam (He, Sang, Chen, & Ding, 2005; Hudnut & Behr, 1998; W. Li & Wang, 2011; Milillo,

Perissin, et al., 2016; Stewart & Tsakiri, 1993). Geotechnical instruments like inclinometers and

piezometers (Erol et al., 2004) are also used to monitor the dam body behaviour and to correlate

with geodetic data.

Despite conventional methods are usually reliable and accurate, they can provide

measurements only for specific points, where special instrumentation has been placed (J. J.

Sousa et al., 2016); moreover, monitoring could be time consuming and requiring field work

(Roque et al., 2015).

One approach to derive structure deformation information emerging in the last decades and

not yet very widespread is related to Synthetic Aperture Radar (SAR), a form of radar system

used for fine resolution mapping (Rosen et al., 2000). The sensor emits microwave radiation and

it records the reflected signal; by measuring the time taken between transmission and reception

of a pulse, the distance between the radar and the target can be exploited to produce an image

of the ground. The measurement of the phase change between two or more SAR acquisitions

over the same area retrieved at different times defines the Interferometric Synthetic Aperture

Radar (InSAR) technique, which has been proved to be a powerful tool for measuring Earth’s

surface changes (W. Zhou, Li, Zhou, & Chang, 2016). A further step is done when there are

several acquisition available; in these cases it is possible to create a series of images which allows

to monitor displacement trends through time with millimeter-level precision, resulting in multi-

temporal InSAR (X. Zhou, Chang, & Li, 2009).

The InSAR technique is very powerful for deformation monitoring but still it encounters

challenges due to availability of the images, image resolution, poor quality of the results in some

cases due to, for example, low signal strength (in the event of highly vegetated areas or if there

are very steep slopes) and systematic errors (i.e. errors associated with faulty equipment).

3

1.2 Problem statement

In Sweden, there are around ten thousands dams varying in shape and size; lot of them are

relatively old (due the an increase of the establishment of hydropower dams in the years 50´s -

70´s) and they are hence facing signs of age (Ekström, 2012; Norstedt, 2013). RIDAS (2012), the

hydropower industry dam safety guidelines, are a series of theoretical and practical indications

concerning the safety of Swedish dams provided by Svensk Energy.

One of the objectives of RIDAS is the guidance on consequence classification. The structures

are grouped into safety classes based on assessment of the consequences to the communities

in case of failure. In the country, there are about two hundred dams that would lead to national

crisis and most of them are in the Norrland region, the northern area (SOU 2012:46). The several

failure episodes happened over the years have led the owners to pay particular attention to the

risk of failure and to provide continuous upgrade of monitoring programs and instruments (Ø.

Lier, 2015).

Therefore, the aging, the safety requirements and the remoteness of many dam sites

induced the demand for cost effective monitoring systems for the dam behaviour. Reliable and

spread data are crucial for safety planning, while current methods need constant calibration and

are localized.

It is important hence to use a method that is precise, accurate, cost-effective and that can give

estimations for the entire structure.

The utilization of the InSAR approach has been introduced quite recently in Scandinavia for

the improving of the reliability and the accuracy of dam monitoring, but it is not yet used on a

regular basis. It has been reported by researchers Lier (2015) and Voege, Frauenfelder and

Larsen (2012), stating the success of the technique.

4

2. Aim

The objective of the project is the further investigation of the effectivity of the multitemporal

InSAR approach through the analysis of the surface deformation of the Ajaure dam, an

embankment dam located in the region of Västerbotten, North Sweden (Figure 1).

Figure 1 Location of the Ajaure dam, retrieved from www.lantamateriet.se.

The studied dam, owned by the company Vattenfall Vattenkraft AB, has been monitored by

geodetic methods and can be a good target for validation of the InSAR approach. For this

purpose, in this study freely available Sentinel 1A images collected between December 2014 and

April 2017 and provided by the European Space Agency will be used.

This work represents a supplementary source for the future improvements of current

monitoring techniques and an occasion for the discussion of advantages and challenges of the

InSAR method for structural deformation monitoring in high latitude and vegetated areas.

Ajaure dam

Tarnaby

20000m

Gardiken Lake

N732805

4

N7250742

E567

181

5

3. Background

3.1 Embankment Dams

In the phase of planning the dam construction, the selection of its type depends on several

factors, including the topography, the geology and the hydrological characteristics of the

selected site, the socio-economic impacts of the project and the availability of the materials

(U.S.Bureau Of Reclamation, 1987). Dams can be categorized based on their function, the

construction materials and the design. According to the classification by construction materials,

dams can be broadly divided in concrete dams, made of mass concrete, and embankment dams,

constructed of earth and/or rock fill (Mays, 2010).

Embankment dams, which represent the most common dam construction type in Sweden as

well as worldwide (ICOLD, n.d.; Norstedt, 2013), can be in turn earth-fill dams or rock-fill dams,

according to the predominant material employed. The first have mostly compacted soil while

the second have coarser material like gravel and crushed rocks.

An embankment dam is generally bank shaped (Figure 2). Zoned type structures are

composed by a central core made of impermeable materials, like clays, flanked by supporting

shells made of sand, gravel, cobbles, or rock, or mixtures of these materials (U.S. Bureau of

Reclamation, 2012). In Sweden is very common the use for the core moraine, which is glacial

derived material with an extensive variety of grain sizes (Rönnqvist, 2007). The core has the main

function of preventing the water to flow from one side to the other. Filter materials, sand and

gravels, placed in between the core and the shells, constitute a protection against material

transport from the core (U S Bureau Of Reclamation, 1987).

Figure 2 Example of structure of zoned embankment dam. Modified from Söderlund (2011)

A- Core

B- Filters

C- Earth- or rock-fill shell

D- Erosion protection

6

Embankment dams are advantageous because a) construction materials are easily found in the

proximities of the build site, b) can adapt to different valley morphologies and c)relatively less

costly (U.S. Bureau of Reclamation, 2011a, 2011b).

On the other side, embankment dams are subjected to internal and external loads that can

produce deformation, with might lead to further deterioration and eventually failure (Emadali

et al., 2017). It is important then to analyze the behavior of the dam to detect early warning

signs of an impending failure establishing a monitoring plan (Michalis, Pytharouli, & Raftopoulos,

2016), accomplishing then an assessment of the dam security and improvement of design

procedures for both existing and future dams (Mizuno & Hirose, 2009).

Conventional control surveying measurements can differ according to the type of dam and

to the parameters needed (for example, they could be displacement, leakage, pore water

pressure and seepage flows). The monitoring surveys should be designed such that the number

and density of the locations of monitored points and the frequency of the measurements are

determined (Szostak-Chrzanowski & Massiéra, 2006) . Usually information are collected in single

points, creating then a network to represent condition over larger areas (Johansson & Watley,

2005). In Sweden, different methods and instruments are used for several applications, allowing

the full coverage of the dam safety monitoring for example optical fibre sensors, used to

measure simultaneously both strain and temperature at spatially distributed points. These non-

geodetic instruments have been installed also in the study area, the Ajaure dam (Johansson &

Watley, 2007), but they have not been used in this research as there were no data available for

the period considered.

Deformation is inevitable for any dam (Wang, Perissin, Rocca, & Liao, 2011). The dams are

deformed by e.g. vertical subsidence due to the weight of the structure and also the horizontal

movements, perpendicular to the centerline of the dam, because of the hydrostatic pressure

caused by the water (Gökalp & Taşçı, 2009). Deformation could be also associated with

settlements provoked by internal erosion, a process which happens when soil particles from the

body of the dam are being detached and carried away by the seepage (Sjödahl, 2006). Vertical

and horizontal displacements are considered by International Commission of Large Dams

(ICOLD) among the most critical parameters to be measured during construction, reservoir filling

and after the first impoundment, and are one of the fundamental indicators of dam

deterioration (ICOLD, 1989). Several factors are involved in the deformation, such as for example

the water load and pressure, construction settings, temperature changes and the geological

characteristics at the foundation (Emadali et al., 2017).

In embankment dams, deformation starts to occur during construction and it develops

during the first reservoir fillings, but then the rate decreases in the long-term (Michalis et al.,

7

2016). Intensity, rate and direction of the displacement vary in a determined point of the crest

or the dam body during the several phases of the dam life, reflecting the variations of stresses

due to differential settlements in the dam (Szostak-Chrzanowski & Massiéra, 2006).

In order to evaluate these displacements, surface marks should be located at regular space

intervals, usually in the dam crest, at the upstream and downstream sides, in the berm or berms,

and less commonly in other places in the downstream slope

3.2 Synthetic Aperture Radar (SAR)

Synthetic Aperture System (SAR) is an imaging system which uses pulses at microwave

frequencies in order to provide electrical and geometrical properties of the surfaces (Simons &

Rosen, 2007). The antenna transmits pulses to the ground and records the echo that comes back

in the antenna look direction (Osmanoglu, Sunar, Wdowinski, & Cabral-Cano, 2016), producing

images of the Earth’s surface and providing information used for measuring ground

deformation.

SAR systems are installed on airborne or spaceborne platforms which follow a linear path

and images can be retrieved in each weather conditions and at each hour of the day (Simons &

Rosen, 2007). The platforms circumnavigate the Earth with a near-polar orbit at an altitude

ranging from 500 to 800 km above the earth’s surface, depending on the satellite platform

(Ferretti, Monti-guarnieri, Prati, & Rocca, 2007). When the satellite travels from the north

toward the South Pole, it is said to be in a descending orbit; when the satellite travels instead

from the south toward he north pole, the direction is referred as ascending orbit. The same area

can be viewed by two angles almost symmetrical (Mazzanti, Perissin, & Rocca, 2015).

Since 1978, when the first civilian satellite for remote sensing SEASAT for ocean observations

was launched by NASA, other countries equipped satellites with SAR sensor for Earth’s surface

monitoring. The availability of global data at various resolution is granted by satellites supported

by different agencies; ERS 1/2, ENVISAT and Sentinel-1 by the European Spatial Agency, ALOS by

Japan Aerospace Exploration Agency, RADARSAT by the Canadian Spatial Agency and COSMO-

Skymed by the Italian Spatial Agency are just some examples of past and present SAR satellite

missions which operate at selected frequencies.

Revisiting time of the different satellites can vary, from one day of high-resolution satellites

(e.g. COSMO-Skymed satellites) to one month or more of middle resolution satellites (e.g. ERS

satellites). Table 1 gives an insight into characteristics of well-known SAR missions.

8

Table 1 Specifications of SAR missions.

SATELLITE MISSION LIFE SPAN

BAND REVISIT TIME [DAYS]

SWATH WIDTH [KM]*

RESLOUTION [M]*

ERS-1 1991-2000 C 35 100 30x30

ERS-2 1995-2011 C 35 100 30x30

ENVISAT 2002-2012 C 35 5-400 (30-150) x (30-150)

RADARSAT-1 1995-2013 C 24 45 – 500 (8-100) x (8-100)

RADARSAT-2 2007-present C 24 18 – 500 (1-100) x (1-100)

SENTINEL 1A 2014-present C 12 20-400 (5-20) x (5-40)

ALOS 2006-2011 L 46 30-350 (10-100) x (10-100)

TERRASAR-X 2007-present X 11 10-100 (1-16) x (1-16)

COSMO-SKYMED

2007-present X 16 10-200 (1-100) x (1-100)

* the values depend on the operating mode

According to the different applications indeed, radar systems can emit signals at different

frequency; within InSAR the most common are L-band (1-2 GHz, ~24 cm wavelength), C-band (4-

8 GHz, ~6 cm wavelength) and X-band (8-12 GHz, ~3 cm wavelength). The figure below (Figure

3) shows the position of microwave area and the relative bands in the electromagnetic

spectrum.

Figure 3 Microwave portion of the electromagnetic spectrum and its division into bands. Retreievd from http://gsp.humboldt.edu/

9

The shorter the wavelength, the greater the ability to detect small features due to the

different interference properties of the objects on the ground, but shorter the penetration

depth (Ferretti et al., 2007). C-band has been demonstrated to be appropriate for the monitoring

of urban areas and engineering structures; on the other hand, when the areas are vegetated, ,

the signal is not coherent because canopy changes quickly, so it is preferable longer wavelengths

(i.e. the L-band), which can penetrate the vegetation and therefore reach the ground (J. J. Sousa

et al., 2016), as seen in the explanatory example in Figure 4.

Figure 4 Interaction of SAR signals with vegetation (Parker (2012)).

Besides to the range of frequency, radar waves can be differently polarized: either vertically (V)

or horizontally (H). It means that, depending on the design of the sensor, the antenna can be set

to transmit and receive the waves with the electric field plane oriented either in a vertical

horizontal direction. By altering the polarization, the penetration depth of the beam can change

and hence different information about features objects can be retrieved (Dallemand,

Lichtenegger, Raney, & Schumann, 1993). Commonly SAR systems are single-polarized, which

means that the antenna has the same polarization settings both when emitting and recording

the signal, but also other configurations exist.

Unlike other optical sensors that look straight down (e.g. aircraft images), SAR sensors

transmit and receive the radiation with side-looking perspective, while the platform is moving.

The antenna is aimed perpendicular to the direction of flight and the beam illuminate an area

offset the nadir, the point on the surface of the Earth directly beneath the platform.

Figure 5 shows the imaging geometry of a typical side looking system. The flight direction is

known as along-track or azimuth direction; perpendicular to the flight path is the across-track

10

or range direction. The angle between the nadir and the direction to which the antenna is

pointing is the look angle. Related to the look angle, there is the incidence angle, the angle

formed by the radar beam and a line perpendicular to the Earth’s surface at the hitting point.

Due to the curvature of the Earth’s surface, it is greater than the look angle.

The swath represents the covered image width in the range dimension. The longitudinal extent

of the swath depends on the platform motion.

Figure 5 SAR imaging geometry. Modified from Ray, Kartikeyan, & Garg (2016).

Bandwidth, along with the beam mode (configuration defined by the swath and the

resolution), polarimetry and the incident angle are examples of operating parameters in which

SAR missions differ among each other and they are important to fully exploit the advantages of

SAR imagery (Parker, 2012).

SAR images are two dimensional representations of the swept area and they consist of a

mosaic of pixels (or resolution cells), each of them associated with a small area of the Earth’s

surface and represented digitally by a complex number.

A resolution cell is specified by range and azimuth resolutions of the system (Mahafza, 2002).

Range resolution (Figure 6) is the minimum distance at which two targets must be so as to

differentiate their backscattered signal in the direction normal to the flight track. In order to do

so, the distance between the objects should be more than half of the pulse length (i.e. duration,

width) so that the echoes would be received accurately at different times. The range resolution

Nadir

Incident angle

11

depends hence on the beam width of the pulse, and to be improved this width should be shorter

as possible.

The azimuth resolution (Figure 7) is the ability to separate two objects in the along track

direction and it depends on the antenna length and the angular width of the radar beam. The

higher the antenna length, the narrower is the beam width and finer the resolution (Natural

Resources Canada, n.d.).

Due to the configuration of side-looking geometry of the SAR sensors, the radar image appears

distorted relative to the ground coordinates (Dallemand et al., 1993). The three principal effects

are: foreshortening, when the image appears compressed because the radar waves hit slopes

and hence lengths are incorrectly represented; layover, when the radar waves reach top of high

objects before the base and hence the top and the base are imaged in inverse order; shadow,

when the radar waves cannot reach areas because of obstacles and hence that portion will

appear dark in the radar image.

As the sensor transmits radar signals to the surface along the radar beam’s line of sight (LOS),

some of the signals are reflected away from the satellite, some are absorbed and some reflect

Figure 6 Range resolution. P is the length of the pulse. Longer the pulse, coarser range resolution. 1 and 2 are not separable. 3 and 4 instead will be recognized as two targets(Natural Resources Canada, n.d.)

Figure 7 Azimuth resolution. A is the angular width of the beam. The azimuth resolution becomes coarser with increasing distance from the antenna. 1and 2 can be distinguished, while 3 and 4 will appear as one object (Natural Resources Canada, n.d.)

12

back to the satellite. Processors on board the satellite integrate these backscattered signals to

form a strip map. The two main properties that characterize the signals and contain information

about the properties of the objects are the phase and the amplitude (Figure 8), and they are

used for determining the distance to the targets (Burgmann, Rosen, & Fielding, 2000).

Figure 8 Wave characteristics. The phase is the position of a point on a waveform cycle and is measured as an angle in degrees or radians, as can be seen in the projected circle. Every cycle goes from 0 to 2π. The amplitude is the height from the center line to the peak. Modified from X. Zhou et al. (2009).

The amplitude is represented by pixel brightness, which indicates how strong (i.e. how much

energy) is the backscattered signal and it depends on the dielectric characteristics, roughness,

size and orientation of the target materials. Urban areas and hard objects in general are highly

reflective and hence the displayed amplitude will be high. Each resolution cell might comprise

several scattering objects, so the signal strength can vary within a resolution cell, creating a

speckle effect (the image is all “salt and pepper”)(McCandless & Jackson, 2004).

The phase depends on the specific properties of the radar signal and it is connected to the

sensor-to-target distance and hence the time spent by the signal to make the two-way travel. It

varies between -π and +π for each wavelength of distance traveled (Osmanoglu et al., 2016).

While the amplitude image appears similar to any other optical image of the ground, the phase

image appears random, as the phase differ from pixel to pixel.

Transmission and reception of the signals are characterized by delays caused by the different

distances between scatterers and radar. Since radar signals can be assumed as sinusoidal waves

in which one cycle corresponds to a wavelength and since in a signal wavelength and phase are

correlated, any change in wavelength (i.e. any delay) corresponds to a change in the phase.

Considering a received signal, the phase φ can be simplified as (Ferretti et al., 2007):

φ

13

𝜑 =

4𝑅𝜋

𝜆 (1)

λ= radar wavelength

R= distance between radar and target

4π = factor since a two-way path is considered

As consequence, the phase of the back-scattered signal is an estimation of the distance

target-sensor in function of the wavelength λ. Phase values are important as they are the key

element in any interferometric measurement (Ferretti et al., 2007), when two SAR images can

be aligned and reveal information about elevation of the target area or movements along the

line of sight direction (Burgmann et al., 2000).

3.2.1 Synthetic Aperture Radar Interferometry (InSAR)

InSAR is the synthesis of the conventional SAR technique (Rosen et al., 2000), as the principle

key is the measurements of the distance between the radar and a point on the ground using the

phase information (J. J. M. de Sousa et al., 2015).

Interferometry refers to the exploitation of constructive and destructive interference of

electromagnetic waves to obtain information and measurements relative to the waves sources

or the waves themselves. SAR Interferometry is the analysis of interference patterns from two

sets of electromagnetic waves reflecting from the ground (Campbell & Wynne, 2011; Simons &

Rosen, 2007).

SAR interferometry can be applied when the images have been acquired by the same sensor

with same mode and characteristics and when the images are separated by a perpendicular

baseline inferior to the so-called critical baseline, a parameter that is defined as the distance at

which two image become decorrelated and it depends on the SAR sensor’s spatial range

resolution, on the radar frequency and on the sensor target distance (Ferretti et al., 2007;

Perissin, 2016; Rosen et al., 2000).

Thanks to the wide coverage and the millimeter accuracy, a series of successful studies based

on the exploitation of signals from different radar acquisitions were initiated (Liu et al., 2012).

InSAR has been used for the first time in 1974 by Graham for topographic mapping using two

antennas mounted in an aircraft (Graham, 1974).

14

The applications of InSAR can be quite broad but can be grouped as (X. Zhou et al., 2009):

- reconstruction of digital elevation model of surfaces, DEM (Zhang, Wang, & Chen, 2012)

- detection of surface deformation connected with natural phenomena, for example

earthquakes (Furuya & Satyabala, 2008), volcanic eruptions (Massonnet, Briole, & Arnaud,

1995), landslides (P. Liu et al., 2012) , glacier motions(Gourmelen et al., 2011).

- measurement of displacement rates of man- made structures, like for examples buildings,

bridges and dams (Lazecky et al., 2015; Milillo, Bürgmann, et al., 2016). Moreover, Human-

made ground deformations such as land subsidence due to mining activities (Paradella et

al., 2015)

There exist two ways in which two SAR images are taken: from two independent passes at

different days and slightly different position of a sensor like in the case of the ERS-1 and ERS-2

(repeat-pass interferometry) or from two antennas mounted on the same platform (single-pass

interferometry).

The InSAR technique consists in the comparison of an interferometric pair (i.e. two complex

SAR images acquired over the same area at different times) exploiting the received signals for

each imaged point (Figure 9) (Simons & Rosen, 2007).

R1

R2

Phase shift caused by

path difference

Figure 9 Geometry of InSAR measurements. Two different signals are recorded at two different times. The distance from SAR 2 to the target to the ground is a measure ∆R longer than the distance from SAR

1 (so Rs,2 = Rs,1 + ∆R). This phase difference can be exploited to estimate the height h. 𝐵⊥and B// are

the perpendicular and parallel component respectively of the spatial distance, or baseline, between SAR1 and SAR2. From (Plank, 2014)

15

The two acquisitions over the same area are co- registered to form an interferogram. The

interferogram is generated by cross multiplying, pixel by pixel, the first SAR image with the

complex conjugate of the second (Ferretti et al., 2007). The image is composed by coloured

fringes, which represent a full phase cycle. The phase cycle of the wave is represented by an

interval 0 to2π radians and after one cycle it returns to 0. The phase difference between two

waves caused by the difference in the distance is proportional to the displacement, and the

colour displayed in one point shows the magnitude of deformation in that location. However,

the phase returns to zero after it reaches 2π radians, creating ambiguity. This means that the

displacement corresponding to the integral multiples of 2π radians has the same phase. So, in

order to get the actual displacement, the number of cycles is counted from a reference point

where deformation is supposed to be zero. This process is called phase unwrapping (Simons &

Rosen, 2007).

There are many phase unwrapping algorithms studied and described in the literature, from

techniques based on statistical cost network flow (Chen & Zebker, 2001), where the entire image

is unwrap, to branch cut algorithms (Goldstein, Zebker, & Werner, 1988) where the image is

unwrapped only in paths in order to locate and isolate phase discontinuities; moreover, there

exist approaches not only in two dimensions but also in three (Andrew Hooper & Zebker, 2007).

The basic InSAR processing can be visually summarized as the flow chart below (Figure 10).

In the step of the interferogram formation, the amplitudes of both images are multiplied and

the phase is represented by the the change of signal phase Δ𝜑𝑖𝑛𝑡 between the two images:

Δ𝜑𝑖𝑛𝑡 = 4𝜋Δ𝑅/ 𝜆 (2)

Figure 10 InSAR basic processing

Single Look Compleximages import

Co-registration

•alignment of the images

Interferogramformation

•complex multiplication pixel by pixel

Phaseunwrapping

• reconstrution of the absolute phase offset

Geocoding

•using DEM

16

Where ΔR = Rs2 – Rs1 id the range change between first and second pass.

The phase difference is hence the difference in the geometric path phases of the two images

and it is measured in radians (Simons & Rosen, 2007); it contains the contribution from different

components , as reported by Veci (2016).

Δ𝜑𝑖𝑛𝑡 = 𝜑𝑓𝑙𝑎𝑡 + 𝜑𝑡𝑜𝑝𝑜 + 𝜑𝑑𝑒𝑓 + 𝜑𝑎𝑡𝑚 + 𝜑𝑛𝑜𝑖𝑠𝑒 (3)

The components are the flat-Earth phase component 𝜑𝑓𝑙𝑎𝑡, which is due to the curvature of the

reference surface, the topographic phase component 𝜑𝑡𝑜𝑝𝑜which is related to the topographic

height of the pixel, the deformation phase component 𝜑𝑑𝑒𝑓, which is related to the ground

deformation, the phase 𝜑𝑎𝑡𝑚, which represents the delay in the signals due to atmospheric

effects, and the noise term φ𝑛𝑜𝑖𝑠𝑒 caused by orbital errors and by spatial and temporal

discrepancies.

The objective during the interferogram formation is to eliminate all the sources of error

through filtering to be left only with the terms of interest, which usually are the topography or

the deformation.

The flat-Earth phase depends on the parallel baseline and it is estimated using the orbital

and metadata information. Through the phase flattening, it can be subtracted from the complex

interferogram (Ai, Liu, Li, & Li, 2008).

The topographic component, equal to

𝜑𝑡𝑜𝑝𝑜 =

4π

λ

B⊥𝐻

Rsinθ (4)

and the deformation component, equal to

𝜑𝑑𝑒𝑓 =

4π

λ 𝑑𝑙𝑜𝑠 (5)

are the major source of investigation (Burgmann et al., 2000; Ferretti et al., 2007; Hellwich,

1999; X. Zhou et al., 2009 ). BII and B⊥ are parallel and perpendicular baseline respectively, H is

the height of the pixel above a reference surface, θ is the incidence angle and 𝑑𝑙𝑜𝑠 is the

displacement.

Digital Elevation Models can be determined considering the topography component as the

variable, assuming that in the moment in which there are no differences in the surface elevation

(Burgmann et al., 2000). The Shuttle Radar Topography Mission (SRTM), for example, consists

17

of a modified radar system placed onboard of the Space Shuttle which was launched in 2000

and was employed for a brief mission to develop DEM with up to 30 m resolution covering 56S

to 60N (NASA, 2017).

The deformation component is due to the movements of the target objects on the ground

between the two acquisitions in the Line of Sight (LOS) direction and it is necessary to know the

other components to be determined. It is the key of the Differential InSAR (DInSar). In these

cases, an external DEM is taken for removing the topographical phase contribution. SRTM DEM

is often used as default as it is automatically integrated in some packages, but others DEM, such

as ASTER, can be used. Regional height models could be also employed, but SRTM has the

advantage that it is global (it covers landmasses between 60N and 54S), it is freely available and

it is accurate.

The different conditions in the atmosphere, and in particular the troposphere, (water

vapour, clouds etc.) at the time of the SAR acquisitions cause a delay of the signals (X. Zhou et

al., 2009), contributing to the interferometric phase as atmospheric phase component 𝜑𝑎𝑡𝑚.

This component can be estimated from the interferograms by applying spatial and temporal

filters, exploiting the concept that atmospheric artefacts are spatially but not temporally

correlated (Hu et al., 2014).

The phase noise term φ𝑛𝑜𝑖𝑠𝑒 can be linked to the interferometric coherence (Osmanoglu

et al., 2016), defined as the degree of the correlation of both amplitude and phase from two

signals (Zhang, T., Zeng, Q., Li, Y., Xiang, 2008). The coherence parameter describes the quality

and the reliability of the interferogram and it can be displayed in an image with values ranging

between 0 and 1, where higher values mean less system noise. Typical sources of decorrelation

are the vegetation, the weathering, and rapid movements (e.g. landslides) (X. Zhou et al., 2009).

The noise term can be estimated and reduced through processing methods like filtering (e.g.

Goldestein filtering), multi-looking techniques and other algorithms (López-Martínez &

Fàbregas, 2002).

3.2.2 Differential Synthetic Aperture Radar Interferometry (DInSAR)

The process of performing displacement measurements is also called Differential InSAR, and it

is based on the fact that if the baseline information and the topography are known, then the

corresponding phase component can be simulated and removed from the interferometric phase

and the displacement in the Line Of Sight (LOS) can be estimated (Ferretti et al., 2007).

18

Calculating the displacement by using an existing external DEM is known as 2-Pass DInSAR

and if topography is estimated first it will be 3-Pass DInSAR.

As seen in Figure 11, a first image is acquired (Master) at a time t1 and measures the

correspondent phase. A second image (Slave) is acquired at a time t2, after an event that causes

the point P moving to P′. Assuming that the surface remains stable, the second measured phase

can be subtracted from the first to obtain the interferometric phase. From here, the deformation

d is extracted and calculated. Deformation is positive when the range between satellites and

ground is increasing, which can be associated to a subsidence of the area.

Figure 11 DInSAR principle, modified after Crosetto (2016).

The DInSAR (and InSAR in general) technologies, despite being advantageous because of the

fine spatial resolution, the capability to work in every condition, the precision of the

measurements and the minor necessity of installing geodetic instrumentations (Hu et al., 2014),

have some limitations. As stated before, limitations are the possible low temporal resolution,

the spatial-temporal decorrelation, the errors due to the atmospheric disturbance on the signal

and the one dimensional estimation of the movements (Hu et al., 2014; P. Liu et al., 2012).

The temporal resolution depends on the revisit time of the satellite considered and on the

nature of the phenomenon observed (see also Table 1 for the revisit time of some sensors).

Temporal decorrelations affect the noise component 𝜑𝑛𝑜𝑖𝑠𝑒 and are due to variations of

the scattering properties of the surface, as vegetation and dielectric properties of soil vary

continuously. These changes in the reflection characteristics results in phase difference.

Master

Slave

19

Geometric decorrelations are due to different imaging geometries, i.e. different incident angle

or rather a too large perpendicular baseline (Roque et al., 2015). Decorrelation is what creates

areas that are irregular over time and causes the difficult interpretation of the measurements

(Simons & Rosen, 2007). For any DInSAR technique, essential aspect is that only pixels with small

𝜑𝑛𝑜𝑖𝑠𝑒 should be identified and analysed, as they can give more reliable results (Michele

Crosetto, Monserrat, Cuevas-González, Devanthéry, & Crippa, 2016).

Atmospheric inhomogeneities affect the interferogram interpretation and they are quite

hard to estimate with just two images. The atmospheric effect is commonly only related in time

on the scale of hours to days. Since the time between two acquisition is around one month, the

atmospheric signal is actually decorrelated in time, and therefore causing disturbance (A. J.

Hooper, 2006). Common practice to reduce the atmospheric influence is the stacking, the

generation and combination of multiple interferograms (X. Zhou et al., 2009).

The one-dimensional estimation of the movement in the LOS direction, added to the

insensitivity to the N-S movements, limits the capacity of getting the entire frame of the ground

motion. However, incorporating multiple radar passes using different geometries (with

ascending and descending orbit as well as different incident angles), the displacement vectors

can be extended up to 3-D (Rocca, 2003).

Figure 12 shows how the combinations of different geometries allow the decomposition of

the LOS motion into the vertical and horizontal components (2D).

Figure 12 Decomposition of LOS motion. The vertical and horizontal (East-West plane) components can be obtained projecting both the ascending and the descending geometries. As consequence, the real vector of displacement can be established. Sketch adopted from Tofani et al., (2013).

20

In order to get the projection vectors, the ascending and descending displacement maps can

be combined to get the vertical (up-down) and horizontal (East-west) deformation projections

according (Milillo, Bürgmann, et al., 2016):

𝑑𝐸𝑎𝑠𝑡−𝑊𝑒𝑠𝑡 =

1

2(

𝑑𝐷𝑆𝐶

𝑠𝑖𝑛𝜃𝐷𝑆𝐶−

𝑑𝐴𝑆𝐶

𝑠𝑖𝑛𝜃𝐴𝑆𝐶)

𝑑𝑈𝑝−𝑑𝑜𝑤𝑛 =1

2(

𝑑𝐷𝑆𝐶

𝑐𝑜𝑠𝜃𝐷𝑆𝐶+

𝑑𝐴𝑆𝐶

𝑐𝑜𝑠𝜃𝐴𝑆𝐶)

(6)

Where 𝑑𝐷𝑆𝐶 and 𝑑𝐴𝑆𝐶 are the deformation in the LOS and 𝜃𝐷𝑆𝐶 and 𝜃𝐴𝑆𝐶 are the incident

angles for the descending and the ascending geometry respectively.

3.2.3 Multi-Temporal InSAR (MT-InSAR)

In order to overcome decorrelation and atmospheric inhomogeneities, the Multitemporal InSAR

(MT-InSAR) technique, also called Advanced DInSAR (Herrera et al., 2007) or Persistent Scatterer

Interferometry or so called PSI (Crosetto et al., 2016) is used to estimate the displacement

without sacrificing accuracy and resolution (Hu et al., 2014). The approach gives the spatio-

temporal evolution of an instability phenomenon, describing hence the deformation history of

the considered area (Li et al., 2014).

In a resolution cell, there are many individual objects and the backscattered signal is the sum of

all the individual ones. When the response to the radar is characterized by a strong signal coming

from a principal reflecting object and it is constant over time, then these pixels are called

Persistent Scatterers (PS); when the response is constant over time, but the signal comes from

different small objects, then the pixels are Distributed Scatterers ( Crosetto et al., 2016), as seen

graphically in Figure 13.

21

Figure 13 Concept of Distributed Scatterer (a) and Persistent Scatterer (b). Whether distributed individual scatters are present in a resolution cell then the back-scattered signal is random, since it is affected by decorrelation. If one of the objects in the resolution cell is brighter than the others, the variation in the received signal due to other scatters is minimal. Hence, the back-scattered signal is stable and can be used for analysis. From Andrew Hooper, Bekaert, Spaans, & Arikan (2012).

The MT-InSAR approach was first developed in Milan by Ferretti et al.( 2000). Since then,

several algorithms have been developed over time, but they all have the objective to exploit

multiple SAR images to produce a continuous record of the displacement in the desired time

period (J. J. M. de Sousa et al., 2015; Osmanoglu et al., 2016).

It is possible to distinguish three categories, according to the rules followed for the selection

of the interferometric pairs (Hu et al., 2014; J. J. Sousa et al., 2016). The techniques that use

interferograms with a single master and exploit only coherent pixel with stable phase or

amplitude are called Persistent Scatterer (PS-InSAR) approaches (for example Ferretti, Prati, &

Rocca, 2001; Andrew Hooper, Zebker, Segall, & Kampes, 2004; Kampes, 2006). The techniques

that use multi-master interferograms with small perpendicular baselines to minimize effects of

baseline decorrelation and exploit distributed pixels are called Small Baseline (SB) approaches

(for example Berardino, Fornaro, Lanari, & Sansosti, 2002; Schmidt & Bürgmann, 2003). The

third category is a mixture of the previous two (e.g. Mora, Mallorqui, & Broquetas, 2003).

The algorithms differ hence in theory and practice, as some of them focus on distributed

scatterers and others on persistent scatterers (Osmanoglu et al., 2016). Since in this study PS-

InSAR approach is employed, an overview of its general characteristics and procedures is

described here.

Persistent Scatterer Interferometry (PSI) techniques locate pixels in the images that contain

stable scatterer in a series of interferograms and then, using appropriate data processing, the

phase displacement is separated from the other components in Eq. 3 (Crosetto et al., 2016; A.

22

Hooper, Segall, & Zebker, 2007). Clearly, for an object to be a Persistent Scatterer, has to remain

physically stable along the time period and has to be smaller than the resolution cell so to avoid

geometric decorrelation (Wang et al., 2011). Good targets that are not affected by geometrical

and temporal decorrelation are buildings, roads, railways as well as exposed rocks (Mazzanti et

al., 2015). With the scatterers hence it is created a network similar to the one created with GPS

measurements and errors can be adjusted (Simons & Rosen, 2007).

After the generation of the single master stack interferograms, the key processing step is the

selection of the PS Candidates (PSC). One of the common ways for a preliminary identification

of the points (Andrew Hooper et al., 2012) is based on amplitude dispersion model, i.e. based

on the amplitude value of the scatterer behaviour with time as proposed by Ferretti et al. (2001).

The amplitude dispersion DA is the measure of the phase stability, and defined as the ratio

between the pixel’s standard deviation 𝜎𝐴 to the mean amplitude value 𝜇𝐴 (Ferretti et al., 2001):

𝐷𝐴 = 𝜎𝐴 /𝜇𝐴 (7)

The principle is that a pixel that has a relatively high amplitude (i.e. small dispersion amplitude

DA) that keeps being similar over the acquisitions is expected to have a minor phase dispersion

(so more reliability) compare to a pixel with relative low amplitude (Kampes, 2006).

Ferretti (2001) reported that pixels should have a dispersion value beneath a threshold, typically

between 0.25 and 0.4, in order to be PS Candidates. The relation enables the selection of PSC

without considering the phase and without considering neighbor pixels, allowing isolated points

to be considered (Kampes, 2006).

After this first selection, the aim is to separate these phase terms previously discussed by

exploiting their different spectral behavior in the framework of a multidimensional (space, time

and normal baseline/acquisition geometry) analysis (Colesanti, Ferretti, Novali, Prati, & Rocca,

2003). The PSC are connected in the form of a network (e.g. Delaunay triangulation) to make

relative observations and data modeling (Kampes, 2006).

For each PSC, the interferometric phase (the difference between master and slave phase) is

compared and subtracted to another PSC’s interferometric phase. The two points should be

close enough (for example within 2-3km) in order to compensate spatially correlated noises (G.

Liu, Luo, Chen, Huang, & Ding, 2008). In doing so a reference point among the candidates

expected to be stable is chosen; the ground velocities that will be calculated will be relative to

this assumed stable point. However, the selection of the rerference point depends on the user

so it is subjective. The stable point can have some initial velocities, but they are assumed to be

zero, because its function is just to define a reference velocity for other points. Therefore,

23

deformation maps generated by different users may look different but expected to be equal in

a relative sense.

At this stage, with this “double-difference”, for each PS candidate the DEM error (which is

the height with respect to the reference external DEM) and the linear deformation phase can be

modelled for the entire time series, since atmospheric and noise are reduced because of the

spatial closeness of the points (Hooper et al., 2012). Linear deformation modelling means that

it is assumed that the target is moving at constant rate. However, the best is to look at the time

series for each individual point and see how velocities vary over time.

The remaining components after the filtering are non-linear deformation, the atmospheric

phase and the noise. This residual phase of each couple of points can be now unwrapped and

through integration, each interferogram will be unwrapped (Colesanti, Ferretti, Prati, & Rocca,

2003).

Deformation is both spatially and temporally correlated, whilst atmospheric artifacts are

spatially but not temporally correlated. This information is exploited to apply a filter to estimate

the atmospheric effects and this is the main advantage of this MTInSAR technique.

The final result is the average annual motion for each PS candidate, assuming the linear

deformations are linear.

The just described approach allows the ground and structure deformation measurements

(Mazzanti et al., 2015). In the field of dam monitoring, several studies have demonstrated the

potential of the MTinSAR technique in different countries, for example in Norway (Ø. Lier,

2015), in Iraq (Milillo, Bürgmann, et al., 2016), in Spain (Herrera et al., 2007), in Italy (Milillo,

Perissin, et al., 2016) in Portugal (Roque et al., 2015) and in China (Wang et al., 2011) and Iran

(Emadali et al., 2017).

3.2.4 Quality of time series analyses

PSI techniques use stacks of SAR images to determine the displacement history of objects

referring temporally to a single reference image. The principal outcome of such analyses are the

motion component in function of time for each PS expressed in millimetres, the deformation

velocity over the time period considered, the Atmospheric Phase Screen (APS) of each SAR image

and topographic error, which is the difference between the true height of the scatterer centre

and the height of the provided DEM (M Crosetto, Monserrat, & Crippa, 2010; Crosetto et al.,

2016).

The parameters are all calculated with reference to one of the points, assumed to be stable

24

The reliability of the estimated parameters is function of the temporal coherence 𝜉𝑝

(Lazecky et al., 2015). The temporal coherence variable describes the quality of the fit between

linear deformation model and phase measurements, or in other words, the effectivity with

which the model removes the noise for each pixel (J. J. M. de Sousa et al., 2015) and it is

measured as (Perissin & Wang, 2012) :

ξp = ∑𝑒𝑗(Δφp

i,k−Δφ̅TOPO−𝑖,𝑘 −Δφ̅𝐷𝐸𝐹

−𝑖,𝑘 )

𝑁

𝑁

𝑖,𝑘=1

(8)

Where 𝑁 is the number of interferograms, Δ𝜑𝑝, is the phase difference of the images i and

k (after “flat earth” removal and filtering), Δ�̅�𝑇𝑂𝑃𝑂, is the topographic term, and Δ�̅�𝐷𝐸𝐹 is the

deformation term.

In few words, it is calculated from the time series phase residues. Although, it has not to be

confused with the spatial coherence, which is the mean interferometric coherence and is

calculated as the average of a set of interferograms.

3.3 Study area

Ajaure is a rockfill dam situated in the northern part of Sweden, 30 km south of the massif Norra

Stortfjället (1767 m a.s.l.) and of the town Tärnaby (see Figure 1 for geographic location).

It gets the waters of the River Ume, a river approximately 470km long with an average discharge

of 450 m3/s (vattenkraft.info, 2017) that goes from the lake Överuman on the border between

Norway and Sweden and it drains into the Gulf of Bothnia in the vicinities of the city of Umeå

(east coast).

3.3.1 Climate and geological settings

North Sweden, according to the Köppen classification, belongs to the group Dfc, a subartic

climatic zone with cold winters and with precipitation on average moderate in all seasons

(Kottek, Grieser, Beck, Rudolf, & Rubel, 2006).

25

The average annual precipitation of the area measured in the period 1961- 1990 is around

600mm, due to the frequent western winds and the orogenic effect provoked by the

mountainous chain on the West (SMHI, 2013). For the same time period, the annual mean

temperature is about zero degrees (SMHI, 2009). A model of the daily data made available by

Swedish Meteorological and Hydrological Institute (SMHI) showed that in the studied area in the

period 2011-2016 the highest rainy period is the end of summer/beginning of autumn, following

the same pattern of the temperatures with exception of the year 2014 (Figure 14).

Figure 14 Precipitation and temperature data for the study area in the period 2010-2016. Data retrieved from SMHI (http://luftweb.smhi.se/)

About the geological settings, the dam is located in an area which, like the rest of Sweden, is

part of the Fennoscandian Shield, an area of old crystalline and metamorphic rocks, commonly

mafic and ultramafic, with local variations (SGU, 2016).

According to the data from the Swedish Meteorological and Hydrological Institute (SMHI, 2017),

the area around the dam is for the major part covered by vegetation and the soil is composed

by till material and frequent morphological features are the moraines, formed thanks to the

glaciation and deglaciations processes that occurred during the Quaternary and that shaped the

landscape.

-20

-15

-10

-05

00

05

10

15

20

00

20

40

60

80

100

120

140

160

20

10

-01

20

10

-05

20

10

-09

20

11

-01

20

11

-05

20

11

-09

20

12

-01

20

12

-05

20

12

-09

20

13

-01

20

13

-05

20

13

-09

20

14

-01

20

14

-05

20

14

-09

20

15

-01

20

15

-05

20

15

-09

20

16

-01

20

16

-05

20

16

-09

tem

per

atu

re

pre

cip

itat

ion

Precipitation [mm] Temperatur [°C]

26

3.3.2 The dam

The dam was constructed in the period 1964- 1966 and since then is aimed to the production

of energy (capacity of 75MW); it is one of the largest owned by Vattenfall. The reservoir has an

extension of 47 km2 and a storage capacity of 209 Mm3 (vattenkraft.info, 2012).

The dam height is 46 m and the length at crest 522 m; the height of the core has been increased

during the phase of construction, so the upstream and downstream slopes are steeper closer to

the crest (Johansson & Watley, 2007). In Figure 15, the view from above of the Ajaure dam.

Figure 15 Ajaure dam, the study area. The spillway structure is the gallery that allows the water to flow out of the reservoir and the intake structure is where the water is captured to be sent further to the hydroelectric powerplant.

The dam suffered since the earlier stages of creeps and movements (Johansson & Watley, 2005).

In the late 80’s survey measurements were indicating horizontal movements of the dam, so in

the 90’s additional supports for protection from the erosion were placed on the downstream

side. From the time of construction to 1989 the total horizontal displacement at the crest was

of approximately 300 mm, and the creep has continued at a rate of 8 mm/year (Nilsson &

Ekström, 2004). Biggest movements toward downstream were recorded by the two

inclinometers installed, which measure the slope change relative to a baseline at different levels

of the bedrock.

In 2001, because of the new flood design imposed by Swedish Guidelines for Flood, the core

has been increased of 5 m in height, using a geomembrane. These works caused little subsidence

but after 2002 the crest settled, keeping a vertical deformation rate of ca 2 mm/year and a

horizontal deformation rate of 3 mm/year (Holm & Nilsson, 2007). After the raise of the crest

210 m

Spillway

structure Intake

structure

27

also the inclinometers registered less movement, as the upper level recorded an annual

inclination of 4mm/m.

Between 2003 and 2007 measurements detected movements on the crest were within the

range of 20 mm (Holm & Nilsson, 2007).

The figure below (Figure 16) shows the vertical and lateral measurements measured on the

crest since the construction, indicating that the lateral motion has been the main component in

the dam deformation (Holm & Nilsson, 2007). The direction of the horizontal movement is

downstream. The jumps in 1994 (yellow circles in the figure) are in correspondence with the

construction of the bank support in the downstream side.

The dam is classified as high consequence dam (class A), meaning that in case of failure the

downstream communities would be severely damaged (RIDAS, 2016).

The dam monitoring using InSAR methods has been carried out by Lier and colleagues ( 2015).

In that study, TerraSAR-X SAR images, both in ascending and descending modes, with high

resolution of 1x1 m covering the years 2013 and 2014 were used. The study showed a total

subsidence of 1 cm in the central part of the crest (O. Lier et al., 2015). That study only covered

a short period of time, i.e. one year. Here, the research is carried on by analysing new satellite

images taken by Sentinel1A in a longer period of time, between 2014 till 2017.

-600

-500

-400

-300

-200

-100

0

mm

Date

Lateral movement

Figure 16 Lateral movement (on the right) and vertical movement (on the left) measured on the crest since the construction. On the y axis the displacement expressed in mm. In the Ajaure dam the horizontal displacements are larger than the subsidence values. Modified after Holm & Nilsson (2007)

-600

-500

-400

-300

-200

-100

0

Date

Subsidence

28

4. Data and methods

In order to analyse the deformation of the dam, it has been employed the Multi-Temporal InSAR

approach, described in section 3.2.3.

The SAR images consist of two datasets (one for the ascending scenes and one for the

descending) of Sentinel1A images and cover the period from December 2014 to April 2017. By

combining InSAR data from two different geometries, it is possible to produce a 2D surface

displacement field and hence to partially overcome the problem of the one-dimensional LOS

motion (Eriksen et al., 2017) so to have a better understanding of the dam and its surrounding

area movements.

The datasets were acquired from the ESA online service and then processed using the software

SARPROZ, a tool developed by Perissin (2016) based on MATLAB (MathWorks, 1996). The

package implements the PS technique (Perissin, Wang, & Wang, 2011) and SAR images from a

wide range of sensors can be processed and lot of functionalities are available. Parameters can

be estimated using the traditional PS algorithm or the Quasi-PS, an implementation developed

by (Perissin & Wang, 2012) that make use of multi-master interferograms. In this study, the

conventional PS methodology was employed.

The obtained results in this study were compared with historical geodetic data provided by

Vattenfall. Comparison of the results with independent data is indeed important for validation

and the acceptability of the PSI method (Crosetto et al., 2016).

In the following sections, the acquisition and the processing of the data for this research are

described in detail.

4.1 Data acquisition

4.1.1 Sentinel 1A TOPS

For this project data from different sensors were firstly considered, but some inconveniences

limited the choice to only Sentinel1 data. COSMO- SkyMED images, promoted by the Italian

Spatial Agency, were not suitable because the only scenes freely available for the study area

were for part of the year 2011, so not enough to elaborate a time series. TERRASAR-X images,

developed by the German Aerospace Center, could give freely only few images and

unfortunately, they could not be delivered in a suitable time for this thesis.

29

Sentinel 1A data were hence chosen because of the free availability, even at high latitudes, and

because of the moderate resolution that still allows to investigate the structural deformation.

The European Space Agency started to develop in 2012 a new family of satellites, called

Sentinels as part of the Copernicus program – the European Commission’s Earth Observation

Program. There are six satellites, aimed to observe both land and ocean environments as well

as atmospheric monitoring. Sentinel1 consists a pair of polar-orbiting satellites (A and B)

equipped with a sensor operating in the C-band. The technical aspects of Sentinel-1 are

summarized in Table 2:

Table 2 Characteristics of Sentinel 1. Information retrieved from (Sentinel-1Team, 2013)

Characteristic Value/description

Launch 2014 (1A) and 2016 (1B)

Orbit Near-polar, sun-synchronous

Orbit height 693 km

Repeat period 12 days repeat cycle (1 satellite); 6 days when

both satellites (A and B) operating

Spatial resolution (5-20) x (5-40) m (depends on acquisition mode)

Flight attitude Right looking

Incident angle 20° - 46° (depends on acquisition mode)

Polarization Single and dual polarization (depends on

acquisition mode)

Frequency 5.405 GHz (C band)

The products are offered with different levels of processing. Simply said, images are Level 0 if

the products are compressed and unprocessed (RAW), Level 1 if they contain orbital information

for the georeferencing (SLC) or if they contain topographical corrections (GRD) and Level 2 if

they are ocean products.

Sentinel-1 can operate in four modes, called Stripmap (SM), Wave (WV), Extra Wide Swath

(EW) and Interferometric Wide Swath (IWS). Each of these modes can work using different

polarization settings (i.e. how the antenna is designed to transmit and receive the EM waves, as

either vertically or horizontally polarised radiation) and their products have diverse geometric

resolution (Table 3).

30

Table 3 Acquisition modes of Sentinel 1 (Sentinel-1Team, 2013).

Mode

Wave

Mode

(WM)

Stripmap

(SM)

Interferometric Wide

Swath (IW)*

Extra Wide

Swath (EW)

Swath (km x km) 20 80 250 400

Geometric resolution

(range x azimuth) 5x5 5x5 5x20 20x40

Polarization single double double double

*this mode was used in this study

The Wave Mode (WM) operates over open oceans and the images are acquired in 20 x 20 km

vignettes at 5 x 5 m resolution. The Stripmap Mode (SM) acquires data with an 80-km swath at

5 x 5 m resolution and it is used to support emergency management actions. The Extra Wide

Swath Mode (EW) employs the TOPSAR technique to acquire data over a 400-km swath. This

mode is employed mainly for ice monitoring over polar areas, oil spill monitoring and security

services.

The Interferometric Wide Swath Mode (IW), the mode used here, is the default mode over land

and has a swath mode of 250 km in across-track direction and the spatial resolution cell is 5x20

m (range x azimuth). This mode images the surface using the Terrain Observation with

Progressive Scans (TOPSAR). It acquires images by recording subsets of echoes, which are called

bursts (Yague-Martinez et al., 2016). Each SLC image is split into three sub-swaths that are in

turn subdivided into several bursts.

The principle behind the IW mode using TOPS is that data are acquired cyclically in bursts

employing a rotation of the antenna both in the range direction (switching from sub-swath to

sub-swath) and in the azimuth direction (from burst to burst) so that the beam is directed

electronically from backward to forward (Figure 17). The technique results in a homogeneous

image quality throughout the swath (De Zan & Guarnieri, 2006).

31

In this project, two datasets of Single Look Complex (SLC) Sentinel 1A ascending and

descending images, consisting of 51 and 47 images respectively, were analysed.

The data were obtained freely from the online service Scientific Data Hub

(https://scihub.copernicus.eu) developed by ESA. In Table 4 the metadata of the images are

reported, while in Appendix I the list of all the scenes is given.

Table 4 Metadata of the images, both ascending and descending

Orbit Ascending Descending

Number of scenes 51 47

Product type SLC SLC

Acquisition mode IW IW

Sub-swath 2 1

Acquisition period 11-12-2014 to 23-4-2017 14-12-2014 to14-04-2017

Track 102 139

Incident angle 38.75° 33.32°

Heading angle (from North)

340° 201°

Polarization VV VV

Figure 17 TOPS acquisition geometry. Tb is the busrt duration. ωr is azimuth beam rotation rate. From

(Meta, Prats, Steinbrecher, Scheiber, & Mittermayer, 2007)

32

4.1.2 In situ data

Vattenfall Vattenkraft AB, the company that owns the Ajaure dam, has collected different

geotechnical and geodetic data and information over the years.

The dam is relatively well instrumented (Holm & Nilsson, 2007) and there are installed

measurement systems addressed to detection of motion, pore pressure in the spillway and

intake galleries, groundwater level, leakage, temperature and precipitation. Despite periodical

maintenance of the instrumentation has been done, along the historical collection there are a

series of shortcomings in the measurements due to the calibration errors, system failures

etcetera. For this study, the data used to compare and validate the results are the air

temperature the reservoir water level and settlement and lateral movements retrieved by

measurement studs placed in the upstream retaining wall along the dam crest (M32, M40 and

M44), see Figure 18. Only three points were selected because they recorded data during the

time interval considered and because, as it will be explained later, these are the approximate

location of the InSAR measurement points.

The air temperature and the water level are both measured hourly while the studs’ data are

retrieved twice a year.

The measurement studs M32, M40 and M44 are just three of the forty-four studs installed

along the wall in 2001. Of these forty-four points, only eleven have been collected data also after

2008. The period considered for the comparison is only 2014-2017.

Figure 18 Plan of the Ajure dam with the measurement points used in this research for the validation of

the results. Modified after Johansson & Watley (2007)

Spillway

Intake

Main building

M32

M44

44 M40

44

33

The inclinometers data were also considered. The automatic measurements started in 2002,

but available data for this study are only for the period 2012-2016. One of them shows a net

variation from 2002 of 0.7 mm/m (i.e. 0.7 in height in mm related to base length of 1 m), while

another shows small movements in the depth of maximum 2 mm/m, with very small variations

toward the downstream side (Holm & Nilsson, 2007). These data were not used for the

comparison, as in the period 2012-2016, any change has been detected.

4.2 Data processing

The ascending and descending datasets were processed separately, and as final step at the end

they were combined in order to estimate the horizontal and vertical velocity components.

As first step, the images were imported and the orbital data along with weather data (e.g.

air temperature and humidity at the time of satellite image acquisition) were downloaded by

the software.

The second step was co-registration. Co-registration means that for each image, the

resolution cells of the same geographic place are superimposed. This is done using orbital

information and then refined. For the PSI analysis, it is used a single image as reference (master)

for the others (slaves), while in other algorithms, such as Small Baseline Subset (SBAS) or

Enhanced SABS (ESBAS) several images as master are employed (Vajedian, Motagh, &

Nilfouroushan, 2015a, 2015b). For the ascending mode, the images were co registered with

respect to a single master image, acquired in 9th of July 2016. For the descending mode, instead,

the chosen master image was acquired in 6th of June 2016. Master images were selected in the

approximate centre of the temporal baseline domain and so that the perpendicular baseline was

minimalized (even if technically for Sentinel TOPS data is not a real issue, as stated by

Devanthéry et al.(2014)). Atmospheric conditions during the time of the acquisition were also

considered, so the master images were chosen so that no precipitation was recorded.

Figure 19 shows the interferometric configuration of both the dataset as a graph describing the

relationship between the images in function of time and perpendicular baseline. Each image is

connected to the chosen Master image, forming an interferometric pair.

In Appendix II the two graphs are reproduced in bigger format for a better reading.

34

Figure 19 Interferometric configuration for ascending (top) and descending (bottom) dataset. On the Y

axis the perpendicular baseline expressed in meters, on the X axis the acquisition date of the images in

format yyyymmdd.

Master

Master

Ascending

Descending

35

In order to make a little bit faster the procedure, only a portion of the subswath underwent

coregistration and its size depends on how the footprints of the images -of the same dataset-

overlap. In this case, the master image for the ascending mode has major area compare to the

descending mode (Figure 20, bottom), but this is not going to influence the time series later on.

Indeed, although the coregistration is a very crucial step in InSAR analyses, it is done to create

the two stacks indipendetely. The important thing is that the dam is covered by both ascending

and descending images so that both datasets can furnish information about the deformation.

Figure 20 shows the disposition of the master images in relation to Sweden (top) and in relation

to the location of the dam (bottom). By means of arrows it also displays the geometry of the

image acquisition for both ascending and descending.

¯

36

Figure 20 Top: Location of the footprints of the master images in Sweden. Bottom: closer look to the footprints of the scene for descending (red) and ascending (blue) modes. The arrows indicate the flight path (340 and 201 for ascending and descending respectively) and the LOS direction for each dataset. The yellow circle represents the dam location.

As third step, the geocoding allowed the projection of SAR coordinates (azimuth,range) into

a conventional cartographic system (longitude, latitude) and the projection of the DEM to the

SAR coordinates.

For the InSAR analysis, the software’s default external DEM is the SRTM but since the dam’s

latitude is over 60° N then an ASTER GDEM (Global Digital Elevation Model) was considered and

downloaded from the United States Geological Survey website (USGS, 2017). The DEM has a

pixel size of 1 arc second, which corresponds to ca. 30 meters.

The stack was geocoded using a Ground Control Point, inserted using geopgraphical coordinate

system with the aid of Google Maps.

After these preliminary operations, the two datasets were processed once again independently

using an apposite module that is specific for PSI analysis of small areas. The scene was then cut

to keep only the dam and the surroundings, for a total area of ca 10 km2.

The starting point of the multi temporal analysis is the reflectivity map, which represents the

average intensity (i.e. amplitude) and it is the result of the average of all the dataset. The

reflectivity is displayed in a grayscale map, where bright areas are indicators of stable reflection

of the radar signals.

Tarnaby

37

For each pixel is calculated the Amplitude Stability Index, which is based on the concept of

the amplitude dispersion DA described in the section 3.2.3, but normalized from 0 to 1.

It is defined as:

𝐴𝑚𝑝. 𝑆𝑡𝑎𝑏. Index = 1 −𝜎/µ = 1 -DA (9)

In simple words, the higher is the standard deviation (i.e. less stable is the point in time), the

lower is the value of the stability index, and vice versa. There is no compulsory value that must

be used as threshold, as it depends on the circumstances, but typically points are considered

reliable with ASI around 0.7-0.75 and higher (Ferretti et al., 2001).

The study area is highly vegetated, so the points which were selected as Permanent

Scatterers had an Amplitude Stability Index which was lower compare to other studies which

use the same method and software but in different regions with less vegetations (e.g. Roque et

al., 2015).

In this study, the PS candidates were identified creating a mask using both the Amplitude

Stability Index threshold, set to 0.4, and a Reflectivity threshold, set to 2 for the descending data

and 1.7 for the ascending data. These values were adopted in order to obtain enough points in

the area especially on the dam structure despite the vegetation. If the ASI threshold would have

been higher, no points would have been available for the analysis. On the other side, the

Reflectivity threshold is different for each dataset simply because in correspondence with the

dam structure the reflectivity range is different in the ascending mode compare to the

descending mode. In the ascending mode 592 candidates were found, while in the descending

mode it decreased to 235 as seen in Figure 21.

In the figure it is possible to see that the PS candidates are placed where areas are more bright

(i.e. where reflectivity is higher), which correspond with roads, exposed rocks, buildings and the

dam itself. On the oter side, vegetated area characterised by low density of PS points, as for

example in the area to the east of the dam. However, the installation in these low-reflectant

zones of corner reflectors, which are special devices that allow to reflect the radar signal back

to the satellite from objects that naturally would not do it, would aid the identification of PS

points and increase the network of measurement points.

38

ASC

DESC

Figure 21 Reflectivity maps of the study area. On the top: graphic representation of the direction of the flight and of the radar beam in respect with the dam. On the bottom, right: ascending mode and left: descending mode. The yellow rectangle indicates the dam location, whereas the red points indicate the points selected as PS. The images use SAR coordinates: Samples indicate range, Lines indicate azimuth.

39

Once the points are selected, the amplitudes of the points are analysed and the deformation

parameters can be calculated by the software. A point assumed to be stable, outside the dam

body is designated as reference point.

The software fits a linear model to determine the deformation rate and DEM errors for each PS

point. As final step, the APS is estimated for the Master image and subtracted to the phase for

each image using a spatio temporal filtering .

At first the results are very noisy (i.e. positive and negative values are randomly distributed,

without any pattern), so a further threshold for the coherence is set to 0.5 to eliminate outliers.

A commonly used standard value is of 0.7 (Bakon, Oliveira, Perissin, Sousa, & Papco, 2016), but

in this case the area is characterized by low coherence, so the threshold has been lowered in

order to still have PS points on the dam.

At the end, two sets of information were then obtained: one for the ascending and one for the

descending mode.

At this point, very important fact to consider is that the deformation rates are displayed in the

LOS. As already mentioned, this means that a location might result uplifting according to the

ascending geometry and subsiding for the descending geometry, or the contrary, or it might be

going in the same way.

Either way, it is possible to have a first understanding of the prevailing direction of the

deformation by visual comparison of the deformation trend for points from the two geometries.

Therefore, time series were produced for points on the body of the dam to understand the

behaviour of the area and to detect causes of such trend.

After the analysis of time series of the single points, the results from both the datasets could

be combined to decompose the signal into vertical and horizontal components to further

investigate the data. Even though technically it is possible to do such operation through the

software, where two different points from different orbits close together via interpolation or

gridding are processed, in this case the resolution of the area and the low density of the PS

points do not allow to have a good alignement and hence a reliable decomposition.

As a solution, also performed for example in Milillo, Bürgmann, et al. (2016) two spatially

close points on the dam were taken and Eq 6 have been applied.

40

5. Results

5.1 InSAR deformation maps and time series

The processing of the InSAR data allowed the production of a map that displays the deformation

velocity for each PS point covering the period from December 2014 to April 2017, .i.e 29 months.

The points and the related information were exported in table format in order to create time

series graphs for selected points. The results, worth to remember, are relative, both spatially

(one point in the map) and temporally (the acquisition date of the master image).

In Figure 22 and Figure 23 the deformation rates in the Line of Sight of the Persistent

Scatterer points determined by processing the ascending and descending datasets respectively

are shown. The rates vary between -10 mm/year to +10 mm/year for both maps. Positive values

indicate a movement of the object toward the sensor (uplift), while negative values indicate the

opposite (subsidence).

Looking at the two maps the first thing that can be observed is that the points from the two

datasets are not in the same locations. The reason is related to the different geometries and the

different settings involved (normal baseline, temporal baseline, terrain properties, look angle,

etc.).

Results obtained using ascending data (Figure 22) show that relatively to the reference point

(pink triangle in the figure), the dam body is slightly affected by subsidence.

41

The structure is indeed subsiding at the centre (ID 13 and ID 14) at rates of approximately 2

mm/year and is more stable toward the extremities (ID 15 and 12), were deformation rate is

approximately zero. In the downstream area, there is a light uplifting, most likely due to the

accumulation that occurs when the sediment is depositing after the transportation by the water

released from the reservoir.

Results from descending mode (Figure 23) show that once again the trend is in general slightly

negative. On the crest of the dam indeed, the points ID 2 and ID5 have a subsiding rate of 0.4

and 0.7 mm/year respectively. ID 3 instead, which is located on the road in the downstream

¯100m

13

14

15

12

Figure 22 Deformation rates for the PS points in the ascending mode for the period 2014-2017. Positive values mean uplift, negative values mean subsidence. The pink triangle is the reference point. The numbered points are the points selected for the time series analysis. GSD- Ortofoto, 1m © Lantmäteriet (2016)

ASCENDING

42

side, has a subsiding rate of 5mm/year. On the downstream area, it is visible the only positive

rate of the entire zone, and this is most likely due to the accumulation, as notices also in the

ascending results.

Interestingly, also in correspondence with the buildings in the right side of the dam (ID 1), there

is a light subsidence.

DESCENDING

¯

2

3

5

Figure 23 Deformation rates for the PS points in the descending mode for the period 2014-2017. Positive values mean uplift, negative values mean subsidence. The pink triangle is the reference point. The numbered point is the points selected for the time series analysis. GSD- Ortofoto, 1m © Lantmäteriet (2016)

100m

1

43

Despite the results visualization using velocity maps constitute a direct way to see the

distribution of the displacement rates in the area, they are not enough and sometime very

useful, as they are mean values and do not reveal the temporal behaviour of the selected points.

Therefore, the time series of the points labelled by numbers in the maps are produced and

shown in figure 24 and 25.

The time series of the ascending points (Figure 24 and Table 5) reveal how the signal

fluctuates along the period of measurements. This fluctuations contain contribution from

different sources, such as, for example, seasonal changes, and some residual noise. As seen also

in the displacement velocity maps, the two points in the middle of the structure (ID 12 and ID

13) have negative trends, while the outermost points (ID 12 and ID 15) have stable semi-flat

trends.

Ascending

Figure 24 Time series of the ascending selected points for the period 2014-2017 (see Figure 22 for their location). The table reassume the temporal coherence, the deformation trend and the associated cumulative displacement for each point.

-20

-10

0

10

20

LOS

dis

pla

cem

en

t [m

m] ID 12

-20

-10

0

10

20

LOS

dis

pla

cem

en

t [m

m] ID 14

ID point

Temporal Coherence

Deformation trend [mm/year] Cumulative displacement

[mm]

12 0.53 0.5±1.34 1.14

13 0.60 -2.4±1.28 -5.79

14 0.57 -2.2±1.27 -5.12

15 0.57 0.9±1.27 2.07

Table 5 Temporal coherence, deformation trend and cumulative displacement for the ascending selected points

-20

-10

0

10

20 ID 13

-20

-10

0

10

20ID 15

44

The time series of the descending points (Figure 25 and Table 6) show that the area of the dam

is subsiding within approximately 0.4 and 5 mm/year. The point ID 3, which is situated on the

downstream slope shows a steeper trend compare to the others. Indeed, ID 3 indicates subsiding

rate of 5 mm/year, whereas ID 1 indicates subsidence at 3 mm/year and ID 2 and 5 show a stable

behavior as the annual rate is close to zero. In correspondence with this point the ground

subsided of approximately 11 mm in three years.

All the four signals considered seemed to have recorded a high peak during December/January

2016 (Figure 25).

Descending

Figure 25 Time series of the descending selected points for the period 2014-2017 (see Figure 23 for their location). The table reassume the temporal coherence, the deformation trend and the associated cumulative displacement for each point.

Table 6 Temporal coherence, deformation trend and cumulative displacement for the descending selected points

ID point

Temporal Coherence

Deformation trend [mm/year] Cumulative displacement [mm]

1 0.55 -3.1±1.24 -7.30

2 0.54 -0.4±1.24 -0.91

3 0.58 -5.1±1.25 -11.78

5 0.57 -0.7±1.13 -1.65

-20

-15

-10

-5

0

5

10

15

LOS

dis

pla

cem

en

t [m

m] ID1

-10

-5

0

5

10

15

20

ID 2

-20

-15

-10

-5

0

5

10 ID 5

-25

-20

-15

-10

-5

0

5

10

15

LOS

dis

pla

cem

en

t [m

m] ID 3

45

As seen then, it is not only the dam that is suffering displacement but also the surrounding area,

as one of the points located in the buildings in the eastern part. The facts to consider however

are that the measurements are relative and that they are just projections of the real movements.

In any case, the area around the dam is characterized by relative low coherence meaning that

the measurements are affected by some noise, making harder for interpretation.

5.2 PSI compared to geodetic measurements

The geodetic data provided by Vattenfall allowed having a clearer delineation of the dam

deformation and a way to assess the performance of the InSAR technique.

Giving as starting point in the year 2014, the geodetic data were compared with the measures

from the points identified on the dam body. Since the instruments in situ detect vertical and

lateral (i.e. perpendicular to the dam) movement, the InSAR data has been projected using Eq.

6 to compare with old measurements

To do so, three points from the descending mode have been coupled with three points from the

ascending mode. The couples have been decided so that the two points are spatially closed

(around 50 m far from each other). As such, the descending points ID 2, 3 and 5 have been

combined with the ascending points ID 15, 14 and 12 respectively getting an approximate

estimation of the vertical and lateral movement of the entire body. The horizontal deformation

(dhor) has been further projected to the direction perpendicular to the dam (dperp), following the

geometry shown in Figure 26.

Figure 26 Projection of the deformation to the axis perpendicular to the dam (shown in horizontal plane).

Dam

E-W PLANE

50°

dhor

dperp

Normal to the dam

46

The results of these estimates are shown schematically in Figure 27. For each couple of the

points, the direction and magnitude of the displacement have been illustrated

Figure 27 Results of the decomposition of the InSAR velocities. All the three couples show a movement toward the reservoir (orange arrows). One couple shows vertical movement toward up, while the other two toward down (red arrows). The arrows are not proportional to the magnitude.

In the following figure the comparison between the results from the combination of the InSAR

points and the geodetic data (see also Figure 18) is displayed. The displacement, expressed in

millimetres, is represented in function of time and negative values mean subsidence (for vertical

measures) and toward downstream (for lateral measures). Technically, InSAR lateral movement

are positive if toward East and negative if toward West, so the opposite to the direction of the

geodetic measurements. Hence here, for sake of readability of the graphs, the sign of the InSAR

results have been inverted.

The InSAR time series have the zero (reference epoch) in correspondence of the first

acquisition time, December 2014, while the geodetic measurements have the zero in 2001.

Therefore, here only the last three years of the data from Vattenfall have been considered. The

measurement studs in the retaining wall (M32, M40 and M44) do not show relevant subsidence,

whilst they detected an upstream movement. As observed in the report, written by Holm &

Nilsson (2007), that describe the instrumentation installed on the dam, the retaining wall which

host the measurement studs is more exposed to external factors like for example collisions, as

Spillway

Intake

Main building

ID 5 -12

ID 3 -14 ID 2-15

Couple ID

(Asc/Desc) Direction

Rate

[mm/year]

ID 15/2 Vertical 0.33

Normal to dam -0.70

ID 14/3 Vertical -4.43

Normal to dam -1.86

ID 12/5 Vertical -0.10

Normal to dam -0.67

47

the dam is open to traffic. In addition, it is more affected by the common practice of snow

plowing.

In Figure 28a the InSAR projections of the couple ID 2 and 15 (both vertical and perpendicular

to the dam) are plotted against the M44 measurement point. In the interval 2014-2017,

according to the measurement stud, the subsidence has not been very relevant (approx. 1 mm

in three years), while the lateral movement has been more significant, with a major trend of 4

mm/year toward upstream direction. From the graph it is possible to see that a part the

fluctuations, the subsidence recorded by the InSAR signal is in agreement with the conventional

measurement.

About the lateral movement, both the methods (i.e. the InSAR results and the geodetic data)

indicate a trend to move upstream, toward the reservoir.

In Figure 28b the vertical and horizontal components of the combined couple ID 3and 14 are

plotted against the M40 measurement point. The stud recorded a stable vertical behaviour in

the period 2014-2017, but recorded a total lateral movement of 1 cm toward the reservoir since

January 2015. The comparison shows that the vertical displacement according to InSAR is bigger,

almost 8 mm in almost three years, and the lateral movement has same direction but less

intense. The combination of the two InSAR points indeed shows that the dam crest is moving

towards upstream at less than 2 mm/year, and since 2014 the net movement has been of 4 mm.

In Figure 28c the InSAR time series of the components calculated from the couple ID 5 and ID 12

are plotted against the measurement stud M32, which is closer to the spillway compare to M40

and M44. The subsidence, once again is no significant, while the lateral movement showed a

tendency toward upstream of ca 3 mm/year since January 2015, which is in agreement with the

tendency displayed by the InSAR component.

48

-15

-10

-5

0

5

10

15

mm

M44_vert InSAR_vertical

-25

-20

-15

-10

-5

0

5

10

15

mm

M40_vert InSAR_vertical

-20

-15

-10

-5

0

5

mm

M32_vert InSAR_vertical

Figure 28 Comparison of InSAR data with geodetic data. In the x axis the time, in y axis the displacement in mm. a) InSAR (couple 2 and 15) vs M44 b) InSAR (couple 3 and 14) vs M40 and c) InSAR (couple 5 and 12) vs M32. The bias of the beginning of the graphs is due to the different “year zero” of the two sets (for InSAR it is 2014, whereas for the geodetic data is 2001).

a

c

b

b

-20

-15

-10

-5

0

5

10

15

20

mm

M44_lateral InSAR_lateral

-15

-10

-5

0

5

10

15

20

mm

M40_lateral InSAR_lateral

-20

0

20

40

60

80

mm

M32_lateral InSAR_lateral

49

6. Discussion

Given the results obtained from the PSI analysis with the InSAR data, it is possible to discuss the

factors involved in the deformation of the dam and the quality of the investigation performed.

6.1 Factors affecting the dam

Dams are subjected to several forces, namely structural, which include the weight of the dam;

hydraulic, like water pressure; and geological (Siddiqui, 2009).

The results show that the structure undergoes subsidence under its own weight, as the negative

trend in the InSAR measures proves.

The time series analysis allowed to look closer at the deformation pattern and they were plotted

against the water level of the reservoir and the air temperature to detect a connection with

external factors that affect the stability of the structure.

In Figure 29,the comparison with the water level is displayed. The water level varies between

431m and 440m asl. The maximum level is reached around end of June; it stays stable until

December-January and then a discharge brings the level to the minimum, where it stays for a

couple of months. Since the water performs a pressure perpendicular to the dam wall, the water

level data have been compared to the InSAR horizontal projections calculated from points on

the dam structure.

It is possible to see that when the water level is lowering, the signal is responding by displaying

a fluctuation, indicating that the structure is adjusting. This suggests the influence of the

pressure of the water acting on the upstream face of the dam. The connection has been also

pointed out by (Nilsson & Ekström, 2004).

The signal however still contains random noise, so it cannot be asserted that these fluctuations

are only due to the water pressure, but there is a clear correlation.

Since the dam is located at high latitudes, this correlation might be also related to the ice that

form on the water surface in winter time and can possibly introduce some noises

50

Figure 29 InSAR projections in direction perpendicular to the dam for the couple ID 3 and 14 on the dam (InSAR) against the variation of the water level of the reservoir. The periodic fluctuations of the water level correlates well with the periodic jumps in the InSAR projections shown by red ellipses. For the last ellipse, there is no data available for water level, but the timing of jumps in INSAR time series is expected to be due to water level changes.

In Figure 30 it is shown that the air temperature registered at the study site from January 2012

to November 2016; temperature has an average of 6 C and it varies reaching peaks of -30 C,

recorded in January-February, and of 30 C, recorded in July.

If there is a correlation between the projected vertical rates of the points on the dam and the

variation of temperature, it is less evident than the connection with the water pressure. There

could be a seasonal variation of the signal, but in this case, it is not really clear.

-15

-10

-5

0

5

10

15

430

432

434

436

438

440

442

mm

Wat

er lv

el [

m a

sl]

water level InSAR

51

Figure 30 InSAR vertical projections for the couple of points ID 3 and 14 on the dam (InSAR) against the variation of the air temperature

It is worth mentioning that Fennoscandia region including Sweden is affected by Glacial Isostatic

Adjustment (GIA) or post glacial rebound phenomenon in which the land is continuously

uplifting due to the adjustment process of the land mass after the melting of the Fennoscandian

Ice Sheet after the last glaciation, ca 15,000 years ago. The model NKG2005LU derived by

observations of GPS stations, gravimetric measurements and tide gauges, indicate that the study

area uplift is of about 4-6 mm/year (Ågren, J. and Svensson, 2007). However, the InSAR results

are relative and the effect of GIA is negligible in such a small area. In fact, InSAR displacements

are relative to a fixed point which is assumed to be stable, so the displacement is set to zero.

Therefore, the GIA signal is annulled in the time series.

Second fact to take into account, the area is relatively low affected by seismic movements,

even though it is known that in the northern part of the country there are some active faults

(Boedvarsson, Lund, Roberts, & Slunga, 2006). In the interval 2014-2017, it has been observed

only one earthquake worth to mention, of magnitude 4.2 in March 2016, on the east cost of the

region (Svenska Nationella Seismiska Nätet, 2017). Even though it has been felt by the dam, it

has not affected the signal enough to be discerned.

Therefore, the geological component can be excluded for the period considered in this study.

In light of the given considerations and of the presented results, the dam deformation is mostly

driven by its own weight, but it is notable also the contribution of the seasonal cycle on the

-30

-20

-10

0

10

20

30

-40

-30

-20

-10

0

10

20

30

40

11/2013 05/2014 12/2014 06/2015 01/2016 07/2016 02/2017

mm

Tem

per

atu

re [°C

]

Date [mm/yyyy]

Temperature InSAR

52

phase and the water level, that during the charge and discharge operations influences the

pressure on the downstream wall of the dam.

The displacement results for the dam region show a wave pattern that cannot be conducted to

a single factor, so it might be possible that there exist other parameters and that needs farther

detailed studies.

6.2 Validation of the results

Concerning the quality of the accomplished analysis, the obtained accuracy of the LOS velocities

ranges from 1.27 to 1.34 mm/year and from 1.13 to 1.25 mm/year for the ascending and

descending geometry respectively. These values are a little bit higher than the accuracy values

of about 1mm/year reported in the study of Sousa (2016), but they are for example in line with

the numbers in the study of Milillo, Bürgmann, et al. (2016), where Sentinel 1 data used for a

dam in Iraq and gives standard deviations up to 4 mm/year.

The temporal coherence for the area is quite low, with an average of 0.4, compare to other

studies about the dam deformation (e.g. Lazecky, Perissin, Lei, Qin, & Scaioni, 2013), where

values of coherence are higher (around 0.8). The reason can be both related to the area and to

the resolution of data used. Indeed, the vegetation and the high latitude that characterized the

study area are factors of disturbance, as also noted by Lier et al. (2015), who analyzed the Ajaure

dam with high resolution data from 2013 to 2014.

The resolution of the Sentinel 1 data in the this case has limited the possibility of finding the PS

points, even though Sentinel images have been already used in similar studies and they have

been described as potentially useful in the monitoring of man-made structures and

infrastructures (e.g. Sousa et al., 2016, ).

To validate the results of this study two things can be considered: a) previous works and b) the

comparison with the geodetic data.

Lier (2015) using high resolution data found that the dam crest has subsided of one centimeter

and has moved toward the reservoir in the period 2013-2014. Numerically it is similar, or even

more, than the results got in this research but the direction of the movement is in agreement.

It is also true that the period analyzed here is longer and the spatial resolution of the images is

different, so it is expected some difference.

About the comparison with the Vattenfall data, the data quite fit if we take into consideration

the direction of the movement. The magnitude might be different, but this is because the

locations of the PS points do not coincide exactly with the geodetic benchmarks. It must also

consider the characteristics of the InSAR method. The orientation of the dam indeed does not

53

favor the movement detection in the direction perpendicular to the dam

upstream/downstream, especially for the ascending geometry, whose LOS is almost linear with

the dam crest (Milillo, Bürgmann, et al., 2016). Nevertheless, the deformation in the LOS has

been projected manually following the Eq. 6, both vertically and in the direction perpendicular

to the dam, and this computation showed a general agreement.

6.3 About the SAR technique

It has been well documented that remote sensing technique provide information to solve social

challenges, given its potentiality of unifying natural hazards assessments with development

planning studies (Bender, 1991). It is in this light then that this study demonstrates the capacity

the InSAR technique, as it gives the chance of getting accurate measurements of the

deformation of the dam decreasing the field work but furnishing reliable observations.

The method, as it was here performed, presents several advantages. First of all, it was not

destructive, as it could be done from any desk at any distance from the location.

It provided measurements with millimetric precision, as the results in this study demonstrate

that the variation of the measured rate values is within 2 mm/year.

It was cost effective, meaning that it was not necessary any measurement in situ and it has

been used SAR satellite images free of charge. Important notation however is that despite the

method provide a way to do less field work, it does not substitute completely the local

measurements with geodetic instrumentation and always recommended to be validated

InSAR has been demonstrated hence to be a very recent powerful tool but some factors

influenced the success of the research.

First of all, the location of study area. In the case of the Ajaure dam, the location is highly

vegetated that brought low density of the PS points, low temporal coherence in the results and

relatively high standard deviation.

The high latitude moreover, constitutes an obstacle from the climatic point of view. The snow

and the ice indeed influence the signal amplitude, and hence the coherence and the results, as

also Lier (2015) observed.

Second consideration, is the choice of the data in terms of resolution and availability. In this

study for example TerraSAR and COSMO images were considered, but for both the sensors the

study area was covered only for a brief interval of time (in 2013 and in 2011 respectively), not

sufficient to develop a time series analysis. With the advancing technology, the satellites are

more numerous and covering all the globe, but according to the purpose of the project, it might

be hard to find the necessary and useful information for all targeted study areas.

54

7. Conclusion

In Sweden, there are around 10000 dams, and lot of them show sign of age and need

maintenance.

In this dissertation the deformation of the Ajaure dam, northern Sweden, has been monitored

in the interval of time December 2014- April 2017 by use of the Multitemporal InSAR technique.

Ascending and descending data were employed to assess the deformation pattern of the dam.

The results showed that the dam body is deforming with a negative LOS rate, meaning that

the dam is subsiding. The combination of the two acquisition modes showed that subsidence is

higher toward the center of the structure, where values are up to 5 mm/year, whereas going

outwards the dam has more stable behavior, with rates ranging from -0.7 to 0.9 mm/year. The

deformation rate has also a horizontal component directed toward West. This horizontal

component, once projected in the direction perpendicular to the dam, revealed that there is a

motion directed toward the reservoir of up to 2 mm/year.

The analysis brought to the light the interplay of the forces acting on the dam, like the weight

of the structure, the water level of the reservoir and the seasonal cycle.

The data were validated through comparison with geodetic measurements and it can be said

that the InSAR data are in agreement with the three geodetic points analyzed. Despite the low

frequency of the geodetic measurements, the comparison showed that the general trends are

similar, supporting the validation of the InSAR methodology.

The study highlights the capacities and the limitations of this technique. InSAR contributes to

a user obtaining data of higher quality (i.e. more frequent in time and in space) and it simplifies

the evaluation of the dam safety. However, if from one side the need of less field work, from the

other the area’s characteristics limit the quality of the results (e.g. orientation of the dam).

Moreover, InSAR cannot supply direct information about processes which occur inside the fam,

such as internal erosion or faults in the foundation.

This study can be further improved in the future by installing on the Ajaure dam artificial

reflectors, instruments that produce a strong amplitude signal that van be detected through all

the satellite images, providing hence coherent data stacks and allowing the estimation of high

accuracy displacement values.

Another hint for additional enhancement of the research is the use of high resolution images,

like COSMO Sky-Med, which are not free of charge, but they would give a higher quality to the

results.

55

This method that involve state-of-the-art satellite imagery and processing is very powerful, but,

like any other work that is about the safety of the society, needs planning and also some training.

56

8. Bibliography

Ågren, J. and Svensson, R. (2007). Postglacial land uplift model and system definition for the new Swedish height system RH 2000. Reports in Geodesy and Geographical Information Systems (Vol. 2007:4). Gävle,Sweden. Retrieved from https://www.lantmateriet.se/globalassets/kartor-och-geografisk-information/gps-och-matning/geodesi/rapporter_publikationer/rapporter/lmv-rapport_2007_4.pdf

Ai, B., Liu, K., Li, X., & Li, D. H. (2008). Flat-earth phase removal algorithm improved with frequency information of interferogram. In Proc. SPIE 7147, Geoinformatics 2008 and Joint Conference on GIS and Built Environment: Classification of Remote Sensing Images (Vol. 7147, p. 71471A–71471A–10). https://doi.org/10.1117/12.813247

Altinbilek, D. (2002). The role of dams in development. In Water Science and Technology (Vol. 45, pp. 169–180). https://doi.org/10.1080/0790062022012162

Bakon, M., Oliveira, I., Perissin, D., Sousa, J., & Papco, J. (2016). A data mining approach for multivariate outlier detection in heterogeneous 2D point clouds: An application to post-processing of multi-temporal InSAR results. In International Geoscience and Remote Sensing Symposium (IGARSS) (Vol. 2016–Novem, pp. 56–59). https://doi.org/10.1109/IGARSS.2016.7729005

Bender, S. (1991). Primer on natural hazard management in integrated regional development planning. Organization of American States, Department of Regional Development and Environment. Executive Secretariat for Economic and Social Affairs, Washington, DC. Retrieved from https://portals.iucn.org/library/node/14455

Berardino, P., Fornaro, G., Lanari, R., & Sansosti, E. (2002). A new algorithm for surface deformation monitoring based on small baseline differential SAR interferograms. IEEE Transactions on Geoscience and Remote Sensing, 40(11), 2375–2383. https://doi.org/10.1109/TGRS.2002.803792

Boedvarsson, R., Lund, B., Roberts, R., & Slunga, R. (2006). Earthquake activity in Sweden. Study in connection with a proposed nuclear waste repository in Forsmark or Oskarshamn. Sweden: SKB. Retrieved from https://inis.iaea.org/search/search.aspx?orig_q=RN:37086963

Bosshard, P. (2010). The dam industry, the World Commission on Dams and the HSAF process. Water Alternatives, 3(2), 58–70.

Brown, P. H., Tullos, D., Tilt, B., Magee, D., & Wolf, A. T. (2009). Modeling the costs and benefits of dam construction from a multidisciplinary perspective. Journal of Environmental Management, 90(SUPPL. 3). https://doi.org/10.1016/j.jenvman.2008.07.025

Burgmann, R., Rosen, P. A., & Fielding, E. J. (2000). Synthetic Aperture Radar Interferometry to Measure Earth’s Surface Topography and Its Deformation. Annual Review of Earth and Planetary Sciences, 28, 169–209. https://doi.org/10.1146/annurev.earth.28.1.169

Campbell, J. B., & Wynne, R. H. (2011). Introduction to remote sensing. Guilford Press.

Chen, W.C., & Zebker, H. A. (2001). Two-dimensional phase unwrapping with use of statistical models for cost functions in nonlinear optimization. Journal of the Optical Society of America. A, Optics, image science, and vision (Vol. 18). https://doi.org/10.1364/JOSAA.18.000338

57

Colesanti, C., Ferretti, A., Novali, F., Prati, C., & Rocca, F. (2003). SAR monitoring of progressive and seasonal ground deformation using the permanent scatterers technique. In IEEE Transactions on Geoscience and Remote Sensing (Vol. 41, pp. 1685–1701). https://doi.org/10.1109/TGRS.2003.813278

Colesanti, C., Ferretti, A., Prati, C., & Rocca, F. (2003). Monitoring landslides and tectonic motions with the Permanent Scatterers Technique. Engineering Geology, 68(1–2), 3–14. https://doi.org/10.1016/S0013-7952(02)00195-3

Crosetto, M., Monserrat, O., & Crippa, B. (2010). Persistent scatterer interferometry: Potentials and limits. Proceedings of the International Society for Photogrammetry and Remote Sensing Congress. Retrieved from http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.409.9625&rep=rep1&type=pdf

Crosetto, M., Monserrat, O., Cuevas-González, M., Devanthéry, N., & Crippa, B. (2016). Persistent Scatterer Interferometry: A review. ISPRS Journal of Photogrammetry and Remote Sensing. https://doi.org/10.1016/j.isprsjprs.2015.10.011

Dallemand, J. F., Lichtenegger, J., Raney, R. K., & Schumann, R. (1993). Radar imagery: theory and interpretation. Lecture notes.

de Sousa, J. J. M., Lazecky, M., Hlavacova, I., Bakon, M., Patricio, G., & Perissin, D. (2015). Satellite SAR interferometry for monitoring dam deformations in Portugal. In Second International Dam World Conference, Portugal, April (pp. 21–24).

De Zan, F., & Guarnieri, A. M. (2006). TOPSAR: Terrain observation by progressive scans. IEEE Transactions on Geoscience and Remote Sensing, 44(9), 2352–2360. https://doi.org/10.1109/TGRS.2006.873853

Devanthéry, N., Crosetto, M., Monserrat, O., Cuevas-González, M., & Crippa, B. (2014). An approach to persistent scatterer interferometry. Remote Sensing, 6(7), 6662–6679. https://doi.org/10.3390/rs6076662

Ekström, I. (2012). Recent dam incidents and failures in Sweden Key words. In ICSE-6 (pp. 281–288). Paris. Retrieved from http://scour-and-erosion.baw.de/icse6-cd/data/articles/000299.pdf

Emadali, L., Motagh, M., & Haghshenas Haghighi, M. (2017). Characterizing post-construction settlement of the Masjed-Soleyman embankment dam, Southwest Iran, using TerraSAR-X SpotLight radar imagery. Engineering Structures, 143, 261–273. https://doi.org/10.1016/j.engstruct.2017.04.009

Eriksen, H. Ø., Lauknes, T. R., Larsen, Y., Corner, G. D., Bergh, S. G., Dehls, J., & Kierulf, H. P. (2017). Visualizing and interpreting surface displacement patterns on unstable slopes using multi-geometry satellite SAR interferometry (2D InSAR). Remote Sensing of Environment, 191, 297–312. https://doi.org/10.1016/j.rse.2016.12.024

Erol, S., Erol, B., & Ayan, T. (2004). A general review of the deformation monitoring techniques and a case study: Analysing deformations using GPS/levelling. A Paper Presented at XXth ISPRS Congress" Geo-Imagery Bridging Continents, 12–23.

Ferretti, A., Monti-guarnieri, A., Prati, C., & Rocca, F. (2007). InSAR Principles : Guidelines for SAR Interferometry Processing and Interpretation.

Ferretti, A., Prati, C., & Rocca, F. (2000). Nonlinear subsidence rate estimation using permanent scatterers in differential SAR interferometry. IEEE Transactions on Geoscience and Remote Sensing, 38(5 I), 2202–2212. https://doi.org/10.1109/36.868878

58

Ferretti, A., Prati, C., & Rocca, F. (2001). Permanent scatterers in SAR interferometry. IEEE Transactions on Geoscience and Remote Sensing, 39(1), 8–20. https://doi.org/10.1109/36.898661

Furuya, M., & Satyabala, S. P. (2008). Slow earthquake in Afghanistan detected by InSAR. Geophysical Research Letters, 35(6). https://doi.org/10.1029/2007GL033049

Gökalp, E., & Taşçı, L. (2009). Deformation Monitoring by GPS at Embankment Dams and Deformation Analysis. Survey Review, 41(311), 86–102. https://doi.org/10.1179/003962608X390021

Goldstein, R. M., Zebker, H. A., & Werner, C. L. (1988). Satellite radar interferometry: Two-dimensional phase unwrapping. Radio Science, 23(4), 713–720.

Gourmelen, N., Kim, S. W., Shepherd, A., Park, J. W., Sundal, A. V., Björnsson, H., & Pálsson, F. (2011). Ice velocity determined using conventional and multiple-aperture InSAR. Earth and Planetary Science Letters, 307(1–2), 156–160. https://doi.org/10.1016/j.epsl.2011.04.026

Graham, L. C. (1974). Synthetic Interferometer Radar For Topographic Mapping. Proceedings of the IEEE, 62(6), 763–768. https://doi.org/10.1109/PROC.1974.9516

He, X., Sang, W., Chen, Y., & Ding, X. (2005). Steep-slope monitoring: GPS multiple-antenna system at Xiaowan Dam. GPS World, 16(11), 20–25.

Hellwich, O. (1999). SAR Interferometry: Principles, Processing, and Perspectives. In Festschrift für Prof. Dr.-Ing. Heinrich Ebner zum 60. Geburtstag (pp. 109–120). https://doi.org/10.1.1.39.9688

Herrera, G., Tomás, R., Lopez-Sanchez, J. M., Delgado, J., Mallorqui, J. J., Duque, S., & Mulas, J. (2007). Advanced DInSAR analysis on mining areas: La Union case study (Murcia, SE Spain). Engineering Geology, 90(3–4), 148–159. https://doi.org/10.1016/j.enggeo.2007.01.001

Holm, D., & Nilsson, Å. (2007). Ajaure - Utvärdering av mätningar på dammar t.o.m. april 2007.

Hooper, A., Bekaert, D., Spaans, K., & Arikan, M. (2012). Recent advances in SAR interferometry time series analysis for measuring crustal deformation. Tectonophysics. https://doi.org/10.1016/j.tecto.2011.10.013

Hooper, A. J. (2006). Persistent scatter radar interferometry for crustal deformation studies and modeling of volcanic deformation.

Hooper, A., Segall, P., & Zebker, H. (2007). Persistent scatterer interferometric synthetic aperture radar for crustal deformation analysis, with application to Volcán Alcedo, Galápagos. Journal of Geophysical Research: Solid Earth, 112(7). https://doi.org/10.1029/2006JB004763

Hooper, A., & Zebker, H. A. (2007). Phase unwrapping in three dimensions with application to InSAR time series. Journal of the Optical Society of America, 24(9), 2737–2747. https://doi.org/10.1364/JOSAA.24.002737

Hooper, A., Zebker, H., Segall, P., & Kampes, B. (2004). A new method for measuring deformation on volcanoes and other natural terrains using InSAR persistent scatterers. Geophysical Research Letters, 31(23), 1–5. https://doi.org/10.1029/2004GL021737

Hu, J., Li, Z. W., Ding, X. L., Zhu, J. J., Zhang, L., & Sun, Q. (2014). Resolving three-dimensional surface displacements from InSAR measurements: A review. Earth-Science Reviews, 133(October), 1–17. https://doi.org/10.1016/j.earscirev.2014.02.005

59

Hudnut, K. W., & Behr, J. A. (1998). Continuous GPS Monitoring of Structural Deformation at Pacoima Dam, California. Seismological Research Letters, 69(4), 299–308. https://doi.org/10.1785/gssrl.69.4.299

ICOLD. (n.d.). General Synthesis. Retrieved June 29, 2017, from http://www.icold-cigb.net/GB/world_register/general_synthesis.asp

ICOLD. (1989). Monitoring of dams and their foundations - state of the art. Paris.

Johansson, S., & Watley, D. (2005). Dam safety Distributed strain measurements for embankment dams – field tests at ajaure dam 2004-05.

Johansson, S., & Watley, D. (2007). Dam safety experiences from distributed strain measurements in five embankment dams. Stockholm. Retrieved from http://www.hydroresearch.se/en/downloads/07_52_rappport_DTSS4.pdf

Kampes, B. M. (2006). Radar interferometry: Persistent scatterer technique (Vol. 12). https://doi.org/10.1007/978-1-4020-4723-7

Kornijów, R. (2009). Controversies around dam reservoirs: benefits, costs and future. Ecohydrology and Hydrobiology, 9(2), 141–148. https://doi.org/10.2478/v10104-010-0001-4

Kottek, M., Grieser, J., Beck, C., Rudolf, B., & Rubel, F. (2006). World map of the Köppen-Geiger climate classification updated. Meteorologische Zeitschrift, 15(3), 259–263. https://doi.org/10.1127/0941-2948/2006/0130

Lazecky, M., Perissin, D., Bakon, M., De Sousa, J. M., Hlavacova, I., & Real, N. (2015). Potential of satellite InSAR techniques for monitoring of bridge deformations. In 2015 Joint Urban Remote Sensing Event, JURSE 2015. https://doi.org/10.1109/JURSE.2015.7120506

Lazecky, M., Perissin, D., Lei, L., Qin, Y., & Scaioni, M. (2013). Plover Cove Dam Monitoring with Spaceborne InSAR Technique in Hong Kong.

Li, T., Liu, G., Lin, H., Jia, H., Zhang, R., Yu, B., & Luo, Q. (2014). A hierarchical multi-temporal insar method for increasing the spatial density of deformation measurements. Remote Sensing, 6(4), 3349–3368. https://doi.org/10.3390/rs6043349

Li, W., & Wang, C. (2011). Procedia Engineering GPS in the tailings dam deformation monitoring. In Procedia Engineering (Vol. 26, pp. 1648–1657). https://doi.org/10.1016/j.proeng.2011.11.2350

Lier, Ø. (2015). Verification of Remote Sensing in Dam Surveillance – Results of the Scandinavian InSAR pilot. Retrieved from http://www.ich.no/Opplastet/Dokumenter/Hydropower15/Lier_sweden.pdf

Lier, O., cetinic, frano, Lauknes, T. R., larsen, yngvar, & ekström, ingvar. (2015). InSAR on Embankment dams.

Liu, G., Luo, X., Chen, Q., Huang, D., & Ding, X. (2008). Detecting land subsidence in Shanghai by PS-networking SAR interferometry. Sensors, 8(8), 4725–4741. https://doi.org/10.3390/s8084725

Liu, P., Li, Z., Hoey, T., Kincal, C., Zhang, J., Zeng, Q., & Muller, J. P. (2012). Using advanced inSAR time series techniques to monitor landslide movements in Badong of the Three Gorges region, China. International Journal of Applied Earth Observation and Geoinformation, 21(1), 253–264. https://doi.org/10.1016/j.jag.2011.10.010

López-Martínez, C., & Fàbregas, X. (2002). Modeling and reduction of SAR interferometric phase noise in the wavelet domain. IEEE Transactions on Geoscience and Remote Sensing,

60

40(12), 2553–2566. https://doi.org/10.1109/TGRS.2002.806997

Mahafza, B. R. (2002). Synthetic Aperture Radar. In Radar Systems Analysis and Design Using MATLAB. Chapman and Hall/CRC. https://doi.org/doi:10.1201/9781584888543.ch12

Massonnet, D., Briole, P., & Arnaud, A. (1995). Deflation of Mount Etna monitored by spaceborne radar interferometry. Nature, 375(6532), 567–570. https://doi.org/10.1038/375567a0

Mays, L. W. (2010). Water resources engineering. John Wiley & Sons.

Mazzanti, P., Perissin, D., & Rocca, A. (2015). Structural health monitoring of dams by advanced satellite SAR interferometry: investigation of past processes and future monitoring perspectives. In 7th Internation Conference on Structural Health Monitoring of Intelligent Infrastructure, Torino, Italy.

McCandless, S. W., & Jackson, C. R. (2004). Principles of synthetic aperture radar. SAR Marine User’s Manual, 1–23. Retrieved from http://www.sarusersmanual.com/ManualPDF/NOAASARManual_CH01_pg001-024.pdf

Meta, A., Prats, P., Steinbrecher, U., Scheiber, R., & Mittermayer, J. (2007). First TOPSAR interferometry results with TerraSAR-X. In Proceedings of FRINGE (p. 8).

Michalis, P., Pytharouli, S. I., & Raftopoulos, S. (2016). Long-term Deformation Patterns of Earth-fill Dams based on Geodetic Monitoring Data: the Pournari I Dam Case Study. In Proc. of 3rd Joint International Symposium on Deformation Monitoring (pp. 1–5).

Milillo, P., Bürgmann, R., Lundgren, P., Salzer, J., Perissin, D., Fielding, E., … Milillo, P. (2016). Space geodetic monitoring of engineered structures: The ongoing destabilization of the Mosul dam, Iraq. Scientific Reports, 6(October), 37408. https://doi.org/10.1038/srep37408

Milillo, P., Perissin, D., Salzer, J. T., Lundgren, P., Lacava, G., Milillo, G., & Serio, C. (2016). Monitoring dam structural health from space: Insights from novel InSAR techniques and multi-parametric modeling applied to the Pertusillo dam Basilicata, Italy. International Journal of Applied Earth Observation and Geoinformation, 52, 221–229. https://doi.org/10.1016/j.jag.2016.06.013

Mizuno, M., & Hirose, T. (2009). Instrumentation and monitoring of dams and reservoirs. Water Storage, Transport, and Distribution, 253.

Mora, O., Mallorqui, J. J., & Broquetas, A. (2003). Linear and nonlinear terrain deformation maps from a reduced set of interferometric SAR images. IEEE Transactions on Geoscience and Remote Sensing, 41(10 PART I), 2243–2253. https://doi.org/10.1109/TGRS.2003.814657

NASA. (2017). Shuttle Radar Topography Mission. Retrieved July 24, 2017, from https://www2.jpl.nasa.gov/srtm/

Natural Resources Canada. (n.d.). Viewing Geometry and Spatial Resolution. Retrieved July 20, 2017, from http://www.nrcan.gc.ca/node/9341

Nilsson, Å., & Ekström, I. (2004). Raising of the Ajaure embankment dam by extending the moraine core with a geomembrane. In Long-term benefits and performance of dams: Proceedings of the 13th Conference of the British Dam Society and the ICOLD European Club meeting held at the University of Kent, Canterbury, UK from 22 to 26 June 2004. (pp. 69–80). Thomas Telford Publishing. Retrieved from http://www.britishdams.org/2004conf/papers/nilsson.pdf

Norstedt, U. (2013). Swedish dams, dam safety and public safety around dams. In International

61

Symposium: Dam engineering in Southeast and Middle Europe Middle Europe - Recent experience and future outlooks (pp. 47–55). Lubjana, Slovenia. Retrieved from http://www.slocold.si/symp20years/106_Z15_47-55_Norstedt.pdf

North, M., Farewell, T., Hallett, S., & Bertelle, A. (2017). Monitoring the Response of Roads and Railways to Seasonal Soil Movement with Persistent Scatterers Interferometry over Six UK Sites. Remote Sensing (Vol. 9). https://doi.org/10.3390/rs9090922

Osmanoglu, B., Sunar, F., Wdowinski, S., & Cabral-Cano, E. (2016). Time series analysis of InSAR data: Methods and trends. ISPRS Journal of Photogrammetry and Remote Sensing. https://doi.org/10.1016/j.isprsjprs.2015.10.003

Paradella, W. R., Ferretti, A., Mura, J. C., Colombo, D., Gama, F. F., Tamburini, A., … Gomes, L. L. (2015). Mapping surface deformation in open pit iron mines of Carajás Province (Amazon Region) using an integrated SAR analysis. Engineering Geology, 193, 61–78. https://doi.org/10.1016/j.enggeo.2015.04.015

Parker, W. (2012). Discover the Benefits of Radar Imaging,‘. EJE Earth Imaging Journal.

Perissin, D. (2016). Interferometric SAR Multitemporal Processing: Techniques and Applications. In Multitemporal Remote Sensing (pp. 145–176). Springer.

Perissin, D., & Wang, T. (2012). Repeat-pass SAR interferometry with partially coherent targets. IEEE Transactions on Geoscience and Remote Sensing, 50(1), 271–280. https://doi.org/10.1109/TGRS.2011.2160644

Perissin, D., Wang, Z., & Wang, T. (2011). The SARPROZ InSAR tool for urban subsidence/manmade structure stability monitoring in China. … of 34th International Symposium on Remote …. Retrieved from http://www.isprs.org/proceedings/2011/ISRSE-34/211104015Final00632.pdf

Plank, S. (2014). Rapid damage assessment by means of multi-temporal SAR-A comprehensive review and outlook to Sentinel-1. Remote Sensing. https://doi.org/10.3390/rs6064870

RIDAS. (2012). Kraftföretagens riktlinjer för dammsäkerhet.

RIDAS. (2016). Kraftföretagens riktlinjer för dammsäkerhet, kapitel 1-3.

Rocca, F. (2003). 3D motion recovery with multi-angle and/or left right interferometry. In Proceedings of the third International Workshop on ERS SAR.

Rönnqvist, H. F. (2007). Assessing the Potential for Internal Erosion in Glacial Moraine Core Embankment Dams. Dam Engineering, 101–117. Retrieved from http://www.ahec.org.in/links/International conference on SHP Kandy Srilanka All Details%5CPapers%5CTechnical Aspects-A%5CA18Revvised.pdf

Roque, D., Perissin, D., Falcão, A. P., Fonseca, A. M., & Maria, J. (2015). Dams regional safety warning using time-series insar techniques. In Second Internatinal Dam World Conference (pp. 21–24). Lisbon: LNEC.

Rosen, P. A., Hensley, S., Joughin, I. R., Li, F. K., Madsen, S. N., Rodriguez, E., & Goldstein, R. M. (2000). Synthetic aperture radar interferometry. Proceedings of the IEEE, 88(3), 333–382. https://doi.org/10.1109/5.838084

Schmidt, D. A., & Bürgmann, R. (2003). Time-dependent land uplift and subsidence in the Santa Clara valley, California, from a large interferometric synthetic aperture radar data set. Journal of Geophysical Research: Solid Earth, 108(B9), 2416. https://doi.org/10.1029/2002jb002267

Sentinel-1Team. (2013). Sentinel-1 User Handbook. Frascati,Italy: ESA. Retrieved from

62

https://sentinel.esa.int/documents/247904/685163/Sentinel-1_User_Handbook

SGU. (2016). Geology of Sweden. Retrieved from https://www.sgu.se/en/geology-of-sweden/

Siddiqui, I. H. (2009). Dams and reservoirs : planning and engineering. Karachi: Oxford University Press.

Simons, M., & Rosen, P. A. (2007). Interferometric Synthetic Aperture Radar Geodesy. In Treatise on Geophysics (Vol. 3, pp. 391–446). https://doi.org/10.1016/B978-044452748-6.00059-6

Sjödahl, P. (2006). Resistivity investigation and monitoring for detection of internal erosion and anomalous seepage in embankment dams. Lund University. Retrieved from http://lup.lub.lu.se/search/ws/files/5949674/546337.pdf

SMHI. (2009). Normal årsmedeltemperatur. Retrieved July 29, 2017, from http://www.smhi.se/klimatdata/meteorologi/temperatur/normal-arsmedeltemperatur-1.3973

SMHI. (2013). Nederbörd. Retrieved July 29, 2017, from https://www.smhi.se/kunskapsbanken/meteorologi/nederbord-1.361

SMHI. (2017). Vattenwebb - Modelldata per område. Retrieved July 29, 2017, from http://vattenwebb.smhi.se/modelarea/

Söderlund, T. (2011). Jordfyllningsdammar. Förnybar Energi, 14–17. Retrieved from http://www.sero.se/Filer/FE/Frnybar Energi_4_2011.pdf

SOU 2012:46. (n.d.). Dammsäkerhet.Tydliga regler och effektiv tillsyn. Stockholm: Fritze.

Sousa, J. J., Ruiz, A. M., Bakoň, M., Lazecky, M., Hlaváčová, I., Patrício, G., … Perissin, D. (2016). Potential of C-Band SAR Interferometry for Dam Monitoring. Procedia Computer Science, 100, 1103–1114. https://doi.org/10.1016/j.procs.2016.09.258

Stewart, M., & Tsakiri, M. (1993). The Application of GPS To Dam Surface Monitoring. Journal of Geospatial Engineering, 3(1), 45–57.

Svenska Nationella Seismiska Nätet. (2017). Svenska skalv som kändes.

Szostak-Chrzanowski, A., & Massiéra, M. (2006). Relation between monitoring and design aspects of large earth dams. In Proceedings of the 3rd IAG Symposium on Geodesy for Geotechnical and Structural Engineering and 12-th FIG Symposium on Deformation Measurements, ed. H. Kahmen and A. Chrzanowski, Baden, Austria, May (pp. 21–24).

Tahmiscioğlu, M., & Anul, N. (2007). Positive and negative impacts of dams on the environment. … Congress on River …, 760–769. Retrieved from https://pdfs.semanticscholar.org/905f/e3883184c9657cb35fe98e92297b29fe3917.pdf

U.S. Bureau of Reclamation. (2011a). General Design Standards. In Design Standards No. 13. Retrieved from https://www.usbr.gov/tsc/techreferences/designstandards-datacollectionguides/finalds-pdfs/DS13-1.pdf

U.S. Bureau of Reclamation. (2011b). Protective filters. In Design Standards No. 13. Retrieved from https://www.usbr.gov/tsc/techreferences/designstandards-datacollectionguides/finalds-pdfs/DS13-5.pdf

U.S. Bureau of Reclamation. (2012). Embankment Design. In Design Standards No. 13. Retrieved from https://www.usbr.gov/tsc/techreferences/designstandards-datacollectionguides/finalds-pdfs/DS13-2.pdf

U.S.Bureau Of Reclamation. (1987). Design of small dams (3rd ed.). Washington D.C.: U.S. Government Printing Office.

63

USGS. (2017). EarthExplorer. Retrieved August 8, 2017, from https://earthexplorer.usgs.gov/

Vajedian, S., Motagh, M., & Nilfouroushan, F. (2015a). Response to Sowter, A.; Cigna, F. On the use of the ISBAS acronym in InSAR applications. Comment on Vajedian, S.; Motagh, M.; Nilfouroushan, F. StaMPS improvement for deformation analysis in mountainous regions: Implications for the damavand volcano and m. Remote Sensing, 7(9), 11324–11325. https://doi.org/10.3390/rs70911324

Vajedian, S., Motagh, M., & Nilfouroushan, F. (2015b). StaMPS improvement for deformation analysis in mountainous regions: Implications for the Damavand volcano and Mosha fault in Alborz. Remote Sensing, 7(7), 8323–8347. https://doi.org/10.3390/rs70708323

vattenkraft.info. (2012). Ajaure. Retrieved July 30, 2017, from http://vattenkraft.info/?id=1

vattenkraft.info. (2017). Umeälven. Retrieved July 29, 2017, from http://vattenkraft.info/?vattenkraft=1&what=Umeälven#vattenkraft

Vattenkraft.info. (2017). Dammtyper. Retrieved June 30, 2017, from https://vattenkraft.info/?page=23

Veci, L. (2016). Sentinel-1 Toolbox TOPS Interferometry Tutorial. Array Systems Computing Inc. Retrieved from http://sentinel1.s3.amazonaws.com/docs/S1TBX TOPSAR Interferometry with Sentinel-1 Tutorial.pdf

Voege, M., Frauenfelder, R., & Larsen, Y. (2012). Displacement monitoring at Svartevatn dam with interferometric SAR. 2012 IEEE International Geoscience and Remote Sensing Symposium. https://doi.org/10.1109/IGARSS.2012.6350561

Wang, T., Perissin, D., Rocca, F., & Liao, M. S. (2011). Three Gorges Dam stability monitoring with time-series InSAR image analysis. Science China Earth Sciences, 54(5), 720–732. https://doi.org/10.1007/s11430-010-4101-1

Yague-Martinez, N., Prats-Iraola, P., Gonzalez, F. R., Brcic, R., Shau, R., Geudtner, D., … Bamler, R. (2016). Interferometric Processing of Sentinel-1 TOPS Data. IEEE Transactions on Geoscience and Remote Sensing, 54(4), 2220–2234. https://doi.org/10.1109/TGRS.2015.2497902

Yang, H., Haynes, M., Winzenread, S., & Okada, K. (1999). The History of Dams. watershed.ucdavis.edu. Retrieved from https://watershed.ucdavis.edu/shed/lund/dams/Dam_History_Page/History.htm

Zebker, H. A., Madsen, S. N., Martin, J., Alberti, G., Vetrella, S., & Cucci, A. (1992). The TOPSAR interferometric radar topographic mapping instrument. International Geoscience and Remote Sensing Symposium (IGARSS), 2(5), 1595–1597. https://doi.org/10.1109/IGARSS.1992.578643

Zhang, T., Zeng, Q., Li, Y., Xiang, Y. (2008). Study on relation between InSAR coherence and soil moisture. Proceedings of the ISPRS Congress, Beijing, China, 3-11th Julychives, 131–134. Retrieved from http://www.isprs.org/proceedings/XXXVII/congress/7_pdf/2_WG-VII-2/11.pdf

Zhang, W., Wang, W., & Chen, L. (2012). Constructing DEM Based on InSAR and the Relationship between InSAR DEM’s Precision and Terrain Factors. In Energy Procedia (Vol. 16, pp. 184–189). https://doi.org/10.1016/j.egypro.2012.01.031

Zhou, W., Li, S., Zhou, Z., & Chang, X. (2016). Remote sensing of deformation of a high concrete-faced rockfill dam using InSAR: A study of the Shuibuya dam, China. Remote Sensing, 8(3), 255.

Zhou, X., Chang, N.-B., & Li, S. (2009). Applications of SAR Interferometry in Earth and

64

Environmental Science Research. Sensors, 9(3), 1876–1912. https://doi.org/10.3390/s90301876

Appendix I – List of images

List of Sentinel 1 scenes employed in this study and their characteristics.

ASCENDING

Acquisition date [yyyymmdd]

Perpendicular baseline [meters]

Temporal baseline [days]

20141211 -58,2395 -576

20141223 -153,2952 -564

20150104 -92,0458 -552

20150128 -42,4825 -528

20150209 -51,9963 -516

20150221 -38,8238 -504

20150305 -103,0964 -492

20150329 -32,5880 -468

20150410 -72,7696 -456

20150422 -89,8619 -444

20150504 -86,6201 -432

20150516 -75,8554 -420

20150609 24,8834 -396

20150703 -70,2399 -372

20150727 11,5730 -348

20150808 41,4529 -336

20150820 -31,0761 -324

20150901 -64,1771 -312

20150925 -56,5494 -288

20151007 1,3330 -276

20151031 -57,4418 -252

20151206 -72,7392 -216

20160123 -142,4969 -168

20160204 -95,3065 -156

20160228 15,8071 -132

20160311 -55,0390 -120

20160323 -84,5757 -108

20160404 -52,9572 -96

20160428 -7,3172 -72

20160510 21,1219 -60

20160603 -81,1794 -36

20160615 -28,4740 -24

20160709 MASTER 0,0000 0

20160721 -125,8776 12

20160802 -77,4350 24

20160826 40,8686 48

65

20160907 -14,8156 60

20160919 -47,1830 72

20161001 -136,1971 84

20161025 -19,3782 108

20161106 -13,0328 120

20161130 -158,2828 144

20161212 -160,6714 156

20161224 -86,3053 168

20170105 -36,2013 180

20170129 -89,1100 204

20170210 -151,3660 216

20170222 -77,5451 228

20170330 -65,3370 264

20170411 -40,2809 276

20170423 -48,8939 288

DESCENDING

Acquisition date [yyyymmdd]

Perpendicular baseline [meters]

Temporal baseline [m]

20141214 87.3487 -540

20141226 63.8509 -528

20150107 -20.8575 -516

20150131 -6.20504 -492

20150212 214.079 -480

20150224 159.849 -468

20150401 48.0614 -432

20150413 103.426 -420

20150507 22.7312 -396

20150531 -19.5307 -372

20150612 46.6075 -360

20150624 97.9774 -348

20150706 105.271 -336

20150730 -14.4635 -312

20150811 90.522 -300

20150823 76.3868 -288

20150916 93.153 -264

20151127 -40.1056 -192

20151209 -61.2597 -180

20151221 63.0414 -168

20160114 85.3208 -144

20160126 51.5721 -132

20160219 34.3682 -108

20160302 135.999 -96

20160407 18.6128 -60

66

20160501 92.2603 -36

20160525 50.8999 -12

20160606 MASTER 0 0

20160630 69.6146 24

20160724 80.3469 48

20160805 -19.3563 60

20160829 107.729 84

20160910 137.579 96

20160922 81.6055 108

20161004 -1.05183 120

20161028 8.026 144

20161109 78.1934 156

20161121 105.996 168

20161203 -7.08544 180

20161227 -81.0861 204

20170108 76.0954 216

20170201 107.38 240

20170225 -6.21335 264

20170309 70.3336 276

20170321 161.206 288

20170402 129.937 300

20170414 74.8141 312

67

Appendix II – Connection graphs

ASCENDING Normal baseline [m]

68

DESCENDING

Normal baseline [m]